Page 1
Electronic copy available at httpssrncomabstract=1009309
1
Do concentrated institutional investors really reduce executive
compensation whilst raising incentives
Gavin S Smith
dagger and Peter L Swan
Dagger
Current Draft February 2 2012
Abstract
Hartzell and Starks (2003) (HS) report that more concentrated institutional investing
associates with higher pay-for-performance sensitivity and lower CEO pay We find
that institutional concentration has no effect on pay-for-performance sensitivity and
increases compensation when we replicate their analysis using the logarithm of firm
size instead of HSrsquo raw firm size as control Moreover HS results are sensitive to
measuring concentration among institutions rather than concentration among
shareholders Finally the HS concentration definition also appears to exacerbate
firm-size effects Overall HS results may be primarily due to what appear to be weak
firm-size controls and not due to institutional monitoring
JEL classification G23 G32 J33
Keywords Executive compensation Monitoring Institutional ownership Principal-
agent Incentives Concentrated ownership
The authors are grateful to the Australian Research Council (DP0346064) for research funding We
also thank the organizers of the First Singapore International Conference on Finance namely the Saw
Centre for Financial Studies and Department of Finance and Accounting NUS Business School
National University for their financial support We wish to thank Reneeacute Adams Rajesh Aggaral
Philip Brown Sudipo Dasgupta Alex Edmans Joseph Fan David Feldman David Gallagher Gerald
Garvey Jay Hartzell Petko Kalev Ron Masulis Sir James Mirrlees Peter Pham Jay Ritter Laura
Starks Jaeyoung Sung David Yermack and seminar participants at the Chinese University of Hong
Kong and University of Illinois Chicago First Singapore International Conference on Finance the
European Finance Association (EFA) Conference 2007 and the 20th
Australasian Finance and
Banking Conference participants for useful comments We are also particularly grateful to Jay Hartzell
and Laura Starks for supplying us with their original dataset daggerBarclays Capital 745 Seventh Avenue New York NY 10019 gavinsmithbarclayscapitalcom
Tel (+1) 212 526 4217 DaggerCorresponding author School of Banking and Finance University of New South Wales Sydney 2052
Australia peterswanunsweduau Phone (+61) (0)2 9385 5871
Electronic copy available at httpssrncomabstract=1009309
2
With 128 cites in Web of Science and 757 cites in Google Scholar Hartzell and Starks
(2003) [hereafter HS] ranks among the most influential corporate finance papers of the
last decade HS find that concentrated institutional monitors provide greater incentives for
executives to perform by granting more options while simultaneously lowering their cash-
based pay with an even greater fall in total direct compensation HS (p2355) define
concentrated institutional holdings [hereafter Top 5 Concentration] by the proportion of
institutional ownership accounted for by the top five institutional investors12
By
contrast our paper argues that several methodological issues in HS alter their finding by
showing that concentration has either no effect on incentives or perhaps reduces them
and in the case of salary and total pay levels reverse their findings Our paper discusses
these issues and then empirically demonstrates the effect that they have on the findings of
HS
HSrsquos findings are supportive of the idea that firms overpay executives as the
skimming or rent-seeking view of executive compensation maintains See for example
Bertrand and Mullainathan (2001) Bebchuk Fried and Walker (2002) Bebchuk and
Fried (2003 2004) Chhaochharia and Grinstein (2009) and Morse Nanda and Seru
(2011) Chhaochharia and Grinstein (2009) find that increased board requirements
reduced CEO pay but note the fragility qualification provided by Guthrie Sokolowsky
and Wan (2012) and Morse Nanda and Seru (2011) find that powerful CEOs rig
performance measures to increase their pay and that these firms then underperform
1 More specifically HS (p2352) find that an increase of one standard deviation in Top 5 institutional
investor concentration is associated with a 20 percent increase in option grant pay-for-performance
sensitivity and a 19 percent drop in total compensation for the average executive
2 HS also obtain similar findings using a Herfindahl index of institutional fractional holdings that they
construct in a very similar fashion to their main concentration measure Our findings for their concentration
measure will also tend to be true of their Herfindahl measure
Electronic copy available at httpssrncomabstract=1009309
3
We find that Institutional Top 5 Concentration according to the HS definition does
not raise option grant pay-for-performance sensitivity as it has no statistically significant
impact either way once one introduces a standard logarithmically transformed size
control Similarly lagged Institutional Concentration has a negative but statistically
insignificant effect on the pay-for-performance sensitivity of cash (salary plus bonus)
compensation when interacted with the lagged change in shareholder wealth HS obtain a
significant positive coefficient using the contemporaneous change in shareholder wealth
but one need replace this by the lagged change in shareholder wealth in keeping with
what influential institutional investors knew at the time Additionally HS find that lagged
concentrated institutional ownership sizably lowers both salary and total direct
compensation and this is statistically significant but we find the reverse ‒ it appears to
raise both and the findings are highly statistically significant also following a standard
logarithmically transformed size control
Upon closer examination several aspects of the HS research design are of interest
First the primary argument in our paper is that the use by HS of the lagged value of the
level of market capitalization as the size control in regressions (HS p2359) rather than
the conventional lagged logarithm of capitalization distorts the coefficient on the
measure of institutional influence3 A number of authors have drawn attention to falling
pay-for-performance sensitivities as firm size increases (eg Jensen and Murphy (1990)
Garen (1994) Hadlock and Lumer (1997) Schaefer (1998) Murphy (1999) Jin (2002)
and Chichello (2005)) and hence the need to have effective size controls In the HS
dataset the raw mean market capitalization ranges from $1181m in decile 1 to $36654m
in decile 10 (see Table III below) a difference of 3104 fold reducing to only 22 with the
3 HS (p2357) also use net assets and total assets as controls but in keeping with the use of raw
capitalization retain them in their raw form
4
logarithmic transformation While this skewed nature of the level of market capitalization
does not per se rule out its use specification tests such as Box-Cox and the replacement
of market capitalization by its cumulative distribution (see Section VI below) indicates a
possible specification error Moreover using the lagged natural logarithm of market
capitalization as a firm size control in the actual HS data with the same dependent
variable and no other changes raises the explanatory power (R-Squared) of the HS model
by between 135 and 1516 percent4
Given that firm size is an important determinant of executive compensation selecting
the correct functional form for firm size is a vexing and important issue Many estimates
including those of Murphy (1985) Rosen (1992) Gibbons and Murphy (1992) Huson
Parrino and Starks (2001) Baker and Hall (2004) Gabaix and Landier (2008) Edmans
Gabaix and Landier (2009) Babenko (2009) and Aggarwal Erel and Ferreira (2011)
use a logarithmic specification with the elasticity between firm size and executive total
pay in the range of 03 and 04 HS seems to employ inadvertently a firm size measure in
levels (also alternative firm size measures also in levels) and consequently imply an
elasticity value (exponent on market capitalization) of unity
Second we question the use of the HS Top 5 Concentration measure of institutional
influence based on the influence of concentrated institutional investors relative to their
peers that appears to lead inadvertently to an exacerbation of the size-control problem In
their companion theoretical model AHS arrive at a different and what might seem a more
natural specification namely Top 5 Ownership which measures concentration relative to
all shareholders not just institutional peers The following example illustrates the
4 The lower result is based on OLS regressions of model 1 of Table II of HS with the R-Squared rising from
00853 to 00968 once Market Capitalization is logged and the higher result model 1 of Table V of HS is
based on an improvement in R-Squared from 04604 to 05302 (see Table VI below)
5
differences between these two methods If there are one hundred shares outstanding in a
small company and the only institution owns one share that company is concentrated
according even though that institution may lack effective monitoring power Instead if
one deflates by shares on issue this results in a concentration ratio of 1100 If on the
other hand ten institutions collectively own 500 shares in a much larger company of
1000 shares and the Top Five collectively own 250 then the HS concentration ratio falls
to only 250500 or 05 and utilizing shares on issue 025 in this case As is confirmed by
the data one expects Top 5 Concentration to correlate negatively with firm size and Top
5 Ownership the reverse given a positive association between institutional ownership and
company size We then re-examine the monitoring hypothesis using this AHS proposed
measure of institutional investor influence that intuitively relates better to institutional
proxy voting power
Third we address issues of reliance on potentially biased self-reported values of option
grants that result in the exclusion of 16 percent of the HS sample size for tests on total
(direct) compensation Forth we argue that tests that HS perform using an alternative
measure of pay-for-performance sensitivity to provide support for the main hypothesis of
the study require one to get the timing right in terms of the lag structure of performance
and concentration Finally we address issues associated with their tests of endogeneity
and reverse causality These issues collectively might suggest overlooking of a
monitoring effect with the main driving forces size related
The next section reviews the HS study in an effort to tease out what might really be
going on Section II examines the literature for studies that in replicating HS provide
independent support for their findings It also reports findings on the actual proxy voting
record of mutual funds and actions of institutional investors in reducing pay Section III
describes the data and methodology Section IV analyzes the relation between
6
institutional influence and measures of pay-for-performance sensitivity Section V
documents the relationship between institutional influence and total and fixed
compensation Section VI examines the robustness of our findings and the final Section
concludes
I Review of Hartzell and Starks Methodology
We comment on five areas of the HS methodology 1) Size control 2) Measurement of
institutional influence 3) Testing the Jensen-Murphy measure of incentives 4) Reliance
on incomplete self-reported valued for option grants and 5) Endogeneity
A Size Control
In our paper we argue that HS unintentionally identify incentive-size and pay-size
effects5 not a monitoring effect It can be seen in Figure II that executive total pay for
HSrsquos Top 5 institutional concentration measure rises from as little as $072m for the
smallest market capitalization decile to $85m for the highest This twelve-fold pay-size
effect in the deciles can cause problems if one utilizes an untransformed size measure
Due to skewness in the size measure using this in its untransformed state appears to
weaken its effectiveness leaving open the possibility that correlations they identify
between incentives and pay levels on the one hand and Top 5 Concentration on the other
are due largely to size-related effects
Table III displays this strong pay-size effect based on our full 1992‒2002 dataset and
graphically in Figure 1 It shows that the incentive measure adopted by HS declines from
$272 per $1000 change in shareholder wealth for the smallest market capitalization
decile of firms to approximately $030 for the largest Table III and Figure 1 show the
5 The multiplicative model of Edmans Gabaix and Landier (2009) finds that incentive pay increases with
firm size (elasticity value of 13) and the empirical elasticity is 037
7
strong relation between institutional concentration as measured by HS and market
capitalization For the smallest decile of market capitalization firms institutional
concentration averages 61 percent declining to about 275 percent for the highest decile
This negative relationship between firm size and institutional concentration seems to arise
because institutional investors do generally favor small firms This causes problems
because in the limit a small company may have only one institutional investor owning
only a small percentage of the company Hence it would appear to be fully concentrated
even though the ability of the institutional investor to influence pay decisions might be
small Thus it is not at all surprising that the correlation between the logarithm of market
capitalization and their institutional concentration measure is both negative and high in
absolute value at 63 percent Their methodology appears to favor small companies that
Table III shows to have low pay levels and high pay for performance sensitivity
Moreover if one does not transform the raw size measure it implicitly assumes that a
given dollar change in firm size has the same absolute impact on pay for a very large
company as it does for a small company Indeed the conventional logarithmic
specification for firm size provides proportionate effects that imply a pay-size elasticity of
approximately one-third instead of unity6
B Measurement of Institutional Influence
We argue that there are additional issues related to the HS measure of institutional
influence ndash institutional concentration In their companion article to HS AHS develop a
model where they capture institutional influence by a single large long-term institutional
6 Note that unlike an earlier version of our paper we leave the HS dependent variables in levels and hence
untransformed While we do this to minimize the alterations that we make to the HS specification ideally
the elasticity specification is preferable especially for the compensation variables Transforming the
dependent variables does not significantly affect the results
8
investor They measure influence relative to total shareholding including private or
atomistic shareholders Presumably they do this to capture relative activism and proxy
voting strength However when it comes to the empirical implementation of the model
AHS adopt the same relative institutional concentration measure as in HS (AHS (p12))
potentially downplaying the relative strength of institutions compared with other
shareholders They justify this because there is generally more than one institutional
investor and there will be free-riding problems with multiple investors that only the
largest institutional investors can overcome While we accept their premise of free-rider
problems it is not obvious why the deflator is lsquoinstitutional shareholdingrsquo rather than
lsquototal shareholdingrsquo from this perspective7 If the strength of the size control can be
queried scaling by total shareholding (ie Top 5 Ownership) produces a measure that
remains relatively unaffected across the size deciles It is 20 percent for the smaller
decile declining to 14 percent for the largest decile whereas for the HS concentration
measure it declines by over 50 percent from 61 percent to 28 percent (see Table III and
Figure 1)
C Testing the Jensen-Murphy Measure of Incentives
The testing of their secondary measure of executive pay-for-performance sensitivity
(see HS Table IV) based on Jensen and Murphy (1990) raises issues other than the size
control HS regress changes in compensation (cash and total) on lagged institutional
influence that has been interacted with the contemporaneous change in shareholder
wealth to capture a pay-for-performance relationship Such a test is appealing ndash
7 AHS do point out one benefit of the institutional concentration measure Institutional investors may prefer
companies with lsquobetterrsquo incentive compensation structures Could using institutional shareholding rather
than total shareholding as the deflator alleviate this problem Presumably both the numerator and
denominator will have the same bias and hence error cancellation is a possibility
9
institutions should increase compensation when the firm has performed well Following
Greene (2000 p326) to execute such a test one needs to include in level form variables
that are then interacted Hence the HS model specification ought to be
1 2 3 1 4 1it it it it it itCompn SW IC SW IC (1)
where itCompn is the change in executive compensation itSW is the contemporaneous
change in shareholder wealth which also appears in the interaction term and 1itIC is
lagged institutional ownership concentration However HS (p2363) omit the
contemporaneous term itSW and the lagged term 1itIC which should appear in levels
as well as in the interaction term Most importantly HS interact their lagged institutional
concentration measure 1itIC with the contemporaneous change in shareholder wealth
itSW As it is after all pay-for-performance concentrated institutions need to condition
their influence on some observed measure of firm performance Contemporaneous firm
performance is unknown at the time institutional influence is measured Reflecting this
timing insight leads to a new Greene (2000) specification
1 2 1 3 1 4 1 1it it it it it itCompn SW IC SW IC (2)
such that all the lagged RHS independent variables that now appear in the interaction
term are included on a stand-alone basis
Reliance on incomplete self-reported values for option grants
Prior to receiving the HS dataset we tried to replicate the findings of HS to validate
our datasets for the HS period and the longer period 1992‒2002 While we initially
focused on the same period as HS (ie 1992‒1997) we found that our sample size was
larger at 47765 observations than HSrsquos at 33928 observations This discrepancy could
have arisen if HS did not replace missing and self-reported option valuations using their
10
Black-Scholes option methodology This may introduce a downward pay bias since self-
reported option-grant pay is subject to possible understatement and there are 13839
observations with likely high option pay dropped altogether
Endogeneity
HSrsquos endogeneity testing raises concerns about reverse causality Institutions may
prefer to hold firms with certain compensation structures such as incentivized executives
HS rightly highlight this concern and use instrumental variables to address this issue HS
assume that institutional ownership reflects any reverse causality Using an instrumental
variables approach to resolve this problem involves finding a variable that influences or
explains institutional ownership but not directly executive compensation They use share
turnover in a firm as an instrument for institutional ownership However theoretical work
by Holmstrom and Tirole (1993) indicates that measures of stock liquidity and trading
activity are indeed important determinants of executive market-based incentives as Kang
and Lui (2008) and Dikolli Kulp and Sedatole (2009) find empirically As such one can
make a case for adopting another instrument rather than share turnover for institutional
ownership
II Previous Research
Prior research fails to address issues concerning possible fragility of the HS results
AHS replicate the HS findings on both incentives and pay levels using the same dataset as
HS and they apply the same definition of institutional concentration and use market
capitalization expressed as a level Cadman Klasa and Matsunaga (2010 Table 6) set out
to replicate HSrsquos (Table II) model of option grant pay-for-performance sensitivity when
they compare a dataset of ExecuComp and Non-ExecuComp firms for the period 2000-
2007 Their ExecuComp results for this sample period are quite similar to our results for
11
the 1992‒2002 period for the same specification controlling for the logarithm of market
capitalization but they do not regard these results as a test of HS as their focus is entirely
on comparing the two groups of stocks Nor do they investigate the impact of
concentration on either cash compensation or total pay Kang and Liu (2008) examine
pay-performance sensitivity use a version of the HS institutional concentration measure
as a control variable in a different framework deriving from Holmstrom and Tirole
(1993) Dikolli Kulp and Sedatole (2009) argue that firms with transient ownership will
issue relatively more equity incentives to counter short-termism with their results robust
to the inclusion of the HS concentration variable
We hoped to find supportive evidence from proxy voting behavior in support of HSs
findings However evidence from actual mutual fund proxy voting records suggests that
most oppose reductions in CEO pay but overall the evidence is mixed8 Davis and Kim
(2005) examine proxy votes of mutual funds for 2004 and find that for a sample of
Fortune 1000 proxy contests in 45 cases that attempted to limit executive pay mutual
funds were unanimously opposed to such attempts Rothberg and Lilien (2006) also
examine proxy-voting data and find that mutual funds voted 66 percent of the time in
managementrsquos favor on issues of executive compensation Ertimur Ferri and Muslu
(2011) find that institutional investors lack sophistication in that they do not target
excessively paid CEOs relative to CEOs with high predicted-pay based on economic
determinants but exert a moderating influence ndash a $23 million reduction ndash for those they
deem ldquoexcessively paidrdquo CEOs
III Data and Methodology
8 We do acknowledge that the HS results are for all institutional investors while the transparency with
respect to proxy voting applies only to mutual funds
12
This Section discusses the data used in the study the construction of variables
descriptive statistics and the methodology Two datasets are used ndash one is the original
dataset generously supplied by Jay Hartzell and Laura Starks which covers 1992‒1997
second a more comprehensive dataset covering 1992‒2002 is computed from the same
data sources used in the HS dataset This is Standard and Poorrsquos (SampP) ExecuComp for
compensation data Thomson Financial CDASpectrum database for institutional holdings
to compute measure of institutional influence SampPrsquos Compustat and the Center for
Research in Stock Prices (CRSP) for firm level accounting and stock price data Our more
comprehensive dataset contains 2559 firms compared with 1914 for the HS dataset Our
full matched sample yields 97679 executive years
A Institutional Influence Measures
To measure institutional influence we employ two ownership-based metrics The first
measure of institutional influence is HSrsquos Institutional Concentration We use this for
direct comparison to the HS analysis The second measure of institutional influence
changes the denominator of the HS measure from institutional shares held to total shares
outstanding in the firm with nomenclature Top 5 Ownership9 We compute both
measures at the fiscal year end of the firm We adjust all compensation measures and
asset prices for inflation to 30 June 2002 prices for our comprehensive sample compared
to the HS dataset that is in the dollars of the day
B Compensation Measures
The ExecuComp database reports several raw measures of executive compensation for
the top five executives (where available) including total salary bonus long-term
9 Top 5 Ownership is included in the HS database but not utilized by them in reported tests as well as
constructed by us for the more comprehensive sample
13
incentives and number of option grants For our more comprehensive dataset we replicate
all the compensation measures used by HS In common with HS we compute the pay-for-
performance sensitivity of option grants using the Yermack (1995) approach This
measure shows the dollar change in value of the executive options for every dollar change
in the value of the firm This involves calculating the option delta using the Black-
Scholes formula adjusted for dividends10
The ratio of the number of shares represented
by the award of options divided by the diluted number of shares outstanding at the
beginning of the year multiplies the option delta with units given by a $1000 change in
firm value Summing salary bonus long-term incentive plan payouts and stock and
option grants provides total direct compensation Options are valued using the Black and
Scholes (1973) formula for European call options adjusted by Merton (1973) to
incorporate dividends
C Descriptive Statistics
Table I presents descriptive statistics of key variables employed in the study according
to our comprehensive dataset11
Institutions own on average 546 percent of the firms in
which they invest This is almost identical to the 531 percent found by HS The
institutions with the five largest holdings account 45 percent of institutional ownership
Once again the figure for HS is almost identical at 44 percent The institutions with the
five largest holdings own on average 229 percent of shares outstanding in a firm (HS
have 223 percent) Average total compensation for executives is $2583 million
considerably larger than that documented by HS ($1250 million) but the HS figure is an
10 When there is more than one grant during the year a weighted average delta is used
11 With the data supplied to us by Hartzell and Starks we are able to replicate exactly their summary data
(HS Table I) for all but cash compensation Salary is included but bonus payments are missing The
replicated table is available on request
14
amalgam of price levels over their sample period 1992‒1997 and ours are in consistent
dollars of June 30 2002 However there has also been significant growth in
compensation levels from 1998 to 2002
ltltINSERT TABLE Igtgt
Based on our comprehensive data in Table II we examine correlations between the
key variables employed in the study Of concern is the large negative correlation of 63
percent which has been discussed previously (Section IIA) between the HS measure of
institutional concentration and the natural logarithm of firm market capitalization
ltltINSERT TABLE IIgtgt
In view of this high correlation Table III provides a breakdown of the three key pay
variables and three institutional ownership variables by size deciles We find that the high
institutional concentration but low institutional ownership share of total shareholdings
mentioned above is evident in the smallest decile of firms by size This decile has low
salary and even lower total compensation in a relative sense but by far the highest pay-
for-performance sensitivity of option grants Small firms award executives a far higher
proportion of market capitalization Since small firm market-capitalization is negligible in
comparison with the highest decile (300 times larger in magnitude) total compensation is
very small in absolute value Moreover due to the domination of the pay-size effect base
salaries are also exceedingly small
ltltINSERT TABLE IIIgtgt
Figure 1 based on Table III shows how as institutional concentration falls with
increases in stock size (but also as the total share of institutional ownership increases)
and the pay-for-performance sensitivity of option grants falls Thus because of what
appears to be a weak size control and size-influenced concentration measure HS
15
potentially pick up a size effect as their main result but not necessarily a monitoring
effect Figure 2 shows that as firm size increases up to decile 8 at least the share of
institutional total ownership increases as does total compensation
ltltINSERT FIGURES 1 AND 2 gtgt
D Methodology
Following HS we use two estimation techniques to examine the effect of institutional
investor influence on executive compensation Due to the censored nature of option grant
pay-for-performance sensitivity this compensation measure most appropriately uses
Tobit model specification whereas an Ordinary Least Squares (OLS) model specification
is used for all other compensation measures We model compensation measures as a
function of contemporaneous and lagged change in shareholder wealth institutional
influence Tobinrsquos Q and market capitalization Controls for CEO industry and year
effects are also used Augmentation occurs to test the effect of institutional influence on
the change in executive cash and total compensation with an interaction term that is the
product of institutional influence and the change in shareholder wealth
The variables used to model executive compensation have the following definitions
The lagged and contemporary measure of change in shareholder wealth is simply the
difference in market capitalization of the firm The measures of institutional influence
used are Institutional Concentration and Top 5 Ownership Use of the two types of
institutional influence occurs in separate models In order to replicate the HS
methodology models using Institutional Concentration have an additional control
consisting of lagged total institutional holdings and Tobinrsquos Q controls for performance
and growth opportunities Market capitalization of the firm is the number of shares
outstanding multiplied by the share price with inclusion both in levels and natural
16
logarithmic form A dummy variable is included to identify whether the executive is the
CEO Dummy variables are also included to control for industry effects This variable
takes a value of 1 for the two-digit Standard Industrial Classification (SIC) in which the
firm operates A year fixed effects (dummy) variable is included taking a value of 1 if the
observation was from the given year
IV Executive Incentives as a Function of Institutional Investor Influence
This section presents the results for the effect of institutional influence on option grant
pay-for-performance sensitivity and changes in compensation
A Option Grant Pay-for-Performance Sensitivity
Table IV presents the results for the effect of institutional influence on option grant
pay-for-performance sensitivity
ltltINSERT TABLE IVgtgt
Model (1) is a direct replication of HS using the HS database as supplied to us by HS
with all the variables computed in the same way and in the same functional form as with
HS The coefficient and t-statistic on lagged institutional concentration interacted with the
change in shareholder wealth is 13994 and 767 that compares to 1358 and 842 in HS
We are unable to account for these minor discrepancies for values that should be
identical In model (2) the only change to the model is that we use the natural logarithm
of market capitalization instead of the level of market capitalization This single change
has a sizeable effect on the coefficient on institutional concentration The coefficient on
institutional concentration is now -00692 and is insignificant The fact that an order
preserving transformation to a control variable can change the results is of interest The
log-transformation to market capitalization also changes the coefficient on Top 5
Ownership in models (3) and (4) The coefficient on Top 5 Ownership is significant in
17
both model (3) and in model (4) but the better size control reduces the coefficient by over
25 percent The better size control does not affect Top 5 Ownership by as much as its
design ensures less sensitivity to what could be inadequate size controls
We also test the effect of institutional influence on option grant pay-for-performance
sensitivity over the period 1992‒2002 This sample is then approximately three times as
large as the sample used by HS Institutional concentration is highly positively significant
in model (5) similar to the results for the period 1992‒1997 When we use the log of
market capitalization in model (6) the coefficient on institutional concentration remains
positive and significant at the 10 percent level with a t-statistic of 166 Of importance
though is that the coefficient on institutional concentration is one-eighth of the coefficient
in model (5) Even for a measure of institutional influence which we believe is less
sensitive to firm size the use of log market capitalization reduces the coefficient on Top 5
Ownership by approximately 50 percent from model (7) to model (8)
B Pay-for-Performance Sensitivity due to Changes in Cash and Total
Compensation
For robustness HS examine the effect that institutional influence conditioned on firm
performance has on a changes in cash and total executive compensation using the general
method of Jensen and Murphy (1990) to compute pay-for-performance sensitivity
ltltINSERT TABLE V gtgt
Table V presents the effect of institutional influence on the change in cash
compensation using the HS size control measure throughout12
Model (1) replicates the
specification in HS for the period of 1992‒1997 and we produce a similar result ndash the
interaction of lagged institutional concentration and contemporaneous change in
12 This test uses our data for the two periods 1992‒1997 and 1992‒2002 We do not use the HS data for
cash compensation as data for bonus is missing in the data supplied to us
18
shareholder wealth positively and significantly relates to a change in executive cash
compensation In model (2) we include the contemporaneous change in shareholder
wealth and the lagged institutional concentration as stand-alone controls as indicated by
Greenersquos (2000) specification of equation (1) above and the interaction term remains
significant In model (3) we interact lagged institutional concentration with the lagged
change in shareholder wealth as in equation (2) above that tries to take account of what
participants could have observed at the time The interaction term is then negative but
insignificant altering the HS result Model (4) provides a similar result where our
preferred proxy Top 5 Ownership captures concentrated institutional influence In
models (5) to (8) we use our full sample In model (5) we find that the lagged
institutional concentration interaction with the contemporaneous change in shareholder
wealth is positive and significant in HSrsquos specification However including change in
contemporaneous shareholder wealth and lagged institutional concentration as separate
control variables as suggested by equation (1)rsquos specification removes the significance
from the interaction term in model (6) Moreover in model (7) when lagged institutional
concentration is interaction with the lagged change in shareholder wealth in accord with
equation (2) this interaction term is now negative and significant reversing the HS
finding We obtain a similar result in model (8) when we measure institutional
concentration influence using Top 5 Ownership13
V The Levels of Executive Fixed and Total Compensation as Functions of
Institutional Investor Influence
13 In unreported results we use the log of market capitalization as a size control These changes to the size
control do not dramatically change the findings from Table V The reason why this size issue is not as
important as in the findings for option grants pay-for-performance is that the change in cash compensation
is not so strongly influenced by firm size In unreported results available on request we also examine the
effect of institutional influence on changes in total compensation with similar results to the change in cash
compensation
19
Two measures of compensation are examined ndash salary and total compensation
ltltINSERT TABLE VI gtgt
Table VI examines the effect of institutional influence on the level of salary In model
(1) we use the HS sample and specification to obtain almost identical results The
coefficient on institutional concentration is (negative) $247509 that is almost identical to
(negative) $247967 in HS Both are highly significant In model (2) we make a single
change to the model ndash we use the natural logarithm of market capitalization This single
change reverses the sign of institutional concentration with the coefficient still being
significant at the 1 percent level In models (3) and (4) we measure institutional influence
using Top 5 Ownership In model (3) where we measure firm size using the level of
market capitalization Top 5 Ownership is insignificant When we use the natural
logarithm of market capitalization in model (4) Top 5 Ownership is significant and
positively relates to the salary level In models (5) to (8) we estimate the same models
using our full sample We find similar results for the full sample as for the period 1992ndash
1997
ltltINSERT TABLE VII gtgt
Table VII examines the effect of institutional influence on total compensation In
model (1) we replicate the HS methodology using the HS data to obtain a similar
significant negative relation between institutional concentration and total compensation of
(negative) $1608633 However after changing to a logarithmic specification of market
capitalization in model (2) institutional concentration increases total compensation by a
significant $32400014
Importantly when institutional influence is measured using Top 5
14 = (5688751 times 0101)
20
Ownership an insignificant relation exists with total compensation in model (3) using the
HS size control but it becomes both positive and significant in (4) with the logged size
control In models (5) and (6) where models are estimated using the full sample we
observe a similar reversal in sign as occurs in models (1) and (2) However now the
magnitude of the reversal is almost equal with a difference of about $67m that is very
large indeed In model (5) the coefficient on institutional concentration is (negative)
$342m and then in model (6) the coefficient is $33m Top 5 Ownership is insignificant
in both models (7) and (8) unlike the results for Salary
VI Robustness
Two issues are examined in is this Section ndash reverse causality and the appropriate
transformation of market capitalization A methodological issue inherent in examining the
effect of ownership-based measures of institutional influence on executive compensation
is endogeneity driven by reverse causality We utilize an instrumental variables approach
where orthogonalized institutional influence measures are regressed on each of the
compensation measures lagged option grant pay-for-performance sensitivity salary and
total compensation In addition market capitalization and a CEO dummy are included in
each regression We use the residual from each regression as an instrument for
institutional influence Such a process removes any preferences for compensation from
the institutional influence measures
ltltINSERT TABLES VIII AND IXgtgt
Table VIII presents the results where we instrument the HS measure lagged
institutional concentration with a compensation-orthogonalized equivalent The analysis
indicates that the instrumented institutional concentration measure is positively and
significantly related to the level of salary and total compensation and insignificantly
21
related to option grant pay-for-performance sensitivity (PPSOG) Table IX presents
results where we instrument Top 5 Ownership with a compensation-orthogonalized
equivalent The instrumented Top 5 Ownership measure insignificantly relates to option
grant pay-for-performance sensitivity and barely negatively relates to salary although it
negatively relates to total compensation suggesting reverse causality is driving a
significant portion of our previous results
ltltINSERT TABLE Xgtgt
Another issue examined in this section is the appropriate transformation of market
capitalization Aneuryn-Evans and Deaton (1980) provide an evaluation of methods
comparing linear and log-linear specifications We use the Box-Cox test to compare
logarithm versus linear specifications In applying the Box-Cox test we specify
transformation of only market capitalization ndash the dependent variable and all other
independent variables remain unchanged Table X presents the lambda coefficient for
each model and the log likelihood score In all models in Panel A and B a lambda of zero
receives a higher (closer to zero) log likelihood score suggesting the log specification is
appropriate
VII Conclusion
Hartzell and Starks (2003) appears to document an effective role for institutional
investors in increasing executive incentives and lowering compensation levels Our paper
re-addresses the issue of concentrated institutional monitoring from a perspective
informed by an additional decade of research since their paper appeared We cannot find
evidence that concentrated institutional investors really have any positive (or negative for
that matter) effect on executive incentive levels While superficially we find evidence
that they may raise executive pay levels this is most likely driven by an institutional
22
preference for highly remunerated executives in large companies and thus by reverse
causality Moreover our findings are not consistent with the rent-extraction hypothesis
suggesting that firms systematically overpay executives in the absence of effective
monitoring by outside parties such as institutional investors
23
References
Almazan Andres Jay C Hartzell and Laura T Starks 2005 ldquoActive Institutional
Shareholders and Costs of Monitoring Evidence from Executive Compensationrdquo
Financial Management Winter 5ndash34
Aneuryn-Evans Gwyn and Angus Deaton 1980 ldquoTesting Linear versus Logarithmic
Regression Modelsrdquo Review of Economic Studies 47 (1) Econometrics Issue 275ndash
291
Babenko Ilona 2009 ldquoShare Repurchases and Pay-Performance Sensitivity of Employee
Compensation Contractsrdquo Journal of Finance 64117-150
Baker George P and Brian J Hall 2004 ldquoCEO Incentives and Firm Sizerdquo Journal of
Labor Market Economics 22 767ndash798
Bebchuk Lucian A Jesse M Fried and David Walker 2002 ldquoManagerial power and
rent extraction in the design of executive compensationrdquo University of Chicago Law
Review 69 751ndash846
Bebchuk Lucian A and Jesse M Fried 2003 ldquoExecutive Compensation as an Agency
Problemrdquo Journal of Economic Perspectives 17 71ndash92
Bebchuk Lucian A and Jesse M Fried 2004 Pay without Performance The Unfulfilled
Promise of Executive Compensation Cambridge MA Harvard University Press
Bertrand Marianne and Sendhil Mullainathan 2001 ldquoAre CEOs Rewarded for Luck
The Ones Without Principals Arerdquo Quarterly Journal of Economics 116 901ndash32
Black Fischer and Myron Scholes 1973 ldquoThe Pricing of Options and Corporate
Liabilitiesrdquo Journal of Political Economy 81 637ndash54
Cadman Brian Sandy Klasa and Steve Matsunaga 2010 ldquoDeterminants of CEO Pay A
Comparison of ExecuComp and Non-ExecuComp Firmsrdquo The Accounting Review 85
1511ndash1543
Chhaochharia Vidhi and Yaniv Grinstein 2009 ldquoCEO Compensation and Board
Structurerdquo Journal of Finance 64 231ndash261
Cichello Michael S 2005 ldquoThe impact of firm size on pay-performance sensitivitiesrdquo
Journal of Corporate Finance 11 609ndash627
Davis Gerald F and E Han Kim 2007 ldquoBusiness ties and proxy voting by mutual
fundsrdquo Journal of Financial Economics 85 552ndash570
Dikolli Shane S Susan L Kulp and Karen L Sedatole 2009 ldquoTransient Institutional
Ownership and CEO Contractingrdquo Accounting Review 84 737ndash770
Edmans Alex Xavier Gabaix and Augustin Landier 2009 ldquoA Multiplicative Model of
Optimal CEO Incentives in Market Equilibriumrdquo Review of Financial Studies 22
4881ndash4917
Edmans Alex 2009 ldquoBlockholder Trading Market Efficiency and Managerial Myopiardquo
Journal of Finance 64 2481ndash2513
Ertimur Yonca Fabrizio Ferri and Volkan Muslu 2011 ldquoShareholder Activism and
CEO Payrdquo Review of Financial Studies 24 535ndash592
24
Gabaix Xavier and Augustin Landier 2008 ldquoWhy Has CEO Pay Increased So Muchrdquo
Quarterly Journal of Economics 123 49ndash100
Garen John 1994 ldquoExecutive compensation and principal-agent theoryrdquo Journal of
Political Economy 102 1175ndash1199
Gibbons Robert and Kevin J Murphy 1992 ldquoOptimal Incentive Contracts in the
Presence of Career Concerns Theory and Evidencerdquo Journal of Political Economy
100 468ndash505
Greene William H 2000 (4th
Edition) Econometric Analysis Prentice-Hall NJ
Guthrie Katherine Jan Sokolowsky and Kam-Ming Wan 2012 ldquoCEO Compensation
and Board Structure Revisitedrdquo Journal of Finance Forthcoming
Hadlock Charles R and Gerald B Lumer 1997 ldquoCompensation turnover and top
management incentives historical evidencerdquo Journal of Business 70 153ndash187
Hartzell Jay C and Laura T Starks 2003 ldquoInstitutional Investors and Executive
Compensationrdquo Journal of Finance 58 2351ndash2374
Huson Mark R Robert Parrino and Laura T Starks 2001 ldquoInternal Monitoring
Mechanisms and CEO Turnover A Long-Term Perspectiverdquo Journal of Finance 56
2265ndash2297
Jensen Michael C and Kevin J Murphy 1990 ldquoPerformance Pay and Top Management
Incentivesrdquo Journal of Political Economy 98 225ndash264
Jin Li 2002 ldquoCEO compensation diversification and incentivesrdquo Journal of Financial
Economics 66 29ndash63
Kang Qiang and Qiao Liu 2008 ldquoStock Market Information Production and Executive
Incentivesrdquo Journal of Corporate Finance 14 484ndash498
Merton Robert 1973 ldquoTheory of Rational Option Pricingrdquo Bell Journal of Economics
and Management Science 4 141-183
Morse Adair Vikram Nanda and Amit Seru 2011 ldquoAre Incentive Contracts Rigged by
Powerful CEOsrdquo Journal of Finance 66 1779ndash1821
Murphy Kevin J 1985 ldquoCorporate performance and managerial remuneration An
empirical analysisrdquo Journal of Accounting and Economics 7 11ndash42
Murphy Kevin J 1999 ldquoExecutive compensationrdquo in Orley Ashenfelter and David
Card Handbook of Labor Economics vol 3 Elsevier
Rosen Sherwin 1992 ldquoContracts and the market for executivesrdquo In Contract
Economics ed Lars Werin and Hans Wijkander 181ndash211 Cambridge MA
Blackwell
Rothberg Burton and Steven Lilien 2006 ldquoMutual Funds and Proxy Voting New
Evidence on Corporate Governancerdquo Journal of Business and Technology Law 1
157ndash184
Schaefer Scott 1998 ldquoThe dependence of payndashperformance sensitivity on the size of the
firmrdquo Review of Economics and Statistics 80 436ndash443
Yermack David 1995 ldquoDo Corporations Award CEO Stock Options Effectivelyrdquo
Journal of Financial Economics 39 237ndash69
25
Table I Univariate Descriptive Statistics
This table reports sample statistics for the key variables used in this study based on our comprehensive
ExecuComp dataset Institutional Concentration is the shares owned by the institutions with the five largest
holdings divided by the shares held by all institutions Top 5 Ownership is the proportion of shares
outstanding held by the institutions with the five largest holdings Total Institutional Ownership is the
proportion of shares held by all institutions Panel B shows the measures executive compensation The pay-
for-performance sensitivity (PPS) of option grants shows the change in executive wealth due to a $1000
change in firm in shareholder wealth Expressions for Salary and Total Compensation are in thousands of
dollars while Market Capitalization is in millions of dollars Panel C shows firm characteristics
Variable Mean Median Std Dev P10 P90
Panel A Institutional Investor Holdings
Institutional Concentration 0450 0424 0151 0282 0656
Top 5 Institutional Ownership 0229 0220 0101 0110 0355
Total Institutional Ownership 0546 0561 0206 0258 0801
Panel B Executive Compensation
PPS of Option Grants 1161 0156 3829 0 2822
Salary 36311 29462 25231 15179 67000
Total Compensation 258313 85878 980701 23875 511022
Panel C Firm Characteristics
Tobins Q 1744 1139 2738 0495 3322
Market Capitalization 560699 112094 1988108 19754 1048754
26
Table II Pairwise Correlation Coefficients
This table presents the correlations between the key variables based on our comprehensive
ExecuComp dataset Institutional Concentration is the shares owned by the institutions with the five
largest holdings divided by the shares held by all institutions Top 5 Ownership is the proportion of
shares outstanding held by the institutions with the five largest holdings Total Institutional
Ownership is the proportion of shares held by all institutions The pay-for-performance sensitivity
(PPS) of option grants is the change in executive wealth due to a $1000 change in shareholder
wealth
(1) (2) (3) (4) (5) (6) (7) (8)
(1) Institutional Concentration 1
(2) Top 5 Ownership 0268 1
(3) Total Ownership -0415 0706 1
(4) PPS of Option Grants 0122 0042 -0024 1
(5) Log Salary -0270 -0012 0171 0038 1
(6) Log Market Capitalization -0630 -0157 0261 -0156 0420 1
(7) Level Market Capitalization -0241 -0157 0007 -0058 0219 0534 1
(8) Tobins Q -0008 -0034 0002 0015 -0083 0066 0121 1
27
Table III
Descriptive Statistics by Market Capitalization Deciles
Market
Capitalization
Deciles
Market
Capitalization
($m)
Total
Compensation
($000)
Salary
($)
PPS Option
Grants
($)
Institutional
Concentration
()
Top 5 Ownership
()
Institutional
Total Ownership
()
1 (Small) 1181 76008 234700 2720 611 200 327
2 2668 90206 257955 1796 495 217 438
3 4193 105946 275120 1517 454 220 485
4 6045 129892 296067 1376 420 206 491
5 8742 155391 315864 1216 394 198 502
6 13086 191888 345942 1072 380 202 531
7 20790 239501 377021 0874 368 196 532
8 34920 314955 427808 0687 354 190 538
9 67747 450970 480319 0569 328 173 528
10 (Big) 366540 846066 614354 0297 275 138 502
28
Table IV Pay-for-Performance Sensitivity of Option Grants as a Function of Institutional Influence
Similar in construction to parts of HS (Table II) this table shows the coefficients from Tobit regressions of option grant pay-for-performance sensitivity per $1000 change in
shareholder wealth against the lagged change and contemporary change in shareholder wealth measures of institutional influence Tobinrsquos Q firm market capitalization and
dummy controls for CEO and year effects We estimate models (1) to (4) over the period 1992‒1997 using data supplied by HS and models (5) to (8) are estimated over the
whole sample Models (1) (2) (5) and (6) use Institutional Concentration as a measure of institutional influence which HS define as the shares owned by the institutions
with the five largest holdings divided by the shares held by all institutions These models control for total institutional ownership Models (3) (4) (7) and (8) use Top 5
Ownership as a measure of institutional influence which we define as the proportion of shares outstanding held by the institutions with the five largest holdings Coefficients
for industry and year controls are not presented One two and three asterisks denote significance at 10 5 and 1 levels
Hartzell and Starks Database 1992‒1997 Full Sample from ExecuComp 1992‒2002
(1) (2) (3) (4) (5) (6) (7) (8)
Change in Shareholder Wealth 77721 191781 89874 137368 -446 -218 -664 -203
(171) (429) (194) (317) (319) (168) (453) (158)
Lag1 Change in Shareholder Wealth 14446 99963 88719 8544 -144 0273 -0342 00852
(027) (212) (166) (184) (071) (015) (016) (005)
Lag1 Institutional Concentration 13994 -00692 26405 03304
(767) (032) (1402) (166)
Lag1 Total Institutional Ownership 08055 10434 07106 0937
(593) (781) (666) (898)
Lag1 Top 5 Institutional Ownership 10567 07457 12116 06293
(399) ( 283) (533) (278)
Lag1 Tobinrsquos Q 0068 00877 00647 00813 00498 00727 00508 00695
(337) (434) (319) (401) (504) (734) (51) (703)
Lag1 Market Capitalization -106202 -173515 -776 -117
(474) (742) (1011) (153)
Lag1 Log Market Capitalization -0297 -02319 -04111 -03908
(1576) (1439) (2997) (2841)
CEO Dummy 13745 13682 13765 13674 28197 28128 28155 28143
(1913) (1918) (1914) (1916) (3388) (3396) (3377) (34)
Number of Observations 33878 33878 33878 33878 97679 97679 97679 97679
29
Table V Pay-for-Performance Sensitivity of Cash Compensation (Salary plus Bonus) as a Function of Institutional Influence - Level Market Capitalization
Similar in parts to models (1) and (2) of HS (Table IV) this table displays the coefficients from OLS regressions of the change in cash compensation against the lagged
change and contemporary change in shareholder wealth measures of institutional influence Tobinrsquos Q firm market capitalization and dummy controls for CEO and year
effects We estimate models (1) to (4) over the period 1992‒1997 based on our ExecuComp sample and models (5) to (8) over the whole sample Models (1) (2) (5) and (6)
use Institutional Concentration as a measure of institutional influence which HS define as the shares owned by the institutions with the five largest holdings divided by the
shares held by all institutions These models control for total institutional ownership Models (4) and (8) use Top 5 Ownership as a measure of institutional influence which
we define as the proportion of shares outstanding held by the institutions with the 5 largest holdings Coefficients for industry and year controls are not presented One two
and three asterisks denote significance at 10 5 and 1 levels
Our Database based on ExecuComp 1992‒1997 Full Sample based on ExecuComp 1992‒2002
(1) (2) (3) (4) (5) (6) (7) (8)
Lag1 Change in Shareholder Wealth 00014 00017 00088 00089 -00006 -00007 00072 00086
(051) (062) (129) (103) (027) (029) (131) (16)
Change in Shareholder Wealth -00102 00102 00105 00032 00102 00101
(097) (292) (293) (055) (611) (608)
Change Shareholder Wealth Lag1 Institutional
Concentration
00435 00764 00363 00253
(463) (258) (689) (144)
Lag1 Change Shareholder Wealth Lag1
Institutional Concentration
-00197 -00283
(122) (197)
Lag1 Institutional Concentration -492182 -288861 400149 453623
(09) (056) (098) (114)
Lag1 Total Institutional Ownership 720688 538372 655091 559587 682069 726226
(276) (232) (281) (314) (308) (322)
Lag1 Change Shareholder Wealth Lag1 Top 5
Institutional Ownership
-0046 -00689
(091) (256)
Lag1 Top 5 Institutional Ownership 583044 1191634
(092) (25)
Lag1 Tobinrsquos Q -32761 -36806 -19422 -18023 -54283 -48681 -29534 -24243
(138) (172) (077) (078) (227) (245) (136) (111)
Lag1 Market Capitalization 00004 00004 00002 00006 00007 00008 00007 00007
(042) (044) (022) (07) (086) (092) (082) (09)
CEO Dummy 768035 765474 768334 773634 616315 61729 619028 618535
(354) (355) (356) (355) (417) (418) (419) (418)
Adjusted R-Sq 00057 0006 00052 0005 00097 00097 00098 00102
Number of Observations 41305 41305 41305 41305 85518 85518 85518 85518
30
Table VI Level of Salary as a Function of Institutional Influence
Similar in parts to models (1) and (2) of HS (Table V) this table displays the coefficients from OLS regressions of salary against the lagged change and contemporary change in
shareholder wealth measures of institutional influence Tobinrsquos Q firm market capitalization and dummy controls for CEO industry and year effects We estimate models (1) to
(4) over the period 1992‒1997 using the actual data supplied by HS and models (5) to (8) over the whole sample 1992‒2002 Models (1) (2) (5) and (6) use Institutional
Concentration as a measure of institutional influence which HS define as the shares owned by the institutions with the five largest holdings divided by the shares held by all
institutions These models control for total institutional ownership Models (3) (4) (7) and (8) use Top 5 Ownership as a measure of institutional influence which we define as
the proportion of shares outstanding held by the institutions with the 5 largest holdings Coefficients for industry and year controls are not presented One two and three
asterisks denote significance at 10 5 and 1 levels
Hartzell and Starks Database 1992‒1997 Full Sample based on ExecuComp 1992‒2002
(1) (2) (3) (4) (5) (6) (7) (8)
Change in Shareholder Wealth 04125E-4 0003 -2439E-4 0003 00021 00012 00024 00012
(057) (464) (031) (468) (1004) (72) (1041) (706)
Lag1 Change in Shareholder Wealth -00062 7856E-4 -0007 9003E-4 -00014 00002 -00014 00003
(594) (098) (592) (112) (522) (123) (5) (145)
Lag1 Institutional Concentration -2475087 555134 -2734983 943317
(3245) (710) (4965) (1671)
Lag1 Total Institutional Ownership 1020109 217488 936337 252464
(1904) (432) (2471) (704)
Lag1 Top 5 Institutional Ownership -148355 31354 -87926 423166
(139) (332) (118) (643)
Lag1 Tobinrsquos Q -195348 -252191 -192344 -25158 -119837 -146373 -122132 -143821
(2812) (3762) (2721) (3752) (248) (274) (2583) (2727)
Lag1 Market Capitalization 844E-06 103E-5 00034 00039
(2781) (3051) (2874) (3104)
Lag1 Log Market Capitalization 750021 723625 776657 732279
(7511) (9285) (11224) (14039)
CEO Dummy 2683978 2702238 2668244 2702417 3281133 3285103 3292175 3283145
(8625) (9277) (8246) (9271) (12739) (13616) (12336) (13605)
Adjusted R-Sq 04604 05302 04187 05295 03856 04588 03476 04574
Number of Observations 36280 36280 36280 36280 97679 97679 97679 97679
31
Table VII Level of Total Direct Compensation as a Function of Institutional Influence
Similar in parts to models (3) and (5) of HS (Table V) this table displays the coefficients from OLS regressions of total direct compensation against the lagged change and
contemporary change in shareholder wealth measures of institutional influence Tobinrsquos Q firm market capitalization and dummy controls for CEO industry and year effects
We estimate models (1) to (4) over the period 1992‒1997 and models (5) to (8) over the whole sample Models (1) (2) (5) and (6) use Institutional Concentration as a measure
of institutional influence which HS define as the shares owned by the institutions with the five largest holdings divided by the shares held by all institutions These models
control for total institutional ownership Models (3) (4) (7) and (8) use Top 5 Ownership as a measure of institutional influence which we define as the proportion of shares
outstanding held by the institutions with the five largest holdings Coefficients for industry and year controls are not presented One two and three asterisks denote significance
at 10 5 and 1 levels
Hartzell and Starks Database 1992‒1997 Full Sample from ExecuComp 1992‒2002
(1) (2) (3) (4) (5) (6) (7) (8)
Change in Shareholder Wealth 005779 007006 005282 006791 00423 00148 0046 00138
(415) (581) (369) (565) (203) (07) (22) (066)
Lag1 Change in Shareholder Wealth 003182 00714 00272 00715 00613 0126 00614 01278
(126) (298) (107) (299) (275) (488) (276) (494)
Lag1 Institutional Concentration -1608633 3248998 -3424892 3334186
(151) (283) (1172) (118)
Lag1 Total Institutional Ownership 104013 5473128 1531013 -2926764
(1645) (831) (876) (158)
Lag1 Top 5 Institutional Ownership 188585 3927986 -2159177 -2640879
(016) (343) (067) (092)
Lag1 Tobinrsquos Q -445692 -803612 -454859 -833566 3257027 2971201 3213936 3086205
(427) (791) (432) (831) (83) (77) (815) (793)
Lag1 Market Capitalization 49E-5 615E-5 01031 01088
(113) (1365) (1555) (168)
Lag1 Log Market Capitalization 4683497 4759377 1645156 1427856
(2775) (3469) (3679) (4239)
CEO Dummy 1501307 1508886 1493688 1508736 4713079 4713629 47278 4703908
(2648) (2684) (2618) (2683) (282) (2828) (2819) (2824)
Adjusted R-Sq 01861 02060 01695 02051 01071 01113 01033 011
Number of Observations 33229 33229 33229 33229 97679 97679 97679 97679
32
Table VIII Two Stage Least Squares (Reverse Causality) Tests - Executive Compensation Measures as a Function of Institutional Concentration
This table shows the coefficients from the Two-Stage Least Squares (2SLS) regressions of executive compensation measures against the lagged change
and contemporary change in shareholder wealth instrumented institutional concentration and total institutional ownership Tobinrsquos Q the natural
logarithm of firm market capitalization and dummy variable controls for CEO and year effects We instrument measures of institutional concentration
and total institutional ownership by orthogonalizing with respect to option grant pay-for-performance sensitivity salary and total compensation
Coefficients for industry and year controls are not presented First stage results are not presented One two and three asterisks denote significance at
10 5 and 1 levels
ExecuComp Sample 1992‒1997 Full Sample from ExecuComp 1992‒2002
PPSOG Salary TC PPSOG Salary TC
Change in Shareholder Wealth 144E-05 0004 02133 -409E-06 00011 00157
(308) (775) (596) (691) (638) (068)
Lag1 Change in Shareholder Wealth -240E-05 -00004 00866 118E-06 00002 01422
(131) (087) (346) (097) (116) (51)
Lag1 Institutional Concentration -00898 668681 3263143 02024 916656 2856417
(052) (801) (176) (163) (1545) (1004)
Lag1 Total Institutional Ownership 01021 -43997 2512587 02983 -104171 -1028347
(09) (084) (153) (4) (269) (487)
Lag1 Tobinrsquos Q 00754 -225606 119297 00655 -158657 2997205
(596) (1649) (053) (834) (2535) (727)
Lag1 Log Market Capitalization -04234 828713 7864046 -04002 825106 1685638
(2195) (8132) (2977) (4106) (11133) (373)
CEO Dummy 15848 3214117 2950274 19149 3205862 4529101
(2223) (9389) (2255) (3802) (1331) (2758)
Overall R-Sq -- 04981 01206 -- 04738 01178
Number of Observations 41305 41305 41305 85518 85518 85518
33
Table IX Two Stage Least Squares (Reverse Causality) Tests - Executive Compensation Measures as a Function of Top 5 Ownership
This table shows the coefficients from 2SLS regressions of executive compensation measures against the lagged change and contemporary change
in shareholder wealth instrumented Top 5 institutional ownership Tobinrsquos Q natural logarithm of firm market capitalization and dummy variable
controls for CEO and year effects We instrument Top 5 institutional ownership by orthogonalizing with respect to option grant pay-for-
performance sensitivity salary and total compensation Coefficients for industry and year controls are not presented First stage results are not
presented One two and three asterisks denote significance at 10 5 and 1 levels
ExecuComp sample 1992‒1997 Full Sample 1992‒2002
PPSOG Salary TC PPSOG Salary TC
Change in Shareholder Wealth 136E-05 00041 02128 -407E-06 00011 00146
(293) (79) (596) (692) (619) (063)
Lag1 Change in Shareholder Wealth -241E-05 -00003 00868 116E-06 00003 01439
(133) (07) (347) (096) (142) (514)
Lag1 Top 5 Institutional Ownership -03774 -187512 -3848531 00179 -104261 -126659
(163) (186) (166) (011) (149) (436)
Lag1 Tobinrsquos Q 00731 -223886 10074 00642 -155467 3115073
(573) (1631) (045) (819) (2509) (75)
Lag1 Log Market Capitalization -0416 775637 7696302 -04015 764632 1462536
(2345) (10084) (3766) (3688) (13708) (4072)
CEO Dummy 15854 3211991 2950158 19152 3203152 4519385
(2223) (9381) (2254) (3804) (133) (2752)
Overall R-Sq -- 04967 01202 -- 04722 01168
Number of Observations 41305 41305 41305 85518 85518 85518
34
Table X Specification Tests utilizing Box-Cox Transformation of Market Capitalization
This table presents results from a Box-Cox transformation of market capitalization In
each model we specify transformation of market capitalization only Panel A shows
results for the period 1992‒1997 and Panel B shows results for the entire period
1992‒2002 The first part of each panel shows the Lambda coefficient that applies to
the transformed market capitalization The second part of each panel shows log
likelihood scores with LR Chi2 Statistics in parentheses ‒ these test for the
appropriateness of a linear or log model specification
Panel A Our ExecuComp sample 1992‒1997
(1) (2) (3)
PPSOG Salary TC
Lambda -05143 01881 02876
(-1914) (2413) (1329)
Test H0 (Log Likelihood Score)
Lambda = -1 -1279748 -3178575 -4795326
(23866) (85952) (99688)
Lambda = 0 -1280152 -3138456 -4791159
(31934) (57147) (16339)
Lambda = 1 -1283157 -3163807 -4793506
(92039) (564156) (63292)
Panel B Full ExecuComp sample 1992‒2002
(1) (2) (4)
PPSOG Salary TC
Lambda -044 01853 04036
(-3015) (40) (3666)
Test H0 (Log Likelihood Score)
Lambda = 0 -2688334 -6476494 -10384490
(66599) (160356) (126236)
Lambda = 1 -2694886 -6538506 -10386795
(197644) (1400605) (172345)
35
Figure 2 Institutional Ownership Measures (LHA) vs Total
Compensation (RHA) by Market Capitalization
0
10
20
30
40
50
60
70
1 2 3 4 5 6 7 8 9 10
$00
$10
$20
$30
$40
$50
$60
$70
$80
$90
Millio
ns
InstitutionalConcentration (LHS)Top5 Ownership (LHS)
Institutional TotalOwnership (LHS)Total Compensation(RHS)
Page 2
Electronic copy available at httpssrncomabstract=1009309
2
With 128 cites in Web of Science and 757 cites in Google Scholar Hartzell and Starks
(2003) [hereafter HS] ranks among the most influential corporate finance papers of the
last decade HS find that concentrated institutional monitors provide greater incentives for
executives to perform by granting more options while simultaneously lowering their cash-
based pay with an even greater fall in total direct compensation HS (p2355) define
concentrated institutional holdings [hereafter Top 5 Concentration] by the proportion of
institutional ownership accounted for by the top five institutional investors12
By
contrast our paper argues that several methodological issues in HS alter their finding by
showing that concentration has either no effect on incentives or perhaps reduces them
and in the case of salary and total pay levels reverse their findings Our paper discusses
these issues and then empirically demonstrates the effect that they have on the findings of
HS
HSrsquos findings are supportive of the idea that firms overpay executives as the
skimming or rent-seeking view of executive compensation maintains See for example
Bertrand and Mullainathan (2001) Bebchuk Fried and Walker (2002) Bebchuk and
Fried (2003 2004) Chhaochharia and Grinstein (2009) and Morse Nanda and Seru
(2011) Chhaochharia and Grinstein (2009) find that increased board requirements
reduced CEO pay but note the fragility qualification provided by Guthrie Sokolowsky
and Wan (2012) and Morse Nanda and Seru (2011) find that powerful CEOs rig
performance measures to increase their pay and that these firms then underperform
1 More specifically HS (p2352) find that an increase of one standard deviation in Top 5 institutional
investor concentration is associated with a 20 percent increase in option grant pay-for-performance
sensitivity and a 19 percent drop in total compensation for the average executive
2 HS also obtain similar findings using a Herfindahl index of institutional fractional holdings that they
construct in a very similar fashion to their main concentration measure Our findings for their concentration
measure will also tend to be true of their Herfindahl measure
Electronic copy available at httpssrncomabstract=1009309
3
We find that Institutional Top 5 Concentration according to the HS definition does
not raise option grant pay-for-performance sensitivity as it has no statistically significant
impact either way once one introduces a standard logarithmically transformed size
control Similarly lagged Institutional Concentration has a negative but statistically
insignificant effect on the pay-for-performance sensitivity of cash (salary plus bonus)
compensation when interacted with the lagged change in shareholder wealth HS obtain a
significant positive coefficient using the contemporaneous change in shareholder wealth
but one need replace this by the lagged change in shareholder wealth in keeping with
what influential institutional investors knew at the time Additionally HS find that lagged
concentrated institutional ownership sizably lowers both salary and total direct
compensation and this is statistically significant but we find the reverse ‒ it appears to
raise both and the findings are highly statistically significant also following a standard
logarithmically transformed size control
Upon closer examination several aspects of the HS research design are of interest
First the primary argument in our paper is that the use by HS of the lagged value of the
level of market capitalization as the size control in regressions (HS p2359) rather than
the conventional lagged logarithm of capitalization distorts the coefficient on the
measure of institutional influence3 A number of authors have drawn attention to falling
pay-for-performance sensitivities as firm size increases (eg Jensen and Murphy (1990)
Garen (1994) Hadlock and Lumer (1997) Schaefer (1998) Murphy (1999) Jin (2002)
and Chichello (2005)) and hence the need to have effective size controls In the HS
dataset the raw mean market capitalization ranges from $1181m in decile 1 to $36654m
in decile 10 (see Table III below) a difference of 3104 fold reducing to only 22 with the
3 HS (p2357) also use net assets and total assets as controls but in keeping with the use of raw
capitalization retain them in their raw form
4
logarithmic transformation While this skewed nature of the level of market capitalization
does not per se rule out its use specification tests such as Box-Cox and the replacement
of market capitalization by its cumulative distribution (see Section VI below) indicates a
possible specification error Moreover using the lagged natural logarithm of market
capitalization as a firm size control in the actual HS data with the same dependent
variable and no other changes raises the explanatory power (R-Squared) of the HS model
by between 135 and 1516 percent4
Given that firm size is an important determinant of executive compensation selecting
the correct functional form for firm size is a vexing and important issue Many estimates
including those of Murphy (1985) Rosen (1992) Gibbons and Murphy (1992) Huson
Parrino and Starks (2001) Baker and Hall (2004) Gabaix and Landier (2008) Edmans
Gabaix and Landier (2009) Babenko (2009) and Aggarwal Erel and Ferreira (2011)
use a logarithmic specification with the elasticity between firm size and executive total
pay in the range of 03 and 04 HS seems to employ inadvertently a firm size measure in
levels (also alternative firm size measures also in levels) and consequently imply an
elasticity value (exponent on market capitalization) of unity
Second we question the use of the HS Top 5 Concentration measure of institutional
influence based on the influence of concentrated institutional investors relative to their
peers that appears to lead inadvertently to an exacerbation of the size-control problem In
their companion theoretical model AHS arrive at a different and what might seem a more
natural specification namely Top 5 Ownership which measures concentration relative to
all shareholders not just institutional peers The following example illustrates the
4 The lower result is based on OLS regressions of model 1 of Table II of HS with the R-Squared rising from
00853 to 00968 once Market Capitalization is logged and the higher result model 1 of Table V of HS is
based on an improvement in R-Squared from 04604 to 05302 (see Table VI below)
5
differences between these two methods If there are one hundred shares outstanding in a
small company and the only institution owns one share that company is concentrated
according even though that institution may lack effective monitoring power Instead if
one deflates by shares on issue this results in a concentration ratio of 1100 If on the
other hand ten institutions collectively own 500 shares in a much larger company of
1000 shares and the Top Five collectively own 250 then the HS concentration ratio falls
to only 250500 or 05 and utilizing shares on issue 025 in this case As is confirmed by
the data one expects Top 5 Concentration to correlate negatively with firm size and Top
5 Ownership the reverse given a positive association between institutional ownership and
company size We then re-examine the monitoring hypothesis using this AHS proposed
measure of institutional investor influence that intuitively relates better to institutional
proxy voting power
Third we address issues of reliance on potentially biased self-reported values of option
grants that result in the exclusion of 16 percent of the HS sample size for tests on total
(direct) compensation Forth we argue that tests that HS perform using an alternative
measure of pay-for-performance sensitivity to provide support for the main hypothesis of
the study require one to get the timing right in terms of the lag structure of performance
and concentration Finally we address issues associated with their tests of endogeneity
and reverse causality These issues collectively might suggest overlooking of a
monitoring effect with the main driving forces size related
The next section reviews the HS study in an effort to tease out what might really be
going on Section II examines the literature for studies that in replicating HS provide
independent support for their findings It also reports findings on the actual proxy voting
record of mutual funds and actions of institutional investors in reducing pay Section III
describes the data and methodology Section IV analyzes the relation between
6
institutional influence and measures of pay-for-performance sensitivity Section V
documents the relationship between institutional influence and total and fixed
compensation Section VI examines the robustness of our findings and the final Section
concludes
I Review of Hartzell and Starks Methodology
We comment on five areas of the HS methodology 1) Size control 2) Measurement of
institutional influence 3) Testing the Jensen-Murphy measure of incentives 4) Reliance
on incomplete self-reported valued for option grants and 5) Endogeneity
A Size Control
In our paper we argue that HS unintentionally identify incentive-size and pay-size
effects5 not a monitoring effect It can be seen in Figure II that executive total pay for
HSrsquos Top 5 institutional concentration measure rises from as little as $072m for the
smallest market capitalization decile to $85m for the highest This twelve-fold pay-size
effect in the deciles can cause problems if one utilizes an untransformed size measure
Due to skewness in the size measure using this in its untransformed state appears to
weaken its effectiveness leaving open the possibility that correlations they identify
between incentives and pay levels on the one hand and Top 5 Concentration on the other
are due largely to size-related effects
Table III displays this strong pay-size effect based on our full 1992‒2002 dataset and
graphically in Figure 1 It shows that the incentive measure adopted by HS declines from
$272 per $1000 change in shareholder wealth for the smallest market capitalization
decile of firms to approximately $030 for the largest Table III and Figure 1 show the
5 The multiplicative model of Edmans Gabaix and Landier (2009) finds that incentive pay increases with
firm size (elasticity value of 13) and the empirical elasticity is 037
7
strong relation between institutional concentration as measured by HS and market
capitalization For the smallest decile of market capitalization firms institutional
concentration averages 61 percent declining to about 275 percent for the highest decile
This negative relationship between firm size and institutional concentration seems to arise
because institutional investors do generally favor small firms This causes problems
because in the limit a small company may have only one institutional investor owning
only a small percentage of the company Hence it would appear to be fully concentrated
even though the ability of the institutional investor to influence pay decisions might be
small Thus it is not at all surprising that the correlation between the logarithm of market
capitalization and their institutional concentration measure is both negative and high in
absolute value at 63 percent Their methodology appears to favor small companies that
Table III shows to have low pay levels and high pay for performance sensitivity
Moreover if one does not transform the raw size measure it implicitly assumes that a
given dollar change in firm size has the same absolute impact on pay for a very large
company as it does for a small company Indeed the conventional logarithmic
specification for firm size provides proportionate effects that imply a pay-size elasticity of
approximately one-third instead of unity6
B Measurement of Institutional Influence
We argue that there are additional issues related to the HS measure of institutional
influence ndash institutional concentration In their companion article to HS AHS develop a
model where they capture institutional influence by a single large long-term institutional
6 Note that unlike an earlier version of our paper we leave the HS dependent variables in levels and hence
untransformed While we do this to minimize the alterations that we make to the HS specification ideally
the elasticity specification is preferable especially for the compensation variables Transforming the
dependent variables does not significantly affect the results
8
investor They measure influence relative to total shareholding including private or
atomistic shareholders Presumably they do this to capture relative activism and proxy
voting strength However when it comes to the empirical implementation of the model
AHS adopt the same relative institutional concentration measure as in HS (AHS (p12))
potentially downplaying the relative strength of institutions compared with other
shareholders They justify this because there is generally more than one institutional
investor and there will be free-riding problems with multiple investors that only the
largest institutional investors can overcome While we accept their premise of free-rider
problems it is not obvious why the deflator is lsquoinstitutional shareholdingrsquo rather than
lsquototal shareholdingrsquo from this perspective7 If the strength of the size control can be
queried scaling by total shareholding (ie Top 5 Ownership) produces a measure that
remains relatively unaffected across the size deciles It is 20 percent for the smaller
decile declining to 14 percent for the largest decile whereas for the HS concentration
measure it declines by over 50 percent from 61 percent to 28 percent (see Table III and
Figure 1)
C Testing the Jensen-Murphy Measure of Incentives
The testing of their secondary measure of executive pay-for-performance sensitivity
(see HS Table IV) based on Jensen and Murphy (1990) raises issues other than the size
control HS regress changes in compensation (cash and total) on lagged institutional
influence that has been interacted with the contemporaneous change in shareholder
wealth to capture a pay-for-performance relationship Such a test is appealing ndash
7 AHS do point out one benefit of the institutional concentration measure Institutional investors may prefer
companies with lsquobetterrsquo incentive compensation structures Could using institutional shareholding rather
than total shareholding as the deflator alleviate this problem Presumably both the numerator and
denominator will have the same bias and hence error cancellation is a possibility
9
institutions should increase compensation when the firm has performed well Following
Greene (2000 p326) to execute such a test one needs to include in level form variables
that are then interacted Hence the HS model specification ought to be
1 2 3 1 4 1it it it it it itCompn SW IC SW IC (1)
where itCompn is the change in executive compensation itSW is the contemporaneous
change in shareholder wealth which also appears in the interaction term and 1itIC is
lagged institutional ownership concentration However HS (p2363) omit the
contemporaneous term itSW and the lagged term 1itIC which should appear in levels
as well as in the interaction term Most importantly HS interact their lagged institutional
concentration measure 1itIC with the contemporaneous change in shareholder wealth
itSW As it is after all pay-for-performance concentrated institutions need to condition
their influence on some observed measure of firm performance Contemporaneous firm
performance is unknown at the time institutional influence is measured Reflecting this
timing insight leads to a new Greene (2000) specification
1 2 1 3 1 4 1 1it it it it it itCompn SW IC SW IC (2)
such that all the lagged RHS independent variables that now appear in the interaction
term are included on a stand-alone basis
Reliance on incomplete self-reported values for option grants
Prior to receiving the HS dataset we tried to replicate the findings of HS to validate
our datasets for the HS period and the longer period 1992‒2002 While we initially
focused on the same period as HS (ie 1992‒1997) we found that our sample size was
larger at 47765 observations than HSrsquos at 33928 observations This discrepancy could
have arisen if HS did not replace missing and self-reported option valuations using their
10
Black-Scholes option methodology This may introduce a downward pay bias since self-
reported option-grant pay is subject to possible understatement and there are 13839
observations with likely high option pay dropped altogether
Endogeneity
HSrsquos endogeneity testing raises concerns about reverse causality Institutions may
prefer to hold firms with certain compensation structures such as incentivized executives
HS rightly highlight this concern and use instrumental variables to address this issue HS
assume that institutional ownership reflects any reverse causality Using an instrumental
variables approach to resolve this problem involves finding a variable that influences or
explains institutional ownership but not directly executive compensation They use share
turnover in a firm as an instrument for institutional ownership However theoretical work
by Holmstrom and Tirole (1993) indicates that measures of stock liquidity and trading
activity are indeed important determinants of executive market-based incentives as Kang
and Lui (2008) and Dikolli Kulp and Sedatole (2009) find empirically As such one can
make a case for adopting another instrument rather than share turnover for institutional
ownership
II Previous Research
Prior research fails to address issues concerning possible fragility of the HS results
AHS replicate the HS findings on both incentives and pay levels using the same dataset as
HS and they apply the same definition of institutional concentration and use market
capitalization expressed as a level Cadman Klasa and Matsunaga (2010 Table 6) set out
to replicate HSrsquos (Table II) model of option grant pay-for-performance sensitivity when
they compare a dataset of ExecuComp and Non-ExecuComp firms for the period 2000-
2007 Their ExecuComp results for this sample period are quite similar to our results for
11
the 1992‒2002 period for the same specification controlling for the logarithm of market
capitalization but they do not regard these results as a test of HS as their focus is entirely
on comparing the two groups of stocks Nor do they investigate the impact of
concentration on either cash compensation or total pay Kang and Liu (2008) examine
pay-performance sensitivity use a version of the HS institutional concentration measure
as a control variable in a different framework deriving from Holmstrom and Tirole
(1993) Dikolli Kulp and Sedatole (2009) argue that firms with transient ownership will
issue relatively more equity incentives to counter short-termism with their results robust
to the inclusion of the HS concentration variable
We hoped to find supportive evidence from proxy voting behavior in support of HSs
findings However evidence from actual mutual fund proxy voting records suggests that
most oppose reductions in CEO pay but overall the evidence is mixed8 Davis and Kim
(2005) examine proxy votes of mutual funds for 2004 and find that for a sample of
Fortune 1000 proxy contests in 45 cases that attempted to limit executive pay mutual
funds were unanimously opposed to such attempts Rothberg and Lilien (2006) also
examine proxy-voting data and find that mutual funds voted 66 percent of the time in
managementrsquos favor on issues of executive compensation Ertimur Ferri and Muslu
(2011) find that institutional investors lack sophistication in that they do not target
excessively paid CEOs relative to CEOs with high predicted-pay based on economic
determinants but exert a moderating influence ndash a $23 million reduction ndash for those they
deem ldquoexcessively paidrdquo CEOs
III Data and Methodology
8 We do acknowledge that the HS results are for all institutional investors while the transparency with
respect to proxy voting applies only to mutual funds
12
This Section discusses the data used in the study the construction of variables
descriptive statistics and the methodology Two datasets are used ndash one is the original
dataset generously supplied by Jay Hartzell and Laura Starks which covers 1992‒1997
second a more comprehensive dataset covering 1992‒2002 is computed from the same
data sources used in the HS dataset This is Standard and Poorrsquos (SampP) ExecuComp for
compensation data Thomson Financial CDASpectrum database for institutional holdings
to compute measure of institutional influence SampPrsquos Compustat and the Center for
Research in Stock Prices (CRSP) for firm level accounting and stock price data Our more
comprehensive dataset contains 2559 firms compared with 1914 for the HS dataset Our
full matched sample yields 97679 executive years
A Institutional Influence Measures
To measure institutional influence we employ two ownership-based metrics The first
measure of institutional influence is HSrsquos Institutional Concentration We use this for
direct comparison to the HS analysis The second measure of institutional influence
changes the denominator of the HS measure from institutional shares held to total shares
outstanding in the firm with nomenclature Top 5 Ownership9 We compute both
measures at the fiscal year end of the firm We adjust all compensation measures and
asset prices for inflation to 30 June 2002 prices for our comprehensive sample compared
to the HS dataset that is in the dollars of the day
B Compensation Measures
The ExecuComp database reports several raw measures of executive compensation for
the top five executives (where available) including total salary bonus long-term
9 Top 5 Ownership is included in the HS database but not utilized by them in reported tests as well as
constructed by us for the more comprehensive sample
13
incentives and number of option grants For our more comprehensive dataset we replicate
all the compensation measures used by HS In common with HS we compute the pay-for-
performance sensitivity of option grants using the Yermack (1995) approach This
measure shows the dollar change in value of the executive options for every dollar change
in the value of the firm This involves calculating the option delta using the Black-
Scholes formula adjusted for dividends10
The ratio of the number of shares represented
by the award of options divided by the diluted number of shares outstanding at the
beginning of the year multiplies the option delta with units given by a $1000 change in
firm value Summing salary bonus long-term incentive plan payouts and stock and
option grants provides total direct compensation Options are valued using the Black and
Scholes (1973) formula for European call options adjusted by Merton (1973) to
incorporate dividends
C Descriptive Statistics
Table I presents descriptive statistics of key variables employed in the study according
to our comprehensive dataset11
Institutions own on average 546 percent of the firms in
which they invest This is almost identical to the 531 percent found by HS The
institutions with the five largest holdings account 45 percent of institutional ownership
Once again the figure for HS is almost identical at 44 percent The institutions with the
five largest holdings own on average 229 percent of shares outstanding in a firm (HS
have 223 percent) Average total compensation for executives is $2583 million
considerably larger than that documented by HS ($1250 million) but the HS figure is an
10 When there is more than one grant during the year a weighted average delta is used
11 With the data supplied to us by Hartzell and Starks we are able to replicate exactly their summary data
(HS Table I) for all but cash compensation Salary is included but bonus payments are missing The
replicated table is available on request
14
amalgam of price levels over their sample period 1992‒1997 and ours are in consistent
dollars of June 30 2002 However there has also been significant growth in
compensation levels from 1998 to 2002
ltltINSERT TABLE Igtgt
Based on our comprehensive data in Table II we examine correlations between the
key variables employed in the study Of concern is the large negative correlation of 63
percent which has been discussed previously (Section IIA) between the HS measure of
institutional concentration and the natural logarithm of firm market capitalization
ltltINSERT TABLE IIgtgt
In view of this high correlation Table III provides a breakdown of the three key pay
variables and three institutional ownership variables by size deciles We find that the high
institutional concentration but low institutional ownership share of total shareholdings
mentioned above is evident in the smallest decile of firms by size This decile has low
salary and even lower total compensation in a relative sense but by far the highest pay-
for-performance sensitivity of option grants Small firms award executives a far higher
proportion of market capitalization Since small firm market-capitalization is negligible in
comparison with the highest decile (300 times larger in magnitude) total compensation is
very small in absolute value Moreover due to the domination of the pay-size effect base
salaries are also exceedingly small
ltltINSERT TABLE IIIgtgt
Figure 1 based on Table III shows how as institutional concentration falls with
increases in stock size (but also as the total share of institutional ownership increases)
and the pay-for-performance sensitivity of option grants falls Thus because of what
appears to be a weak size control and size-influenced concentration measure HS
15
potentially pick up a size effect as their main result but not necessarily a monitoring
effect Figure 2 shows that as firm size increases up to decile 8 at least the share of
institutional total ownership increases as does total compensation
ltltINSERT FIGURES 1 AND 2 gtgt
D Methodology
Following HS we use two estimation techniques to examine the effect of institutional
investor influence on executive compensation Due to the censored nature of option grant
pay-for-performance sensitivity this compensation measure most appropriately uses
Tobit model specification whereas an Ordinary Least Squares (OLS) model specification
is used for all other compensation measures We model compensation measures as a
function of contemporaneous and lagged change in shareholder wealth institutional
influence Tobinrsquos Q and market capitalization Controls for CEO industry and year
effects are also used Augmentation occurs to test the effect of institutional influence on
the change in executive cash and total compensation with an interaction term that is the
product of institutional influence and the change in shareholder wealth
The variables used to model executive compensation have the following definitions
The lagged and contemporary measure of change in shareholder wealth is simply the
difference in market capitalization of the firm The measures of institutional influence
used are Institutional Concentration and Top 5 Ownership Use of the two types of
institutional influence occurs in separate models In order to replicate the HS
methodology models using Institutional Concentration have an additional control
consisting of lagged total institutional holdings and Tobinrsquos Q controls for performance
and growth opportunities Market capitalization of the firm is the number of shares
outstanding multiplied by the share price with inclusion both in levels and natural
16
logarithmic form A dummy variable is included to identify whether the executive is the
CEO Dummy variables are also included to control for industry effects This variable
takes a value of 1 for the two-digit Standard Industrial Classification (SIC) in which the
firm operates A year fixed effects (dummy) variable is included taking a value of 1 if the
observation was from the given year
IV Executive Incentives as a Function of Institutional Investor Influence
This section presents the results for the effect of institutional influence on option grant
pay-for-performance sensitivity and changes in compensation
A Option Grant Pay-for-Performance Sensitivity
Table IV presents the results for the effect of institutional influence on option grant
pay-for-performance sensitivity
ltltINSERT TABLE IVgtgt
Model (1) is a direct replication of HS using the HS database as supplied to us by HS
with all the variables computed in the same way and in the same functional form as with
HS The coefficient and t-statistic on lagged institutional concentration interacted with the
change in shareholder wealth is 13994 and 767 that compares to 1358 and 842 in HS
We are unable to account for these minor discrepancies for values that should be
identical In model (2) the only change to the model is that we use the natural logarithm
of market capitalization instead of the level of market capitalization This single change
has a sizeable effect on the coefficient on institutional concentration The coefficient on
institutional concentration is now -00692 and is insignificant The fact that an order
preserving transformation to a control variable can change the results is of interest The
log-transformation to market capitalization also changes the coefficient on Top 5
Ownership in models (3) and (4) The coefficient on Top 5 Ownership is significant in
17
both model (3) and in model (4) but the better size control reduces the coefficient by over
25 percent The better size control does not affect Top 5 Ownership by as much as its
design ensures less sensitivity to what could be inadequate size controls
We also test the effect of institutional influence on option grant pay-for-performance
sensitivity over the period 1992‒2002 This sample is then approximately three times as
large as the sample used by HS Institutional concentration is highly positively significant
in model (5) similar to the results for the period 1992‒1997 When we use the log of
market capitalization in model (6) the coefficient on institutional concentration remains
positive and significant at the 10 percent level with a t-statistic of 166 Of importance
though is that the coefficient on institutional concentration is one-eighth of the coefficient
in model (5) Even for a measure of institutional influence which we believe is less
sensitive to firm size the use of log market capitalization reduces the coefficient on Top 5
Ownership by approximately 50 percent from model (7) to model (8)
B Pay-for-Performance Sensitivity due to Changes in Cash and Total
Compensation
For robustness HS examine the effect that institutional influence conditioned on firm
performance has on a changes in cash and total executive compensation using the general
method of Jensen and Murphy (1990) to compute pay-for-performance sensitivity
ltltINSERT TABLE V gtgt
Table V presents the effect of institutional influence on the change in cash
compensation using the HS size control measure throughout12
Model (1) replicates the
specification in HS for the period of 1992‒1997 and we produce a similar result ndash the
interaction of lagged institutional concentration and contemporaneous change in
12 This test uses our data for the two periods 1992‒1997 and 1992‒2002 We do not use the HS data for
cash compensation as data for bonus is missing in the data supplied to us
18
shareholder wealth positively and significantly relates to a change in executive cash
compensation In model (2) we include the contemporaneous change in shareholder
wealth and the lagged institutional concentration as stand-alone controls as indicated by
Greenersquos (2000) specification of equation (1) above and the interaction term remains
significant In model (3) we interact lagged institutional concentration with the lagged
change in shareholder wealth as in equation (2) above that tries to take account of what
participants could have observed at the time The interaction term is then negative but
insignificant altering the HS result Model (4) provides a similar result where our
preferred proxy Top 5 Ownership captures concentrated institutional influence In
models (5) to (8) we use our full sample In model (5) we find that the lagged
institutional concentration interaction with the contemporaneous change in shareholder
wealth is positive and significant in HSrsquos specification However including change in
contemporaneous shareholder wealth and lagged institutional concentration as separate
control variables as suggested by equation (1)rsquos specification removes the significance
from the interaction term in model (6) Moreover in model (7) when lagged institutional
concentration is interaction with the lagged change in shareholder wealth in accord with
equation (2) this interaction term is now negative and significant reversing the HS
finding We obtain a similar result in model (8) when we measure institutional
concentration influence using Top 5 Ownership13
V The Levels of Executive Fixed and Total Compensation as Functions of
Institutional Investor Influence
13 In unreported results we use the log of market capitalization as a size control These changes to the size
control do not dramatically change the findings from Table V The reason why this size issue is not as
important as in the findings for option grants pay-for-performance is that the change in cash compensation
is not so strongly influenced by firm size In unreported results available on request we also examine the
effect of institutional influence on changes in total compensation with similar results to the change in cash
compensation
19
Two measures of compensation are examined ndash salary and total compensation
ltltINSERT TABLE VI gtgt
Table VI examines the effect of institutional influence on the level of salary In model
(1) we use the HS sample and specification to obtain almost identical results The
coefficient on institutional concentration is (negative) $247509 that is almost identical to
(negative) $247967 in HS Both are highly significant In model (2) we make a single
change to the model ndash we use the natural logarithm of market capitalization This single
change reverses the sign of institutional concentration with the coefficient still being
significant at the 1 percent level In models (3) and (4) we measure institutional influence
using Top 5 Ownership In model (3) where we measure firm size using the level of
market capitalization Top 5 Ownership is insignificant When we use the natural
logarithm of market capitalization in model (4) Top 5 Ownership is significant and
positively relates to the salary level In models (5) to (8) we estimate the same models
using our full sample We find similar results for the full sample as for the period 1992ndash
1997
ltltINSERT TABLE VII gtgt
Table VII examines the effect of institutional influence on total compensation In
model (1) we replicate the HS methodology using the HS data to obtain a similar
significant negative relation between institutional concentration and total compensation of
(negative) $1608633 However after changing to a logarithmic specification of market
capitalization in model (2) institutional concentration increases total compensation by a
significant $32400014
Importantly when institutional influence is measured using Top 5
14 = (5688751 times 0101)
20
Ownership an insignificant relation exists with total compensation in model (3) using the
HS size control but it becomes both positive and significant in (4) with the logged size
control In models (5) and (6) where models are estimated using the full sample we
observe a similar reversal in sign as occurs in models (1) and (2) However now the
magnitude of the reversal is almost equal with a difference of about $67m that is very
large indeed In model (5) the coefficient on institutional concentration is (negative)
$342m and then in model (6) the coefficient is $33m Top 5 Ownership is insignificant
in both models (7) and (8) unlike the results for Salary
VI Robustness
Two issues are examined in is this Section ndash reverse causality and the appropriate
transformation of market capitalization A methodological issue inherent in examining the
effect of ownership-based measures of institutional influence on executive compensation
is endogeneity driven by reverse causality We utilize an instrumental variables approach
where orthogonalized institutional influence measures are regressed on each of the
compensation measures lagged option grant pay-for-performance sensitivity salary and
total compensation In addition market capitalization and a CEO dummy are included in
each regression We use the residual from each regression as an instrument for
institutional influence Such a process removes any preferences for compensation from
the institutional influence measures
ltltINSERT TABLES VIII AND IXgtgt
Table VIII presents the results where we instrument the HS measure lagged
institutional concentration with a compensation-orthogonalized equivalent The analysis
indicates that the instrumented institutional concentration measure is positively and
significantly related to the level of salary and total compensation and insignificantly
21
related to option grant pay-for-performance sensitivity (PPSOG) Table IX presents
results where we instrument Top 5 Ownership with a compensation-orthogonalized
equivalent The instrumented Top 5 Ownership measure insignificantly relates to option
grant pay-for-performance sensitivity and barely negatively relates to salary although it
negatively relates to total compensation suggesting reverse causality is driving a
significant portion of our previous results
ltltINSERT TABLE Xgtgt
Another issue examined in this section is the appropriate transformation of market
capitalization Aneuryn-Evans and Deaton (1980) provide an evaluation of methods
comparing linear and log-linear specifications We use the Box-Cox test to compare
logarithm versus linear specifications In applying the Box-Cox test we specify
transformation of only market capitalization ndash the dependent variable and all other
independent variables remain unchanged Table X presents the lambda coefficient for
each model and the log likelihood score In all models in Panel A and B a lambda of zero
receives a higher (closer to zero) log likelihood score suggesting the log specification is
appropriate
VII Conclusion
Hartzell and Starks (2003) appears to document an effective role for institutional
investors in increasing executive incentives and lowering compensation levels Our paper
re-addresses the issue of concentrated institutional monitoring from a perspective
informed by an additional decade of research since their paper appeared We cannot find
evidence that concentrated institutional investors really have any positive (or negative for
that matter) effect on executive incentive levels While superficially we find evidence
that they may raise executive pay levels this is most likely driven by an institutional
22
preference for highly remunerated executives in large companies and thus by reverse
causality Moreover our findings are not consistent with the rent-extraction hypothesis
suggesting that firms systematically overpay executives in the absence of effective
monitoring by outside parties such as institutional investors
23
References
Almazan Andres Jay C Hartzell and Laura T Starks 2005 ldquoActive Institutional
Shareholders and Costs of Monitoring Evidence from Executive Compensationrdquo
Financial Management Winter 5ndash34
Aneuryn-Evans Gwyn and Angus Deaton 1980 ldquoTesting Linear versus Logarithmic
Regression Modelsrdquo Review of Economic Studies 47 (1) Econometrics Issue 275ndash
291
Babenko Ilona 2009 ldquoShare Repurchases and Pay-Performance Sensitivity of Employee
Compensation Contractsrdquo Journal of Finance 64117-150
Baker George P and Brian J Hall 2004 ldquoCEO Incentives and Firm Sizerdquo Journal of
Labor Market Economics 22 767ndash798
Bebchuk Lucian A Jesse M Fried and David Walker 2002 ldquoManagerial power and
rent extraction in the design of executive compensationrdquo University of Chicago Law
Review 69 751ndash846
Bebchuk Lucian A and Jesse M Fried 2003 ldquoExecutive Compensation as an Agency
Problemrdquo Journal of Economic Perspectives 17 71ndash92
Bebchuk Lucian A and Jesse M Fried 2004 Pay without Performance The Unfulfilled
Promise of Executive Compensation Cambridge MA Harvard University Press
Bertrand Marianne and Sendhil Mullainathan 2001 ldquoAre CEOs Rewarded for Luck
The Ones Without Principals Arerdquo Quarterly Journal of Economics 116 901ndash32
Black Fischer and Myron Scholes 1973 ldquoThe Pricing of Options and Corporate
Liabilitiesrdquo Journal of Political Economy 81 637ndash54
Cadman Brian Sandy Klasa and Steve Matsunaga 2010 ldquoDeterminants of CEO Pay A
Comparison of ExecuComp and Non-ExecuComp Firmsrdquo The Accounting Review 85
1511ndash1543
Chhaochharia Vidhi and Yaniv Grinstein 2009 ldquoCEO Compensation and Board
Structurerdquo Journal of Finance 64 231ndash261
Cichello Michael S 2005 ldquoThe impact of firm size on pay-performance sensitivitiesrdquo
Journal of Corporate Finance 11 609ndash627
Davis Gerald F and E Han Kim 2007 ldquoBusiness ties and proxy voting by mutual
fundsrdquo Journal of Financial Economics 85 552ndash570
Dikolli Shane S Susan L Kulp and Karen L Sedatole 2009 ldquoTransient Institutional
Ownership and CEO Contractingrdquo Accounting Review 84 737ndash770
Edmans Alex Xavier Gabaix and Augustin Landier 2009 ldquoA Multiplicative Model of
Optimal CEO Incentives in Market Equilibriumrdquo Review of Financial Studies 22
4881ndash4917
Edmans Alex 2009 ldquoBlockholder Trading Market Efficiency and Managerial Myopiardquo
Journal of Finance 64 2481ndash2513
Ertimur Yonca Fabrizio Ferri and Volkan Muslu 2011 ldquoShareholder Activism and
CEO Payrdquo Review of Financial Studies 24 535ndash592
24
Gabaix Xavier and Augustin Landier 2008 ldquoWhy Has CEO Pay Increased So Muchrdquo
Quarterly Journal of Economics 123 49ndash100
Garen John 1994 ldquoExecutive compensation and principal-agent theoryrdquo Journal of
Political Economy 102 1175ndash1199
Gibbons Robert and Kevin J Murphy 1992 ldquoOptimal Incentive Contracts in the
Presence of Career Concerns Theory and Evidencerdquo Journal of Political Economy
100 468ndash505
Greene William H 2000 (4th
Edition) Econometric Analysis Prentice-Hall NJ
Guthrie Katherine Jan Sokolowsky and Kam-Ming Wan 2012 ldquoCEO Compensation
and Board Structure Revisitedrdquo Journal of Finance Forthcoming
Hadlock Charles R and Gerald B Lumer 1997 ldquoCompensation turnover and top
management incentives historical evidencerdquo Journal of Business 70 153ndash187
Hartzell Jay C and Laura T Starks 2003 ldquoInstitutional Investors and Executive
Compensationrdquo Journal of Finance 58 2351ndash2374
Huson Mark R Robert Parrino and Laura T Starks 2001 ldquoInternal Monitoring
Mechanisms and CEO Turnover A Long-Term Perspectiverdquo Journal of Finance 56
2265ndash2297
Jensen Michael C and Kevin J Murphy 1990 ldquoPerformance Pay and Top Management
Incentivesrdquo Journal of Political Economy 98 225ndash264
Jin Li 2002 ldquoCEO compensation diversification and incentivesrdquo Journal of Financial
Economics 66 29ndash63
Kang Qiang and Qiao Liu 2008 ldquoStock Market Information Production and Executive
Incentivesrdquo Journal of Corporate Finance 14 484ndash498
Merton Robert 1973 ldquoTheory of Rational Option Pricingrdquo Bell Journal of Economics
and Management Science 4 141-183
Morse Adair Vikram Nanda and Amit Seru 2011 ldquoAre Incentive Contracts Rigged by
Powerful CEOsrdquo Journal of Finance 66 1779ndash1821
Murphy Kevin J 1985 ldquoCorporate performance and managerial remuneration An
empirical analysisrdquo Journal of Accounting and Economics 7 11ndash42
Murphy Kevin J 1999 ldquoExecutive compensationrdquo in Orley Ashenfelter and David
Card Handbook of Labor Economics vol 3 Elsevier
Rosen Sherwin 1992 ldquoContracts and the market for executivesrdquo In Contract
Economics ed Lars Werin and Hans Wijkander 181ndash211 Cambridge MA
Blackwell
Rothberg Burton and Steven Lilien 2006 ldquoMutual Funds and Proxy Voting New
Evidence on Corporate Governancerdquo Journal of Business and Technology Law 1
157ndash184
Schaefer Scott 1998 ldquoThe dependence of payndashperformance sensitivity on the size of the
firmrdquo Review of Economics and Statistics 80 436ndash443
Yermack David 1995 ldquoDo Corporations Award CEO Stock Options Effectivelyrdquo
Journal of Financial Economics 39 237ndash69
25
Table I Univariate Descriptive Statistics
This table reports sample statistics for the key variables used in this study based on our comprehensive
ExecuComp dataset Institutional Concentration is the shares owned by the institutions with the five largest
holdings divided by the shares held by all institutions Top 5 Ownership is the proportion of shares
outstanding held by the institutions with the five largest holdings Total Institutional Ownership is the
proportion of shares held by all institutions Panel B shows the measures executive compensation The pay-
for-performance sensitivity (PPS) of option grants shows the change in executive wealth due to a $1000
change in firm in shareholder wealth Expressions for Salary and Total Compensation are in thousands of
dollars while Market Capitalization is in millions of dollars Panel C shows firm characteristics
Variable Mean Median Std Dev P10 P90
Panel A Institutional Investor Holdings
Institutional Concentration 0450 0424 0151 0282 0656
Top 5 Institutional Ownership 0229 0220 0101 0110 0355
Total Institutional Ownership 0546 0561 0206 0258 0801
Panel B Executive Compensation
PPS of Option Grants 1161 0156 3829 0 2822
Salary 36311 29462 25231 15179 67000
Total Compensation 258313 85878 980701 23875 511022
Panel C Firm Characteristics
Tobins Q 1744 1139 2738 0495 3322
Market Capitalization 560699 112094 1988108 19754 1048754
26
Table II Pairwise Correlation Coefficients
This table presents the correlations between the key variables based on our comprehensive
ExecuComp dataset Institutional Concentration is the shares owned by the institutions with the five
largest holdings divided by the shares held by all institutions Top 5 Ownership is the proportion of
shares outstanding held by the institutions with the five largest holdings Total Institutional
Ownership is the proportion of shares held by all institutions The pay-for-performance sensitivity
(PPS) of option grants is the change in executive wealth due to a $1000 change in shareholder
wealth
(1) (2) (3) (4) (5) (6) (7) (8)
(1) Institutional Concentration 1
(2) Top 5 Ownership 0268 1
(3) Total Ownership -0415 0706 1
(4) PPS of Option Grants 0122 0042 -0024 1
(5) Log Salary -0270 -0012 0171 0038 1
(6) Log Market Capitalization -0630 -0157 0261 -0156 0420 1
(7) Level Market Capitalization -0241 -0157 0007 -0058 0219 0534 1
(8) Tobins Q -0008 -0034 0002 0015 -0083 0066 0121 1
27
Table III
Descriptive Statistics by Market Capitalization Deciles
Market
Capitalization
Deciles
Market
Capitalization
($m)
Total
Compensation
($000)
Salary
($)
PPS Option
Grants
($)
Institutional
Concentration
()
Top 5 Ownership
()
Institutional
Total Ownership
()
1 (Small) 1181 76008 234700 2720 611 200 327
2 2668 90206 257955 1796 495 217 438
3 4193 105946 275120 1517 454 220 485
4 6045 129892 296067 1376 420 206 491
5 8742 155391 315864 1216 394 198 502
6 13086 191888 345942 1072 380 202 531
7 20790 239501 377021 0874 368 196 532
8 34920 314955 427808 0687 354 190 538
9 67747 450970 480319 0569 328 173 528
10 (Big) 366540 846066 614354 0297 275 138 502
28
Table IV Pay-for-Performance Sensitivity of Option Grants as a Function of Institutional Influence
Similar in construction to parts of HS (Table II) this table shows the coefficients from Tobit regressions of option grant pay-for-performance sensitivity per $1000 change in
shareholder wealth against the lagged change and contemporary change in shareholder wealth measures of institutional influence Tobinrsquos Q firm market capitalization and
dummy controls for CEO and year effects We estimate models (1) to (4) over the period 1992‒1997 using data supplied by HS and models (5) to (8) are estimated over the
whole sample Models (1) (2) (5) and (6) use Institutional Concentration as a measure of institutional influence which HS define as the shares owned by the institutions
with the five largest holdings divided by the shares held by all institutions These models control for total institutional ownership Models (3) (4) (7) and (8) use Top 5
Ownership as a measure of institutional influence which we define as the proportion of shares outstanding held by the institutions with the five largest holdings Coefficients
for industry and year controls are not presented One two and three asterisks denote significance at 10 5 and 1 levels
Hartzell and Starks Database 1992‒1997 Full Sample from ExecuComp 1992‒2002
(1) (2) (3) (4) (5) (6) (7) (8)
Change in Shareholder Wealth 77721 191781 89874 137368 -446 -218 -664 -203
(171) (429) (194) (317) (319) (168) (453) (158)
Lag1 Change in Shareholder Wealth 14446 99963 88719 8544 -144 0273 -0342 00852
(027) (212) (166) (184) (071) (015) (016) (005)
Lag1 Institutional Concentration 13994 -00692 26405 03304
(767) (032) (1402) (166)
Lag1 Total Institutional Ownership 08055 10434 07106 0937
(593) (781) (666) (898)
Lag1 Top 5 Institutional Ownership 10567 07457 12116 06293
(399) ( 283) (533) (278)
Lag1 Tobinrsquos Q 0068 00877 00647 00813 00498 00727 00508 00695
(337) (434) (319) (401) (504) (734) (51) (703)
Lag1 Market Capitalization -106202 -173515 -776 -117
(474) (742) (1011) (153)
Lag1 Log Market Capitalization -0297 -02319 -04111 -03908
(1576) (1439) (2997) (2841)
CEO Dummy 13745 13682 13765 13674 28197 28128 28155 28143
(1913) (1918) (1914) (1916) (3388) (3396) (3377) (34)
Number of Observations 33878 33878 33878 33878 97679 97679 97679 97679
29
Table V Pay-for-Performance Sensitivity of Cash Compensation (Salary plus Bonus) as a Function of Institutional Influence - Level Market Capitalization
Similar in parts to models (1) and (2) of HS (Table IV) this table displays the coefficients from OLS regressions of the change in cash compensation against the lagged
change and contemporary change in shareholder wealth measures of institutional influence Tobinrsquos Q firm market capitalization and dummy controls for CEO and year
effects We estimate models (1) to (4) over the period 1992‒1997 based on our ExecuComp sample and models (5) to (8) over the whole sample Models (1) (2) (5) and (6)
use Institutional Concentration as a measure of institutional influence which HS define as the shares owned by the institutions with the five largest holdings divided by the
shares held by all institutions These models control for total institutional ownership Models (4) and (8) use Top 5 Ownership as a measure of institutional influence which
we define as the proportion of shares outstanding held by the institutions with the 5 largest holdings Coefficients for industry and year controls are not presented One two
and three asterisks denote significance at 10 5 and 1 levels
Our Database based on ExecuComp 1992‒1997 Full Sample based on ExecuComp 1992‒2002
(1) (2) (3) (4) (5) (6) (7) (8)
Lag1 Change in Shareholder Wealth 00014 00017 00088 00089 -00006 -00007 00072 00086
(051) (062) (129) (103) (027) (029) (131) (16)
Change in Shareholder Wealth -00102 00102 00105 00032 00102 00101
(097) (292) (293) (055) (611) (608)
Change Shareholder Wealth Lag1 Institutional
Concentration
00435 00764 00363 00253
(463) (258) (689) (144)
Lag1 Change Shareholder Wealth Lag1
Institutional Concentration
-00197 -00283
(122) (197)
Lag1 Institutional Concentration -492182 -288861 400149 453623
(09) (056) (098) (114)
Lag1 Total Institutional Ownership 720688 538372 655091 559587 682069 726226
(276) (232) (281) (314) (308) (322)
Lag1 Change Shareholder Wealth Lag1 Top 5
Institutional Ownership
-0046 -00689
(091) (256)
Lag1 Top 5 Institutional Ownership 583044 1191634
(092) (25)
Lag1 Tobinrsquos Q -32761 -36806 -19422 -18023 -54283 -48681 -29534 -24243
(138) (172) (077) (078) (227) (245) (136) (111)
Lag1 Market Capitalization 00004 00004 00002 00006 00007 00008 00007 00007
(042) (044) (022) (07) (086) (092) (082) (09)
CEO Dummy 768035 765474 768334 773634 616315 61729 619028 618535
(354) (355) (356) (355) (417) (418) (419) (418)
Adjusted R-Sq 00057 0006 00052 0005 00097 00097 00098 00102
Number of Observations 41305 41305 41305 41305 85518 85518 85518 85518
30
Table VI Level of Salary as a Function of Institutional Influence
Similar in parts to models (1) and (2) of HS (Table V) this table displays the coefficients from OLS regressions of salary against the lagged change and contemporary change in
shareholder wealth measures of institutional influence Tobinrsquos Q firm market capitalization and dummy controls for CEO industry and year effects We estimate models (1) to
(4) over the period 1992‒1997 using the actual data supplied by HS and models (5) to (8) over the whole sample 1992‒2002 Models (1) (2) (5) and (6) use Institutional
Concentration as a measure of institutional influence which HS define as the shares owned by the institutions with the five largest holdings divided by the shares held by all
institutions These models control for total institutional ownership Models (3) (4) (7) and (8) use Top 5 Ownership as a measure of institutional influence which we define as
the proportion of shares outstanding held by the institutions with the 5 largest holdings Coefficients for industry and year controls are not presented One two and three
asterisks denote significance at 10 5 and 1 levels
Hartzell and Starks Database 1992‒1997 Full Sample based on ExecuComp 1992‒2002
(1) (2) (3) (4) (5) (6) (7) (8)
Change in Shareholder Wealth 04125E-4 0003 -2439E-4 0003 00021 00012 00024 00012
(057) (464) (031) (468) (1004) (72) (1041) (706)
Lag1 Change in Shareholder Wealth -00062 7856E-4 -0007 9003E-4 -00014 00002 -00014 00003
(594) (098) (592) (112) (522) (123) (5) (145)
Lag1 Institutional Concentration -2475087 555134 -2734983 943317
(3245) (710) (4965) (1671)
Lag1 Total Institutional Ownership 1020109 217488 936337 252464
(1904) (432) (2471) (704)
Lag1 Top 5 Institutional Ownership -148355 31354 -87926 423166
(139) (332) (118) (643)
Lag1 Tobinrsquos Q -195348 -252191 -192344 -25158 -119837 -146373 -122132 -143821
(2812) (3762) (2721) (3752) (248) (274) (2583) (2727)
Lag1 Market Capitalization 844E-06 103E-5 00034 00039
(2781) (3051) (2874) (3104)
Lag1 Log Market Capitalization 750021 723625 776657 732279
(7511) (9285) (11224) (14039)
CEO Dummy 2683978 2702238 2668244 2702417 3281133 3285103 3292175 3283145
(8625) (9277) (8246) (9271) (12739) (13616) (12336) (13605)
Adjusted R-Sq 04604 05302 04187 05295 03856 04588 03476 04574
Number of Observations 36280 36280 36280 36280 97679 97679 97679 97679
31
Table VII Level of Total Direct Compensation as a Function of Institutional Influence
Similar in parts to models (3) and (5) of HS (Table V) this table displays the coefficients from OLS regressions of total direct compensation against the lagged change and
contemporary change in shareholder wealth measures of institutional influence Tobinrsquos Q firm market capitalization and dummy controls for CEO industry and year effects
We estimate models (1) to (4) over the period 1992‒1997 and models (5) to (8) over the whole sample Models (1) (2) (5) and (6) use Institutional Concentration as a measure
of institutional influence which HS define as the shares owned by the institutions with the five largest holdings divided by the shares held by all institutions These models
control for total institutional ownership Models (3) (4) (7) and (8) use Top 5 Ownership as a measure of institutional influence which we define as the proportion of shares
outstanding held by the institutions with the five largest holdings Coefficients for industry and year controls are not presented One two and three asterisks denote significance
at 10 5 and 1 levels
Hartzell and Starks Database 1992‒1997 Full Sample from ExecuComp 1992‒2002
(1) (2) (3) (4) (5) (6) (7) (8)
Change in Shareholder Wealth 005779 007006 005282 006791 00423 00148 0046 00138
(415) (581) (369) (565) (203) (07) (22) (066)
Lag1 Change in Shareholder Wealth 003182 00714 00272 00715 00613 0126 00614 01278
(126) (298) (107) (299) (275) (488) (276) (494)
Lag1 Institutional Concentration -1608633 3248998 -3424892 3334186
(151) (283) (1172) (118)
Lag1 Total Institutional Ownership 104013 5473128 1531013 -2926764
(1645) (831) (876) (158)
Lag1 Top 5 Institutional Ownership 188585 3927986 -2159177 -2640879
(016) (343) (067) (092)
Lag1 Tobinrsquos Q -445692 -803612 -454859 -833566 3257027 2971201 3213936 3086205
(427) (791) (432) (831) (83) (77) (815) (793)
Lag1 Market Capitalization 49E-5 615E-5 01031 01088
(113) (1365) (1555) (168)
Lag1 Log Market Capitalization 4683497 4759377 1645156 1427856
(2775) (3469) (3679) (4239)
CEO Dummy 1501307 1508886 1493688 1508736 4713079 4713629 47278 4703908
(2648) (2684) (2618) (2683) (282) (2828) (2819) (2824)
Adjusted R-Sq 01861 02060 01695 02051 01071 01113 01033 011
Number of Observations 33229 33229 33229 33229 97679 97679 97679 97679
32
Table VIII Two Stage Least Squares (Reverse Causality) Tests - Executive Compensation Measures as a Function of Institutional Concentration
This table shows the coefficients from the Two-Stage Least Squares (2SLS) regressions of executive compensation measures against the lagged change
and contemporary change in shareholder wealth instrumented institutional concentration and total institutional ownership Tobinrsquos Q the natural
logarithm of firm market capitalization and dummy variable controls for CEO and year effects We instrument measures of institutional concentration
and total institutional ownership by orthogonalizing with respect to option grant pay-for-performance sensitivity salary and total compensation
Coefficients for industry and year controls are not presented First stage results are not presented One two and three asterisks denote significance at
10 5 and 1 levels
ExecuComp Sample 1992‒1997 Full Sample from ExecuComp 1992‒2002
PPSOG Salary TC PPSOG Salary TC
Change in Shareholder Wealth 144E-05 0004 02133 -409E-06 00011 00157
(308) (775) (596) (691) (638) (068)
Lag1 Change in Shareholder Wealth -240E-05 -00004 00866 118E-06 00002 01422
(131) (087) (346) (097) (116) (51)
Lag1 Institutional Concentration -00898 668681 3263143 02024 916656 2856417
(052) (801) (176) (163) (1545) (1004)
Lag1 Total Institutional Ownership 01021 -43997 2512587 02983 -104171 -1028347
(09) (084) (153) (4) (269) (487)
Lag1 Tobinrsquos Q 00754 -225606 119297 00655 -158657 2997205
(596) (1649) (053) (834) (2535) (727)
Lag1 Log Market Capitalization -04234 828713 7864046 -04002 825106 1685638
(2195) (8132) (2977) (4106) (11133) (373)
CEO Dummy 15848 3214117 2950274 19149 3205862 4529101
(2223) (9389) (2255) (3802) (1331) (2758)
Overall R-Sq -- 04981 01206 -- 04738 01178
Number of Observations 41305 41305 41305 85518 85518 85518
33
Table IX Two Stage Least Squares (Reverse Causality) Tests - Executive Compensation Measures as a Function of Top 5 Ownership
This table shows the coefficients from 2SLS regressions of executive compensation measures against the lagged change and contemporary change
in shareholder wealth instrumented Top 5 institutional ownership Tobinrsquos Q natural logarithm of firm market capitalization and dummy variable
controls for CEO and year effects We instrument Top 5 institutional ownership by orthogonalizing with respect to option grant pay-for-
performance sensitivity salary and total compensation Coefficients for industry and year controls are not presented First stage results are not
presented One two and three asterisks denote significance at 10 5 and 1 levels
ExecuComp sample 1992‒1997 Full Sample 1992‒2002
PPSOG Salary TC PPSOG Salary TC
Change in Shareholder Wealth 136E-05 00041 02128 -407E-06 00011 00146
(293) (79) (596) (692) (619) (063)
Lag1 Change in Shareholder Wealth -241E-05 -00003 00868 116E-06 00003 01439
(133) (07) (347) (096) (142) (514)
Lag1 Top 5 Institutional Ownership -03774 -187512 -3848531 00179 -104261 -126659
(163) (186) (166) (011) (149) (436)
Lag1 Tobinrsquos Q 00731 -223886 10074 00642 -155467 3115073
(573) (1631) (045) (819) (2509) (75)
Lag1 Log Market Capitalization -0416 775637 7696302 -04015 764632 1462536
(2345) (10084) (3766) (3688) (13708) (4072)
CEO Dummy 15854 3211991 2950158 19152 3203152 4519385
(2223) (9381) (2254) (3804) (133) (2752)
Overall R-Sq -- 04967 01202 -- 04722 01168
Number of Observations 41305 41305 41305 85518 85518 85518
34
Table X Specification Tests utilizing Box-Cox Transformation of Market Capitalization
This table presents results from a Box-Cox transformation of market capitalization In
each model we specify transformation of market capitalization only Panel A shows
results for the period 1992‒1997 and Panel B shows results for the entire period
1992‒2002 The first part of each panel shows the Lambda coefficient that applies to
the transformed market capitalization The second part of each panel shows log
likelihood scores with LR Chi2 Statistics in parentheses ‒ these test for the
appropriateness of a linear or log model specification
Panel A Our ExecuComp sample 1992‒1997
(1) (2) (3)
PPSOG Salary TC
Lambda -05143 01881 02876
(-1914) (2413) (1329)
Test H0 (Log Likelihood Score)
Lambda = -1 -1279748 -3178575 -4795326
(23866) (85952) (99688)
Lambda = 0 -1280152 -3138456 -4791159
(31934) (57147) (16339)
Lambda = 1 -1283157 -3163807 -4793506
(92039) (564156) (63292)
Panel B Full ExecuComp sample 1992‒2002
(1) (2) (4)
PPSOG Salary TC
Lambda -044 01853 04036
(-3015) (40) (3666)
Test H0 (Log Likelihood Score)
Lambda = 0 -2688334 -6476494 -10384490
(66599) (160356) (126236)
Lambda = 1 -2694886 -6538506 -10386795
(197644) (1400605) (172345)
35
Figure 2 Institutional Ownership Measures (LHA) vs Total
Compensation (RHA) by Market Capitalization
0
10
20
30
40
50
60
70
1 2 3 4 5 6 7 8 9 10
$00
$10
$20
$30
$40
$50
$60
$70
$80
$90
Millio
ns
InstitutionalConcentration (LHS)Top5 Ownership (LHS)
Institutional TotalOwnership (LHS)Total Compensation(RHS)
Page 3
Electronic copy available at httpssrncomabstract=1009309
3
We find that Institutional Top 5 Concentration according to the HS definition does
not raise option grant pay-for-performance sensitivity as it has no statistically significant
impact either way once one introduces a standard logarithmically transformed size
control Similarly lagged Institutional Concentration has a negative but statistically
insignificant effect on the pay-for-performance sensitivity of cash (salary plus bonus)
compensation when interacted with the lagged change in shareholder wealth HS obtain a
significant positive coefficient using the contemporaneous change in shareholder wealth
but one need replace this by the lagged change in shareholder wealth in keeping with
what influential institutional investors knew at the time Additionally HS find that lagged
concentrated institutional ownership sizably lowers both salary and total direct
compensation and this is statistically significant but we find the reverse ‒ it appears to
raise both and the findings are highly statistically significant also following a standard
logarithmically transformed size control
Upon closer examination several aspects of the HS research design are of interest
First the primary argument in our paper is that the use by HS of the lagged value of the
level of market capitalization as the size control in regressions (HS p2359) rather than
the conventional lagged logarithm of capitalization distorts the coefficient on the
measure of institutional influence3 A number of authors have drawn attention to falling
pay-for-performance sensitivities as firm size increases (eg Jensen and Murphy (1990)
Garen (1994) Hadlock and Lumer (1997) Schaefer (1998) Murphy (1999) Jin (2002)
and Chichello (2005)) and hence the need to have effective size controls In the HS
dataset the raw mean market capitalization ranges from $1181m in decile 1 to $36654m
in decile 10 (see Table III below) a difference of 3104 fold reducing to only 22 with the
3 HS (p2357) also use net assets and total assets as controls but in keeping with the use of raw
capitalization retain them in their raw form
4
logarithmic transformation While this skewed nature of the level of market capitalization
does not per se rule out its use specification tests such as Box-Cox and the replacement
of market capitalization by its cumulative distribution (see Section VI below) indicates a
possible specification error Moreover using the lagged natural logarithm of market
capitalization as a firm size control in the actual HS data with the same dependent
variable and no other changes raises the explanatory power (R-Squared) of the HS model
by between 135 and 1516 percent4
Given that firm size is an important determinant of executive compensation selecting
the correct functional form for firm size is a vexing and important issue Many estimates
including those of Murphy (1985) Rosen (1992) Gibbons and Murphy (1992) Huson
Parrino and Starks (2001) Baker and Hall (2004) Gabaix and Landier (2008) Edmans
Gabaix and Landier (2009) Babenko (2009) and Aggarwal Erel and Ferreira (2011)
use a logarithmic specification with the elasticity between firm size and executive total
pay in the range of 03 and 04 HS seems to employ inadvertently a firm size measure in
levels (also alternative firm size measures also in levels) and consequently imply an
elasticity value (exponent on market capitalization) of unity
Second we question the use of the HS Top 5 Concentration measure of institutional
influence based on the influence of concentrated institutional investors relative to their
peers that appears to lead inadvertently to an exacerbation of the size-control problem In
their companion theoretical model AHS arrive at a different and what might seem a more
natural specification namely Top 5 Ownership which measures concentration relative to
all shareholders not just institutional peers The following example illustrates the
4 The lower result is based on OLS regressions of model 1 of Table II of HS with the R-Squared rising from
00853 to 00968 once Market Capitalization is logged and the higher result model 1 of Table V of HS is
based on an improvement in R-Squared from 04604 to 05302 (see Table VI below)
5
differences between these two methods If there are one hundred shares outstanding in a
small company and the only institution owns one share that company is concentrated
according even though that institution may lack effective monitoring power Instead if
one deflates by shares on issue this results in a concentration ratio of 1100 If on the
other hand ten institutions collectively own 500 shares in a much larger company of
1000 shares and the Top Five collectively own 250 then the HS concentration ratio falls
to only 250500 or 05 and utilizing shares on issue 025 in this case As is confirmed by
the data one expects Top 5 Concentration to correlate negatively with firm size and Top
5 Ownership the reverse given a positive association between institutional ownership and
company size We then re-examine the monitoring hypothesis using this AHS proposed
measure of institutional investor influence that intuitively relates better to institutional
proxy voting power
Third we address issues of reliance on potentially biased self-reported values of option
grants that result in the exclusion of 16 percent of the HS sample size for tests on total
(direct) compensation Forth we argue that tests that HS perform using an alternative
measure of pay-for-performance sensitivity to provide support for the main hypothesis of
the study require one to get the timing right in terms of the lag structure of performance
and concentration Finally we address issues associated with their tests of endogeneity
and reverse causality These issues collectively might suggest overlooking of a
monitoring effect with the main driving forces size related
The next section reviews the HS study in an effort to tease out what might really be
going on Section II examines the literature for studies that in replicating HS provide
independent support for their findings It also reports findings on the actual proxy voting
record of mutual funds and actions of institutional investors in reducing pay Section III
describes the data and methodology Section IV analyzes the relation between
6
institutional influence and measures of pay-for-performance sensitivity Section V
documents the relationship between institutional influence and total and fixed
compensation Section VI examines the robustness of our findings and the final Section
concludes
I Review of Hartzell and Starks Methodology
We comment on five areas of the HS methodology 1) Size control 2) Measurement of
institutional influence 3) Testing the Jensen-Murphy measure of incentives 4) Reliance
on incomplete self-reported valued for option grants and 5) Endogeneity
A Size Control
In our paper we argue that HS unintentionally identify incentive-size and pay-size
effects5 not a monitoring effect It can be seen in Figure II that executive total pay for
HSrsquos Top 5 institutional concentration measure rises from as little as $072m for the
smallest market capitalization decile to $85m for the highest This twelve-fold pay-size
effect in the deciles can cause problems if one utilizes an untransformed size measure
Due to skewness in the size measure using this in its untransformed state appears to
weaken its effectiveness leaving open the possibility that correlations they identify
between incentives and pay levels on the one hand and Top 5 Concentration on the other
are due largely to size-related effects
Table III displays this strong pay-size effect based on our full 1992‒2002 dataset and
graphically in Figure 1 It shows that the incentive measure adopted by HS declines from
$272 per $1000 change in shareholder wealth for the smallest market capitalization
decile of firms to approximately $030 for the largest Table III and Figure 1 show the
5 The multiplicative model of Edmans Gabaix and Landier (2009) finds that incentive pay increases with
firm size (elasticity value of 13) and the empirical elasticity is 037
7
strong relation between institutional concentration as measured by HS and market
capitalization For the smallest decile of market capitalization firms institutional
concentration averages 61 percent declining to about 275 percent for the highest decile
This negative relationship between firm size and institutional concentration seems to arise
because institutional investors do generally favor small firms This causes problems
because in the limit a small company may have only one institutional investor owning
only a small percentage of the company Hence it would appear to be fully concentrated
even though the ability of the institutional investor to influence pay decisions might be
small Thus it is not at all surprising that the correlation between the logarithm of market
capitalization and their institutional concentration measure is both negative and high in
absolute value at 63 percent Their methodology appears to favor small companies that
Table III shows to have low pay levels and high pay for performance sensitivity
Moreover if one does not transform the raw size measure it implicitly assumes that a
given dollar change in firm size has the same absolute impact on pay for a very large
company as it does for a small company Indeed the conventional logarithmic
specification for firm size provides proportionate effects that imply a pay-size elasticity of
approximately one-third instead of unity6
B Measurement of Institutional Influence
We argue that there are additional issues related to the HS measure of institutional
influence ndash institutional concentration In their companion article to HS AHS develop a
model where they capture institutional influence by a single large long-term institutional
6 Note that unlike an earlier version of our paper we leave the HS dependent variables in levels and hence
untransformed While we do this to minimize the alterations that we make to the HS specification ideally
the elasticity specification is preferable especially for the compensation variables Transforming the
dependent variables does not significantly affect the results
8
investor They measure influence relative to total shareholding including private or
atomistic shareholders Presumably they do this to capture relative activism and proxy
voting strength However when it comes to the empirical implementation of the model
AHS adopt the same relative institutional concentration measure as in HS (AHS (p12))
potentially downplaying the relative strength of institutions compared with other
shareholders They justify this because there is generally more than one institutional
investor and there will be free-riding problems with multiple investors that only the
largest institutional investors can overcome While we accept their premise of free-rider
problems it is not obvious why the deflator is lsquoinstitutional shareholdingrsquo rather than
lsquototal shareholdingrsquo from this perspective7 If the strength of the size control can be
queried scaling by total shareholding (ie Top 5 Ownership) produces a measure that
remains relatively unaffected across the size deciles It is 20 percent for the smaller
decile declining to 14 percent for the largest decile whereas for the HS concentration
measure it declines by over 50 percent from 61 percent to 28 percent (see Table III and
Figure 1)
C Testing the Jensen-Murphy Measure of Incentives
The testing of their secondary measure of executive pay-for-performance sensitivity
(see HS Table IV) based on Jensen and Murphy (1990) raises issues other than the size
control HS regress changes in compensation (cash and total) on lagged institutional
influence that has been interacted with the contemporaneous change in shareholder
wealth to capture a pay-for-performance relationship Such a test is appealing ndash
7 AHS do point out one benefit of the institutional concentration measure Institutional investors may prefer
companies with lsquobetterrsquo incentive compensation structures Could using institutional shareholding rather
than total shareholding as the deflator alleviate this problem Presumably both the numerator and
denominator will have the same bias and hence error cancellation is a possibility
9
institutions should increase compensation when the firm has performed well Following
Greene (2000 p326) to execute such a test one needs to include in level form variables
that are then interacted Hence the HS model specification ought to be
1 2 3 1 4 1it it it it it itCompn SW IC SW IC (1)
where itCompn is the change in executive compensation itSW is the contemporaneous
change in shareholder wealth which also appears in the interaction term and 1itIC is
lagged institutional ownership concentration However HS (p2363) omit the
contemporaneous term itSW and the lagged term 1itIC which should appear in levels
as well as in the interaction term Most importantly HS interact their lagged institutional
concentration measure 1itIC with the contemporaneous change in shareholder wealth
itSW As it is after all pay-for-performance concentrated institutions need to condition
their influence on some observed measure of firm performance Contemporaneous firm
performance is unknown at the time institutional influence is measured Reflecting this
timing insight leads to a new Greene (2000) specification
1 2 1 3 1 4 1 1it it it it it itCompn SW IC SW IC (2)
such that all the lagged RHS independent variables that now appear in the interaction
term are included on a stand-alone basis
Reliance on incomplete self-reported values for option grants
Prior to receiving the HS dataset we tried to replicate the findings of HS to validate
our datasets for the HS period and the longer period 1992‒2002 While we initially
focused on the same period as HS (ie 1992‒1997) we found that our sample size was
larger at 47765 observations than HSrsquos at 33928 observations This discrepancy could
have arisen if HS did not replace missing and self-reported option valuations using their
10
Black-Scholes option methodology This may introduce a downward pay bias since self-
reported option-grant pay is subject to possible understatement and there are 13839
observations with likely high option pay dropped altogether
Endogeneity
HSrsquos endogeneity testing raises concerns about reverse causality Institutions may
prefer to hold firms with certain compensation structures such as incentivized executives
HS rightly highlight this concern and use instrumental variables to address this issue HS
assume that institutional ownership reflects any reverse causality Using an instrumental
variables approach to resolve this problem involves finding a variable that influences or
explains institutional ownership but not directly executive compensation They use share
turnover in a firm as an instrument for institutional ownership However theoretical work
by Holmstrom and Tirole (1993) indicates that measures of stock liquidity and trading
activity are indeed important determinants of executive market-based incentives as Kang
and Lui (2008) and Dikolli Kulp and Sedatole (2009) find empirically As such one can
make a case for adopting another instrument rather than share turnover for institutional
ownership
II Previous Research
Prior research fails to address issues concerning possible fragility of the HS results
AHS replicate the HS findings on both incentives and pay levels using the same dataset as
HS and they apply the same definition of institutional concentration and use market
capitalization expressed as a level Cadman Klasa and Matsunaga (2010 Table 6) set out
to replicate HSrsquos (Table II) model of option grant pay-for-performance sensitivity when
they compare a dataset of ExecuComp and Non-ExecuComp firms for the period 2000-
2007 Their ExecuComp results for this sample period are quite similar to our results for
11
the 1992‒2002 period for the same specification controlling for the logarithm of market
capitalization but they do not regard these results as a test of HS as their focus is entirely
on comparing the two groups of stocks Nor do they investigate the impact of
concentration on either cash compensation or total pay Kang and Liu (2008) examine
pay-performance sensitivity use a version of the HS institutional concentration measure
as a control variable in a different framework deriving from Holmstrom and Tirole
(1993) Dikolli Kulp and Sedatole (2009) argue that firms with transient ownership will
issue relatively more equity incentives to counter short-termism with their results robust
to the inclusion of the HS concentration variable
We hoped to find supportive evidence from proxy voting behavior in support of HSs
findings However evidence from actual mutual fund proxy voting records suggests that
most oppose reductions in CEO pay but overall the evidence is mixed8 Davis and Kim
(2005) examine proxy votes of mutual funds for 2004 and find that for a sample of
Fortune 1000 proxy contests in 45 cases that attempted to limit executive pay mutual
funds were unanimously opposed to such attempts Rothberg and Lilien (2006) also
examine proxy-voting data and find that mutual funds voted 66 percent of the time in
managementrsquos favor on issues of executive compensation Ertimur Ferri and Muslu
(2011) find that institutional investors lack sophistication in that they do not target
excessively paid CEOs relative to CEOs with high predicted-pay based on economic
determinants but exert a moderating influence ndash a $23 million reduction ndash for those they
deem ldquoexcessively paidrdquo CEOs
III Data and Methodology
8 We do acknowledge that the HS results are for all institutional investors while the transparency with
respect to proxy voting applies only to mutual funds
12
This Section discusses the data used in the study the construction of variables
descriptive statistics and the methodology Two datasets are used ndash one is the original
dataset generously supplied by Jay Hartzell and Laura Starks which covers 1992‒1997
second a more comprehensive dataset covering 1992‒2002 is computed from the same
data sources used in the HS dataset This is Standard and Poorrsquos (SampP) ExecuComp for
compensation data Thomson Financial CDASpectrum database for institutional holdings
to compute measure of institutional influence SampPrsquos Compustat and the Center for
Research in Stock Prices (CRSP) for firm level accounting and stock price data Our more
comprehensive dataset contains 2559 firms compared with 1914 for the HS dataset Our
full matched sample yields 97679 executive years
A Institutional Influence Measures
To measure institutional influence we employ two ownership-based metrics The first
measure of institutional influence is HSrsquos Institutional Concentration We use this for
direct comparison to the HS analysis The second measure of institutional influence
changes the denominator of the HS measure from institutional shares held to total shares
outstanding in the firm with nomenclature Top 5 Ownership9 We compute both
measures at the fiscal year end of the firm We adjust all compensation measures and
asset prices for inflation to 30 June 2002 prices for our comprehensive sample compared
to the HS dataset that is in the dollars of the day
B Compensation Measures
The ExecuComp database reports several raw measures of executive compensation for
the top five executives (where available) including total salary bonus long-term
9 Top 5 Ownership is included in the HS database but not utilized by them in reported tests as well as
constructed by us for the more comprehensive sample
13
incentives and number of option grants For our more comprehensive dataset we replicate
all the compensation measures used by HS In common with HS we compute the pay-for-
performance sensitivity of option grants using the Yermack (1995) approach This
measure shows the dollar change in value of the executive options for every dollar change
in the value of the firm This involves calculating the option delta using the Black-
Scholes formula adjusted for dividends10
The ratio of the number of shares represented
by the award of options divided by the diluted number of shares outstanding at the
beginning of the year multiplies the option delta with units given by a $1000 change in
firm value Summing salary bonus long-term incentive plan payouts and stock and
option grants provides total direct compensation Options are valued using the Black and
Scholes (1973) formula for European call options adjusted by Merton (1973) to
incorporate dividends
C Descriptive Statistics
Table I presents descriptive statistics of key variables employed in the study according
to our comprehensive dataset11
Institutions own on average 546 percent of the firms in
which they invest This is almost identical to the 531 percent found by HS The
institutions with the five largest holdings account 45 percent of institutional ownership
Once again the figure for HS is almost identical at 44 percent The institutions with the
five largest holdings own on average 229 percent of shares outstanding in a firm (HS
have 223 percent) Average total compensation for executives is $2583 million
considerably larger than that documented by HS ($1250 million) but the HS figure is an
10 When there is more than one grant during the year a weighted average delta is used
11 With the data supplied to us by Hartzell and Starks we are able to replicate exactly their summary data
(HS Table I) for all but cash compensation Salary is included but bonus payments are missing The
replicated table is available on request
14
amalgam of price levels over their sample period 1992‒1997 and ours are in consistent
dollars of June 30 2002 However there has also been significant growth in
compensation levels from 1998 to 2002
ltltINSERT TABLE Igtgt
Based on our comprehensive data in Table II we examine correlations between the
key variables employed in the study Of concern is the large negative correlation of 63
percent which has been discussed previously (Section IIA) between the HS measure of
institutional concentration and the natural logarithm of firm market capitalization
ltltINSERT TABLE IIgtgt
In view of this high correlation Table III provides a breakdown of the three key pay
variables and three institutional ownership variables by size deciles We find that the high
institutional concentration but low institutional ownership share of total shareholdings
mentioned above is evident in the smallest decile of firms by size This decile has low
salary and even lower total compensation in a relative sense but by far the highest pay-
for-performance sensitivity of option grants Small firms award executives a far higher
proportion of market capitalization Since small firm market-capitalization is negligible in
comparison with the highest decile (300 times larger in magnitude) total compensation is
very small in absolute value Moreover due to the domination of the pay-size effect base
salaries are also exceedingly small
ltltINSERT TABLE IIIgtgt
Figure 1 based on Table III shows how as institutional concentration falls with
increases in stock size (but also as the total share of institutional ownership increases)
and the pay-for-performance sensitivity of option grants falls Thus because of what
appears to be a weak size control and size-influenced concentration measure HS
15
potentially pick up a size effect as their main result but not necessarily a monitoring
effect Figure 2 shows that as firm size increases up to decile 8 at least the share of
institutional total ownership increases as does total compensation
ltltINSERT FIGURES 1 AND 2 gtgt
D Methodology
Following HS we use two estimation techniques to examine the effect of institutional
investor influence on executive compensation Due to the censored nature of option grant
pay-for-performance sensitivity this compensation measure most appropriately uses
Tobit model specification whereas an Ordinary Least Squares (OLS) model specification
is used for all other compensation measures We model compensation measures as a
function of contemporaneous and lagged change in shareholder wealth institutional
influence Tobinrsquos Q and market capitalization Controls for CEO industry and year
effects are also used Augmentation occurs to test the effect of institutional influence on
the change in executive cash and total compensation with an interaction term that is the
product of institutional influence and the change in shareholder wealth
The variables used to model executive compensation have the following definitions
The lagged and contemporary measure of change in shareholder wealth is simply the
difference in market capitalization of the firm The measures of institutional influence
used are Institutional Concentration and Top 5 Ownership Use of the two types of
institutional influence occurs in separate models In order to replicate the HS
methodology models using Institutional Concentration have an additional control
consisting of lagged total institutional holdings and Tobinrsquos Q controls for performance
and growth opportunities Market capitalization of the firm is the number of shares
outstanding multiplied by the share price with inclusion both in levels and natural
16
logarithmic form A dummy variable is included to identify whether the executive is the
CEO Dummy variables are also included to control for industry effects This variable
takes a value of 1 for the two-digit Standard Industrial Classification (SIC) in which the
firm operates A year fixed effects (dummy) variable is included taking a value of 1 if the
observation was from the given year
IV Executive Incentives as a Function of Institutional Investor Influence
This section presents the results for the effect of institutional influence on option grant
pay-for-performance sensitivity and changes in compensation
A Option Grant Pay-for-Performance Sensitivity
Table IV presents the results for the effect of institutional influence on option grant
pay-for-performance sensitivity
ltltINSERT TABLE IVgtgt
Model (1) is a direct replication of HS using the HS database as supplied to us by HS
with all the variables computed in the same way and in the same functional form as with
HS The coefficient and t-statistic on lagged institutional concentration interacted with the
change in shareholder wealth is 13994 and 767 that compares to 1358 and 842 in HS
We are unable to account for these minor discrepancies for values that should be
identical In model (2) the only change to the model is that we use the natural logarithm
of market capitalization instead of the level of market capitalization This single change
has a sizeable effect on the coefficient on institutional concentration The coefficient on
institutional concentration is now -00692 and is insignificant The fact that an order
preserving transformation to a control variable can change the results is of interest The
log-transformation to market capitalization also changes the coefficient on Top 5
Ownership in models (3) and (4) The coefficient on Top 5 Ownership is significant in
17
both model (3) and in model (4) but the better size control reduces the coefficient by over
25 percent The better size control does not affect Top 5 Ownership by as much as its
design ensures less sensitivity to what could be inadequate size controls
We also test the effect of institutional influence on option grant pay-for-performance
sensitivity over the period 1992‒2002 This sample is then approximately three times as
large as the sample used by HS Institutional concentration is highly positively significant
in model (5) similar to the results for the period 1992‒1997 When we use the log of
market capitalization in model (6) the coefficient on institutional concentration remains
positive and significant at the 10 percent level with a t-statistic of 166 Of importance
though is that the coefficient on institutional concentration is one-eighth of the coefficient
in model (5) Even for a measure of institutional influence which we believe is less
sensitive to firm size the use of log market capitalization reduces the coefficient on Top 5
Ownership by approximately 50 percent from model (7) to model (8)
B Pay-for-Performance Sensitivity due to Changes in Cash and Total
Compensation
For robustness HS examine the effect that institutional influence conditioned on firm
performance has on a changes in cash and total executive compensation using the general
method of Jensen and Murphy (1990) to compute pay-for-performance sensitivity
ltltINSERT TABLE V gtgt
Table V presents the effect of institutional influence on the change in cash
compensation using the HS size control measure throughout12
Model (1) replicates the
specification in HS for the period of 1992‒1997 and we produce a similar result ndash the
interaction of lagged institutional concentration and contemporaneous change in
12 This test uses our data for the two periods 1992‒1997 and 1992‒2002 We do not use the HS data for
cash compensation as data for bonus is missing in the data supplied to us
18
shareholder wealth positively and significantly relates to a change in executive cash
compensation In model (2) we include the contemporaneous change in shareholder
wealth and the lagged institutional concentration as stand-alone controls as indicated by
Greenersquos (2000) specification of equation (1) above and the interaction term remains
significant In model (3) we interact lagged institutional concentration with the lagged
change in shareholder wealth as in equation (2) above that tries to take account of what
participants could have observed at the time The interaction term is then negative but
insignificant altering the HS result Model (4) provides a similar result where our
preferred proxy Top 5 Ownership captures concentrated institutional influence In
models (5) to (8) we use our full sample In model (5) we find that the lagged
institutional concentration interaction with the contemporaneous change in shareholder
wealth is positive and significant in HSrsquos specification However including change in
contemporaneous shareholder wealth and lagged institutional concentration as separate
control variables as suggested by equation (1)rsquos specification removes the significance
from the interaction term in model (6) Moreover in model (7) when lagged institutional
concentration is interaction with the lagged change in shareholder wealth in accord with
equation (2) this interaction term is now negative and significant reversing the HS
finding We obtain a similar result in model (8) when we measure institutional
concentration influence using Top 5 Ownership13
V The Levels of Executive Fixed and Total Compensation as Functions of
Institutional Investor Influence
13 In unreported results we use the log of market capitalization as a size control These changes to the size
control do not dramatically change the findings from Table V The reason why this size issue is not as
important as in the findings for option grants pay-for-performance is that the change in cash compensation
is not so strongly influenced by firm size In unreported results available on request we also examine the
effect of institutional influence on changes in total compensation with similar results to the change in cash
compensation
19
Two measures of compensation are examined ndash salary and total compensation
ltltINSERT TABLE VI gtgt
Table VI examines the effect of institutional influence on the level of salary In model
(1) we use the HS sample and specification to obtain almost identical results The
coefficient on institutional concentration is (negative) $247509 that is almost identical to
(negative) $247967 in HS Both are highly significant In model (2) we make a single
change to the model ndash we use the natural logarithm of market capitalization This single
change reverses the sign of institutional concentration with the coefficient still being
significant at the 1 percent level In models (3) and (4) we measure institutional influence
using Top 5 Ownership In model (3) where we measure firm size using the level of
market capitalization Top 5 Ownership is insignificant When we use the natural
logarithm of market capitalization in model (4) Top 5 Ownership is significant and
positively relates to the salary level In models (5) to (8) we estimate the same models
using our full sample We find similar results for the full sample as for the period 1992ndash
1997
ltltINSERT TABLE VII gtgt
Table VII examines the effect of institutional influence on total compensation In
model (1) we replicate the HS methodology using the HS data to obtain a similar
significant negative relation between institutional concentration and total compensation of
(negative) $1608633 However after changing to a logarithmic specification of market
capitalization in model (2) institutional concentration increases total compensation by a
significant $32400014
Importantly when institutional influence is measured using Top 5
14 = (5688751 times 0101)
20
Ownership an insignificant relation exists with total compensation in model (3) using the
HS size control but it becomes both positive and significant in (4) with the logged size
control In models (5) and (6) where models are estimated using the full sample we
observe a similar reversal in sign as occurs in models (1) and (2) However now the
magnitude of the reversal is almost equal with a difference of about $67m that is very
large indeed In model (5) the coefficient on institutional concentration is (negative)
$342m and then in model (6) the coefficient is $33m Top 5 Ownership is insignificant
in both models (7) and (8) unlike the results for Salary
VI Robustness
Two issues are examined in is this Section ndash reverse causality and the appropriate
transformation of market capitalization A methodological issue inherent in examining the
effect of ownership-based measures of institutional influence on executive compensation
is endogeneity driven by reverse causality We utilize an instrumental variables approach
where orthogonalized institutional influence measures are regressed on each of the
compensation measures lagged option grant pay-for-performance sensitivity salary and
total compensation In addition market capitalization and a CEO dummy are included in
each regression We use the residual from each regression as an instrument for
institutional influence Such a process removes any preferences for compensation from
the institutional influence measures
ltltINSERT TABLES VIII AND IXgtgt
Table VIII presents the results where we instrument the HS measure lagged
institutional concentration with a compensation-orthogonalized equivalent The analysis
indicates that the instrumented institutional concentration measure is positively and
significantly related to the level of salary and total compensation and insignificantly
21
related to option grant pay-for-performance sensitivity (PPSOG) Table IX presents
results where we instrument Top 5 Ownership with a compensation-orthogonalized
equivalent The instrumented Top 5 Ownership measure insignificantly relates to option
grant pay-for-performance sensitivity and barely negatively relates to salary although it
negatively relates to total compensation suggesting reverse causality is driving a
significant portion of our previous results
ltltINSERT TABLE Xgtgt
Another issue examined in this section is the appropriate transformation of market
capitalization Aneuryn-Evans and Deaton (1980) provide an evaluation of methods
comparing linear and log-linear specifications We use the Box-Cox test to compare
logarithm versus linear specifications In applying the Box-Cox test we specify
transformation of only market capitalization ndash the dependent variable and all other
independent variables remain unchanged Table X presents the lambda coefficient for
each model and the log likelihood score In all models in Panel A and B a lambda of zero
receives a higher (closer to zero) log likelihood score suggesting the log specification is
appropriate
VII Conclusion
Hartzell and Starks (2003) appears to document an effective role for institutional
investors in increasing executive incentives and lowering compensation levels Our paper
re-addresses the issue of concentrated institutional monitoring from a perspective
informed by an additional decade of research since their paper appeared We cannot find
evidence that concentrated institutional investors really have any positive (or negative for
that matter) effect on executive incentive levels While superficially we find evidence
that they may raise executive pay levels this is most likely driven by an institutional
22
preference for highly remunerated executives in large companies and thus by reverse
causality Moreover our findings are not consistent with the rent-extraction hypothesis
suggesting that firms systematically overpay executives in the absence of effective
monitoring by outside parties such as institutional investors
23
References
Almazan Andres Jay C Hartzell and Laura T Starks 2005 ldquoActive Institutional
Shareholders and Costs of Monitoring Evidence from Executive Compensationrdquo
Financial Management Winter 5ndash34
Aneuryn-Evans Gwyn and Angus Deaton 1980 ldquoTesting Linear versus Logarithmic
Regression Modelsrdquo Review of Economic Studies 47 (1) Econometrics Issue 275ndash
291
Babenko Ilona 2009 ldquoShare Repurchases and Pay-Performance Sensitivity of Employee
Compensation Contractsrdquo Journal of Finance 64117-150
Baker George P and Brian J Hall 2004 ldquoCEO Incentives and Firm Sizerdquo Journal of
Labor Market Economics 22 767ndash798
Bebchuk Lucian A Jesse M Fried and David Walker 2002 ldquoManagerial power and
rent extraction in the design of executive compensationrdquo University of Chicago Law
Review 69 751ndash846
Bebchuk Lucian A and Jesse M Fried 2003 ldquoExecutive Compensation as an Agency
Problemrdquo Journal of Economic Perspectives 17 71ndash92
Bebchuk Lucian A and Jesse M Fried 2004 Pay without Performance The Unfulfilled
Promise of Executive Compensation Cambridge MA Harvard University Press
Bertrand Marianne and Sendhil Mullainathan 2001 ldquoAre CEOs Rewarded for Luck
The Ones Without Principals Arerdquo Quarterly Journal of Economics 116 901ndash32
Black Fischer and Myron Scholes 1973 ldquoThe Pricing of Options and Corporate
Liabilitiesrdquo Journal of Political Economy 81 637ndash54
Cadman Brian Sandy Klasa and Steve Matsunaga 2010 ldquoDeterminants of CEO Pay A
Comparison of ExecuComp and Non-ExecuComp Firmsrdquo The Accounting Review 85
1511ndash1543
Chhaochharia Vidhi and Yaniv Grinstein 2009 ldquoCEO Compensation and Board
Structurerdquo Journal of Finance 64 231ndash261
Cichello Michael S 2005 ldquoThe impact of firm size on pay-performance sensitivitiesrdquo
Journal of Corporate Finance 11 609ndash627
Davis Gerald F and E Han Kim 2007 ldquoBusiness ties and proxy voting by mutual
fundsrdquo Journal of Financial Economics 85 552ndash570
Dikolli Shane S Susan L Kulp and Karen L Sedatole 2009 ldquoTransient Institutional
Ownership and CEO Contractingrdquo Accounting Review 84 737ndash770
Edmans Alex Xavier Gabaix and Augustin Landier 2009 ldquoA Multiplicative Model of
Optimal CEO Incentives in Market Equilibriumrdquo Review of Financial Studies 22
4881ndash4917
Edmans Alex 2009 ldquoBlockholder Trading Market Efficiency and Managerial Myopiardquo
Journal of Finance 64 2481ndash2513
Ertimur Yonca Fabrizio Ferri and Volkan Muslu 2011 ldquoShareholder Activism and
CEO Payrdquo Review of Financial Studies 24 535ndash592
24
Gabaix Xavier and Augustin Landier 2008 ldquoWhy Has CEO Pay Increased So Muchrdquo
Quarterly Journal of Economics 123 49ndash100
Garen John 1994 ldquoExecutive compensation and principal-agent theoryrdquo Journal of
Political Economy 102 1175ndash1199
Gibbons Robert and Kevin J Murphy 1992 ldquoOptimal Incentive Contracts in the
Presence of Career Concerns Theory and Evidencerdquo Journal of Political Economy
100 468ndash505
Greene William H 2000 (4th
Edition) Econometric Analysis Prentice-Hall NJ
Guthrie Katherine Jan Sokolowsky and Kam-Ming Wan 2012 ldquoCEO Compensation
and Board Structure Revisitedrdquo Journal of Finance Forthcoming
Hadlock Charles R and Gerald B Lumer 1997 ldquoCompensation turnover and top
management incentives historical evidencerdquo Journal of Business 70 153ndash187
Hartzell Jay C and Laura T Starks 2003 ldquoInstitutional Investors and Executive
Compensationrdquo Journal of Finance 58 2351ndash2374
Huson Mark R Robert Parrino and Laura T Starks 2001 ldquoInternal Monitoring
Mechanisms and CEO Turnover A Long-Term Perspectiverdquo Journal of Finance 56
2265ndash2297
Jensen Michael C and Kevin J Murphy 1990 ldquoPerformance Pay and Top Management
Incentivesrdquo Journal of Political Economy 98 225ndash264
Jin Li 2002 ldquoCEO compensation diversification and incentivesrdquo Journal of Financial
Economics 66 29ndash63
Kang Qiang and Qiao Liu 2008 ldquoStock Market Information Production and Executive
Incentivesrdquo Journal of Corporate Finance 14 484ndash498
Merton Robert 1973 ldquoTheory of Rational Option Pricingrdquo Bell Journal of Economics
and Management Science 4 141-183
Morse Adair Vikram Nanda and Amit Seru 2011 ldquoAre Incentive Contracts Rigged by
Powerful CEOsrdquo Journal of Finance 66 1779ndash1821
Murphy Kevin J 1985 ldquoCorporate performance and managerial remuneration An
empirical analysisrdquo Journal of Accounting and Economics 7 11ndash42
Murphy Kevin J 1999 ldquoExecutive compensationrdquo in Orley Ashenfelter and David
Card Handbook of Labor Economics vol 3 Elsevier
Rosen Sherwin 1992 ldquoContracts and the market for executivesrdquo In Contract
Economics ed Lars Werin and Hans Wijkander 181ndash211 Cambridge MA
Blackwell
Rothberg Burton and Steven Lilien 2006 ldquoMutual Funds and Proxy Voting New
Evidence on Corporate Governancerdquo Journal of Business and Technology Law 1
157ndash184
Schaefer Scott 1998 ldquoThe dependence of payndashperformance sensitivity on the size of the
firmrdquo Review of Economics and Statistics 80 436ndash443
Yermack David 1995 ldquoDo Corporations Award CEO Stock Options Effectivelyrdquo
Journal of Financial Economics 39 237ndash69
25
Table I Univariate Descriptive Statistics
This table reports sample statistics for the key variables used in this study based on our comprehensive
ExecuComp dataset Institutional Concentration is the shares owned by the institutions with the five largest
holdings divided by the shares held by all institutions Top 5 Ownership is the proportion of shares
outstanding held by the institutions with the five largest holdings Total Institutional Ownership is the
proportion of shares held by all institutions Panel B shows the measures executive compensation The pay-
for-performance sensitivity (PPS) of option grants shows the change in executive wealth due to a $1000
change in firm in shareholder wealth Expressions for Salary and Total Compensation are in thousands of
dollars while Market Capitalization is in millions of dollars Panel C shows firm characteristics
Variable Mean Median Std Dev P10 P90
Panel A Institutional Investor Holdings
Institutional Concentration 0450 0424 0151 0282 0656
Top 5 Institutional Ownership 0229 0220 0101 0110 0355
Total Institutional Ownership 0546 0561 0206 0258 0801
Panel B Executive Compensation
PPS of Option Grants 1161 0156 3829 0 2822
Salary 36311 29462 25231 15179 67000
Total Compensation 258313 85878 980701 23875 511022
Panel C Firm Characteristics
Tobins Q 1744 1139 2738 0495 3322
Market Capitalization 560699 112094 1988108 19754 1048754
26
Table II Pairwise Correlation Coefficients
This table presents the correlations between the key variables based on our comprehensive
ExecuComp dataset Institutional Concentration is the shares owned by the institutions with the five
largest holdings divided by the shares held by all institutions Top 5 Ownership is the proportion of
shares outstanding held by the institutions with the five largest holdings Total Institutional
Ownership is the proportion of shares held by all institutions The pay-for-performance sensitivity
(PPS) of option grants is the change in executive wealth due to a $1000 change in shareholder
wealth
(1) (2) (3) (4) (5) (6) (7) (8)
(1) Institutional Concentration 1
(2) Top 5 Ownership 0268 1
(3) Total Ownership -0415 0706 1
(4) PPS of Option Grants 0122 0042 -0024 1
(5) Log Salary -0270 -0012 0171 0038 1
(6) Log Market Capitalization -0630 -0157 0261 -0156 0420 1
(7) Level Market Capitalization -0241 -0157 0007 -0058 0219 0534 1
(8) Tobins Q -0008 -0034 0002 0015 -0083 0066 0121 1
27
Table III
Descriptive Statistics by Market Capitalization Deciles
Market
Capitalization
Deciles
Market
Capitalization
($m)
Total
Compensation
($000)
Salary
($)
PPS Option
Grants
($)
Institutional
Concentration
()
Top 5 Ownership
()
Institutional
Total Ownership
()
1 (Small) 1181 76008 234700 2720 611 200 327
2 2668 90206 257955 1796 495 217 438
3 4193 105946 275120 1517 454 220 485
4 6045 129892 296067 1376 420 206 491
5 8742 155391 315864 1216 394 198 502
6 13086 191888 345942 1072 380 202 531
7 20790 239501 377021 0874 368 196 532
8 34920 314955 427808 0687 354 190 538
9 67747 450970 480319 0569 328 173 528
10 (Big) 366540 846066 614354 0297 275 138 502
28
Table IV Pay-for-Performance Sensitivity of Option Grants as a Function of Institutional Influence
Similar in construction to parts of HS (Table II) this table shows the coefficients from Tobit regressions of option grant pay-for-performance sensitivity per $1000 change in
shareholder wealth against the lagged change and contemporary change in shareholder wealth measures of institutional influence Tobinrsquos Q firm market capitalization and
dummy controls for CEO and year effects We estimate models (1) to (4) over the period 1992‒1997 using data supplied by HS and models (5) to (8) are estimated over the
whole sample Models (1) (2) (5) and (6) use Institutional Concentration as a measure of institutional influence which HS define as the shares owned by the institutions
with the five largest holdings divided by the shares held by all institutions These models control for total institutional ownership Models (3) (4) (7) and (8) use Top 5
Ownership as a measure of institutional influence which we define as the proportion of shares outstanding held by the institutions with the five largest holdings Coefficients
for industry and year controls are not presented One two and three asterisks denote significance at 10 5 and 1 levels
Hartzell and Starks Database 1992‒1997 Full Sample from ExecuComp 1992‒2002
(1) (2) (3) (4) (5) (6) (7) (8)
Change in Shareholder Wealth 77721 191781 89874 137368 -446 -218 -664 -203
(171) (429) (194) (317) (319) (168) (453) (158)
Lag1 Change in Shareholder Wealth 14446 99963 88719 8544 -144 0273 -0342 00852
(027) (212) (166) (184) (071) (015) (016) (005)
Lag1 Institutional Concentration 13994 -00692 26405 03304
(767) (032) (1402) (166)
Lag1 Total Institutional Ownership 08055 10434 07106 0937
(593) (781) (666) (898)
Lag1 Top 5 Institutional Ownership 10567 07457 12116 06293
(399) ( 283) (533) (278)
Lag1 Tobinrsquos Q 0068 00877 00647 00813 00498 00727 00508 00695
(337) (434) (319) (401) (504) (734) (51) (703)
Lag1 Market Capitalization -106202 -173515 -776 -117
(474) (742) (1011) (153)
Lag1 Log Market Capitalization -0297 -02319 -04111 -03908
(1576) (1439) (2997) (2841)
CEO Dummy 13745 13682 13765 13674 28197 28128 28155 28143
(1913) (1918) (1914) (1916) (3388) (3396) (3377) (34)
Number of Observations 33878 33878 33878 33878 97679 97679 97679 97679
29
Table V Pay-for-Performance Sensitivity of Cash Compensation (Salary plus Bonus) as a Function of Institutional Influence - Level Market Capitalization
Similar in parts to models (1) and (2) of HS (Table IV) this table displays the coefficients from OLS regressions of the change in cash compensation against the lagged
change and contemporary change in shareholder wealth measures of institutional influence Tobinrsquos Q firm market capitalization and dummy controls for CEO and year
effects We estimate models (1) to (4) over the period 1992‒1997 based on our ExecuComp sample and models (5) to (8) over the whole sample Models (1) (2) (5) and (6)
use Institutional Concentration as a measure of institutional influence which HS define as the shares owned by the institutions with the five largest holdings divided by the
shares held by all institutions These models control for total institutional ownership Models (4) and (8) use Top 5 Ownership as a measure of institutional influence which
we define as the proportion of shares outstanding held by the institutions with the 5 largest holdings Coefficients for industry and year controls are not presented One two
and three asterisks denote significance at 10 5 and 1 levels
Our Database based on ExecuComp 1992‒1997 Full Sample based on ExecuComp 1992‒2002
(1) (2) (3) (4) (5) (6) (7) (8)
Lag1 Change in Shareholder Wealth 00014 00017 00088 00089 -00006 -00007 00072 00086
(051) (062) (129) (103) (027) (029) (131) (16)
Change in Shareholder Wealth -00102 00102 00105 00032 00102 00101
(097) (292) (293) (055) (611) (608)
Change Shareholder Wealth Lag1 Institutional
Concentration
00435 00764 00363 00253
(463) (258) (689) (144)
Lag1 Change Shareholder Wealth Lag1
Institutional Concentration
-00197 -00283
(122) (197)
Lag1 Institutional Concentration -492182 -288861 400149 453623
(09) (056) (098) (114)
Lag1 Total Institutional Ownership 720688 538372 655091 559587 682069 726226
(276) (232) (281) (314) (308) (322)
Lag1 Change Shareholder Wealth Lag1 Top 5
Institutional Ownership
-0046 -00689
(091) (256)
Lag1 Top 5 Institutional Ownership 583044 1191634
(092) (25)
Lag1 Tobinrsquos Q -32761 -36806 -19422 -18023 -54283 -48681 -29534 -24243
(138) (172) (077) (078) (227) (245) (136) (111)
Lag1 Market Capitalization 00004 00004 00002 00006 00007 00008 00007 00007
(042) (044) (022) (07) (086) (092) (082) (09)
CEO Dummy 768035 765474 768334 773634 616315 61729 619028 618535
(354) (355) (356) (355) (417) (418) (419) (418)
Adjusted R-Sq 00057 0006 00052 0005 00097 00097 00098 00102
Number of Observations 41305 41305 41305 41305 85518 85518 85518 85518
30
Table VI Level of Salary as a Function of Institutional Influence
Similar in parts to models (1) and (2) of HS (Table V) this table displays the coefficients from OLS regressions of salary against the lagged change and contemporary change in
shareholder wealth measures of institutional influence Tobinrsquos Q firm market capitalization and dummy controls for CEO industry and year effects We estimate models (1) to
(4) over the period 1992‒1997 using the actual data supplied by HS and models (5) to (8) over the whole sample 1992‒2002 Models (1) (2) (5) and (6) use Institutional
Concentration as a measure of institutional influence which HS define as the shares owned by the institutions with the five largest holdings divided by the shares held by all
institutions These models control for total institutional ownership Models (3) (4) (7) and (8) use Top 5 Ownership as a measure of institutional influence which we define as
the proportion of shares outstanding held by the institutions with the 5 largest holdings Coefficients for industry and year controls are not presented One two and three
asterisks denote significance at 10 5 and 1 levels
Hartzell and Starks Database 1992‒1997 Full Sample based on ExecuComp 1992‒2002
(1) (2) (3) (4) (5) (6) (7) (8)
Change in Shareholder Wealth 04125E-4 0003 -2439E-4 0003 00021 00012 00024 00012
(057) (464) (031) (468) (1004) (72) (1041) (706)
Lag1 Change in Shareholder Wealth -00062 7856E-4 -0007 9003E-4 -00014 00002 -00014 00003
(594) (098) (592) (112) (522) (123) (5) (145)
Lag1 Institutional Concentration -2475087 555134 -2734983 943317
(3245) (710) (4965) (1671)
Lag1 Total Institutional Ownership 1020109 217488 936337 252464
(1904) (432) (2471) (704)
Lag1 Top 5 Institutional Ownership -148355 31354 -87926 423166
(139) (332) (118) (643)
Lag1 Tobinrsquos Q -195348 -252191 -192344 -25158 -119837 -146373 -122132 -143821
(2812) (3762) (2721) (3752) (248) (274) (2583) (2727)
Lag1 Market Capitalization 844E-06 103E-5 00034 00039
(2781) (3051) (2874) (3104)
Lag1 Log Market Capitalization 750021 723625 776657 732279
(7511) (9285) (11224) (14039)
CEO Dummy 2683978 2702238 2668244 2702417 3281133 3285103 3292175 3283145
(8625) (9277) (8246) (9271) (12739) (13616) (12336) (13605)
Adjusted R-Sq 04604 05302 04187 05295 03856 04588 03476 04574
Number of Observations 36280 36280 36280 36280 97679 97679 97679 97679
31
Table VII Level of Total Direct Compensation as a Function of Institutional Influence
Similar in parts to models (3) and (5) of HS (Table V) this table displays the coefficients from OLS regressions of total direct compensation against the lagged change and
contemporary change in shareholder wealth measures of institutional influence Tobinrsquos Q firm market capitalization and dummy controls for CEO industry and year effects
We estimate models (1) to (4) over the period 1992‒1997 and models (5) to (8) over the whole sample Models (1) (2) (5) and (6) use Institutional Concentration as a measure
of institutional influence which HS define as the shares owned by the institutions with the five largest holdings divided by the shares held by all institutions These models
control for total institutional ownership Models (3) (4) (7) and (8) use Top 5 Ownership as a measure of institutional influence which we define as the proportion of shares
outstanding held by the institutions with the five largest holdings Coefficients for industry and year controls are not presented One two and three asterisks denote significance
at 10 5 and 1 levels
Hartzell and Starks Database 1992‒1997 Full Sample from ExecuComp 1992‒2002
(1) (2) (3) (4) (5) (6) (7) (8)
Change in Shareholder Wealth 005779 007006 005282 006791 00423 00148 0046 00138
(415) (581) (369) (565) (203) (07) (22) (066)
Lag1 Change in Shareholder Wealth 003182 00714 00272 00715 00613 0126 00614 01278
(126) (298) (107) (299) (275) (488) (276) (494)
Lag1 Institutional Concentration -1608633 3248998 -3424892 3334186
(151) (283) (1172) (118)
Lag1 Total Institutional Ownership 104013 5473128 1531013 -2926764
(1645) (831) (876) (158)
Lag1 Top 5 Institutional Ownership 188585 3927986 -2159177 -2640879
(016) (343) (067) (092)
Lag1 Tobinrsquos Q -445692 -803612 -454859 -833566 3257027 2971201 3213936 3086205
(427) (791) (432) (831) (83) (77) (815) (793)
Lag1 Market Capitalization 49E-5 615E-5 01031 01088
(113) (1365) (1555) (168)
Lag1 Log Market Capitalization 4683497 4759377 1645156 1427856
(2775) (3469) (3679) (4239)
CEO Dummy 1501307 1508886 1493688 1508736 4713079 4713629 47278 4703908
(2648) (2684) (2618) (2683) (282) (2828) (2819) (2824)
Adjusted R-Sq 01861 02060 01695 02051 01071 01113 01033 011
Number of Observations 33229 33229 33229 33229 97679 97679 97679 97679
32
Table VIII Two Stage Least Squares (Reverse Causality) Tests - Executive Compensation Measures as a Function of Institutional Concentration
This table shows the coefficients from the Two-Stage Least Squares (2SLS) regressions of executive compensation measures against the lagged change
and contemporary change in shareholder wealth instrumented institutional concentration and total institutional ownership Tobinrsquos Q the natural
logarithm of firm market capitalization and dummy variable controls for CEO and year effects We instrument measures of institutional concentration
and total institutional ownership by orthogonalizing with respect to option grant pay-for-performance sensitivity salary and total compensation
Coefficients for industry and year controls are not presented First stage results are not presented One two and three asterisks denote significance at
10 5 and 1 levels
ExecuComp Sample 1992‒1997 Full Sample from ExecuComp 1992‒2002
PPSOG Salary TC PPSOG Salary TC
Change in Shareholder Wealth 144E-05 0004 02133 -409E-06 00011 00157
(308) (775) (596) (691) (638) (068)
Lag1 Change in Shareholder Wealth -240E-05 -00004 00866 118E-06 00002 01422
(131) (087) (346) (097) (116) (51)
Lag1 Institutional Concentration -00898 668681 3263143 02024 916656 2856417
(052) (801) (176) (163) (1545) (1004)
Lag1 Total Institutional Ownership 01021 -43997 2512587 02983 -104171 -1028347
(09) (084) (153) (4) (269) (487)
Lag1 Tobinrsquos Q 00754 -225606 119297 00655 -158657 2997205
(596) (1649) (053) (834) (2535) (727)
Lag1 Log Market Capitalization -04234 828713 7864046 -04002 825106 1685638
(2195) (8132) (2977) (4106) (11133) (373)
CEO Dummy 15848 3214117 2950274 19149 3205862 4529101
(2223) (9389) (2255) (3802) (1331) (2758)
Overall R-Sq -- 04981 01206 -- 04738 01178
Number of Observations 41305 41305 41305 85518 85518 85518
33
Table IX Two Stage Least Squares (Reverse Causality) Tests - Executive Compensation Measures as a Function of Top 5 Ownership
This table shows the coefficients from 2SLS regressions of executive compensation measures against the lagged change and contemporary change
in shareholder wealth instrumented Top 5 institutional ownership Tobinrsquos Q natural logarithm of firm market capitalization and dummy variable
controls for CEO and year effects We instrument Top 5 institutional ownership by orthogonalizing with respect to option grant pay-for-
performance sensitivity salary and total compensation Coefficients for industry and year controls are not presented First stage results are not
presented One two and three asterisks denote significance at 10 5 and 1 levels
ExecuComp sample 1992‒1997 Full Sample 1992‒2002
PPSOG Salary TC PPSOG Salary TC
Change in Shareholder Wealth 136E-05 00041 02128 -407E-06 00011 00146
(293) (79) (596) (692) (619) (063)
Lag1 Change in Shareholder Wealth -241E-05 -00003 00868 116E-06 00003 01439
(133) (07) (347) (096) (142) (514)
Lag1 Top 5 Institutional Ownership -03774 -187512 -3848531 00179 -104261 -126659
(163) (186) (166) (011) (149) (436)
Lag1 Tobinrsquos Q 00731 -223886 10074 00642 -155467 3115073
(573) (1631) (045) (819) (2509) (75)
Lag1 Log Market Capitalization -0416 775637 7696302 -04015 764632 1462536
(2345) (10084) (3766) (3688) (13708) (4072)
CEO Dummy 15854 3211991 2950158 19152 3203152 4519385
(2223) (9381) (2254) (3804) (133) (2752)
Overall R-Sq -- 04967 01202 -- 04722 01168
Number of Observations 41305 41305 41305 85518 85518 85518
34
Table X Specification Tests utilizing Box-Cox Transformation of Market Capitalization
This table presents results from a Box-Cox transformation of market capitalization In
each model we specify transformation of market capitalization only Panel A shows
results for the period 1992‒1997 and Panel B shows results for the entire period
1992‒2002 The first part of each panel shows the Lambda coefficient that applies to
the transformed market capitalization The second part of each panel shows log
likelihood scores with LR Chi2 Statistics in parentheses ‒ these test for the
appropriateness of a linear or log model specification
Panel A Our ExecuComp sample 1992‒1997
(1) (2) (3)
PPSOG Salary TC
Lambda -05143 01881 02876
(-1914) (2413) (1329)
Test H0 (Log Likelihood Score)
Lambda = -1 -1279748 -3178575 -4795326
(23866) (85952) (99688)
Lambda = 0 -1280152 -3138456 -4791159
(31934) (57147) (16339)
Lambda = 1 -1283157 -3163807 -4793506
(92039) (564156) (63292)
Panel B Full ExecuComp sample 1992‒2002
(1) (2) (4)
PPSOG Salary TC
Lambda -044 01853 04036
(-3015) (40) (3666)
Test H0 (Log Likelihood Score)
Lambda = 0 -2688334 -6476494 -10384490
(66599) (160356) (126236)
Lambda = 1 -2694886 -6538506 -10386795
(197644) (1400605) (172345)
35
Figure 2 Institutional Ownership Measures (LHA) vs Total
Compensation (RHA) by Market Capitalization
0
10
20
30
40
50
60
70
1 2 3 4 5 6 7 8 9 10
$00
$10
$20
$30
$40
$50
$60
$70
$80
$90
Millio
ns
InstitutionalConcentration (LHS)Top5 Ownership (LHS)
Institutional TotalOwnership (LHS)Total Compensation(RHS)
Page 4
4
logarithmic transformation While this skewed nature of the level of market capitalization
does not per se rule out its use specification tests such as Box-Cox and the replacement
of market capitalization by its cumulative distribution (see Section VI below) indicates a
possible specification error Moreover using the lagged natural logarithm of market
capitalization as a firm size control in the actual HS data with the same dependent
variable and no other changes raises the explanatory power (R-Squared) of the HS model
by between 135 and 1516 percent4
Given that firm size is an important determinant of executive compensation selecting
the correct functional form for firm size is a vexing and important issue Many estimates
including those of Murphy (1985) Rosen (1992) Gibbons and Murphy (1992) Huson
Parrino and Starks (2001) Baker and Hall (2004) Gabaix and Landier (2008) Edmans
Gabaix and Landier (2009) Babenko (2009) and Aggarwal Erel and Ferreira (2011)
use a logarithmic specification with the elasticity between firm size and executive total
pay in the range of 03 and 04 HS seems to employ inadvertently a firm size measure in
levels (also alternative firm size measures also in levels) and consequently imply an
elasticity value (exponent on market capitalization) of unity
Second we question the use of the HS Top 5 Concentration measure of institutional
influence based on the influence of concentrated institutional investors relative to their
peers that appears to lead inadvertently to an exacerbation of the size-control problem In
their companion theoretical model AHS arrive at a different and what might seem a more
natural specification namely Top 5 Ownership which measures concentration relative to
all shareholders not just institutional peers The following example illustrates the
4 The lower result is based on OLS regressions of model 1 of Table II of HS with the R-Squared rising from
00853 to 00968 once Market Capitalization is logged and the higher result model 1 of Table V of HS is
based on an improvement in R-Squared from 04604 to 05302 (see Table VI below)
5
differences between these two methods If there are one hundred shares outstanding in a
small company and the only institution owns one share that company is concentrated
according even though that institution may lack effective monitoring power Instead if
one deflates by shares on issue this results in a concentration ratio of 1100 If on the
other hand ten institutions collectively own 500 shares in a much larger company of
1000 shares and the Top Five collectively own 250 then the HS concentration ratio falls
to only 250500 or 05 and utilizing shares on issue 025 in this case As is confirmed by
the data one expects Top 5 Concentration to correlate negatively with firm size and Top
5 Ownership the reverse given a positive association between institutional ownership and
company size We then re-examine the monitoring hypothesis using this AHS proposed
measure of institutional investor influence that intuitively relates better to institutional
proxy voting power
Third we address issues of reliance on potentially biased self-reported values of option
grants that result in the exclusion of 16 percent of the HS sample size for tests on total
(direct) compensation Forth we argue that tests that HS perform using an alternative
measure of pay-for-performance sensitivity to provide support for the main hypothesis of
the study require one to get the timing right in terms of the lag structure of performance
and concentration Finally we address issues associated with their tests of endogeneity
and reverse causality These issues collectively might suggest overlooking of a
monitoring effect with the main driving forces size related
The next section reviews the HS study in an effort to tease out what might really be
going on Section II examines the literature for studies that in replicating HS provide
independent support for their findings It also reports findings on the actual proxy voting
record of mutual funds and actions of institutional investors in reducing pay Section III
describes the data and methodology Section IV analyzes the relation between
6
institutional influence and measures of pay-for-performance sensitivity Section V
documents the relationship between institutional influence and total and fixed
compensation Section VI examines the robustness of our findings and the final Section
concludes
I Review of Hartzell and Starks Methodology
We comment on five areas of the HS methodology 1) Size control 2) Measurement of
institutional influence 3) Testing the Jensen-Murphy measure of incentives 4) Reliance
on incomplete self-reported valued for option grants and 5) Endogeneity
A Size Control
In our paper we argue that HS unintentionally identify incentive-size and pay-size
effects5 not a monitoring effect It can be seen in Figure II that executive total pay for
HSrsquos Top 5 institutional concentration measure rises from as little as $072m for the
smallest market capitalization decile to $85m for the highest This twelve-fold pay-size
effect in the deciles can cause problems if one utilizes an untransformed size measure
Due to skewness in the size measure using this in its untransformed state appears to
weaken its effectiveness leaving open the possibility that correlations they identify
between incentives and pay levels on the one hand and Top 5 Concentration on the other
are due largely to size-related effects
Table III displays this strong pay-size effect based on our full 1992‒2002 dataset and
graphically in Figure 1 It shows that the incentive measure adopted by HS declines from
$272 per $1000 change in shareholder wealth for the smallest market capitalization
decile of firms to approximately $030 for the largest Table III and Figure 1 show the
5 The multiplicative model of Edmans Gabaix and Landier (2009) finds that incentive pay increases with
firm size (elasticity value of 13) and the empirical elasticity is 037
7
strong relation between institutional concentration as measured by HS and market
capitalization For the smallest decile of market capitalization firms institutional
concentration averages 61 percent declining to about 275 percent for the highest decile
This negative relationship between firm size and institutional concentration seems to arise
because institutional investors do generally favor small firms This causes problems
because in the limit a small company may have only one institutional investor owning
only a small percentage of the company Hence it would appear to be fully concentrated
even though the ability of the institutional investor to influence pay decisions might be
small Thus it is not at all surprising that the correlation between the logarithm of market
capitalization and their institutional concentration measure is both negative and high in
absolute value at 63 percent Their methodology appears to favor small companies that
Table III shows to have low pay levels and high pay for performance sensitivity
Moreover if one does not transform the raw size measure it implicitly assumes that a
given dollar change in firm size has the same absolute impact on pay for a very large
company as it does for a small company Indeed the conventional logarithmic
specification for firm size provides proportionate effects that imply a pay-size elasticity of
approximately one-third instead of unity6
B Measurement of Institutional Influence
We argue that there are additional issues related to the HS measure of institutional
influence ndash institutional concentration In their companion article to HS AHS develop a
model where they capture institutional influence by a single large long-term institutional
6 Note that unlike an earlier version of our paper we leave the HS dependent variables in levels and hence
untransformed While we do this to minimize the alterations that we make to the HS specification ideally
the elasticity specification is preferable especially for the compensation variables Transforming the
dependent variables does not significantly affect the results
8
investor They measure influence relative to total shareholding including private or
atomistic shareholders Presumably they do this to capture relative activism and proxy
voting strength However when it comes to the empirical implementation of the model
AHS adopt the same relative institutional concentration measure as in HS (AHS (p12))
potentially downplaying the relative strength of institutions compared with other
shareholders They justify this because there is generally more than one institutional
investor and there will be free-riding problems with multiple investors that only the
largest institutional investors can overcome While we accept their premise of free-rider
problems it is not obvious why the deflator is lsquoinstitutional shareholdingrsquo rather than
lsquototal shareholdingrsquo from this perspective7 If the strength of the size control can be
queried scaling by total shareholding (ie Top 5 Ownership) produces a measure that
remains relatively unaffected across the size deciles It is 20 percent for the smaller
decile declining to 14 percent for the largest decile whereas for the HS concentration
measure it declines by over 50 percent from 61 percent to 28 percent (see Table III and
Figure 1)
C Testing the Jensen-Murphy Measure of Incentives
The testing of their secondary measure of executive pay-for-performance sensitivity
(see HS Table IV) based on Jensen and Murphy (1990) raises issues other than the size
control HS regress changes in compensation (cash and total) on lagged institutional
influence that has been interacted with the contemporaneous change in shareholder
wealth to capture a pay-for-performance relationship Such a test is appealing ndash
7 AHS do point out one benefit of the institutional concentration measure Institutional investors may prefer
companies with lsquobetterrsquo incentive compensation structures Could using institutional shareholding rather
than total shareholding as the deflator alleviate this problem Presumably both the numerator and
denominator will have the same bias and hence error cancellation is a possibility
9
institutions should increase compensation when the firm has performed well Following
Greene (2000 p326) to execute such a test one needs to include in level form variables
that are then interacted Hence the HS model specification ought to be
1 2 3 1 4 1it it it it it itCompn SW IC SW IC (1)
where itCompn is the change in executive compensation itSW is the contemporaneous
change in shareholder wealth which also appears in the interaction term and 1itIC is
lagged institutional ownership concentration However HS (p2363) omit the
contemporaneous term itSW and the lagged term 1itIC which should appear in levels
as well as in the interaction term Most importantly HS interact their lagged institutional
concentration measure 1itIC with the contemporaneous change in shareholder wealth
itSW As it is after all pay-for-performance concentrated institutions need to condition
their influence on some observed measure of firm performance Contemporaneous firm
performance is unknown at the time institutional influence is measured Reflecting this
timing insight leads to a new Greene (2000) specification
1 2 1 3 1 4 1 1it it it it it itCompn SW IC SW IC (2)
such that all the lagged RHS independent variables that now appear in the interaction
term are included on a stand-alone basis
Reliance on incomplete self-reported values for option grants
Prior to receiving the HS dataset we tried to replicate the findings of HS to validate
our datasets for the HS period and the longer period 1992‒2002 While we initially
focused on the same period as HS (ie 1992‒1997) we found that our sample size was
larger at 47765 observations than HSrsquos at 33928 observations This discrepancy could
have arisen if HS did not replace missing and self-reported option valuations using their
10
Black-Scholes option methodology This may introduce a downward pay bias since self-
reported option-grant pay is subject to possible understatement and there are 13839
observations with likely high option pay dropped altogether
Endogeneity
HSrsquos endogeneity testing raises concerns about reverse causality Institutions may
prefer to hold firms with certain compensation structures such as incentivized executives
HS rightly highlight this concern and use instrumental variables to address this issue HS
assume that institutional ownership reflects any reverse causality Using an instrumental
variables approach to resolve this problem involves finding a variable that influences or
explains institutional ownership but not directly executive compensation They use share
turnover in a firm as an instrument for institutional ownership However theoretical work
by Holmstrom and Tirole (1993) indicates that measures of stock liquidity and trading
activity are indeed important determinants of executive market-based incentives as Kang
and Lui (2008) and Dikolli Kulp and Sedatole (2009) find empirically As such one can
make a case for adopting another instrument rather than share turnover for institutional
ownership
II Previous Research
Prior research fails to address issues concerning possible fragility of the HS results
AHS replicate the HS findings on both incentives and pay levels using the same dataset as
HS and they apply the same definition of institutional concentration and use market
capitalization expressed as a level Cadman Klasa and Matsunaga (2010 Table 6) set out
to replicate HSrsquos (Table II) model of option grant pay-for-performance sensitivity when
they compare a dataset of ExecuComp and Non-ExecuComp firms for the period 2000-
2007 Their ExecuComp results for this sample period are quite similar to our results for
11
the 1992‒2002 period for the same specification controlling for the logarithm of market
capitalization but they do not regard these results as a test of HS as their focus is entirely
on comparing the two groups of stocks Nor do they investigate the impact of
concentration on either cash compensation or total pay Kang and Liu (2008) examine
pay-performance sensitivity use a version of the HS institutional concentration measure
as a control variable in a different framework deriving from Holmstrom and Tirole
(1993) Dikolli Kulp and Sedatole (2009) argue that firms with transient ownership will
issue relatively more equity incentives to counter short-termism with their results robust
to the inclusion of the HS concentration variable
We hoped to find supportive evidence from proxy voting behavior in support of HSs
findings However evidence from actual mutual fund proxy voting records suggests that
most oppose reductions in CEO pay but overall the evidence is mixed8 Davis and Kim
(2005) examine proxy votes of mutual funds for 2004 and find that for a sample of
Fortune 1000 proxy contests in 45 cases that attempted to limit executive pay mutual
funds were unanimously opposed to such attempts Rothberg and Lilien (2006) also
examine proxy-voting data and find that mutual funds voted 66 percent of the time in
managementrsquos favor on issues of executive compensation Ertimur Ferri and Muslu
(2011) find that institutional investors lack sophistication in that they do not target
excessively paid CEOs relative to CEOs with high predicted-pay based on economic
determinants but exert a moderating influence ndash a $23 million reduction ndash for those they
deem ldquoexcessively paidrdquo CEOs
III Data and Methodology
8 We do acknowledge that the HS results are for all institutional investors while the transparency with
respect to proxy voting applies only to mutual funds
12
This Section discusses the data used in the study the construction of variables
descriptive statistics and the methodology Two datasets are used ndash one is the original
dataset generously supplied by Jay Hartzell and Laura Starks which covers 1992‒1997
second a more comprehensive dataset covering 1992‒2002 is computed from the same
data sources used in the HS dataset This is Standard and Poorrsquos (SampP) ExecuComp for
compensation data Thomson Financial CDASpectrum database for institutional holdings
to compute measure of institutional influence SampPrsquos Compustat and the Center for
Research in Stock Prices (CRSP) for firm level accounting and stock price data Our more
comprehensive dataset contains 2559 firms compared with 1914 for the HS dataset Our
full matched sample yields 97679 executive years
A Institutional Influence Measures
To measure institutional influence we employ two ownership-based metrics The first
measure of institutional influence is HSrsquos Institutional Concentration We use this for
direct comparison to the HS analysis The second measure of institutional influence
changes the denominator of the HS measure from institutional shares held to total shares
outstanding in the firm with nomenclature Top 5 Ownership9 We compute both
measures at the fiscal year end of the firm We adjust all compensation measures and
asset prices for inflation to 30 June 2002 prices for our comprehensive sample compared
to the HS dataset that is in the dollars of the day
B Compensation Measures
The ExecuComp database reports several raw measures of executive compensation for
the top five executives (where available) including total salary bonus long-term
9 Top 5 Ownership is included in the HS database but not utilized by them in reported tests as well as
constructed by us for the more comprehensive sample
13
incentives and number of option grants For our more comprehensive dataset we replicate
all the compensation measures used by HS In common with HS we compute the pay-for-
performance sensitivity of option grants using the Yermack (1995) approach This
measure shows the dollar change in value of the executive options for every dollar change
in the value of the firm This involves calculating the option delta using the Black-
Scholes formula adjusted for dividends10
The ratio of the number of shares represented
by the award of options divided by the diluted number of shares outstanding at the
beginning of the year multiplies the option delta with units given by a $1000 change in
firm value Summing salary bonus long-term incentive plan payouts and stock and
option grants provides total direct compensation Options are valued using the Black and
Scholes (1973) formula for European call options adjusted by Merton (1973) to
incorporate dividends
C Descriptive Statistics
Table I presents descriptive statistics of key variables employed in the study according
to our comprehensive dataset11
Institutions own on average 546 percent of the firms in
which they invest This is almost identical to the 531 percent found by HS The
institutions with the five largest holdings account 45 percent of institutional ownership
Once again the figure for HS is almost identical at 44 percent The institutions with the
five largest holdings own on average 229 percent of shares outstanding in a firm (HS
have 223 percent) Average total compensation for executives is $2583 million
considerably larger than that documented by HS ($1250 million) but the HS figure is an
10 When there is more than one grant during the year a weighted average delta is used
11 With the data supplied to us by Hartzell and Starks we are able to replicate exactly their summary data
(HS Table I) for all but cash compensation Salary is included but bonus payments are missing The
replicated table is available on request
14
amalgam of price levels over their sample period 1992‒1997 and ours are in consistent
dollars of June 30 2002 However there has also been significant growth in
compensation levels from 1998 to 2002
ltltINSERT TABLE Igtgt
Based on our comprehensive data in Table II we examine correlations between the
key variables employed in the study Of concern is the large negative correlation of 63
percent which has been discussed previously (Section IIA) between the HS measure of
institutional concentration and the natural logarithm of firm market capitalization
ltltINSERT TABLE IIgtgt
In view of this high correlation Table III provides a breakdown of the three key pay
variables and three institutional ownership variables by size deciles We find that the high
institutional concentration but low institutional ownership share of total shareholdings
mentioned above is evident in the smallest decile of firms by size This decile has low
salary and even lower total compensation in a relative sense but by far the highest pay-
for-performance sensitivity of option grants Small firms award executives a far higher
proportion of market capitalization Since small firm market-capitalization is negligible in
comparison with the highest decile (300 times larger in magnitude) total compensation is
very small in absolute value Moreover due to the domination of the pay-size effect base
salaries are also exceedingly small
ltltINSERT TABLE IIIgtgt
Figure 1 based on Table III shows how as institutional concentration falls with
increases in stock size (but also as the total share of institutional ownership increases)
and the pay-for-performance sensitivity of option grants falls Thus because of what
appears to be a weak size control and size-influenced concentration measure HS
15
potentially pick up a size effect as their main result but not necessarily a monitoring
effect Figure 2 shows that as firm size increases up to decile 8 at least the share of
institutional total ownership increases as does total compensation
ltltINSERT FIGURES 1 AND 2 gtgt
D Methodology
Following HS we use two estimation techniques to examine the effect of institutional
investor influence on executive compensation Due to the censored nature of option grant
pay-for-performance sensitivity this compensation measure most appropriately uses
Tobit model specification whereas an Ordinary Least Squares (OLS) model specification
is used for all other compensation measures We model compensation measures as a
function of contemporaneous and lagged change in shareholder wealth institutional
influence Tobinrsquos Q and market capitalization Controls for CEO industry and year
effects are also used Augmentation occurs to test the effect of institutional influence on
the change in executive cash and total compensation with an interaction term that is the
product of institutional influence and the change in shareholder wealth
The variables used to model executive compensation have the following definitions
The lagged and contemporary measure of change in shareholder wealth is simply the
difference in market capitalization of the firm The measures of institutional influence
used are Institutional Concentration and Top 5 Ownership Use of the two types of
institutional influence occurs in separate models In order to replicate the HS
methodology models using Institutional Concentration have an additional control
consisting of lagged total institutional holdings and Tobinrsquos Q controls for performance
and growth opportunities Market capitalization of the firm is the number of shares
outstanding multiplied by the share price with inclusion both in levels and natural
16
logarithmic form A dummy variable is included to identify whether the executive is the
CEO Dummy variables are also included to control for industry effects This variable
takes a value of 1 for the two-digit Standard Industrial Classification (SIC) in which the
firm operates A year fixed effects (dummy) variable is included taking a value of 1 if the
observation was from the given year
IV Executive Incentives as a Function of Institutional Investor Influence
This section presents the results for the effect of institutional influence on option grant
pay-for-performance sensitivity and changes in compensation
A Option Grant Pay-for-Performance Sensitivity
Table IV presents the results for the effect of institutional influence on option grant
pay-for-performance sensitivity
ltltINSERT TABLE IVgtgt
Model (1) is a direct replication of HS using the HS database as supplied to us by HS
with all the variables computed in the same way and in the same functional form as with
HS The coefficient and t-statistic on lagged institutional concentration interacted with the
change in shareholder wealth is 13994 and 767 that compares to 1358 and 842 in HS
We are unable to account for these minor discrepancies for values that should be
identical In model (2) the only change to the model is that we use the natural logarithm
of market capitalization instead of the level of market capitalization This single change
has a sizeable effect on the coefficient on institutional concentration The coefficient on
institutional concentration is now -00692 and is insignificant The fact that an order
preserving transformation to a control variable can change the results is of interest The
log-transformation to market capitalization also changes the coefficient on Top 5
Ownership in models (3) and (4) The coefficient on Top 5 Ownership is significant in
17
both model (3) and in model (4) but the better size control reduces the coefficient by over
25 percent The better size control does not affect Top 5 Ownership by as much as its
design ensures less sensitivity to what could be inadequate size controls
We also test the effect of institutional influence on option grant pay-for-performance
sensitivity over the period 1992‒2002 This sample is then approximately three times as
large as the sample used by HS Institutional concentration is highly positively significant
in model (5) similar to the results for the period 1992‒1997 When we use the log of
market capitalization in model (6) the coefficient on institutional concentration remains
positive and significant at the 10 percent level with a t-statistic of 166 Of importance
though is that the coefficient on institutional concentration is one-eighth of the coefficient
in model (5) Even for a measure of institutional influence which we believe is less
sensitive to firm size the use of log market capitalization reduces the coefficient on Top 5
Ownership by approximately 50 percent from model (7) to model (8)
B Pay-for-Performance Sensitivity due to Changes in Cash and Total
Compensation
For robustness HS examine the effect that institutional influence conditioned on firm
performance has on a changes in cash and total executive compensation using the general
method of Jensen and Murphy (1990) to compute pay-for-performance sensitivity
ltltINSERT TABLE V gtgt
Table V presents the effect of institutional influence on the change in cash
compensation using the HS size control measure throughout12
Model (1) replicates the
specification in HS for the period of 1992‒1997 and we produce a similar result ndash the
interaction of lagged institutional concentration and contemporaneous change in
12 This test uses our data for the two periods 1992‒1997 and 1992‒2002 We do not use the HS data for
cash compensation as data for bonus is missing in the data supplied to us
18
shareholder wealth positively and significantly relates to a change in executive cash
compensation In model (2) we include the contemporaneous change in shareholder
wealth and the lagged institutional concentration as stand-alone controls as indicated by
Greenersquos (2000) specification of equation (1) above and the interaction term remains
significant In model (3) we interact lagged institutional concentration with the lagged
change in shareholder wealth as in equation (2) above that tries to take account of what
participants could have observed at the time The interaction term is then negative but
insignificant altering the HS result Model (4) provides a similar result where our
preferred proxy Top 5 Ownership captures concentrated institutional influence In
models (5) to (8) we use our full sample In model (5) we find that the lagged
institutional concentration interaction with the contemporaneous change in shareholder
wealth is positive and significant in HSrsquos specification However including change in
contemporaneous shareholder wealth and lagged institutional concentration as separate
control variables as suggested by equation (1)rsquos specification removes the significance
from the interaction term in model (6) Moreover in model (7) when lagged institutional
concentration is interaction with the lagged change in shareholder wealth in accord with
equation (2) this interaction term is now negative and significant reversing the HS
finding We obtain a similar result in model (8) when we measure institutional
concentration influence using Top 5 Ownership13
V The Levels of Executive Fixed and Total Compensation as Functions of
Institutional Investor Influence
13 In unreported results we use the log of market capitalization as a size control These changes to the size
control do not dramatically change the findings from Table V The reason why this size issue is not as
important as in the findings for option grants pay-for-performance is that the change in cash compensation
is not so strongly influenced by firm size In unreported results available on request we also examine the
effect of institutional influence on changes in total compensation with similar results to the change in cash
compensation
19
Two measures of compensation are examined ndash salary and total compensation
ltltINSERT TABLE VI gtgt
Table VI examines the effect of institutional influence on the level of salary In model
(1) we use the HS sample and specification to obtain almost identical results The
coefficient on institutional concentration is (negative) $247509 that is almost identical to
(negative) $247967 in HS Both are highly significant In model (2) we make a single
change to the model ndash we use the natural logarithm of market capitalization This single
change reverses the sign of institutional concentration with the coefficient still being
significant at the 1 percent level In models (3) and (4) we measure institutional influence
using Top 5 Ownership In model (3) where we measure firm size using the level of
market capitalization Top 5 Ownership is insignificant When we use the natural
logarithm of market capitalization in model (4) Top 5 Ownership is significant and
positively relates to the salary level In models (5) to (8) we estimate the same models
using our full sample We find similar results for the full sample as for the period 1992ndash
1997
ltltINSERT TABLE VII gtgt
Table VII examines the effect of institutional influence on total compensation In
model (1) we replicate the HS methodology using the HS data to obtain a similar
significant negative relation between institutional concentration and total compensation of
(negative) $1608633 However after changing to a logarithmic specification of market
capitalization in model (2) institutional concentration increases total compensation by a
significant $32400014
Importantly when institutional influence is measured using Top 5
14 = (5688751 times 0101)
20
Ownership an insignificant relation exists with total compensation in model (3) using the
HS size control but it becomes both positive and significant in (4) with the logged size
control In models (5) and (6) where models are estimated using the full sample we
observe a similar reversal in sign as occurs in models (1) and (2) However now the
magnitude of the reversal is almost equal with a difference of about $67m that is very
large indeed In model (5) the coefficient on institutional concentration is (negative)
$342m and then in model (6) the coefficient is $33m Top 5 Ownership is insignificant
in both models (7) and (8) unlike the results for Salary
VI Robustness
Two issues are examined in is this Section ndash reverse causality and the appropriate
transformation of market capitalization A methodological issue inherent in examining the
effect of ownership-based measures of institutional influence on executive compensation
is endogeneity driven by reverse causality We utilize an instrumental variables approach
where orthogonalized institutional influence measures are regressed on each of the
compensation measures lagged option grant pay-for-performance sensitivity salary and
total compensation In addition market capitalization and a CEO dummy are included in
each regression We use the residual from each regression as an instrument for
institutional influence Such a process removes any preferences for compensation from
the institutional influence measures
ltltINSERT TABLES VIII AND IXgtgt
Table VIII presents the results where we instrument the HS measure lagged
institutional concentration with a compensation-orthogonalized equivalent The analysis
indicates that the instrumented institutional concentration measure is positively and
significantly related to the level of salary and total compensation and insignificantly
21
related to option grant pay-for-performance sensitivity (PPSOG) Table IX presents
results where we instrument Top 5 Ownership with a compensation-orthogonalized
equivalent The instrumented Top 5 Ownership measure insignificantly relates to option
grant pay-for-performance sensitivity and barely negatively relates to salary although it
negatively relates to total compensation suggesting reverse causality is driving a
significant portion of our previous results
ltltINSERT TABLE Xgtgt
Another issue examined in this section is the appropriate transformation of market
capitalization Aneuryn-Evans and Deaton (1980) provide an evaluation of methods
comparing linear and log-linear specifications We use the Box-Cox test to compare
logarithm versus linear specifications In applying the Box-Cox test we specify
transformation of only market capitalization ndash the dependent variable and all other
independent variables remain unchanged Table X presents the lambda coefficient for
each model and the log likelihood score In all models in Panel A and B a lambda of zero
receives a higher (closer to zero) log likelihood score suggesting the log specification is
appropriate
VII Conclusion
Hartzell and Starks (2003) appears to document an effective role for institutional
investors in increasing executive incentives and lowering compensation levels Our paper
re-addresses the issue of concentrated institutional monitoring from a perspective
informed by an additional decade of research since their paper appeared We cannot find
evidence that concentrated institutional investors really have any positive (or negative for
that matter) effect on executive incentive levels While superficially we find evidence
that they may raise executive pay levels this is most likely driven by an institutional
22
preference for highly remunerated executives in large companies and thus by reverse
causality Moreover our findings are not consistent with the rent-extraction hypothesis
suggesting that firms systematically overpay executives in the absence of effective
monitoring by outside parties such as institutional investors
23
References
Almazan Andres Jay C Hartzell and Laura T Starks 2005 ldquoActive Institutional
Shareholders and Costs of Monitoring Evidence from Executive Compensationrdquo
Financial Management Winter 5ndash34
Aneuryn-Evans Gwyn and Angus Deaton 1980 ldquoTesting Linear versus Logarithmic
Regression Modelsrdquo Review of Economic Studies 47 (1) Econometrics Issue 275ndash
291
Babenko Ilona 2009 ldquoShare Repurchases and Pay-Performance Sensitivity of Employee
Compensation Contractsrdquo Journal of Finance 64117-150
Baker George P and Brian J Hall 2004 ldquoCEO Incentives and Firm Sizerdquo Journal of
Labor Market Economics 22 767ndash798
Bebchuk Lucian A Jesse M Fried and David Walker 2002 ldquoManagerial power and
rent extraction in the design of executive compensationrdquo University of Chicago Law
Review 69 751ndash846
Bebchuk Lucian A and Jesse M Fried 2003 ldquoExecutive Compensation as an Agency
Problemrdquo Journal of Economic Perspectives 17 71ndash92
Bebchuk Lucian A and Jesse M Fried 2004 Pay without Performance The Unfulfilled
Promise of Executive Compensation Cambridge MA Harvard University Press
Bertrand Marianne and Sendhil Mullainathan 2001 ldquoAre CEOs Rewarded for Luck
The Ones Without Principals Arerdquo Quarterly Journal of Economics 116 901ndash32
Black Fischer and Myron Scholes 1973 ldquoThe Pricing of Options and Corporate
Liabilitiesrdquo Journal of Political Economy 81 637ndash54
Cadman Brian Sandy Klasa and Steve Matsunaga 2010 ldquoDeterminants of CEO Pay A
Comparison of ExecuComp and Non-ExecuComp Firmsrdquo The Accounting Review 85
1511ndash1543
Chhaochharia Vidhi and Yaniv Grinstein 2009 ldquoCEO Compensation and Board
Structurerdquo Journal of Finance 64 231ndash261
Cichello Michael S 2005 ldquoThe impact of firm size on pay-performance sensitivitiesrdquo
Journal of Corporate Finance 11 609ndash627
Davis Gerald F and E Han Kim 2007 ldquoBusiness ties and proxy voting by mutual
fundsrdquo Journal of Financial Economics 85 552ndash570
Dikolli Shane S Susan L Kulp and Karen L Sedatole 2009 ldquoTransient Institutional
Ownership and CEO Contractingrdquo Accounting Review 84 737ndash770
Edmans Alex Xavier Gabaix and Augustin Landier 2009 ldquoA Multiplicative Model of
Optimal CEO Incentives in Market Equilibriumrdquo Review of Financial Studies 22
4881ndash4917
Edmans Alex 2009 ldquoBlockholder Trading Market Efficiency and Managerial Myopiardquo
Journal of Finance 64 2481ndash2513
Ertimur Yonca Fabrizio Ferri and Volkan Muslu 2011 ldquoShareholder Activism and
CEO Payrdquo Review of Financial Studies 24 535ndash592
24
Gabaix Xavier and Augustin Landier 2008 ldquoWhy Has CEO Pay Increased So Muchrdquo
Quarterly Journal of Economics 123 49ndash100
Garen John 1994 ldquoExecutive compensation and principal-agent theoryrdquo Journal of
Political Economy 102 1175ndash1199
Gibbons Robert and Kevin J Murphy 1992 ldquoOptimal Incentive Contracts in the
Presence of Career Concerns Theory and Evidencerdquo Journal of Political Economy
100 468ndash505
Greene William H 2000 (4th
Edition) Econometric Analysis Prentice-Hall NJ
Guthrie Katherine Jan Sokolowsky and Kam-Ming Wan 2012 ldquoCEO Compensation
and Board Structure Revisitedrdquo Journal of Finance Forthcoming
Hadlock Charles R and Gerald B Lumer 1997 ldquoCompensation turnover and top
management incentives historical evidencerdquo Journal of Business 70 153ndash187
Hartzell Jay C and Laura T Starks 2003 ldquoInstitutional Investors and Executive
Compensationrdquo Journal of Finance 58 2351ndash2374
Huson Mark R Robert Parrino and Laura T Starks 2001 ldquoInternal Monitoring
Mechanisms and CEO Turnover A Long-Term Perspectiverdquo Journal of Finance 56
2265ndash2297
Jensen Michael C and Kevin J Murphy 1990 ldquoPerformance Pay and Top Management
Incentivesrdquo Journal of Political Economy 98 225ndash264
Jin Li 2002 ldquoCEO compensation diversification and incentivesrdquo Journal of Financial
Economics 66 29ndash63
Kang Qiang and Qiao Liu 2008 ldquoStock Market Information Production and Executive
Incentivesrdquo Journal of Corporate Finance 14 484ndash498
Merton Robert 1973 ldquoTheory of Rational Option Pricingrdquo Bell Journal of Economics
and Management Science 4 141-183
Morse Adair Vikram Nanda and Amit Seru 2011 ldquoAre Incentive Contracts Rigged by
Powerful CEOsrdquo Journal of Finance 66 1779ndash1821
Murphy Kevin J 1985 ldquoCorporate performance and managerial remuneration An
empirical analysisrdquo Journal of Accounting and Economics 7 11ndash42
Murphy Kevin J 1999 ldquoExecutive compensationrdquo in Orley Ashenfelter and David
Card Handbook of Labor Economics vol 3 Elsevier
Rosen Sherwin 1992 ldquoContracts and the market for executivesrdquo In Contract
Economics ed Lars Werin and Hans Wijkander 181ndash211 Cambridge MA
Blackwell
Rothberg Burton and Steven Lilien 2006 ldquoMutual Funds and Proxy Voting New
Evidence on Corporate Governancerdquo Journal of Business and Technology Law 1
157ndash184
Schaefer Scott 1998 ldquoThe dependence of payndashperformance sensitivity on the size of the
firmrdquo Review of Economics and Statistics 80 436ndash443
Yermack David 1995 ldquoDo Corporations Award CEO Stock Options Effectivelyrdquo
Journal of Financial Economics 39 237ndash69
25
Table I Univariate Descriptive Statistics
This table reports sample statistics for the key variables used in this study based on our comprehensive
ExecuComp dataset Institutional Concentration is the shares owned by the institutions with the five largest
holdings divided by the shares held by all institutions Top 5 Ownership is the proportion of shares
outstanding held by the institutions with the five largest holdings Total Institutional Ownership is the
proportion of shares held by all institutions Panel B shows the measures executive compensation The pay-
for-performance sensitivity (PPS) of option grants shows the change in executive wealth due to a $1000
change in firm in shareholder wealth Expressions for Salary and Total Compensation are in thousands of
dollars while Market Capitalization is in millions of dollars Panel C shows firm characteristics
Variable Mean Median Std Dev P10 P90
Panel A Institutional Investor Holdings
Institutional Concentration 0450 0424 0151 0282 0656
Top 5 Institutional Ownership 0229 0220 0101 0110 0355
Total Institutional Ownership 0546 0561 0206 0258 0801
Panel B Executive Compensation
PPS of Option Grants 1161 0156 3829 0 2822
Salary 36311 29462 25231 15179 67000
Total Compensation 258313 85878 980701 23875 511022
Panel C Firm Characteristics
Tobins Q 1744 1139 2738 0495 3322
Market Capitalization 560699 112094 1988108 19754 1048754
26
Table II Pairwise Correlation Coefficients
This table presents the correlations between the key variables based on our comprehensive
ExecuComp dataset Institutional Concentration is the shares owned by the institutions with the five
largest holdings divided by the shares held by all institutions Top 5 Ownership is the proportion of
shares outstanding held by the institutions with the five largest holdings Total Institutional
Ownership is the proportion of shares held by all institutions The pay-for-performance sensitivity
(PPS) of option grants is the change in executive wealth due to a $1000 change in shareholder
wealth
(1) (2) (3) (4) (5) (6) (7) (8)
(1) Institutional Concentration 1
(2) Top 5 Ownership 0268 1
(3) Total Ownership -0415 0706 1
(4) PPS of Option Grants 0122 0042 -0024 1
(5) Log Salary -0270 -0012 0171 0038 1
(6) Log Market Capitalization -0630 -0157 0261 -0156 0420 1
(7) Level Market Capitalization -0241 -0157 0007 -0058 0219 0534 1
(8) Tobins Q -0008 -0034 0002 0015 -0083 0066 0121 1
27
Table III
Descriptive Statistics by Market Capitalization Deciles
Market
Capitalization
Deciles
Market
Capitalization
($m)
Total
Compensation
($000)
Salary
($)
PPS Option
Grants
($)
Institutional
Concentration
()
Top 5 Ownership
()
Institutional
Total Ownership
()
1 (Small) 1181 76008 234700 2720 611 200 327
2 2668 90206 257955 1796 495 217 438
3 4193 105946 275120 1517 454 220 485
4 6045 129892 296067 1376 420 206 491
5 8742 155391 315864 1216 394 198 502
6 13086 191888 345942 1072 380 202 531
7 20790 239501 377021 0874 368 196 532
8 34920 314955 427808 0687 354 190 538
9 67747 450970 480319 0569 328 173 528
10 (Big) 366540 846066 614354 0297 275 138 502
28
Table IV Pay-for-Performance Sensitivity of Option Grants as a Function of Institutional Influence
Similar in construction to parts of HS (Table II) this table shows the coefficients from Tobit regressions of option grant pay-for-performance sensitivity per $1000 change in
shareholder wealth against the lagged change and contemporary change in shareholder wealth measures of institutional influence Tobinrsquos Q firm market capitalization and
dummy controls for CEO and year effects We estimate models (1) to (4) over the period 1992‒1997 using data supplied by HS and models (5) to (8) are estimated over the
whole sample Models (1) (2) (5) and (6) use Institutional Concentration as a measure of institutional influence which HS define as the shares owned by the institutions
with the five largest holdings divided by the shares held by all institutions These models control for total institutional ownership Models (3) (4) (7) and (8) use Top 5
Ownership as a measure of institutional influence which we define as the proportion of shares outstanding held by the institutions with the five largest holdings Coefficients
for industry and year controls are not presented One two and three asterisks denote significance at 10 5 and 1 levels
Hartzell and Starks Database 1992‒1997 Full Sample from ExecuComp 1992‒2002
(1) (2) (3) (4) (5) (6) (7) (8)
Change in Shareholder Wealth 77721 191781 89874 137368 -446 -218 -664 -203
(171) (429) (194) (317) (319) (168) (453) (158)
Lag1 Change in Shareholder Wealth 14446 99963 88719 8544 -144 0273 -0342 00852
(027) (212) (166) (184) (071) (015) (016) (005)
Lag1 Institutional Concentration 13994 -00692 26405 03304
(767) (032) (1402) (166)
Lag1 Total Institutional Ownership 08055 10434 07106 0937
(593) (781) (666) (898)
Lag1 Top 5 Institutional Ownership 10567 07457 12116 06293
(399) ( 283) (533) (278)
Lag1 Tobinrsquos Q 0068 00877 00647 00813 00498 00727 00508 00695
(337) (434) (319) (401) (504) (734) (51) (703)
Lag1 Market Capitalization -106202 -173515 -776 -117
(474) (742) (1011) (153)
Lag1 Log Market Capitalization -0297 -02319 -04111 -03908
(1576) (1439) (2997) (2841)
CEO Dummy 13745 13682 13765 13674 28197 28128 28155 28143
(1913) (1918) (1914) (1916) (3388) (3396) (3377) (34)
Number of Observations 33878 33878 33878 33878 97679 97679 97679 97679
29
Table V Pay-for-Performance Sensitivity of Cash Compensation (Salary plus Bonus) as a Function of Institutional Influence - Level Market Capitalization
Similar in parts to models (1) and (2) of HS (Table IV) this table displays the coefficients from OLS regressions of the change in cash compensation against the lagged
change and contemporary change in shareholder wealth measures of institutional influence Tobinrsquos Q firm market capitalization and dummy controls for CEO and year
effects We estimate models (1) to (4) over the period 1992‒1997 based on our ExecuComp sample and models (5) to (8) over the whole sample Models (1) (2) (5) and (6)
use Institutional Concentration as a measure of institutional influence which HS define as the shares owned by the institutions with the five largest holdings divided by the
shares held by all institutions These models control for total institutional ownership Models (4) and (8) use Top 5 Ownership as a measure of institutional influence which
we define as the proportion of shares outstanding held by the institutions with the 5 largest holdings Coefficients for industry and year controls are not presented One two
and three asterisks denote significance at 10 5 and 1 levels
Our Database based on ExecuComp 1992‒1997 Full Sample based on ExecuComp 1992‒2002
(1) (2) (3) (4) (5) (6) (7) (8)
Lag1 Change in Shareholder Wealth 00014 00017 00088 00089 -00006 -00007 00072 00086
(051) (062) (129) (103) (027) (029) (131) (16)
Change in Shareholder Wealth -00102 00102 00105 00032 00102 00101
(097) (292) (293) (055) (611) (608)
Change Shareholder Wealth Lag1 Institutional
Concentration
00435 00764 00363 00253
(463) (258) (689) (144)
Lag1 Change Shareholder Wealth Lag1
Institutional Concentration
-00197 -00283
(122) (197)
Lag1 Institutional Concentration -492182 -288861 400149 453623
(09) (056) (098) (114)
Lag1 Total Institutional Ownership 720688 538372 655091 559587 682069 726226
(276) (232) (281) (314) (308) (322)
Lag1 Change Shareholder Wealth Lag1 Top 5
Institutional Ownership
-0046 -00689
(091) (256)
Lag1 Top 5 Institutional Ownership 583044 1191634
(092) (25)
Lag1 Tobinrsquos Q -32761 -36806 -19422 -18023 -54283 -48681 -29534 -24243
(138) (172) (077) (078) (227) (245) (136) (111)
Lag1 Market Capitalization 00004 00004 00002 00006 00007 00008 00007 00007
(042) (044) (022) (07) (086) (092) (082) (09)
CEO Dummy 768035 765474 768334 773634 616315 61729 619028 618535
(354) (355) (356) (355) (417) (418) (419) (418)
Adjusted R-Sq 00057 0006 00052 0005 00097 00097 00098 00102
Number of Observations 41305 41305 41305 41305 85518 85518 85518 85518
30
Table VI Level of Salary as a Function of Institutional Influence
Similar in parts to models (1) and (2) of HS (Table V) this table displays the coefficients from OLS regressions of salary against the lagged change and contemporary change in
shareholder wealth measures of institutional influence Tobinrsquos Q firm market capitalization and dummy controls for CEO industry and year effects We estimate models (1) to
(4) over the period 1992‒1997 using the actual data supplied by HS and models (5) to (8) over the whole sample 1992‒2002 Models (1) (2) (5) and (6) use Institutional
Concentration as a measure of institutional influence which HS define as the shares owned by the institutions with the five largest holdings divided by the shares held by all
institutions These models control for total institutional ownership Models (3) (4) (7) and (8) use Top 5 Ownership as a measure of institutional influence which we define as
the proportion of shares outstanding held by the institutions with the 5 largest holdings Coefficients for industry and year controls are not presented One two and three
asterisks denote significance at 10 5 and 1 levels
Hartzell and Starks Database 1992‒1997 Full Sample based on ExecuComp 1992‒2002
(1) (2) (3) (4) (5) (6) (7) (8)
Change in Shareholder Wealth 04125E-4 0003 -2439E-4 0003 00021 00012 00024 00012
(057) (464) (031) (468) (1004) (72) (1041) (706)
Lag1 Change in Shareholder Wealth -00062 7856E-4 -0007 9003E-4 -00014 00002 -00014 00003
(594) (098) (592) (112) (522) (123) (5) (145)
Lag1 Institutional Concentration -2475087 555134 -2734983 943317
(3245) (710) (4965) (1671)
Lag1 Total Institutional Ownership 1020109 217488 936337 252464
(1904) (432) (2471) (704)
Lag1 Top 5 Institutional Ownership -148355 31354 -87926 423166
(139) (332) (118) (643)
Lag1 Tobinrsquos Q -195348 -252191 -192344 -25158 -119837 -146373 -122132 -143821
(2812) (3762) (2721) (3752) (248) (274) (2583) (2727)
Lag1 Market Capitalization 844E-06 103E-5 00034 00039
(2781) (3051) (2874) (3104)
Lag1 Log Market Capitalization 750021 723625 776657 732279
(7511) (9285) (11224) (14039)
CEO Dummy 2683978 2702238 2668244 2702417 3281133 3285103 3292175 3283145
(8625) (9277) (8246) (9271) (12739) (13616) (12336) (13605)
Adjusted R-Sq 04604 05302 04187 05295 03856 04588 03476 04574
Number of Observations 36280 36280 36280 36280 97679 97679 97679 97679
31
Table VII Level of Total Direct Compensation as a Function of Institutional Influence
Similar in parts to models (3) and (5) of HS (Table V) this table displays the coefficients from OLS regressions of total direct compensation against the lagged change and
contemporary change in shareholder wealth measures of institutional influence Tobinrsquos Q firm market capitalization and dummy controls for CEO industry and year effects
We estimate models (1) to (4) over the period 1992‒1997 and models (5) to (8) over the whole sample Models (1) (2) (5) and (6) use Institutional Concentration as a measure
of institutional influence which HS define as the shares owned by the institutions with the five largest holdings divided by the shares held by all institutions These models
control for total institutional ownership Models (3) (4) (7) and (8) use Top 5 Ownership as a measure of institutional influence which we define as the proportion of shares
outstanding held by the institutions with the five largest holdings Coefficients for industry and year controls are not presented One two and three asterisks denote significance
at 10 5 and 1 levels
Hartzell and Starks Database 1992‒1997 Full Sample from ExecuComp 1992‒2002
(1) (2) (3) (4) (5) (6) (7) (8)
Change in Shareholder Wealth 005779 007006 005282 006791 00423 00148 0046 00138
(415) (581) (369) (565) (203) (07) (22) (066)
Lag1 Change in Shareholder Wealth 003182 00714 00272 00715 00613 0126 00614 01278
(126) (298) (107) (299) (275) (488) (276) (494)
Lag1 Institutional Concentration -1608633 3248998 -3424892 3334186
(151) (283) (1172) (118)
Lag1 Total Institutional Ownership 104013 5473128 1531013 -2926764
(1645) (831) (876) (158)
Lag1 Top 5 Institutional Ownership 188585 3927986 -2159177 -2640879
(016) (343) (067) (092)
Lag1 Tobinrsquos Q -445692 -803612 -454859 -833566 3257027 2971201 3213936 3086205
(427) (791) (432) (831) (83) (77) (815) (793)
Lag1 Market Capitalization 49E-5 615E-5 01031 01088
(113) (1365) (1555) (168)
Lag1 Log Market Capitalization 4683497 4759377 1645156 1427856
(2775) (3469) (3679) (4239)
CEO Dummy 1501307 1508886 1493688 1508736 4713079 4713629 47278 4703908
(2648) (2684) (2618) (2683) (282) (2828) (2819) (2824)
Adjusted R-Sq 01861 02060 01695 02051 01071 01113 01033 011
Number of Observations 33229 33229 33229 33229 97679 97679 97679 97679
32
Table VIII Two Stage Least Squares (Reverse Causality) Tests - Executive Compensation Measures as a Function of Institutional Concentration
This table shows the coefficients from the Two-Stage Least Squares (2SLS) regressions of executive compensation measures against the lagged change
and contemporary change in shareholder wealth instrumented institutional concentration and total institutional ownership Tobinrsquos Q the natural
logarithm of firm market capitalization and dummy variable controls for CEO and year effects We instrument measures of institutional concentration
and total institutional ownership by orthogonalizing with respect to option grant pay-for-performance sensitivity salary and total compensation
Coefficients for industry and year controls are not presented First stage results are not presented One two and three asterisks denote significance at
10 5 and 1 levels
ExecuComp Sample 1992‒1997 Full Sample from ExecuComp 1992‒2002
PPSOG Salary TC PPSOG Salary TC
Change in Shareholder Wealth 144E-05 0004 02133 -409E-06 00011 00157
(308) (775) (596) (691) (638) (068)
Lag1 Change in Shareholder Wealth -240E-05 -00004 00866 118E-06 00002 01422
(131) (087) (346) (097) (116) (51)
Lag1 Institutional Concentration -00898 668681 3263143 02024 916656 2856417
(052) (801) (176) (163) (1545) (1004)
Lag1 Total Institutional Ownership 01021 -43997 2512587 02983 -104171 -1028347
(09) (084) (153) (4) (269) (487)
Lag1 Tobinrsquos Q 00754 -225606 119297 00655 -158657 2997205
(596) (1649) (053) (834) (2535) (727)
Lag1 Log Market Capitalization -04234 828713 7864046 -04002 825106 1685638
(2195) (8132) (2977) (4106) (11133) (373)
CEO Dummy 15848 3214117 2950274 19149 3205862 4529101
(2223) (9389) (2255) (3802) (1331) (2758)
Overall R-Sq -- 04981 01206 -- 04738 01178
Number of Observations 41305 41305 41305 85518 85518 85518
33
Table IX Two Stage Least Squares (Reverse Causality) Tests - Executive Compensation Measures as a Function of Top 5 Ownership
This table shows the coefficients from 2SLS regressions of executive compensation measures against the lagged change and contemporary change
in shareholder wealth instrumented Top 5 institutional ownership Tobinrsquos Q natural logarithm of firm market capitalization and dummy variable
controls for CEO and year effects We instrument Top 5 institutional ownership by orthogonalizing with respect to option grant pay-for-
performance sensitivity salary and total compensation Coefficients for industry and year controls are not presented First stage results are not
presented One two and three asterisks denote significance at 10 5 and 1 levels
ExecuComp sample 1992‒1997 Full Sample 1992‒2002
PPSOG Salary TC PPSOG Salary TC
Change in Shareholder Wealth 136E-05 00041 02128 -407E-06 00011 00146
(293) (79) (596) (692) (619) (063)
Lag1 Change in Shareholder Wealth -241E-05 -00003 00868 116E-06 00003 01439
(133) (07) (347) (096) (142) (514)
Lag1 Top 5 Institutional Ownership -03774 -187512 -3848531 00179 -104261 -126659
(163) (186) (166) (011) (149) (436)
Lag1 Tobinrsquos Q 00731 -223886 10074 00642 -155467 3115073
(573) (1631) (045) (819) (2509) (75)
Lag1 Log Market Capitalization -0416 775637 7696302 -04015 764632 1462536
(2345) (10084) (3766) (3688) (13708) (4072)
CEO Dummy 15854 3211991 2950158 19152 3203152 4519385
(2223) (9381) (2254) (3804) (133) (2752)
Overall R-Sq -- 04967 01202 -- 04722 01168
Number of Observations 41305 41305 41305 85518 85518 85518
34
Table X Specification Tests utilizing Box-Cox Transformation of Market Capitalization
This table presents results from a Box-Cox transformation of market capitalization In
each model we specify transformation of market capitalization only Panel A shows
results for the period 1992‒1997 and Panel B shows results for the entire period
1992‒2002 The first part of each panel shows the Lambda coefficient that applies to
the transformed market capitalization The second part of each panel shows log
likelihood scores with LR Chi2 Statistics in parentheses ‒ these test for the
appropriateness of a linear or log model specification
Panel A Our ExecuComp sample 1992‒1997
(1) (2) (3)
PPSOG Salary TC
Lambda -05143 01881 02876
(-1914) (2413) (1329)
Test H0 (Log Likelihood Score)
Lambda = -1 -1279748 -3178575 -4795326
(23866) (85952) (99688)
Lambda = 0 -1280152 -3138456 -4791159
(31934) (57147) (16339)
Lambda = 1 -1283157 -3163807 -4793506
(92039) (564156) (63292)
Panel B Full ExecuComp sample 1992‒2002
(1) (2) (4)
PPSOG Salary TC
Lambda -044 01853 04036
(-3015) (40) (3666)
Test H0 (Log Likelihood Score)
Lambda = 0 -2688334 -6476494 -10384490
(66599) (160356) (126236)
Lambda = 1 -2694886 -6538506 -10386795
(197644) (1400605) (172345)
35
Figure 2 Institutional Ownership Measures (LHA) vs Total
Compensation (RHA) by Market Capitalization
0
10
20
30
40
50
60
70
1 2 3 4 5 6 7 8 9 10
$00
$10
$20
$30
$40
$50
$60
$70
$80
$90
Millio
ns
InstitutionalConcentration (LHS)Top5 Ownership (LHS)
Institutional TotalOwnership (LHS)Total Compensation(RHS)
Page 5
5
differences between these two methods If there are one hundred shares outstanding in a
small company and the only institution owns one share that company is concentrated
according even though that institution may lack effective monitoring power Instead if
one deflates by shares on issue this results in a concentration ratio of 1100 If on the
other hand ten institutions collectively own 500 shares in a much larger company of
1000 shares and the Top Five collectively own 250 then the HS concentration ratio falls
to only 250500 or 05 and utilizing shares on issue 025 in this case As is confirmed by
the data one expects Top 5 Concentration to correlate negatively with firm size and Top
5 Ownership the reverse given a positive association between institutional ownership and
company size We then re-examine the monitoring hypothesis using this AHS proposed
measure of institutional investor influence that intuitively relates better to institutional
proxy voting power
Third we address issues of reliance on potentially biased self-reported values of option
grants that result in the exclusion of 16 percent of the HS sample size for tests on total
(direct) compensation Forth we argue that tests that HS perform using an alternative
measure of pay-for-performance sensitivity to provide support for the main hypothesis of
the study require one to get the timing right in terms of the lag structure of performance
and concentration Finally we address issues associated with their tests of endogeneity
and reverse causality These issues collectively might suggest overlooking of a
monitoring effect with the main driving forces size related
The next section reviews the HS study in an effort to tease out what might really be
going on Section II examines the literature for studies that in replicating HS provide
independent support for their findings It also reports findings on the actual proxy voting
record of mutual funds and actions of institutional investors in reducing pay Section III
describes the data and methodology Section IV analyzes the relation between
6
institutional influence and measures of pay-for-performance sensitivity Section V
documents the relationship between institutional influence and total and fixed
compensation Section VI examines the robustness of our findings and the final Section
concludes
I Review of Hartzell and Starks Methodology
We comment on five areas of the HS methodology 1) Size control 2) Measurement of
institutional influence 3) Testing the Jensen-Murphy measure of incentives 4) Reliance
on incomplete self-reported valued for option grants and 5) Endogeneity
A Size Control
In our paper we argue that HS unintentionally identify incentive-size and pay-size
effects5 not a monitoring effect It can be seen in Figure II that executive total pay for
HSrsquos Top 5 institutional concentration measure rises from as little as $072m for the
smallest market capitalization decile to $85m for the highest This twelve-fold pay-size
effect in the deciles can cause problems if one utilizes an untransformed size measure
Due to skewness in the size measure using this in its untransformed state appears to
weaken its effectiveness leaving open the possibility that correlations they identify
between incentives and pay levels on the one hand and Top 5 Concentration on the other
are due largely to size-related effects
Table III displays this strong pay-size effect based on our full 1992‒2002 dataset and
graphically in Figure 1 It shows that the incentive measure adopted by HS declines from
$272 per $1000 change in shareholder wealth for the smallest market capitalization
decile of firms to approximately $030 for the largest Table III and Figure 1 show the
5 The multiplicative model of Edmans Gabaix and Landier (2009) finds that incentive pay increases with
firm size (elasticity value of 13) and the empirical elasticity is 037
7
strong relation between institutional concentration as measured by HS and market
capitalization For the smallest decile of market capitalization firms institutional
concentration averages 61 percent declining to about 275 percent for the highest decile
This negative relationship between firm size and institutional concentration seems to arise
because institutional investors do generally favor small firms This causes problems
because in the limit a small company may have only one institutional investor owning
only a small percentage of the company Hence it would appear to be fully concentrated
even though the ability of the institutional investor to influence pay decisions might be
small Thus it is not at all surprising that the correlation between the logarithm of market
capitalization and their institutional concentration measure is both negative and high in
absolute value at 63 percent Their methodology appears to favor small companies that
Table III shows to have low pay levels and high pay for performance sensitivity
Moreover if one does not transform the raw size measure it implicitly assumes that a
given dollar change in firm size has the same absolute impact on pay for a very large
company as it does for a small company Indeed the conventional logarithmic
specification for firm size provides proportionate effects that imply a pay-size elasticity of
approximately one-third instead of unity6
B Measurement of Institutional Influence
We argue that there are additional issues related to the HS measure of institutional
influence ndash institutional concentration In their companion article to HS AHS develop a
model where they capture institutional influence by a single large long-term institutional
6 Note that unlike an earlier version of our paper we leave the HS dependent variables in levels and hence
untransformed While we do this to minimize the alterations that we make to the HS specification ideally
the elasticity specification is preferable especially for the compensation variables Transforming the
dependent variables does not significantly affect the results
8
investor They measure influence relative to total shareholding including private or
atomistic shareholders Presumably they do this to capture relative activism and proxy
voting strength However when it comes to the empirical implementation of the model
AHS adopt the same relative institutional concentration measure as in HS (AHS (p12))
potentially downplaying the relative strength of institutions compared with other
shareholders They justify this because there is generally more than one institutional
investor and there will be free-riding problems with multiple investors that only the
largest institutional investors can overcome While we accept their premise of free-rider
problems it is not obvious why the deflator is lsquoinstitutional shareholdingrsquo rather than
lsquototal shareholdingrsquo from this perspective7 If the strength of the size control can be
queried scaling by total shareholding (ie Top 5 Ownership) produces a measure that
remains relatively unaffected across the size deciles It is 20 percent for the smaller
decile declining to 14 percent for the largest decile whereas for the HS concentration
measure it declines by over 50 percent from 61 percent to 28 percent (see Table III and
Figure 1)
C Testing the Jensen-Murphy Measure of Incentives
The testing of their secondary measure of executive pay-for-performance sensitivity
(see HS Table IV) based on Jensen and Murphy (1990) raises issues other than the size
control HS regress changes in compensation (cash and total) on lagged institutional
influence that has been interacted with the contemporaneous change in shareholder
wealth to capture a pay-for-performance relationship Such a test is appealing ndash
7 AHS do point out one benefit of the institutional concentration measure Institutional investors may prefer
companies with lsquobetterrsquo incentive compensation structures Could using institutional shareholding rather
than total shareholding as the deflator alleviate this problem Presumably both the numerator and
denominator will have the same bias and hence error cancellation is a possibility
9
institutions should increase compensation when the firm has performed well Following
Greene (2000 p326) to execute such a test one needs to include in level form variables
that are then interacted Hence the HS model specification ought to be
1 2 3 1 4 1it it it it it itCompn SW IC SW IC (1)
where itCompn is the change in executive compensation itSW is the contemporaneous
change in shareholder wealth which also appears in the interaction term and 1itIC is
lagged institutional ownership concentration However HS (p2363) omit the
contemporaneous term itSW and the lagged term 1itIC which should appear in levels
as well as in the interaction term Most importantly HS interact their lagged institutional
concentration measure 1itIC with the contemporaneous change in shareholder wealth
itSW As it is after all pay-for-performance concentrated institutions need to condition
their influence on some observed measure of firm performance Contemporaneous firm
performance is unknown at the time institutional influence is measured Reflecting this
timing insight leads to a new Greene (2000) specification
1 2 1 3 1 4 1 1it it it it it itCompn SW IC SW IC (2)
such that all the lagged RHS independent variables that now appear in the interaction
term are included on a stand-alone basis
Reliance on incomplete self-reported values for option grants
Prior to receiving the HS dataset we tried to replicate the findings of HS to validate
our datasets for the HS period and the longer period 1992‒2002 While we initially
focused on the same period as HS (ie 1992‒1997) we found that our sample size was
larger at 47765 observations than HSrsquos at 33928 observations This discrepancy could
have arisen if HS did not replace missing and self-reported option valuations using their
10
Black-Scholes option methodology This may introduce a downward pay bias since self-
reported option-grant pay is subject to possible understatement and there are 13839
observations with likely high option pay dropped altogether
Endogeneity
HSrsquos endogeneity testing raises concerns about reverse causality Institutions may
prefer to hold firms with certain compensation structures such as incentivized executives
HS rightly highlight this concern and use instrumental variables to address this issue HS
assume that institutional ownership reflects any reverse causality Using an instrumental
variables approach to resolve this problem involves finding a variable that influences or
explains institutional ownership but not directly executive compensation They use share
turnover in a firm as an instrument for institutional ownership However theoretical work
by Holmstrom and Tirole (1993) indicates that measures of stock liquidity and trading
activity are indeed important determinants of executive market-based incentives as Kang
and Lui (2008) and Dikolli Kulp and Sedatole (2009) find empirically As such one can
make a case for adopting another instrument rather than share turnover for institutional
ownership
II Previous Research
Prior research fails to address issues concerning possible fragility of the HS results
AHS replicate the HS findings on both incentives and pay levels using the same dataset as
HS and they apply the same definition of institutional concentration and use market
capitalization expressed as a level Cadman Klasa and Matsunaga (2010 Table 6) set out
to replicate HSrsquos (Table II) model of option grant pay-for-performance sensitivity when
they compare a dataset of ExecuComp and Non-ExecuComp firms for the period 2000-
2007 Their ExecuComp results for this sample period are quite similar to our results for
11
the 1992‒2002 period for the same specification controlling for the logarithm of market
capitalization but they do not regard these results as a test of HS as their focus is entirely
on comparing the two groups of stocks Nor do they investigate the impact of
concentration on either cash compensation or total pay Kang and Liu (2008) examine
pay-performance sensitivity use a version of the HS institutional concentration measure
as a control variable in a different framework deriving from Holmstrom and Tirole
(1993) Dikolli Kulp and Sedatole (2009) argue that firms with transient ownership will
issue relatively more equity incentives to counter short-termism with their results robust
to the inclusion of the HS concentration variable
We hoped to find supportive evidence from proxy voting behavior in support of HSs
findings However evidence from actual mutual fund proxy voting records suggests that
most oppose reductions in CEO pay but overall the evidence is mixed8 Davis and Kim
(2005) examine proxy votes of mutual funds for 2004 and find that for a sample of
Fortune 1000 proxy contests in 45 cases that attempted to limit executive pay mutual
funds were unanimously opposed to such attempts Rothberg and Lilien (2006) also
examine proxy-voting data and find that mutual funds voted 66 percent of the time in
managementrsquos favor on issues of executive compensation Ertimur Ferri and Muslu
(2011) find that institutional investors lack sophistication in that they do not target
excessively paid CEOs relative to CEOs with high predicted-pay based on economic
determinants but exert a moderating influence ndash a $23 million reduction ndash for those they
deem ldquoexcessively paidrdquo CEOs
III Data and Methodology
8 We do acknowledge that the HS results are for all institutional investors while the transparency with
respect to proxy voting applies only to mutual funds
12
This Section discusses the data used in the study the construction of variables
descriptive statistics and the methodology Two datasets are used ndash one is the original
dataset generously supplied by Jay Hartzell and Laura Starks which covers 1992‒1997
second a more comprehensive dataset covering 1992‒2002 is computed from the same
data sources used in the HS dataset This is Standard and Poorrsquos (SampP) ExecuComp for
compensation data Thomson Financial CDASpectrum database for institutional holdings
to compute measure of institutional influence SampPrsquos Compustat and the Center for
Research in Stock Prices (CRSP) for firm level accounting and stock price data Our more
comprehensive dataset contains 2559 firms compared with 1914 for the HS dataset Our
full matched sample yields 97679 executive years
A Institutional Influence Measures
To measure institutional influence we employ two ownership-based metrics The first
measure of institutional influence is HSrsquos Institutional Concentration We use this for
direct comparison to the HS analysis The second measure of institutional influence
changes the denominator of the HS measure from institutional shares held to total shares
outstanding in the firm with nomenclature Top 5 Ownership9 We compute both
measures at the fiscal year end of the firm We adjust all compensation measures and
asset prices for inflation to 30 June 2002 prices for our comprehensive sample compared
to the HS dataset that is in the dollars of the day
B Compensation Measures
The ExecuComp database reports several raw measures of executive compensation for
the top five executives (where available) including total salary bonus long-term
9 Top 5 Ownership is included in the HS database but not utilized by them in reported tests as well as
constructed by us for the more comprehensive sample
13
incentives and number of option grants For our more comprehensive dataset we replicate
all the compensation measures used by HS In common with HS we compute the pay-for-
performance sensitivity of option grants using the Yermack (1995) approach This
measure shows the dollar change in value of the executive options for every dollar change
in the value of the firm This involves calculating the option delta using the Black-
Scholes formula adjusted for dividends10
The ratio of the number of shares represented
by the award of options divided by the diluted number of shares outstanding at the
beginning of the year multiplies the option delta with units given by a $1000 change in
firm value Summing salary bonus long-term incentive plan payouts and stock and
option grants provides total direct compensation Options are valued using the Black and
Scholes (1973) formula for European call options adjusted by Merton (1973) to
incorporate dividends
C Descriptive Statistics
Table I presents descriptive statistics of key variables employed in the study according
to our comprehensive dataset11
Institutions own on average 546 percent of the firms in
which they invest This is almost identical to the 531 percent found by HS The
institutions with the five largest holdings account 45 percent of institutional ownership
Once again the figure for HS is almost identical at 44 percent The institutions with the
five largest holdings own on average 229 percent of shares outstanding in a firm (HS
have 223 percent) Average total compensation for executives is $2583 million
considerably larger than that documented by HS ($1250 million) but the HS figure is an
10 When there is more than one grant during the year a weighted average delta is used
11 With the data supplied to us by Hartzell and Starks we are able to replicate exactly their summary data
(HS Table I) for all but cash compensation Salary is included but bonus payments are missing The
replicated table is available on request
14
amalgam of price levels over their sample period 1992‒1997 and ours are in consistent
dollars of June 30 2002 However there has also been significant growth in
compensation levels from 1998 to 2002
ltltINSERT TABLE Igtgt
Based on our comprehensive data in Table II we examine correlations between the
key variables employed in the study Of concern is the large negative correlation of 63
percent which has been discussed previously (Section IIA) between the HS measure of
institutional concentration and the natural logarithm of firm market capitalization
ltltINSERT TABLE IIgtgt
In view of this high correlation Table III provides a breakdown of the three key pay
variables and three institutional ownership variables by size deciles We find that the high
institutional concentration but low institutional ownership share of total shareholdings
mentioned above is evident in the smallest decile of firms by size This decile has low
salary and even lower total compensation in a relative sense but by far the highest pay-
for-performance sensitivity of option grants Small firms award executives a far higher
proportion of market capitalization Since small firm market-capitalization is negligible in
comparison with the highest decile (300 times larger in magnitude) total compensation is
very small in absolute value Moreover due to the domination of the pay-size effect base
salaries are also exceedingly small
ltltINSERT TABLE IIIgtgt
Figure 1 based on Table III shows how as institutional concentration falls with
increases in stock size (but also as the total share of institutional ownership increases)
and the pay-for-performance sensitivity of option grants falls Thus because of what
appears to be a weak size control and size-influenced concentration measure HS
15
potentially pick up a size effect as their main result but not necessarily a monitoring
effect Figure 2 shows that as firm size increases up to decile 8 at least the share of
institutional total ownership increases as does total compensation
ltltINSERT FIGURES 1 AND 2 gtgt
D Methodology
Following HS we use two estimation techniques to examine the effect of institutional
investor influence on executive compensation Due to the censored nature of option grant
pay-for-performance sensitivity this compensation measure most appropriately uses
Tobit model specification whereas an Ordinary Least Squares (OLS) model specification
is used for all other compensation measures We model compensation measures as a
function of contemporaneous and lagged change in shareholder wealth institutional
influence Tobinrsquos Q and market capitalization Controls for CEO industry and year
effects are also used Augmentation occurs to test the effect of institutional influence on
the change in executive cash and total compensation with an interaction term that is the
product of institutional influence and the change in shareholder wealth
The variables used to model executive compensation have the following definitions
The lagged and contemporary measure of change in shareholder wealth is simply the
difference in market capitalization of the firm The measures of institutional influence
used are Institutional Concentration and Top 5 Ownership Use of the two types of
institutional influence occurs in separate models In order to replicate the HS
methodology models using Institutional Concentration have an additional control
consisting of lagged total institutional holdings and Tobinrsquos Q controls for performance
and growth opportunities Market capitalization of the firm is the number of shares
outstanding multiplied by the share price with inclusion both in levels and natural
16
logarithmic form A dummy variable is included to identify whether the executive is the
CEO Dummy variables are also included to control for industry effects This variable
takes a value of 1 for the two-digit Standard Industrial Classification (SIC) in which the
firm operates A year fixed effects (dummy) variable is included taking a value of 1 if the
observation was from the given year
IV Executive Incentives as a Function of Institutional Investor Influence
This section presents the results for the effect of institutional influence on option grant
pay-for-performance sensitivity and changes in compensation
A Option Grant Pay-for-Performance Sensitivity
Table IV presents the results for the effect of institutional influence on option grant
pay-for-performance sensitivity
ltltINSERT TABLE IVgtgt
Model (1) is a direct replication of HS using the HS database as supplied to us by HS
with all the variables computed in the same way and in the same functional form as with
HS The coefficient and t-statistic on lagged institutional concentration interacted with the
change in shareholder wealth is 13994 and 767 that compares to 1358 and 842 in HS
We are unable to account for these minor discrepancies for values that should be
identical In model (2) the only change to the model is that we use the natural logarithm
of market capitalization instead of the level of market capitalization This single change
has a sizeable effect on the coefficient on institutional concentration The coefficient on
institutional concentration is now -00692 and is insignificant The fact that an order
preserving transformation to a control variable can change the results is of interest The
log-transformation to market capitalization also changes the coefficient on Top 5
Ownership in models (3) and (4) The coefficient on Top 5 Ownership is significant in
17
both model (3) and in model (4) but the better size control reduces the coefficient by over
25 percent The better size control does not affect Top 5 Ownership by as much as its
design ensures less sensitivity to what could be inadequate size controls
We also test the effect of institutional influence on option grant pay-for-performance
sensitivity over the period 1992‒2002 This sample is then approximately three times as
large as the sample used by HS Institutional concentration is highly positively significant
in model (5) similar to the results for the period 1992‒1997 When we use the log of
market capitalization in model (6) the coefficient on institutional concentration remains
positive and significant at the 10 percent level with a t-statistic of 166 Of importance
though is that the coefficient on institutional concentration is one-eighth of the coefficient
in model (5) Even for a measure of institutional influence which we believe is less
sensitive to firm size the use of log market capitalization reduces the coefficient on Top 5
Ownership by approximately 50 percent from model (7) to model (8)
B Pay-for-Performance Sensitivity due to Changes in Cash and Total
Compensation
For robustness HS examine the effect that institutional influence conditioned on firm
performance has on a changes in cash and total executive compensation using the general
method of Jensen and Murphy (1990) to compute pay-for-performance sensitivity
ltltINSERT TABLE V gtgt
Table V presents the effect of institutional influence on the change in cash
compensation using the HS size control measure throughout12
Model (1) replicates the
specification in HS for the period of 1992‒1997 and we produce a similar result ndash the
interaction of lagged institutional concentration and contemporaneous change in
12 This test uses our data for the two periods 1992‒1997 and 1992‒2002 We do not use the HS data for
cash compensation as data for bonus is missing in the data supplied to us
18
shareholder wealth positively and significantly relates to a change in executive cash
compensation In model (2) we include the contemporaneous change in shareholder
wealth and the lagged institutional concentration as stand-alone controls as indicated by
Greenersquos (2000) specification of equation (1) above and the interaction term remains
significant In model (3) we interact lagged institutional concentration with the lagged
change in shareholder wealth as in equation (2) above that tries to take account of what
participants could have observed at the time The interaction term is then negative but
insignificant altering the HS result Model (4) provides a similar result where our
preferred proxy Top 5 Ownership captures concentrated institutional influence In
models (5) to (8) we use our full sample In model (5) we find that the lagged
institutional concentration interaction with the contemporaneous change in shareholder
wealth is positive and significant in HSrsquos specification However including change in
contemporaneous shareholder wealth and lagged institutional concentration as separate
control variables as suggested by equation (1)rsquos specification removes the significance
from the interaction term in model (6) Moreover in model (7) when lagged institutional
concentration is interaction with the lagged change in shareholder wealth in accord with
equation (2) this interaction term is now negative and significant reversing the HS
finding We obtain a similar result in model (8) when we measure institutional
concentration influence using Top 5 Ownership13
V The Levels of Executive Fixed and Total Compensation as Functions of
Institutional Investor Influence
13 In unreported results we use the log of market capitalization as a size control These changes to the size
control do not dramatically change the findings from Table V The reason why this size issue is not as
important as in the findings for option grants pay-for-performance is that the change in cash compensation
is not so strongly influenced by firm size In unreported results available on request we also examine the
effect of institutional influence on changes in total compensation with similar results to the change in cash
compensation
19
Two measures of compensation are examined ndash salary and total compensation
ltltINSERT TABLE VI gtgt
Table VI examines the effect of institutional influence on the level of salary In model
(1) we use the HS sample and specification to obtain almost identical results The
coefficient on institutional concentration is (negative) $247509 that is almost identical to
(negative) $247967 in HS Both are highly significant In model (2) we make a single
change to the model ndash we use the natural logarithm of market capitalization This single
change reverses the sign of institutional concentration with the coefficient still being
significant at the 1 percent level In models (3) and (4) we measure institutional influence
using Top 5 Ownership In model (3) where we measure firm size using the level of
market capitalization Top 5 Ownership is insignificant When we use the natural
logarithm of market capitalization in model (4) Top 5 Ownership is significant and
positively relates to the salary level In models (5) to (8) we estimate the same models
using our full sample We find similar results for the full sample as for the period 1992ndash
1997
ltltINSERT TABLE VII gtgt
Table VII examines the effect of institutional influence on total compensation In
model (1) we replicate the HS methodology using the HS data to obtain a similar
significant negative relation between institutional concentration and total compensation of
(negative) $1608633 However after changing to a logarithmic specification of market
capitalization in model (2) institutional concentration increases total compensation by a
significant $32400014
Importantly when institutional influence is measured using Top 5
14 = (5688751 times 0101)
20
Ownership an insignificant relation exists with total compensation in model (3) using the
HS size control but it becomes both positive and significant in (4) with the logged size
control In models (5) and (6) where models are estimated using the full sample we
observe a similar reversal in sign as occurs in models (1) and (2) However now the
magnitude of the reversal is almost equal with a difference of about $67m that is very
large indeed In model (5) the coefficient on institutional concentration is (negative)
$342m and then in model (6) the coefficient is $33m Top 5 Ownership is insignificant
in both models (7) and (8) unlike the results for Salary
VI Robustness
Two issues are examined in is this Section ndash reverse causality and the appropriate
transformation of market capitalization A methodological issue inherent in examining the
effect of ownership-based measures of institutional influence on executive compensation
is endogeneity driven by reverse causality We utilize an instrumental variables approach
where orthogonalized institutional influence measures are regressed on each of the
compensation measures lagged option grant pay-for-performance sensitivity salary and
total compensation In addition market capitalization and a CEO dummy are included in
each regression We use the residual from each regression as an instrument for
institutional influence Such a process removes any preferences for compensation from
the institutional influence measures
ltltINSERT TABLES VIII AND IXgtgt
Table VIII presents the results where we instrument the HS measure lagged
institutional concentration with a compensation-orthogonalized equivalent The analysis
indicates that the instrumented institutional concentration measure is positively and
significantly related to the level of salary and total compensation and insignificantly
21
related to option grant pay-for-performance sensitivity (PPSOG) Table IX presents
results where we instrument Top 5 Ownership with a compensation-orthogonalized
equivalent The instrumented Top 5 Ownership measure insignificantly relates to option
grant pay-for-performance sensitivity and barely negatively relates to salary although it
negatively relates to total compensation suggesting reverse causality is driving a
significant portion of our previous results
ltltINSERT TABLE Xgtgt
Another issue examined in this section is the appropriate transformation of market
capitalization Aneuryn-Evans and Deaton (1980) provide an evaluation of methods
comparing linear and log-linear specifications We use the Box-Cox test to compare
logarithm versus linear specifications In applying the Box-Cox test we specify
transformation of only market capitalization ndash the dependent variable and all other
independent variables remain unchanged Table X presents the lambda coefficient for
each model and the log likelihood score In all models in Panel A and B a lambda of zero
receives a higher (closer to zero) log likelihood score suggesting the log specification is
appropriate
VII Conclusion
Hartzell and Starks (2003) appears to document an effective role for institutional
investors in increasing executive incentives and lowering compensation levels Our paper
re-addresses the issue of concentrated institutional monitoring from a perspective
informed by an additional decade of research since their paper appeared We cannot find
evidence that concentrated institutional investors really have any positive (or negative for
that matter) effect on executive incentive levels While superficially we find evidence
that they may raise executive pay levels this is most likely driven by an institutional
22
preference for highly remunerated executives in large companies and thus by reverse
causality Moreover our findings are not consistent with the rent-extraction hypothesis
suggesting that firms systematically overpay executives in the absence of effective
monitoring by outside parties such as institutional investors
23
References
Almazan Andres Jay C Hartzell and Laura T Starks 2005 ldquoActive Institutional
Shareholders and Costs of Monitoring Evidence from Executive Compensationrdquo
Financial Management Winter 5ndash34
Aneuryn-Evans Gwyn and Angus Deaton 1980 ldquoTesting Linear versus Logarithmic
Regression Modelsrdquo Review of Economic Studies 47 (1) Econometrics Issue 275ndash
291
Babenko Ilona 2009 ldquoShare Repurchases and Pay-Performance Sensitivity of Employee
Compensation Contractsrdquo Journal of Finance 64117-150
Baker George P and Brian J Hall 2004 ldquoCEO Incentives and Firm Sizerdquo Journal of
Labor Market Economics 22 767ndash798
Bebchuk Lucian A Jesse M Fried and David Walker 2002 ldquoManagerial power and
rent extraction in the design of executive compensationrdquo University of Chicago Law
Review 69 751ndash846
Bebchuk Lucian A and Jesse M Fried 2003 ldquoExecutive Compensation as an Agency
Problemrdquo Journal of Economic Perspectives 17 71ndash92
Bebchuk Lucian A and Jesse M Fried 2004 Pay without Performance The Unfulfilled
Promise of Executive Compensation Cambridge MA Harvard University Press
Bertrand Marianne and Sendhil Mullainathan 2001 ldquoAre CEOs Rewarded for Luck
The Ones Without Principals Arerdquo Quarterly Journal of Economics 116 901ndash32
Black Fischer and Myron Scholes 1973 ldquoThe Pricing of Options and Corporate
Liabilitiesrdquo Journal of Political Economy 81 637ndash54
Cadman Brian Sandy Klasa and Steve Matsunaga 2010 ldquoDeterminants of CEO Pay A
Comparison of ExecuComp and Non-ExecuComp Firmsrdquo The Accounting Review 85
1511ndash1543
Chhaochharia Vidhi and Yaniv Grinstein 2009 ldquoCEO Compensation and Board
Structurerdquo Journal of Finance 64 231ndash261
Cichello Michael S 2005 ldquoThe impact of firm size on pay-performance sensitivitiesrdquo
Journal of Corporate Finance 11 609ndash627
Davis Gerald F and E Han Kim 2007 ldquoBusiness ties and proxy voting by mutual
fundsrdquo Journal of Financial Economics 85 552ndash570
Dikolli Shane S Susan L Kulp and Karen L Sedatole 2009 ldquoTransient Institutional
Ownership and CEO Contractingrdquo Accounting Review 84 737ndash770
Edmans Alex Xavier Gabaix and Augustin Landier 2009 ldquoA Multiplicative Model of
Optimal CEO Incentives in Market Equilibriumrdquo Review of Financial Studies 22
4881ndash4917
Edmans Alex 2009 ldquoBlockholder Trading Market Efficiency and Managerial Myopiardquo
Journal of Finance 64 2481ndash2513
Ertimur Yonca Fabrizio Ferri and Volkan Muslu 2011 ldquoShareholder Activism and
CEO Payrdquo Review of Financial Studies 24 535ndash592
24
Gabaix Xavier and Augustin Landier 2008 ldquoWhy Has CEO Pay Increased So Muchrdquo
Quarterly Journal of Economics 123 49ndash100
Garen John 1994 ldquoExecutive compensation and principal-agent theoryrdquo Journal of
Political Economy 102 1175ndash1199
Gibbons Robert and Kevin J Murphy 1992 ldquoOptimal Incentive Contracts in the
Presence of Career Concerns Theory and Evidencerdquo Journal of Political Economy
100 468ndash505
Greene William H 2000 (4th
Edition) Econometric Analysis Prentice-Hall NJ
Guthrie Katherine Jan Sokolowsky and Kam-Ming Wan 2012 ldquoCEO Compensation
and Board Structure Revisitedrdquo Journal of Finance Forthcoming
Hadlock Charles R and Gerald B Lumer 1997 ldquoCompensation turnover and top
management incentives historical evidencerdquo Journal of Business 70 153ndash187
Hartzell Jay C and Laura T Starks 2003 ldquoInstitutional Investors and Executive
Compensationrdquo Journal of Finance 58 2351ndash2374
Huson Mark R Robert Parrino and Laura T Starks 2001 ldquoInternal Monitoring
Mechanisms and CEO Turnover A Long-Term Perspectiverdquo Journal of Finance 56
2265ndash2297
Jensen Michael C and Kevin J Murphy 1990 ldquoPerformance Pay and Top Management
Incentivesrdquo Journal of Political Economy 98 225ndash264
Jin Li 2002 ldquoCEO compensation diversification and incentivesrdquo Journal of Financial
Economics 66 29ndash63
Kang Qiang and Qiao Liu 2008 ldquoStock Market Information Production and Executive
Incentivesrdquo Journal of Corporate Finance 14 484ndash498
Merton Robert 1973 ldquoTheory of Rational Option Pricingrdquo Bell Journal of Economics
and Management Science 4 141-183
Morse Adair Vikram Nanda and Amit Seru 2011 ldquoAre Incentive Contracts Rigged by
Powerful CEOsrdquo Journal of Finance 66 1779ndash1821
Murphy Kevin J 1985 ldquoCorporate performance and managerial remuneration An
empirical analysisrdquo Journal of Accounting and Economics 7 11ndash42
Murphy Kevin J 1999 ldquoExecutive compensationrdquo in Orley Ashenfelter and David
Card Handbook of Labor Economics vol 3 Elsevier
Rosen Sherwin 1992 ldquoContracts and the market for executivesrdquo In Contract
Economics ed Lars Werin and Hans Wijkander 181ndash211 Cambridge MA
Blackwell
Rothberg Burton and Steven Lilien 2006 ldquoMutual Funds and Proxy Voting New
Evidence on Corporate Governancerdquo Journal of Business and Technology Law 1
157ndash184
Schaefer Scott 1998 ldquoThe dependence of payndashperformance sensitivity on the size of the
firmrdquo Review of Economics and Statistics 80 436ndash443
Yermack David 1995 ldquoDo Corporations Award CEO Stock Options Effectivelyrdquo
Journal of Financial Economics 39 237ndash69
25
Table I Univariate Descriptive Statistics
This table reports sample statistics for the key variables used in this study based on our comprehensive
ExecuComp dataset Institutional Concentration is the shares owned by the institutions with the five largest
holdings divided by the shares held by all institutions Top 5 Ownership is the proportion of shares
outstanding held by the institutions with the five largest holdings Total Institutional Ownership is the
proportion of shares held by all institutions Panel B shows the measures executive compensation The pay-
for-performance sensitivity (PPS) of option grants shows the change in executive wealth due to a $1000
change in firm in shareholder wealth Expressions for Salary and Total Compensation are in thousands of
dollars while Market Capitalization is in millions of dollars Panel C shows firm characteristics
Variable Mean Median Std Dev P10 P90
Panel A Institutional Investor Holdings
Institutional Concentration 0450 0424 0151 0282 0656
Top 5 Institutional Ownership 0229 0220 0101 0110 0355
Total Institutional Ownership 0546 0561 0206 0258 0801
Panel B Executive Compensation
PPS of Option Grants 1161 0156 3829 0 2822
Salary 36311 29462 25231 15179 67000
Total Compensation 258313 85878 980701 23875 511022
Panel C Firm Characteristics
Tobins Q 1744 1139 2738 0495 3322
Market Capitalization 560699 112094 1988108 19754 1048754
26
Table II Pairwise Correlation Coefficients
This table presents the correlations between the key variables based on our comprehensive
ExecuComp dataset Institutional Concentration is the shares owned by the institutions with the five
largest holdings divided by the shares held by all institutions Top 5 Ownership is the proportion of
shares outstanding held by the institutions with the five largest holdings Total Institutional
Ownership is the proportion of shares held by all institutions The pay-for-performance sensitivity
(PPS) of option grants is the change in executive wealth due to a $1000 change in shareholder
wealth
(1) (2) (3) (4) (5) (6) (7) (8)
(1) Institutional Concentration 1
(2) Top 5 Ownership 0268 1
(3) Total Ownership -0415 0706 1
(4) PPS of Option Grants 0122 0042 -0024 1
(5) Log Salary -0270 -0012 0171 0038 1
(6) Log Market Capitalization -0630 -0157 0261 -0156 0420 1
(7) Level Market Capitalization -0241 -0157 0007 -0058 0219 0534 1
(8) Tobins Q -0008 -0034 0002 0015 -0083 0066 0121 1
27
Table III
Descriptive Statistics by Market Capitalization Deciles
Market
Capitalization
Deciles
Market
Capitalization
($m)
Total
Compensation
($000)
Salary
($)
PPS Option
Grants
($)
Institutional
Concentration
()
Top 5 Ownership
()
Institutional
Total Ownership
()
1 (Small) 1181 76008 234700 2720 611 200 327
2 2668 90206 257955 1796 495 217 438
3 4193 105946 275120 1517 454 220 485
4 6045 129892 296067 1376 420 206 491
5 8742 155391 315864 1216 394 198 502
6 13086 191888 345942 1072 380 202 531
7 20790 239501 377021 0874 368 196 532
8 34920 314955 427808 0687 354 190 538
9 67747 450970 480319 0569 328 173 528
10 (Big) 366540 846066 614354 0297 275 138 502
28
Table IV Pay-for-Performance Sensitivity of Option Grants as a Function of Institutional Influence
Similar in construction to parts of HS (Table II) this table shows the coefficients from Tobit regressions of option grant pay-for-performance sensitivity per $1000 change in
shareholder wealth against the lagged change and contemporary change in shareholder wealth measures of institutional influence Tobinrsquos Q firm market capitalization and
dummy controls for CEO and year effects We estimate models (1) to (4) over the period 1992‒1997 using data supplied by HS and models (5) to (8) are estimated over the
whole sample Models (1) (2) (5) and (6) use Institutional Concentration as a measure of institutional influence which HS define as the shares owned by the institutions
with the five largest holdings divided by the shares held by all institutions These models control for total institutional ownership Models (3) (4) (7) and (8) use Top 5
Ownership as a measure of institutional influence which we define as the proportion of shares outstanding held by the institutions with the five largest holdings Coefficients
for industry and year controls are not presented One two and three asterisks denote significance at 10 5 and 1 levels
Hartzell and Starks Database 1992‒1997 Full Sample from ExecuComp 1992‒2002
(1) (2) (3) (4) (5) (6) (7) (8)
Change in Shareholder Wealth 77721 191781 89874 137368 -446 -218 -664 -203
(171) (429) (194) (317) (319) (168) (453) (158)
Lag1 Change in Shareholder Wealth 14446 99963 88719 8544 -144 0273 -0342 00852
(027) (212) (166) (184) (071) (015) (016) (005)
Lag1 Institutional Concentration 13994 -00692 26405 03304
(767) (032) (1402) (166)
Lag1 Total Institutional Ownership 08055 10434 07106 0937
(593) (781) (666) (898)
Lag1 Top 5 Institutional Ownership 10567 07457 12116 06293
(399) ( 283) (533) (278)
Lag1 Tobinrsquos Q 0068 00877 00647 00813 00498 00727 00508 00695
(337) (434) (319) (401) (504) (734) (51) (703)
Lag1 Market Capitalization -106202 -173515 -776 -117
(474) (742) (1011) (153)
Lag1 Log Market Capitalization -0297 -02319 -04111 -03908
(1576) (1439) (2997) (2841)
CEO Dummy 13745 13682 13765 13674 28197 28128 28155 28143
(1913) (1918) (1914) (1916) (3388) (3396) (3377) (34)
Number of Observations 33878 33878 33878 33878 97679 97679 97679 97679
29
Table V Pay-for-Performance Sensitivity of Cash Compensation (Salary plus Bonus) as a Function of Institutional Influence - Level Market Capitalization
Similar in parts to models (1) and (2) of HS (Table IV) this table displays the coefficients from OLS regressions of the change in cash compensation against the lagged
change and contemporary change in shareholder wealth measures of institutional influence Tobinrsquos Q firm market capitalization and dummy controls for CEO and year
effects We estimate models (1) to (4) over the period 1992‒1997 based on our ExecuComp sample and models (5) to (8) over the whole sample Models (1) (2) (5) and (6)
use Institutional Concentration as a measure of institutional influence which HS define as the shares owned by the institutions with the five largest holdings divided by the
shares held by all institutions These models control for total institutional ownership Models (4) and (8) use Top 5 Ownership as a measure of institutional influence which
we define as the proportion of shares outstanding held by the institutions with the 5 largest holdings Coefficients for industry and year controls are not presented One two
and three asterisks denote significance at 10 5 and 1 levels
Our Database based on ExecuComp 1992‒1997 Full Sample based on ExecuComp 1992‒2002
(1) (2) (3) (4) (5) (6) (7) (8)
Lag1 Change in Shareholder Wealth 00014 00017 00088 00089 -00006 -00007 00072 00086
(051) (062) (129) (103) (027) (029) (131) (16)
Change in Shareholder Wealth -00102 00102 00105 00032 00102 00101
(097) (292) (293) (055) (611) (608)
Change Shareholder Wealth Lag1 Institutional
Concentration
00435 00764 00363 00253
(463) (258) (689) (144)
Lag1 Change Shareholder Wealth Lag1
Institutional Concentration
-00197 -00283
(122) (197)
Lag1 Institutional Concentration -492182 -288861 400149 453623
(09) (056) (098) (114)
Lag1 Total Institutional Ownership 720688 538372 655091 559587 682069 726226
(276) (232) (281) (314) (308) (322)
Lag1 Change Shareholder Wealth Lag1 Top 5
Institutional Ownership
-0046 -00689
(091) (256)
Lag1 Top 5 Institutional Ownership 583044 1191634
(092) (25)
Lag1 Tobinrsquos Q -32761 -36806 -19422 -18023 -54283 -48681 -29534 -24243
(138) (172) (077) (078) (227) (245) (136) (111)
Lag1 Market Capitalization 00004 00004 00002 00006 00007 00008 00007 00007
(042) (044) (022) (07) (086) (092) (082) (09)
CEO Dummy 768035 765474 768334 773634 616315 61729 619028 618535
(354) (355) (356) (355) (417) (418) (419) (418)
Adjusted R-Sq 00057 0006 00052 0005 00097 00097 00098 00102
Number of Observations 41305 41305 41305 41305 85518 85518 85518 85518
30
Table VI Level of Salary as a Function of Institutional Influence
Similar in parts to models (1) and (2) of HS (Table V) this table displays the coefficients from OLS regressions of salary against the lagged change and contemporary change in
shareholder wealth measures of institutional influence Tobinrsquos Q firm market capitalization and dummy controls for CEO industry and year effects We estimate models (1) to
(4) over the period 1992‒1997 using the actual data supplied by HS and models (5) to (8) over the whole sample 1992‒2002 Models (1) (2) (5) and (6) use Institutional
Concentration as a measure of institutional influence which HS define as the shares owned by the institutions with the five largest holdings divided by the shares held by all
institutions These models control for total institutional ownership Models (3) (4) (7) and (8) use Top 5 Ownership as a measure of institutional influence which we define as
the proportion of shares outstanding held by the institutions with the 5 largest holdings Coefficients for industry and year controls are not presented One two and three
asterisks denote significance at 10 5 and 1 levels
Hartzell and Starks Database 1992‒1997 Full Sample based on ExecuComp 1992‒2002
(1) (2) (3) (4) (5) (6) (7) (8)
Change in Shareholder Wealth 04125E-4 0003 -2439E-4 0003 00021 00012 00024 00012
(057) (464) (031) (468) (1004) (72) (1041) (706)
Lag1 Change in Shareholder Wealth -00062 7856E-4 -0007 9003E-4 -00014 00002 -00014 00003
(594) (098) (592) (112) (522) (123) (5) (145)
Lag1 Institutional Concentration -2475087 555134 -2734983 943317
(3245) (710) (4965) (1671)
Lag1 Total Institutional Ownership 1020109 217488 936337 252464
(1904) (432) (2471) (704)
Lag1 Top 5 Institutional Ownership -148355 31354 -87926 423166
(139) (332) (118) (643)
Lag1 Tobinrsquos Q -195348 -252191 -192344 -25158 -119837 -146373 -122132 -143821
(2812) (3762) (2721) (3752) (248) (274) (2583) (2727)
Lag1 Market Capitalization 844E-06 103E-5 00034 00039
(2781) (3051) (2874) (3104)
Lag1 Log Market Capitalization 750021 723625 776657 732279
(7511) (9285) (11224) (14039)
CEO Dummy 2683978 2702238 2668244 2702417 3281133 3285103 3292175 3283145
(8625) (9277) (8246) (9271) (12739) (13616) (12336) (13605)
Adjusted R-Sq 04604 05302 04187 05295 03856 04588 03476 04574
Number of Observations 36280 36280 36280 36280 97679 97679 97679 97679
31
Table VII Level of Total Direct Compensation as a Function of Institutional Influence
Similar in parts to models (3) and (5) of HS (Table V) this table displays the coefficients from OLS regressions of total direct compensation against the lagged change and
contemporary change in shareholder wealth measures of institutional influence Tobinrsquos Q firm market capitalization and dummy controls for CEO industry and year effects
We estimate models (1) to (4) over the period 1992‒1997 and models (5) to (8) over the whole sample Models (1) (2) (5) and (6) use Institutional Concentration as a measure
of institutional influence which HS define as the shares owned by the institutions with the five largest holdings divided by the shares held by all institutions These models
control for total institutional ownership Models (3) (4) (7) and (8) use Top 5 Ownership as a measure of institutional influence which we define as the proportion of shares
outstanding held by the institutions with the five largest holdings Coefficients for industry and year controls are not presented One two and three asterisks denote significance
at 10 5 and 1 levels
Hartzell and Starks Database 1992‒1997 Full Sample from ExecuComp 1992‒2002
(1) (2) (3) (4) (5) (6) (7) (8)
Change in Shareholder Wealth 005779 007006 005282 006791 00423 00148 0046 00138
(415) (581) (369) (565) (203) (07) (22) (066)
Lag1 Change in Shareholder Wealth 003182 00714 00272 00715 00613 0126 00614 01278
(126) (298) (107) (299) (275) (488) (276) (494)
Lag1 Institutional Concentration -1608633 3248998 -3424892 3334186
(151) (283) (1172) (118)
Lag1 Total Institutional Ownership 104013 5473128 1531013 -2926764
(1645) (831) (876) (158)
Lag1 Top 5 Institutional Ownership 188585 3927986 -2159177 -2640879
(016) (343) (067) (092)
Lag1 Tobinrsquos Q -445692 -803612 -454859 -833566 3257027 2971201 3213936 3086205
(427) (791) (432) (831) (83) (77) (815) (793)
Lag1 Market Capitalization 49E-5 615E-5 01031 01088
(113) (1365) (1555) (168)
Lag1 Log Market Capitalization 4683497 4759377 1645156 1427856
(2775) (3469) (3679) (4239)
CEO Dummy 1501307 1508886 1493688 1508736 4713079 4713629 47278 4703908
(2648) (2684) (2618) (2683) (282) (2828) (2819) (2824)
Adjusted R-Sq 01861 02060 01695 02051 01071 01113 01033 011
Number of Observations 33229 33229 33229 33229 97679 97679 97679 97679
32
Table VIII Two Stage Least Squares (Reverse Causality) Tests - Executive Compensation Measures as a Function of Institutional Concentration
This table shows the coefficients from the Two-Stage Least Squares (2SLS) regressions of executive compensation measures against the lagged change
and contemporary change in shareholder wealth instrumented institutional concentration and total institutional ownership Tobinrsquos Q the natural
logarithm of firm market capitalization and dummy variable controls for CEO and year effects We instrument measures of institutional concentration
and total institutional ownership by orthogonalizing with respect to option grant pay-for-performance sensitivity salary and total compensation
Coefficients for industry and year controls are not presented First stage results are not presented One two and three asterisks denote significance at
10 5 and 1 levels
ExecuComp Sample 1992‒1997 Full Sample from ExecuComp 1992‒2002
PPSOG Salary TC PPSOG Salary TC
Change in Shareholder Wealth 144E-05 0004 02133 -409E-06 00011 00157
(308) (775) (596) (691) (638) (068)
Lag1 Change in Shareholder Wealth -240E-05 -00004 00866 118E-06 00002 01422
(131) (087) (346) (097) (116) (51)
Lag1 Institutional Concentration -00898 668681 3263143 02024 916656 2856417
(052) (801) (176) (163) (1545) (1004)
Lag1 Total Institutional Ownership 01021 -43997 2512587 02983 -104171 -1028347
(09) (084) (153) (4) (269) (487)
Lag1 Tobinrsquos Q 00754 -225606 119297 00655 -158657 2997205
(596) (1649) (053) (834) (2535) (727)
Lag1 Log Market Capitalization -04234 828713 7864046 -04002 825106 1685638
(2195) (8132) (2977) (4106) (11133) (373)
CEO Dummy 15848 3214117 2950274 19149 3205862 4529101
(2223) (9389) (2255) (3802) (1331) (2758)
Overall R-Sq -- 04981 01206 -- 04738 01178
Number of Observations 41305 41305 41305 85518 85518 85518
33
Table IX Two Stage Least Squares (Reverse Causality) Tests - Executive Compensation Measures as a Function of Top 5 Ownership
This table shows the coefficients from 2SLS regressions of executive compensation measures against the lagged change and contemporary change
in shareholder wealth instrumented Top 5 institutional ownership Tobinrsquos Q natural logarithm of firm market capitalization and dummy variable
controls for CEO and year effects We instrument Top 5 institutional ownership by orthogonalizing with respect to option grant pay-for-
performance sensitivity salary and total compensation Coefficients for industry and year controls are not presented First stage results are not
presented One two and three asterisks denote significance at 10 5 and 1 levels
ExecuComp sample 1992‒1997 Full Sample 1992‒2002
PPSOG Salary TC PPSOG Salary TC
Change in Shareholder Wealth 136E-05 00041 02128 -407E-06 00011 00146
(293) (79) (596) (692) (619) (063)
Lag1 Change in Shareholder Wealth -241E-05 -00003 00868 116E-06 00003 01439
(133) (07) (347) (096) (142) (514)
Lag1 Top 5 Institutional Ownership -03774 -187512 -3848531 00179 -104261 -126659
(163) (186) (166) (011) (149) (436)
Lag1 Tobinrsquos Q 00731 -223886 10074 00642 -155467 3115073
(573) (1631) (045) (819) (2509) (75)
Lag1 Log Market Capitalization -0416 775637 7696302 -04015 764632 1462536
(2345) (10084) (3766) (3688) (13708) (4072)
CEO Dummy 15854 3211991 2950158 19152 3203152 4519385
(2223) (9381) (2254) (3804) (133) (2752)
Overall R-Sq -- 04967 01202 -- 04722 01168
Number of Observations 41305 41305 41305 85518 85518 85518
34
Table X Specification Tests utilizing Box-Cox Transformation of Market Capitalization
This table presents results from a Box-Cox transformation of market capitalization In
each model we specify transformation of market capitalization only Panel A shows
results for the period 1992‒1997 and Panel B shows results for the entire period
1992‒2002 The first part of each panel shows the Lambda coefficient that applies to
the transformed market capitalization The second part of each panel shows log
likelihood scores with LR Chi2 Statistics in parentheses ‒ these test for the
appropriateness of a linear or log model specification
Panel A Our ExecuComp sample 1992‒1997
(1) (2) (3)
PPSOG Salary TC
Lambda -05143 01881 02876
(-1914) (2413) (1329)
Test H0 (Log Likelihood Score)
Lambda = -1 -1279748 -3178575 -4795326
(23866) (85952) (99688)
Lambda = 0 -1280152 -3138456 -4791159
(31934) (57147) (16339)
Lambda = 1 -1283157 -3163807 -4793506
(92039) (564156) (63292)
Panel B Full ExecuComp sample 1992‒2002
(1) (2) (4)
PPSOG Salary TC
Lambda -044 01853 04036
(-3015) (40) (3666)
Test H0 (Log Likelihood Score)
Lambda = 0 -2688334 -6476494 -10384490
(66599) (160356) (126236)
Lambda = 1 -2694886 -6538506 -10386795
(197644) (1400605) (172345)
35
Figure 2 Institutional Ownership Measures (LHA) vs Total
Compensation (RHA) by Market Capitalization
0
10
20
30
40
50
60
70
1 2 3 4 5 6 7 8 9 10
$00
$10
$20
$30
$40
$50
$60
$70
$80
$90
Millio
ns
InstitutionalConcentration (LHS)Top5 Ownership (LHS)
Institutional TotalOwnership (LHS)Total Compensation(RHS)
Page 6
6
institutional influence and measures of pay-for-performance sensitivity Section V
documents the relationship between institutional influence and total and fixed
compensation Section VI examines the robustness of our findings and the final Section
concludes
I Review of Hartzell and Starks Methodology
We comment on five areas of the HS methodology 1) Size control 2) Measurement of
institutional influence 3) Testing the Jensen-Murphy measure of incentives 4) Reliance
on incomplete self-reported valued for option grants and 5) Endogeneity
A Size Control
In our paper we argue that HS unintentionally identify incentive-size and pay-size
effects5 not a monitoring effect It can be seen in Figure II that executive total pay for
HSrsquos Top 5 institutional concentration measure rises from as little as $072m for the
smallest market capitalization decile to $85m for the highest This twelve-fold pay-size
effect in the deciles can cause problems if one utilizes an untransformed size measure
Due to skewness in the size measure using this in its untransformed state appears to
weaken its effectiveness leaving open the possibility that correlations they identify
between incentives and pay levels on the one hand and Top 5 Concentration on the other
are due largely to size-related effects
Table III displays this strong pay-size effect based on our full 1992‒2002 dataset and
graphically in Figure 1 It shows that the incentive measure adopted by HS declines from
$272 per $1000 change in shareholder wealth for the smallest market capitalization
decile of firms to approximately $030 for the largest Table III and Figure 1 show the
5 The multiplicative model of Edmans Gabaix and Landier (2009) finds that incentive pay increases with
firm size (elasticity value of 13) and the empirical elasticity is 037
7
strong relation between institutional concentration as measured by HS and market
capitalization For the smallest decile of market capitalization firms institutional
concentration averages 61 percent declining to about 275 percent for the highest decile
This negative relationship between firm size and institutional concentration seems to arise
because institutional investors do generally favor small firms This causes problems
because in the limit a small company may have only one institutional investor owning
only a small percentage of the company Hence it would appear to be fully concentrated
even though the ability of the institutional investor to influence pay decisions might be
small Thus it is not at all surprising that the correlation between the logarithm of market
capitalization and their institutional concentration measure is both negative and high in
absolute value at 63 percent Their methodology appears to favor small companies that
Table III shows to have low pay levels and high pay for performance sensitivity
Moreover if one does not transform the raw size measure it implicitly assumes that a
given dollar change in firm size has the same absolute impact on pay for a very large
company as it does for a small company Indeed the conventional logarithmic
specification for firm size provides proportionate effects that imply a pay-size elasticity of
approximately one-third instead of unity6
B Measurement of Institutional Influence
We argue that there are additional issues related to the HS measure of institutional
influence ndash institutional concentration In their companion article to HS AHS develop a
model where they capture institutional influence by a single large long-term institutional
6 Note that unlike an earlier version of our paper we leave the HS dependent variables in levels and hence
untransformed While we do this to minimize the alterations that we make to the HS specification ideally
the elasticity specification is preferable especially for the compensation variables Transforming the
dependent variables does not significantly affect the results
8
investor They measure influence relative to total shareholding including private or
atomistic shareholders Presumably they do this to capture relative activism and proxy
voting strength However when it comes to the empirical implementation of the model
AHS adopt the same relative institutional concentration measure as in HS (AHS (p12))
potentially downplaying the relative strength of institutions compared with other
shareholders They justify this because there is generally more than one institutional
investor and there will be free-riding problems with multiple investors that only the
largest institutional investors can overcome While we accept their premise of free-rider
problems it is not obvious why the deflator is lsquoinstitutional shareholdingrsquo rather than
lsquototal shareholdingrsquo from this perspective7 If the strength of the size control can be
queried scaling by total shareholding (ie Top 5 Ownership) produces a measure that
remains relatively unaffected across the size deciles It is 20 percent for the smaller
decile declining to 14 percent for the largest decile whereas for the HS concentration
measure it declines by over 50 percent from 61 percent to 28 percent (see Table III and
Figure 1)
C Testing the Jensen-Murphy Measure of Incentives
The testing of their secondary measure of executive pay-for-performance sensitivity
(see HS Table IV) based on Jensen and Murphy (1990) raises issues other than the size
control HS regress changes in compensation (cash and total) on lagged institutional
influence that has been interacted with the contemporaneous change in shareholder
wealth to capture a pay-for-performance relationship Such a test is appealing ndash
7 AHS do point out one benefit of the institutional concentration measure Institutional investors may prefer
companies with lsquobetterrsquo incentive compensation structures Could using institutional shareholding rather
than total shareholding as the deflator alleviate this problem Presumably both the numerator and
denominator will have the same bias and hence error cancellation is a possibility
9
institutions should increase compensation when the firm has performed well Following
Greene (2000 p326) to execute such a test one needs to include in level form variables
that are then interacted Hence the HS model specification ought to be
1 2 3 1 4 1it it it it it itCompn SW IC SW IC (1)
where itCompn is the change in executive compensation itSW is the contemporaneous
change in shareholder wealth which also appears in the interaction term and 1itIC is
lagged institutional ownership concentration However HS (p2363) omit the
contemporaneous term itSW and the lagged term 1itIC which should appear in levels
as well as in the interaction term Most importantly HS interact their lagged institutional
concentration measure 1itIC with the contemporaneous change in shareholder wealth
itSW As it is after all pay-for-performance concentrated institutions need to condition
their influence on some observed measure of firm performance Contemporaneous firm
performance is unknown at the time institutional influence is measured Reflecting this
timing insight leads to a new Greene (2000) specification
1 2 1 3 1 4 1 1it it it it it itCompn SW IC SW IC (2)
such that all the lagged RHS independent variables that now appear in the interaction
term are included on a stand-alone basis
Reliance on incomplete self-reported values for option grants
Prior to receiving the HS dataset we tried to replicate the findings of HS to validate
our datasets for the HS period and the longer period 1992‒2002 While we initially
focused on the same period as HS (ie 1992‒1997) we found that our sample size was
larger at 47765 observations than HSrsquos at 33928 observations This discrepancy could
have arisen if HS did not replace missing and self-reported option valuations using their
10
Black-Scholes option methodology This may introduce a downward pay bias since self-
reported option-grant pay is subject to possible understatement and there are 13839
observations with likely high option pay dropped altogether
Endogeneity
HSrsquos endogeneity testing raises concerns about reverse causality Institutions may
prefer to hold firms with certain compensation structures such as incentivized executives
HS rightly highlight this concern and use instrumental variables to address this issue HS
assume that institutional ownership reflects any reverse causality Using an instrumental
variables approach to resolve this problem involves finding a variable that influences or
explains institutional ownership but not directly executive compensation They use share
turnover in a firm as an instrument for institutional ownership However theoretical work
by Holmstrom and Tirole (1993) indicates that measures of stock liquidity and trading
activity are indeed important determinants of executive market-based incentives as Kang
and Lui (2008) and Dikolli Kulp and Sedatole (2009) find empirically As such one can
make a case for adopting another instrument rather than share turnover for institutional
ownership
II Previous Research
Prior research fails to address issues concerning possible fragility of the HS results
AHS replicate the HS findings on both incentives and pay levels using the same dataset as
HS and they apply the same definition of institutional concentration and use market
capitalization expressed as a level Cadman Klasa and Matsunaga (2010 Table 6) set out
to replicate HSrsquos (Table II) model of option grant pay-for-performance sensitivity when
they compare a dataset of ExecuComp and Non-ExecuComp firms for the period 2000-
2007 Their ExecuComp results for this sample period are quite similar to our results for
11
the 1992‒2002 period for the same specification controlling for the logarithm of market
capitalization but they do not regard these results as a test of HS as their focus is entirely
on comparing the two groups of stocks Nor do they investigate the impact of
concentration on either cash compensation or total pay Kang and Liu (2008) examine
pay-performance sensitivity use a version of the HS institutional concentration measure
as a control variable in a different framework deriving from Holmstrom and Tirole
(1993) Dikolli Kulp and Sedatole (2009) argue that firms with transient ownership will
issue relatively more equity incentives to counter short-termism with their results robust
to the inclusion of the HS concentration variable
We hoped to find supportive evidence from proxy voting behavior in support of HSs
findings However evidence from actual mutual fund proxy voting records suggests that
most oppose reductions in CEO pay but overall the evidence is mixed8 Davis and Kim
(2005) examine proxy votes of mutual funds for 2004 and find that for a sample of
Fortune 1000 proxy contests in 45 cases that attempted to limit executive pay mutual
funds were unanimously opposed to such attempts Rothberg and Lilien (2006) also
examine proxy-voting data and find that mutual funds voted 66 percent of the time in
managementrsquos favor on issues of executive compensation Ertimur Ferri and Muslu
(2011) find that institutional investors lack sophistication in that they do not target
excessively paid CEOs relative to CEOs with high predicted-pay based on economic
determinants but exert a moderating influence ndash a $23 million reduction ndash for those they
deem ldquoexcessively paidrdquo CEOs
III Data and Methodology
8 We do acknowledge that the HS results are for all institutional investors while the transparency with
respect to proxy voting applies only to mutual funds
12
This Section discusses the data used in the study the construction of variables
descriptive statistics and the methodology Two datasets are used ndash one is the original
dataset generously supplied by Jay Hartzell and Laura Starks which covers 1992‒1997
second a more comprehensive dataset covering 1992‒2002 is computed from the same
data sources used in the HS dataset This is Standard and Poorrsquos (SampP) ExecuComp for
compensation data Thomson Financial CDASpectrum database for institutional holdings
to compute measure of institutional influence SampPrsquos Compustat and the Center for
Research in Stock Prices (CRSP) for firm level accounting and stock price data Our more
comprehensive dataset contains 2559 firms compared with 1914 for the HS dataset Our
full matched sample yields 97679 executive years
A Institutional Influence Measures
To measure institutional influence we employ two ownership-based metrics The first
measure of institutional influence is HSrsquos Institutional Concentration We use this for
direct comparison to the HS analysis The second measure of institutional influence
changes the denominator of the HS measure from institutional shares held to total shares
outstanding in the firm with nomenclature Top 5 Ownership9 We compute both
measures at the fiscal year end of the firm We adjust all compensation measures and
asset prices for inflation to 30 June 2002 prices for our comprehensive sample compared
to the HS dataset that is in the dollars of the day
B Compensation Measures
The ExecuComp database reports several raw measures of executive compensation for
the top five executives (where available) including total salary bonus long-term
9 Top 5 Ownership is included in the HS database but not utilized by them in reported tests as well as
constructed by us for the more comprehensive sample
13
incentives and number of option grants For our more comprehensive dataset we replicate
all the compensation measures used by HS In common with HS we compute the pay-for-
performance sensitivity of option grants using the Yermack (1995) approach This
measure shows the dollar change in value of the executive options for every dollar change
in the value of the firm This involves calculating the option delta using the Black-
Scholes formula adjusted for dividends10
The ratio of the number of shares represented
by the award of options divided by the diluted number of shares outstanding at the
beginning of the year multiplies the option delta with units given by a $1000 change in
firm value Summing salary bonus long-term incentive plan payouts and stock and
option grants provides total direct compensation Options are valued using the Black and
Scholes (1973) formula for European call options adjusted by Merton (1973) to
incorporate dividends
C Descriptive Statistics
Table I presents descriptive statistics of key variables employed in the study according
to our comprehensive dataset11
Institutions own on average 546 percent of the firms in
which they invest This is almost identical to the 531 percent found by HS The
institutions with the five largest holdings account 45 percent of institutional ownership
Once again the figure for HS is almost identical at 44 percent The institutions with the
five largest holdings own on average 229 percent of shares outstanding in a firm (HS
have 223 percent) Average total compensation for executives is $2583 million
considerably larger than that documented by HS ($1250 million) but the HS figure is an
10 When there is more than one grant during the year a weighted average delta is used
11 With the data supplied to us by Hartzell and Starks we are able to replicate exactly their summary data
(HS Table I) for all but cash compensation Salary is included but bonus payments are missing The
replicated table is available on request
14
amalgam of price levels over their sample period 1992‒1997 and ours are in consistent
dollars of June 30 2002 However there has also been significant growth in
compensation levels from 1998 to 2002
ltltINSERT TABLE Igtgt
Based on our comprehensive data in Table II we examine correlations between the
key variables employed in the study Of concern is the large negative correlation of 63
percent which has been discussed previously (Section IIA) between the HS measure of
institutional concentration and the natural logarithm of firm market capitalization
ltltINSERT TABLE IIgtgt
In view of this high correlation Table III provides a breakdown of the three key pay
variables and three institutional ownership variables by size deciles We find that the high
institutional concentration but low institutional ownership share of total shareholdings
mentioned above is evident in the smallest decile of firms by size This decile has low
salary and even lower total compensation in a relative sense but by far the highest pay-
for-performance sensitivity of option grants Small firms award executives a far higher
proportion of market capitalization Since small firm market-capitalization is negligible in
comparison with the highest decile (300 times larger in magnitude) total compensation is
very small in absolute value Moreover due to the domination of the pay-size effect base
salaries are also exceedingly small
ltltINSERT TABLE IIIgtgt
Figure 1 based on Table III shows how as institutional concentration falls with
increases in stock size (but also as the total share of institutional ownership increases)
and the pay-for-performance sensitivity of option grants falls Thus because of what
appears to be a weak size control and size-influenced concentration measure HS
15
potentially pick up a size effect as their main result but not necessarily a monitoring
effect Figure 2 shows that as firm size increases up to decile 8 at least the share of
institutional total ownership increases as does total compensation
ltltINSERT FIGURES 1 AND 2 gtgt
D Methodology
Following HS we use two estimation techniques to examine the effect of institutional
investor influence on executive compensation Due to the censored nature of option grant
pay-for-performance sensitivity this compensation measure most appropriately uses
Tobit model specification whereas an Ordinary Least Squares (OLS) model specification
is used for all other compensation measures We model compensation measures as a
function of contemporaneous and lagged change in shareholder wealth institutional
influence Tobinrsquos Q and market capitalization Controls for CEO industry and year
effects are also used Augmentation occurs to test the effect of institutional influence on
the change in executive cash and total compensation with an interaction term that is the
product of institutional influence and the change in shareholder wealth
The variables used to model executive compensation have the following definitions
The lagged and contemporary measure of change in shareholder wealth is simply the
difference in market capitalization of the firm The measures of institutional influence
used are Institutional Concentration and Top 5 Ownership Use of the two types of
institutional influence occurs in separate models In order to replicate the HS
methodology models using Institutional Concentration have an additional control
consisting of lagged total institutional holdings and Tobinrsquos Q controls for performance
and growth opportunities Market capitalization of the firm is the number of shares
outstanding multiplied by the share price with inclusion both in levels and natural
16
logarithmic form A dummy variable is included to identify whether the executive is the
CEO Dummy variables are also included to control for industry effects This variable
takes a value of 1 for the two-digit Standard Industrial Classification (SIC) in which the
firm operates A year fixed effects (dummy) variable is included taking a value of 1 if the
observation was from the given year
IV Executive Incentives as a Function of Institutional Investor Influence
This section presents the results for the effect of institutional influence on option grant
pay-for-performance sensitivity and changes in compensation
A Option Grant Pay-for-Performance Sensitivity
Table IV presents the results for the effect of institutional influence on option grant
pay-for-performance sensitivity
ltltINSERT TABLE IVgtgt
Model (1) is a direct replication of HS using the HS database as supplied to us by HS
with all the variables computed in the same way and in the same functional form as with
HS The coefficient and t-statistic on lagged institutional concentration interacted with the
change in shareholder wealth is 13994 and 767 that compares to 1358 and 842 in HS
We are unable to account for these minor discrepancies for values that should be
identical In model (2) the only change to the model is that we use the natural logarithm
of market capitalization instead of the level of market capitalization This single change
has a sizeable effect on the coefficient on institutional concentration The coefficient on
institutional concentration is now -00692 and is insignificant The fact that an order
preserving transformation to a control variable can change the results is of interest The
log-transformation to market capitalization also changes the coefficient on Top 5
Ownership in models (3) and (4) The coefficient on Top 5 Ownership is significant in
17
both model (3) and in model (4) but the better size control reduces the coefficient by over
25 percent The better size control does not affect Top 5 Ownership by as much as its
design ensures less sensitivity to what could be inadequate size controls
We also test the effect of institutional influence on option grant pay-for-performance
sensitivity over the period 1992‒2002 This sample is then approximately three times as
large as the sample used by HS Institutional concentration is highly positively significant
in model (5) similar to the results for the period 1992‒1997 When we use the log of
market capitalization in model (6) the coefficient on institutional concentration remains
positive and significant at the 10 percent level with a t-statistic of 166 Of importance
though is that the coefficient on institutional concentration is one-eighth of the coefficient
in model (5) Even for a measure of institutional influence which we believe is less
sensitive to firm size the use of log market capitalization reduces the coefficient on Top 5
Ownership by approximately 50 percent from model (7) to model (8)
B Pay-for-Performance Sensitivity due to Changes in Cash and Total
Compensation
For robustness HS examine the effect that institutional influence conditioned on firm
performance has on a changes in cash and total executive compensation using the general
method of Jensen and Murphy (1990) to compute pay-for-performance sensitivity
ltltINSERT TABLE V gtgt
Table V presents the effect of institutional influence on the change in cash
compensation using the HS size control measure throughout12
Model (1) replicates the
specification in HS for the period of 1992‒1997 and we produce a similar result ndash the
interaction of lagged institutional concentration and contemporaneous change in
12 This test uses our data for the two periods 1992‒1997 and 1992‒2002 We do not use the HS data for
cash compensation as data for bonus is missing in the data supplied to us
18
shareholder wealth positively and significantly relates to a change in executive cash
compensation In model (2) we include the contemporaneous change in shareholder
wealth and the lagged institutional concentration as stand-alone controls as indicated by
Greenersquos (2000) specification of equation (1) above and the interaction term remains
significant In model (3) we interact lagged institutional concentration with the lagged
change in shareholder wealth as in equation (2) above that tries to take account of what
participants could have observed at the time The interaction term is then negative but
insignificant altering the HS result Model (4) provides a similar result where our
preferred proxy Top 5 Ownership captures concentrated institutional influence In
models (5) to (8) we use our full sample In model (5) we find that the lagged
institutional concentration interaction with the contemporaneous change in shareholder
wealth is positive and significant in HSrsquos specification However including change in
contemporaneous shareholder wealth and lagged institutional concentration as separate
control variables as suggested by equation (1)rsquos specification removes the significance
from the interaction term in model (6) Moreover in model (7) when lagged institutional
concentration is interaction with the lagged change in shareholder wealth in accord with
equation (2) this interaction term is now negative and significant reversing the HS
finding We obtain a similar result in model (8) when we measure institutional
concentration influence using Top 5 Ownership13
V The Levels of Executive Fixed and Total Compensation as Functions of
Institutional Investor Influence
13 In unreported results we use the log of market capitalization as a size control These changes to the size
control do not dramatically change the findings from Table V The reason why this size issue is not as
important as in the findings for option grants pay-for-performance is that the change in cash compensation
is not so strongly influenced by firm size In unreported results available on request we also examine the
effect of institutional influence on changes in total compensation with similar results to the change in cash
compensation
19
Two measures of compensation are examined ndash salary and total compensation
ltltINSERT TABLE VI gtgt
Table VI examines the effect of institutional influence on the level of salary In model
(1) we use the HS sample and specification to obtain almost identical results The
coefficient on institutional concentration is (negative) $247509 that is almost identical to
(negative) $247967 in HS Both are highly significant In model (2) we make a single
change to the model ndash we use the natural logarithm of market capitalization This single
change reverses the sign of institutional concentration with the coefficient still being
significant at the 1 percent level In models (3) and (4) we measure institutional influence
using Top 5 Ownership In model (3) where we measure firm size using the level of
market capitalization Top 5 Ownership is insignificant When we use the natural
logarithm of market capitalization in model (4) Top 5 Ownership is significant and
positively relates to the salary level In models (5) to (8) we estimate the same models
using our full sample We find similar results for the full sample as for the period 1992ndash
1997
ltltINSERT TABLE VII gtgt
Table VII examines the effect of institutional influence on total compensation In
model (1) we replicate the HS methodology using the HS data to obtain a similar
significant negative relation between institutional concentration and total compensation of
(negative) $1608633 However after changing to a logarithmic specification of market
capitalization in model (2) institutional concentration increases total compensation by a
significant $32400014
Importantly when institutional influence is measured using Top 5
14 = (5688751 times 0101)
20
Ownership an insignificant relation exists with total compensation in model (3) using the
HS size control but it becomes both positive and significant in (4) with the logged size
control In models (5) and (6) where models are estimated using the full sample we
observe a similar reversal in sign as occurs in models (1) and (2) However now the
magnitude of the reversal is almost equal with a difference of about $67m that is very
large indeed In model (5) the coefficient on institutional concentration is (negative)
$342m and then in model (6) the coefficient is $33m Top 5 Ownership is insignificant
in both models (7) and (8) unlike the results for Salary
VI Robustness
Two issues are examined in is this Section ndash reverse causality and the appropriate
transformation of market capitalization A methodological issue inherent in examining the
effect of ownership-based measures of institutional influence on executive compensation
is endogeneity driven by reverse causality We utilize an instrumental variables approach
where orthogonalized institutional influence measures are regressed on each of the
compensation measures lagged option grant pay-for-performance sensitivity salary and
total compensation In addition market capitalization and a CEO dummy are included in
each regression We use the residual from each regression as an instrument for
institutional influence Such a process removes any preferences for compensation from
the institutional influence measures
ltltINSERT TABLES VIII AND IXgtgt
Table VIII presents the results where we instrument the HS measure lagged
institutional concentration with a compensation-orthogonalized equivalent The analysis
indicates that the instrumented institutional concentration measure is positively and
significantly related to the level of salary and total compensation and insignificantly
21
related to option grant pay-for-performance sensitivity (PPSOG) Table IX presents
results where we instrument Top 5 Ownership with a compensation-orthogonalized
equivalent The instrumented Top 5 Ownership measure insignificantly relates to option
grant pay-for-performance sensitivity and barely negatively relates to salary although it
negatively relates to total compensation suggesting reverse causality is driving a
significant portion of our previous results
ltltINSERT TABLE Xgtgt
Another issue examined in this section is the appropriate transformation of market
capitalization Aneuryn-Evans and Deaton (1980) provide an evaluation of methods
comparing linear and log-linear specifications We use the Box-Cox test to compare
logarithm versus linear specifications In applying the Box-Cox test we specify
transformation of only market capitalization ndash the dependent variable and all other
independent variables remain unchanged Table X presents the lambda coefficient for
each model and the log likelihood score In all models in Panel A and B a lambda of zero
receives a higher (closer to zero) log likelihood score suggesting the log specification is
appropriate
VII Conclusion
Hartzell and Starks (2003) appears to document an effective role for institutional
investors in increasing executive incentives and lowering compensation levels Our paper
re-addresses the issue of concentrated institutional monitoring from a perspective
informed by an additional decade of research since their paper appeared We cannot find
evidence that concentrated institutional investors really have any positive (or negative for
that matter) effect on executive incentive levels While superficially we find evidence
that they may raise executive pay levels this is most likely driven by an institutional
22
preference for highly remunerated executives in large companies and thus by reverse
causality Moreover our findings are not consistent with the rent-extraction hypothesis
suggesting that firms systematically overpay executives in the absence of effective
monitoring by outside parties such as institutional investors
23
References
Almazan Andres Jay C Hartzell and Laura T Starks 2005 ldquoActive Institutional
Shareholders and Costs of Monitoring Evidence from Executive Compensationrdquo
Financial Management Winter 5ndash34
Aneuryn-Evans Gwyn and Angus Deaton 1980 ldquoTesting Linear versus Logarithmic
Regression Modelsrdquo Review of Economic Studies 47 (1) Econometrics Issue 275ndash
291
Babenko Ilona 2009 ldquoShare Repurchases and Pay-Performance Sensitivity of Employee
Compensation Contractsrdquo Journal of Finance 64117-150
Baker George P and Brian J Hall 2004 ldquoCEO Incentives and Firm Sizerdquo Journal of
Labor Market Economics 22 767ndash798
Bebchuk Lucian A Jesse M Fried and David Walker 2002 ldquoManagerial power and
rent extraction in the design of executive compensationrdquo University of Chicago Law
Review 69 751ndash846
Bebchuk Lucian A and Jesse M Fried 2003 ldquoExecutive Compensation as an Agency
Problemrdquo Journal of Economic Perspectives 17 71ndash92
Bebchuk Lucian A and Jesse M Fried 2004 Pay without Performance The Unfulfilled
Promise of Executive Compensation Cambridge MA Harvard University Press
Bertrand Marianne and Sendhil Mullainathan 2001 ldquoAre CEOs Rewarded for Luck
The Ones Without Principals Arerdquo Quarterly Journal of Economics 116 901ndash32
Black Fischer and Myron Scholes 1973 ldquoThe Pricing of Options and Corporate
Liabilitiesrdquo Journal of Political Economy 81 637ndash54
Cadman Brian Sandy Klasa and Steve Matsunaga 2010 ldquoDeterminants of CEO Pay A
Comparison of ExecuComp and Non-ExecuComp Firmsrdquo The Accounting Review 85
1511ndash1543
Chhaochharia Vidhi and Yaniv Grinstein 2009 ldquoCEO Compensation and Board
Structurerdquo Journal of Finance 64 231ndash261
Cichello Michael S 2005 ldquoThe impact of firm size on pay-performance sensitivitiesrdquo
Journal of Corporate Finance 11 609ndash627
Davis Gerald F and E Han Kim 2007 ldquoBusiness ties and proxy voting by mutual
fundsrdquo Journal of Financial Economics 85 552ndash570
Dikolli Shane S Susan L Kulp and Karen L Sedatole 2009 ldquoTransient Institutional
Ownership and CEO Contractingrdquo Accounting Review 84 737ndash770
Edmans Alex Xavier Gabaix and Augustin Landier 2009 ldquoA Multiplicative Model of
Optimal CEO Incentives in Market Equilibriumrdquo Review of Financial Studies 22
4881ndash4917
Edmans Alex 2009 ldquoBlockholder Trading Market Efficiency and Managerial Myopiardquo
Journal of Finance 64 2481ndash2513
Ertimur Yonca Fabrizio Ferri and Volkan Muslu 2011 ldquoShareholder Activism and
CEO Payrdquo Review of Financial Studies 24 535ndash592
24
Gabaix Xavier and Augustin Landier 2008 ldquoWhy Has CEO Pay Increased So Muchrdquo
Quarterly Journal of Economics 123 49ndash100
Garen John 1994 ldquoExecutive compensation and principal-agent theoryrdquo Journal of
Political Economy 102 1175ndash1199
Gibbons Robert and Kevin J Murphy 1992 ldquoOptimal Incentive Contracts in the
Presence of Career Concerns Theory and Evidencerdquo Journal of Political Economy
100 468ndash505
Greene William H 2000 (4th
Edition) Econometric Analysis Prentice-Hall NJ
Guthrie Katherine Jan Sokolowsky and Kam-Ming Wan 2012 ldquoCEO Compensation
and Board Structure Revisitedrdquo Journal of Finance Forthcoming
Hadlock Charles R and Gerald B Lumer 1997 ldquoCompensation turnover and top
management incentives historical evidencerdquo Journal of Business 70 153ndash187
Hartzell Jay C and Laura T Starks 2003 ldquoInstitutional Investors and Executive
Compensationrdquo Journal of Finance 58 2351ndash2374
Huson Mark R Robert Parrino and Laura T Starks 2001 ldquoInternal Monitoring
Mechanisms and CEO Turnover A Long-Term Perspectiverdquo Journal of Finance 56
2265ndash2297
Jensen Michael C and Kevin J Murphy 1990 ldquoPerformance Pay and Top Management
Incentivesrdquo Journal of Political Economy 98 225ndash264
Jin Li 2002 ldquoCEO compensation diversification and incentivesrdquo Journal of Financial
Economics 66 29ndash63
Kang Qiang and Qiao Liu 2008 ldquoStock Market Information Production and Executive
Incentivesrdquo Journal of Corporate Finance 14 484ndash498
Merton Robert 1973 ldquoTheory of Rational Option Pricingrdquo Bell Journal of Economics
and Management Science 4 141-183
Morse Adair Vikram Nanda and Amit Seru 2011 ldquoAre Incentive Contracts Rigged by
Powerful CEOsrdquo Journal of Finance 66 1779ndash1821
Murphy Kevin J 1985 ldquoCorporate performance and managerial remuneration An
empirical analysisrdquo Journal of Accounting and Economics 7 11ndash42
Murphy Kevin J 1999 ldquoExecutive compensationrdquo in Orley Ashenfelter and David
Card Handbook of Labor Economics vol 3 Elsevier
Rosen Sherwin 1992 ldquoContracts and the market for executivesrdquo In Contract
Economics ed Lars Werin and Hans Wijkander 181ndash211 Cambridge MA
Blackwell
Rothberg Burton and Steven Lilien 2006 ldquoMutual Funds and Proxy Voting New
Evidence on Corporate Governancerdquo Journal of Business and Technology Law 1
157ndash184
Schaefer Scott 1998 ldquoThe dependence of payndashperformance sensitivity on the size of the
firmrdquo Review of Economics and Statistics 80 436ndash443
Yermack David 1995 ldquoDo Corporations Award CEO Stock Options Effectivelyrdquo
Journal of Financial Economics 39 237ndash69
25
Table I Univariate Descriptive Statistics
This table reports sample statistics for the key variables used in this study based on our comprehensive
ExecuComp dataset Institutional Concentration is the shares owned by the institutions with the five largest
holdings divided by the shares held by all institutions Top 5 Ownership is the proportion of shares
outstanding held by the institutions with the five largest holdings Total Institutional Ownership is the
proportion of shares held by all institutions Panel B shows the measures executive compensation The pay-
for-performance sensitivity (PPS) of option grants shows the change in executive wealth due to a $1000
change in firm in shareholder wealth Expressions for Salary and Total Compensation are in thousands of
dollars while Market Capitalization is in millions of dollars Panel C shows firm characteristics
Variable Mean Median Std Dev P10 P90
Panel A Institutional Investor Holdings
Institutional Concentration 0450 0424 0151 0282 0656
Top 5 Institutional Ownership 0229 0220 0101 0110 0355
Total Institutional Ownership 0546 0561 0206 0258 0801
Panel B Executive Compensation
PPS of Option Grants 1161 0156 3829 0 2822
Salary 36311 29462 25231 15179 67000
Total Compensation 258313 85878 980701 23875 511022
Panel C Firm Characteristics
Tobins Q 1744 1139 2738 0495 3322
Market Capitalization 560699 112094 1988108 19754 1048754
26
Table II Pairwise Correlation Coefficients
This table presents the correlations between the key variables based on our comprehensive
ExecuComp dataset Institutional Concentration is the shares owned by the institutions with the five
largest holdings divided by the shares held by all institutions Top 5 Ownership is the proportion of
shares outstanding held by the institutions with the five largest holdings Total Institutional
Ownership is the proportion of shares held by all institutions The pay-for-performance sensitivity
(PPS) of option grants is the change in executive wealth due to a $1000 change in shareholder
wealth
(1) (2) (3) (4) (5) (6) (7) (8)
(1) Institutional Concentration 1
(2) Top 5 Ownership 0268 1
(3) Total Ownership -0415 0706 1
(4) PPS of Option Grants 0122 0042 -0024 1
(5) Log Salary -0270 -0012 0171 0038 1
(6) Log Market Capitalization -0630 -0157 0261 -0156 0420 1
(7) Level Market Capitalization -0241 -0157 0007 -0058 0219 0534 1
(8) Tobins Q -0008 -0034 0002 0015 -0083 0066 0121 1
27
Table III
Descriptive Statistics by Market Capitalization Deciles
Market
Capitalization
Deciles
Market
Capitalization
($m)
Total
Compensation
($000)
Salary
($)
PPS Option
Grants
($)
Institutional
Concentration
()
Top 5 Ownership
()
Institutional
Total Ownership
()
1 (Small) 1181 76008 234700 2720 611 200 327
2 2668 90206 257955 1796 495 217 438
3 4193 105946 275120 1517 454 220 485
4 6045 129892 296067 1376 420 206 491
5 8742 155391 315864 1216 394 198 502
6 13086 191888 345942 1072 380 202 531
7 20790 239501 377021 0874 368 196 532
8 34920 314955 427808 0687 354 190 538
9 67747 450970 480319 0569 328 173 528
10 (Big) 366540 846066 614354 0297 275 138 502
28
Table IV Pay-for-Performance Sensitivity of Option Grants as a Function of Institutional Influence
Similar in construction to parts of HS (Table II) this table shows the coefficients from Tobit regressions of option grant pay-for-performance sensitivity per $1000 change in
shareholder wealth against the lagged change and contemporary change in shareholder wealth measures of institutional influence Tobinrsquos Q firm market capitalization and
dummy controls for CEO and year effects We estimate models (1) to (4) over the period 1992‒1997 using data supplied by HS and models (5) to (8) are estimated over the
whole sample Models (1) (2) (5) and (6) use Institutional Concentration as a measure of institutional influence which HS define as the shares owned by the institutions
with the five largest holdings divided by the shares held by all institutions These models control for total institutional ownership Models (3) (4) (7) and (8) use Top 5
Ownership as a measure of institutional influence which we define as the proportion of shares outstanding held by the institutions with the five largest holdings Coefficients
for industry and year controls are not presented One two and three asterisks denote significance at 10 5 and 1 levels
Hartzell and Starks Database 1992‒1997 Full Sample from ExecuComp 1992‒2002
(1) (2) (3) (4) (5) (6) (7) (8)
Change in Shareholder Wealth 77721 191781 89874 137368 -446 -218 -664 -203
(171) (429) (194) (317) (319) (168) (453) (158)
Lag1 Change in Shareholder Wealth 14446 99963 88719 8544 -144 0273 -0342 00852
(027) (212) (166) (184) (071) (015) (016) (005)
Lag1 Institutional Concentration 13994 -00692 26405 03304
(767) (032) (1402) (166)
Lag1 Total Institutional Ownership 08055 10434 07106 0937
(593) (781) (666) (898)
Lag1 Top 5 Institutional Ownership 10567 07457 12116 06293
(399) ( 283) (533) (278)
Lag1 Tobinrsquos Q 0068 00877 00647 00813 00498 00727 00508 00695
(337) (434) (319) (401) (504) (734) (51) (703)
Lag1 Market Capitalization -106202 -173515 -776 -117
(474) (742) (1011) (153)
Lag1 Log Market Capitalization -0297 -02319 -04111 -03908
(1576) (1439) (2997) (2841)
CEO Dummy 13745 13682 13765 13674 28197 28128 28155 28143
(1913) (1918) (1914) (1916) (3388) (3396) (3377) (34)
Number of Observations 33878 33878 33878 33878 97679 97679 97679 97679
29
Table V Pay-for-Performance Sensitivity of Cash Compensation (Salary plus Bonus) as a Function of Institutional Influence - Level Market Capitalization
Similar in parts to models (1) and (2) of HS (Table IV) this table displays the coefficients from OLS regressions of the change in cash compensation against the lagged
change and contemporary change in shareholder wealth measures of institutional influence Tobinrsquos Q firm market capitalization and dummy controls for CEO and year
effects We estimate models (1) to (4) over the period 1992‒1997 based on our ExecuComp sample and models (5) to (8) over the whole sample Models (1) (2) (5) and (6)
use Institutional Concentration as a measure of institutional influence which HS define as the shares owned by the institutions with the five largest holdings divided by the
shares held by all institutions These models control for total institutional ownership Models (4) and (8) use Top 5 Ownership as a measure of institutional influence which
we define as the proportion of shares outstanding held by the institutions with the 5 largest holdings Coefficients for industry and year controls are not presented One two
and three asterisks denote significance at 10 5 and 1 levels
Our Database based on ExecuComp 1992‒1997 Full Sample based on ExecuComp 1992‒2002
(1) (2) (3) (4) (5) (6) (7) (8)
Lag1 Change in Shareholder Wealth 00014 00017 00088 00089 -00006 -00007 00072 00086
(051) (062) (129) (103) (027) (029) (131) (16)
Change in Shareholder Wealth -00102 00102 00105 00032 00102 00101
(097) (292) (293) (055) (611) (608)
Change Shareholder Wealth Lag1 Institutional
Concentration
00435 00764 00363 00253
(463) (258) (689) (144)
Lag1 Change Shareholder Wealth Lag1
Institutional Concentration
-00197 -00283
(122) (197)
Lag1 Institutional Concentration -492182 -288861 400149 453623
(09) (056) (098) (114)
Lag1 Total Institutional Ownership 720688 538372 655091 559587 682069 726226
(276) (232) (281) (314) (308) (322)
Lag1 Change Shareholder Wealth Lag1 Top 5
Institutional Ownership
-0046 -00689
(091) (256)
Lag1 Top 5 Institutional Ownership 583044 1191634
(092) (25)
Lag1 Tobinrsquos Q -32761 -36806 -19422 -18023 -54283 -48681 -29534 -24243
(138) (172) (077) (078) (227) (245) (136) (111)
Lag1 Market Capitalization 00004 00004 00002 00006 00007 00008 00007 00007
(042) (044) (022) (07) (086) (092) (082) (09)
CEO Dummy 768035 765474 768334 773634 616315 61729 619028 618535
(354) (355) (356) (355) (417) (418) (419) (418)
Adjusted R-Sq 00057 0006 00052 0005 00097 00097 00098 00102
Number of Observations 41305 41305 41305 41305 85518 85518 85518 85518
30
Table VI Level of Salary as a Function of Institutional Influence
Similar in parts to models (1) and (2) of HS (Table V) this table displays the coefficients from OLS regressions of salary against the lagged change and contemporary change in
shareholder wealth measures of institutional influence Tobinrsquos Q firm market capitalization and dummy controls for CEO industry and year effects We estimate models (1) to
(4) over the period 1992‒1997 using the actual data supplied by HS and models (5) to (8) over the whole sample 1992‒2002 Models (1) (2) (5) and (6) use Institutional
Concentration as a measure of institutional influence which HS define as the shares owned by the institutions with the five largest holdings divided by the shares held by all
institutions These models control for total institutional ownership Models (3) (4) (7) and (8) use Top 5 Ownership as a measure of institutional influence which we define as
the proportion of shares outstanding held by the institutions with the 5 largest holdings Coefficients for industry and year controls are not presented One two and three
asterisks denote significance at 10 5 and 1 levels
Hartzell and Starks Database 1992‒1997 Full Sample based on ExecuComp 1992‒2002
(1) (2) (3) (4) (5) (6) (7) (8)
Change in Shareholder Wealth 04125E-4 0003 -2439E-4 0003 00021 00012 00024 00012
(057) (464) (031) (468) (1004) (72) (1041) (706)
Lag1 Change in Shareholder Wealth -00062 7856E-4 -0007 9003E-4 -00014 00002 -00014 00003
(594) (098) (592) (112) (522) (123) (5) (145)
Lag1 Institutional Concentration -2475087 555134 -2734983 943317
(3245) (710) (4965) (1671)
Lag1 Total Institutional Ownership 1020109 217488 936337 252464
(1904) (432) (2471) (704)
Lag1 Top 5 Institutional Ownership -148355 31354 -87926 423166
(139) (332) (118) (643)
Lag1 Tobinrsquos Q -195348 -252191 -192344 -25158 -119837 -146373 -122132 -143821
(2812) (3762) (2721) (3752) (248) (274) (2583) (2727)
Lag1 Market Capitalization 844E-06 103E-5 00034 00039
(2781) (3051) (2874) (3104)
Lag1 Log Market Capitalization 750021 723625 776657 732279
(7511) (9285) (11224) (14039)
CEO Dummy 2683978 2702238 2668244 2702417 3281133 3285103 3292175 3283145
(8625) (9277) (8246) (9271) (12739) (13616) (12336) (13605)
Adjusted R-Sq 04604 05302 04187 05295 03856 04588 03476 04574
Number of Observations 36280 36280 36280 36280 97679 97679 97679 97679
31
Table VII Level of Total Direct Compensation as a Function of Institutional Influence
Similar in parts to models (3) and (5) of HS (Table V) this table displays the coefficients from OLS regressions of total direct compensation against the lagged change and
contemporary change in shareholder wealth measures of institutional influence Tobinrsquos Q firm market capitalization and dummy controls for CEO industry and year effects
We estimate models (1) to (4) over the period 1992‒1997 and models (5) to (8) over the whole sample Models (1) (2) (5) and (6) use Institutional Concentration as a measure
of institutional influence which HS define as the shares owned by the institutions with the five largest holdings divided by the shares held by all institutions These models
control for total institutional ownership Models (3) (4) (7) and (8) use Top 5 Ownership as a measure of institutional influence which we define as the proportion of shares
outstanding held by the institutions with the five largest holdings Coefficients for industry and year controls are not presented One two and three asterisks denote significance
at 10 5 and 1 levels
Hartzell and Starks Database 1992‒1997 Full Sample from ExecuComp 1992‒2002
(1) (2) (3) (4) (5) (6) (7) (8)
Change in Shareholder Wealth 005779 007006 005282 006791 00423 00148 0046 00138
(415) (581) (369) (565) (203) (07) (22) (066)
Lag1 Change in Shareholder Wealth 003182 00714 00272 00715 00613 0126 00614 01278
(126) (298) (107) (299) (275) (488) (276) (494)
Lag1 Institutional Concentration -1608633 3248998 -3424892 3334186
(151) (283) (1172) (118)
Lag1 Total Institutional Ownership 104013 5473128 1531013 -2926764
(1645) (831) (876) (158)
Lag1 Top 5 Institutional Ownership 188585 3927986 -2159177 -2640879
(016) (343) (067) (092)
Lag1 Tobinrsquos Q -445692 -803612 -454859 -833566 3257027 2971201 3213936 3086205
(427) (791) (432) (831) (83) (77) (815) (793)
Lag1 Market Capitalization 49E-5 615E-5 01031 01088
(113) (1365) (1555) (168)
Lag1 Log Market Capitalization 4683497 4759377 1645156 1427856
(2775) (3469) (3679) (4239)
CEO Dummy 1501307 1508886 1493688 1508736 4713079 4713629 47278 4703908
(2648) (2684) (2618) (2683) (282) (2828) (2819) (2824)
Adjusted R-Sq 01861 02060 01695 02051 01071 01113 01033 011
Number of Observations 33229 33229 33229 33229 97679 97679 97679 97679
32
Table VIII Two Stage Least Squares (Reverse Causality) Tests - Executive Compensation Measures as a Function of Institutional Concentration
This table shows the coefficients from the Two-Stage Least Squares (2SLS) regressions of executive compensation measures against the lagged change
and contemporary change in shareholder wealth instrumented institutional concentration and total institutional ownership Tobinrsquos Q the natural
logarithm of firm market capitalization and dummy variable controls for CEO and year effects We instrument measures of institutional concentration
and total institutional ownership by orthogonalizing with respect to option grant pay-for-performance sensitivity salary and total compensation
Coefficients for industry and year controls are not presented First stage results are not presented One two and three asterisks denote significance at
10 5 and 1 levels
ExecuComp Sample 1992‒1997 Full Sample from ExecuComp 1992‒2002
PPSOG Salary TC PPSOG Salary TC
Change in Shareholder Wealth 144E-05 0004 02133 -409E-06 00011 00157
(308) (775) (596) (691) (638) (068)
Lag1 Change in Shareholder Wealth -240E-05 -00004 00866 118E-06 00002 01422
(131) (087) (346) (097) (116) (51)
Lag1 Institutional Concentration -00898 668681 3263143 02024 916656 2856417
(052) (801) (176) (163) (1545) (1004)
Lag1 Total Institutional Ownership 01021 -43997 2512587 02983 -104171 -1028347
(09) (084) (153) (4) (269) (487)
Lag1 Tobinrsquos Q 00754 -225606 119297 00655 -158657 2997205
(596) (1649) (053) (834) (2535) (727)
Lag1 Log Market Capitalization -04234 828713 7864046 -04002 825106 1685638
(2195) (8132) (2977) (4106) (11133) (373)
CEO Dummy 15848 3214117 2950274 19149 3205862 4529101
(2223) (9389) (2255) (3802) (1331) (2758)
Overall R-Sq -- 04981 01206 -- 04738 01178
Number of Observations 41305 41305 41305 85518 85518 85518
33
Table IX Two Stage Least Squares (Reverse Causality) Tests - Executive Compensation Measures as a Function of Top 5 Ownership
This table shows the coefficients from 2SLS regressions of executive compensation measures against the lagged change and contemporary change
in shareholder wealth instrumented Top 5 institutional ownership Tobinrsquos Q natural logarithm of firm market capitalization and dummy variable
controls for CEO and year effects We instrument Top 5 institutional ownership by orthogonalizing with respect to option grant pay-for-
performance sensitivity salary and total compensation Coefficients for industry and year controls are not presented First stage results are not
presented One two and three asterisks denote significance at 10 5 and 1 levels
ExecuComp sample 1992‒1997 Full Sample 1992‒2002
PPSOG Salary TC PPSOG Salary TC
Change in Shareholder Wealth 136E-05 00041 02128 -407E-06 00011 00146
(293) (79) (596) (692) (619) (063)
Lag1 Change in Shareholder Wealth -241E-05 -00003 00868 116E-06 00003 01439
(133) (07) (347) (096) (142) (514)
Lag1 Top 5 Institutional Ownership -03774 -187512 -3848531 00179 -104261 -126659
(163) (186) (166) (011) (149) (436)
Lag1 Tobinrsquos Q 00731 -223886 10074 00642 -155467 3115073
(573) (1631) (045) (819) (2509) (75)
Lag1 Log Market Capitalization -0416 775637 7696302 -04015 764632 1462536
(2345) (10084) (3766) (3688) (13708) (4072)
CEO Dummy 15854 3211991 2950158 19152 3203152 4519385
(2223) (9381) (2254) (3804) (133) (2752)
Overall R-Sq -- 04967 01202 -- 04722 01168
Number of Observations 41305 41305 41305 85518 85518 85518
34
Table X Specification Tests utilizing Box-Cox Transformation of Market Capitalization
This table presents results from a Box-Cox transformation of market capitalization In
each model we specify transformation of market capitalization only Panel A shows
results for the period 1992‒1997 and Panel B shows results for the entire period
1992‒2002 The first part of each panel shows the Lambda coefficient that applies to
the transformed market capitalization The second part of each panel shows log
likelihood scores with LR Chi2 Statistics in parentheses ‒ these test for the
appropriateness of a linear or log model specification
Panel A Our ExecuComp sample 1992‒1997
(1) (2) (3)
PPSOG Salary TC
Lambda -05143 01881 02876
(-1914) (2413) (1329)
Test H0 (Log Likelihood Score)
Lambda = -1 -1279748 -3178575 -4795326
(23866) (85952) (99688)
Lambda = 0 -1280152 -3138456 -4791159
(31934) (57147) (16339)
Lambda = 1 -1283157 -3163807 -4793506
(92039) (564156) (63292)
Panel B Full ExecuComp sample 1992‒2002
(1) (2) (4)
PPSOG Salary TC
Lambda -044 01853 04036
(-3015) (40) (3666)
Test H0 (Log Likelihood Score)
Lambda = 0 -2688334 -6476494 -10384490
(66599) (160356) (126236)
Lambda = 1 -2694886 -6538506 -10386795
(197644) (1400605) (172345)
35
Figure 2 Institutional Ownership Measures (LHA) vs Total
Compensation (RHA) by Market Capitalization
0
10
20
30
40
50
60
70
1 2 3 4 5 6 7 8 9 10
$00
$10
$20
$30
$40
$50
$60
$70
$80
$90
Millio
ns
InstitutionalConcentration (LHS)Top5 Ownership (LHS)
Institutional TotalOwnership (LHS)Total Compensation(RHS)
Page 7
7
strong relation between institutional concentration as measured by HS and market
capitalization For the smallest decile of market capitalization firms institutional
concentration averages 61 percent declining to about 275 percent for the highest decile
This negative relationship between firm size and institutional concentration seems to arise
because institutional investors do generally favor small firms This causes problems
because in the limit a small company may have only one institutional investor owning
only a small percentage of the company Hence it would appear to be fully concentrated
even though the ability of the institutional investor to influence pay decisions might be
small Thus it is not at all surprising that the correlation between the logarithm of market
capitalization and their institutional concentration measure is both negative and high in
absolute value at 63 percent Their methodology appears to favor small companies that
Table III shows to have low pay levels and high pay for performance sensitivity
Moreover if one does not transform the raw size measure it implicitly assumes that a
given dollar change in firm size has the same absolute impact on pay for a very large
company as it does for a small company Indeed the conventional logarithmic
specification for firm size provides proportionate effects that imply a pay-size elasticity of
approximately one-third instead of unity6
B Measurement of Institutional Influence
We argue that there are additional issues related to the HS measure of institutional
influence ndash institutional concentration In their companion article to HS AHS develop a
model where they capture institutional influence by a single large long-term institutional
6 Note that unlike an earlier version of our paper we leave the HS dependent variables in levels and hence
untransformed While we do this to minimize the alterations that we make to the HS specification ideally
the elasticity specification is preferable especially for the compensation variables Transforming the
dependent variables does not significantly affect the results
8
investor They measure influence relative to total shareholding including private or
atomistic shareholders Presumably they do this to capture relative activism and proxy
voting strength However when it comes to the empirical implementation of the model
AHS adopt the same relative institutional concentration measure as in HS (AHS (p12))
potentially downplaying the relative strength of institutions compared with other
shareholders They justify this because there is generally more than one institutional
investor and there will be free-riding problems with multiple investors that only the
largest institutional investors can overcome While we accept their premise of free-rider
problems it is not obvious why the deflator is lsquoinstitutional shareholdingrsquo rather than
lsquototal shareholdingrsquo from this perspective7 If the strength of the size control can be
queried scaling by total shareholding (ie Top 5 Ownership) produces a measure that
remains relatively unaffected across the size deciles It is 20 percent for the smaller
decile declining to 14 percent for the largest decile whereas for the HS concentration
measure it declines by over 50 percent from 61 percent to 28 percent (see Table III and
Figure 1)
C Testing the Jensen-Murphy Measure of Incentives
The testing of their secondary measure of executive pay-for-performance sensitivity
(see HS Table IV) based on Jensen and Murphy (1990) raises issues other than the size
control HS regress changes in compensation (cash and total) on lagged institutional
influence that has been interacted with the contemporaneous change in shareholder
wealth to capture a pay-for-performance relationship Such a test is appealing ndash
7 AHS do point out one benefit of the institutional concentration measure Institutional investors may prefer
companies with lsquobetterrsquo incentive compensation structures Could using institutional shareholding rather
than total shareholding as the deflator alleviate this problem Presumably both the numerator and
denominator will have the same bias and hence error cancellation is a possibility
9
institutions should increase compensation when the firm has performed well Following
Greene (2000 p326) to execute such a test one needs to include in level form variables
that are then interacted Hence the HS model specification ought to be
1 2 3 1 4 1it it it it it itCompn SW IC SW IC (1)
where itCompn is the change in executive compensation itSW is the contemporaneous
change in shareholder wealth which also appears in the interaction term and 1itIC is
lagged institutional ownership concentration However HS (p2363) omit the
contemporaneous term itSW and the lagged term 1itIC which should appear in levels
as well as in the interaction term Most importantly HS interact their lagged institutional
concentration measure 1itIC with the contemporaneous change in shareholder wealth
itSW As it is after all pay-for-performance concentrated institutions need to condition
their influence on some observed measure of firm performance Contemporaneous firm
performance is unknown at the time institutional influence is measured Reflecting this
timing insight leads to a new Greene (2000) specification
1 2 1 3 1 4 1 1it it it it it itCompn SW IC SW IC (2)
such that all the lagged RHS independent variables that now appear in the interaction
term are included on a stand-alone basis
Reliance on incomplete self-reported values for option grants
Prior to receiving the HS dataset we tried to replicate the findings of HS to validate
our datasets for the HS period and the longer period 1992‒2002 While we initially
focused on the same period as HS (ie 1992‒1997) we found that our sample size was
larger at 47765 observations than HSrsquos at 33928 observations This discrepancy could
have arisen if HS did not replace missing and self-reported option valuations using their
10
Black-Scholes option methodology This may introduce a downward pay bias since self-
reported option-grant pay is subject to possible understatement and there are 13839
observations with likely high option pay dropped altogether
Endogeneity
HSrsquos endogeneity testing raises concerns about reverse causality Institutions may
prefer to hold firms with certain compensation structures such as incentivized executives
HS rightly highlight this concern and use instrumental variables to address this issue HS
assume that institutional ownership reflects any reverse causality Using an instrumental
variables approach to resolve this problem involves finding a variable that influences or
explains institutional ownership but not directly executive compensation They use share
turnover in a firm as an instrument for institutional ownership However theoretical work
by Holmstrom and Tirole (1993) indicates that measures of stock liquidity and trading
activity are indeed important determinants of executive market-based incentives as Kang
and Lui (2008) and Dikolli Kulp and Sedatole (2009) find empirically As such one can
make a case for adopting another instrument rather than share turnover for institutional
ownership
II Previous Research
Prior research fails to address issues concerning possible fragility of the HS results
AHS replicate the HS findings on both incentives and pay levels using the same dataset as
HS and they apply the same definition of institutional concentration and use market
capitalization expressed as a level Cadman Klasa and Matsunaga (2010 Table 6) set out
to replicate HSrsquos (Table II) model of option grant pay-for-performance sensitivity when
they compare a dataset of ExecuComp and Non-ExecuComp firms for the period 2000-
2007 Their ExecuComp results for this sample period are quite similar to our results for
11
the 1992‒2002 period for the same specification controlling for the logarithm of market
capitalization but they do not regard these results as a test of HS as their focus is entirely
on comparing the two groups of stocks Nor do they investigate the impact of
concentration on either cash compensation or total pay Kang and Liu (2008) examine
pay-performance sensitivity use a version of the HS institutional concentration measure
as a control variable in a different framework deriving from Holmstrom and Tirole
(1993) Dikolli Kulp and Sedatole (2009) argue that firms with transient ownership will
issue relatively more equity incentives to counter short-termism with their results robust
to the inclusion of the HS concentration variable
We hoped to find supportive evidence from proxy voting behavior in support of HSs
findings However evidence from actual mutual fund proxy voting records suggests that
most oppose reductions in CEO pay but overall the evidence is mixed8 Davis and Kim
(2005) examine proxy votes of mutual funds for 2004 and find that for a sample of
Fortune 1000 proxy contests in 45 cases that attempted to limit executive pay mutual
funds were unanimously opposed to such attempts Rothberg and Lilien (2006) also
examine proxy-voting data and find that mutual funds voted 66 percent of the time in
managementrsquos favor on issues of executive compensation Ertimur Ferri and Muslu
(2011) find that institutional investors lack sophistication in that they do not target
excessively paid CEOs relative to CEOs with high predicted-pay based on economic
determinants but exert a moderating influence ndash a $23 million reduction ndash for those they
deem ldquoexcessively paidrdquo CEOs
III Data and Methodology
8 We do acknowledge that the HS results are for all institutional investors while the transparency with
respect to proxy voting applies only to mutual funds
12
This Section discusses the data used in the study the construction of variables
descriptive statistics and the methodology Two datasets are used ndash one is the original
dataset generously supplied by Jay Hartzell and Laura Starks which covers 1992‒1997
second a more comprehensive dataset covering 1992‒2002 is computed from the same
data sources used in the HS dataset This is Standard and Poorrsquos (SampP) ExecuComp for
compensation data Thomson Financial CDASpectrum database for institutional holdings
to compute measure of institutional influence SampPrsquos Compustat and the Center for
Research in Stock Prices (CRSP) for firm level accounting and stock price data Our more
comprehensive dataset contains 2559 firms compared with 1914 for the HS dataset Our
full matched sample yields 97679 executive years
A Institutional Influence Measures
To measure institutional influence we employ two ownership-based metrics The first
measure of institutional influence is HSrsquos Institutional Concentration We use this for
direct comparison to the HS analysis The second measure of institutional influence
changes the denominator of the HS measure from institutional shares held to total shares
outstanding in the firm with nomenclature Top 5 Ownership9 We compute both
measures at the fiscal year end of the firm We adjust all compensation measures and
asset prices for inflation to 30 June 2002 prices for our comprehensive sample compared
to the HS dataset that is in the dollars of the day
B Compensation Measures
The ExecuComp database reports several raw measures of executive compensation for
the top five executives (where available) including total salary bonus long-term
9 Top 5 Ownership is included in the HS database but not utilized by them in reported tests as well as
constructed by us for the more comprehensive sample
13
incentives and number of option grants For our more comprehensive dataset we replicate
all the compensation measures used by HS In common with HS we compute the pay-for-
performance sensitivity of option grants using the Yermack (1995) approach This
measure shows the dollar change in value of the executive options for every dollar change
in the value of the firm This involves calculating the option delta using the Black-
Scholes formula adjusted for dividends10
The ratio of the number of shares represented
by the award of options divided by the diluted number of shares outstanding at the
beginning of the year multiplies the option delta with units given by a $1000 change in
firm value Summing salary bonus long-term incentive plan payouts and stock and
option grants provides total direct compensation Options are valued using the Black and
Scholes (1973) formula for European call options adjusted by Merton (1973) to
incorporate dividends
C Descriptive Statistics
Table I presents descriptive statistics of key variables employed in the study according
to our comprehensive dataset11
Institutions own on average 546 percent of the firms in
which they invest This is almost identical to the 531 percent found by HS The
institutions with the five largest holdings account 45 percent of institutional ownership
Once again the figure for HS is almost identical at 44 percent The institutions with the
five largest holdings own on average 229 percent of shares outstanding in a firm (HS
have 223 percent) Average total compensation for executives is $2583 million
considerably larger than that documented by HS ($1250 million) but the HS figure is an
10 When there is more than one grant during the year a weighted average delta is used
11 With the data supplied to us by Hartzell and Starks we are able to replicate exactly their summary data
(HS Table I) for all but cash compensation Salary is included but bonus payments are missing The
replicated table is available on request
14
amalgam of price levels over their sample period 1992‒1997 and ours are in consistent
dollars of June 30 2002 However there has also been significant growth in
compensation levels from 1998 to 2002
ltltINSERT TABLE Igtgt
Based on our comprehensive data in Table II we examine correlations between the
key variables employed in the study Of concern is the large negative correlation of 63
percent which has been discussed previously (Section IIA) between the HS measure of
institutional concentration and the natural logarithm of firm market capitalization
ltltINSERT TABLE IIgtgt
In view of this high correlation Table III provides a breakdown of the three key pay
variables and three institutional ownership variables by size deciles We find that the high
institutional concentration but low institutional ownership share of total shareholdings
mentioned above is evident in the smallest decile of firms by size This decile has low
salary and even lower total compensation in a relative sense but by far the highest pay-
for-performance sensitivity of option grants Small firms award executives a far higher
proportion of market capitalization Since small firm market-capitalization is negligible in
comparison with the highest decile (300 times larger in magnitude) total compensation is
very small in absolute value Moreover due to the domination of the pay-size effect base
salaries are also exceedingly small
ltltINSERT TABLE IIIgtgt
Figure 1 based on Table III shows how as institutional concentration falls with
increases in stock size (but also as the total share of institutional ownership increases)
and the pay-for-performance sensitivity of option grants falls Thus because of what
appears to be a weak size control and size-influenced concentration measure HS
15
potentially pick up a size effect as their main result but not necessarily a monitoring
effect Figure 2 shows that as firm size increases up to decile 8 at least the share of
institutional total ownership increases as does total compensation
ltltINSERT FIGURES 1 AND 2 gtgt
D Methodology
Following HS we use two estimation techniques to examine the effect of institutional
investor influence on executive compensation Due to the censored nature of option grant
pay-for-performance sensitivity this compensation measure most appropriately uses
Tobit model specification whereas an Ordinary Least Squares (OLS) model specification
is used for all other compensation measures We model compensation measures as a
function of contemporaneous and lagged change in shareholder wealth institutional
influence Tobinrsquos Q and market capitalization Controls for CEO industry and year
effects are also used Augmentation occurs to test the effect of institutional influence on
the change in executive cash and total compensation with an interaction term that is the
product of institutional influence and the change in shareholder wealth
The variables used to model executive compensation have the following definitions
The lagged and contemporary measure of change in shareholder wealth is simply the
difference in market capitalization of the firm The measures of institutional influence
used are Institutional Concentration and Top 5 Ownership Use of the two types of
institutional influence occurs in separate models In order to replicate the HS
methodology models using Institutional Concentration have an additional control
consisting of lagged total institutional holdings and Tobinrsquos Q controls for performance
and growth opportunities Market capitalization of the firm is the number of shares
outstanding multiplied by the share price with inclusion both in levels and natural
16
logarithmic form A dummy variable is included to identify whether the executive is the
CEO Dummy variables are also included to control for industry effects This variable
takes a value of 1 for the two-digit Standard Industrial Classification (SIC) in which the
firm operates A year fixed effects (dummy) variable is included taking a value of 1 if the
observation was from the given year
IV Executive Incentives as a Function of Institutional Investor Influence
This section presents the results for the effect of institutional influence on option grant
pay-for-performance sensitivity and changes in compensation
A Option Grant Pay-for-Performance Sensitivity
Table IV presents the results for the effect of institutional influence on option grant
pay-for-performance sensitivity
ltltINSERT TABLE IVgtgt
Model (1) is a direct replication of HS using the HS database as supplied to us by HS
with all the variables computed in the same way and in the same functional form as with
HS The coefficient and t-statistic on lagged institutional concentration interacted with the
change in shareholder wealth is 13994 and 767 that compares to 1358 and 842 in HS
We are unable to account for these minor discrepancies for values that should be
identical In model (2) the only change to the model is that we use the natural logarithm
of market capitalization instead of the level of market capitalization This single change
has a sizeable effect on the coefficient on institutional concentration The coefficient on
institutional concentration is now -00692 and is insignificant The fact that an order
preserving transformation to a control variable can change the results is of interest The
log-transformation to market capitalization also changes the coefficient on Top 5
Ownership in models (3) and (4) The coefficient on Top 5 Ownership is significant in
17
both model (3) and in model (4) but the better size control reduces the coefficient by over
25 percent The better size control does not affect Top 5 Ownership by as much as its
design ensures less sensitivity to what could be inadequate size controls
We also test the effect of institutional influence on option grant pay-for-performance
sensitivity over the period 1992‒2002 This sample is then approximately three times as
large as the sample used by HS Institutional concentration is highly positively significant
in model (5) similar to the results for the period 1992‒1997 When we use the log of
market capitalization in model (6) the coefficient on institutional concentration remains
positive and significant at the 10 percent level with a t-statistic of 166 Of importance
though is that the coefficient on institutional concentration is one-eighth of the coefficient
in model (5) Even for a measure of institutional influence which we believe is less
sensitive to firm size the use of log market capitalization reduces the coefficient on Top 5
Ownership by approximately 50 percent from model (7) to model (8)
B Pay-for-Performance Sensitivity due to Changes in Cash and Total
Compensation
For robustness HS examine the effect that institutional influence conditioned on firm
performance has on a changes in cash and total executive compensation using the general
method of Jensen and Murphy (1990) to compute pay-for-performance sensitivity
ltltINSERT TABLE V gtgt
Table V presents the effect of institutional influence on the change in cash
compensation using the HS size control measure throughout12
Model (1) replicates the
specification in HS for the period of 1992‒1997 and we produce a similar result ndash the
interaction of lagged institutional concentration and contemporaneous change in
12 This test uses our data for the two periods 1992‒1997 and 1992‒2002 We do not use the HS data for
cash compensation as data for bonus is missing in the data supplied to us
18
shareholder wealth positively and significantly relates to a change in executive cash
compensation In model (2) we include the contemporaneous change in shareholder
wealth and the lagged institutional concentration as stand-alone controls as indicated by
Greenersquos (2000) specification of equation (1) above and the interaction term remains
significant In model (3) we interact lagged institutional concentration with the lagged
change in shareholder wealth as in equation (2) above that tries to take account of what
participants could have observed at the time The interaction term is then negative but
insignificant altering the HS result Model (4) provides a similar result where our
preferred proxy Top 5 Ownership captures concentrated institutional influence In
models (5) to (8) we use our full sample In model (5) we find that the lagged
institutional concentration interaction with the contemporaneous change in shareholder
wealth is positive and significant in HSrsquos specification However including change in
contemporaneous shareholder wealth and lagged institutional concentration as separate
control variables as suggested by equation (1)rsquos specification removes the significance
from the interaction term in model (6) Moreover in model (7) when lagged institutional
concentration is interaction with the lagged change in shareholder wealth in accord with
equation (2) this interaction term is now negative and significant reversing the HS
finding We obtain a similar result in model (8) when we measure institutional
concentration influence using Top 5 Ownership13
V The Levels of Executive Fixed and Total Compensation as Functions of
Institutional Investor Influence
13 In unreported results we use the log of market capitalization as a size control These changes to the size
control do not dramatically change the findings from Table V The reason why this size issue is not as
important as in the findings for option grants pay-for-performance is that the change in cash compensation
is not so strongly influenced by firm size In unreported results available on request we also examine the
effect of institutional influence on changes in total compensation with similar results to the change in cash
compensation
19
Two measures of compensation are examined ndash salary and total compensation
ltltINSERT TABLE VI gtgt
Table VI examines the effect of institutional influence on the level of salary In model
(1) we use the HS sample and specification to obtain almost identical results The
coefficient on institutional concentration is (negative) $247509 that is almost identical to
(negative) $247967 in HS Both are highly significant In model (2) we make a single
change to the model ndash we use the natural logarithm of market capitalization This single
change reverses the sign of institutional concentration with the coefficient still being
significant at the 1 percent level In models (3) and (4) we measure institutional influence
using Top 5 Ownership In model (3) where we measure firm size using the level of
market capitalization Top 5 Ownership is insignificant When we use the natural
logarithm of market capitalization in model (4) Top 5 Ownership is significant and
positively relates to the salary level In models (5) to (8) we estimate the same models
using our full sample We find similar results for the full sample as for the period 1992ndash
1997
ltltINSERT TABLE VII gtgt
Table VII examines the effect of institutional influence on total compensation In
model (1) we replicate the HS methodology using the HS data to obtain a similar
significant negative relation between institutional concentration and total compensation of
(negative) $1608633 However after changing to a logarithmic specification of market
capitalization in model (2) institutional concentration increases total compensation by a
significant $32400014
Importantly when institutional influence is measured using Top 5
14 = (5688751 times 0101)
20
Ownership an insignificant relation exists with total compensation in model (3) using the
HS size control but it becomes both positive and significant in (4) with the logged size
control In models (5) and (6) where models are estimated using the full sample we
observe a similar reversal in sign as occurs in models (1) and (2) However now the
magnitude of the reversal is almost equal with a difference of about $67m that is very
large indeed In model (5) the coefficient on institutional concentration is (negative)
$342m and then in model (6) the coefficient is $33m Top 5 Ownership is insignificant
in both models (7) and (8) unlike the results for Salary
VI Robustness
Two issues are examined in is this Section ndash reverse causality and the appropriate
transformation of market capitalization A methodological issue inherent in examining the
effect of ownership-based measures of institutional influence on executive compensation
is endogeneity driven by reverse causality We utilize an instrumental variables approach
where orthogonalized institutional influence measures are regressed on each of the
compensation measures lagged option grant pay-for-performance sensitivity salary and
total compensation In addition market capitalization and a CEO dummy are included in
each regression We use the residual from each regression as an instrument for
institutional influence Such a process removes any preferences for compensation from
the institutional influence measures
ltltINSERT TABLES VIII AND IXgtgt
Table VIII presents the results where we instrument the HS measure lagged
institutional concentration with a compensation-orthogonalized equivalent The analysis
indicates that the instrumented institutional concentration measure is positively and
significantly related to the level of salary and total compensation and insignificantly
21
related to option grant pay-for-performance sensitivity (PPSOG) Table IX presents
results where we instrument Top 5 Ownership with a compensation-orthogonalized
equivalent The instrumented Top 5 Ownership measure insignificantly relates to option
grant pay-for-performance sensitivity and barely negatively relates to salary although it
negatively relates to total compensation suggesting reverse causality is driving a
significant portion of our previous results
ltltINSERT TABLE Xgtgt
Another issue examined in this section is the appropriate transformation of market
capitalization Aneuryn-Evans and Deaton (1980) provide an evaluation of methods
comparing linear and log-linear specifications We use the Box-Cox test to compare
logarithm versus linear specifications In applying the Box-Cox test we specify
transformation of only market capitalization ndash the dependent variable and all other
independent variables remain unchanged Table X presents the lambda coefficient for
each model and the log likelihood score In all models in Panel A and B a lambda of zero
receives a higher (closer to zero) log likelihood score suggesting the log specification is
appropriate
VII Conclusion
Hartzell and Starks (2003) appears to document an effective role for institutional
investors in increasing executive incentives and lowering compensation levels Our paper
re-addresses the issue of concentrated institutional monitoring from a perspective
informed by an additional decade of research since their paper appeared We cannot find
evidence that concentrated institutional investors really have any positive (or negative for
that matter) effect on executive incentive levels While superficially we find evidence
that they may raise executive pay levels this is most likely driven by an institutional
22
preference for highly remunerated executives in large companies and thus by reverse
causality Moreover our findings are not consistent with the rent-extraction hypothesis
suggesting that firms systematically overpay executives in the absence of effective
monitoring by outside parties such as institutional investors
23
References
Almazan Andres Jay C Hartzell and Laura T Starks 2005 ldquoActive Institutional
Shareholders and Costs of Monitoring Evidence from Executive Compensationrdquo
Financial Management Winter 5ndash34
Aneuryn-Evans Gwyn and Angus Deaton 1980 ldquoTesting Linear versus Logarithmic
Regression Modelsrdquo Review of Economic Studies 47 (1) Econometrics Issue 275ndash
291
Babenko Ilona 2009 ldquoShare Repurchases and Pay-Performance Sensitivity of Employee
Compensation Contractsrdquo Journal of Finance 64117-150
Baker George P and Brian J Hall 2004 ldquoCEO Incentives and Firm Sizerdquo Journal of
Labor Market Economics 22 767ndash798
Bebchuk Lucian A Jesse M Fried and David Walker 2002 ldquoManagerial power and
rent extraction in the design of executive compensationrdquo University of Chicago Law
Review 69 751ndash846
Bebchuk Lucian A and Jesse M Fried 2003 ldquoExecutive Compensation as an Agency
Problemrdquo Journal of Economic Perspectives 17 71ndash92
Bebchuk Lucian A and Jesse M Fried 2004 Pay without Performance The Unfulfilled
Promise of Executive Compensation Cambridge MA Harvard University Press
Bertrand Marianne and Sendhil Mullainathan 2001 ldquoAre CEOs Rewarded for Luck
The Ones Without Principals Arerdquo Quarterly Journal of Economics 116 901ndash32
Black Fischer and Myron Scholes 1973 ldquoThe Pricing of Options and Corporate
Liabilitiesrdquo Journal of Political Economy 81 637ndash54
Cadman Brian Sandy Klasa and Steve Matsunaga 2010 ldquoDeterminants of CEO Pay A
Comparison of ExecuComp and Non-ExecuComp Firmsrdquo The Accounting Review 85
1511ndash1543
Chhaochharia Vidhi and Yaniv Grinstein 2009 ldquoCEO Compensation and Board
Structurerdquo Journal of Finance 64 231ndash261
Cichello Michael S 2005 ldquoThe impact of firm size on pay-performance sensitivitiesrdquo
Journal of Corporate Finance 11 609ndash627
Davis Gerald F and E Han Kim 2007 ldquoBusiness ties and proxy voting by mutual
fundsrdquo Journal of Financial Economics 85 552ndash570
Dikolli Shane S Susan L Kulp and Karen L Sedatole 2009 ldquoTransient Institutional
Ownership and CEO Contractingrdquo Accounting Review 84 737ndash770
Edmans Alex Xavier Gabaix and Augustin Landier 2009 ldquoA Multiplicative Model of
Optimal CEO Incentives in Market Equilibriumrdquo Review of Financial Studies 22
4881ndash4917
Edmans Alex 2009 ldquoBlockholder Trading Market Efficiency and Managerial Myopiardquo
Journal of Finance 64 2481ndash2513
Ertimur Yonca Fabrizio Ferri and Volkan Muslu 2011 ldquoShareholder Activism and
CEO Payrdquo Review of Financial Studies 24 535ndash592
24
Gabaix Xavier and Augustin Landier 2008 ldquoWhy Has CEO Pay Increased So Muchrdquo
Quarterly Journal of Economics 123 49ndash100
Garen John 1994 ldquoExecutive compensation and principal-agent theoryrdquo Journal of
Political Economy 102 1175ndash1199
Gibbons Robert and Kevin J Murphy 1992 ldquoOptimal Incentive Contracts in the
Presence of Career Concerns Theory and Evidencerdquo Journal of Political Economy
100 468ndash505
Greene William H 2000 (4th
Edition) Econometric Analysis Prentice-Hall NJ
Guthrie Katherine Jan Sokolowsky and Kam-Ming Wan 2012 ldquoCEO Compensation
and Board Structure Revisitedrdquo Journal of Finance Forthcoming
Hadlock Charles R and Gerald B Lumer 1997 ldquoCompensation turnover and top
management incentives historical evidencerdquo Journal of Business 70 153ndash187
Hartzell Jay C and Laura T Starks 2003 ldquoInstitutional Investors and Executive
Compensationrdquo Journal of Finance 58 2351ndash2374
Huson Mark R Robert Parrino and Laura T Starks 2001 ldquoInternal Monitoring
Mechanisms and CEO Turnover A Long-Term Perspectiverdquo Journal of Finance 56
2265ndash2297
Jensen Michael C and Kevin J Murphy 1990 ldquoPerformance Pay and Top Management
Incentivesrdquo Journal of Political Economy 98 225ndash264
Jin Li 2002 ldquoCEO compensation diversification and incentivesrdquo Journal of Financial
Economics 66 29ndash63
Kang Qiang and Qiao Liu 2008 ldquoStock Market Information Production and Executive
Incentivesrdquo Journal of Corporate Finance 14 484ndash498
Merton Robert 1973 ldquoTheory of Rational Option Pricingrdquo Bell Journal of Economics
and Management Science 4 141-183
Morse Adair Vikram Nanda and Amit Seru 2011 ldquoAre Incentive Contracts Rigged by
Powerful CEOsrdquo Journal of Finance 66 1779ndash1821
Murphy Kevin J 1985 ldquoCorporate performance and managerial remuneration An
empirical analysisrdquo Journal of Accounting and Economics 7 11ndash42
Murphy Kevin J 1999 ldquoExecutive compensationrdquo in Orley Ashenfelter and David
Card Handbook of Labor Economics vol 3 Elsevier
Rosen Sherwin 1992 ldquoContracts and the market for executivesrdquo In Contract
Economics ed Lars Werin and Hans Wijkander 181ndash211 Cambridge MA
Blackwell
Rothberg Burton and Steven Lilien 2006 ldquoMutual Funds and Proxy Voting New
Evidence on Corporate Governancerdquo Journal of Business and Technology Law 1
157ndash184
Schaefer Scott 1998 ldquoThe dependence of payndashperformance sensitivity on the size of the
firmrdquo Review of Economics and Statistics 80 436ndash443
Yermack David 1995 ldquoDo Corporations Award CEO Stock Options Effectivelyrdquo
Journal of Financial Economics 39 237ndash69
25
Table I Univariate Descriptive Statistics
This table reports sample statistics for the key variables used in this study based on our comprehensive
ExecuComp dataset Institutional Concentration is the shares owned by the institutions with the five largest
holdings divided by the shares held by all institutions Top 5 Ownership is the proportion of shares
outstanding held by the institutions with the five largest holdings Total Institutional Ownership is the
proportion of shares held by all institutions Panel B shows the measures executive compensation The pay-
for-performance sensitivity (PPS) of option grants shows the change in executive wealth due to a $1000
change in firm in shareholder wealth Expressions for Salary and Total Compensation are in thousands of
dollars while Market Capitalization is in millions of dollars Panel C shows firm characteristics
Variable Mean Median Std Dev P10 P90
Panel A Institutional Investor Holdings
Institutional Concentration 0450 0424 0151 0282 0656
Top 5 Institutional Ownership 0229 0220 0101 0110 0355
Total Institutional Ownership 0546 0561 0206 0258 0801
Panel B Executive Compensation
PPS of Option Grants 1161 0156 3829 0 2822
Salary 36311 29462 25231 15179 67000
Total Compensation 258313 85878 980701 23875 511022
Panel C Firm Characteristics
Tobins Q 1744 1139 2738 0495 3322
Market Capitalization 560699 112094 1988108 19754 1048754
26
Table II Pairwise Correlation Coefficients
This table presents the correlations between the key variables based on our comprehensive
ExecuComp dataset Institutional Concentration is the shares owned by the institutions with the five
largest holdings divided by the shares held by all institutions Top 5 Ownership is the proportion of
shares outstanding held by the institutions with the five largest holdings Total Institutional
Ownership is the proportion of shares held by all institutions The pay-for-performance sensitivity
(PPS) of option grants is the change in executive wealth due to a $1000 change in shareholder
wealth
(1) (2) (3) (4) (5) (6) (7) (8)
(1) Institutional Concentration 1
(2) Top 5 Ownership 0268 1
(3) Total Ownership -0415 0706 1
(4) PPS of Option Grants 0122 0042 -0024 1
(5) Log Salary -0270 -0012 0171 0038 1
(6) Log Market Capitalization -0630 -0157 0261 -0156 0420 1
(7) Level Market Capitalization -0241 -0157 0007 -0058 0219 0534 1
(8) Tobins Q -0008 -0034 0002 0015 -0083 0066 0121 1
27
Table III
Descriptive Statistics by Market Capitalization Deciles
Market
Capitalization
Deciles
Market
Capitalization
($m)
Total
Compensation
($000)
Salary
($)
PPS Option
Grants
($)
Institutional
Concentration
()
Top 5 Ownership
()
Institutional
Total Ownership
()
1 (Small) 1181 76008 234700 2720 611 200 327
2 2668 90206 257955 1796 495 217 438
3 4193 105946 275120 1517 454 220 485
4 6045 129892 296067 1376 420 206 491
5 8742 155391 315864 1216 394 198 502
6 13086 191888 345942 1072 380 202 531
7 20790 239501 377021 0874 368 196 532
8 34920 314955 427808 0687 354 190 538
9 67747 450970 480319 0569 328 173 528
10 (Big) 366540 846066 614354 0297 275 138 502
28
Table IV Pay-for-Performance Sensitivity of Option Grants as a Function of Institutional Influence
Similar in construction to parts of HS (Table II) this table shows the coefficients from Tobit regressions of option grant pay-for-performance sensitivity per $1000 change in
shareholder wealth against the lagged change and contemporary change in shareholder wealth measures of institutional influence Tobinrsquos Q firm market capitalization and
dummy controls for CEO and year effects We estimate models (1) to (4) over the period 1992‒1997 using data supplied by HS and models (5) to (8) are estimated over the
whole sample Models (1) (2) (5) and (6) use Institutional Concentration as a measure of institutional influence which HS define as the shares owned by the institutions
with the five largest holdings divided by the shares held by all institutions These models control for total institutional ownership Models (3) (4) (7) and (8) use Top 5
Ownership as a measure of institutional influence which we define as the proportion of shares outstanding held by the institutions with the five largest holdings Coefficients
for industry and year controls are not presented One two and three asterisks denote significance at 10 5 and 1 levels
Hartzell and Starks Database 1992‒1997 Full Sample from ExecuComp 1992‒2002
(1) (2) (3) (4) (5) (6) (7) (8)
Change in Shareholder Wealth 77721 191781 89874 137368 -446 -218 -664 -203
(171) (429) (194) (317) (319) (168) (453) (158)
Lag1 Change in Shareholder Wealth 14446 99963 88719 8544 -144 0273 -0342 00852
(027) (212) (166) (184) (071) (015) (016) (005)
Lag1 Institutional Concentration 13994 -00692 26405 03304
(767) (032) (1402) (166)
Lag1 Total Institutional Ownership 08055 10434 07106 0937
(593) (781) (666) (898)
Lag1 Top 5 Institutional Ownership 10567 07457 12116 06293
(399) ( 283) (533) (278)
Lag1 Tobinrsquos Q 0068 00877 00647 00813 00498 00727 00508 00695
(337) (434) (319) (401) (504) (734) (51) (703)
Lag1 Market Capitalization -106202 -173515 -776 -117
(474) (742) (1011) (153)
Lag1 Log Market Capitalization -0297 -02319 -04111 -03908
(1576) (1439) (2997) (2841)
CEO Dummy 13745 13682 13765 13674 28197 28128 28155 28143
(1913) (1918) (1914) (1916) (3388) (3396) (3377) (34)
Number of Observations 33878 33878 33878 33878 97679 97679 97679 97679
29
Table V Pay-for-Performance Sensitivity of Cash Compensation (Salary plus Bonus) as a Function of Institutional Influence - Level Market Capitalization
Similar in parts to models (1) and (2) of HS (Table IV) this table displays the coefficients from OLS regressions of the change in cash compensation against the lagged
change and contemporary change in shareholder wealth measures of institutional influence Tobinrsquos Q firm market capitalization and dummy controls for CEO and year
effects We estimate models (1) to (4) over the period 1992‒1997 based on our ExecuComp sample and models (5) to (8) over the whole sample Models (1) (2) (5) and (6)
use Institutional Concentration as a measure of institutional influence which HS define as the shares owned by the institutions with the five largest holdings divided by the
shares held by all institutions These models control for total institutional ownership Models (4) and (8) use Top 5 Ownership as a measure of institutional influence which
we define as the proportion of shares outstanding held by the institutions with the 5 largest holdings Coefficients for industry and year controls are not presented One two
and three asterisks denote significance at 10 5 and 1 levels
Our Database based on ExecuComp 1992‒1997 Full Sample based on ExecuComp 1992‒2002
(1) (2) (3) (4) (5) (6) (7) (8)
Lag1 Change in Shareholder Wealth 00014 00017 00088 00089 -00006 -00007 00072 00086
(051) (062) (129) (103) (027) (029) (131) (16)
Change in Shareholder Wealth -00102 00102 00105 00032 00102 00101
(097) (292) (293) (055) (611) (608)
Change Shareholder Wealth Lag1 Institutional
Concentration
00435 00764 00363 00253
(463) (258) (689) (144)
Lag1 Change Shareholder Wealth Lag1
Institutional Concentration
-00197 -00283
(122) (197)
Lag1 Institutional Concentration -492182 -288861 400149 453623
(09) (056) (098) (114)
Lag1 Total Institutional Ownership 720688 538372 655091 559587 682069 726226
(276) (232) (281) (314) (308) (322)
Lag1 Change Shareholder Wealth Lag1 Top 5
Institutional Ownership
-0046 -00689
(091) (256)
Lag1 Top 5 Institutional Ownership 583044 1191634
(092) (25)
Lag1 Tobinrsquos Q -32761 -36806 -19422 -18023 -54283 -48681 -29534 -24243
(138) (172) (077) (078) (227) (245) (136) (111)
Lag1 Market Capitalization 00004 00004 00002 00006 00007 00008 00007 00007
(042) (044) (022) (07) (086) (092) (082) (09)
CEO Dummy 768035 765474 768334 773634 616315 61729 619028 618535
(354) (355) (356) (355) (417) (418) (419) (418)
Adjusted R-Sq 00057 0006 00052 0005 00097 00097 00098 00102
Number of Observations 41305 41305 41305 41305 85518 85518 85518 85518
30
Table VI Level of Salary as a Function of Institutional Influence
Similar in parts to models (1) and (2) of HS (Table V) this table displays the coefficients from OLS regressions of salary against the lagged change and contemporary change in
shareholder wealth measures of institutional influence Tobinrsquos Q firm market capitalization and dummy controls for CEO industry and year effects We estimate models (1) to
(4) over the period 1992‒1997 using the actual data supplied by HS and models (5) to (8) over the whole sample 1992‒2002 Models (1) (2) (5) and (6) use Institutional
Concentration as a measure of institutional influence which HS define as the shares owned by the institutions with the five largest holdings divided by the shares held by all
institutions These models control for total institutional ownership Models (3) (4) (7) and (8) use Top 5 Ownership as a measure of institutional influence which we define as
the proportion of shares outstanding held by the institutions with the 5 largest holdings Coefficients for industry and year controls are not presented One two and three
asterisks denote significance at 10 5 and 1 levels
Hartzell and Starks Database 1992‒1997 Full Sample based on ExecuComp 1992‒2002
(1) (2) (3) (4) (5) (6) (7) (8)
Change in Shareholder Wealth 04125E-4 0003 -2439E-4 0003 00021 00012 00024 00012
(057) (464) (031) (468) (1004) (72) (1041) (706)
Lag1 Change in Shareholder Wealth -00062 7856E-4 -0007 9003E-4 -00014 00002 -00014 00003
(594) (098) (592) (112) (522) (123) (5) (145)
Lag1 Institutional Concentration -2475087 555134 -2734983 943317
(3245) (710) (4965) (1671)
Lag1 Total Institutional Ownership 1020109 217488 936337 252464
(1904) (432) (2471) (704)
Lag1 Top 5 Institutional Ownership -148355 31354 -87926 423166
(139) (332) (118) (643)
Lag1 Tobinrsquos Q -195348 -252191 -192344 -25158 -119837 -146373 -122132 -143821
(2812) (3762) (2721) (3752) (248) (274) (2583) (2727)
Lag1 Market Capitalization 844E-06 103E-5 00034 00039
(2781) (3051) (2874) (3104)
Lag1 Log Market Capitalization 750021 723625 776657 732279
(7511) (9285) (11224) (14039)
CEO Dummy 2683978 2702238 2668244 2702417 3281133 3285103 3292175 3283145
(8625) (9277) (8246) (9271) (12739) (13616) (12336) (13605)
Adjusted R-Sq 04604 05302 04187 05295 03856 04588 03476 04574
Number of Observations 36280 36280 36280 36280 97679 97679 97679 97679
31
Table VII Level of Total Direct Compensation as a Function of Institutional Influence
Similar in parts to models (3) and (5) of HS (Table V) this table displays the coefficients from OLS regressions of total direct compensation against the lagged change and
contemporary change in shareholder wealth measures of institutional influence Tobinrsquos Q firm market capitalization and dummy controls for CEO industry and year effects
We estimate models (1) to (4) over the period 1992‒1997 and models (5) to (8) over the whole sample Models (1) (2) (5) and (6) use Institutional Concentration as a measure
of institutional influence which HS define as the shares owned by the institutions with the five largest holdings divided by the shares held by all institutions These models
control for total institutional ownership Models (3) (4) (7) and (8) use Top 5 Ownership as a measure of institutional influence which we define as the proportion of shares
outstanding held by the institutions with the five largest holdings Coefficients for industry and year controls are not presented One two and three asterisks denote significance
at 10 5 and 1 levels
Hartzell and Starks Database 1992‒1997 Full Sample from ExecuComp 1992‒2002
(1) (2) (3) (4) (5) (6) (7) (8)
Change in Shareholder Wealth 005779 007006 005282 006791 00423 00148 0046 00138
(415) (581) (369) (565) (203) (07) (22) (066)
Lag1 Change in Shareholder Wealth 003182 00714 00272 00715 00613 0126 00614 01278
(126) (298) (107) (299) (275) (488) (276) (494)
Lag1 Institutional Concentration -1608633 3248998 -3424892 3334186
(151) (283) (1172) (118)
Lag1 Total Institutional Ownership 104013 5473128 1531013 -2926764
(1645) (831) (876) (158)
Lag1 Top 5 Institutional Ownership 188585 3927986 -2159177 -2640879
(016) (343) (067) (092)
Lag1 Tobinrsquos Q -445692 -803612 -454859 -833566 3257027 2971201 3213936 3086205
(427) (791) (432) (831) (83) (77) (815) (793)
Lag1 Market Capitalization 49E-5 615E-5 01031 01088
(113) (1365) (1555) (168)
Lag1 Log Market Capitalization 4683497 4759377 1645156 1427856
(2775) (3469) (3679) (4239)
CEO Dummy 1501307 1508886 1493688 1508736 4713079 4713629 47278 4703908
(2648) (2684) (2618) (2683) (282) (2828) (2819) (2824)
Adjusted R-Sq 01861 02060 01695 02051 01071 01113 01033 011
Number of Observations 33229 33229 33229 33229 97679 97679 97679 97679
32
Table VIII Two Stage Least Squares (Reverse Causality) Tests - Executive Compensation Measures as a Function of Institutional Concentration
This table shows the coefficients from the Two-Stage Least Squares (2SLS) regressions of executive compensation measures against the lagged change
and contemporary change in shareholder wealth instrumented institutional concentration and total institutional ownership Tobinrsquos Q the natural
logarithm of firm market capitalization and dummy variable controls for CEO and year effects We instrument measures of institutional concentration
and total institutional ownership by orthogonalizing with respect to option grant pay-for-performance sensitivity salary and total compensation
Coefficients for industry and year controls are not presented First stage results are not presented One two and three asterisks denote significance at
10 5 and 1 levels
ExecuComp Sample 1992‒1997 Full Sample from ExecuComp 1992‒2002
PPSOG Salary TC PPSOG Salary TC
Change in Shareholder Wealth 144E-05 0004 02133 -409E-06 00011 00157
(308) (775) (596) (691) (638) (068)
Lag1 Change in Shareholder Wealth -240E-05 -00004 00866 118E-06 00002 01422
(131) (087) (346) (097) (116) (51)
Lag1 Institutional Concentration -00898 668681 3263143 02024 916656 2856417
(052) (801) (176) (163) (1545) (1004)
Lag1 Total Institutional Ownership 01021 -43997 2512587 02983 -104171 -1028347
(09) (084) (153) (4) (269) (487)
Lag1 Tobinrsquos Q 00754 -225606 119297 00655 -158657 2997205
(596) (1649) (053) (834) (2535) (727)
Lag1 Log Market Capitalization -04234 828713 7864046 -04002 825106 1685638
(2195) (8132) (2977) (4106) (11133) (373)
CEO Dummy 15848 3214117 2950274 19149 3205862 4529101
(2223) (9389) (2255) (3802) (1331) (2758)
Overall R-Sq -- 04981 01206 -- 04738 01178
Number of Observations 41305 41305 41305 85518 85518 85518
33
Table IX Two Stage Least Squares (Reverse Causality) Tests - Executive Compensation Measures as a Function of Top 5 Ownership
This table shows the coefficients from 2SLS regressions of executive compensation measures against the lagged change and contemporary change
in shareholder wealth instrumented Top 5 institutional ownership Tobinrsquos Q natural logarithm of firm market capitalization and dummy variable
controls for CEO and year effects We instrument Top 5 institutional ownership by orthogonalizing with respect to option grant pay-for-
performance sensitivity salary and total compensation Coefficients for industry and year controls are not presented First stage results are not
presented One two and three asterisks denote significance at 10 5 and 1 levels
ExecuComp sample 1992‒1997 Full Sample 1992‒2002
PPSOG Salary TC PPSOG Salary TC
Change in Shareholder Wealth 136E-05 00041 02128 -407E-06 00011 00146
(293) (79) (596) (692) (619) (063)
Lag1 Change in Shareholder Wealth -241E-05 -00003 00868 116E-06 00003 01439
(133) (07) (347) (096) (142) (514)
Lag1 Top 5 Institutional Ownership -03774 -187512 -3848531 00179 -104261 -126659
(163) (186) (166) (011) (149) (436)
Lag1 Tobinrsquos Q 00731 -223886 10074 00642 -155467 3115073
(573) (1631) (045) (819) (2509) (75)
Lag1 Log Market Capitalization -0416 775637 7696302 -04015 764632 1462536
(2345) (10084) (3766) (3688) (13708) (4072)
CEO Dummy 15854 3211991 2950158 19152 3203152 4519385
(2223) (9381) (2254) (3804) (133) (2752)
Overall R-Sq -- 04967 01202 -- 04722 01168
Number of Observations 41305 41305 41305 85518 85518 85518
34
Table X Specification Tests utilizing Box-Cox Transformation of Market Capitalization
This table presents results from a Box-Cox transformation of market capitalization In
each model we specify transformation of market capitalization only Panel A shows
results for the period 1992‒1997 and Panel B shows results for the entire period
1992‒2002 The first part of each panel shows the Lambda coefficient that applies to
the transformed market capitalization The second part of each panel shows log
likelihood scores with LR Chi2 Statistics in parentheses ‒ these test for the
appropriateness of a linear or log model specification
Panel A Our ExecuComp sample 1992‒1997
(1) (2) (3)
PPSOG Salary TC
Lambda -05143 01881 02876
(-1914) (2413) (1329)
Test H0 (Log Likelihood Score)
Lambda = -1 -1279748 -3178575 -4795326
(23866) (85952) (99688)
Lambda = 0 -1280152 -3138456 -4791159
(31934) (57147) (16339)
Lambda = 1 -1283157 -3163807 -4793506
(92039) (564156) (63292)
Panel B Full ExecuComp sample 1992‒2002
(1) (2) (4)
PPSOG Salary TC
Lambda -044 01853 04036
(-3015) (40) (3666)
Test H0 (Log Likelihood Score)
Lambda = 0 -2688334 -6476494 -10384490
(66599) (160356) (126236)
Lambda = 1 -2694886 -6538506 -10386795
(197644) (1400605) (172345)
35
Figure 2 Institutional Ownership Measures (LHA) vs Total
Compensation (RHA) by Market Capitalization
0
10
20
30
40
50
60
70
1 2 3 4 5 6 7 8 9 10
$00
$10
$20
$30
$40
$50
$60
$70
$80
$90
Millio
ns
InstitutionalConcentration (LHS)Top5 Ownership (LHS)
Institutional TotalOwnership (LHS)Total Compensation(RHS)
Page 8
8
investor They measure influence relative to total shareholding including private or
atomistic shareholders Presumably they do this to capture relative activism and proxy
voting strength However when it comes to the empirical implementation of the model
AHS adopt the same relative institutional concentration measure as in HS (AHS (p12))
potentially downplaying the relative strength of institutions compared with other
shareholders They justify this because there is generally more than one institutional
investor and there will be free-riding problems with multiple investors that only the
largest institutional investors can overcome While we accept their premise of free-rider
problems it is not obvious why the deflator is lsquoinstitutional shareholdingrsquo rather than
lsquototal shareholdingrsquo from this perspective7 If the strength of the size control can be
queried scaling by total shareholding (ie Top 5 Ownership) produces a measure that
remains relatively unaffected across the size deciles It is 20 percent for the smaller
decile declining to 14 percent for the largest decile whereas for the HS concentration
measure it declines by over 50 percent from 61 percent to 28 percent (see Table III and
Figure 1)
C Testing the Jensen-Murphy Measure of Incentives
The testing of their secondary measure of executive pay-for-performance sensitivity
(see HS Table IV) based on Jensen and Murphy (1990) raises issues other than the size
control HS regress changes in compensation (cash and total) on lagged institutional
influence that has been interacted with the contemporaneous change in shareholder
wealth to capture a pay-for-performance relationship Such a test is appealing ndash
7 AHS do point out one benefit of the institutional concentration measure Institutional investors may prefer
companies with lsquobetterrsquo incentive compensation structures Could using institutional shareholding rather
than total shareholding as the deflator alleviate this problem Presumably both the numerator and
denominator will have the same bias and hence error cancellation is a possibility
9
institutions should increase compensation when the firm has performed well Following
Greene (2000 p326) to execute such a test one needs to include in level form variables
that are then interacted Hence the HS model specification ought to be
1 2 3 1 4 1it it it it it itCompn SW IC SW IC (1)
where itCompn is the change in executive compensation itSW is the contemporaneous
change in shareholder wealth which also appears in the interaction term and 1itIC is
lagged institutional ownership concentration However HS (p2363) omit the
contemporaneous term itSW and the lagged term 1itIC which should appear in levels
as well as in the interaction term Most importantly HS interact their lagged institutional
concentration measure 1itIC with the contemporaneous change in shareholder wealth
itSW As it is after all pay-for-performance concentrated institutions need to condition
their influence on some observed measure of firm performance Contemporaneous firm
performance is unknown at the time institutional influence is measured Reflecting this
timing insight leads to a new Greene (2000) specification
1 2 1 3 1 4 1 1it it it it it itCompn SW IC SW IC (2)
such that all the lagged RHS independent variables that now appear in the interaction
term are included on a stand-alone basis
Reliance on incomplete self-reported values for option grants
Prior to receiving the HS dataset we tried to replicate the findings of HS to validate
our datasets for the HS period and the longer period 1992‒2002 While we initially
focused on the same period as HS (ie 1992‒1997) we found that our sample size was
larger at 47765 observations than HSrsquos at 33928 observations This discrepancy could
have arisen if HS did not replace missing and self-reported option valuations using their
10
Black-Scholes option methodology This may introduce a downward pay bias since self-
reported option-grant pay is subject to possible understatement and there are 13839
observations with likely high option pay dropped altogether
Endogeneity
HSrsquos endogeneity testing raises concerns about reverse causality Institutions may
prefer to hold firms with certain compensation structures such as incentivized executives
HS rightly highlight this concern and use instrumental variables to address this issue HS
assume that institutional ownership reflects any reverse causality Using an instrumental
variables approach to resolve this problem involves finding a variable that influences or
explains institutional ownership but not directly executive compensation They use share
turnover in a firm as an instrument for institutional ownership However theoretical work
by Holmstrom and Tirole (1993) indicates that measures of stock liquidity and trading
activity are indeed important determinants of executive market-based incentives as Kang
and Lui (2008) and Dikolli Kulp and Sedatole (2009) find empirically As such one can
make a case for adopting another instrument rather than share turnover for institutional
ownership
II Previous Research
Prior research fails to address issues concerning possible fragility of the HS results
AHS replicate the HS findings on both incentives and pay levels using the same dataset as
HS and they apply the same definition of institutional concentration and use market
capitalization expressed as a level Cadman Klasa and Matsunaga (2010 Table 6) set out
to replicate HSrsquos (Table II) model of option grant pay-for-performance sensitivity when
they compare a dataset of ExecuComp and Non-ExecuComp firms for the period 2000-
2007 Their ExecuComp results for this sample period are quite similar to our results for
11
the 1992‒2002 period for the same specification controlling for the logarithm of market
capitalization but they do not regard these results as a test of HS as their focus is entirely
on comparing the two groups of stocks Nor do they investigate the impact of
concentration on either cash compensation or total pay Kang and Liu (2008) examine
pay-performance sensitivity use a version of the HS institutional concentration measure
as a control variable in a different framework deriving from Holmstrom and Tirole
(1993) Dikolli Kulp and Sedatole (2009) argue that firms with transient ownership will
issue relatively more equity incentives to counter short-termism with their results robust
to the inclusion of the HS concentration variable
We hoped to find supportive evidence from proxy voting behavior in support of HSs
findings However evidence from actual mutual fund proxy voting records suggests that
most oppose reductions in CEO pay but overall the evidence is mixed8 Davis and Kim
(2005) examine proxy votes of mutual funds for 2004 and find that for a sample of
Fortune 1000 proxy contests in 45 cases that attempted to limit executive pay mutual
funds were unanimously opposed to such attempts Rothberg and Lilien (2006) also
examine proxy-voting data and find that mutual funds voted 66 percent of the time in
managementrsquos favor on issues of executive compensation Ertimur Ferri and Muslu
(2011) find that institutional investors lack sophistication in that they do not target
excessively paid CEOs relative to CEOs with high predicted-pay based on economic
determinants but exert a moderating influence ndash a $23 million reduction ndash for those they
deem ldquoexcessively paidrdquo CEOs
III Data and Methodology
8 We do acknowledge that the HS results are for all institutional investors while the transparency with
respect to proxy voting applies only to mutual funds
12
This Section discusses the data used in the study the construction of variables
descriptive statistics and the methodology Two datasets are used ndash one is the original
dataset generously supplied by Jay Hartzell and Laura Starks which covers 1992‒1997
second a more comprehensive dataset covering 1992‒2002 is computed from the same
data sources used in the HS dataset This is Standard and Poorrsquos (SampP) ExecuComp for
compensation data Thomson Financial CDASpectrum database for institutional holdings
to compute measure of institutional influence SampPrsquos Compustat and the Center for
Research in Stock Prices (CRSP) for firm level accounting and stock price data Our more
comprehensive dataset contains 2559 firms compared with 1914 for the HS dataset Our
full matched sample yields 97679 executive years
A Institutional Influence Measures
To measure institutional influence we employ two ownership-based metrics The first
measure of institutional influence is HSrsquos Institutional Concentration We use this for
direct comparison to the HS analysis The second measure of institutional influence
changes the denominator of the HS measure from institutional shares held to total shares
outstanding in the firm with nomenclature Top 5 Ownership9 We compute both
measures at the fiscal year end of the firm We adjust all compensation measures and
asset prices for inflation to 30 June 2002 prices for our comprehensive sample compared
to the HS dataset that is in the dollars of the day
B Compensation Measures
The ExecuComp database reports several raw measures of executive compensation for
the top five executives (where available) including total salary bonus long-term
9 Top 5 Ownership is included in the HS database but not utilized by them in reported tests as well as
constructed by us for the more comprehensive sample
13
incentives and number of option grants For our more comprehensive dataset we replicate
all the compensation measures used by HS In common with HS we compute the pay-for-
performance sensitivity of option grants using the Yermack (1995) approach This
measure shows the dollar change in value of the executive options for every dollar change
in the value of the firm This involves calculating the option delta using the Black-
Scholes formula adjusted for dividends10
The ratio of the number of shares represented
by the award of options divided by the diluted number of shares outstanding at the
beginning of the year multiplies the option delta with units given by a $1000 change in
firm value Summing salary bonus long-term incentive plan payouts and stock and
option grants provides total direct compensation Options are valued using the Black and
Scholes (1973) formula for European call options adjusted by Merton (1973) to
incorporate dividends
C Descriptive Statistics
Table I presents descriptive statistics of key variables employed in the study according
to our comprehensive dataset11
Institutions own on average 546 percent of the firms in
which they invest This is almost identical to the 531 percent found by HS The
institutions with the five largest holdings account 45 percent of institutional ownership
Once again the figure for HS is almost identical at 44 percent The institutions with the
five largest holdings own on average 229 percent of shares outstanding in a firm (HS
have 223 percent) Average total compensation for executives is $2583 million
considerably larger than that documented by HS ($1250 million) but the HS figure is an
10 When there is more than one grant during the year a weighted average delta is used
11 With the data supplied to us by Hartzell and Starks we are able to replicate exactly their summary data
(HS Table I) for all but cash compensation Salary is included but bonus payments are missing The
replicated table is available on request
14
amalgam of price levels over their sample period 1992‒1997 and ours are in consistent
dollars of June 30 2002 However there has also been significant growth in
compensation levels from 1998 to 2002
ltltINSERT TABLE Igtgt
Based on our comprehensive data in Table II we examine correlations between the
key variables employed in the study Of concern is the large negative correlation of 63
percent which has been discussed previously (Section IIA) between the HS measure of
institutional concentration and the natural logarithm of firm market capitalization
ltltINSERT TABLE IIgtgt
In view of this high correlation Table III provides a breakdown of the three key pay
variables and three institutional ownership variables by size deciles We find that the high
institutional concentration but low institutional ownership share of total shareholdings
mentioned above is evident in the smallest decile of firms by size This decile has low
salary and even lower total compensation in a relative sense but by far the highest pay-
for-performance sensitivity of option grants Small firms award executives a far higher
proportion of market capitalization Since small firm market-capitalization is negligible in
comparison with the highest decile (300 times larger in magnitude) total compensation is
very small in absolute value Moreover due to the domination of the pay-size effect base
salaries are also exceedingly small
ltltINSERT TABLE IIIgtgt
Figure 1 based on Table III shows how as institutional concentration falls with
increases in stock size (but also as the total share of institutional ownership increases)
and the pay-for-performance sensitivity of option grants falls Thus because of what
appears to be a weak size control and size-influenced concentration measure HS
15
potentially pick up a size effect as their main result but not necessarily a monitoring
effect Figure 2 shows that as firm size increases up to decile 8 at least the share of
institutional total ownership increases as does total compensation
ltltINSERT FIGURES 1 AND 2 gtgt
D Methodology
Following HS we use two estimation techniques to examine the effect of institutional
investor influence on executive compensation Due to the censored nature of option grant
pay-for-performance sensitivity this compensation measure most appropriately uses
Tobit model specification whereas an Ordinary Least Squares (OLS) model specification
is used for all other compensation measures We model compensation measures as a
function of contemporaneous and lagged change in shareholder wealth institutional
influence Tobinrsquos Q and market capitalization Controls for CEO industry and year
effects are also used Augmentation occurs to test the effect of institutional influence on
the change in executive cash and total compensation with an interaction term that is the
product of institutional influence and the change in shareholder wealth
The variables used to model executive compensation have the following definitions
The lagged and contemporary measure of change in shareholder wealth is simply the
difference in market capitalization of the firm The measures of institutional influence
used are Institutional Concentration and Top 5 Ownership Use of the two types of
institutional influence occurs in separate models In order to replicate the HS
methodology models using Institutional Concentration have an additional control
consisting of lagged total institutional holdings and Tobinrsquos Q controls for performance
and growth opportunities Market capitalization of the firm is the number of shares
outstanding multiplied by the share price with inclusion both in levels and natural
16
logarithmic form A dummy variable is included to identify whether the executive is the
CEO Dummy variables are also included to control for industry effects This variable
takes a value of 1 for the two-digit Standard Industrial Classification (SIC) in which the
firm operates A year fixed effects (dummy) variable is included taking a value of 1 if the
observation was from the given year
IV Executive Incentives as a Function of Institutional Investor Influence
This section presents the results for the effect of institutional influence on option grant
pay-for-performance sensitivity and changes in compensation
A Option Grant Pay-for-Performance Sensitivity
Table IV presents the results for the effect of institutional influence on option grant
pay-for-performance sensitivity
ltltINSERT TABLE IVgtgt
Model (1) is a direct replication of HS using the HS database as supplied to us by HS
with all the variables computed in the same way and in the same functional form as with
HS The coefficient and t-statistic on lagged institutional concentration interacted with the
change in shareholder wealth is 13994 and 767 that compares to 1358 and 842 in HS
We are unable to account for these minor discrepancies for values that should be
identical In model (2) the only change to the model is that we use the natural logarithm
of market capitalization instead of the level of market capitalization This single change
has a sizeable effect on the coefficient on institutional concentration The coefficient on
institutional concentration is now -00692 and is insignificant The fact that an order
preserving transformation to a control variable can change the results is of interest The
log-transformation to market capitalization also changes the coefficient on Top 5
Ownership in models (3) and (4) The coefficient on Top 5 Ownership is significant in
17
both model (3) and in model (4) but the better size control reduces the coefficient by over
25 percent The better size control does not affect Top 5 Ownership by as much as its
design ensures less sensitivity to what could be inadequate size controls
We also test the effect of institutional influence on option grant pay-for-performance
sensitivity over the period 1992‒2002 This sample is then approximately three times as
large as the sample used by HS Institutional concentration is highly positively significant
in model (5) similar to the results for the period 1992‒1997 When we use the log of
market capitalization in model (6) the coefficient on institutional concentration remains
positive and significant at the 10 percent level with a t-statistic of 166 Of importance
though is that the coefficient on institutional concentration is one-eighth of the coefficient
in model (5) Even for a measure of institutional influence which we believe is less
sensitive to firm size the use of log market capitalization reduces the coefficient on Top 5
Ownership by approximately 50 percent from model (7) to model (8)
B Pay-for-Performance Sensitivity due to Changes in Cash and Total
Compensation
For robustness HS examine the effect that institutional influence conditioned on firm
performance has on a changes in cash and total executive compensation using the general
method of Jensen and Murphy (1990) to compute pay-for-performance sensitivity
ltltINSERT TABLE V gtgt
Table V presents the effect of institutional influence on the change in cash
compensation using the HS size control measure throughout12
Model (1) replicates the
specification in HS for the period of 1992‒1997 and we produce a similar result ndash the
interaction of lagged institutional concentration and contemporaneous change in
12 This test uses our data for the two periods 1992‒1997 and 1992‒2002 We do not use the HS data for
cash compensation as data for bonus is missing in the data supplied to us
18
shareholder wealth positively and significantly relates to a change in executive cash
compensation In model (2) we include the contemporaneous change in shareholder
wealth and the lagged institutional concentration as stand-alone controls as indicated by
Greenersquos (2000) specification of equation (1) above and the interaction term remains
significant In model (3) we interact lagged institutional concentration with the lagged
change in shareholder wealth as in equation (2) above that tries to take account of what
participants could have observed at the time The interaction term is then negative but
insignificant altering the HS result Model (4) provides a similar result where our
preferred proxy Top 5 Ownership captures concentrated institutional influence In
models (5) to (8) we use our full sample In model (5) we find that the lagged
institutional concentration interaction with the contemporaneous change in shareholder
wealth is positive and significant in HSrsquos specification However including change in
contemporaneous shareholder wealth and lagged institutional concentration as separate
control variables as suggested by equation (1)rsquos specification removes the significance
from the interaction term in model (6) Moreover in model (7) when lagged institutional
concentration is interaction with the lagged change in shareholder wealth in accord with
equation (2) this interaction term is now negative and significant reversing the HS
finding We obtain a similar result in model (8) when we measure institutional
concentration influence using Top 5 Ownership13
V The Levels of Executive Fixed and Total Compensation as Functions of
Institutional Investor Influence
13 In unreported results we use the log of market capitalization as a size control These changes to the size
control do not dramatically change the findings from Table V The reason why this size issue is not as
important as in the findings for option grants pay-for-performance is that the change in cash compensation
is not so strongly influenced by firm size In unreported results available on request we also examine the
effect of institutional influence on changes in total compensation with similar results to the change in cash
compensation
19
Two measures of compensation are examined ndash salary and total compensation
ltltINSERT TABLE VI gtgt
Table VI examines the effect of institutional influence on the level of salary In model
(1) we use the HS sample and specification to obtain almost identical results The
coefficient on institutional concentration is (negative) $247509 that is almost identical to
(negative) $247967 in HS Both are highly significant In model (2) we make a single
change to the model ndash we use the natural logarithm of market capitalization This single
change reverses the sign of institutional concentration with the coefficient still being
significant at the 1 percent level In models (3) and (4) we measure institutional influence
using Top 5 Ownership In model (3) where we measure firm size using the level of
market capitalization Top 5 Ownership is insignificant When we use the natural
logarithm of market capitalization in model (4) Top 5 Ownership is significant and
positively relates to the salary level In models (5) to (8) we estimate the same models
using our full sample We find similar results for the full sample as for the period 1992ndash
1997
ltltINSERT TABLE VII gtgt
Table VII examines the effect of institutional influence on total compensation In
model (1) we replicate the HS methodology using the HS data to obtain a similar
significant negative relation between institutional concentration and total compensation of
(negative) $1608633 However after changing to a logarithmic specification of market
capitalization in model (2) institutional concentration increases total compensation by a
significant $32400014
Importantly when institutional influence is measured using Top 5
14 = (5688751 times 0101)
20
Ownership an insignificant relation exists with total compensation in model (3) using the
HS size control but it becomes both positive and significant in (4) with the logged size
control In models (5) and (6) where models are estimated using the full sample we
observe a similar reversal in sign as occurs in models (1) and (2) However now the
magnitude of the reversal is almost equal with a difference of about $67m that is very
large indeed In model (5) the coefficient on institutional concentration is (negative)
$342m and then in model (6) the coefficient is $33m Top 5 Ownership is insignificant
in both models (7) and (8) unlike the results for Salary
VI Robustness
Two issues are examined in is this Section ndash reverse causality and the appropriate
transformation of market capitalization A methodological issue inherent in examining the
effect of ownership-based measures of institutional influence on executive compensation
is endogeneity driven by reverse causality We utilize an instrumental variables approach
where orthogonalized institutional influence measures are regressed on each of the
compensation measures lagged option grant pay-for-performance sensitivity salary and
total compensation In addition market capitalization and a CEO dummy are included in
each regression We use the residual from each regression as an instrument for
institutional influence Such a process removes any preferences for compensation from
the institutional influence measures
ltltINSERT TABLES VIII AND IXgtgt
Table VIII presents the results where we instrument the HS measure lagged
institutional concentration with a compensation-orthogonalized equivalent The analysis
indicates that the instrumented institutional concentration measure is positively and
significantly related to the level of salary and total compensation and insignificantly
21
related to option grant pay-for-performance sensitivity (PPSOG) Table IX presents
results where we instrument Top 5 Ownership with a compensation-orthogonalized
equivalent The instrumented Top 5 Ownership measure insignificantly relates to option
grant pay-for-performance sensitivity and barely negatively relates to salary although it
negatively relates to total compensation suggesting reverse causality is driving a
significant portion of our previous results
ltltINSERT TABLE Xgtgt
Another issue examined in this section is the appropriate transformation of market
capitalization Aneuryn-Evans and Deaton (1980) provide an evaluation of methods
comparing linear and log-linear specifications We use the Box-Cox test to compare
logarithm versus linear specifications In applying the Box-Cox test we specify
transformation of only market capitalization ndash the dependent variable and all other
independent variables remain unchanged Table X presents the lambda coefficient for
each model and the log likelihood score In all models in Panel A and B a lambda of zero
receives a higher (closer to zero) log likelihood score suggesting the log specification is
appropriate
VII Conclusion
Hartzell and Starks (2003) appears to document an effective role for institutional
investors in increasing executive incentives and lowering compensation levels Our paper
re-addresses the issue of concentrated institutional monitoring from a perspective
informed by an additional decade of research since their paper appeared We cannot find
evidence that concentrated institutional investors really have any positive (or negative for
that matter) effect on executive incentive levels While superficially we find evidence
that they may raise executive pay levels this is most likely driven by an institutional
22
preference for highly remunerated executives in large companies and thus by reverse
causality Moreover our findings are not consistent with the rent-extraction hypothesis
suggesting that firms systematically overpay executives in the absence of effective
monitoring by outside parties such as institutional investors
23
References
Almazan Andres Jay C Hartzell and Laura T Starks 2005 ldquoActive Institutional
Shareholders and Costs of Monitoring Evidence from Executive Compensationrdquo
Financial Management Winter 5ndash34
Aneuryn-Evans Gwyn and Angus Deaton 1980 ldquoTesting Linear versus Logarithmic
Regression Modelsrdquo Review of Economic Studies 47 (1) Econometrics Issue 275ndash
291
Babenko Ilona 2009 ldquoShare Repurchases and Pay-Performance Sensitivity of Employee
Compensation Contractsrdquo Journal of Finance 64117-150
Baker George P and Brian J Hall 2004 ldquoCEO Incentives and Firm Sizerdquo Journal of
Labor Market Economics 22 767ndash798
Bebchuk Lucian A Jesse M Fried and David Walker 2002 ldquoManagerial power and
rent extraction in the design of executive compensationrdquo University of Chicago Law
Review 69 751ndash846
Bebchuk Lucian A and Jesse M Fried 2003 ldquoExecutive Compensation as an Agency
Problemrdquo Journal of Economic Perspectives 17 71ndash92
Bebchuk Lucian A and Jesse M Fried 2004 Pay without Performance The Unfulfilled
Promise of Executive Compensation Cambridge MA Harvard University Press
Bertrand Marianne and Sendhil Mullainathan 2001 ldquoAre CEOs Rewarded for Luck
The Ones Without Principals Arerdquo Quarterly Journal of Economics 116 901ndash32
Black Fischer and Myron Scholes 1973 ldquoThe Pricing of Options and Corporate
Liabilitiesrdquo Journal of Political Economy 81 637ndash54
Cadman Brian Sandy Klasa and Steve Matsunaga 2010 ldquoDeterminants of CEO Pay A
Comparison of ExecuComp and Non-ExecuComp Firmsrdquo The Accounting Review 85
1511ndash1543
Chhaochharia Vidhi and Yaniv Grinstein 2009 ldquoCEO Compensation and Board
Structurerdquo Journal of Finance 64 231ndash261
Cichello Michael S 2005 ldquoThe impact of firm size on pay-performance sensitivitiesrdquo
Journal of Corporate Finance 11 609ndash627
Davis Gerald F and E Han Kim 2007 ldquoBusiness ties and proxy voting by mutual
fundsrdquo Journal of Financial Economics 85 552ndash570
Dikolli Shane S Susan L Kulp and Karen L Sedatole 2009 ldquoTransient Institutional
Ownership and CEO Contractingrdquo Accounting Review 84 737ndash770
Edmans Alex Xavier Gabaix and Augustin Landier 2009 ldquoA Multiplicative Model of
Optimal CEO Incentives in Market Equilibriumrdquo Review of Financial Studies 22
4881ndash4917
Edmans Alex 2009 ldquoBlockholder Trading Market Efficiency and Managerial Myopiardquo
Journal of Finance 64 2481ndash2513
Ertimur Yonca Fabrizio Ferri and Volkan Muslu 2011 ldquoShareholder Activism and
CEO Payrdquo Review of Financial Studies 24 535ndash592
24
Gabaix Xavier and Augustin Landier 2008 ldquoWhy Has CEO Pay Increased So Muchrdquo
Quarterly Journal of Economics 123 49ndash100
Garen John 1994 ldquoExecutive compensation and principal-agent theoryrdquo Journal of
Political Economy 102 1175ndash1199
Gibbons Robert and Kevin J Murphy 1992 ldquoOptimal Incentive Contracts in the
Presence of Career Concerns Theory and Evidencerdquo Journal of Political Economy
100 468ndash505
Greene William H 2000 (4th
Edition) Econometric Analysis Prentice-Hall NJ
Guthrie Katherine Jan Sokolowsky and Kam-Ming Wan 2012 ldquoCEO Compensation
and Board Structure Revisitedrdquo Journal of Finance Forthcoming
Hadlock Charles R and Gerald B Lumer 1997 ldquoCompensation turnover and top
management incentives historical evidencerdquo Journal of Business 70 153ndash187
Hartzell Jay C and Laura T Starks 2003 ldquoInstitutional Investors and Executive
Compensationrdquo Journal of Finance 58 2351ndash2374
Huson Mark R Robert Parrino and Laura T Starks 2001 ldquoInternal Monitoring
Mechanisms and CEO Turnover A Long-Term Perspectiverdquo Journal of Finance 56
2265ndash2297
Jensen Michael C and Kevin J Murphy 1990 ldquoPerformance Pay and Top Management
Incentivesrdquo Journal of Political Economy 98 225ndash264
Jin Li 2002 ldquoCEO compensation diversification and incentivesrdquo Journal of Financial
Economics 66 29ndash63
Kang Qiang and Qiao Liu 2008 ldquoStock Market Information Production and Executive
Incentivesrdquo Journal of Corporate Finance 14 484ndash498
Merton Robert 1973 ldquoTheory of Rational Option Pricingrdquo Bell Journal of Economics
and Management Science 4 141-183
Morse Adair Vikram Nanda and Amit Seru 2011 ldquoAre Incentive Contracts Rigged by
Powerful CEOsrdquo Journal of Finance 66 1779ndash1821
Murphy Kevin J 1985 ldquoCorporate performance and managerial remuneration An
empirical analysisrdquo Journal of Accounting and Economics 7 11ndash42
Murphy Kevin J 1999 ldquoExecutive compensationrdquo in Orley Ashenfelter and David
Card Handbook of Labor Economics vol 3 Elsevier
Rosen Sherwin 1992 ldquoContracts and the market for executivesrdquo In Contract
Economics ed Lars Werin and Hans Wijkander 181ndash211 Cambridge MA
Blackwell
Rothberg Burton and Steven Lilien 2006 ldquoMutual Funds and Proxy Voting New
Evidence on Corporate Governancerdquo Journal of Business and Technology Law 1
157ndash184
Schaefer Scott 1998 ldquoThe dependence of payndashperformance sensitivity on the size of the
firmrdquo Review of Economics and Statistics 80 436ndash443
Yermack David 1995 ldquoDo Corporations Award CEO Stock Options Effectivelyrdquo
Journal of Financial Economics 39 237ndash69
25
Table I Univariate Descriptive Statistics
This table reports sample statistics for the key variables used in this study based on our comprehensive
ExecuComp dataset Institutional Concentration is the shares owned by the institutions with the five largest
holdings divided by the shares held by all institutions Top 5 Ownership is the proportion of shares
outstanding held by the institutions with the five largest holdings Total Institutional Ownership is the
proportion of shares held by all institutions Panel B shows the measures executive compensation The pay-
for-performance sensitivity (PPS) of option grants shows the change in executive wealth due to a $1000
change in firm in shareholder wealth Expressions for Salary and Total Compensation are in thousands of
dollars while Market Capitalization is in millions of dollars Panel C shows firm characteristics
Variable Mean Median Std Dev P10 P90
Panel A Institutional Investor Holdings
Institutional Concentration 0450 0424 0151 0282 0656
Top 5 Institutional Ownership 0229 0220 0101 0110 0355
Total Institutional Ownership 0546 0561 0206 0258 0801
Panel B Executive Compensation
PPS of Option Grants 1161 0156 3829 0 2822
Salary 36311 29462 25231 15179 67000
Total Compensation 258313 85878 980701 23875 511022
Panel C Firm Characteristics
Tobins Q 1744 1139 2738 0495 3322
Market Capitalization 560699 112094 1988108 19754 1048754
26
Table II Pairwise Correlation Coefficients
This table presents the correlations between the key variables based on our comprehensive
ExecuComp dataset Institutional Concentration is the shares owned by the institutions with the five
largest holdings divided by the shares held by all institutions Top 5 Ownership is the proportion of
shares outstanding held by the institutions with the five largest holdings Total Institutional
Ownership is the proportion of shares held by all institutions The pay-for-performance sensitivity
(PPS) of option grants is the change in executive wealth due to a $1000 change in shareholder
wealth
(1) (2) (3) (4) (5) (6) (7) (8)
(1) Institutional Concentration 1
(2) Top 5 Ownership 0268 1
(3) Total Ownership -0415 0706 1
(4) PPS of Option Grants 0122 0042 -0024 1
(5) Log Salary -0270 -0012 0171 0038 1
(6) Log Market Capitalization -0630 -0157 0261 -0156 0420 1
(7) Level Market Capitalization -0241 -0157 0007 -0058 0219 0534 1
(8) Tobins Q -0008 -0034 0002 0015 -0083 0066 0121 1
27
Table III
Descriptive Statistics by Market Capitalization Deciles
Market
Capitalization
Deciles
Market
Capitalization
($m)
Total
Compensation
($000)
Salary
($)
PPS Option
Grants
($)
Institutional
Concentration
()
Top 5 Ownership
()
Institutional
Total Ownership
()
1 (Small) 1181 76008 234700 2720 611 200 327
2 2668 90206 257955 1796 495 217 438
3 4193 105946 275120 1517 454 220 485
4 6045 129892 296067 1376 420 206 491
5 8742 155391 315864 1216 394 198 502
6 13086 191888 345942 1072 380 202 531
7 20790 239501 377021 0874 368 196 532
8 34920 314955 427808 0687 354 190 538
9 67747 450970 480319 0569 328 173 528
10 (Big) 366540 846066 614354 0297 275 138 502
28
Table IV Pay-for-Performance Sensitivity of Option Grants as a Function of Institutional Influence
Similar in construction to parts of HS (Table II) this table shows the coefficients from Tobit regressions of option grant pay-for-performance sensitivity per $1000 change in
shareholder wealth against the lagged change and contemporary change in shareholder wealth measures of institutional influence Tobinrsquos Q firm market capitalization and
dummy controls for CEO and year effects We estimate models (1) to (4) over the period 1992‒1997 using data supplied by HS and models (5) to (8) are estimated over the
whole sample Models (1) (2) (5) and (6) use Institutional Concentration as a measure of institutional influence which HS define as the shares owned by the institutions
with the five largest holdings divided by the shares held by all institutions These models control for total institutional ownership Models (3) (4) (7) and (8) use Top 5
Ownership as a measure of institutional influence which we define as the proportion of shares outstanding held by the institutions with the five largest holdings Coefficients
for industry and year controls are not presented One two and three asterisks denote significance at 10 5 and 1 levels
Hartzell and Starks Database 1992‒1997 Full Sample from ExecuComp 1992‒2002
(1) (2) (3) (4) (5) (6) (7) (8)
Change in Shareholder Wealth 77721 191781 89874 137368 -446 -218 -664 -203
(171) (429) (194) (317) (319) (168) (453) (158)
Lag1 Change in Shareholder Wealth 14446 99963 88719 8544 -144 0273 -0342 00852
(027) (212) (166) (184) (071) (015) (016) (005)
Lag1 Institutional Concentration 13994 -00692 26405 03304
(767) (032) (1402) (166)
Lag1 Total Institutional Ownership 08055 10434 07106 0937
(593) (781) (666) (898)
Lag1 Top 5 Institutional Ownership 10567 07457 12116 06293
(399) ( 283) (533) (278)
Lag1 Tobinrsquos Q 0068 00877 00647 00813 00498 00727 00508 00695
(337) (434) (319) (401) (504) (734) (51) (703)
Lag1 Market Capitalization -106202 -173515 -776 -117
(474) (742) (1011) (153)
Lag1 Log Market Capitalization -0297 -02319 -04111 -03908
(1576) (1439) (2997) (2841)
CEO Dummy 13745 13682 13765 13674 28197 28128 28155 28143
(1913) (1918) (1914) (1916) (3388) (3396) (3377) (34)
Number of Observations 33878 33878 33878 33878 97679 97679 97679 97679
29
Table V Pay-for-Performance Sensitivity of Cash Compensation (Salary plus Bonus) as a Function of Institutional Influence - Level Market Capitalization
Similar in parts to models (1) and (2) of HS (Table IV) this table displays the coefficients from OLS regressions of the change in cash compensation against the lagged
change and contemporary change in shareholder wealth measures of institutional influence Tobinrsquos Q firm market capitalization and dummy controls for CEO and year
effects We estimate models (1) to (4) over the period 1992‒1997 based on our ExecuComp sample and models (5) to (8) over the whole sample Models (1) (2) (5) and (6)
use Institutional Concentration as a measure of institutional influence which HS define as the shares owned by the institutions with the five largest holdings divided by the
shares held by all institutions These models control for total institutional ownership Models (4) and (8) use Top 5 Ownership as a measure of institutional influence which
we define as the proportion of shares outstanding held by the institutions with the 5 largest holdings Coefficients for industry and year controls are not presented One two
and three asterisks denote significance at 10 5 and 1 levels
Our Database based on ExecuComp 1992‒1997 Full Sample based on ExecuComp 1992‒2002
(1) (2) (3) (4) (5) (6) (7) (8)
Lag1 Change in Shareholder Wealth 00014 00017 00088 00089 -00006 -00007 00072 00086
(051) (062) (129) (103) (027) (029) (131) (16)
Change in Shareholder Wealth -00102 00102 00105 00032 00102 00101
(097) (292) (293) (055) (611) (608)
Change Shareholder Wealth Lag1 Institutional
Concentration
00435 00764 00363 00253
(463) (258) (689) (144)
Lag1 Change Shareholder Wealth Lag1
Institutional Concentration
-00197 -00283
(122) (197)
Lag1 Institutional Concentration -492182 -288861 400149 453623
(09) (056) (098) (114)
Lag1 Total Institutional Ownership 720688 538372 655091 559587 682069 726226
(276) (232) (281) (314) (308) (322)
Lag1 Change Shareholder Wealth Lag1 Top 5
Institutional Ownership
-0046 -00689
(091) (256)
Lag1 Top 5 Institutional Ownership 583044 1191634
(092) (25)
Lag1 Tobinrsquos Q -32761 -36806 -19422 -18023 -54283 -48681 -29534 -24243
(138) (172) (077) (078) (227) (245) (136) (111)
Lag1 Market Capitalization 00004 00004 00002 00006 00007 00008 00007 00007
(042) (044) (022) (07) (086) (092) (082) (09)
CEO Dummy 768035 765474 768334 773634 616315 61729 619028 618535
(354) (355) (356) (355) (417) (418) (419) (418)
Adjusted R-Sq 00057 0006 00052 0005 00097 00097 00098 00102
Number of Observations 41305 41305 41305 41305 85518 85518 85518 85518
30
Table VI Level of Salary as a Function of Institutional Influence
Similar in parts to models (1) and (2) of HS (Table V) this table displays the coefficients from OLS regressions of salary against the lagged change and contemporary change in
shareholder wealth measures of institutional influence Tobinrsquos Q firm market capitalization and dummy controls for CEO industry and year effects We estimate models (1) to
(4) over the period 1992‒1997 using the actual data supplied by HS and models (5) to (8) over the whole sample 1992‒2002 Models (1) (2) (5) and (6) use Institutional
Concentration as a measure of institutional influence which HS define as the shares owned by the institutions with the five largest holdings divided by the shares held by all
institutions These models control for total institutional ownership Models (3) (4) (7) and (8) use Top 5 Ownership as a measure of institutional influence which we define as
the proportion of shares outstanding held by the institutions with the 5 largest holdings Coefficients for industry and year controls are not presented One two and three
asterisks denote significance at 10 5 and 1 levels
Hartzell and Starks Database 1992‒1997 Full Sample based on ExecuComp 1992‒2002
(1) (2) (3) (4) (5) (6) (7) (8)
Change in Shareholder Wealth 04125E-4 0003 -2439E-4 0003 00021 00012 00024 00012
(057) (464) (031) (468) (1004) (72) (1041) (706)
Lag1 Change in Shareholder Wealth -00062 7856E-4 -0007 9003E-4 -00014 00002 -00014 00003
(594) (098) (592) (112) (522) (123) (5) (145)
Lag1 Institutional Concentration -2475087 555134 -2734983 943317
(3245) (710) (4965) (1671)
Lag1 Total Institutional Ownership 1020109 217488 936337 252464
(1904) (432) (2471) (704)
Lag1 Top 5 Institutional Ownership -148355 31354 -87926 423166
(139) (332) (118) (643)
Lag1 Tobinrsquos Q -195348 -252191 -192344 -25158 -119837 -146373 -122132 -143821
(2812) (3762) (2721) (3752) (248) (274) (2583) (2727)
Lag1 Market Capitalization 844E-06 103E-5 00034 00039
(2781) (3051) (2874) (3104)
Lag1 Log Market Capitalization 750021 723625 776657 732279
(7511) (9285) (11224) (14039)
CEO Dummy 2683978 2702238 2668244 2702417 3281133 3285103 3292175 3283145
(8625) (9277) (8246) (9271) (12739) (13616) (12336) (13605)
Adjusted R-Sq 04604 05302 04187 05295 03856 04588 03476 04574
Number of Observations 36280 36280 36280 36280 97679 97679 97679 97679
31
Table VII Level of Total Direct Compensation as a Function of Institutional Influence
Similar in parts to models (3) and (5) of HS (Table V) this table displays the coefficients from OLS regressions of total direct compensation against the lagged change and
contemporary change in shareholder wealth measures of institutional influence Tobinrsquos Q firm market capitalization and dummy controls for CEO industry and year effects
We estimate models (1) to (4) over the period 1992‒1997 and models (5) to (8) over the whole sample Models (1) (2) (5) and (6) use Institutional Concentration as a measure
of institutional influence which HS define as the shares owned by the institutions with the five largest holdings divided by the shares held by all institutions These models
control for total institutional ownership Models (3) (4) (7) and (8) use Top 5 Ownership as a measure of institutional influence which we define as the proportion of shares
outstanding held by the institutions with the five largest holdings Coefficients for industry and year controls are not presented One two and three asterisks denote significance
at 10 5 and 1 levels
Hartzell and Starks Database 1992‒1997 Full Sample from ExecuComp 1992‒2002
(1) (2) (3) (4) (5) (6) (7) (8)
Change in Shareholder Wealth 005779 007006 005282 006791 00423 00148 0046 00138
(415) (581) (369) (565) (203) (07) (22) (066)
Lag1 Change in Shareholder Wealth 003182 00714 00272 00715 00613 0126 00614 01278
(126) (298) (107) (299) (275) (488) (276) (494)
Lag1 Institutional Concentration -1608633 3248998 -3424892 3334186
(151) (283) (1172) (118)
Lag1 Total Institutional Ownership 104013 5473128 1531013 -2926764
(1645) (831) (876) (158)
Lag1 Top 5 Institutional Ownership 188585 3927986 -2159177 -2640879
(016) (343) (067) (092)
Lag1 Tobinrsquos Q -445692 -803612 -454859 -833566 3257027 2971201 3213936 3086205
(427) (791) (432) (831) (83) (77) (815) (793)
Lag1 Market Capitalization 49E-5 615E-5 01031 01088
(113) (1365) (1555) (168)
Lag1 Log Market Capitalization 4683497 4759377 1645156 1427856
(2775) (3469) (3679) (4239)
CEO Dummy 1501307 1508886 1493688 1508736 4713079 4713629 47278 4703908
(2648) (2684) (2618) (2683) (282) (2828) (2819) (2824)
Adjusted R-Sq 01861 02060 01695 02051 01071 01113 01033 011
Number of Observations 33229 33229 33229 33229 97679 97679 97679 97679
32
Table VIII Two Stage Least Squares (Reverse Causality) Tests - Executive Compensation Measures as a Function of Institutional Concentration
This table shows the coefficients from the Two-Stage Least Squares (2SLS) regressions of executive compensation measures against the lagged change
and contemporary change in shareholder wealth instrumented institutional concentration and total institutional ownership Tobinrsquos Q the natural
logarithm of firm market capitalization and dummy variable controls for CEO and year effects We instrument measures of institutional concentration
and total institutional ownership by orthogonalizing with respect to option grant pay-for-performance sensitivity salary and total compensation
Coefficients for industry and year controls are not presented First stage results are not presented One two and three asterisks denote significance at
10 5 and 1 levels
ExecuComp Sample 1992‒1997 Full Sample from ExecuComp 1992‒2002
PPSOG Salary TC PPSOG Salary TC
Change in Shareholder Wealth 144E-05 0004 02133 -409E-06 00011 00157
(308) (775) (596) (691) (638) (068)
Lag1 Change in Shareholder Wealth -240E-05 -00004 00866 118E-06 00002 01422
(131) (087) (346) (097) (116) (51)
Lag1 Institutional Concentration -00898 668681 3263143 02024 916656 2856417
(052) (801) (176) (163) (1545) (1004)
Lag1 Total Institutional Ownership 01021 -43997 2512587 02983 -104171 -1028347
(09) (084) (153) (4) (269) (487)
Lag1 Tobinrsquos Q 00754 -225606 119297 00655 -158657 2997205
(596) (1649) (053) (834) (2535) (727)
Lag1 Log Market Capitalization -04234 828713 7864046 -04002 825106 1685638
(2195) (8132) (2977) (4106) (11133) (373)
CEO Dummy 15848 3214117 2950274 19149 3205862 4529101
(2223) (9389) (2255) (3802) (1331) (2758)
Overall R-Sq -- 04981 01206 -- 04738 01178
Number of Observations 41305 41305 41305 85518 85518 85518
33
Table IX Two Stage Least Squares (Reverse Causality) Tests - Executive Compensation Measures as a Function of Top 5 Ownership
This table shows the coefficients from 2SLS regressions of executive compensation measures against the lagged change and contemporary change
in shareholder wealth instrumented Top 5 institutional ownership Tobinrsquos Q natural logarithm of firm market capitalization and dummy variable
controls for CEO and year effects We instrument Top 5 institutional ownership by orthogonalizing with respect to option grant pay-for-
performance sensitivity salary and total compensation Coefficients for industry and year controls are not presented First stage results are not
presented One two and three asterisks denote significance at 10 5 and 1 levels
ExecuComp sample 1992‒1997 Full Sample 1992‒2002
PPSOG Salary TC PPSOG Salary TC
Change in Shareholder Wealth 136E-05 00041 02128 -407E-06 00011 00146
(293) (79) (596) (692) (619) (063)
Lag1 Change in Shareholder Wealth -241E-05 -00003 00868 116E-06 00003 01439
(133) (07) (347) (096) (142) (514)
Lag1 Top 5 Institutional Ownership -03774 -187512 -3848531 00179 -104261 -126659
(163) (186) (166) (011) (149) (436)
Lag1 Tobinrsquos Q 00731 -223886 10074 00642 -155467 3115073
(573) (1631) (045) (819) (2509) (75)
Lag1 Log Market Capitalization -0416 775637 7696302 -04015 764632 1462536
(2345) (10084) (3766) (3688) (13708) (4072)
CEO Dummy 15854 3211991 2950158 19152 3203152 4519385
(2223) (9381) (2254) (3804) (133) (2752)
Overall R-Sq -- 04967 01202 -- 04722 01168
Number of Observations 41305 41305 41305 85518 85518 85518
34
Table X Specification Tests utilizing Box-Cox Transformation of Market Capitalization
This table presents results from a Box-Cox transformation of market capitalization In
each model we specify transformation of market capitalization only Panel A shows
results for the period 1992‒1997 and Panel B shows results for the entire period
1992‒2002 The first part of each panel shows the Lambda coefficient that applies to
the transformed market capitalization The second part of each panel shows log
likelihood scores with LR Chi2 Statistics in parentheses ‒ these test for the
appropriateness of a linear or log model specification
Panel A Our ExecuComp sample 1992‒1997
(1) (2) (3)
PPSOG Salary TC
Lambda -05143 01881 02876
(-1914) (2413) (1329)
Test H0 (Log Likelihood Score)
Lambda = -1 -1279748 -3178575 -4795326
(23866) (85952) (99688)
Lambda = 0 -1280152 -3138456 -4791159
(31934) (57147) (16339)
Lambda = 1 -1283157 -3163807 -4793506
(92039) (564156) (63292)
Panel B Full ExecuComp sample 1992‒2002
(1) (2) (4)
PPSOG Salary TC
Lambda -044 01853 04036
(-3015) (40) (3666)
Test H0 (Log Likelihood Score)
Lambda = 0 -2688334 -6476494 -10384490
(66599) (160356) (126236)
Lambda = 1 -2694886 -6538506 -10386795
(197644) (1400605) (172345)
35
Figure 2 Institutional Ownership Measures (LHA) vs Total
Compensation (RHA) by Market Capitalization
0
10
20
30
40
50
60
70
1 2 3 4 5 6 7 8 9 10
$00
$10
$20
$30
$40
$50
$60
$70
$80
$90
Millio
ns
InstitutionalConcentration (LHS)Top5 Ownership (LHS)
Institutional TotalOwnership (LHS)Total Compensation(RHS)
Page 9
9
institutions should increase compensation when the firm has performed well Following
Greene (2000 p326) to execute such a test one needs to include in level form variables
that are then interacted Hence the HS model specification ought to be
1 2 3 1 4 1it it it it it itCompn SW IC SW IC (1)
where itCompn is the change in executive compensation itSW is the contemporaneous
change in shareholder wealth which also appears in the interaction term and 1itIC is
lagged institutional ownership concentration However HS (p2363) omit the
contemporaneous term itSW and the lagged term 1itIC which should appear in levels
as well as in the interaction term Most importantly HS interact their lagged institutional
concentration measure 1itIC with the contemporaneous change in shareholder wealth
itSW As it is after all pay-for-performance concentrated institutions need to condition
their influence on some observed measure of firm performance Contemporaneous firm
performance is unknown at the time institutional influence is measured Reflecting this
timing insight leads to a new Greene (2000) specification
1 2 1 3 1 4 1 1it it it it it itCompn SW IC SW IC (2)
such that all the lagged RHS independent variables that now appear in the interaction
term are included on a stand-alone basis
Reliance on incomplete self-reported values for option grants
Prior to receiving the HS dataset we tried to replicate the findings of HS to validate
our datasets for the HS period and the longer period 1992‒2002 While we initially
focused on the same period as HS (ie 1992‒1997) we found that our sample size was
larger at 47765 observations than HSrsquos at 33928 observations This discrepancy could
have arisen if HS did not replace missing and self-reported option valuations using their
10
Black-Scholes option methodology This may introduce a downward pay bias since self-
reported option-grant pay is subject to possible understatement and there are 13839
observations with likely high option pay dropped altogether
Endogeneity
HSrsquos endogeneity testing raises concerns about reverse causality Institutions may
prefer to hold firms with certain compensation structures such as incentivized executives
HS rightly highlight this concern and use instrumental variables to address this issue HS
assume that institutional ownership reflects any reverse causality Using an instrumental
variables approach to resolve this problem involves finding a variable that influences or
explains institutional ownership but not directly executive compensation They use share
turnover in a firm as an instrument for institutional ownership However theoretical work
by Holmstrom and Tirole (1993) indicates that measures of stock liquidity and trading
activity are indeed important determinants of executive market-based incentives as Kang
and Lui (2008) and Dikolli Kulp and Sedatole (2009) find empirically As such one can
make a case for adopting another instrument rather than share turnover for institutional
ownership
II Previous Research
Prior research fails to address issues concerning possible fragility of the HS results
AHS replicate the HS findings on both incentives and pay levels using the same dataset as
HS and they apply the same definition of institutional concentration and use market
capitalization expressed as a level Cadman Klasa and Matsunaga (2010 Table 6) set out
to replicate HSrsquos (Table II) model of option grant pay-for-performance sensitivity when
they compare a dataset of ExecuComp and Non-ExecuComp firms for the period 2000-
2007 Their ExecuComp results for this sample period are quite similar to our results for
11
the 1992‒2002 period for the same specification controlling for the logarithm of market
capitalization but they do not regard these results as a test of HS as their focus is entirely
on comparing the two groups of stocks Nor do they investigate the impact of
concentration on either cash compensation or total pay Kang and Liu (2008) examine
pay-performance sensitivity use a version of the HS institutional concentration measure
as a control variable in a different framework deriving from Holmstrom and Tirole
(1993) Dikolli Kulp and Sedatole (2009) argue that firms with transient ownership will
issue relatively more equity incentives to counter short-termism with their results robust
to the inclusion of the HS concentration variable
We hoped to find supportive evidence from proxy voting behavior in support of HSs
findings However evidence from actual mutual fund proxy voting records suggests that
most oppose reductions in CEO pay but overall the evidence is mixed8 Davis and Kim
(2005) examine proxy votes of mutual funds for 2004 and find that for a sample of
Fortune 1000 proxy contests in 45 cases that attempted to limit executive pay mutual
funds were unanimously opposed to such attempts Rothberg and Lilien (2006) also
examine proxy-voting data and find that mutual funds voted 66 percent of the time in
managementrsquos favor on issues of executive compensation Ertimur Ferri and Muslu
(2011) find that institutional investors lack sophistication in that they do not target
excessively paid CEOs relative to CEOs with high predicted-pay based on economic
determinants but exert a moderating influence ndash a $23 million reduction ndash for those they
deem ldquoexcessively paidrdquo CEOs
III Data and Methodology
8 We do acknowledge that the HS results are for all institutional investors while the transparency with
respect to proxy voting applies only to mutual funds
12
This Section discusses the data used in the study the construction of variables
descriptive statistics and the methodology Two datasets are used ndash one is the original
dataset generously supplied by Jay Hartzell and Laura Starks which covers 1992‒1997
second a more comprehensive dataset covering 1992‒2002 is computed from the same
data sources used in the HS dataset This is Standard and Poorrsquos (SampP) ExecuComp for
compensation data Thomson Financial CDASpectrum database for institutional holdings
to compute measure of institutional influence SampPrsquos Compustat and the Center for
Research in Stock Prices (CRSP) for firm level accounting and stock price data Our more
comprehensive dataset contains 2559 firms compared with 1914 for the HS dataset Our
full matched sample yields 97679 executive years
A Institutional Influence Measures
To measure institutional influence we employ two ownership-based metrics The first
measure of institutional influence is HSrsquos Institutional Concentration We use this for
direct comparison to the HS analysis The second measure of institutional influence
changes the denominator of the HS measure from institutional shares held to total shares
outstanding in the firm with nomenclature Top 5 Ownership9 We compute both
measures at the fiscal year end of the firm We adjust all compensation measures and
asset prices for inflation to 30 June 2002 prices for our comprehensive sample compared
to the HS dataset that is in the dollars of the day
B Compensation Measures
The ExecuComp database reports several raw measures of executive compensation for
the top five executives (where available) including total salary bonus long-term
9 Top 5 Ownership is included in the HS database but not utilized by them in reported tests as well as
constructed by us for the more comprehensive sample
13
incentives and number of option grants For our more comprehensive dataset we replicate
all the compensation measures used by HS In common with HS we compute the pay-for-
performance sensitivity of option grants using the Yermack (1995) approach This
measure shows the dollar change in value of the executive options for every dollar change
in the value of the firm This involves calculating the option delta using the Black-
Scholes formula adjusted for dividends10
The ratio of the number of shares represented
by the award of options divided by the diluted number of shares outstanding at the
beginning of the year multiplies the option delta with units given by a $1000 change in
firm value Summing salary bonus long-term incentive plan payouts and stock and
option grants provides total direct compensation Options are valued using the Black and
Scholes (1973) formula for European call options adjusted by Merton (1973) to
incorporate dividends
C Descriptive Statistics
Table I presents descriptive statistics of key variables employed in the study according
to our comprehensive dataset11
Institutions own on average 546 percent of the firms in
which they invest This is almost identical to the 531 percent found by HS The
institutions with the five largest holdings account 45 percent of institutional ownership
Once again the figure for HS is almost identical at 44 percent The institutions with the
five largest holdings own on average 229 percent of shares outstanding in a firm (HS
have 223 percent) Average total compensation for executives is $2583 million
considerably larger than that documented by HS ($1250 million) but the HS figure is an
10 When there is more than one grant during the year a weighted average delta is used
11 With the data supplied to us by Hartzell and Starks we are able to replicate exactly their summary data
(HS Table I) for all but cash compensation Salary is included but bonus payments are missing The
replicated table is available on request
14
amalgam of price levels over their sample period 1992‒1997 and ours are in consistent
dollars of June 30 2002 However there has also been significant growth in
compensation levels from 1998 to 2002
ltltINSERT TABLE Igtgt
Based on our comprehensive data in Table II we examine correlations between the
key variables employed in the study Of concern is the large negative correlation of 63
percent which has been discussed previously (Section IIA) between the HS measure of
institutional concentration and the natural logarithm of firm market capitalization
ltltINSERT TABLE IIgtgt
In view of this high correlation Table III provides a breakdown of the three key pay
variables and three institutional ownership variables by size deciles We find that the high
institutional concentration but low institutional ownership share of total shareholdings
mentioned above is evident in the smallest decile of firms by size This decile has low
salary and even lower total compensation in a relative sense but by far the highest pay-
for-performance sensitivity of option grants Small firms award executives a far higher
proportion of market capitalization Since small firm market-capitalization is negligible in
comparison with the highest decile (300 times larger in magnitude) total compensation is
very small in absolute value Moreover due to the domination of the pay-size effect base
salaries are also exceedingly small
ltltINSERT TABLE IIIgtgt
Figure 1 based on Table III shows how as institutional concentration falls with
increases in stock size (but also as the total share of institutional ownership increases)
and the pay-for-performance sensitivity of option grants falls Thus because of what
appears to be a weak size control and size-influenced concentration measure HS
15
potentially pick up a size effect as their main result but not necessarily a monitoring
effect Figure 2 shows that as firm size increases up to decile 8 at least the share of
institutional total ownership increases as does total compensation
ltltINSERT FIGURES 1 AND 2 gtgt
D Methodology
Following HS we use two estimation techniques to examine the effect of institutional
investor influence on executive compensation Due to the censored nature of option grant
pay-for-performance sensitivity this compensation measure most appropriately uses
Tobit model specification whereas an Ordinary Least Squares (OLS) model specification
is used for all other compensation measures We model compensation measures as a
function of contemporaneous and lagged change in shareholder wealth institutional
influence Tobinrsquos Q and market capitalization Controls for CEO industry and year
effects are also used Augmentation occurs to test the effect of institutional influence on
the change in executive cash and total compensation with an interaction term that is the
product of institutional influence and the change in shareholder wealth
The variables used to model executive compensation have the following definitions
The lagged and contemporary measure of change in shareholder wealth is simply the
difference in market capitalization of the firm The measures of institutional influence
used are Institutional Concentration and Top 5 Ownership Use of the two types of
institutional influence occurs in separate models In order to replicate the HS
methodology models using Institutional Concentration have an additional control
consisting of lagged total institutional holdings and Tobinrsquos Q controls for performance
and growth opportunities Market capitalization of the firm is the number of shares
outstanding multiplied by the share price with inclusion both in levels and natural
16
logarithmic form A dummy variable is included to identify whether the executive is the
CEO Dummy variables are also included to control for industry effects This variable
takes a value of 1 for the two-digit Standard Industrial Classification (SIC) in which the
firm operates A year fixed effects (dummy) variable is included taking a value of 1 if the
observation was from the given year
IV Executive Incentives as a Function of Institutional Investor Influence
This section presents the results for the effect of institutional influence on option grant
pay-for-performance sensitivity and changes in compensation
A Option Grant Pay-for-Performance Sensitivity
Table IV presents the results for the effect of institutional influence on option grant
pay-for-performance sensitivity
ltltINSERT TABLE IVgtgt
Model (1) is a direct replication of HS using the HS database as supplied to us by HS
with all the variables computed in the same way and in the same functional form as with
HS The coefficient and t-statistic on lagged institutional concentration interacted with the
change in shareholder wealth is 13994 and 767 that compares to 1358 and 842 in HS
We are unable to account for these minor discrepancies for values that should be
identical In model (2) the only change to the model is that we use the natural logarithm
of market capitalization instead of the level of market capitalization This single change
has a sizeable effect on the coefficient on institutional concentration The coefficient on
institutional concentration is now -00692 and is insignificant The fact that an order
preserving transformation to a control variable can change the results is of interest The
log-transformation to market capitalization also changes the coefficient on Top 5
Ownership in models (3) and (4) The coefficient on Top 5 Ownership is significant in
17
both model (3) and in model (4) but the better size control reduces the coefficient by over
25 percent The better size control does not affect Top 5 Ownership by as much as its
design ensures less sensitivity to what could be inadequate size controls
We also test the effect of institutional influence on option grant pay-for-performance
sensitivity over the period 1992‒2002 This sample is then approximately three times as
large as the sample used by HS Institutional concentration is highly positively significant
in model (5) similar to the results for the period 1992‒1997 When we use the log of
market capitalization in model (6) the coefficient on institutional concentration remains
positive and significant at the 10 percent level with a t-statistic of 166 Of importance
though is that the coefficient on institutional concentration is one-eighth of the coefficient
in model (5) Even for a measure of institutional influence which we believe is less
sensitive to firm size the use of log market capitalization reduces the coefficient on Top 5
Ownership by approximately 50 percent from model (7) to model (8)
B Pay-for-Performance Sensitivity due to Changes in Cash and Total
Compensation
For robustness HS examine the effect that institutional influence conditioned on firm
performance has on a changes in cash and total executive compensation using the general
method of Jensen and Murphy (1990) to compute pay-for-performance sensitivity
ltltINSERT TABLE V gtgt
Table V presents the effect of institutional influence on the change in cash
compensation using the HS size control measure throughout12
Model (1) replicates the
specification in HS for the period of 1992‒1997 and we produce a similar result ndash the
interaction of lagged institutional concentration and contemporaneous change in
12 This test uses our data for the two periods 1992‒1997 and 1992‒2002 We do not use the HS data for
cash compensation as data for bonus is missing in the data supplied to us
18
shareholder wealth positively and significantly relates to a change in executive cash
compensation In model (2) we include the contemporaneous change in shareholder
wealth and the lagged institutional concentration as stand-alone controls as indicated by
Greenersquos (2000) specification of equation (1) above and the interaction term remains
significant In model (3) we interact lagged institutional concentration with the lagged
change in shareholder wealth as in equation (2) above that tries to take account of what
participants could have observed at the time The interaction term is then negative but
insignificant altering the HS result Model (4) provides a similar result where our
preferred proxy Top 5 Ownership captures concentrated institutional influence In
models (5) to (8) we use our full sample In model (5) we find that the lagged
institutional concentration interaction with the contemporaneous change in shareholder
wealth is positive and significant in HSrsquos specification However including change in
contemporaneous shareholder wealth and lagged institutional concentration as separate
control variables as suggested by equation (1)rsquos specification removes the significance
from the interaction term in model (6) Moreover in model (7) when lagged institutional
concentration is interaction with the lagged change in shareholder wealth in accord with
equation (2) this interaction term is now negative and significant reversing the HS
finding We obtain a similar result in model (8) when we measure institutional
concentration influence using Top 5 Ownership13
V The Levels of Executive Fixed and Total Compensation as Functions of
Institutional Investor Influence
13 In unreported results we use the log of market capitalization as a size control These changes to the size
control do not dramatically change the findings from Table V The reason why this size issue is not as
important as in the findings for option grants pay-for-performance is that the change in cash compensation
is not so strongly influenced by firm size In unreported results available on request we also examine the
effect of institutional influence on changes in total compensation with similar results to the change in cash
compensation
19
Two measures of compensation are examined ndash salary and total compensation
ltltINSERT TABLE VI gtgt
Table VI examines the effect of institutional influence on the level of salary In model
(1) we use the HS sample and specification to obtain almost identical results The
coefficient on institutional concentration is (negative) $247509 that is almost identical to
(negative) $247967 in HS Both are highly significant In model (2) we make a single
change to the model ndash we use the natural logarithm of market capitalization This single
change reverses the sign of institutional concentration with the coefficient still being
significant at the 1 percent level In models (3) and (4) we measure institutional influence
using Top 5 Ownership In model (3) where we measure firm size using the level of
market capitalization Top 5 Ownership is insignificant When we use the natural
logarithm of market capitalization in model (4) Top 5 Ownership is significant and
positively relates to the salary level In models (5) to (8) we estimate the same models
using our full sample We find similar results for the full sample as for the period 1992ndash
1997
ltltINSERT TABLE VII gtgt
Table VII examines the effect of institutional influence on total compensation In
model (1) we replicate the HS methodology using the HS data to obtain a similar
significant negative relation between institutional concentration and total compensation of
(negative) $1608633 However after changing to a logarithmic specification of market
capitalization in model (2) institutional concentration increases total compensation by a
significant $32400014
Importantly when institutional influence is measured using Top 5
14 = (5688751 times 0101)
20
Ownership an insignificant relation exists with total compensation in model (3) using the
HS size control but it becomes both positive and significant in (4) with the logged size
control In models (5) and (6) where models are estimated using the full sample we
observe a similar reversal in sign as occurs in models (1) and (2) However now the
magnitude of the reversal is almost equal with a difference of about $67m that is very
large indeed In model (5) the coefficient on institutional concentration is (negative)
$342m and then in model (6) the coefficient is $33m Top 5 Ownership is insignificant
in both models (7) and (8) unlike the results for Salary
VI Robustness
Two issues are examined in is this Section ndash reverse causality and the appropriate
transformation of market capitalization A methodological issue inherent in examining the
effect of ownership-based measures of institutional influence on executive compensation
is endogeneity driven by reverse causality We utilize an instrumental variables approach
where orthogonalized institutional influence measures are regressed on each of the
compensation measures lagged option grant pay-for-performance sensitivity salary and
total compensation In addition market capitalization and a CEO dummy are included in
each regression We use the residual from each regression as an instrument for
institutional influence Such a process removes any preferences for compensation from
the institutional influence measures
ltltINSERT TABLES VIII AND IXgtgt
Table VIII presents the results where we instrument the HS measure lagged
institutional concentration with a compensation-orthogonalized equivalent The analysis
indicates that the instrumented institutional concentration measure is positively and
significantly related to the level of salary and total compensation and insignificantly
21
related to option grant pay-for-performance sensitivity (PPSOG) Table IX presents
results where we instrument Top 5 Ownership with a compensation-orthogonalized
equivalent The instrumented Top 5 Ownership measure insignificantly relates to option
grant pay-for-performance sensitivity and barely negatively relates to salary although it
negatively relates to total compensation suggesting reverse causality is driving a
significant portion of our previous results
ltltINSERT TABLE Xgtgt
Another issue examined in this section is the appropriate transformation of market
capitalization Aneuryn-Evans and Deaton (1980) provide an evaluation of methods
comparing linear and log-linear specifications We use the Box-Cox test to compare
logarithm versus linear specifications In applying the Box-Cox test we specify
transformation of only market capitalization ndash the dependent variable and all other
independent variables remain unchanged Table X presents the lambda coefficient for
each model and the log likelihood score In all models in Panel A and B a lambda of zero
receives a higher (closer to zero) log likelihood score suggesting the log specification is
appropriate
VII Conclusion
Hartzell and Starks (2003) appears to document an effective role for institutional
investors in increasing executive incentives and lowering compensation levels Our paper
re-addresses the issue of concentrated institutional monitoring from a perspective
informed by an additional decade of research since their paper appeared We cannot find
evidence that concentrated institutional investors really have any positive (or negative for
that matter) effect on executive incentive levels While superficially we find evidence
that they may raise executive pay levels this is most likely driven by an institutional
22
preference for highly remunerated executives in large companies and thus by reverse
causality Moreover our findings are not consistent with the rent-extraction hypothesis
suggesting that firms systematically overpay executives in the absence of effective
monitoring by outside parties such as institutional investors
23
References
Almazan Andres Jay C Hartzell and Laura T Starks 2005 ldquoActive Institutional
Shareholders and Costs of Monitoring Evidence from Executive Compensationrdquo
Financial Management Winter 5ndash34
Aneuryn-Evans Gwyn and Angus Deaton 1980 ldquoTesting Linear versus Logarithmic
Regression Modelsrdquo Review of Economic Studies 47 (1) Econometrics Issue 275ndash
291
Babenko Ilona 2009 ldquoShare Repurchases and Pay-Performance Sensitivity of Employee
Compensation Contractsrdquo Journal of Finance 64117-150
Baker George P and Brian J Hall 2004 ldquoCEO Incentives and Firm Sizerdquo Journal of
Labor Market Economics 22 767ndash798
Bebchuk Lucian A Jesse M Fried and David Walker 2002 ldquoManagerial power and
rent extraction in the design of executive compensationrdquo University of Chicago Law
Review 69 751ndash846
Bebchuk Lucian A and Jesse M Fried 2003 ldquoExecutive Compensation as an Agency
Problemrdquo Journal of Economic Perspectives 17 71ndash92
Bebchuk Lucian A and Jesse M Fried 2004 Pay without Performance The Unfulfilled
Promise of Executive Compensation Cambridge MA Harvard University Press
Bertrand Marianne and Sendhil Mullainathan 2001 ldquoAre CEOs Rewarded for Luck
The Ones Without Principals Arerdquo Quarterly Journal of Economics 116 901ndash32
Black Fischer and Myron Scholes 1973 ldquoThe Pricing of Options and Corporate
Liabilitiesrdquo Journal of Political Economy 81 637ndash54
Cadman Brian Sandy Klasa and Steve Matsunaga 2010 ldquoDeterminants of CEO Pay A
Comparison of ExecuComp and Non-ExecuComp Firmsrdquo The Accounting Review 85
1511ndash1543
Chhaochharia Vidhi and Yaniv Grinstein 2009 ldquoCEO Compensation and Board
Structurerdquo Journal of Finance 64 231ndash261
Cichello Michael S 2005 ldquoThe impact of firm size on pay-performance sensitivitiesrdquo
Journal of Corporate Finance 11 609ndash627
Davis Gerald F and E Han Kim 2007 ldquoBusiness ties and proxy voting by mutual
fundsrdquo Journal of Financial Economics 85 552ndash570
Dikolli Shane S Susan L Kulp and Karen L Sedatole 2009 ldquoTransient Institutional
Ownership and CEO Contractingrdquo Accounting Review 84 737ndash770
Edmans Alex Xavier Gabaix and Augustin Landier 2009 ldquoA Multiplicative Model of
Optimal CEO Incentives in Market Equilibriumrdquo Review of Financial Studies 22
4881ndash4917
Edmans Alex 2009 ldquoBlockholder Trading Market Efficiency and Managerial Myopiardquo
Journal of Finance 64 2481ndash2513
Ertimur Yonca Fabrizio Ferri and Volkan Muslu 2011 ldquoShareholder Activism and
CEO Payrdquo Review of Financial Studies 24 535ndash592
24
Gabaix Xavier and Augustin Landier 2008 ldquoWhy Has CEO Pay Increased So Muchrdquo
Quarterly Journal of Economics 123 49ndash100
Garen John 1994 ldquoExecutive compensation and principal-agent theoryrdquo Journal of
Political Economy 102 1175ndash1199
Gibbons Robert and Kevin J Murphy 1992 ldquoOptimal Incentive Contracts in the
Presence of Career Concerns Theory and Evidencerdquo Journal of Political Economy
100 468ndash505
Greene William H 2000 (4th
Edition) Econometric Analysis Prentice-Hall NJ
Guthrie Katherine Jan Sokolowsky and Kam-Ming Wan 2012 ldquoCEO Compensation
and Board Structure Revisitedrdquo Journal of Finance Forthcoming
Hadlock Charles R and Gerald B Lumer 1997 ldquoCompensation turnover and top
management incentives historical evidencerdquo Journal of Business 70 153ndash187
Hartzell Jay C and Laura T Starks 2003 ldquoInstitutional Investors and Executive
Compensationrdquo Journal of Finance 58 2351ndash2374
Huson Mark R Robert Parrino and Laura T Starks 2001 ldquoInternal Monitoring
Mechanisms and CEO Turnover A Long-Term Perspectiverdquo Journal of Finance 56
2265ndash2297
Jensen Michael C and Kevin J Murphy 1990 ldquoPerformance Pay and Top Management
Incentivesrdquo Journal of Political Economy 98 225ndash264
Jin Li 2002 ldquoCEO compensation diversification and incentivesrdquo Journal of Financial
Economics 66 29ndash63
Kang Qiang and Qiao Liu 2008 ldquoStock Market Information Production and Executive
Incentivesrdquo Journal of Corporate Finance 14 484ndash498
Merton Robert 1973 ldquoTheory of Rational Option Pricingrdquo Bell Journal of Economics
and Management Science 4 141-183
Morse Adair Vikram Nanda and Amit Seru 2011 ldquoAre Incentive Contracts Rigged by
Powerful CEOsrdquo Journal of Finance 66 1779ndash1821
Murphy Kevin J 1985 ldquoCorporate performance and managerial remuneration An
empirical analysisrdquo Journal of Accounting and Economics 7 11ndash42
Murphy Kevin J 1999 ldquoExecutive compensationrdquo in Orley Ashenfelter and David
Card Handbook of Labor Economics vol 3 Elsevier
Rosen Sherwin 1992 ldquoContracts and the market for executivesrdquo In Contract
Economics ed Lars Werin and Hans Wijkander 181ndash211 Cambridge MA
Blackwell
Rothberg Burton and Steven Lilien 2006 ldquoMutual Funds and Proxy Voting New
Evidence on Corporate Governancerdquo Journal of Business and Technology Law 1
157ndash184
Schaefer Scott 1998 ldquoThe dependence of payndashperformance sensitivity on the size of the
firmrdquo Review of Economics and Statistics 80 436ndash443
Yermack David 1995 ldquoDo Corporations Award CEO Stock Options Effectivelyrdquo
Journal of Financial Economics 39 237ndash69
25
Table I Univariate Descriptive Statistics
This table reports sample statistics for the key variables used in this study based on our comprehensive
ExecuComp dataset Institutional Concentration is the shares owned by the institutions with the five largest
holdings divided by the shares held by all institutions Top 5 Ownership is the proportion of shares
outstanding held by the institutions with the five largest holdings Total Institutional Ownership is the
proportion of shares held by all institutions Panel B shows the measures executive compensation The pay-
for-performance sensitivity (PPS) of option grants shows the change in executive wealth due to a $1000
change in firm in shareholder wealth Expressions for Salary and Total Compensation are in thousands of
dollars while Market Capitalization is in millions of dollars Panel C shows firm characteristics
Variable Mean Median Std Dev P10 P90
Panel A Institutional Investor Holdings
Institutional Concentration 0450 0424 0151 0282 0656
Top 5 Institutional Ownership 0229 0220 0101 0110 0355
Total Institutional Ownership 0546 0561 0206 0258 0801
Panel B Executive Compensation
PPS of Option Grants 1161 0156 3829 0 2822
Salary 36311 29462 25231 15179 67000
Total Compensation 258313 85878 980701 23875 511022
Panel C Firm Characteristics
Tobins Q 1744 1139 2738 0495 3322
Market Capitalization 560699 112094 1988108 19754 1048754
26
Table II Pairwise Correlation Coefficients
This table presents the correlations between the key variables based on our comprehensive
ExecuComp dataset Institutional Concentration is the shares owned by the institutions with the five
largest holdings divided by the shares held by all institutions Top 5 Ownership is the proportion of
shares outstanding held by the institutions with the five largest holdings Total Institutional
Ownership is the proportion of shares held by all institutions The pay-for-performance sensitivity
(PPS) of option grants is the change in executive wealth due to a $1000 change in shareholder
wealth
(1) (2) (3) (4) (5) (6) (7) (8)
(1) Institutional Concentration 1
(2) Top 5 Ownership 0268 1
(3) Total Ownership -0415 0706 1
(4) PPS of Option Grants 0122 0042 -0024 1
(5) Log Salary -0270 -0012 0171 0038 1
(6) Log Market Capitalization -0630 -0157 0261 -0156 0420 1
(7) Level Market Capitalization -0241 -0157 0007 -0058 0219 0534 1
(8) Tobins Q -0008 -0034 0002 0015 -0083 0066 0121 1
27
Table III
Descriptive Statistics by Market Capitalization Deciles
Market
Capitalization
Deciles
Market
Capitalization
($m)
Total
Compensation
($000)
Salary
($)
PPS Option
Grants
($)
Institutional
Concentration
()
Top 5 Ownership
()
Institutional
Total Ownership
()
1 (Small) 1181 76008 234700 2720 611 200 327
2 2668 90206 257955 1796 495 217 438
3 4193 105946 275120 1517 454 220 485
4 6045 129892 296067 1376 420 206 491
5 8742 155391 315864 1216 394 198 502
6 13086 191888 345942 1072 380 202 531
7 20790 239501 377021 0874 368 196 532
8 34920 314955 427808 0687 354 190 538
9 67747 450970 480319 0569 328 173 528
10 (Big) 366540 846066 614354 0297 275 138 502
28
Table IV Pay-for-Performance Sensitivity of Option Grants as a Function of Institutional Influence
Similar in construction to parts of HS (Table II) this table shows the coefficients from Tobit regressions of option grant pay-for-performance sensitivity per $1000 change in
shareholder wealth against the lagged change and contemporary change in shareholder wealth measures of institutional influence Tobinrsquos Q firm market capitalization and
dummy controls for CEO and year effects We estimate models (1) to (4) over the period 1992‒1997 using data supplied by HS and models (5) to (8) are estimated over the
whole sample Models (1) (2) (5) and (6) use Institutional Concentration as a measure of institutional influence which HS define as the shares owned by the institutions
with the five largest holdings divided by the shares held by all institutions These models control for total institutional ownership Models (3) (4) (7) and (8) use Top 5
Ownership as a measure of institutional influence which we define as the proportion of shares outstanding held by the institutions with the five largest holdings Coefficients
for industry and year controls are not presented One two and three asterisks denote significance at 10 5 and 1 levels
Hartzell and Starks Database 1992‒1997 Full Sample from ExecuComp 1992‒2002
(1) (2) (3) (4) (5) (6) (7) (8)
Change in Shareholder Wealth 77721 191781 89874 137368 -446 -218 -664 -203
(171) (429) (194) (317) (319) (168) (453) (158)
Lag1 Change in Shareholder Wealth 14446 99963 88719 8544 -144 0273 -0342 00852
(027) (212) (166) (184) (071) (015) (016) (005)
Lag1 Institutional Concentration 13994 -00692 26405 03304
(767) (032) (1402) (166)
Lag1 Total Institutional Ownership 08055 10434 07106 0937
(593) (781) (666) (898)
Lag1 Top 5 Institutional Ownership 10567 07457 12116 06293
(399) ( 283) (533) (278)
Lag1 Tobinrsquos Q 0068 00877 00647 00813 00498 00727 00508 00695
(337) (434) (319) (401) (504) (734) (51) (703)
Lag1 Market Capitalization -106202 -173515 -776 -117
(474) (742) (1011) (153)
Lag1 Log Market Capitalization -0297 -02319 -04111 -03908
(1576) (1439) (2997) (2841)
CEO Dummy 13745 13682 13765 13674 28197 28128 28155 28143
(1913) (1918) (1914) (1916) (3388) (3396) (3377) (34)
Number of Observations 33878 33878 33878 33878 97679 97679 97679 97679
29
Table V Pay-for-Performance Sensitivity of Cash Compensation (Salary plus Bonus) as a Function of Institutional Influence - Level Market Capitalization
Similar in parts to models (1) and (2) of HS (Table IV) this table displays the coefficients from OLS regressions of the change in cash compensation against the lagged
change and contemporary change in shareholder wealth measures of institutional influence Tobinrsquos Q firm market capitalization and dummy controls for CEO and year
effects We estimate models (1) to (4) over the period 1992‒1997 based on our ExecuComp sample and models (5) to (8) over the whole sample Models (1) (2) (5) and (6)
use Institutional Concentration as a measure of institutional influence which HS define as the shares owned by the institutions with the five largest holdings divided by the
shares held by all institutions These models control for total institutional ownership Models (4) and (8) use Top 5 Ownership as a measure of institutional influence which
we define as the proportion of shares outstanding held by the institutions with the 5 largest holdings Coefficients for industry and year controls are not presented One two
and three asterisks denote significance at 10 5 and 1 levels
Our Database based on ExecuComp 1992‒1997 Full Sample based on ExecuComp 1992‒2002
(1) (2) (3) (4) (5) (6) (7) (8)
Lag1 Change in Shareholder Wealth 00014 00017 00088 00089 -00006 -00007 00072 00086
(051) (062) (129) (103) (027) (029) (131) (16)
Change in Shareholder Wealth -00102 00102 00105 00032 00102 00101
(097) (292) (293) (055) (611) (608)
Change Shareholder Wealth Lag1 Institutional
Concentration
00435 00764 00363 00253
(463) (258) (689) (144)
Lag1 Change Shareholder Wealth Lag1
Institutional Concentration
-00197 -00283
(122) (197)
Lag1 Institutional Concentration -492182 -288861 400149 453623
(09) (056) (098) (114)
Lag1 Total Institutional Ownership 720688 538372 655091 559587 682069 726226
(276) (232) (281) (314) (308) (322)
Lag1 Change Shareholder Wealth Lag1 Top 5
Institutional Ownership
-0046 -00689
(091) (256)
Lag1 Top 5 Institutional Ownership 583044 1191634
(092) (25)
Lag1 Tobinrsquos Q -32761 -36806 -19422 -18023 -54283 -48681 -29534 -24243
(138) (172) (077) (078) (227) (245) (136) (111)
Lag1 Market Capitalization 00004 00004 00002 00006 00007 00008 00007 00007
(042) (044) (022) (07) (086) (092) (082) (09)
CEO Dummy 768035 765474 768334 773634 616315 61729 619028 618535
(354) (355) (356) (355) (417) (418) (419) (418)
Adjusted R-Sq 00057 0006 00052 0005 00097 00097 00098 00102
Number of Observations 41305 41305 41305 41305 85518 85518 85518 85518
30
Table VI Level of Salary as a Function of Institutional Influence
Similar in parts to models (1) and (2) of HS (Table V) this table displays the coefficients from OLS regressions of salary against the lagged change and contemporary change in
shareholder wealth measures of institutional influence Tobinrsquos Q firm market capitalization and dummy controls for CEO industry and year effects We estimate models (1) to
(4) over the period 1992‒1997 using the actual data supplied by HS and models (5) to (8) over the whole sample 1992‒2002 Models (1) (2) (5) and (6) use Institutional
Concentration as a measure of institutional influence which HS define as the shares owned by the institutions with the five largest holdings divided by the shares held by all
institutions These models control for total institutional ownership Models (3) (4) (7) and (8) use Top 5 Ownership as a measure of institutional influence which we define as
the proportion of shares outstanding held by the institutions with the 5 largest holdings Coefficients for industry and year controls are not presented One two and three
asterisks denote significance at 10 5 and 1 levels
Hartzell and Starks Database 1992‒1997 Full Sample based on ExecuComp 1992‒2002
(1) (2) (3) (4) (5) (6) (7) (8)
Change in Shareholder Wealth 04125E-4 0003 -2439E-4 0003 00021 00012 00024 00012
(057) (464) (031) (468) (1004) (72) (1041) (706)
Lag1 Change in Shareholder Wealth -00062 7856E-4 -0007 9003E-4 -00014 00002 -00014 00003
(594) (098) (592) (112) (522) (123) (5) (145)
Lag1 Institutional Concentration -2475087 555134 -2734983 943317
(3245) (710) (4965) (1671)
Lag1 Total Institutional Ownership 1020109 217488 936337 252464
(1904) (432) (2471) (704)
Lag1 Top 5 Institutional Ownership -148355 31354 -87926 423166
(139) (332) (118) (643)
Lag1 Tobinrsquos Q -195348 -252191 -192344 -25158 -119837 -146373 -122132 -143821
(2812) (3762) (2721) (3752) (248) (274) (2583) (2727)
Lag1 Market Capitalization 844E-06 103E-5 00034 00039
(2781) (3051) (2874) (3104)
Lag1 Log Market Capitalization 750021 723625 776657 732279
(7511) (9285) (11224) (14039)
CEO Dummy 2683978 2702238 2668244 2702417 3281133 3285103 3292175 3283145
(8625) (9277) (8246) (9271) (12739) (13616) (12336) (13605)
Adjusted R-Sq 04604 05302 04187 05295 03856 04588 03476 04574
Number of Observations 36280 36280 36280 36280 97679 97679 97679 97679
31
Table VII Level of Total Direct Compensation as a Function of Institutional Influence
Similar in parts to models (3) and (5) of HS (Table V) this table displays the coefficients from OLS regressions of total direct compensation against the lagged change and
contemporary change in shareholder wealth measures of institutional influence Tobinrsquos Q firm market capitalization and dummy controls for CEO industry and year effects
We estimate models (1) to (4) over the period 1992‒1997 and models (5) to (8) over the whole sample Models (1) (2) (5) and (6) use Institutional Concentration as a measure
of institutional influence which HS define as the shares owned by the institutions with the five largest holdings divided by the shares held by all institutions These models
control for total institutional ownership Models (3) (4) (7) and (8) use Top 5 Ownership as a measure of institutional influence which we define as the proportion of shares
outstanding held by the institutions with the five largest holdings Coefficients for industry and year controls are not presented One two and three asterisks denote significance
at 10 5 and 1 levels
Hartzell and Starks Database 1992‒1997 Full Sample from ExecuComp 1992‒2002
(1) (2) (3) (4) (5) (6) (7) (8)
Change in Shareholder Wealth 005779 007006 005282 006791 00423 00148 0046 00138
(415) (581) (369) (565) (203) (07) (22) (066)
Lag1 Change in Shareholder Wealth 003182 00714 00272 00715 00613 0126 00614 01278
(126) (298) (107) (299) (275) (488) (276) (494)
Lag1 Institutional Concentration -1608633 3248998 -3424892 3334186
(151) (283) (1172) (118)
Lag1 Total Institutional Ownership 104013 5473128 1531013 -2926764
(1645) (831) (876) (158)
Lag1 Top 5 Institutional Ownership 188585 3927986 -2159177 -2640879
(016) (343) (067) (092)
Lag1 Tobinrsquos Q -445692 -803612 -454859 -833566 3257027 2971201 3213936 3086205
(427) (791) (432) (831) (83) (77) (815) (793)
Lag1 Market Capitalization 49E-5 615E-5 01031 01088
(113) (1365) (1555) (168)
Lag1 Log Market Capitalization 4683497 4759377 1645156 1427856
(2775) (3469) (3679) (4239)
CEO Dummy 1501307 1508886 1493688 1508736 4713079 4713629 47278 4703908
(2648) (2684) (2618) (2683) (282) (2828) (2819) (2824)
Adjusted R-Sq 01861 02060 01695 02051 01071 01113 01033 011
Number of Observations 33229 33229 33229 33229 97679 97679 97679 97679
32
Table VIII Two Stage Least Squares (Reverse Causality) Tests - Executive Compensation Measures as a Function of Institutional Concentration
This table shows the coefficients from the Two-Stage Least Squares (2SLS) regressions of executive compensation measures against the lagged change
and contemporary change in shareholder wealth instrumented institutional concentration and total institutional ownership Tobinrsquos Q the natural
logarithm of firm market capitalization and dummy variable controls for CEO and year effects We instrument measures of institutional concentration
and total institutional ownership by orthogonalizing with respect to option grant pay-for-performance sensitivity salary and total compensation
Coefficients for industry and year controls are not presented First stage results are not presented One two and three asterisks denote significance at
10 5 and 1 levels
ExecuComp Sample 1992‒1997 Full Sample from ExecuComp 1992‒2002
PPSOG Salary TC PPSOG Salary TC
Change in Shareholder Wealth 144E-05 0004 02133 -409E-06 00011 00157
(308) (775) (596) (691) (638) (068)
Lag1 Change in Shareholder Wealth -240E-05 -00004 00866 118E-06 00002 01422
(131) (087) (346) (097) (116) (51)
Lag1 Institutional Concentration -00898 668681 3263143 02024 916656 2856417
(052) (801) (176) (163) (1545) (1004)
Lag1 Total Institutional Ownership 01021 -43997 2512587 02983 -104171 -1028347
(09) (084) (153) (4) (269) (487)
Lag1 Tobinrsquos Q 00754 -225606 119297 00655 -158657 2997205
(596) (1649) (053) (834) (2535) (727)
Lag1 Log Market Capitalization -04234 828713 7864046 -04002 825106 1685638
(2195) (8132) (2977) (4106) (11133) (373)
CEO Dummy 15848 3214117 2950274 19149 3205862 4529101
(2223) (9389) (2255) (3802) (1331) (2758)
Overall R-Sq -- 04981 01206 -- 04738 01178
Number of Observations 41305 41305 41305 85518 85518 85518
33
Table IX Two Stage Least Squares (Reverse Causality) Tests - Executive Compensation Measures as a Function of Top 5 Ownership
This table shows the coefficients from 2SLS regressions of executive compensation measures against the lagged change and contemporary change
in shareholder wealth instrumented Top 5 institutional ownership Tobinrsquos Q natural logarithm of firm market capitalization and dummy variable
controls for CEO and year effects We instrument Top 5 institutional ownership by orthogonalizing with respect to option grant pay-for-
performance sensitivity salary and total compensation Coefficients for industry and year controls are not presented First stage results are not
presented One two and three asterisks denote significance at 10 5 and 1 levels
ExecuComp sample 1992‒1997 Full Sample 1992‒2002
PPSOG Salary TC PPSOG Salary TC
Change in Shareholder Wealth 136E-05 00041 02128 -407E-06 00011 00146
(293) (79) (596) (692) (619) (063)
Lag1 Change in Shareholder Wealth -241E-05 -00003 00868 116E-06 00003 01439
(133) (07) (347) (096) (142) (514)
Lag1 Top 5 Institutional Ownership -03774 -187512 -3848531 00179 -104261 -126659
(163) (186) (166) (011) (149) (436)
Lag1 Tobinrsquos Q 00731 -223886 10074 00642 -155467 3115073
(573) (1631) (045) (819) (2509) (75)
Lag1 Log Market Capitalization -0416 775637 7696302 -04015 764632 1462536
(2345) (10084) (3766) (3688) (13708) (4072)
CEO Dummy 15854 3211991 2950158 19152 3203152 4519385
(2223) (9381) (2254) (3804) (133) (2752)
Overall R-Sq -- 04967 01202 -- 04722 01168
Number of Observations 41305 41305 41305 85518 85518 85518
34
Table X Specification Tests utilizing Box-Cox Transformation of Market Capitalization
This table presents results from a Box-Cox transformation of market capitalization In
each model we specify transformation of market capitalization only Panel A shows
results for the period 1992‒1997 and Panel B shows results for the entire period
1992‒2002 The first part of each panel shows the Lambda coefficient that applies to
the transformed market capitalization The second part of each panel shows log
likelihood scores with LR Chi2 Statistics in parentheses ‒ these test for the
appropriateness of a linear or log model specification
Panel A Our ExecuComp sample 1992‒1997
(1) (2) (3)
PPSOG Salary TC
Lambda -05143 01881 02876
(-1914) (2413) (1329)
Test H0 (Log Likelihood Score)
Lambda = -1 -1279748 -3178575 -4795326
(23866) (85952) (99688)
Lambda = 0 -1280152 -3138456 -4791159
(31934) (57147) (16339)
Lambda = 1 -1283157 -3163807 -4793506
(92039) (564156) (63292)
Panel B Full ExecuComp sample 1992‒2002
(1) (2) (4)
PPSOG Salary TC
Lambda -044 01853 04036
(-3015) (40) (3666)
Test H0 (Log Likelihood Score)
Lambda = 0 -2688334 -6476494 -10384490
(66599) (160356) (126236)
Lambda = 1 -2694886 -6538506 -10386795
(197644) (1400605) (172345)
35
Figure 2 Institutional Ownership Measures (LHA) vs Total
Compensation (RHA) by Market Capitalization
0
10
20
30
40
50
60
70
1 2 3 4 5 6 7 8 9 10
$00
$10
$20
$30
$40
$50
$60
$70
$80
$90
Millio
ns
InstitutionalConcentration (LHS)Top5 Ownership (LHS)
Institutional TotalOwnership (LHS)Total Compensation(RHS)
Page 10
10
Black-Scholes option methodology This may introduce a downward pay bias since self-
reported option-grant pay is subject to possible understatement and there are 13839
observations with likely high option pay dropped altogether
Endogeneity
HSrsquos endogeneity testing raises concerns about reverse causality Institutions may
prefer to hold firms with certain compensation structures such as incentivized executives
HS rightly highlight this concern and use instrumental variables to address this issue HS
assume that institutional ownership reflects any reverse causality Using an instrumental
variables approach to resolve this problem involves finding a variable that influences or
explains institutional ownership but not directly executive compensation They use share
turnover in a firm as an instrument for institutional ownership However theoretical work
by Holmstrom and Tirole (1993) indicates that measures of stock liquidity and trading
activity are indeed important determinants of executive market-based incentives as Kang
and Lui (2008) and Dikolli Kulp and Sedatole (2009) find empirically As such one can
make a case for adopting another instrument rather than share turnover for institutional
ownership
II Previous Research
Prior research fails to address issues concerning possible fragility of the HS results
AHS replicate the HS findings on both incentives and pay levels using the same dataset as
HS and they apply the same definition of institutional concentration and use market
capitalization expressed as a level Cadman Klasa and Matsunaga (2010 Table 6) set out
to replicate HSrsquos (Table II) model of option grant pay-for-performance sensitivity when
they compare a dataset of ExecuComp and Non-ExecuComp firms for the period 2000-
2007 Their ExecuComp results for this sample period are quite similar to our results for
11
the 1992‒2002 period for the same specification controlling for the logarithm of market
capitalization but they do not regard these results as a test of HS as their focus is entirely
on comparing the two groups of stocks Nor do they investigate the impact of
concentration on either cash compensation or total pay Kang and Liu (2008) examine
pay-performance sensitivity use a version of the HS institutional concentration measure
as a control variable in a different framework deriving from Holmstrom and Tirole
(1993) Dikolli Kulp and Sedatole (2009) argue that firms with transient ownership will
issue relatively more equity incentives to counter short-termism with their results robust
to the inclusion of the HS concentration variable
We hoped to find supportive evidence from proxy voting behavior in support of HSs
findings However evidence from actual mutual fund proxy voting records suggests that
most oppose reductions in CEO pay but overall the evidence is mixed8 Davis and Kim
(2005) examine proxy votes of mutual funds for 2004 and find that for a sample of
Fortune 1000 proxy contests in 45 cases that attempted to limit executive pay mutual
funds were unanimously opposed to such attempts Rothberg and Lilien (2006) also
examine proxy-voting data and find that mutual funds voted 66 percent of the time in
managementrsquos favor on issues of executive compensation Ertimur Ferri and Muslu
(2011) find that institutional investors lack sophistication in that they do not target
excessively paid CEOs relative to CEOs with high predicted-pay based on economic
determinants but exert a moderating influence ndash a $23 million reduction ndash for those they
deem ldquoexcessively paidrdquo CEOs
III Data and Methodology
8 We do acknowledge that the HS results are for all institutional investors while the transparency with
respect to proxy voting applies only to mutual funds
12
This Section discusses the data used in the study the construction of variables
descriptive statistics and the methodology Two datasets are used ndash one is the original
dataset generously supplied by Jay Hartzell and Laura Starks which covers 1992‒1997
second a more comprehensive dataset covering 1992‒2002 is computed from the same
data sources used in the HS dataset This is Standard and Poorrsquos (SampP) ExecuComp for
compensation data Thomson Financial CDASpectrum database for institutional holdings
to compute measure of institutional influence SampPrsquos Compustat and the Center for
Research in Stock Prices (CRSP) for firm level accounting and stock price data Our more
comprehensive dataset contains 2559 firms compared with 1914 for the HS dataset Our
full matched sample yields 97679 executive years
A Institutional Influence Measures
To measure institutional influence we employ two ownership-based metrics The first
measure of institutional influence is HSrsquos Institutional Concentration We use this for
direct comparison to the HS analysis The second measure of institutional influence
changes the denominator of the HS measure from institutional shares held to total shares
outstanding in the firm with nomenclature Top 5 Ownership9 We compute both
measures at the fiscal year end of the firm We adjust all compensation measures and
asset prices for inflation to 30 June 2002 prices for our comprehensive sample compared
to the HS dataset that is in the dollars of the day
B Compensation Measures
The ExecuComp database reports several raw measures of executive compensation for
the top five executives (where available) including total salary bonus long-term
9 Top 5 Ownership is included in the HS database but not utilized by them in reported tests as well as
constructed by us for the more comprehensive sample
13
incentives and number of option grants For our more comprehensive dataset we replicate
all the compensation measures used by HS In common with HS we compute the pay-for-
performance sensitivity of option grants using the Yermack (1995) approach This
measure shows the dollar change in value of the executive options for every dollar change
in the value of the firm This involves calculating the option delta using the Black-
Scholes formula adjusted for dividends10
The ratio of the number of shares represented
by the award of options divided by the diluted number of shares outstanding at the
beginning of the year multiplies the option delta with units given by a $1000 change in
firm value Summing salary bonus long-term incentive plan payouts and stock and
option grants provides total direct compensation Options are valued using the Black and
Scholes (1973) formula for European call options adjusted by Merton (1973) to
incorporate dividends
C Descriptive Statistics
Table I presents descriptive statistics of key variables employed in the study according
to our comprehensive dataset11
Institutions own on average 546 percent of the firms in
which they invest This is almost identical to the 531 percent found by HS The
institutions with the five largest holdings account 45 percent of institutional ownership
Once again the figure for HS is almost identical at 44 percent The institutions with the
five largest holdings own on average 229 percent of shares outstanding in a firm (HS
have 223 percent) Average total compensation for executives is $2583 million
considerably larger than that documented by HS ($1250 million) but the HS figure is an
10 When there is more than one grant during the year a weighted average delta is used
11 With the data supplied to us by Hartzell and Starks we are able to replicate exactly their summary data
(HS Table I) for all but cash compensation Salary is included but bonus payments are missing The
replicated table is available on request
14
amalgam of price levels over their sample period 1992‒1997 and ours are in consistent
dollars of June 30 2002 However there has also been significant growth in
compensation levels from 1998 to 2002
ltltINSERT TABLE Igtgt
Based on our comprehensive data in Table II we examine correlations between the
key variables employed in the study Of concern is the large negative correlation of 63
percent which has been discussed previously (Section IIA) between the HS measure of
institutional concentration and the natural logarithm of firm market capitalization
ltltINSERT TABLE IIgtgt
In view of this high correlation Table III provides a breakdown of the three key pay
variables and three institutional ownership variables by size deciles We find that the high
institutional concentration but low institutional ownership share of total shareholdings
mentioned above is evident in the smallest decile of firms by size This decile has low
salary and even lower total compensation in a relative sense but by far the highest pay-
for-performance sensitivity of option grants Small firms award executives a far higher
proportion of market capitalization Since small firm market-capitalization is negligible in
comparison with the highest decile (300 times larger in magnitude) total compensation is
very small in absolute value Moreover due to the domination of the pay-size effect base
salaries are also exceedingly small
ltltINSERT TABLE IIIgtgt
Figure 1 based on Table III shows how as institutional concentration falls with
increases in stock size (but also as the total share of institutional ownership increases)
and the pay-for-performance sensitivity of option grants falls Thus because of what
appears to be a weak size control and size-influenced concentration measure HS
15
potentially pick up a size effect as their main result but not necessarily a monitoring
effect Figure 2 shows that as firm size increases up to decile 8 at least the share of
institutional total ownership increases as does total compensation
ltltINSERT FIGURES 1 AND 2 gtgt
D Methodology
Following HS we use two estimation techniques to examine the effect of institutional
investor influence on executive compensation Due to the censored nature of option grant
pay-for-performance sensitivity this compensation measure most appropriately uses
Tobit model specification whereas an Ordinary Least Squares (OLS) model specification
is used for all other compensation measures We model compensation measures as a
function of contemporaneous and lagged change in shareholder wealth institutional
influence Tobinrsquos Q and market capitalization Controls for CEO industry and year
effects are also used Augmentation occurs to test the effect of institutional influence on
the change in executive cash and total compensation with an interaction term that is the
product of institutional influence and the change in shareholder wealth
The variables used to model executive compensation have the following definitions
The lagged and contemporary measure of change in shareholder wealth is simply the
difference in market capitalization of the firm The measures of institutional influence
used are Institutional Concentration and Top 5 Ownership Use of the two types of
institutional influence occurs in separate models In order to replicate the HS
methodology models using Institutional Concentration have an additional control
consisting of lagged total institutional holdings and Tobinrsquos Q controls for performance
and growth opportunities Market capitalization of the firm is the number of shares
outstanding multiplied by the share price with inclusion both in levels and natural
16
logarithmic form A dummy variable is included to identify whether the executive is the
CEO Dummy variables are also included to control for industry effects This variable
takes a value of 1 for the two-digit Standard Industrial Classification (SIC) in which the
firm operates A year fixed effects (dummy) variable is included taking a value of 1 if the
observation was from the given year
IV Executive Incentives as a Function of Institutional Investor Influence
This section presents the results for the effect of institutional influence on option grant
pay-for-performance sensitivity and changes in compensation
A Option Grant Pay-for-Performance Sensitivity
Table IV presents the results for the effect of institutional influence on option grant
pay-for-performance sensitivity
ltltINSERT TABLE IVgtgt
Model (1) is a direct replication of HS using the HS database as supplied to us by HS
with all the variables computed in the same way and in the same functional form as with
HS The coefficient and t-statistic on lagged institutional concentration interacted with the
change in shareholder wealth is 13994 and 767 that compares to 1358 and 842 in HS
We are unable to account for these minor discrepancies for values that should be
identical In model (2) the only change to the model is that we use the natural logarithm
of market capitalization instead of the level of market capitalization This single change
has a sizeable effect on the coefficient on institutional concentration The coefficient on
institutional concentration is now -00692 and is insignificant The fact that an order
preserving transformation to a control variable can change the results is of interest The
log-transformation to market capitalization also changes the coefficient on Top 5
Ownership in models (3) and (4) The coefficient on Top 5 Ownership is significant in
17
both model (3) and in model (4) but the better size control reduces the coefficient by over
25 percent The better size control does not affect Top 5 Ownership by as much as its
design ensures less sensitivity to what could be inadequate size controls
We also test the effect of institutional influence on option grant pay-for-performance
sensitivity over the period 1992‒2002 This sample is then approximately three times as
large as the sample used by HS Institutional concentration is highly positively significant
in model (5) similar to the results for the period 1992‒1997 When we use the log of
market capitalization in model (6) the coefficient on institutional concentration remains
positive and significant at the 10 percent level with a t-statistic of 166 Of importance
though is that the coefficient on institutional concentration is one-eighth of the coefficient
in model (5) Even for a measure of institutional influence which we believe is less
sensitive to firm size the use of log market capitalization reduces the coefficient on Top 5
Ownership by approximately 50 percent from model (7) to model (8)
B Pay-for-Performance Sensitivity due to Changes in Cash and Total
Compensation
For robustness HS examine the effect that institutional influence conditioned on firm
performance has on a changes in cash and total executive compensation using the general
method of Jensen and Murphy (1990) to compute pay-for-performance sensitivity
ltltINSERT TABLE V gtgt
Table V presents the effect of institutional influence on the change in cash
compensation using the HS size control measure throughout12
Model (1) replicates the
specification in HS for the period of 1992‒1997 and we produce a similar result ndash the
interaction of lagged institutional concentration and contemporaneous change in
12 This test uses our data for the two periods 1992‒1997 and 1992‒2002 We do not use the HS data for
cash compensation as data for bonus is missing in the data supplied to us
18
shareholder wealth positively and significantly relates to a change in executive cash
compensation In model (2) we include the contemporaneous change in shareholder
wealth and the lagged institutional concentration as stand-alone controls as indicated by
Greenersquos (2000) specification of equation (1) above and the interaction term remains
significant In model (3) we interact lagged institutional concentration with the lagged
change in shareholder wealth as in equation (2) above that tries to take account of what
participants could have observed at the time The interaction term is then negative but
insignificant altering the HS result Model (4) provides a similar result where our
preferred proxy Top 5 Ownership captures concentrated institutional influence In
models (5) to (8) we use our full sample In model (5) we find that the lagged
institutional concentration interaction with the contemporaneous change in shareholder
wealth is positive and significant in HSrsquos specification However including change in
contemporaneous shareholder wealth and lagged institutional concentration as separate
control variables as suggested by equation (1)rsquos specification removes the significance
from the interaction term in model (6) Moreover in model (7) when lagged institutional
concentration is interaction with the lagged change in shareholder wealth in accord with
equation (2) this interaction term is now negative and significant reversing the HS
finding We obtain a similar result in model (8) when we measure institutional
concentration influence using Top 5 Ownership13
V The Levels of Executive Fixed and Total Compensation as Functions of
Institutional Investor Influence
13 In unreported results we use the log of market capitalization as a size control These changes to the size
control do not dramatically change the findings from Table V The reason why this size issue is not as
important as in the findings for option grants pay-for-performance is that the change in cash compensation
is not so strongly influenced by firm size In unreported results available on request we also examine the
effect of institutional influence on changes in total compensation with similar results to the change in cash
compensation
19
Two measures of compensation are examined ndash salary and total compensation
ltltINSERT TABLE VI gtgt
Table VI examines the effect of institutional influence on the level of salary In model
(1) we use the HS sample and specification to obtain almost identical results The
coefficient on institutional concentration is (negative) $247509 that is almost identical to
(negative) $247967 in HS Both are highly significant In model (2) we make a single
change to the model ndash we use the natural logarithm of market capitalization This single
change reverses the sign of institutional concentration with the coefficient still being
significant at the 1 percent level In models (3) and (4) we measure institutional influence
using Top 5 Ownership In model (3) where we measure firm size using the level of
market capitalization Top 5 Ownership is insignificant When we use the natural
logarithm of market capitalization in model (4) Top 5 Ownership is significant and
positively relates to the salary level In models (5) to (8) we estimate the same models
using our full sample We find similar results for the full sample as for the period 1992ndash
1997
ltltINSERT TABLE VII gtgt
Table VII examines the effect of institutional influence on total compensation In
model (1) we replicate the HS methodology using the HS data to obtain a similar
significant negative relation between institutional concentration and total compensation of
(negative) $1608633 However after changing to a logarithmic specification of market
capitalization in model (2) institutional concentration increases total compensation by a
significant $32400014
Importantly when institutional influence is measured using Top 5
14 = (5688751 times 0101)
20
Ownership an insignificant relation exists with total compensation in model (3) using the
HS size control but it becomes both positive and significant in (4) with the logged size
control In models (5) and (6) where models are estimated using the full sample we
observe a similar reversal in sign as occurs in models (1) and (2) However now the
magnitude of the reversal is almost equal with a difference of about $67m that is very
large indeed In model (5) the coefficient on institutional concentration is (negative)
$342m and then in model (6) the coefficient is $33m Top 5 Ownership is insignificant
in both models (7) and (8) unlike the results for Salary
VI Robustness
Two issues are examined in is this Section ndash reverse causality and the appropriate
transformation of market capitalization A methodological issue inherent in examining the
effect of ownership-based measures of institutional influence on executive compensation
is endogeneity driven by reverse causality We utilize an instrumental variables approach
where orthogonalized institutional influence measures are regressed on each of the
compensation measures lagged option grant pay-for-performance sensitivity salary and
total compensation In addition market capitalization and a CEO dummy are included in
each regression We use the residual from each regression as an instrument for
institutional influence Such a process removes any preferences for compensation from
the institutional influence measures
ltltINSERT TABLES VIII AND IXgtgt
Table VIII presents the results where we instrument the HS measure lagged
institutional concentration with a compensation-orthogonalized equivalent The analysis
indicates that the instrumented institutional concentration measure is positively and
significantly related to the level of salary and total compensation and insignificantly
21
related to option grant pay-for-performance sensitivity (PPSOG) Table IX presents
results where we instrument Top 5 Ownership with a compensation-orthogonalized
equivalent The instrumented Top 5 Ownership measure insignificantly relates to option
grant pay-for-performance sensitivity and barely negatively relates to salary although it
negatively relates to total compensation suggesting reverse causality is driving a
significant portion of our previous results
ltltINSERT TABLE Xgtgt
Another issue examined in this section is the appropriate transformation of market
capitalization Aneuryn-Evans and Deaton (1980) provide an evaluation of methods
comparing linear and log-linear specifications We use the Box-Cox test to compare
logarithm versus linear specifications In applying the Box-Cox test we specify
transformation of only market capitalization ndash the dependent variable and all other
independent variables remain unchanged Table X presents the lambda coefficient for
each model and the log likelihood score In all models in Panel A and B a lambda of zero
receives a higher (closer to zero) log likelihood score suggesting the log specification is
appropriate
VII Conclusion
Hartzell and Starks (2003) appears to document an effective role for institutional
investors in increasing executive incentives and lowering compensation levels Our paper
re-addresses the issue of concentrated institutional monitoring from a perspective
informed by an additional decade of research since their paper appeared We cannot find
evidence that concentrated institutional investors really have any positive (or negative for
that matter) effect on executive incentive levels While superficially we find evidence
that they may raise executive pay levels this is most likely driven by an institutional
22
preference for highly remunerated executives in large companies and thus by reverse
causality Moreover our findings are not consistent with the rent-extraction hypothesis
suggesting that firms systematically overpay executives in the absence of effective
monitoring by outside parties such as institutional investors
23
References
Almazan Andres Jay C Hartzell and Laura T Starks 2005 ldquoActive Institutional
Shareholders and Costs of Monitoring Evidence from Executive Compensationrdquo
Financial Management Winter 5ndash34
Aneuryn-Evans Gwyn and Angus Deaton 1980 ldquoTesting Linear versus Logarithmic
Regression Modelsrdquo Review of Economic Studies 47 (1) Econometrics Issue 275ndash
291
Babenko Ilona 2009 ldquoShare Repurchases and Pay-Performance Sensitivity of Employee
Compensation Contractsrdquo Journal of Finance 64117-150
Baker George P and Brian J Hall 2004 ldquoCEO Incentives and Firm Sizerdquo Journal of
Labor Market Economics 22 767ndash798
Bebchuk Lucian A Jesse M Fried and David Walker 2002 ldquoManagerial power and
rent extraction in the design of executive compensationrdquo University of Chicago Law
Review 69 751ndash846
Bebchuk Lucian A and Jesse M Fried 2003 ldquoExecutive Compensation as an Agency
Problemrdquo Journal of Economic Perspectives 17 71ndash92
Bebchuk Lucian A and Jesse M Fried 2004 Pay without Performance The Unfulfilled
Promise of Executive Compensation Cambridge MA Harvard University Press
Bertrand Marianne and Sendhil Mullainathan 2001 ldquoAre CEOs Rewarded for Luck
The Ones Without Principals Arerdquo Quarterly Journal of Economics 116 901ndash32
Black Fischer and Myron Scholes 1973 ldquoThe Pricing of Options and Corporate
Liabilitiesrdquo Journal of Political Economy 81 637ndash54
Cadman Brian Sandy Klasa and Steve Matsunaga 2010 ldquoDeterminants of CEO Pay A
Comparison of ExecuComp and Non-ExecuComp Firmsrdquo The Accounting Review 85
1511ndash1543
Chhaochharia Vidhi and Yaniv Grinstein 2009 ldquoCEO Compensation and Board
Structurerdquo Journal of Finance 64 231ndash261
Cichello Michael S 2005 ldquoThe impact of firm size on pay-performance sensitivitiesrdquo
Journal of Corporate Finance 11 609ndash627
Davis Gerald F and E Han Kim 2007 ldquoBusiness ties and proxy voting by mutual
fundsrdquo Journal of Financial Economics 85 552ndash570
Dikolli Shane S Susan L Kulp and Karen L Sedatole 2009 ldquoTransient Institutional
Ownership and CEO Contractingrdquo Accounting Review 84 737ndash770
Edmans Alex Xavier Gabaix and Augustin Landier 2009 ldquoA Multiplicative Model of
Optimal CEO Incentives in Market Equilibriumrdquo Review of Financial Studies 22
4881ndash4917
Edmans Alex 2009 ldquoBlockholder Trading Market Efficiency and Managerial Myopiardquo
Journal of Finance 64 2481ndash2513
Ertimur Yonca Fabrizio Ferri and Volkan Muslu 2011 ldquoShareholder Activism and
CEO Payrdquo Review of Financial Studies 24 535ndash592
24
Gabaix Xavier and Augustin Landier 2008 ldquoWhy Has CEO Pay Increased So Muchrdquo
Quarterly Journal of Economics 123 49ndash100
Garen John 1994 ldquoExecutive compensation and principal-agent theoryrdquo Journal of
Political Economy 102 1175ndash1199
Gibbons Robert and Kevin J Murphy 1992 ldquoOptimal Incentive Contracts in the
Presence of Career Concerns Theory and Evidencerdquo Journal of Political Economy
100 468ndash505
Greene William H 2000 (4th
Edition) Econometric Analysis Prentice-Hall NJ
Guthrie Katherine Jan Sokolowsky and Kam-Ming Wan 2012 ldquoCEO Compensation
and Board Structure Revisitedrdquo Journal of Finance Forthcoming
Hadlock Charles R and Gerald B Lumer 1997 ldquoCompensation turnover and top
management incentives historical evidencerdquo Journal of Business 70 153ndash187
Hartzell Jay C and Laura T Starks 2003 ldquoInstitutional Investors and Executive
Compensationrdquo Journal of Finance 58 2351ndash2374
Huson Mark R Robert Parrino and Laura T Starks 2001 ldquoInternal Monitoring
Mechanisms and CEO Turnover A Long-Term Perspectiverdquo Journal of Finance 56
2265ndash2297
Jensen Michael C and Kevin J Murphy 1990 ldquoPerformance Pay and Top Management
Incentivesrdquo Journal of Political Economy 98 225ndash264
Jin Li 2002 ldquoCEO compensation diversification and incentivesrdquo Journal of Financial
Economics 66 29ndash63
Kang Qiang and Qiao Liu 2008 ldquoStock Market Information Production and Executive
Incentivesrdquo Journal of Corporate Finance 14 484ndash498
Merton Robert 1973 ldquoTheory of Rational Option Pricingrdquo Bell Journal of Economics
and Management Science 4 141-183
Morse Adair Vikram Nanda and Amit Seru 2011 ldquoAre Incentive Contracts Rigged by
Powerful CEOsrdquo Journal of Finance 66 1779ndash1821
Murphy Kevin J 1985 ldquoCorporate performance and managerial remuneration An
empirical analysisrdquo Journal of Accounting and Economics 7 11ndash42
Murphy Kevin J 1999 ldquoExecutive compensationrdquo in Orley Ashenfelter and David
Card Handbook of Labor Economics vol 3 Elsevier
Rosen Sherwin 1992 ldquoContracts and the market for executivesrdquo In Contract
Economics ed Lars Werin and Hans Wijkander 181ndash211 Cambridge MA
Blackwell
Rothberg Burton and Steven Lilien 2006 ldquoMutual Funds and Proxy Voting New
Evidence on Corporate Governancerdquo Journal of Business and Technology Law 1
157ndash184
Schaefer Scott 1998 ldquoThe dependence of payndashperformance sensitivity on the size of the
firmrdquo Review of Economics and Statistics 80 436ndash443
Yermack David 1995 ldquoDo Corporations Award CEO Stock Options Effectivelyrdquo
Journal of Financial Economics 39 237ndash69
25
Table I Univariate Descriptive Statistics
This table reports sample statistics for the key variables used in this study based on our comprehensive
ExecuComp dataset Institutional Concentration is the shares owned by the institutions with the five largest
holdings divided by the shares held by all institutions Top 5 Ownership is the proportion of shares
outstanding held by the institutions with the five largest holdings Total Institutional Ownership is the
proportion of shares held by all institutions Panel B shows the measures executive compensation The pay-
for-performance sensitivity (PPS) of option grants shows the change in executive wealth due to a $1000
change in firm in shareholder wealth Expressions for Salary and Total Compensation are in thousands of
dollars while Market Capitalization is in millions of dollars Panel C shows firm characteristics
Variable Mean Median Std Dev P10 P90
Panel A Institutional Investor Holdings
Institutional Concentration 0450 0424 0151 0282 0656
Top 5 Institutional Ownership 0229 0220 0101 0110 0355
Total Institutional Ownership 0546 0561 0206 0258 0801
Panel B Executive Compensation
PPS of Option Grants 1161 0156 3829 0 2822
Salary 36311 29462 25231 15179 67000
Total Compensation 258313 85878 980701 23875 511022
Panel C Firm Characteristics
Tobins Q 1744 1139 2738 0495 3322
Market Capitalization 560699 112094 1988108 19754 1048754
26
Table II Pairwise Correlation Coefficients
This table presents the correlations between the key variables based on our comprehensive
ExecuComp dataset Institutional Concentration is the shares owned by the institutions with the five
largest holdings divided by the shares held by all institutions Top 5 Ownership is the proportion of
shares outstanding held by the institutions with the five largest holdings Total Institutional
Ownership is the proportion of shares held by all institutions The pay-for-performance sensitivity
(PPS) of option grants is the change in executive wealth due to a $1000 change in shareholder
wealth
(1) (2) (3) (4) (5) (6) (7) (8)
(1) Institutional Concentration 1
(2) Top 5 Ownership 0268 1
(3) Total Ownership -0415 0706 1
(4) PPS of Option Grants 0122 0042 -0024 1
(5) Log Salary -0270 -0012 0171 0038 1
(6) Log Market Capitalization -0630 -0157 0261 -0156 0420 1
(7) Level Market Capitalization -0241 -0157 0007 -0058 0219 0534 1
(8) Tobins Q -0008 -0034 0002 0015 -0083 0066 0121 1
27
Table III
Descriptive Statistics by Market Capitalization Deciles
Market
Capitalization
Deciles
Market
Capitalization
($m)
Total
Compensation
($000)
Salary
($)
PPS Option
Grants
($)
Institutional
Concentration
()
Top 5 Ownership
()
Institutional
Total Ownership
()
1 (Small) 1181 76008 234700 2720 611 200 327
2 2668 90206 257955 1796 495 217 438
3 4193 105946 275120 1517 454 220 485
4 6045 129892 296067 1376 420 206 491
5 8742 155391 315864 1216 394 198 502
6 13086 191888 345942 1072 380 202 531
7 20790 239501 377021 0874 368 196 532
8 34920 314955 427808 0687 354 190 538
9 67747 450970 480319 0569 328 173 528
10 (Big) 366540 846066 614354 0297 275 138 502
28
Table IV Pay-for-Performance Sensitivity of Option Grants as a Function of Institutional Influence
Similar in construction to parts of HS (Table II) this table shows the coefficients from Tobit regressions of option grant pay-for-performance sensitivity per $1000 change in
shareholder wealth against the lagged change and contemporary change in shareholder wealth measures of institutional influence Tobinrsquos Q firm market capitalization and
dummy controls for CEO and year effects We estimate models (1) to (4) over the period 1992‒1997 using data supplied by HS and models (5) to (8) are estimated over the
whole sample Models (1) (2) (5) and (6) use Institutional Concentration as a measure of institutional influence which HS define as the shares owned by the institutions
with the five largest holdings divided by the shares held by all institutions These models control for total institutional ownership Models (3) (4) (7) and (8) use Top 5
Ownership as a measure of institutional influence which we define as the proportion of shares outstanding held by the institutions with the five largest holdings Coefficients
for industry and year controls are not presented One two and three asterisks denote significance at 10 5 and 1 levels
Hartzell and Starks Database 1992‒1997 Full Sample from ExecuComp 1992‒2002
(1) (2) (3) (4) (5) (6) (7) (8)
Change in Shareholder Wealth 77721 191781 89874 137368 -446 -218 -664 -203
(171) (429) (194) (317) (319) (168) (453) (158)
Lag1 Change in Shareholder Wealth 14446 99963 88719 8544 -144 0273 -0342 00852
(027) (212) (166) (184) (071) (015) (016) (005)
Lag1 Institutional Concentration 13994 -00692 26405 03304
(767) (032) (1402) (166)
Lag1 Total Institutional Ownership 08055 10434 07106 0937
(593) (781) (666) (898)
Lag1 Top 5 Institutional Ownership 10567 07457 12116 06293
(399) ( 283) (533) (278)
Lag1 Tobinrsquos Q 0068 00877 00647 00813 00498 00727 00508 00695
(337) (434) (319) (401) (504) (734) (51) (703)
Lag1 Market Capitalization -106202 -173515 -776 -117
(474) (742) (1011) (153)
Lag1 Log Market Capitalization -0297 -02319 -04111 -03908
(1576) (1439) (2997) (2841)
CEO Dummy 13745 13682 13765 13674 28197 28128 28155 28143
(1913) (1918) (1914) (1916) (3388) (3396) (3377) (34)
Number of Observations 33878 33878 33878 33878 97679 97679 97679 97679
29
Table V Pay-for-Performance Sensitivity of Cash Compensation (Salary plus Bonus) as a Function of Institutional Influence - Level Market Capitalization
Similar in parts to models (1) and (2) of HS (Table IV) this table displays the coefficients from OLS regressions of the change in cash compensation against the lagged
change and contemporary change in shareholder wealth measures of institutional influence Tobinrsquos Q firm market capitalization and dummy controls for CEO and year
effects We estimate models (1) to (4) over the period 1992‒1997 based on our ExecuComp sample and models (5) to (8) over the whole sample Models (1) (2) (5) and (6)
use Institutional Concentration as a measure of institutional influence which HS define as the shares owned by the institutions with the five largest holdings divided by the
shares held by all institutions These models control for total institutional ownership Models (4) and (8) use Top 5 Ownership as a measure of institutional influence which
we define as the proportion of shares outstanding held by the institutions with the 5 largest holdings Coefficients for industry and year controls are not presented One two
and three asterisks denote significance at 10 5 and 1 levels
Our Database based on ExecuComp 1992‒1997 Full Sample based on ExecuComp 1992‒2002
(1) (2) (3) (4) (5) (6) (7) (8)
Lag1 Change in Shareholder Wealth 00014 00017 00088 00089 -00006 -00007 00072 00086
(051) (062) (129) (103) (027) (029) (131) (16)
Change in Shareholder Wealth -00102 00102 00105 00032 00102 00101
(097) (292) (293) (055) (611) (608)
Change Shareholder Wealth Lag1 Institutional
Concentration
00435 00764 00363 00253
(463) (258) (689) (144)
Lag1 Change Shareholder Wealth Lag1
Institutional Concentration
-00197 -00283
(122) (197)
Lag1 Institutional Concentration -492182 -288861 400149 453623
(09) (056) (098) (114)
Lag1 Total Institutional Ownership 720688 538372 655091 559587 682069 726226
(276) (232) (281) (314) (308) (322)
Lag1 Change Shareholder Wealth Lag1 Top 5
Institutional Ownership
-0046 -00689
(091) (256)
Lag1 Top 5 Institutional Ownership 583044 1191634
(092) (25)
Lag1 Tobinrsquos Q -32761 -36806 -19422 -18023 -54283 -48681 -29534 -24243
(138) (172) (077) (078) (227) (245) (136) (111)
Lag1 Market Capitalization 00004 00004 00002 00006 00007 00008 00007 00007
(042) (044) (022) (07) (086) (092) (082) (09)
CEO Dummy 768035 765474 768334 773634 616315 61729 619028 618535
(354) (355) (356) (355) (417) (418) (419) (418)
Adjusted R-Sq 00057 0006 00052 0005 00097 00097 00098 00102
Number of Observations 41305 41305 41305 41305 85518 85518 85518 85518
30
Table VI Level of Salary as a Function of Institutional Influence
Similar in parts to models (1) and (2) of HS (Table V) this table displays the coefficients from OLS regressions of salary against the lagged change and contemporary change in
shareholder wealth measures of institutional influence Tobinrsquos Q firm market capitalization and dummy controls for CEO industry and year effects We estimate models (1) to
(4) over the period 1992‒1997 using the actual data supplied by HS and models (5) to (8) over the whole sample 1992‒2002 Models (1) (2) (5) and (6) use Institutional
Concentration as a measure of institutional influence which HS define as the shares owned by the institutions with the five largest holdings divided by the shares held by all
institutions These models control for total institutional ownership Models (3) (4) (7) and (8) use Top 5 Ownership as a measure of institutional influence which we define as
the proportion of shares outstanding held by the institutions with the 5 largest holdings Coefficients for industry and year controls are not presented One two and three
asterisks denote significance at 10 5 and 1 levels
Hartzell and Starks Database 1992‒1997 Full Sample based on ExecuComp 1992‒2002
(1) (2) (3) (4) (5) (6) (7) (8)
Change in Shareholder Wealth 04125E-4 0003 -2439E-4 0003 00021 00012 00024 00012
(057) (464) (031) (468) (1004) (72) (1041) (706)
Lag1 Change in Shareholder Wealth -00062 7856E-4 -0007 9003E-4 -00014 00002 -00014 00003
(594) (098) (592) (112) (522) (123) (5) (145)
Lag1 Institutional Concentration -2475087 555134 -2734983 943317
(3245) (710) (4965) (1671)
Lag1 Total Institutional Ownership 1020109 217488 936337 252464
(1904) (432) (2471) (704)
Lag1 Top 5 Institutional Ownership -148355 31354 -87926 423166
(139) (332) (118) (643)
Lag1 Tobinrsquos Q -195348 -252191 -192344 -25158 -119837 -146373 -122132 -143821
(2812) (3762) (2721) (3752) (248) (274) (2583) (2727)
Lag1 Market Capitalization 844E-06 103E-5 00034 00039
(2781) (3051) (2874) (3104)
Lag1 Log Market Capitalization 750021 723625 776657 732279
(7511) (9285) (11224) (14039)
CEO Dummy 2683978 2702238 2668244 2702417 3281133 3285103 3292175 3283145
(8625) (9277) (8246) (9271) (12739) (13616) (12336) (13605)
Adjusted R-Sq 04604 05302 04187 05295 03856 04588 03476 04574
Number of Observations 36280 36280 36280 36280 97679 97679 97679 97679
31
Table VII Level of Total Direct Compensation as a Function of Institutional Influence
Similar in parts to models (3) and (5) of HS (Table V) this table displays the coefficients from OLS regressions of total direct compensation against the lagged change and
contemporary change in shareholder wealth measures of institutional influence Tobinrsquos Q firm market capitalization and dummy controls for CEO industry and year effects
We estimate models (1) to (4) over the period 1992‒1997 and models (5) to (8) over the whole sample Models (1) (2) (5) and (6) use Institutional Concentration as a measure
of institutional influence which HS define as the shares owned by the institutions with the five largest holdings divided by the shares held by all institutions These models
control for total institutional ownership Models (3) (4) (7) and (8) use Top 5 Ownership as a measure of institutional influence which we define as the proportion of shares
outstanding held by the institutions with the five largest holdings Coefficients for industry and year controls are not presented One two and three asterisks denote significance
at 10 5 and 1 levels
Hartzell and Starks Database 1992‒1997 Full Sample from ExecuComp 1992‒2002
(1) (2) (3) (4) (5) (6) (7) (8)
Change in Shareholder Wealth 005779 007006 005282 006791 00423 00148 0046 00138
(415) (581) (369) (565) (203) (07) (22) (066)
Lag1 Change in Shareholder Wealth 003182 00714 00272 00715 00613 0126 00614 01278
(126) (298) (107) (299) (275) (488) (276) (494)
Lag1 Institutional Concentration -1608633 3248998 -3424892 3334186
(151) (283) (1172) (118)
Lag1 Total Institutional Ownership 104013 5473128 1531013 -2926764
(1645) (831) (876) (158)
Lag1 Top 5 Institutional Ownership 188585 3927986 -2159177 -2640879
(016) (343) (067) (092)
Lag1 Tobinrsquos Q -445692 -803612 -454859 -833566 3257027 2971201 3213936 3086205
(427) (791) (432) (831) (83) (77) (815) (793)
Lag1 Market Capitalization 49E-5 615E-5 01031 01088
(113) (1365) (1555) (168)
Lag1 Log Market Capitalization 4683497 4759377 1645156 1427856
(2775) (3469) (3679) (4239)
CEO Dummy 1501307 1508886 1493688 1508736 4713079 4713629 47278 4703908
(2648) (2684) (2618) (2683) (282) (2828) (2819) (2824)
Adjusted R-Sq 01861 02060 01695 02051 01071 01113 01033 011
Number of Observations 33229 33229 33229 33229 97679 97679 97679 97679
32
Table VIII Two Stage Least Squares (Reverse Causality) Tests - Executive Compensation Measures as a Function of Institutional Concentration
This table shows the coefficients from the Two-Stage Least Squares (2SLS) regressions of executive compensation measures against the lagged change
and contemporary change in shareholder wealth instrumented institutional concentration and total institutional ownership Tobinrsquos Q the natural
logarithm of firm market capitalization and dummy variable controls for CEO and year effects We instrument measures of institutional concentration
and total institutional ownership by orthogonalizing with respect to option grant pay-for-performance sensitivity salary and total compensation
Coefficients for industry and year controls are not presented First stage results are not presented One two and three asterisks denote significance at
10 5 and 1 levels
ExecuComp Sample 1992‒1997 Full Sample from ExecuComp 1992‒2002
PPSOG Salary TC PPSOG Salary TC
Change in Shareholder Wealth 144E-05 0004 02133 -409E-06 00011 00157
(308) (775) (596) (691) (638) (068)
Lag1 Change in Shareholder Wealth -240E-05 -00004 00866 118E-06 00002 01422
(131) (087) (346) (097) (116) (51)
Lag1 Institutional Concentration -00898 668681 3263143 02024 916656 2856417
(052) (801) (176) (163) (1545) (1004)
Lag1 Total Institutional Ownership 01021 -43997 2512587 02983 -104171 -1028347
(09) (084) (153) (4) (269) (487)
Lag1 Tobinrsquos Q 00754 -225606 119297 00655 -158657 2997205
(596) (1649) (053) (834) (2535) (727)
Lag1 Log Market Capitalization -04234 828713 7864046 -04002 825106 1685638
(2195) (8132) (2977) (4106) (11133) (373)
CEO Dummy 15848 3214117 2950274 19149 3205862 4529101
(2223) (9389) (2255) (3802) (1331) (2758)
Overall R-Sq -- 04981 01206 -- 04738 01178
Number of Observations 41305 41305 41305 85518 85518 85518
33
Table IX Two Stage Least Squares (Reverse Causality) Tests - Executive Compensation Measures as a Function of Top 5 Ownership
This table shows the coefficients from 2SLS regressions of executive compensation measures against the lagged change and contemporary change
in shareholder wealth instrumented Top 5 institutional ownership Tobinrsquos Q natural logarithm of firm market capitalization and dummy variable
controls for CEO and year effects We instrument Top 5 institutional ownership by orthogonalizing with respect to option grant pay-for-
performance sensitivity salary and total compensation Coefficients for industry and year controls are not presented First stage results are not
presented One two and three asterisks denote significance at 10 5 and 1 levels
ExecuComp sample 1992‒1997 Full Sample 1992‒2002
PPSOG Salary TC PPSOG Salary TC
Change in Shareholder Wealth 136E-05 00041 02128 -407E-06 00011 00146
(293) (79) (596) (692) (619) (063)
Lag1 Change in Shareholder Wealth -241E-05 -00003 00868 116E-06 00003 01439
(133) (07) (347) (096) (142) (514)
Lag1 Top 5 Institutional Ownership -03774 -187512 -3848531 00179 -104261 -126659
(163) (186) (166) (011) (149) (436)
Lag1 Tobinrsquos Q 00731 -223886 10074 00642 -155467 3115073
(573) (1631) (045) (819) (2509) (75)
Lag1 Log Market Capitalization -0416 775637 7696302 -04015 764632 1462536
(2345) (10084) (3766) (3688) (13708) (4072)
CEO Dummy 15854 3211991 2950158 19152 3203152 4519385
(2223) (9381) (2254) (3804) (133) (2752)
Overall R-Sq -- 04967 01202 -- 04722 01168
Number of Observations 41305 41305 41305 85518 85518 85518
34
Table X Specification Tests utilizing Box-Cox Transformation of Market Capitalization
This table presents results from a Box-Cox transformation of market capitalization In
each model we specify transformation of market capitalization only Panel A shows
results for the period 1992‒1997 and Panel B shows results for the entire period
1992‒2002 The first part of each panel shows the Lambda coefficient that applies to
the transformed market capitalization The second part of each panel shows log
likelihood scores with LR Chi2 Statistics in parentheses ‒ these test for the
appropriateness of a linear or log model specification
Panel A Our ExecuComp sample 1992‒1997
(1) (2) (3)
PPSOG Salary TC
Lambda -05143 01881 02876
(-1914) (2413) (1329)
Test H0 (Log Likelihood Score)
Lambda = -1 -1279748 -3178575 -4795326
(23866) (85952) (99688)
Lambda = 0 -1280152 -3138456 -4791159
(31934) (57147) (16339)
Lambda = 1 -1283157 -3163807 -4793506
(92039) (564156) (63292)
Panel B Full ExecuComp sample 1992‒2002
(1) (2) (4)
PPSOG Salary TC
Lambda -044 01853 04036
(-3015) (40) (3666)
Test H0 (Log Likelihood Score)
Lambda = 0 -2688334 -6476494 -10384490
(66599) (160356) (126236)
Lambda = 1 -2694886 -6538506 -10386795
(197644) (1400605) (172345)
35
Figure 2 Institutional Ownership Measures (LHA) vs Total
Compensation (RHA) by Market Capitalization
0
10
20
30
40
50
60
70
1 2 3 4 5 6 7 8 9 10
$00
$10
$20
$30
$40
$50
$60
$70
$80
$90
Millio
ns
InstitutionalConcentration (LHS)Top5 Ownership (LHS)
Institutional TotalOwnership (LHS)Total Compensation(RHS)
Page 11
11
the 1992‒2002 period for the same specification controlling for the logarithm of market
capitalization but they do not regard these results as a test of HS as their focus is entirely
on comparing the two groups of stocks Nor do they investigate the impact of
concentration on either cash compensation or total pay Kang and Liu (2008) examine
pay-performance sensitivity use a version of the HS institutional concentration measure
as a control variable in a different framework deriving from Holmstrom and Tirole
(1993) Dikolli Kulp and Sedatole (2009) argue that firms with transient ownership will
issue relatively more equity incentives to counter short-termism with their results robust
to the inclusion of the HS concentration variable
We hoped to find supportive evidence from proxy voting behavior in support of HSs
findings However evidence from actual mutual fund proxy voting records suggests that
most oppose reductions in CEO pay but overall the evidence is mixed8 Davis and Kim
(2005) examine proxy votes of mutual funds for 2004 and find that for a sample of
Fortune 1000 proxy contests in 45 cases that attempted to limit executive pay mutual
funds were unanimously opposed to such attempts Rothberg and Lilien (2006) also
examine proxy-voting data and find that mutual funds voted 66 percent of the time in
managementrsquos favor on issues of executive compensation Ertimur Ferri and Muslu
(2011) find that institutional investors lack sophistication in that they do not target
excessively paid CEOs relative to CEOs with high predicted-pay based on economic
determinants but exert a moderating influence ndash a $23 million reduction ndash for those they
deem ldquoexcessively paidrdquo CEOs
III Data and Methodology
8 We do acknowledge that the HS results are for all institutional investors while the transparency with
respect to proxy voting applies only to mutual funds
12
This Section discusses the data used in the study the construction of variables
descriptive statistics and the methodology Two datasets are used ndash one is the original
dataset generously supplied by Jay Hartzell and Laura Starks which covers 1992‒1997
second a more comprehensive dataset covering 1992‒2002 is computed from the same
data sources used in the HS dataset This is Standard and Poorrsquos (SampP) ExecuComp for
compensation data Thomson Financial CDASpectrum database for institutional holdings
to compute measure of institutional influence SampPrsquos Compustat and the Center for
Research in Stock Prices (CRSP) for firm level accounting and stock price data Our more
comprehensive dataset contains 2559 firms compared with 1914 for the HS dataset Our
full matched sample yields 97679 executive years
A Institutional Influence Measures
To measure institutional influence we employ two ownership-based metrics The first
measure of institutional influence is HSrsquos Institutional Concentration We use this for
direct comparison to the HS analysis The second measure of institutional influence
changes the denominator of the HS measure from institutional shares held to total shares
outstanding in the firm with nomenclature Top 5 Ownership9 We compute both
measures at the fiscal year end of the firm We adjust all compensation measures and
asset prices for inflation to 30 June 2002 prices for our comprehensive sample compared
to the HS dataset that is in the dollars of the day
B Compensation Measures
The ExecuComp database reports several raw measures of executive compensation for
the top five executives (where available) including total salary bonus long-term
9 Top 5 Ownership is included in the HS database but not utilized by them in reported tests as well as
constructed by us for the more comprehensive sample
13
incentives and number of option grants For our more comprehensive dataset we replicate
all the compensation measures used by HS In common with HS we compute the pay-for-
performance sensitivity of option grants using the Yermack (1995) approach This
measure shows the dollar change in value of the executive options for every dollar change
in the value of the firm This involves calculating the option delta using the Black-
Scholes formula adjusted for dividends10
The ratio of the number of shares represented
by the award of options divided by the diluted number of shares outstanding at the
beginning of the year multiplies the option delta with units given by a $1000 change in
firm value Summing salary bonus long-term incentive plan payouts and stock and
option grants provides total direct compensation Options are valued using the Black and
Scholes (1973) formula for European call options adjusted by Merton (1973) to
incorporate dividends
C Descriptive Statistics
Table I presents descriptive statistics of key variables employed in the study according
to our comprehensive dataset11
Institutions own on average 546 percent of the firms in
which they invest This is almost identical to the 531 percent found by HS The
institutions with the five largest holdings account 45 percent of institutional ownership
Once again the figure for HS is almost identical at 44 percent The institutions with the
five largest holdings own on average 229 percent of shares outstanding in a firm (HS
have 223 percent) Average total compensation for executives is $2583 million
considerably larger than that documented by HS ($1250 million) but the HS figure is an
10 When there is more than one grant during the year a weighted average delta is used
11 With the data supplied to us by Hartzell and Starks we are able to replicate exactly their summary data
(HS Table I) for all but cash compensation Salary is included but bonus payments are missing The
replicated table is available on request
14
amalgam of price levels over their sample period 1992‒1997 and ours are in consistent
dollars of June 30 2002 However there has also been significant growth in
compensation levels from 1998 to 2002
ltltINSERT TABLE Igtgt
Based on our comprehensive data in Table II we examine correlations between the
key variables employed in the study Of concern is the large negative correlation of 63
percent which has been discussed previously (Section IIA) between the HS measure of
institutional concentration and the natural logarithm of firm market capitalization
ltltINSERT TABLE IIgtgt
In view of this high correlation Table III provides a breakdown of the three key pay
variables and three institutional ownership variables by size deciles We find that the high
institutional concentration but low institutional ownership share of total shareholdings
mentioned above is evident in the smallest decile of firms by size This decile has low
salary and even lower total compensation in a relative sense but by far the highest pay-
for-performance sensitivity of option grants Small firms award executives a far higher
proportion of market capitalization Since small firm market-capitalization is negligible in
comparison with the highest decile (300 times larger in magnitude) total compensation is
very small in absolute value Moreover due to the domination of the pay-size effect base
salaries are also exceedingly small
ltltINSERT TABLE IIIgtgt
Figure 1 based on Table III shows how as institutional concentration falls with
increases in stock size (but also as the total share of institutional ownership increases)
and the pay-for-performance sensitivity of option grants falls Thus because of what
appears to be a weak size control and size-influenced concentration measure HS
15
potentially pick up a size effect as their main result but not necessarily a monitoring
effect Figure 2 shows that as firm size increases up to decile 8 at least the share of
institutional total ownership increases as does total compensation
ltltINSERT FIGURES 1 AND 2 gtgt
D Methodology
Following HS we use two estimation techniques to examine the effect of institutional
investor influence on executive compensation Due to the censored nature of option grant
pay-for-performance sensitivity this compensation measure most appropriately uses
Tobit model specification whereas an Ordinary Least Squares (OLS) model specification
is used for all other compensation measures We model compensation measures as a
function of contemporaneous and lagged change in shareholder wealth institutional
influence Tobinrsquos Q and market capitalization Controls for CEO industry and year
effects are also used Augmentation occurs to test the effect of institutional influence on
the change in executive cash and total compensation with an interaction term that is the
product of institutional influence and the change in shareholder wealth
The variables used to model executive compensation have the following definitions
The lagged and contemporary measure of change in shareholder wealth is simply the
difference in market capitalization of the firm The measures of institutional influence
used are Institutional Concentration and Top 5 Ownership Use of the two types of
institutional influence occurs in separate models In order to replicate the HS
methodology models using Institutional Concentration have an additional control
consisting of lagged total institutional holdings and Tobinrsquos Q controls for performance
and growth opportunities Market capitalization of the firm is the number of shares
outstanding multiplied by the share price with inclusion both in levels and natural
16
logarithmic form A dummy variable is included to identify whether the executive is the
CEO Dummy variables are also included to control for industry effects This variable
takes a value of 1 for the two-digit Standard Industrial Classification (SIC) in which the
firm operates A year fixed effects (dummy) variable is included taking a value of 1 if the
observation was from the given year
IV Executive Incentives as a Function of Institutional Investor Influence
This section presents the results for the effect of institutional influence on option grant
pay-for-performance sensitivity and changes in compensation
A Option Grant Pay-for-Performance Sensitivity
Table IV presents the results for the effect of institutional influence on option grant
pay-for-performance sensitivity
ltltINSERT TABLE IVgtgt
Model (1) is a direct replication of HS using the HS database as supplied to us by HS
with all the variables computed in the same way and in the same functional form as with
HS The coefficient and t-statistic on lagged institutional concentration interacted with the
change in shareholder wealth is 13994 and 767 that compares to 1358 and 842 in HS
We are unable to account for these minor discrepancies for values that should be
identical In model (2) the only change to the model is that we use the natural logarithm
of market capitalization instead of the level of market capitalization This single change
has a sizeable effect on the coefficient on institutional concentration The coefficient on
institutional concentration is now -00692 and is insignificant The fact that an order
preserving transformation to a control variable can change the results is of interest The
log-transformation to market capitalization also changes the coefficient on Top 5
Ownership in models (3) and (4) The coefficient on Top 5 Ownership is significant in
17
both model (3) and in model (4) but the better size control reduces the coefficient by over
25 percent The better size control does not affect Top 5 Ownership by as much as its
design ensures less sensitivity to what could be inadequate size controls
We also test the effect of institutional influence on option grant pay-for-performance
sensitivity over the period 1992‒2002 This sample is then approximately three times as
large as the sample used by HS Institutional concentration is highly positively significant
in model (5) similar to the results for the period 1992‒1997 When we use the log of
market capitalization in model (6) the coefficient on institutional concentration remains
positive and significant at the 10 percent level with a t-statistic of 166 Of importance
though is that the coefficient on institutional concentration is one-eighth of the coefficient
in model (5) Even for a measure of institutional influence which we believe is less
sensitive to firm size the use of log market capitalization reduces the coefficient on Top 5
Ownership by approximately 50 percent from model (7) to model (8)
B Pay-for-Performance Sensitivity due to Changes in Cash and Total
Compensation
For robustness HS examine the effect that institutional influence conditioned on firm
performance has on a changes in cash and total executive compensation using the general
method of Jensen and Murphy (1990) to compute pay-for-performance sensitivity
ltltINSERT TABLE V gtgt
Table V presents the effect of institutional influence on the change in cash
compensation using the HS size control measure throughout12
Model (1) replicates the
specification in HS for the period of 1992‒1997 and we produce a similar result ndash the
interaction of lagged institutional concentration and contemporaneous change in
12 This test uses our data for the two periods 1992‒1997 and 1992‒2002 We do not use the HS data for
cash compensation as data for bonus is missing in the data supplied to us
18
shareholder wealth positively and significantly relates to a change in executive cash
compensation In model (2) we include the contemporaneous change in shareholder
wealth and the lagged institutional concentration as stand-alone controls as indicated by
Greenersquos (2000) specification of equation (1) above and the interaction term remains
significant In model (3) we interact lagged institutional concentration with the lagged
change in shareholder wealth as in equation (2) above that tries to take account of what
participants could have observed at the time The interaction term is then negative but
insignificant altering the HS result Model (4) provides a similar result where our
preferred proxy Top 5 Ownership captures concentrated institutional influence In
models (5) to (8) we use our full sample In model (5) we find that the lagged
institutional concentration interaction with the contemporaneous change in shareholder
wealth is positive and significant in HSrsquos specification However including change in
contemporaneous shareholder wealth and lagged institutional concentration as separate
control variables as suggested by equation (1)rsquos specification removes the significance
from the interaction term in model (6) Moreover in model (7) when lagged institutional
concentration is interaction with the lagged change in shareholder wealth in accord with
equation (2) this interaction term is now negative and significant reversing the HS
finding We obtain a similar result in model (8) when we measure institutional
concentration influence using Top 5 Ownership13
V The Levels of Executive Fixed and Total Compensation as Functions of
Institutional Investor Influence
13 In unreported results we use the log of market capitalization as a size control These changes to the size
control do not dramatically change the findings from Table V The reason why this size issue is not as
important as in the findings for option grants pay-for-performance is that the change in cash compensation
is not so strongly influenced by firm size In unreported results available on request we also examine the
effect of institutional influence on changes in total compensation with similar results to the change in cash
compensation
19
Two measures of compensation are examined ndash salary and total compensation
ltltINSERT TABLE VI gtgt
Table VI examines the effect of institutional influence on the level of salary In model
(1) we use the HS sample and specification to obtain almost identical results The
coefficient on institutional concentration is (negative) $247509 that is almost identical to
(negative) $247967 in HS Both are highly significant In model (2) we make a single
change to the model ndash we use the natural logarithm of market capitalization This single
change reverses the sign of institutional concentration with the coefficient still being
significant at the 1 percent level In models (3) and (4) we measure institutional influence
using Top 5 Ownership In model (3) where we measure firm size using the level of
market capitalization Top 5 Ownership is insignificant When we use the natural
logarithm of market capitalization in model (4) Top 5 Ownership is significant and
positively relates to the salary level In models (5) to (8) we estimate the same models
using our full sample We find similar results for the full sample as for the period 1992ndash
1997
ltltINSERT TABLE VII gtgt
Table VII examines the effect of institutional influence on total compensation In
model (1) we replicate the HS methodology using the HS data to obtain a similar
significant negative relation between institutional concentration and total compensation of
(negative) $1608633 However after changing to a logarithmic specification of market
capitalization in model (2) institutional concentration increases total compensation by a
significant $32400014
Importantly when institutional influence is measured using Top 5
14 = (5688751 times 0101)
20
Ownership an insignificant relation exists with total compensation in model (3) using the
HS size control but it becomes both positive and significant in (4) with the logged size
control In models (5) and (6) where models are estimated using the full sample we
observe a similar reversal in sign as occurs in models (1) and (2) However now the
magnitude of the reversal is almost equal with a difference of about $67m that is very
large indeed In model (5) the coefficient on institutional concentration is (negative)
$342m and then in model (6) the coefficient is $33m Top 5 Ownership is insignificant
in both models (7) and (8) unlike the results for Salary
VI Robustness
Two issues are examined in is this Section ndash reverse causality and the appropriate
transformation of market capitalization A methodological issue inherent in examining the
effect of ownership-based measures of institutional influence on executive compensation
is endogeneity driven by reverse causality We utilize an instrumental variables approach
where orthogonalized institutional influence measures are regressed on each of the
compensation measures lagged option grant pay-for-performance sensitivity salary and
total compensation In addition market capitalization and a CEO dummy are included in
each regression We use the residual from each regression as an instrument for
institutional influence Such a process removes any preferences for compensation from
the institutional influence measures
ltltINSERT TABLES VIII AND IXgtgt
Table VIII presents the results where we instrument the HS measure lagged
institutional concentration with a compensation-orthogonalized equivalent The analysis
indicates that the instrumented institutional concentration measure is positively and
significantly related to the level of salary and total compensation and insignificantly
21
related to option grant pay-for-performance sensitivity (PPSOG) Table IX presents
results where we instrument Top 5 Ownership with a compensation-orthogonalized
equivalent The instrumented Top 5 Ownership measure insignificantly relates to option
grant pay-for-performance sensitivity and barely negatively relates to salary although it
negatively relates to total compensation suggesting reverse causality is driving a
significant portion of our previous results
ltltINSERT TABLE Xgtgt
Another issue examined in this section is the appropriate transformation of market
capitalization Aneuryn-Evans and Deaton (1980) provide an evaluation of methods
comparing linear and log-linear specifications We use the Box-Cox test to compare
logarithm versus linear specifications In applying the Box-Cox test we specify
transformation of only market capitalization ndash the dependent variable and all other
independent variables remain unchanged Table X presents the lambda coefficient for
each model and the log likelihood score In all models in Panel A and B a lambda of zero
receives a higher (closer to zero) log likelihood score suggesting the log specification is
appropriate
VII Conclusion
Hartzell and Starks (2003) appears to document an effective role for institutional
investors in increasing executive incentives and lowering compensation levels Our paper
re-addresses the issue of concentrated institutional monitoring from a perspective
informed by an additional decade of research since their paper appeared We cannot find
evidence that concentrated institutional investors really have any positive (or negative for
that matter) effect on executive incentive levels While superficially we find evidence
that they may raise executive pay levels this is most likely driven by an institutional
22
preference for highly remunerated executives in large companies and thus by reverse
causality Moreover our findings are not consistent with the rent-extraction hypothesis
suggesting that firms systematically overpay executives in the absence of effective
monitoring by outside parties such as institutional investors
23
References
Almazan Andres Jay C Hartzell and Laura T Starks 2005 ldquoActive Institutional
Shareholders and Costs of Monitoring Evidence from Executive Compensationrdquo
Financial Management Winter 5ndash34
Aneuryn-Evans Gwyn and Angus Deaton 1980 ldquoTesting Linear versus Logarithmic
Regression Modelsrdquo Review of Economic Studies 47 (1) Econometrics Issue 275ndash
291
Babenko Ilona 2009 ldquoShare Repurchases and Pay-Performance Sensitivity of Employee
Compensation Contractsrdquo Journal of Finance 64117-150
Baker George P and Brian J Hall 2004 ldquoCEO Incentives and Firm Sizerdquo Journal of
Labor Market Economics 22 767ndash798
Bebchuk Lucian A Jesse M Fried and David Walker 2002 ldquoManagerial power and
rent extraction in the design of executive compensationrdquo University of Chicago Law
Review 69 751ndash846
Bebchuk Lucian A and Jesse M Fried 2003 ldquoExecutive Compensation as an Agency
Problemrdquo Journal of Economic Perspectives 17 71ndash92
Bebchuk Lucian A and Jesse M Fried 2004 Pay without Performance The Unfulfilled
Promise of Executive Compensation Cambridge MA Harvard University Press
Bertrand Marianne and Sendhil Mullainathan 2001 ldquoAre CEOs Rewarded for Luck
The Ones Without Principals Arerdquo Quarterly Journal of Economics 116 901ndash32
Black Fischer and Myron Scholes 1973 ldquoThe Pricing of Options and Corporate
Liabilitiesrdquo Journal of Political Economy 81 637ndash54
Cadman Brian Sandy Klasa and Steve Matsunaga 2010 ldquoDeterminants of CEO Pay A
Comparison of ExecuComp and Non-ExecuComp Firmsrdquo The Accounting Review 85
1511ndash1543
Chhaochharia Vidhi and Yaniv Grinstein 2009 ldquoCEO Compensation and Board
Structurerdquo Journal of Finance 64 231ndash261
Cichello Michael S 2005 ldquoThe impact of firm size on pay-performance sensitivitiesrdquo
Journal of Corporate Finance 11 609ndash627
Davis Gerald F and E Han Kim 2007 ldquoBusiness ties and proxy voting by mutual
fundsrdquo Journal of Financial Economics 85 552ndash570
Dikolli Shane S Susan L Kulp and Karen L Sedatole 2009 ldquoTransient Institutional
Ownership and CEO Contractingrdquo Accounting Review 84 737ndash770
Edmans Alex Xavier Gabaix and Augustin Landier 2009 ldquoA Multiplicative Model of
Optimal CEO Incentives in Market Equilibriumrdquo Review of Financial Studies 22
4881ndash4917
Edmans Alex 2009 ldquoBlockholder Trading Market Efficiency and Managerial Myopiardquo
Journal of Finance 64 2481ndash2513
Ertimur Yonca Fabrizio Ferri and Volkan Muslu 2011 ldquoShareholder Activism and
CEO Payrdquo Review of Financial Studies 24 535ndash592
24
Gabaix Xavier and Augustin Landier 2008 ldquoWhy Has CEO Pay Increased So Muchrdquo
Quarterly Journal of Economics 123 49ndash100
Garen John 1994 ldquoExecutive compensation and principal-agent theoryrdquo Journal of
Political Economy 102 1175ndash1199
Gibbons Robert and Kevin J Murphy 1992 ldquoOptimal Incentive Contracts in the
Presence of Career Concerns Theory and Evidencerdquo Journal of Political Economy
100 468ndash505
Greene William H 2000 (4th
Edition) Econometric Analysis Prentice-Hall NJ
Guthrie Katherine Jan Sokolowsky and Kam-Ming Wan 2012 ldquoCEO Compensation
and Board Structure Revisitedrdquo Journal of Finance Forthcoming
Hadlock Charles R and Gerald B Lumer 1997 ldquoCompensation turnover and top
management incentives historical evidencerdquo Journal of Business 70 153ndash187
Hartzell Jay C and Laura T Starks 2003 ldquoInstitutional Investors and Executive
Compensationrdquo Journal of Finance 58 2351ndash2374
Huson Mark R Robert Parrino and Laura T Starks 2001 ldquoInternal Monitoring
Mechanisms and CEO Turnover A Long-Term Perspectiverdquo Journal of Finance 56
2265ndash2297
Jensen Michael C and Kevin J Murphy 1990 ldquoPerformance Pay and Top Management
Incentivesrdquo Journal of Political Economy 98 225ndash264
Jin Li 2002 ldquoCEO compensation diversification and incentivesrdquo Journal of Financial
Economics 66 29ndash63
Kang Qiang and Qiao Liu 2008 ldquoStock Market Information Production and Executive
Incentivesrdquo Journal of Corporate Finance 14 484ndash498
Merton Robert 1973 ldquoTheory of Rational Option Pricingrdquo Bell Journal of Economics
and Management Science 4 141-183
Morse Adair Vikram Nanda and Amit Seru 2011 ldquoAre Incentive Contracts Rigged by
Powerful CEOsrdquo Journal of Finance 66 1779ndash1821
Murphy Kevin J 1985 ldquoCorporate performance and managerial remuneration An
empirical analysisrdquo Journal of Accounting and Economics 7 11ndash42
Murphy Kevin J 1999 ldquoExecutive compensationrdquo in Orley Ashenfelter and David
Card Handbook of Labor Economics vol 3 Elsevier
Rosen Sherwin 1992 ldquoContracts and the market for executivesrdquo In Contract
Economics ed Lars Werin and Hans Wijkander 181ndash211 Cambridge MA
Blackwell
Rothberg Burton and Steven Lilien 2006 ldquoMutual Funds and Proxy Voting New
Evidence on Corporate Governancerdquo Journal of Business and Technology Law 1
157ndash184
Schaefer Scott 1998 ldquoThe dependence of payndashperformance sensitivity on the size of the
firmrdquo Review of Economics and Statistics 80 436ndash443
Yermack David 1995 ldquoDo Corporations Award CEO Stock Options Effectivelyrdquo
Journal of Financial Economics 39 237ndash69
25
Table I Univariate Descriptive Statistics
This table reports sample statistics for the key variables used in this study based on our comprehensive
ExecuComp dataset Institutional Concentration is the shares owned by the institutions with the five largest
holdings divided by the shares held by all institutions Top 5 Ownership is the proportion of shares
outstanding held by the institutions with the five largest holdings Total Institutional Ownership is the
proportion of shares held by all institutions Panel B shows the measures executive compensation The pay-
for-performance sensitivity (PPS) of option grants shows the change in executive wealth due to a $1000
change in firm in shareholder wealth Expressions for Salary and Total Compensation are in thousands of
dollars while Market Capitalization is in millions of dollars Panel C shows firm characteristics
Variable Mean Median Std Dev P10 P90
Panel A Institutional Investor Holdings
Institutional Concentration 0450 0424 0151 0282 0656
Top 5 Institutional Ownership 0229 0220 0101 0110 0355
Total Institutional Ownership 0546 0561 0206 0258 0801
Panel B Executive Compensation
PPS of Option Grants 1161 0156 3829 0 2822
Salary 36311 29462 25231 15179 67000
Total Compensation 258313 85878 980701 23875 511022
Panel C Firm Characteristics
Tobins Q 1744 1139 2738 0495 3322
Market Capitalization 560699 112094 1988108 19754 1048754
26
Table II Pairwise Correlation Coefficients
This table presents the correlations between the key variables based on our comprehensive
ExecuComp dataset Institutional Concentration is the shares owned by the institutions with the five
largest holdings divided by the shares held by all institutions Top 5 Ownership is the proportion of
shares outstanding held by the institutions with the five largest holdings Total Institutional
Ownership is the proportion of shares held by all institutions The pay-for-performance sensitivity
(PPS) of option grants is the change in executive wealth due to a $1000 change in shareholder
wealth
(1) (2) (3) (4) (5) (6) (7) (8)
(1) Institutional Concentration 1
(2) Top 5 Ownership 0268 1
(3) Total Ownership -0415 0706 1
(4) PPS of Option Grants 0122 0042 -0024 1
(5) Log Salary -0270 -0012 0171 0038 1
(6) Log Market Capitalization -0630 -0157 0261 -0156 0420 1
(7) Level Market Capitalization -0241 -0157 0007 -0058 0219 0534 1
(8) Tobins Q -0008 -0034 0002 0015 -0083 0066 0121 1
27
Table III
Descriptive Statistics by Market Capitalization Deciles
Market
Capitalization
Deciles
Market
Capitalization
($m)
Total
Compensation
($000)
Salary
($)
PPS Option
Grants
($)
Institutional
Concentration
()
Top 5 Ownership
()
Institutional
Total Ownership
()
1 (Small) 1181 76008 234700 2720 611 200 327
2 2668 90206 257955 1796 495 217 438
3 4193 105946 275120 1517 454 220 485
4 6045 129892 296067 1376 420 206 491
5 8742 155391 315864 1216 394 198 502
6 13086 191888 345942 1072 380 202 531
7 20790 239501 377021 0874 368 196 532
8 34920 314955 427808 0687 354 190 538
9 67747 450970 480319 0569 328 173 528
10 (Big) 366540 846066 614354 0297 275 138 502
28
Table IV Pay-for-Performance Sensitivity of Option Grants as a Function of Institutional Influence
Similar in construction to parts of HS (Table II) this table shows the coefficients from Tobit regressions of option grant pay-for-performance sensitivity per $1000 change in
shareholder wealth against the lagged change and contemporary change in shareholder wealth measures of institutional influence Tobinrsquos Q firm market capitalization and
dummy controls for CEO and year effects We estimate models (1) to (4) over the period 1992‒1997 using data supplied by HS and models (5) to (8) are estimated over the
whole sample Models (1) (2) (5) and (6) use Institutional Concentration as a measure of institutional influence which HS define as the shares owned by the institutions
with the five largest holdings divided by the shares held by all institutions These models control for total institutional ownership Models (3) (4) (7) and (8) use Top 5
Ownership as a measure of institutional influence which we define as the proportion of shares outstanding held by the institutions with the five largest holdings Coefficients
for industry and year controls are not presented One two and three asterisks denote significance at 10 5 and 1 levels
Hartzell and Starks Database 1992‒1997 Full Sample from ExecuComp 1992‒2002
(1) (2) (3) (4) (5) (6) (7) (8)
Change in Shareholder Wealth 77721 191781 89874 137368 -446 -218 -664 -203
(171) (429) (194) (317) (319) (168) (453) (158)
Lag1 Change in Shareholder Wealth 14446 99963 88719 8544 -144 0273 -0342 00852
(027) (212) (166) (184) (071) (015) (016) (005)
Lag1 Institutional Concentration 13994 -00692 26405 03304
(767) (032) (1402) (166)
Lag1 Total Institutional Ownership 08055 10434 07106 0937
(593) (781) (666) (898)
Lag1 Top 5 Institutional Ownership 10567 07457 12116 06293
(399) ( 283) (533) (278)
Lag1 Tobinrsquos Q 0068 00877 00647 00813 00498 00727 00508 00695
(337) (434) (319) (401) (504) (734) (51) (703)
Lag1 Market Capitalization -106202 -173515 -776 -117
(474) (742) (1011) (153)
Lag1 Log Market Capitalization -0297 -02319 -04111 -03908
(1576) (1439) (2997) (2841)
CEO Dummy 13745 13682 13765 13674 28197 28128 28155 28143
(1913) (1918) (1914) (1916) (3388) (3396) (3377) (34)
Number of Observations 33878 33878 33878 33878 97679 97679 97679 97679
29
Table V Pay-for-Performance Sensitivity of Cash Compensation (Salary plus Bonus) as a Function of Institutional Influence - Level Market Capitalization
Similar in parts to models (1) and (2) of HS (Table IV) this table displays the coefficients from OLS regressions of the change in cash compensation against the lagged
change and contemporary change in shareholder wealth measures of institutional influence Tobinrsquos Q firm market capitalization and dummy controls for CEO and year
effects We estimate models (1) to (4) over the period 1992‒1997 based on our ExecuComp sample and models (5) to (8) over the whole sample Models (1) (2) (5) and (6)
use Institutional Concentration as a measure of institutional influence which HS define as the shares owned by the institutions with the five largest holdings divided by the
shares held by all institutions These models control for total institutional ownership Models (4) and (8) use Top 5 Ownership as a measure of institutional influence which
we define as the proportion of shares outstanding held by the institutions with the 5 largest holdings Coefficients for industry and year controls are not presented One two
and three asterisks denote significance at 10 5 and 1 levels
Our Database based on ExecuComp 1992‒1997 Full Sample based on ExecuComp 1992‒2002
(1) (2) (3) (4) (5) (6) (7) (8)
Lag1 Change in Shareholder Wealth 00014 00017 00088 00089 -00006 -00007 00072 00086
(051) (062) (129) (103) (027) (029) (131) (16)
Change in Shareholder Wealth -00102 00102 00105 00032 00102 00101
(097) (292) (293) (055) (611) (608)
Change Shareholder Wealth Lag1 Institutional
Concentration
00435 00764 00363 00253
(463) (258) (689) (144)
Lag1 Change Shareholder Wealth Lag1
Institutional Concentration
-00197 -00283
(122) (197)
Lag1 Institutional Concentration -492182 -288861 400149 453623
(09) (056) (098) (114)
Lag1 Total Institutional Ownership 720688 538372 655091 559587 682069 726226
(276) (232) (281) (314) (308) (322)
Lag1 Change Shareholder Wealth Lag1 Top 5
Institutional Ownership
-0046 -00689
(091) (256)
Lag1 Top 5 Institutional Ownership 583044 1191634
(092) (25)
Lag1 Tobinrsquos Q -32761 -36806 -19422 -18023 -54283 -48681 -29534 -24243
(138) (172) (077) (078) (227) (245) (136) (111)
Lag1 Market Capitalization 00004 00004 00002 00006 00007 00008 00007 00007
(042) (044) (022) (07) (086) (092) (082) (09)
CEO Dummy 768035 765474 768334 773634 616315 61729 619028 618535
(354) (355) (356) (355) (417) (418) (419) (418)
Adjusted R-Sq 00057 0006 00052 0005 00097 00097 00098 00102
Number of Observations 41305 41305 41305 41305 85518 85518 85518 85518
30
Table VI Level of Salary as a Function of Institutional Influence
Similar in parts to models (1) and (2) of HS (Table V) this table displays the coefficients from OLS regressions of salary against the lagged change and contemporary change in
shareholder wealth measures of institutional influence Tobinrsquos Q firm market capitalization and dummy controls for CEO industry and year effects We estimate models (1) to
(4) over the period 1992‒1997 using the actual data supplied by HS and models (5) to (8) over the whole sample 1992‒2002 Models (1) (2) (5) and (6) use Institutional
Concentration as a measure of institutional influence which HS define as the shares owned by the institutions with the five largest holdings divided by the shares held by all
institutions These models control for total institutional ownership Models (3) (4) (7) and (8) use Top 5 Ownership as a measure of institutional influence which we define as
the proportion of shares outstanding held by the institutions with the 5 largest holdings Coefficients for industry and year controls are not presented One two and three
asterisks denote significance at 10 5 and 1 levels
Hartzell and Starks Database 1992‒1997 Full Sample based on ExecuComp 1992‒2002
(1) (2) (3) (4) (5) (6) (7) (8)
Change in Shareholder Wealth 04125E-4 0003 -2439E-4 0003 00021 00012 00024 00012
(057) (464) (031) (468) (1004) (72) (1041) (706)
Lag1 Change in Shareholder Wealth -00062 7856E-4 -0007 9003E-4 -00014 00002 -00014 00003
(594) (098) (592) (112) (522) (123) (5) (145)
Lag1 Institutional Concentration -2475087 555134 -2734983 943317
(3245) (710) (4965) (1671)
Lag1 Total Institutional Ownership 1020109 217488 936337 252464
(1904) (432) (2471) (704)
Lag1 Top 5 Institutional Ownership -148355 31354 -87926 423166
(139) (332) (118) (643)
Lag1 Tobinrsquos Q -195348 -252191 -192344 -25158 -119837 -146373 -122132 -143821
(2812) (3762) (2721) (3752) (248) (274) (2583) (2727)
Lag1 Market Capitalization 844E-06 103E-5 00034 00039
(2781) (3051) (2874) (3104)
Lag1 Log Market Capitalization 750021 723625 776657 732279
(7511) (9285) (11224) (14039)
CEO Dummy 2683978 2702238 2668244 2702417 3281133 3285103 3292175 3283145
(8625) (9277) (8246) (9271) (12739) (13616) (12336) (13605)
Adjusted R-Sq 04604 05302 04187 05295 03856 04588 03476 04574
Number of Observations 36280 36280 36280 36280 97679 97679 97679 97679
31
Table VII Level of Total Direct Compensation as a Function of Institutional Influence
Similar in parts to models (3) and (5) of HS (Table V) this table displays the coefficients from OLS regressions of total direct compensation against the lagged change and
contemporary change in shareholder wealth measures of institutional influence Tobinrsquos Q firm market capitalization and dummy controls for CEO industry and year effects
We estimate models (1) to (4) over the period 1992‒1997 and models (5) to (8) over the whole sample Models (1) (2) (5) and (6) use Institutional Concentration as a measure
of institutional influence which HS define as the shares owned by the institutions with the five largest holdings divided by the shares held by all institutions These models
control for total institutional ownership Models (3) (4) (7) and (8) use Top 5 Ownership as a measure of institutional influence which we define as the proportion of shares
outstanding held by the institutions with the five largest holdings Coefficients for industry and year controls are not presented One two and three asterisks denote significance
at 10 5 and 1 levels
Hartzell and Starks Database 1992‒1997 Full Sample from ExecuComp 1992‒2002
(1) (2) (3) (4) (5) (6) (7) (8)
Change in Shareholder Wealth 005779 007006 005282 006791 00423 00148 0046 00138
(415) (581) (369) (565) (203) (07) (22) (066)
Lag1 Change in Shareholder Wealth 003182 00714 00272 00715 00613 0126 00614 01278
(126) (298) (107) (299) (275) (488) (276) (494)
Lag1 Institutional Concentration -1608633 3248998 -3424892 3334186
(151) (283) (1172) (118)
Lag1 Total Institutional Ownership 104013 5473128 1531013 -2926764
(1645) (831) (876) (158)
Lag1 Top 5 Institutional Ownership 188585 3927986 -2159177 -2640879
(016) (343) (067) (092)
Lag1 Tobinrsquos Q -445692 -803612 -454859 -833566 3257027 2971201 3213936 3086205
(427) (791) (432) (831) (83) (77) (815) (793)
Lag1 Market Capitalization 49E-5 615E-5 01031 01088
(113) (1365) (1555) (168)
Lag1 Log Market Capitalization 4683497 4759377 1645156 1427856
(2775) (3469) (3679) (4239)
CEO Dummy 1501307 1508886 1493688 1508736 4713079 4713629 47278 4703908
(2648) (2684) (2618) (2683) (282) (2828) (2819) (2824)
Adjusted R-Sq 01861 02060 01695 02051 01071 01113 01033 011
Number of Observations 33229 33229 33229 33229 97679 97679 97679 97679
32
Table VIII Two Stage Least Squares (Reverse Causality) Tests - Executive Compensation Measures as a Function of Institutional Concentration
This table shows the coefficients from the Two-Stage Least Squares (2SLS) regressions of executive compensation measures against the lagged change
and contemporary change in shareholder wealth instrumented institutional concentration and total institutional ownership Tobinrsquos Q the natural
logarithm of firm market capitalization and dummy variable controls for CEO and year effects We instrument measures of institutional concentration
and total institutional ownership by orthogonalizing with respect to option grant pay-for-performance sensitivity salary and total compensation
Coefficients for industry and year controls are not presented First stage results are not presented One two and three asterisks denote significance at
10 5 and 1 levels
ExecuComp Sample 1992‒1997 Full Sample from ExecuComp 1992‒2002
PPSOG Salary TC PPSOG Salary TC
Change in Shareholder Wealth 144E-05 0004 02133 -409E-06 00011 00157
(308) (775) (596) (691) (638) (068)
Lag1 Change in Shareholder Wealth -240E-05 -00004 00866 118E-06 00002 01422
(131) (087) (346) (097) (116) (51)
Lag1 Institutional Concentration -00898 668681 3263143 02024 916656 2856417
(052) (801) (176) (163) (1545) (1004)
Lag1 Total Institutional Ownership 01021 -43997 2512587 02983 -104171 -1028347
(09) (084) (153) (4) (269) (487)
Lag1 Tobinrsquos Q 00754 -225606 119297 00655 -158657 2997205
(596) (1649) (053) (834) (2535) (727)
Lag1 Log Market Capitalization -04234 828713 7864046 -04002 825106 1685638
(2195) (8132) (2977) (4106) (11133) (373)
CEO Dummy 15848 3214117 2950274 19149 3205862 4529101
(2223) (9389) (2255) (3802) (1331) (2758)
Overall R-Sq -- 04981 01206 -- 04738 01178
Number of Observations 41305 41305 41305 85518 85518 85518
33
Table IX Two Stage Least Squares (Reverse Causality) Tests - Executive Compensation Measures as a Function of Top 5 Ownership
This table shows the coefficients from 2SLS regressions of executive compensation measures against the lagged change and contemporary change
in shareholder wealth instrumented Top 5 institutional ownership Tobinrsquos Q natural logarithm of firm market capitalization and dummy variable
controls for CEO and year effects We instrument Top 5 institutional ownership by orthogonalizing with respect to option grant pay-for-
performance sensitivity salary and total compensation Coefficients for industry and year controls are not presented First stage results are not
presented One two and three asterisks denote significance at 10 5 and 1 levels
ExecuComp sample 1992‒1997 Full Sample 1992‒2002
PPSOG Salary TC PPSOG Salary TC
Change in Shareholder Wealth 136E-05 00041 02128 -407E-06 00011 00146
(293) (79) (596) (692) (619) (063)
Lag1 Change in Shareholder Wealth -241E-05 -00003 00868 116E-06 00003 01439
(133) (07) (347) (096) (142) (514)
Lag1 Top 5 Institutional Ownership -03774 -187512 -3848531 00179 -104261 -126659
(163) (186) (166) (011) (149) (436)
Lag1 Tobinrsquos Q 00731 -223886 10074 00642 -155467 3115073
(573) (1631) (045) (819) (2509) (75)
Lag1 Log Market Capitalization -0416 775637 7696302 -04015 764632 1462536
(2345) (10084) (3766) (3688) (13708) (4072)
CEO Dummy 15854 3211991 2950158 19152 3203152 4519385
(2223) (9381) (2254) (3804) (133) (2752)
Overall R-Sq -- 04967 01202 -- 04722 01168
Number of Observations 41305 41305 41305 85518 85518 85518
34
Table X Specification Tests utilizing Box-Cox Transformation of Market Capitalization
This table presents results from a Box-Cox transformation of market capitalization In
each model we specify transformation of market capitalization only Panel A shows
results for the period 1992‒1997 and Panel B shows results for the entire period
1992‒2002 The first part of each panel shows the Lambda coefficient that applies to
the transformed market capitalization The second part of each panel shows log
likelihood scores with LR Chi2 Statistics in parentheses ‒ these test for the
appropriateness of a linear or log model specification
Panel A Our ExecuComp sample 1992‒1997
(1) (2) (3)
PPSOG Salary TC
Lambda -05143 01881 02876
(-1914) (2413) (1329)
Test H0 (Log Likelihood Score)
Lambda = -1 -1279748 -3178575 -4795326
(23866) (85952) (99688)
Lambda = 0 -1280152 -3138456 -4791159
(31934) (57147) (16339)
Lambda = 1 -1283157 -3163807 -4793506
(92039) (564156) (63292)
Panel B Full ExecuComp sample 1992‒2002
(1) (2) (4)
PPSOG Salary TC
Lambda -044 01853 04036
(-3015) (40) (3666)
Test H0 (Log Likelihood Score)
Lambda = 0 -2688334 -6476494 -10384490
(66599) (160356) (126236)
Lambda = 1 -2694886 -6538506 -10386795
(197644) (1400605) (172345)
35
Figure 2 Institutional Ownership Measures (LHA) vs Total
Compensation (RHA) by Market Capitalization
0
10
20
30
40
50
60
70
1 2 3 4 5 6 7 8 9 10
$00
$10
$20
$30
$40
$50
$60
$70
$80
$90
Millio
ns
InstitutionalConcentration (LHS)Top5 Ownership (LHS)
Institutional TotalOwnership (LHS)Total Compensation(RHS)
Page 12
12
This Section discusses the data used in the study the construction of variables
descriptive statistics and the methodology Two datasets are used ndash one is the original
dataset generously supplied by Jay Hartzell and Laura Starks which covers 1992‒1997
second a more comprehensive dataset covering 1992‒2002 is computed from the same
data sources used in the HS dataset This is Standard and Poorrsquos (SampP) ExecuComp for
compensation data Thomson Financial CDASpectrum database for institutional holdings
to compute measure of institutional influence SampPrsquos Compustat and the Center for
Research in Stock Prices (CRSP) for firm level accounting and stock price data Our more
comprehensive dataset contains 2559 firms compared with 1914 for the HS dataset Our
full matched sample yields 97679 executive years
A Institutional Influence Measures
To measure institutional influence we employ two ownership-based metrics The first
measure of institutional influence is HSrsquos Institutional Concentration We use this for
direct comparison to the HS analysis The second measure of institutional influence
changes the denominator of the HS measure from institutional shares held to total shares
outstanding in the firm with nomenclature Top 5 Ownership9 We compute both
measures at the fiscal year end of the firm We adjust all compensation measures and
asset prices for inflation to 30 June 2002 prices for our comprehensive sample compared
to the HS dataset that is in the dollars of the day
B Compensation Measures
The ExecuComp database reports several raw measures of executive compensation for
the top five executives (where available) including total salary bonus long-term
9 Top 5 Ownership is included in the HS database but not utilized by them in reported tests as well as
constructed by us for the more comprehensive sample
13
incentives and number of option grants For our more comprehensive dataset we replicate
all the compensation measures used by HS In common with HS we compute the pay-for-
performance sensitivity of option grants using the Yermack (1995) approach This
measure shows the dollar change in value of the executive options for every dollar change
in the value of the firm This involves calculating the option delta using the Black-
Scholes formula adjusted for dividends10
The ratio of the number of shares represented
by the award of options divided by the diluted number of shares outstanding at the
beginning of the year multiplies the option delta with units given by a $1000 change in
firm value Summing salary bonus long-term incentive plan payouts and stock and
option grants provides total direct compensation Options are valued using the Black and
Scholes (1973) formula for European call options adjusted by Merton (1973) to
incorporate dividends
C Descriptive Statistics
Table I presents descriptive statistics of key variables employed in the study according
to our comprehensive dataset11
Institutions own on average 546 percent of the firms in
which they invest This is almost identical to the 531 percent found by HS The
institutions with the five largest holdings account 45 percent of institutional ownership
Once again the figure for HS is almost identical at 44 percent The institutions with the
five largest holdings own on average 229 percent of shares outstanding in a firm (HS
have 223 percent) Average total compensation for executives is $2583 million
considerably larger than that documented by HS ($1250 million) but the HS figure is an
10 When there is more than one grant during the year a weighted average delta is used
11 With the data supplied to us by Hartzell and Starks we are able to replicate exactly their summary data
(HS Table I) for all but cash compensation Salary is included but bonus payments are missing The
replicated table is available on request
14
amalgam of price levels over their sample period 1992‒1997 and ours are in consistent
dollars of June 30 2002 However there has also been significant growth in
compensation levels from 1998 to 2002
ltltINSERT TABLE Igtgt
Based on our comprehensive data in Table II we examine correlations between the
key variables employed in the study Of concern is the large negative correlation of 63
percent which has been discussed previously (Section IIA) between the HS measure of
institutional concentration and the natural logarithm of firm market capitalization
ltltINSERT TABLE IIgtgt
In view of this high correlation Table III provides a breakdown of the three key pay
variables and three institutional ownership variables by size deciles We find that the high
institutional concentration but low institutional ownership share of total shareholdings
mentioned above is evident in the smallest decile of firms by size This decile has low
salary and even lower total compensation in a relative sense but by far the highest pay-
for-performance sensitivity of option grants Small firms award executives a far higher
proportion of market capitalization Since small firm market-capitalization is negligible in
comparison with the highest decile (300 times larger in magnitude) total compensation is
very small in absolute value Moreover due to the domination of the pay-size effect base
salaries are also exceedingly small
ltltINSERT TABLE IIIgtgt
Figure 1 based on Table III shows how as institutional concentration falls with
increases in stock size (but also as the total share of institutional ownership increases)
and the pay-for-performance sensitivity of option grants falls Thus because of what
appears to be a weak size control and size-influenced concentration measure HS
15
potentially pick up a size effect as their main result but not necessarily a monitoring
effect Figure 2 shows that as firm size increases up to decile 8 at least the share of
institutional total ownership increases as does total compensation
ltltINSERT FIGURES 1 AND 2 gtgt
D Methodology
Following HS we use two estimation techniques to examine the effect of institutional
investor influence on executive compensation Due to the censored nature of option grant
pay-for-performance sensitivity this compensation measure most appropriately uses
Tobit model specification whereas an Ordinary Least Squares (OLS) model specification
is used for all other compensation measures We model compensation measures as a
function of contemporaneous and lagged change in shareholder wealth institutional
influence Tobinrsquos Q and market capitalization Controls for CEO industry and year
effects are also used Augmentation occurs to test the effect of institutional influence on
the change in executive cash and total compensation with an interaction term that is the
product of institutional influence and the change in shareholder wealth
The variables used to model executive compensation have the following definitions
The lagged and contemporary measure of change in shareholder wealth is simply the
difference in market capitalization of the firm The measures of institutional influence
used are Institutional Concentration and Top 5 Ownership Use of the two types of
institutional influence occurs in separate models In order to replicate the HS
methodology models using Institutional Concentration have an additional control
consisting of lagged total institutional holdings and Tobinrsquos Q controls for performance
and growth opportunities Market capitalization of the firm is the number of shares
outstanding multiplied by the share price with inclusion both in levels and natural
16
logarithmic form A dummy variable is included to identify whether the executive is the
CEO Dummy variables are also included to control for industry effects This variable
takes a value of 1 for the two-digit Standard Industrial Classification (SIC) in which the
firm operates A year fixed effects (dummy) variable is included taking a value of 1 if the
observation was from the given year
IV Executive Incentives as a Function of Institutional Investor Influence
This section presents the results for the effect of institutional influence on option grant
pay-for-performance sensitivity and changes in compensation
A Option Grant Pay-for-Performance Sensitivity
Table IV presents the results for the effect of institutional influence on option grant
pay-for-performance sensitivity
ltltINSERT TABLE IVgtgt
Model (1) is a direct replication of HS using the HS database as supplied to us by HS
with all the variables computed in the same way and in the same functional form as with
HS The coefficient and t-statistic on lagged institutional concentration interacted with the
change in shareholder wealth is 13994 and 767 that compares to 1358 and 842 in HS
We are unable to account for these minor discrepancies for values that should be
identical In model (2) the only change to the model is that we use the natural logarithm
of market capitalization instead of the level of market capitalization This single change
has a sizeable effect on the coefficient on institutional concentration The coefficient on
institutional concentration is now -00692 and is insignificant The fact that an order
preserving transformation to a control variable can change the results is of interest The
log-transformation to market capitalization also changes the coefficient on Top 5
Ownership in models (3) and (4) The coefficient on Top 5 Ownership is significant in
17
both model (3) and in model (4) but the better size control reduces the coefficient by over
25 percent The better size control does not affect Top 5 Ownership by as much as its
design ensures less sensitivity to what could be inadequate size controls
We also test the effect of institutional influence on option grant pay-for-performance
sensitivity over the period 1992‒2002 This sample is then approximately three times as
large as the sample used by HS Institutional concentration is highly positively significant
in model (5) similar to the results for the period 1992‒1997 When we use the log of
market capitalization in model (6) the coefficient on institutional concentration remains
positive and significant at the 10 percent level with a t-statistic of 166 Of importance
though is that the coefficient on institutional concentration is one-eighth of the coefficient
in model (5) Even for a measure of institutional influence which we believe is less
sensitive to firm size the use of log market capitalization reduces the coefficient on Top 5
Ownership by approximately 50 percent from model (7) to model (8)
B Pay-for-Performance Sensitivity due to Changes in Cash and Total
Compensation
For robustness HS examine the effect that institutional influence conditioned on firm
performance has on a changes in cash and total executive compensation using the general
method of Jensen and Murphy (1990) to compute pay-for-performance sensitivity
ltltINSERT TABLE V gtgt
Table V presents the effect of institutional influence on the change in cash
compensation using the HS size control measure throughout12
Model (1) replicates the
specification in HS for the period of 1992‒1997 and we produce a similar result ndash the
interaction of lagged institutional concentration and contemporaneous change in
12 This test uses our data for the two periods 1992‒1997 and 1992‒2002 We do not use the HS data for
cash compensation as data for bonus is missing in the data supplied to us
18
shareholder wealth positively and significantly relates to a change in executive cash
compensation In model (2) we include the contemporaneous change in shareholder
wealth and the lagged institutional concentration as stand-alone controls as indicated by
Greenersquos (2000) specification of equation (1) above and the interaction term remains
significant In model (3) we interact lagged institutional concentration with the lagged
change in shareholder wealth as in equation (2) above that tries to take account of what
participants could have observed at the time The interaction term is then negative but
insignificant altering the HS result Model (4) provides a similar result where our
preferred proxy Top 5 Ownership captures concentrated institutional influence In
models (5) to (8) we use our full sample In model (5) we find that the lagged
institutional concentration interaction with the contemporaneous change in shareholder
wealth is positive and significant in HSrsquos specification However including change in
contemporaneous shareholder wealth and lagged institutional concentration as separate
control variables as suggested by equation (1)rsquos specification removes the significance
from the interaction term in model (6) Moreover in model (7) when lagged institutional
concentration is interaction with the lagged change in shareholder wealth in accord with
equation (2) this interaction term is now negative and significant reversing the HS
finding We obtain a similar result in model (8) when we measure institutional
concentration influence using Top 5 Ownership13
V The Levels of Executive Fixed and Total Compensation as Functions of
Institutional Investor Influence
13 In unreported results we use the log of market capitalization as a size control These changes to the size
control do not dramatically change the findings from Table V The reason why this size issue is not as
important as in the findings for option grants pay-for-performance is that the change in cash compensation
is not so strongly influenced by firm size In unreported results available on request we also examine the
effect of institutional influence on changes in total compensation with similar results to the change in cash
compensation
19
Two measures of compensation are examined ndash salary and total compensation
ltltINSERT TABLE VI gtgt
Table VI examines the effect of institutional influence on the level of salary In model
(1) we use the HS sample and specification to obtain almost identical results The
coefficient on institutional concentration is (negative) $247509 that is almost identical to
(negative) $247967 in HS Both are highly significant In model (2) we make a single
change to the model ndash we use the natural logarithm of market capitalization This single
change reverses the sign of institutional concentration with the coefficient still being
significant at the 1 percent level In models (3) and (4) we measure institutional influence
using Top 5 Ownership In model (3) where we measure firm size using the level of
market capitalization Top 5 Ownership is insignificant When we use the natural
logarithm of market capitalization in model (4) Top 5 Ownership is significant and
positively relates to the salary level In models (5) to (8) we estimate the same models
using our full sample We find similar results for the full sample as for the period 1992ndash
1997
ltltINSERT TABLE VII gtgt
Table VII examines the effect of institutional influence on total compensation In
model (1) we replicate the HS methodology using the HS data to obtain a similar
significant negative relation between institutional concentration and total compensation of
(negative) $1608633 However after changing to a logarithmic specification of market
capitalization in model (2) institutional concentration increases total compensation by a
significant $32400014
Importantly when institutional influence is measured using Top 5
14 = (5688751 times 0101)
20
Ownership an insignificant relation exists with total compensation in model (3) using the
HS size control but it becomes both positive and significant in (4) with the logged size
control In models (5) and (6) where models are estimated using the full sample we
observe a similar reversal in sign as occurs in models (1) and (2) However now the
magnitude of the reversal is almost equal with a difference of about $67m that is very
large indeed In model (5) the coefficient on institutional concentration is (negative)
$342m and then in model (6) the coefficient is $33m Top 5 Ownership is insignificant
in both models (7) and (8) unlike the results for Salary
VI Robustness
Two issues are examined in is this Section ndash reverse causality and the appropriate
transformation of market capitalization A methodological issue inherent in examining the
effect of ownership-based measures of institutional influence on executive compensation
is endogeneity driven by reverse causality We utilize an instrumental variables approach
where orthogonalized institutional influence measures are regressed on each of the
compensation measures lagged option grant pay-for-performance sensitivity salary and
total compensation In addition market capitalization and a CEO dummy are included in
each regression We use the residual from each regression as an instrument for
institutional influence Such a process removes any preferences for compensation from
the institutional influence measures
ltltINSERT TABLES VIII AND IXgtgt
Table VIII presents the results where we instrument the HS measure lagged
institutional concentration with a compensation-orthogonalized equivalent The analysis
indicates that the instrumented institutional concentration measure is positively and
significantly related to the level of salary and total compensation and insignificantly
21
related to option grant pay-for-performance sensitivity (PPSOG) Table IX presents
results where we instrument Top 5 Ownership with a compensation-orthogonalized
equivalent The instrumented Top 5 Ownership measure insignificantly relates to option
grant pay-for-performance sensitivity and barely negatively relates to salary although it
negatively relates to total compensation suggesting reverse causality is driving a
significant portion of our previous results
ltltINSERT TABLE Xgtgt
Another issue examined in this section is the appropriate transformation of market
capitalization Aneuryn-Evans and Deaton (1980) provide an evaluation of methods
comparing linear and log-linear specifications We use the Box-Cox test to compare
logarithm versus linear specifications In applying the Box-Cox test we specify
transformation of only market capitalization ndash the dependent variable and all other
independent variables remain unchanged Table X presents the lambda coefficient for
each model and the log likelihood score In all models in Panel A and B a lambda of zero
receives a higher (closer to zero) log likelihood score suggesting the log specification is
appropriate
VII Conclusion
Hartzell and Starks (2003) appears to document an effective role for institutional
investors in increasing executive incentives and lowering compensation levels Our paper
re-addresses the issue of concentrated institutional monitoring from a perspective
informed by an additional decade of research since their paper appeared We cannot find
evidence that concentrated institutional investors really have any positive (or negative for
that matter) effect on executive incentive levels While superficially we find evidence
that they may raise executive pay levels this is most likely driven by an institutional
22
preference for highly remunerated executives in large companies and thus by reverse
causality Moreover our findings are not consistent with the rent-extraction hypothesis
suggesting that firms systematically overpay executives in the absence of effective
monitoring by outside parties such as institutional investors
23
References
Almazan Andres Jay C Hartzell and Laura T Starks 2005 ldquoActive Institutional
Shareholders and Costs of Monitoring Evidence from Executive Compensationrdquo
Financial Management Winter 5ndash34
Aneuryn-Evans Gwyn and Angus Deaton 1980 ldquoTesting Linear versus Logarithmic
Regression Modelsrdquo Review of Economic Studies 47 (1) Econometrics Issue 275ndash
291
Babenko Ilona 2009 ldquoShare Repurchases and Pay-Performance Sensitivity of Employee
Compensation Contractsrdquo Journal of Finance 64117-150
Baker George P and Brian J Hall 2004 ldquoCEO Incentives and Firm Sizerdquo Journal of
Labor Market Economics 22 767ndash798
Bebchuk Lucian A Jesse M Fried and David Walker 2002 ldquoManagerial power and
rent extraction in the design of executive compensationrdquo University of Chicago Law
Review 69 751ndash846
Bebchuk Lucian A and Jesse M Fried 2003 ldquoExecutive Compensation as an Agency
Problemrdquo Journal of Economic Perspectives 17 71ndash92
Bebchuk Lucian A and Jesse M Fried 2004 Pay without Performance The Unfulfilled
Promise of Executive Compensation Cambridge MA Harvard University Press
Bertrand Marianne and Sendhil Mullainathan 2001 ldquoAre CEOs Rewarded for Luck
The Ones Without Principals Arerdquo Quarterly Journal of Economics 116 901ndash32
Black Fischer and Myron Scholes 1973 ldquoThe Pricing of Options and Corporate
Liabilitiesrdquo Journal of Political Economy 81 637ndash54
Cadman Brian Sandy Klasa and Steve Matsunaga 2010 ldquoDeterminants of CEO Pay A
Comparison of ExecuComp and Non-ExecuComp Firmsrdquo The Accounting Review 85
1511ndash1543
Chhaochharia Vidhi and Yaniv Grinstein 2009 ldquoCEO Compensation and Board
Structurerdquo Journal of Finance 64 231ndash261
Cichello Michael S 2005 ldquoThe impact of firm size on pay-performance sensitivitiesrdquo
Journal of Corporate Finance 11 609ndash627
Davis Gerald F and E Han Kim 2007 ldquoBusiness ties and proxy voting by mutual
fundsrdquo Journal of Financial Economics 85 552ndash570
Dikolli Shane S Susan L Kulp and Karen L Sedatole 2009 ldquoTransient Institutional
Ownership and CEO Contractingrdquo Accounting Review 84 737ndash770
Edmans Alex Xavier Gabaix and Augustin Landier 2009 ldquoA Multiplicative Model of
Optimal CEO Incentives in Market Equilibriumrdquo Review of Financial Studies 22
4881ndash4917
Edmans Alex 2009 ldquoBlockholder Trading Market Efficiency and Managerial Myopiardquo
Journal of Finance 64 2481ndash2513
Ertimur Yonca Fabrizio Ferri and Volkan Muslu 2011 ldquoShareholder Activism and
CEO Payrdquo Review of Financial Studies 24 535ndash592
24
Gabaix Xavier and Augustin Landier 2008 ldquoWhy Has CEO Pay Increased So Muchrdquo
Quarterly Journal of Economics 123 49ndash100
Garen John 1994 ldquoExecutive compensation and principal-agent theoryrdquo Journal of
Political Economy 102 1175ndash1199
Gibbons Robert and Kevin J Murphy 1992 ldquoOptimal Incentive Contracts in the
Presence of Career Concerns Theory and Evidencerdquo Journal of Political Economy
100 468ndash505
Greene William H 2000 (4th
Edition) Econometric Analysis Prentice-Hall NJ
Guthrie Katherine Jan Sokolowsky and Kam-Ming Wan 2012 ldquoCEO Compensation
and Board Structure Revisitedrdquo Journal of Finance Forthcoming
Hadlock Charles R and Gerald B Lumer 1997 ldquoCompensation turnover and top
management incentives historical evidencerdquo Journal of Business 70 153ndash187
Hartzell Jay C and Laura T Starks 2003 ldquoInstitutional Investors and Executive
Compensationrdquo Journal of Finance 58 2351ndash2374
Huson Mark R Robert Parrino and Laura T Starks 2001 ldquoInternal Monitoring
Mechanisms and CEO Turnover A Long-Term Perspectiverdquo Journal of Finance 56
2265ndash2297
Jensen Michael C and Kevin J Murphy 1990 ldquoPerformance Pay and Top Management
Incentivesrdquo Journal of Political Economy 98 225ndash264
Jin Li 2002 ldquoCEO compensation diversification and incentivesrdquo Journal of Financial
Economics 66 29ndash63
Kang Qiang and Qiao Liu 2008 ldquoStock Market Information Production and Executive
Incentivesrdquo Journal of Corporate Finance 14 484ndash498
Merton Robert 1973 ldquoTheory of Rational Option Pricingrdquo Bell Journal of Economics
and Management Science 4 141-183
Morse Adair Vikram Nanda and Amit Seru 2011 ldquoAre Incentive Contracts Rigged by
Powerful CEOsrdquo Journal of Finance 66 1779ndash1821
Murphy Kevin J 1985 ldquoCorporate performance and managerial remuneration An
empirical analysisrdquo Journal of Accounting and Economics 7 11ndash42
Murphy Kevin J 1999 ldquoExecutive compensationrdquo in Orley Ashenfelter and David
Card Handbook of Labor Economics vol 3 Elsevier
Rosen Sherwin 1992 ldquoContracts and the market for executivesrdquo In Contract
Economics ed Lars Werin and Hans Wijkander 181ndash211 Cambridge MA
Blackwell
Rothberg Burton and Steven Lilien 2006 ldquoMutual Funds and Proxy Voting New
Evidence on Corporate Governancerdquo Journal of Business and Technology Law 1
157ndash184
Schaefer Scott 1998 ldquoThe dependence of payndashperformance sensitivity on the size of the
firmrdquo Review of Economics and Statistics 80 436ndash443
Yermack David 1995 ldquoDo Corporations Award CEO Stock Options Effectivelyrdquo
Journal of Financial Economics 39 237ndash69
25
Table I Univariate Descriptive Statistics
This table reports sample statistics for the key variables used in this study based on our comprehensive
ExecuComp dataset Institutional Concentration is the shares owned by the institutions with the five largest
holdings divided by the shares held by all institutions Top 5 Ownership is the proportion of shares
outstanding held by the institutions with the five largest holdings Total Institutional Ownership is the
proportion of shares held by all institutions Panel B shows the measures executive compensation The pay-
for-performance sensitivity (PPS) of option grants shows the change in executive wealth due to a $1000
change in firm in shareholder wealth Expressions for Salary and Total Compensation are in thousands of
dollars while Market Capitalization is in millions of dollars Panel C shows firm characteristics
Variable Mean Median Std Dev P10 P90
Panel A Institutional Investor Holdings
Institutional Concentration 0450 0424 0151 0282 0656
Top 5 Institutional Ownership 0229 0220 0101 0110 0355
Total Institutional Ownership 0546 0561 0206 0258 0801
Panel B Executive Compensation
PPS of Option Grants 1161 0156 3829 0 2822
Salary 36311 29462 25231 15179 67000
Total Compensation 258313 85878 980701 23875 511022
Panel C Firm Characteristics
Tobins Q 1744 1139 2738 0495 3322
Market Capitalization 560699 112094 1988108 19754 1048754
26
Table II Pairwise Correlation Coefficients
This table presents the correlations between the key variables based on our comprehensive
ExecuComp dataset Institutional Concentration is the shares owned by the institutions with the five
largest holdings divided by the shares held by all institutions Top 5 Ownership is the proportion of
shares outstanding held by the institutions with the five largest holdings Total Institutional
Ownership is the proportion of shares held by all institutions The pay-for-performance sensitivity
(PPS) of option grants is the change in executive wealth due to a $1000 change in shareholder
wealth
(1) (2) (3) (4) (5) (6) (7) (8)
(1) Institutional Concentration 1
(2) Top 5 Ownership 0268 1
(3) Total Ownership -0415 0706 1
(4) PPS of Option Grants 0122 0042 -0024 1
(5) Log Salary -0270 -0012 0171 0038 1
(6) Log Market Capitalization -0630 -0157 0261 -0156 0420 1
(7) Level Market Capitalization -0241 -0157 0007 -0058 0219 0534 1
(8) Tobins Q -0008 -0034 0002 0015 -0083 0066 0121 1
27
Table III
Descriptive Statistics by Market Capitalization Deciles
Market
Capitalization
Deciles
Market
Capitalization
($m)
Total
Compensation
($000)
Salary
($)
PPS Option
Grants
($)
Institutional
Concentration
()
Top 5 Ownership
()
Institutional
Total Ownership
()
1 (Small) 1181 76008 234700 2720 611 200 327
2 2668 90206 257955 1796 495 217 438
3 4193 105946 275120 1517 454 220 485
4 6045 129892 296067 1376 420 206 491
5 8742 155391 315864 1216 394 198 502
6 13086 191888 345942 1072 380 202 531
7 20790 239501 377021 0874 368 196 532
8 34920 314955 427808 0687 354 190 538
9 67747 450970 480319 0569 328 173 528
10 (Big) 366540 846066 614354 0297 275 138 502
28
Table IV Pay-for-Performance Sensitivity of Option Grants as a Function of Institutional Influence
Similar in construction to parts of HS (Table II) this table shows the coefficients from Tobit regressions of option grant pay-for-performance sensitivity per $1000 change in
shareholder wealth against the lagged change and contemporary change in shareholder wealth measures of institutional influence Tobinrsquos Q firm market capitalization and
dummy controls for CEO and year effects We estimate models (1) to (4) over the period 1992‒1997 using data supplied by HS and models (5) to (8) are estimated over the
whole sample Models (1) (2) (5) and (6) use Institutional Concentration as a measure of institutional influence which HS define as the shares owned by the institutions
with the five largest holdings divided by the shares held by all institutions These models control for total institutional ownership Models (3) (4) (7) and (8) use Top 5
Ownership as a measure of institutional influence which we define as the proportion of shares outstanding held by the institutions with the five largest holdings Coefficients
for industry and year controls are not presented One two and three asterisks denote significance at 10 5 and 1 levels
Hartzell and Starks Database 1992‒1997 Full Sample from ExecuComp 1992‒2002
(1) (2) (3) (4) (5) (6) (7) (8)
Change in Shareholder Wealth 77721 191781 89874 137368 -446 -218 -664 -203
(171) (429) (194) (317) (319) (168) (453) (158)
Lag1 Change in Shareholder Wealth 14446 99963 88719 8544 -144 0273 -0342 00852
(027) (212) (166) (184) (071) (015) (016) (005)
Lag1 Institutional Concentration 13994 -00692 26405 03304
(767) (032) (1402) (166)
Lag1 Total Institutional Ownership 08055 10434 07106 0937
(593) (781) (666) (898)
Lag1 Top 5 Institutional Ownership 10567 07457 12116 06293
(399) ( 283) (533) (278)
Lag1 Tobinrsquos Q 0068 00877 00647 00813 00498 00727 00508 00695
(337) (434) (319) (401) (504) (734) (51) (703)
Lag1 Market Capitalization -106202 -173515 -776 -117
(474) (742) (1011) (153)
Lag1 Log Market Capitalization -0297 -02319 -04111 -03908
(1576) (1439) (2997) (2841)
CEO Dummy 13745 13682 13765 13674 28197 28128 28155 28143
(1913) (1918) (1914) (1916) (3388) (3396) (3377) (34)
Number of Observations 33878 33878 33878 33878 97679 97679 97679 97679
29
Table V Pay-for-Performance Sensitivity of Cash Compensation (Salary plus Bonus) as a Function of Institutional Influence - Level Market Capitalization
Similar in parts to models (1) and (2) of HS (Table IV) this table displays the coefficients from OLS regressions of the change in cash compensation against the lagged
change and contemporary change in shareholder wealth measures of institutional influence Tobinrsquos Q firm market capitalization and dummy controls for CEO and year
effects We estimate models (1) to (4) over the period 1992‒1997 based on our ExecuComp sample and models (5) to (8) over the whole sample Models (1) (2) (5) and (6)
use Institutional Concentration as a measure of institutional influence which HS define as the shares owned by the institutions with the five largest holdings divided by the
shares held by all institutions These models control for total institutional ownership Models (4) and (8) use Top 5 Ownership as a measure of institutional influence which
we define as the proportion of shares outstanding held by the institutions with the 5 largest holdings Coefficients for industry and year controls are not presented One two
and three asterisks denote significance at 10 5 and 1 levels
Our Database based on ExecuComp 1992‒1997 Full Sample based on ExecuComp 1992‒2002
(1) (2) (3) (4) (5) (6) (7) (8)
Lag1 Change in Shareholder Wealth 00014 00017 00088 00089 -00006 -00007 00072 00086
(051) (062) (129) (103) (027) (029) (131) (16)
Change in Shareholder Wealth -00102 00102 00105 00032 00102 00101
(097) (292) (293) (055) (611) (608)
Change Shareholder Wealth Lag1 Institutional
Concentration
00435 00764 00363 00253
(463) (258) (689) (144)
Lag1 Change Shareholder Wealth Lag1
Institutional Concentration
-00197 -00283
(122) (197)
Lag1 Institutional Concentration -492182 -288861 400149 453623
(09) (056) (098) (114)
Lag1 Total Institutional Ownership 720688 538372 655091 559587 682069 726226
(276) (232) (281) (314) (308) (322)
Lag1 Change Shareholder Wealth Lag1 Top 5
Institutional Ownership
-0046 -00689
(091) (256)
Lag1 Top 5 Institutional Ownership 583044 1191634
(092) (25)
Lag1 Tobinrsquos Q -32761 -36806 -19422 -18023 -54283 -48681 -29534 -24243
(138) (172) (077) (078) (227) (245) (136) (111)
Lag1 Market Capitalization 00004 00004 00002 00006 00007 00008 00007 00007
(042) (044) (022) (07) (086) (092) (082) (09)
CEO Dummy 768035 765474 768334 773634 616315 61729 619028 618535
(354) (355) (356) (355) (417) (418) (419) (418)
Adjusted R-Sq 00057 0006 00052 0005 00097 00097 00098 00102
Number of Observations 41305 41305 41305 41305 85518 85518 85518 85518
30
Table VI Level of Salary as a Function of Institutional Influence
Similar in parts to models (1) and (2) of HS (Table V) this table displays the coefficients from OLS regressions of salary against the lagged change and contemporary change in
shareholder wealth measures of institutional influence Tobinrsquos Q firm market capitalization and dummy controls for CEO industry and year effects We estimate models (1) to
(4) over the period 1992‒1997 using the actual data supplied by HS and models (5) to (8) over the whole sample 1992‒2002 Models (1) (2) (5) and (6) use Institutional
Concentration as a measure of institutional influence which HS define as the shares owned by the institutions with the five largest holdings divided by the shares held by all
institutions These models control for total institutional ownership Models (3) (4) (7) and (8) use Top 5 Ownership as a measure of institutional influence which we define as
the proportion of shares outstanding held by the institutions with the 5 largest holdings Coefficients for industry and year controls are not presented One two and three
asterisks denote significance at 10 5 and 1 levels
Hartzell and Starks Database 1992‒1997 Full Sample based on ExecuComp 1992‒2002
(1) (2) (3) (4) (5) (6) (7) (8)
Change in Shareholder Wealth 04125E-4 0003 -2439E-4 0003 00021 00012 00024 00012
(057) (464) (031) (468) (1004) (72) (1041) (706)
Lag1 Change in Shareholder Wealth -00062 7856E-4 -0007 9003E-4 -00014 00002 -00014 00003
(594) (098) (592) (112) (522) (123) (5) (145)
Lag1 Institutional Concentration -2475087 555134 -2734983 943317
(3245) (710) (4965) (1671)
Lag1 Total Institutional Ownership 1020109 217488 936337 252464
(1904) (432) (2471) (704)
Lag1 Top 5 Institutional Ownership -148355 31354 -87926 423166
(139) (332) (118) (643)
Lag1 Tobinrsquos Q -195348 -252191 -192344 -25158 -119837 -146373 -122132 -143821
(2812) (3762) (2721) (3752) (248) (274) (2583) (2727)
Lag1 Market Capitalization 844E-06 103E-5 00034 00039
(2781) (3051) (2874) (3104)
Lag1 Log Market Capitalization 750021 723625 776657 732279
(7511) (9285) (11224) (14039)
CEO Dummy 2683978 2702238 2668244 2702417 3281133 3285103 3292175 3283145
(8625) (9277) (8246) (9271) (12739) (13616) (12336) (13605)
Adjusted R-Sq 04604 05302 04187 05295 03856 04588 03476 04574
Number of Observations 36280 36280 36280 36280 97679 97679 97679 97679
31
Table VII Level of Total Direct Compensation as a Function of Institutional Influence
Similar in parts to models (3) and (5) of HS (Table V) this table displays the coefficients from OLS regressions of total direct compensation against the lagged change and
contemporary change in shareholder wealth measures of institutional influence Tobinrsquos Q firm market capitalization and dummy controls for CEO industry and year effects
We estimate models (1) to (4) over the period 1992‒1997 and models (5) to (8) over the whole sample Models (1) (2) (5) and (6) use Institutional Concentration as a measure
of institutional influence which HS define as the shares owned by the institutions with the five largest holdings divided by the shares held by all institutions These models
control for total institutional ownership Models (3) (4) (7) and (8) use Top 5 Ownership as a measure of institutional influence which we define as the proportion of shares
outstanding held by the institutions with the five largest holdings Coefficients for industry and year controls are not presented One two and three asterisks denote significance
at 10 5 and 1 levels
Hartzell and Starks Database 1992‒1997 Full Sample from ExecuComp 1992‒2002
(1) (2) (3) (4) (5) (6) (7) (8)
Change in Shareholder Wealth 005779 007006 005282 006791 00423 00148 0046 00138
(415) (581) (369) (565) (203) (07) (22) (066)
Lag1 Change in Shareholder Wealth 003182 00714 00272 00715 00613 0126 00614 01278
(126) (298) (107) (299) (275) (488) (276) (494)
Lag1 Institutional Concentration -1608633 3248998 -3424892 3334186
(151) (283) (1172) (118)
Lag1 Total Institutional Ownership 104013 5473128 1531013 -2926764
(1645) (831) (876) (158)
Lag1 Top 5 Institutional Ownership 188585 3927986 -2159177 -2640879
(016) (343) (067) (092)
Lag1 Tobinrsquos Q -445692 -803612 -454859 -833566 3257027 2971201 3213936 3086205
(427) (791) (432) (831) (83) (77) (815) (793)
Lag1 Market Capitalization 49E-5 615E-5 01031 01088
(113) (1365) (1555) (168)
Lag1 Log Market Capitalization 4683497 4759377 1645156 1427856
(2775) (3469) (3679) (4239)
CEO Dummy 1501307 1508886 1493688 1508736 4713079 4713629 47278 4703908
(2648) (2684) (2618) (2683) (282) (2828) (2819) (2824)
Adjusted R-Sq 01861 02060 01695 02051 01071 01113 01033 011
Number of Observations 33229 33229 33229 33229 97679 97679 97679 97679
32
Table VIII Two Stage Least Squares (Reverse Causality) Tests - Executive Compensation Measures as a Function of Institutional Concentration
This table shows the coefficients from the Two-Stage Least Squares (2SLS) regressions of executive compensation measures against the lagged change
and contemporary change in shareholder wealth instrumented institutional concentration and total institutional ownership Tobinrsquos Q the natural
logarithm of firm market capitalization and dummy variable controls for CEO and year effects We instrument measures of institutional concentration
and total institutional ownership by orthogonalizing with respect to option grant pay-for-performance sensitivity salary and total compensation
Coefficients for industry and year controls are not presented First stage results are not presented One two and three asterisks denote significance at
10 5 and 1 levels
ExecuComp Sample 1992‒1997 Full Sample from ExecuComp 1992‒2002
PPSOG Salary TC PPSOG Salary TC
Change in Shareholder Wealth 144E-05 0004 02133 -409E-06 00011 00157
(308) (775) (596) (691) (638) (068)
Lag1 Change in Shareholder Wealth -240E-05 -00004 00866 118E-06 00002 01422
(131) (087) (346) (097) (116) (51)
Lag1 Institutional Concentration -00898 668681 3263143 02024 916656 2856417
(052) (801) (176) (163) (1545) (1004)
Lag1 Total Institutional Ownership 01021 -43997 2512587 02983 -104171 -1028347
(09) (084) (153) (4) (269) (487)
Lag1 Tobinrsquos Q 00754 -225606 119297 00655 -158657 2997205
(596) (1649) (053) (834) (2535) (727)
Lag1 Log Market Capitalization -04234 828713 7864046 -04002 825106 1685638
(2195) (8132) (2977) (4106) (11133) (373)
CEO Dummy 15848 3214117 2950274 19149 3205862 4529101
(2223) (9389) (2255) (3802) (1331) (2758)
Overall R-Sq -- 04981 01206 -- 04738 01178
Number of Observations 41305 41305 41305 85518 85518 85518
33
Table IX Two Stage Least Squares (Reverse Causality) Tests - Executive Compensation Measures as a Function of Top 5 Ownership
This table shows the coefficients from 2SLS regressions of executive compensation measures against the lagged change and contemporary change
in shareholder wealth instrumented Top 5 institutional ownership Tobinrsquos Q natural logarithm of firm market capitalization and dummy variable
controls for CEO and year effects We instrument Top 5 institutional ownership by orthogonalizing with respect to option grant pay-for-
performance sensitivity salary and total compensation Coefficients for industry and year controls are not presented First stage results are not
presented One two and three asterisks denote significance at 10 5 and 1 levels
ExecuComp sample 1992‒1997 Full Sample 1992‒2002
PPSOG Salary TC PPSOG Salary TC
Change in Shareholder Wealth 136E-05 00041 02128 -407E-06 00011 00146
(293) (79) (596) (692) (619) (063)
Lag1 Change in Shareholder Wealth -241E-05 -00003 00868 116E-06 00003 01439
(133) (07) (347) (096) (142) (514)
Lag1 Top 5 Institutional Ownership -03774 -187512 -3848531 00179 -104261 -126659
(163) (186) (166) (011) (149) (436)
Lag1 Tobinrsquos Q 00731 -223886 10074 00642 -155467 3115073
(573) (1631) (045) (819) (2509) (75)
Lag1 Log Market Capitalization -0416 775637 7696302 -04015 764632 1462536
(2345) (10084) (3766) (3688) (13708) (4072)
CEO Dummy 15854 3211991 2950158 19152 3203152 4519385
(2223) (9381) (2254) (3804) (133) (2752)
Overall R-Sq -- 04967 01202 -- 04722 01168
Number of Observations 41305 41305 41305 85518 85518 85518
34
Table X Specification Tests utilizing Box-Cox Transformation of Market Capitalization
This table presents results from a Box-Cox transformation of market capitalization In
each model we specify transformation of market capitalization only Panel A shows
results for the period 1992‒1997 and Panel B shows results for the entire period
1992‒2002 The first part of each panel shows the Lambda coefficient that applies to
the transformed market capitalization The second part of each panel shows log
likelihood scores with LR Chi2 Statistics in parentheses ‒ these test for the
appropriateness of a linear or log model specification
Panel A Our ExecuComp sample 1992‒1997
(1) (2) (3)
PPSOG Salary TC
Lambda -05143 01881 02876
(-1914) (2413) (1329)
Test H0 (Log Likelihood Score)
Lambda = -1 -1279748 -3178575 -4795326
(23866) (85952) (99688)
Lambda = 0 -1280152 -3138456 -4791159
(31934) (57147) (16339)
Lambda = 1 -1283157 -3163807 -4793506
(92039) (564156) (63292)
Panel B Full ExecuComp sample 1992‒2002
(1) (2) (4)
PPSOG Salary TC
Lambda -044 01853 04036
(-3015) (40) (3666)
Test H0 (Log Likelihood Score)
Lambda = 0 -2688334 -6476494 -10384490
(66599) (160356) (126236)
Lambda = 1 -2694886 -6538506 -10386795
(197644) (1400605) (172345)
35
Figure 2 Institutional Ownership Measures (LHA) vs Total
Compensation (RHA) by Market Capitalization
0
10
20
30
40
50
60
70
1 2 3 4 5 6 7 8 9 10
$00
$10
$20
$30
$40
$50
$60
$70
$80
$90
Millio
ns
InstitutionalConcentration (LHS)Top5 Ownership (LHS)
Institutional TotalOwnership (LHS)Total Compensation(RHS)
Page 13
13
incentives and number of option grants For our more comprehensive dataset we replicate
all the compensation measures used by HS In common with HS we compute the pay-for-
performance sensitivity of option grants using the Yermack (1995) approach This
measure shows the dollar change in value of the executive options for every dollar change
in the value of the firm This involves calculating the option delta using the Black-
Scholes formula adjusted for dividends10
The ratio of the number of shares represented
by the award of options divided by the diluted number of shares outstanding at the
beginning of the year multiplies the option delta with units given by a $1000 change in
firm value Summing salary bonus long-term incentive plan payouts and stock and
option grants provides total direct compensation Options are valued using the Black and
Scholes (1973) formula for European call options adjusted by Merton (1973) to
incorporate dividends
C Descriptive Statistics
Table I presents descriptive statistics of key variables employed in the study according
to our comprehensive dataset11
Institutions own on average 546 percent of the firms in
which they invest This is almost identical to the 531 percent found by HS The
institutions with the five largest holdings account 45 percent of institutional ownership
Once again the figure for HS is almost identical at 44 percent The institutions with the
five largest holdings own on average 229 percent of shares outstanding in a firm (HS
have 223 percent) Average total compensation for executives is $2583 million
considerably larger than that documented by HS ($1250 million) but the HS figure is an
10 When there is more than one grant during the year a weighted average delta is used
11 With the data supplied to us by Hartzell and Starks we are able to replicate exactly their summary data
(HS Table I) for all but cash compensation Salary is included but bonus payments are missing The
replicated table is available on request
14
amalgam of price levels over their sample period 1992‒1997 and ours are in consistent
dollars of June 30 2002 However there has also been significant growth in
compensation levels from 1998 to 2002
ltltINSERT TABLE Igtgt
Based on our comprehensive data in Table II we examine correlations between the
key variables employed in the study Of concern is the large negative correlation of 63
percent which has been discussed previously (Section IIA) between the HS measure of
institutional concentration and the natural logarithm of firm market capitalization
ltltINSERT TABLE IIgtgt
In view of this high correlation Table III provides a breakdown of the three key pay
variables and three institutional ownership variables by size deciles We find that the high
institutional concentration but low institutional ownership share of total shareholdings
mentioned above is evident in the smallest decile of firms by size This decile has low
salary and even lower total compensation in a relative sense but by far the highest pay-
for-performance sensitivity of option grants Small firms award executives a far higher
proportion of market capitalization Since small firm market-capitalization is negligible in
comparison with the highest decile (300 times larger in magnitude) total compensation is
very small in absolute value Moreover due to the domination of the pay-size effect base
salaries are also exceedingly small
ltltINSERT TABLE IIIgtgt
Figure 1 based on Table III shows how as institutional concentration falls with
increases in stock size (but also as the total share of institutional ownership increases)
and the pay-for-performance sensitivity of option grants falls Thus because of what
appears to be a weak size control and size-influenced concentration measure HS
15
potentially pick up a size effect as their main result but not necessarily a monitoring
effect Figure 2 shows that as firm size increases up to decile 8 at least the share of
institutional total ownership increases as does total compensation
ltltINSERT FIGURES 1 AND 2 gtgt
D Methodology
Following HS we use two estimation techniques to examine the effect of institutional
investor influence on executive compensation Due to the censored nature of option grant
pay-for-performance sensitivity this compensation measure most appropriately uses
Tobit model specification whereas an Ordinary Least Squares (OLS) model specification
is used for all other compensation measures We model compensation measures as a
function of contemporaneous and lagged change in shareholder wealth institutional
influence Tobinrsquos Q and market capitalization Controls for CEO industry and year
effects are also used Augmentation occurs to test the effect of institutional influence on
the change in executive cash and total compensation with an interaction term that is the
product of institutional influence and the change in shareholder wealth
The variables used to model executive compensation have the following definitions
The lagged and contemporary measure of change in shareholder wealth is simply the
difference in market capitalization of the firm The measures of institutional influence
used are Institutional Concentration and Top 5 Ownership Use of the two types of
institutional influence occurs in separate models In order to replicate the HS
methodology models using Institutional Concentration have an additional control
consisting of lagged total institutional holdings and Tobinrsquos Q controls for performance
and growth opportunities Market capitalization of the firm is the number of shares
outstanding multiplied by the share price with inclusion both in levels and natural
16
logarithmic form A dummy variable is included to identify whether the executive is the
CEO Dummy variables are also included to control for industry effects This variable
takes a value of 1 for the two-digit Standard Industrial Classification (SIC) in which the
firm operates A year fixed effects (dummy) variable is included taking a value of 1 if the
observation was from the given year
IV Executive Incentives as a Function of Institutional Investor Influence
This section presents the results for the effect of institutional influence on option grant
pay-for-performance sensitivity and changes in compensation
A Option Grant Pay-for-Performance Sensitivity
Table IV presents the results for the effect of institutional influence on option grant
pay-for-performance sensitivity
ltltINSERT TABLE IVgtgt
Model (1) is a direct replication of HS using the HS database as supplied to us by HS
with all the variables computed in the same way and in the same functional form as with
HS The coefficient and t-statistic on lagged institutional concentration interacted with the
change in shareholder wealth is 13994 and 767 that compares to 1358 and 842 in HS
We are unable to account for these minor discrepancies for values that should be
identical In model (2) the only change to the model is that we use the natural logarithm
of market capitalization instead of the level of market capitalization This single change
has a sizeable effect on the coefficient on institutional concentration The coefficient on
institutional concentration is now -00692 and is insignificant The fact that an order
preserving transformation to a control variable can change the results is of interest The
log-transformation to market capitalization also changes the coefficient on Top 5
Ownership in models (3) and (4) The coefficient on Top 5 Ownership is significant in
17
both model (3) and in model (4) but the better size control reduces the coefficient by over
25 percent The better size control does not affect Top 5 Ownership by as much as its
design ensures less sensitivity to what could be inadequate size controls
We also test the effect of institutional influence on option grant pay-for-performance
sensitivity over the period 1992‒2002 This sample is then approximately three times as
large as the sample used by HS Institutional concentration is highly positively significant
in model (5) similar to the results for the period 1992‒1997 When we use the log of
market capitalization in model (6) the coefficient on institutional concentration remains
positive and significant at the 10 percent level with a t-statistic of 166 Of importance
though is that the coefficient on institutional concentration is one-eighth of the coefficient
in model (5) Even for a measure of institutional influence which we believe is less
sensitive to firm size the use of log market capitalization reduces the coefficient on Top 5
Ownership by approximately 50 percent from model (7) to model (8)
B Pay-for-Performance Sensitivity due to Changes in Cash and Total
Compensation
For robustness HS examine the effect that institutional influence conditioned on firm
performance has on a changes in cash and total executive compensation using the general
method of Jensen and Murphy (1990) to compute pay-for-performance sensitivity
ltltINSERT TABLE V gtgt
Table V presents the effect of institutional influence on the change in cash
compensation using the HS size control measure throughout12
Model (1) replicates the
specification in HS for the period of 1992‒1997 and we produce a similar result ndash the
interaction of lagged institutional concentration and contemporaneous change in
12 This test uses our data for the two periods 1992‒1997 and 1992‒2002 We do not use the HS data for
cash compensation as data for bonus is missing in the data supplied to us
18
shareholder wealth positively and significantly relates to a change in executive cash
compensation In model (2) we include the contemporaneous change in shareholder
wealth and the lagged institutional concentration as stand-alone controls as indicated by
Greenersquos (2000) specification of equation (1) above and the interaction term remains
significant In model (3) we interact lagged institutional concentration with the lagged
change in shareholder wealth as in equation (2) above that tries to take account of what
participants could have observed at the time The interaction term is then negative but
insignificant altering the HS result Model (4) provides a similar result where our
preferred proxy Top 5 Ownership captures concentrated institutional influence In
models (5) to (8) we use our full sample In model (5) we find that the lagged
institutional concentration interaction with the contemporaneous change in shareholder
wealth is positive and significant in HSrsquos specification However including change in
contemporaneous shareholder wealth and lagged institutional concentration as separate
control variables as suggested by equation (1)rsquos specification removes the significance
from the interaction term in model (6) Moreover in model (7) when lagged institutional
concentration is interaction with the lagged change in shareholder wealth in accord with
equation (2) this interaction term is now negative and significant reversing the HS
finding We obtain a similar result in model (8) when we measure institutional
concentration influence using Top 5 Ownership13
V The Levels of Executive Fixed and Total Compensation as Functions of
Institutional Investor Influence
13 In unreported results we use the log of market capitalization as a size control These changes to the size
control do not dramatically change the findings from Table V The reason why this size issue is not as
important as in the findings for option grants pay-for-performance is that the change in cash compensation
is not so strongly influenced by firm size In unreported results available on request we also examine the
effect of institutional influence on changes in total compensation with similar results to the change in cash
compensation
19
Two measures of compensation are examined ndash salary and total compensation
ltltINSERT TABLE VI gtgt
Table VI examines the effect of institutional influence on the level of salary In model
(1) we use the HS sample and specification to obtain almost identical results The
coefficient on institutional concentration is (negative) $247509 that is almost identical to
(negative) $247967 in HS Both are highly significant In model (2) we make a single
change to the model ndash we use the natural logarithm of market capitalization This single
change reverses the sign of institutional concentration with the coefficient still being
significant at the 1 percent level In models (3) and (4) we measure institutional influence
using Top 5 Ownership In model (3) where we measure firm size using the level of
market capitalization Top 5 Ownership is insignificant When we use the natural
logarithm of market capitalization in model (4) Top 5 Ownership is significant and
positively relates to the salary level In models (5) to (8) we estimate the same models
using our full sample We find similar results for the full sample as for the period 1992ndash
1997
ltltINSERT TABLE VII gtgt
Table VII examines the effect of institutional influence on total compensation In
model (1) we replicate the HS methodology using the HS data to obtain a similar
significant negative relation between institutional concentration and total compensation of
(negative) $1608633 However after changing to a logarithmic specification of market
capitalization in model (2) institutional concentration increases total compensation by a
significant $32400014
Importantly when institutional influence is measured using Top 5
14 = (5688751 times 0101)
20
Ownership an insignificant relation exists with total compensation in model (3) using the
HS size control but it becomes both positive and significant in (4) with the logged size
control In models (5) and (6) where models are estimated using the full sample we
observe a similar reversal in sign as occurs in models (1) and (2) However now the
magnitude of the reversal is almost equal with a difference of about $67m that is very
large indeed In model (5) the coefficient on institutional concentration is (negative)
$342m and then in model (6) the coefficient is $33m Top 5 Ownership is insignificant
in both models (7) and (8) unlike the results for Salary
VI Robustness
Two issues are examined in is this Section ndash reverse causality and the appropriate
transformation of market capitalization A methodological issue inherent in examining the
effect of ownership-based measures of institutional influence on executive compensation
is endogeneity driven by reverse causality We utilize an instrumental variables approach
where orthogonalized institutional influence measures are regressed on each of the
compensation measures lagged option grant pay-for-performance sensitivity salary and
total compensation In addition market capitalization and a CEO dummy are included in
each regression We use the residual from each regression as an instrument for
institutional influence Such a process removes any preferences for compensation from
the institutional influence measures
ltltINSERT TABLES VIII AND IXgtgt
Table VIII presents the results where we instrument the HS measure lagged
institutional concentration with a compensation-orthogonalized equivalent The analysis
indicates that the instrumented institutional concentration measure is positively and
significantly related to the level of salary and total compensation and insignificantly
21
related to option grant pay-for-performance sensitivity (PPSOG) Table IX presents
results where we instrument Top 5 Ownership with a compensation-orthogonalized
equivalent The instrumented Top 5 Ownership measure insignificantly relates to option
grant pay-for-performance sensitivity and barely negatively relates to salary although it
negatively relates to total compensation suggesting reverse causality is driving a
significant portion of our previous results
ltltINSERT TABLE Xgtgt
Another issue examined in this section is the appropriate transformation of market
capitalization Aneuryn-Evans and Deaton (1980) provide an evaluation of methods
comparing linear and log-linear specifications We use the Box-Cox test to compare
logarithm versus linear specifications In applying the Box-Cox test we specify
transformation of only market capitalization ndash the dependent variable and all other
independent variables remain unchanged Table X presents the lambda coefficient for
each model and the log likelihood score In all models in Panel A and B a lambda of zero
receives a higher (closer to zero) log likelihood score suggesting the log specification is
appropriate
VII Conclusion
Hartzell and Starks (2003) appears to document an effective role for institutional
investors in increasing executive incentives and lowering compensation levels Our paper
re-addresses the issue of concentrated institutional monitoring from a perspective
informed by an additional decade of research since their paper appeared We cannot find
evidence that concentrated institutional investors really have any positive (or negative for
that matter) effect on executive incentive levels While superficially we find evidence
that they may raise executive pay levels this is most likely driven by an institutional
22
preference for highly remunerated executives in large companies and thus by reverse
causality Moreover our findings are not consistent with the rent-extraction hypothesis
suggesting that firms systematically overpay executives in the absence of effective
monitoring by outside parties such as institutional investors
23
References
Almazan Andres Jay C Hartzell and Laura T Starks 2005 ldquoActive Institutional
Shareholders and Costs of Monitoring Evidence from Executive Compensationrdquo
Financial Management Winter 5ndash34
Aneuryn-Evans Gwyn and Angus Deaton 1980 ldquoTesting Linear versus Logarithmic
Regression Modelsrdquo Review of Economic Studies 47 (1) Econometrics Issue 275ndash
291
Babenko Ilona 2009 ldquoShare Repurchases and Pay-Performance Sensitivity of Employee
Compensation Contractsrdquo Journal of Finance 64117-150
Baker George P and Brian J Hall 2004 ldquoCEO Incentives and Firm Sizerdquo Journal of
Labor Market Economics 22 767ndash798
Bebchuk Lucian A Jesse M Fried and David Walker 2002 ldquoManagerial power and
rent extraction in the design of executive compensationrdquo University of Chicago Law
Review 69 751ndash846
Bebchuk Lucian A and Jesse M Fried 2003 ldquoExecutive Compensation as an Agency
Problemrdquo Journal of Economic Perspectives 17 71ndash92
Bebchuk Lucian A and Jesse M Fried 2004 Pay without Performance The Unfulfilled
Promise of Executive Compensation Cambridge MA Harvard University Press
Bertrand Marianne and Sendhil Mullainathan 2001 ldquoAre CEOs Rewarded for Luck
The Ones Without Principals Arerdquo Quarterly Journal of Economics 116 901ndash32
Black Fischer and Myron Scholes 1973 ldquoThe Pricing of Options and Corporate
Liabilitiesrdquo Journal of Political Economy 81 637ndash54
Cadman Brian Sandy Klasa and Steve Matsunaga 2010 ldquoDeterminants of CEO Pay A
Comparison of ExecuComp and Non-ExecuComp Firmsrdquo The Accounting Review 85
1511ndash1543
Chhaochharia Vidhi and Yaniv Grinstein 2009 ldquoCEO Compensation and Board
Structurerdquo Journal of Finance 64 231ndash261
Cichello Michael S 2005 ldquoThe impact of firm size on pay-performance sensitivitiesrdquo
Journal of Corporate Finance 11 609ndash627
Davis Gerald F and E Han Kim 2007 ldquoBusiness ties and proxy voting by mutual
fundsrdquo Journal of Financial Economics 85 552ndash570
Dikolli Shane S Susan L Kulp and Karen L Sedatole 2009 ldquoTransient Institutional
Ownership and CEO Contractingrdquo Accounting Review 84 737ndash770
Edmans Alex Xavier Gabaix and Augustin Landier 2009 ldquoA Multiplicative Model of
Optimal CEO Incentives in Market Equilibriumrdquo Review of Financial Studies 22
4881ndash4917
Edmans Alex 2009 ldquoBlockholder Trading Market Efficiency and Managerial Myopiardquo
Journal of Finance 64 2481ndash2513
Ertimur Yonca Fabrizio Ferri and Volkan Muslu 2011 ldquoShareholder Activism and
CEO Payrdquo Review of Financial Studies 24 535ndash592
24
Gabaix Xavier and Augustin Landier 2008 ldquoWhy Has CEO Pay Increased So Muchrdquo
Quarterly Journal of Economics 123 49ndash100
Garen John 1994 ldquoExecutive compensation and principal-agent theoryrdquo Journal of
Political Economy 102 1175ndash1199
Gibbons Robert and Kevin J Murphy 1992 ldquoOptimal Incentive Contracts in the
Presence of Career Concerns Theory and Evidencerdquo Journal of Political Economy
100 468ndash505
Greene William H 2000 (4th
Edition) Econometric Analysis Prentice-Hall NJ
Guthrie Katherine Jan Sokolowsky and Kam-Ming Wan 2012 ldquoCEO Compensation
and Board Structure Revisitedrdquo Journal of Finance Forthcoming
Hadlock Charles R and Gerald B Lumer 1997 ldquoCompensation turnover and top
management incentives historical evidencerdquo Journal of Business 70 153ndash187
Hartzell Jay C and Laura T Starks 2003 ldquoInstitutional Investors and Executive
Compensationrdquo Journal of Finance 58 2351ndash2374
Huson Mark R Robert Parrino and Laura T Starks 2001 ldquoInternal Monitoring
Mechanisms and CEO Turnover A Long-Term Perspectiverdquo Journal of Finance 56
2265ndash2297
Jensen Michael C and Kevin J Murphy 1990 ldquoPerformance Pay and Top Management
Incentivesrdquo Journal of Political Economy 98 225ndash264
Jin Li 2002 ldquoCEO compensation diversification and incentivesrdquo Journal of Financial
Economics 66 29ndash63
Kang Qiang and Qiao Liu 2008 ldquoStock Market Information Production and Executive
Incentivesrdquo Journal of Corporate Finance 14 484ndash498
Merton Robert 1973 ldquoTheory of Rational Option Pricingrdquo Bell Journal of Economics
and Management Science 4 141-183
Morse Adair Vikram Nanda and Amit Seru 2011 ldquoAre Incentive Contracts Rigged by
Powerful CEOsrdquo Journal of Finance 66 1779ndash1821
Murphy Kevin J 1985 ldquoCorporate performance and managerial remuneration An
empirical analysisrdquo Journal of Accounting and Economics 7 11ndash42
Murphy Kevin J 1999 ldquoExecutive compensationrdquo in Orley Ashenfelter and David
Card Handbook of Labor Economics vol 3 Elsevier
Rosen Sherwin 1992 ldquoContracts and the market for executivesrdquo In Contract
Economics ed Lars Werin and Hans Wijkander 181ndash211 Cambridge MA
Blackwell
Rothberg Burton and Steven Lilien 2006 ldquoMutual Funds and Proxy Voting New
Evidence on Corporate Governancerdquo Journal of Business and Technology Law 1
157ndash184
Schaefer Scott 1998 ldquoThe dependence of payndashperformance sensitivity on the size of the
firmrdquo Review of Economics and Statistics 80 436ndash443
Yermack David 1995 ldquoDo Corporations Award CEO Stock Options Effectivelyrdquo
Journal of Financial Economics 39 237ndash69
25
Table I Univariate Descriptive Statistics
This table reports sample statistics for the key variables used in this study based on our comprehensive
ExecuComp dataset Institutional Concentration is the shares owned by the institutions with the five largest
holdings divided by the shares held by all institutions Top 5 Ownership is the proportion of shares
outstanding held by the institutions with the five largest holdings Total Institutional Ownership is the
proportion of shares held by all institutions Panel B shows the measures executive compensation The pay-
for-performance sensitivity (PPS) of option grants shows the change in executive wealth due to a $1000
change in firm in shareholder wealth Expressions for Salary and Total Compensation are in thousands of
dollars while Market Capitalization is in millions of dollars Panel C shows firm characteristics
Variable Mean Median Std Dev P10 P90
Panel A Institutional Investor Holdings
Institutional Concentration 0450 0424 0151 0282 0656
Top 5 Institutional Ownership 0229 0220 0101 0110 0355
Total Institutional Ownership 0546 0561 0206 0258 0801
Panel B Executive Compensation
PPS of Option Grants 1161 0156 3829 0 2822
Salary 36311 29462 25231 15179 67000
Total Compensation 258313 85878 980701 23875 511022
Panel C Firm Characteristics
Tobins Q 1744 1139 2738 0495 3322
Market Capitalization 560699 112094 1988108 19754 1048754
26
Table II Pairwise Correlation Coefficients
This table presents the correlations between the key variables based on our comprehensive
ExecuComp dataset Institutional Concentration is the shares owned by the institutions with the five
largest holdings divided by the shares held by all institutions Top 5 Ownership is the proportion of
shares outstanding held by the institutions with the five largest holdings Total Institutional
Ownership is the proportion of shares held by all institutions The pay-for-performance sensitivity
(PPS) of option grants is the change in executive wealth due to a $1000 change in shareholder
wealth
(1) (2) (3) (4) (5) (6) (7) (8)
(1) Institutional Concentration 1
(2) Top 5 Ownership 0268 1
(3) Total Ownership -0415 0706 1
(4) PPS of Option Grants 0122 0042 -0024 1
(5) Log Salary -0270 -0012 0171 0038 1
(6) Log Market Capitalization -0630 -0157 0261 -0156 0420 1
(7) Level Market Capitalization -0241 -0157 0007 -0058 0219 0534 1
(8) Tobins Q -0008 -0034 0002 0015 -0083 0066 0121 1
27
Table III
Descriptive Statistics by Market Capitalization Deciles
Market
Capitalization
Deciles
Market
Capitalization
($m)
Total
Compensation
($000)
Salary
($)
PPS Option
Grants
($)
Institutional
Concentration
()
Top 5 Ownership
()
Institutional
Total Ownership
()
1 (Small) 1181 76008 234700 2720 611 200 327
2 2668 90206 257955 1796 495 217 438
3 4193 105946 275120 1517 454 220 485
4 6045 129892 296067 1376 420 206 491
5 8742 155391 315864 1216 394 198 502
6 13086 191888 345942 1072 380 202 531
7 20790 239501 377021 0874 368 196 532
8 34920 314955 427808 0687 354 190 538
9 67747 450970 480319 0569 328 173 528
10 (Big) 366540 846066 614354 0297 275 138 502
28
Table IV Pay-for-Performance Sensitivity of Option Grants as a Function of Institutional Influence
Similar in construction to parts of HS (Table II) this table shows the coefficients from Tobit regressions of option grant pay-for-performance sensitivity per $1000 change in
shareholder wealth against the lagged change and contemporary change in shareholder wealth measures of institutional influence Tobinrsquos Q firm market capitalization and
dummy controls for CEO and year effects We estimate models (1) to (4) over the period 1992‒1997 using data supplied by HS and models (5) to (8) are estimated over the
whole sample Models (1) (2) (5) and (6) use Institutional Concentration as a measure of institutional influence which HS define as the shares owned by the institutions
with the five largest holdings divided by the shares held by all institutions These models control for total institutional ownership Models (3) (4) (7) and (8) use Top 5
Ownership as a measure of institutional influence which we define as the proportion of shares outstanding held by the institutions with the five largest holdings Coefficients
for industry and year controls are not presented One two and three asterisks denote significance at 10 5 and 1 levels
Hartzell and Starks Database 1992‒1997 Full Sample from ExecuComp 1992‒2002
(1) (2) (3) (4) (5) (6) (7) (8)
Change in Shareholder Wealth 77721 191781 89874 137368 -446 -218 -664 -203
(171) (429) (194) (317) (319) (168) (453) (158)
Lag1 Change in Shareholder Wealth 14446 99963 88719 8544 -144 0273 -0342 00852
(027) (212) (166) (184) (071) (015) (016) (005)
Lag1 Institutional Concentration 13994 -00692 26405 03304
(767) (032) (1402) (166)
Lag1 Total Institutional Ownership 08055 10434 07106 0937
(593) (781) (666) (898)
Lag1 Top 5 Institutional Ownership 10567 07457 12116 06293
(399) ( 283) (533) (278)
Lag1 Tobinrsquos Q 0068 00877 00647 00813 00498 00727 00508 00695
(337) (434) (319) (401) (504) (734) (51) (703)
Lag1 Market Capitalization -106202 -173515 -776 -117
(474) (742) (1011) (153)
Lag1 Log Market Capitalization -0297 -02319 -04111 -03908
(1576) (1439) (2997) (2841)
CEO Dummy 13745 13682 13765 13674 28197 28128 28155 28143
(1913) (1918) (1914) (1916) (3388) (3396) (3377) (34)
Number of Observations 33878 33878 33878 33878 97679 97679 97679 97679
29
Table V Pay-for-Performance Sensitivity of Cash Compensation (Salary plus Bonus) as a Function of Institutional Influence - Level Market Capitalization
Similar in parts to models (1) and (2) of HS (Table IV) this table displays the coefficients from OLS regressions of the change in cash compensation against the lagged
change and contemporary change in shareholder wealth measures of institutional influence Tobinrsquos Q firm market capitalization and dummy controls for CEO and year
effects We estimate models (1) to (4) over the period 1992‒1997 based on our ExecuComp sample and models (5) to (8) over the whole sample Models (1) (2) (5) and (6)
use Institutional Concentration as a measure of institutional influence which HS define as the shares owned by the institutions with the five largest holdings divided by the
shares held by all institutions These models control for total institutional ownership Models (4) and (8) use Top 5 Ownership as a measure of institutional influence which
we define as the proportion of shares outstanding held by the institutions with the 5 largest holdings Coefficients for industry and year controls are not presented One two
and three asterisks denote significance at 10 5 and 1 levels
Our Database based on ExecuComp 1992‒1997 Full Sample based on ExecuComp 1992‒2002
(1) (2) (3) (4) (5) (6) (7) (8)
Lag1 Change in Shareholder Wealth 00014 00017 00088 00089 -00006 -00007 00072 00086
(051) (062) (129) (103) (027) (029) (131) (16)
Change in Shareholder Wealth -00102 00102 00105 00032 00102 00101
(097) (292) (293) (055) (611) (608)
Change Shareholder Wealth Lag1 Institutional
Concentration
00435 00764 00363 00253
(463) (258) (689) (144)
Lag1 Change Shareholder Wealth Lag1
Institutional Concentration
-00197 -00283
(122) (197)
Lag1 Institutional Concentration -492182 -288861 400149 453623
(09) (056) (098) (114)
Lag1 Total Institutional Ownership 720688 538372 655091 559587 682069 726226
(276) (232) (281) (314) (308) (322)
Lag1 Change Shareholder Wealth Lag1 Top 5
Institutional Ownership
-0046 -00689
(091) (256)
Lag1 Top 5 Institutional Ownership 583044 1191634
(092) (25)
Lag1 Tobinrsquos Q -32761 -36806 -19422 -18023 -54283 -48681 -29534 -24243
(138) (172) (077) (078) (227) (245) (136) (111)
Lag1 Market Capitalization 00004 00004 00002 00006 00007 00008 00007 00007
(042) (044) (022) (07) (086) (092) (082) (09)
CEO Dummy 768035 765474 768334 773634 616315 61729 619028 618535
(354) (355) (356) (355) (417) (418) (419) (418)
Adjusted R-Sq 00057 0006 00052 0005 00097 00097 00098 00102
Number of Observations 41305 41305 41305 41305 85518 85518 85518 85518
30
Table VI Level of Salary as a Function of Institutional Influence
Similar in parts to models (1) and (2) of HS (Table V) this table displays the coefficients from OLS regressions of salary against the lagged change and contemporary change in
shareholder wealth measures of institutional influence Tobinrsquos Q firm market capitalization and dummy controls for CEO industry and year effects We estimate models (1) to
(4) over the period 1992‒1997 using the actual data supplied by HS and models (5) to (8) over the whole sample 1992‒2002 Models (1) (2) (5) and (6) use Institutional
Concentration as a measure of institutional influence which HS define as the shares owned by the institutions with the five largest holdings divided by the shares held by all
institutions These models control for total institutional ownership Models (3) (4) (7) and (8) use Top 5 Ownership as a measure of institutional influence which we define as
the proportion of shares outstanding held by the institutions with the 5 largest holdings Coefficients for industry and year controls are not presented One two and three
asterisks denote significance at 10 5 and 1 levels
Hartzell and Starks Database 1992‒1997 Full Sample based on ExecuComp 1992‒2002
(1) (2) (3) (4) (5) (6) (7) (8)
Change in Shareholder Wealth 04125E-4 0003 -2439E-4 0003 00021 00012 00024 00012
(057) (464) (031) (468) (1004) (72) (1041) (706)
Lag1 Change in Shareholder Wealth -00062 7856E-4 -0007 9003E-4 -00014 00002 -00014 00003
(594) (098) (592) (112) (522) (123) (5) (145)
Lag1 Institutional Concentration -2475087 555134 -2734983 943317
(3245) (710) (4965) (1671)
Lag1 Total Institutional Ownership 1020109 217488 936337 252464
(1904) (432) (2471) (704)
Lag1 Top 5 Institutional Ownership -148355 31354 -87926 423166
(139) (332) (118) (643)
Lag1 Tobinrsquos Q -195348 -252191 -192344 -25158 -119837 -146373 -122132 -143821
(2812) (3762) (2721) (3752) (248) (274) (2583) (2727)
Lag1 Market Capitalization 844E-06 103E-5 00034 00039
(2781) (3051) (2874) (3104)
Lag1 Log Market Capitalization 750021 723625 776657 732279
(7511) (9285) (11224) (14039)
CEO Dummy 2683978 2702238 2668244 2702417 3281133 3285103 3292175 3283145
(8625) (9277) (8246) (9271) (12739) (13616) (12336) (13605)
Adjusted R-Sq 04604 05302 04187 05295 03856 04588 03476 04574
Number of Observations 36280 36280 36280 36280 97679 97679 97679 97679
31
Table VII Level of Total Direct Compensation as a Function of Institutional Influence
Similar in parts to models (3) and (5) of HS (Table V) this table displays the coefficients from OLS regressions of total direct compensation against the lagged change and
contemporary change in shareholder wealth measures of institutional influence Tobinrsquos Q firm market capitalization and dummy controls for CEO industry and year effects
We estimate models (1) to (4) over the period 1992‒1997 and models (5) to (8) over the whole sample Models (1) (2) (5) and (6) use Institutional Concentration as a measure
of institutional influence which HS define as the shares owned by the institutions with the five largest holdings divided by the shares held by all institutions These models
control for total institutional ownership Models (3) (4) (7) and (8) use Top 5 Ownership as a measure of institutional influence which we define as the proportion of shares
outstanding held by the institutions with the five largest holdings Coefficients for industry and year controls are not presented One two and three asterisks denote significance
at 10 5 and 1 levels
Hartzell and Starks Database 1992‒1997 Full Sample from ExecuComp 1992‒2002
(1) (2) (3) (4) (5) (6) (7) (8)
Change in Shareholder Wealth 005779 007006 005282 006791 00423 00148 0046 00138
(415) (581) (369) (565) (203) (07) (22) (066)
Lag1 Change in Shareholder Wealth 003182 00714 00272 00715 00613 0126 00614 01278
(126) (298) (107) (299) (275) (488) (276) (494)
Lag1 Institutional Concentration -1608633 3248998 -3424892 3334186
(151) (283) (1172) (118)
Lag1 Total Institutional Ownership 104013 5473128 1531013 -2926764
(1645) (831) (876) (158)
Lag1 Top 5 Institutional Ownership 188585 3927986 -2159177 -2640879
(016) (343) (067) (092)
Lag1 Tobinrsquos Q -445692 -803612 -454859 -833566 3257027 2971201 3213936 3086205
(427) (791) (432) (831) (83) (77) (815) (793)
Lag1 Market Capitalization 49E-5 615E-5 01031 01088
(113) (1365) (1555) (168)
Lag1 Log Market Capitalization 4683497 4759377 1645156 1427856
(2775) (3469) (3679) (4239)
CEO Dummy 1501307 1508886 1493688 1508736 4713079 4713629 47278 4703908
(2648) (2684) (2618) (2683) (282) (2828) (2819) (2824)
Adjusted R-Sq 01861 02060 01695 02051 01071 01113 01033 011
Number of Observations 33229 33229 33229 33229 97679 97679 97679 97679
32
Table VIII Two Stage Least Squares (Reverse Causality) Tests - Executive Compensation Measures as a Function of Institutional Concentration
This table shows the coefficients from the Two-Stage Least Squares (2SLS) regressions of executive compensation measures against the lagged change
and contemporary change in shareholder wealth instrumented institutional concentration and total institutional ownership Tobinrsquos Q the natural
logarithm of firm market capitalization and dummy variable controls for CEO and year effects We instrument measures of institutional concentration
and total institutional ownership by orthogonalizing with respect to option grant pay-for-performance sensitivity salary and total compensation
Coefficients for industry and year controls are not presented First stage results are not presented One two and three asterisks denote significance at
10 5 and 1 levels
ExecuComp Sample 1992‒1997 Full Sample from ExecuComp 1992‒2002
PPSOG Salary TC PPSOG Salary TC
Change in Shareholder Wealth 144E-05 0004 02133 -409E-06 00011 00157
(308) (775) (596) (691) (638) (068)
Lag1 Change in Shareholder Wealth -240E-05 -00004 00866 118E-06 00002 01422
(131) (087) (346) (097) (116) (51)
Lag1 Institutional Concentration -00898 668681 3263143 02024 916656 2856417
(052) (801) (176) (163) (1545) (1004)
Lag1 Total Institutional Ownership 01021 -43997 2512587 02983 -104171 -1028347
(09) (084) (153) (4) (269) (487)
Lag1 Tobinrsquos Q 00754 -225606 119297 00655 -158657 2997205
(596) (1649) (053) (834) (2535) (727)
Lag1 Log Market Capitalization -04234 828713 7864046 -04002 825106 1685638
(2195) (8132) (2977) (4106) (11133) (373)
CEO Dummy 15848 3214117 2950274 19149 3205862 4529101
(2223) (9389) (2255) (3802) (1331) (2758)
Overall R-Sq -- 04981 01206 -- 04738 01178
Number of Observations 41305 41305 41305 85518 85518 85518
33
Table IX Two Stage Least Squares (Reverse Causality) Tests - Executive Compensation Measures as a Function of Top 5 Ownership
This table shows the coefficients from 2SLS regressions of executive compensation measures against the lagged change and contemporary change
in shareholder wealth instrumented Top 5 institutional ownership Tobinrsquos Q natural logarithm of firm market capitalization and dummy variable
controls for CEO and year effects We instrument Top 5 institutional ownership by orthogonalizing with respect to option grant pay-for-
performance sensitivity salary and total compensation Coefficients for industry and year controls are not presented First stage results are not
presented One two and three asterisks denote significance at 10 5 and 1 levels
ExecuComp sample 1992‒1997 Full Sample 1992‒2002
PPSOG Salary TC PPSOG Salary TC
Change in Shareholder Wealth 136E-05 00041 02128 -407E-06 00011 00146
(293) (79) (596) (692) (619) (063)
Lag1 Change in Shareholder Wealth -241E-05 -00003 00868 116E-06 00003 01439
(133) (07) (347) (096) (142) (514)
Lag1 Top 5 Institutional Ownership -03774 -187512 -3848531 00179 -104261 -126659
(163) (186) (166) (011) (149) (436)
Lag1 Tobinrsquos Q 00731 -223886 10074 00642 -155467 3115073
(573) (1631) (045) (819) (2509) (75)
Lag1 Log Market Capitalization -0416 775637 7696302 -04015 764632 1462536
(2345) (10084) (3766) (3688) (13708) (4072)
CEO Dummy 15854 3211991 2950158 19152 3203152 4519385
(2223) (9381) (2254) (3804) (133) (2752)
Overall R-Sq -- 04967 01202 -- 04722 01168
Number of Observations 41305 41305 41305 85518 85518 85518
34
Table X Specification Tests utilizing Box-Cox Transformation of Market Capitalization
This table presents results from a Box-Cox transformation of market capitalization In
each model we specify transformation of market capitalization only Panel A shows
results for the period 1992‒1997 and Panel B shows results for the entire period
1992‒2002 The first part of each panel shows the Lambda coefficient that applies to
the transformed market capitalization The second part of each panel shows log
likelihood scores with LR Chi2 Statistics in parentheses ‒ these test for the
appropriateness of a linear or log model specification
Panel A Our ExecuComp sample 1992‒1997
(1) (2) (3)
PPSOG Salary TC
Lambda -05143 01881 02876
(-1914) (2413) (1329)
Test H0 (Log Likelihood Score)
Lambda = -1 -1279748 -3178575 -4795326
(23866) (85952) (99688)
Lambda = 0 -1280152 -3138456 -4791159
(31934) (57147) (16339)
Lambda = 1 -1283157 -3163807 -4793506
(92039) (564156) (63292)
Panel B Full ExecuComp sample 1992‒2002
(1) (2) (4)
PPSOG Salary TC
Lambda -044 01853 04036
(-3015) (40) (3666)
Test H0 (Log Likelihood Score)
Lambda = 0 -2688334 -6476494 -10384490
(66599) (160356) (126236)
Lambda = 1 -2694886 -6538506 -10386795
(197644) (1400605) (172345)
35
Figure 2 Institutional Ownership Measures (LHA) vs Total
Compensation (RHA) by Market Capitalization
0
10
20
30
40
50
60
70
1 2 3 4 5 6 7 8 9 10
$00
$10
$20
$30
$40
$50
$60
$70
$80
$90
Millio
ns
InstitutionalConcentration (LHS)Top5 Ownership (LHS)
Institutional TotalOwnership (LHS)Total Compensation(RHS)
Page 14
14
amalgam of price levels over their sample period 1992‒1997 and ours are in consistent
dollars of June 30 2002 However there has also been significant growth in
compensation levels from 1998 to 2002
ltltINSERT TABLE Igtgt
Based on our comprehensive data in Table II we examine correlations between the
key variables employed in the study Of concern is the large negative correlation of 63
percent which has been discussed previously (Section IIA) between the HS measure of
institutional concentration and the natural logarithm of firm market capitalization
ltltINSERT TABLE IIgtgt
In view of this high correlation Table III provides a breakdown of the three key pay
variables and three institutional ownership variables by size deciles We find that the high
institutional concentration but low institutional ownership share of total shareholdings
mentioned above is evident in the smallest decile of firms by size This decile has low
salary and even lower total compensation in a relative sense but by far the highest pay-
for-performance sensitivity of option grants Small firms award executives a far higher
proportion of market capitalization Since small firm market-capitalization is negligible in
comparison with the highest decile (300 times larger in magnitude) total compensation is
very small in absolute value Moreover due to the domination of the pay-size effect base
salaries are also exceedingly small
ltltINSERT TABLE IIIgtgt
Figure 1 based on Table III shows how as institutional concentration falls with
increases in stock size (but also as the total share of institutional ownership increases)
and the pay-for-performance sensitivity of option grants falls Thus because of what
appears to be a weak size control and size-influenced concentration measure HS
15
potentially pick up a size effect as their main result but not necessarily a monitoring
effect Figure 2 shows that as firm size increases up to decile 8 at least the share of
institutional total ownership increases as does total compensation
ltltINSERT FIGURES 1 AND 2 gtgt
D Methodology
Following HS we use two estimation techniques to examine the effect of institutional
investor influence on executive compensation Due to the censored nature of option grant
pay-for-performance sensitivity this compensation measure most appropriately uses
Tobit model specification whereas an Ordinary Least Squares (OLS) model specification
is used for all other compensation measures We model compensation measures as a
function of contemporaneous and lagged change in shareholder wealth institutional
influence Tobinrsquos Q and market capitalization Controls for CEO industry and year
effects are also used Augmentation occurs to test the effect of institutional influence on
the change in executive cash and total compensation with an interaction term that is the
product of institutional influence and the change in shareholder wealth
The variables used to model executive compensation have the following definitions
The lagged and contemporary measure of change in shareholder wealth is simply the
difference in market capitalization of the firm The measures of institutional influence
used are Institutional Concentration and Top 5 Ownership Use of the two types of
institutional influence occurs in separate models In order to replicate the HS
methodology models using Institutional Concentration have an additional control
consisting of lagged total institutional holdings and Tobinrsquos Q controls for performance
and growth opportunities Market capitalization of the firm is the number of shares
outstanding multiplied by the share price with inclusion both in levels and natural
16
logarithmic form A dummy variable is included to identify whether the executive is the
CEO Dummy variables are also included to control for industry effects This variable
takes a value of 1 for the two-digit Standard Industrial Classification (SIC) in which the
firm operates A year fixed effects (dummy) variable is included taking a value of 1 if the
observation was from the given year
IV Executive Incentives as a Function of Institutional Investor Influence
This section presents the results for the effect of institutional influence on option grant
pay-for-performance sensitivity and changes in compensation
A Option Grant Pay-for-Performance Sensitivity
Table IV presents the results for the effect of institutional influence on option grant
pay-for-performance sensitivity
ltltINSERT TABLE IVgtgt
Model (1) is a direct replication of HS using the HS database as supplied to us by HS
with all the variables computed in the same way and in the same functional form as with
HS The coefficient and t-statistic on lagged institutional concentration interacted with the
change in shareholder wealth is 13994 and 767 that compares to 1358 and 842 in HS
We are unable to account for these minor discrepancies for values that should be
identical In model (2) the only change to the model is that we use the natural logarithm
of market capitalization instead of the level of market capitalization This single change
has a sizeable effect on the coefficient on institutional concentration The coefficient on
institutional concentration is now -00692 and is insignificant The fact that an order
preserving transformation to a control variable can change the results is of interest The
log-transformation to market capitalization also changes the coefficient on Top 5
Ownership in models (3) and (4) The coefficient on Top 5 Ownership is significant in
17
both model (3) and in model (4) but the better size control reduces the coefficient by over
25 percent The better size control does not affect Top 5 Ownership by as much as its
design ensures less sensitivity to what could be inadequate size controls
We also test the effect of institutional influence on option grant pay-for-performance
sensitivity over the period 1992‒2002 This sample is then approximately three times as
large as the sample used by HS Institutional concentration is highly positively significant
in model (5) similar to the results for the period 1992‒1997 When we use the log of
market capitalization in model (6) the coefficient on institutional concentration remains
positive and significant at the 10 percent level with a t-statistic of 166 Of importance
though is that the coefficient on institutional concentration is one-eighth of the coefficient
in model (5) Even for a measure of institutional influence which we believe is less
sensitive to firm size the use of log market capitalization reduces the coefficient on Top 5
Ownership by approximately 50 percent from model (7) to model (8)
B Pay-for-Performance Sensitivity due to Changes in Cash and Total
Compensation
For robustness HS examine the effect that institutional influence conditioned on firm
performance has on a changes in cash and total executive compensation using the general
method of Jensen and Murphy (1990) to compute pay-for-performance sensitivity
ltltINSERT TABLE V gtgt
Table V presents the effect of institutional influence on the change in cash
compensation using the HS size control measure throughout12
Model (1) replicates the
specification in HS for the period of 1992‒1997 and we produce a similar result ndash the
interaction of lagged institutional concentration and contemporaneous change in
12 This test uses our data for the two periods 1992‒1997 and 1992‒2002 We do not use the HS data for
cash compensation as data for bonus is missing in the data supplied to us
18
shareholder wealth positively and significantly relates to a change in executive cash
compensation In model (2) we include the contemporaneous change in shareholder
wealth and the lagged institutional concentration as stand-alone controls as indicated by
Greenersquos (2000) specification of equation (1) above and the interaction term remains
significant In model (3) we interact lagged institutional concentration with the lagged
change in shareholder wealth as in equation (2) above that tries to take account of what
participants could have observed at the time The interaction term is then negative but
insignificant altering the HS result Model (4) provides a similar result where our
preferred proxy Top 5 Ownership captures concentrated institutional influence In
models (5) to (8) we use our full sample In model (5) we find that the lagged
institutional concentration interaction with the contemporaneous change in shareholder
wealth is positive and significant in HSrsquos specification However including change in
contemporaneous shareholder wealth and lagged institutional concentration as separate
control variables as suggested by equation (1)rsquos specification removes the significance
from the interaction term in model (6) Moreover in model (7) when lagged institutional
concentration is interaction with the lagged change in shareholder wealth in accord with
equation (2) this interaction term is now negative and significant reversing the HS
finding We obtain a similar result in model (8) when we measure institutional
concentration influence using Top 5 Ownership13
V The Levels of Executive Fixed and Total Compensation as Functions of
Institutional Investor Influence
13 In unreported results we use the log of market capitalization as a size control These changes to the size
control do not dramatically change the findings from Table V The reason why this size issue is not as
important as in the findings for option grants pay-for-performance is that the change in cash compensation
is not so strongly influenced by firm size In unreported results available on request we also examine the
effect of institutional influence on changes in total compensation with similar results to the change in cash
compensation
19
Two measures of compensation are examined ndash salary and total compensation
ltltINSERT TABLE VI gtgt
Table VI examines the effect of institutional influence on the level of salary In model
(1) we use the HS sample and specification to obtain almost identical results The
coefficient on institutional concentration is (negative) $247509 that is almost identical to
(negative) $247967 in HS Both are highly significant In model (2) we make a single
change to the model ndash we use the natural logarithm of market capitalization This single
change reverses the sign of institutional concentration with the coefficient still being
significant at the 1 percent level In models (3) and (4) we measure institutional influence
using Top 5 Ownership In model (3) where we measure firm size using the level of
market capitalization Top 5 Ownership is insignificant When we use the natural
logarithm of market capitalization in model (4) Top 5 Ownership is significant and
positively relates to the salary level In models (5) to (8) we estimate the same models
using our full sample We find similar results for the full sample as for the period 1992ndash
1997
ltltINSERT TABLE VII gtgt
Table VII examines the effect of institutional influence on total compensation In
model (1) we replicate the HS methodology using the HS data to obtain a similar
significant negative relation between institutional concentration and total compensation of
(negative) $1608633 However after changing to a logarithmic specification of market
capitalization in model (2) institutional concentration increases total compensation by a
significant $32400014
Importantly when institutional influence is measured using Top 5
14 = (5688751 times 0101)
20
Ownership an insignificant relation exists with total compensation in model (3) using the
HS size control but it becomes both positive and significant in (4) with the logged size
control In models (5) and (6) where models are estimated using the full sample we
observe a similar reversal in sign as occurs in models (1) and (2) However now the
magnitude of the reversal is almost equal with a difference of about $67m that is very
large indeed In model (5) the coefficient on institutional concentration is (negative)
$342m and then in model (6) the coefficient is $33m Top 5 Ownership is insignificant
in both models (7) and (8) unlike the results for Salary
VI Robustness
Two issues are examined in is this Section ndash reverse causality and the appropriate
transformation of market capitalization A methodological issue inherent in examining the
effect of ownership-based measures of institutional influence on executive compensation
is endogeneity driven by reverse causality We utilize an instrumental variables approach
where orthogonalized institutional influence measures are regressed on each of the
compensation measures lagged option grant pay-for-performance sensitivity salary and
total compensation In addition market capitalization and a CEO dummy are included in
each regression We use the residual from each regression as an instrument for
institutional influence Such a process removes any preferences for compensation from
the institutional influence measures
ltltINSERT TABLES VIII AND IXgtgt
Table VIII presents the results where we instrument the HS measure lagged
institutional concentration with a compensation-orthogonalized equivalent The analysis
indicates that the instrumented institutional concentration measure is positively and
significantly related to the level of salary and total compensation and insignificantly
21
related to option grant pay-for-performance sensitivity (PPSOG) Table IX presents
results where we instrument Top 5 Ownership with a compensation-orthogonalized
equivalent The instrumented Top 5 Ownership measure insignificantly relates to option
grant pay-for-performance sensitivity and barely negatively relates to salary although it
negatively relates to total compensation suggesting reverse causality is driving a
significant portion of our previous results
ltltINSERT TABLE Xgtgt
Another issue examined in this section is the appropriate transformation of market
capitalization Aneuryn-Evans and Deaton (1980) provide an evaluation of methods
comparing linear and log-linear specifications We use the Box-Cox test to compare
logarithm versus linear specifications In applying the Box-Cox test we specify
transformation of only market capitalization ndash the dependent variable and all other
independent variables remain unchanged Table X presents the lambda coefficient for
each model and the log likelihood score In all models in Panel A and B a lambda of zero
receives a higher (closer to zero) log likelihood score suggesting the log specification is
appropriate
VII Conclusion
Hartzell and Starks (2003) appears to document an effective role for institutional
investors in increasing executive incentives and lowering compensation levels Our paper
re-addresses the issue of concentrated institutional monitoring from a perspective
informed by an additional decade of research since their paper appeared We cannot find
evidence that concentrated institutional investors really have any positive (or negative for
that matter) effect on executive incentive levels While superficially we find evidence
that they may raise executive pay levels this is most likely driven by an institutional
22
preference for highly remunerated executives in large companies and thus by reverse
causality Moreover our findings are not consistent with the rent-extraction hypothesis
suggesting that firms systematically overpay executives in the absence of effective
monitoring by outside parties such as institutional investors
23
References
Almazan Andres Jay C Hartzell and Laura T Starks 2005 ldquoActive Institutional
Shareholders and Costs of Monitoring Evidence from Executive Compensationrdquo
Financial Management Winter 5ndash34
Aneuryn-Evans Gwyn and Angus Deaton 1980 ldquoTesting Linear versus Logarithmic
Regression Modelsrdquo Review of Economic Studies 47 (1) Econometrics Issue 275ndash
291
Babenko Ilona 2009 ldquoShare Repurchases and Pay-Performance Sensitivity of Employee
Compensation Contractsrdquo Journal of Finance 64117-150
Baker George P and Brian J Hall 2004 ldquoCEO Incentives and Firm Sizerdquo Journal of
Labor Market Economics 22 767ndash798
Bebchuk Lucian A Jesse M Fried and David Walker 2002 ldquoManagerial power and
rent extraction in the design of executive compensationrdquo University of Chicago Law
Review 69 751ndash846
Bebchuk Lucian A and Jesse M Fried 2003 ldquoExecutive Compensation as an Agency
Problemrdquo Journal of Economic Perspectives 17 71ndash92
Bebchuk Lucian A and Jesse M Fried 2004 Pay without Performance The Unfulfilled
Promise of Executive Compensation Cambridge MA Harvard University Press
Bertrand Marianne and Sendhil Mullainathan 2001 ldquoAre CEOs Rewarded for Luck
The Ones Without Principals Arerdquo Quarterly Journal of Economics 116 901ndash32
Black Fischer and Myron Scholes 1973 ldquoThe Pricing of Options and Corporate
Liabilitiesrdquo Journal of Political Economy 81 637ndash54
Cadman Brian Sandy Klasa and Steve Matsunaga 2010 ldquoDeterminants of CEO Pay A
Comparison of ExecuComp and Non-ExecuComp Firmsrdquo The Accounting Review 85
1511ndash1543
Chhaochharia Vidhi and Yaniv Grinstein 2009 ldquoCEO Compensation and Board
Structurerdquo Journal of Finance 64 231ndash261
Cichello Michael S 2005 ldquoThe impact of firm size on pay-performance sensitivitiesrdquo
Journal of Corporate Finance 11 609ndash627
Davis Gerald F and E Han Kim 2007 ldquoBusiness ties and proxy voting by mutual
fundsrdquo Journal of Financial Economics 85 552ndash570
Dikolli Shane S Susan L Kulp and Karen L Sedatole 2009 ldquoTransient Institutional
Ownership and CEO Contractingrdquo Accounting Review 84 737ndash770
Edmans Alex Xavier Gabaix and Augustin Landier 2009 ldquoA Multiplicative Model of
Optimal CEO Incentives in Market Equilibriumrdquo Review of Financial Studies 22
4881ndash4917
Edmans Alex 2009 ldquoBlockholder Trading Market Efficiency and Managerial Myopiardquo
Journal of Finance 64 2481ndash2513
Ertimur Yonca Fabrizio Ferri and Volkan Muslu 2011 ldquoShareholder Activism and
CEO Payrdquo Review of Financial Studies 24 535ndash592
24
Gabaix Xavier and Augustin Landier 2008 ldquoWhy Has CEO Pay Increased So Muchrdquo
Quarterly Journal of Economics 123 49ndash100
Garen John 1994 ldquoExecutive compensation and principal-agent theoryrdquo Journal of
Political Economy 102 1175ndash1199
Gibbons Robert and Kevin J Murphy 1992 ldquoOptimal Incentive Contracts in the
Presence of Career Concerns Theory and Evidencerdquo Journal of Political Economy
100 468ndash505
Greene William H 2000 (4th
Edition) Econometric Analysis Prentice-Hall NJ
Guthrie Katherine Jan Sokolowsky and Kam-Ming Wan 2012 ldquoCEO Compensation
and Board Structure Revisitedrdquo Journal of Finance Forthcoming
Hadlock Charles R and Gerald B Lumer 1997 ldquoCompensation turnover and top
management incentives historical evidencerdquo Journal of Business 70 153ndash187
Hartzell Jay C and Laura T Starks 2003 ldquoInstitutional Investors and Executive
Compensationrdquo Journal of Finance 58 2351ndash2374
Huson Mark R Robert Parrino and Laura T Starks 2001 ldquoInternal Monitoring
Mechanisms and CEO Turnover A Long-Term Perspectiverdquo Journal of Finance 56
2265ndash2297
Jensen Michael C and Kevin J Murphy 1990 ldquoPerformance Pay and Top Management
Incentivesrdquo Journal of Political Economy 98 225ndash264
Jin Li 2002 ldquoCEO compensation diversification and incentivesrdquo Journal of Financial
Economics 66 29ndash63
Kang Qiang and Qiao Liu 2008 ldquoStock Market Information Production and Executive
Incentivesrdquo Journal of Corporate Finance 14 484ndash498
Merton Robert 1973 ldquoTheory of Rational Option Pricingrdquo Bell Journal of Economics
and Management Science 4 141-183
Morse Adair Vikram Nanda and Amit Seru 2011 ldquoAre Incentive Contracts Rigged by
Powerful CEOsrdquo Journal of Finance 66 1779ndash1821
Murphy Kevin J 1985 ldquoCorporate performance and managerial remuneration An
empirical analysisrdquo Journal of Accounting and Economics 7 11ndash42
Murphy Kevin J 1999 ldquoExecutive compensationrdquo in Orley Ashenfelter and David
Card Handbook of Labor Economics vol 3 Elsevier
Rosen Sherwin 1992 ldquoContracts and the market for executivesrdquo In Contract
Economics ed Lars Werin and Hans Wijkander 181ndash211 Cambridge MA
Blackwell
Rothberg Burton and Steven Lilien 2006 ldquoMutual Funds and Proxy Voting New
Evidence on Corporate Governancerdquo Journal of Business and Technology Law 1
157ndash184
Schaefer Scott 1998 ldquoThe dependence of payndashperformance sensitivity on the size of the
firmrdquo Review of Economics and Statistics 80 436ndash443
Yermack David 1995 ldquoDo Corporations Award CEO Stock Options Effectivelyrdquo
Journal of Financial Economics 39 237ndash69
25
Table I Univariate Descriptive Statistics
This table reports sample statistics for the key variables used in this study based on our comprehensive
ExecuComp dataset Institutional Concentration is the shares owned by the institutions with the five largest
holdings divided by the shares held by all institutions Top 5 Ownership is the proportion of shares
outstanding held by the institutions with the five largest holdings Total Institutional Ownership is the
proportion of shares held by all institutions Panel B shows the measures executive compensation The pay-
for-performance sensitivity (PPS) of option grants shows the change in executive wealth due to a $1000
change in firm in shareholder wealth Expressions for Salary and Total Compensation are in thousands of
dollars while Market Capitalization is in millions of dollars Panel C shows firm characteristics
Variable Mean Median Std Dev P10 P90
Panel A Institutional Investor Holdings
Institutional Concentration 0450 0424 0151 0282 0656
Top 5 Institutional Ownership 0229 0220 0101 0110 0355
Total Institutional Ownership 0546 0561 0206 0258 0801
Panel B Executive Compensation
PPS of Option Grants 1161 0156 3829 0 2822
Salary 36311 29462 25231 15179 67000
Total Compensation 258313 85878 980701 23875 511022
Panel C Firm Characteristics
Tobins Q 1744 1139 2738 0495 3322
Market Capitalization 560699 112094 1988108 19754 1048754
26
Table II Pairwise Correlation Coefficients
This table presents the correlations between the key variables based on our comprehensive
ExecuComp dataset Institutional Concentration is the shares owned by the institutions with the five
largest holdings divided by the shares held by all institutions Top 5 Ownership is the proportion of
shares outstanding held by the institutions with the five largest holdings Total Institutional
Ownership is the proportion of shares held by all institutions The pay-for-performance sensitivity
(PPS) of option grants is the change in executive wealth due to a $1000 change in shareholder
wealth
(1) (2) (3) (4) (5) (6) (7) (8)
(1) Institutional Concentration 1
(2) Top 5 Ownership 0268 1
(3) Total Ownership -0415 0706 1
(4) PPS of Option Grants 0122 0042 -0024 1
(5) Log Salary -0270 -0012 0171 0038 1
(6) Log Market Capitalization -0630 -0157 0261 -0156 0420 1
(7) Level Market Capitalization -0241 -0157 0007 -0058 0219 0534 1
(8) Tobins Q -0008 -0034 0002 0015 -0083 0066 0121 1
27
Table III
Descriptive Statistics by Market Capitalization Deciles
Market
Capitalization
Deciles
Market
Capitalization
($m)
Total
Compensation
($000)
Salary
($)
PPS Option
Grants
($)
Institutional
Concentration
()
Top 5 Ownership
()
Institutional
Total Ownership
()
1 (Small) 1181 76008 234700 2720 611 200 327
2 2668 90206 257955 1796 495 217 438
3 4193 105946 275120 1517 454 220 485
4 6045 129892 296067 1376 420 206 491
5 8742 155391 315864 1216 394 198 502
6 13086 191888 345942 1072 380 202 531
7 20790 239501 377021 0874 368 196 532
8 34920 314955 427808 0687 354 190 538
9 67747 450970 480319 0569 328 173 528
10 (Big) 366540 846066 614354 0297 275 138 502
28
Table IV Pay-for-Performance Sensitivity of Option Grants as a Function of Institutional Influence
Similar in construction to parts of HS (Table II) this table shows the coefficients from Tobit regressions of option grant pay-for-performance sensitivity per $1000 change in
shareholder wealth against the lagged change and contemporary change in shareholder wealth measures of institutional influence Tobinrsquos Q firm market capitalization and
dummy controls for CEO and year effects We estimate models (1) to (4) over the period 1992‒1997 using data supplied by HS and models (5) to (8) are estimated over the
whole sample Models (1) (2) (5) and (6) use Institutional Concentration as a measure of institutional influence which HS define as the shares owned by the institutions
with the five largest holdings divided by the shares held by all institutions These models control for total institutional ownership Models (3) (4) (7) and (8) use Top 5
Ownership as a measure of institutional influence which we define as the proportion of shares outstanding held by the institutions with the five largest holdings Coefficients
for industry and year controls are not presented One two and three asterisks denote significance at 10 5 and 1 levels
Hartzell and Starks Database 1992‒1997 Full Sample from ExecuComp 1992‒2002
(1) (2) (3) (4) (5) (6) (7) (8)
Change in Shareholder Wealth 77721 191781 89874 137368 -446 -218 -664 -203
(171) (429) (194) (317) (319) (168) (453) (158)
Lag1 Change in Shareholder Wealth 14446 99963 88719 8544 -144 0273 -0342 00852
(027) (212) (166) (184) (071) (015) (016) (005)
Lag1 Institutional Concentration 13994 -00692 26405 03304
(767) (032) (1402) (166)
Lag1 Total Institutional Ownership 08055 10434 07106 0937
(593) (781) (666) (898)
Lag1 Top 5 Institutional Ownership 10567 07457 12116 06293
(399) ( 283) (533) (278)
Lag1 Tobinrsquos Q 0068 00877 00647 00813 00498 00727 00508 00695
(337) (434) (319) (401) (504) (734) (51) (703)
Lag1 Market Capitalization -106202 -173515 -776 -117
(474) (742) (1011) (153)
Lag1 Log Market Capitalization -0297 -02319 -04111 -03908
(1576) (1439) (2997) (2841)
CEO Dummy 13745 13682 13765 13674 28197 28128 28155 28143
(1913) (1918) (1914) (1916) (3388) (3396) (3377) (34)
Number of Observations 33878 33878 33878 33878 97679 97679 97679 97679
29
Table V Pay-for-Performance Sensitivity of Cash Compensation (Salary plus Bonus) as a Function of Institutional Influence - Level Market Capitalization
Similar in parts to models (1) and (2) of HS (Table IV) this table displays the coefficients from OLS regressions of the change in cash compensation against the lagged
change and contemporary change in shareholder wealth measures of institutional influence Tobinrsquos Q firm market capitalization and dummy controls for CEO and year
effects We estimate models (1) to (4) over the period 1992‒1997 based on our ExecuComp sample and models (5) to (8) over the whole sample Models (1) (2) (5) and (6)
use Institutional Concentration as a measure of institutional influence which HS define as the shares owned by the institutions with the five largest holdings divided by the
shares held by all institutions These models control for total institutional ownership Models (4) and (8) use Top 5 Ownership as a measure of institutional influence which
we define as the proportion of shares outstanding held by the institutions with the 5 largest holdings Coefficients for industry and year controls are not presented One two
and three asterisks denote significance at 10 5 and 1 levels
Our Database based on ExecuComp 1992‒1997 Full Sample based on ExecuComp 1992‒2002
(1) (2) (3) (4) (5) (6) (7) (8)
Lag1 Change in Shareholder Wealth 00014 00017 00088 00089 -00006 -00007 00072 00086
(051) (062) (129) (103) (027) (029) (131) (16)
Change in Shareholder Wealth -00102 00102 00105 00032 00102 00101
(097) (292) (293) (055) (611) (608)
Change Shareholder Wealth Lag1 Institutional
Concentration
00435 00764 00363 00253
(463) (258) (689) (144)
Lag1 Change Shareholder Wealth Lag1
Institutional Concentration
-00197 -00283
(122) (197)
Lag1 Institutional Concentration -492182 -288861 400149 453623
(09) (056) (098) (114)
Lag1 Total Institutional Ownership 720688 538372 655091 559587 682069 726226
(276) (232) (281) (314) (308) (322)
Lag1 Change Shareholder Wealth Lag1 Top 5
Institutional Ownership
-0046 -00689
(091) (256)
Lag1 Top 5 Institutional Ownership 583044 1191634
(092) (25)
Lag1 Tobinrsquos Q -32761 -36806 -19422 -18023 -54283 -48681 -29534 -24243
(138) (172) (077) (078) (227) (245) (136) (111)
Lag1 Market Capitalization 00004 00004 00002 00006 00007 00008 00007 00007
(042) (044) (022) (07) (086) (092) (082) (09)
CEO Dummy 768035 765474 768334 773634 616315 61729 619028 618535
(354) (355) (356) (355) (417) (418) (419) (418)
Adjusted R-Sq 00057 0006 00052 0005 00097 00097 00098 00102
Number of Observations 41305 41305 41305 41305 85518 85518 85518 85518
30
Table VI Level of Salary as a Function of Institutional Influence
Similar in parts to models (1) and (2) of HS (Table V) this table displays the coefficients from OLS regressions of salary against the lagged change and contemporary change in
shareholder wealth measures of institutional influence Tobinrsquos Q firm market capitalization and dummy controls for CEO industry and year effects We estimate models (1) to
(4) over the period 1992‒1997 using the actual data supplied by HS and models (5) to (8) over the whole sample 1992‒2002 Models (1) (2) (5) and (6) use Institutional
Concentration as a measure of institutional influence which HS define as the shares owned by the institutions with the five largest holdings divided by the shares held by all
institutions These models control for total institutional ownership Models (3) (4) (7) and (8) use Top 5 Ownership as a measure of institutional influence which we define as
the proportion of shares outstanding held by the institutions with the 5 largest holdings Coefficients for industry and year controls are not presented One two and three
asterisks denote significance at 10 5 and 1 levels
Hartzell and Starks Database 1992‒1997 Full Sample based on ExecuComp 1992‒2002
(1) (2) (3) (4) (5) (6) (7) (8)
Change in Shareholder Wealth 04125E-4 0003 -2439E-4 0003 00021 00012 00024 00012
(057) (464) (031) (468) (1004) (72) (1041) (706)
Lag1 Change in Shareholder Wealth -00062 7856E-4 -0007 9003E-4 -00014 00002 -00014 00003
(594) (098) (592) (112) (522) (123) (5) (145)
Lag1 Institutional Concentration -2475087 555134 -2734983 943317
(3245) (710) (4965) (1671)
Lag1 Total Institutional Ownership 1020109 217488 936337 252464
(1904) (432) (2471) (704)
Lag1 Top 5 Institutional Ownership -148355 31354 -87926 423166
(139) (332) (118) (643)
Lag1 Tobinrsquos Q -195348 -252191 -192344 -25158 -119837 -146373 -122132 -143821
(2812) (3762) (2721) (3752) (248) (274) (2583) (2727)
Lag1 Market Capitalization 844E-06 103E-5 00034 00039
(2781) (3051) (2874) (3104)
Lag1 Log Market Capitalization 750021 723625 776657 732279
(7511) (9285) (11224) (14039)
CEO Dummy 2683978 2702238 2668244 2702417 3281133 3285103 3292175 3283145
(8625) (9277) (8246) (9271) (12739) (13616) (12336) (13605)
Adjusted R-Sq 04604 05302 04187 05295 03856 04588 03476 04574
Number of Observations 36280 36280 36280 36280 97679 97679 97679 97679
31
Table VII Level of Total Direct Compensation as a Function of Institutional Influence
Similar in parts to models (3) and (5) of HS (Table V) this table displays the coefficients from OLS regressions of total direct compensation against the lagged change and
contemporary change in shareholder wealth measures of institutional influence Tobinrsquos Q firm market capitalization and dummy controls for CEO industry and year effects
We estimate models (1) to (4) over the period 1992‒1997 and models (5) to (8) over the whole sample Models (1) (2) (5) and (6) use Institutional Concentration as a measure
of institutional influence which HS define as the shares owned by the institutions with the five largest holdings divided by the shares held by all institutions These models
control for total institutional ownership Models (3) (4) (7) and (8) use Top 5 Ownership as a measure of institutional influence which we define as the proportion of shares
outstanding held by the institutions with the five largest holdings Coefficients for industry and year controls are not presented One two and three asterisks denote significance
at 10 5 and 1 levels
Hartzell and Starks Database 1992‒1997 Full Sample from ExecuComp 1992‒2002
(1) (2) (3) (4) (5) (6) (7) (8)
Change in Shareholder Wealth 005779 007006 005282 006791 00423 00148 0046 00138
(415) (581) (369) (565) (203) (07) (22) (066)
Lag1 Change in Shareholder Wealth 003182 00714 00272 00715 00613 0126 00614 01278
(126) (298) (107) (299) (275) (488) (276) (494)
Lag1 Institutional Concentration -1608633 3248998 -3424892 3334186
(151) (283) (1172) (118)
Lag1 Total Institutional Ownership 104013 5473128 1531013 -2926764
(1645) (831) (876) (158)
Lag1 Top 5 Institutional Ownership 188585 3927986 -2159177 -2640879
(016) (343) (067) (092)
Lag1 Tobinrsquos Q -445692 -803612 -454859 -833566 3257027 2971201 3213936 3086205
(427) (791) (432) (831) (83) (77) (815) (793)
Lag1 Market Capitalization 49E-5 615E-5 01031 01088
(113) (1365) (1555) (168)
Lag1 Log Market Capitalization 4683497 4759377 1645156 1427856
(2775) (3469) (3679) (4239)
CEO Dummy 1501307 1508886 1493688 1508736 4713079 4713629 47278 4703908
(2648) (2684) (2618) (2683) (282) (2828) (2819) (2824)
Adjusted R-Sq 01861 02060 01695 02051 01071 01113 01033 011
Number of Observations 33229 33229 33229 33229 97679 97679 97679 97679
32
Table VIII Two Stage Least Squares (Reverse Causality) Tests - Executive Compensation Measures as a Function of Institutional Concentration
This table shows the coefficients from the Two-Stage Least Squares (2SLS) regressions of executive compensation measures against the lagged change
and contemporary change in shareholder wealth instrumented institutional concentration and total institutional ownership Tobinrsquos Q the natural
logarithm of firm market capitalization and dummy variable controls for CEO and year effects We instrument measures of institutional concentration
and total institutional ownership by orthogonalizing with respect to option grant pay-for-performance sensitivity salary and total compensation
Coefficients for industry and year controls are not presented First stage results are not presented One two and three asterisks denote significance at
10 5 and 1 levels
ExecuComp Sample 1992‒1997 Full Sample from ExecuComp 1992‒2002
PPSOG Salary TC PPSOG Salary TC
Change in Shareholder Wealth 144E-05 0004 02133 -409E-06 00011 00157
(308) (775) (596) (691) (638) (068)
Lag1 Change in Shareholder Wealth -240E-05 -00004 00866 118E-06 00002 01422
(131) (087) (346) (097) (116) (51)
Lag1 Institutional Concentration -00898 668681 3263143 02024 916656 2856417
(052) (801) (176) (163) (1545) (1004)
Lag1 Total Institutional Ownership 01021 -43997 2512587 02983 -104171 -1028347
(09) (084) (153) (4) (269) (487)
Lag1 Tobinrsquos Q 00754 -225606 119297 00655 -158657 2997205
(596) (1649) (053) (834) (2535) (727)
Lag1 Log Market Capitalization -04234 828713 7864046 -04002 825106 1685638
(2195) (8132) (2977) (4106) (11133) (373)
CEO Dummy 15848 3214117 2950274 19149 3205862 4529101
(2223) (9389) (2255) (3802) (1331) (2758)
Overall R-Sq -- 04981 01206 -- 04738 01178
Number of Observations 41305 41305 41305 85518 85518 85518
33
Table IX Two Stage Least Squares (Reverse Causality) Tests - Executive Compensation Measures as a Function of Top 5 Ownership
This table shows the coefficients from 2SLS regressions of executive compensation measures against the lagged change and contemporary change
in shareholder wealth instrumented Top 5 institutional ownership Tobinrsquos Q natural logarithm of firm market capitalization and dummy variable
controls for CEO and year effects We instrument Top 5 institutional ownership by orthogonalizing with respect to option grant pay-for-
performance sensitivity salary and total compensation Coefficients for industry and year controls are not presented First stage results are not
presented One two and three asterisks denote significance at 10 5 and 1 levels
ExecuComp sample 1992‒1997 Full Sample 1992‒2002
PPSOG Salary TC PPSOG Salary TC
Change in Shareholder Wealth 136E-05 00041 02128 -407E-06 00011 00146
(293) (79) (596) (692) (619) (063)
Lag1 Change in Shareholder Wealth -241E-05 -00003 00868 116E-06 00003 01439
(133) (07) (347) (096) (142) (514)
Lag1 Top 5 Institutional Ownership -03774 -187512 -3848531 00179 -104261 -126659
(163) (186) (166) (011) (149) (436)
Lag1 Tobinrsquos Q 00731 -223886 10074 00642 -155467 3115073
(573) (1631) (045) (819) (2509) (75)
Lag1 Log Market Capitalization -0416 775637 7696302 -04015 764632 1462536
(2345) (10084) (3766) (3688) (13708) (4072)
CEO Dummy 15854 3211991 2950158 19152 3203152 4519385
(2223) (9381) (2254) (3804) (133) (2752)
Overall R-Sq -- 04967 01202 -- 04722 01168
Number of Observations 41305 41305 41305 85518 85518 85518
34
Table X Specification Tests utilizing Box-Cox Transformation of Market Capitalization
This table presents results from a Box-Cox transformation of market capitalization In
each model we specify transformation of market capitalization only Panel A shows
results for the period 1992‒1997 and Panel B shows results for the entire period
1992‒2002 The first part of each panel shows the Lambda coefficient that applies to
the transformed market capitalization The second part of each panel shows log
likelihood scores with LR Chi2 Statistics in parentheses ‒ these test for the
appropriateness of a linear or log model specification
Panel A Our ExecuComp sample 1992‒1997
(1) (2) (3)
PPSOG Salary TC
Lambda -05143 01881 02876
(-1914) (2413) (1329)
Test H0 (Log Likelihood Score)
Lambda = -1 -1279748 -3178575 -4795326
(23866) (85952) (99688)
Lambda = 0 -1280152 -3138456 -4791159
(31934) (57147) (16339)
Lambda = 1 -1283157 -3163807 -4793506
(92039) (564156) (63292)
Panel B Full ExecuComp sample 1992‒2002
(1) (2) (4)
PPSOG Salary TC
Lambda -044 01853 04036
(-3015) (40) (3666)
Test H0 (Log Likelihood Score)
Lambda = 0 -2688334 -6476494 -10384490
(66599) (160356) (126236)
Lambda = 1 -2694886 -6538506 -10386795
(197644) (1400605) (172345)
35
Figure 2 Institutional Ownership Measures (LHA) vs Total
Compensation (RHA) by Market Capitalization
0
10
20
30
40
50
60
70
1 2 3 4 5 6 7 8 9 10
$00
$10
$20
$30
$40
$50
$60
$70
$80
$90
Millio
ns
InstitutionalConcentration (LHS)Top5 Ownership (LHS)
Institutional TotalOwnership (LHS)Total Compensation(RHS)
Page 15
15
potentially pick up a size effect as their main result but not necessarily a monitoring
effect Figure 2 shows that as firm size increases up to decile 8 at least the share of
institutional total ownership increases as does total compensation
ltltINSERT FIGURES 1 AND 2 gtgt
D Methodology
Following HS we use two estimation techniques to examine the effect of institutional
investor influence on executive compensation Due to the censored nature of option grant
pay-for-performance sensitivity this compensation measure most appropriately uses
Tobit model specification whereas an Ordinary Least Squares (OLS) model specification
is used for all other compensation measures We model compensation measures as a
function of contemporaneous and lagged change in shareholder wealth institutional
influence Tobinrsquos Q and market capitalization Controls for CEO industry and year
effects are also used Augmentation occurs to test the effect of institutional influence on
the change in executive cash and total compensation with an interaction term that is the
product of institutional influence and the change in shareholder wealth
The variables used to model executive compensation have the following definitions
The lagged and contemporary measure of change in shareholder wealth is simply the
difference in market capitalization of the firm The measures of institutional influence
used are Institutional Concentration and Top 5 Ownership Use of the two types of
institutional influence occurs in separate models In order to replicate the HS
methodology models using Institutional Concentration have an additional control
consisting of lagged total institutional holdings and Tobinrsquos Q controls for performance
and growth opportunities Market capitalization of the firm is the number of shares
outstanding multiplied by the share price with inclusion both in levels and natural
16
logarithmic form A dummy variable is included to identify whether the executive is the
CEO Dummy variables are also included to control for industry effects This variable
takes a value of 1 for the two-digit Standard Industrial Classification (SIC) in which the
firm operates A year fixed effects (dummy) variable is included taking a value of 1 if the
observation was from the given year
IV Executive Incentives as a Function of Institutional Investor Influence
This section presents the results for the effect of institutional influence on option grant
pay-for-performance sensitivity and changes in compensation
A Option Grant Pay-for-Performance Sensitivity
Table IV presents the results for the effect of institutional influence on option grant
pay-for-performance sensitivity
ltltINSERT TABLE IVgtgt
Model (1) is a direct replication of HS using the HS database as supplied to us by HS
with all the variables computed in the same way and in the same functional form as with
HS The coefficient and t-statistic on lagged institutional concentration interacted with the
change in shareholder wealth is 13994 and 767 that compares to 1358 and 842 in HS
We are unable to account for these minor discrepancies for values that should be
identical In model (2) the only change to the model is that we use the natural logarithm
of market capitalization instead of the level of market capitalization This single change
has a sizeable effect on the coefficient on institutional concentration The coefficient on
institutional concentration is now -00692 and is insignificant The fact that an order
preserving transformation to a control variable can change the results is of interest The
log-transformation to market capitalization also changes the coefficient on Top 5
Ownership in models (3) and (4) The coefficient on Top 5 Ownership is significant in
17
both model (3) and in model (4) but the better size control reduces the coefficient by over
25 percent The better size control does not affect Top 5 Ownership by as much as its
design ensures less sensitivity to what could be inadequate size controls
We also test the effect of institutional influence on option grant pay-for-performance
sensitivity over the period 1992‒2002 This sample is then approximately three times as
large as the sample used by HS Institutional concentration is highly positively significant
in model (5) similar to the results for the period 1992‒1997 When we use the log of
market capitalization in model (6) the coefficient on institutional concentration remains
positive and significant at the 10 percent level with a t-statistic of 166 Of importance
though is that the coefficient on institutional concentration is one-eighth of the coefficient
in model (5) Even for a measure of institutional influence which we believe is less
sensitive to firm size the use of log market capitalization reduces the coefficient on Top 5
Ownership by approximately 50 percent from model (7) to model (8)
B Pay-for-Performance Sensitivity due to Changes in Cash and Total
Compensation
For robustness HS examine the effect that institutional influence conditioned on firm
performance has on a changes in cash and total executive compensation using the general
method of Jensen and Murphy (1990) to compute pay-for-performance sensitivity
ltltINSERT TABLE V gtgt
Table V presents the effect of institutional influence on the change in cash
compensation using the HS size control measure throughout12
Model (1) replicates the
specification in HS for the period of 1992‒1997 and we produce a similar result ndash the
interaction of lagged institutional concentration and contemporaneous change in
12 This test uses our data for the two periods 1992‒1997 and 1992‒2002 We do not use the HS data for
cash compensation as data for bonus is missing in the data supplied to us
18
shareholder wealth positively and significantly relates to a change in executive cash
compensation In model (2) we include the contemporaneous change in shareholder
wealth and the lagged institutional concentration as stand-alone controls as indicated by
Greenersquos (2000) specification of equation (1) above and the interaction term remains
significant In model (3) we interact lagged institutional concentration with the lagged
change in shareholder wealth as in equation (2) above that tries to take account of what
participants could have observed at the time The interaction term is then negative but
insignificant altering the HS result Model (4) provides a similar result where our
preferred proxy Top 5 Ownership captures concentrated institutional influence In
models (5) to (8) we use our full sample In model (5) we find that the lagged
institutional concentration interaction with the contemporaneous change in shareholder
wealth is positive and significant in HSrsquos specification However including change in
contemporaneous shareholder wealth and lagged institutional concentration as separate
control variables as suggested by equation (1)rsquos specification removes the significance
from the interaction term in model (6) Moreover in model (7) when lagged institutional
concentration is interaction with the lagged change in shareholder wealth in accord with
equation (2) this interaction term is now negative and significant reversing the HS
finding We obtain a similar result in model (8) when we measure institutional
concentration influence using Top 5 Ownership13
V The Levels of Executive Fixed and Total Compensation as Functions of
Institutional Investor Influence
13 In unreported results we use the log of market capitalization as a size control These changes to the size
control do not dramatically change the findings from Table V The reason why this size issue is not as
important as in the findings for option grants pay-for-performance is that the change in cash compensation
is not so strongly influenced by firm size In unreported results available on request we also examine the
effect of institutional influence on changes in total compensation with similar results to the change in cash
compensation
19
Two measures of compensation are examined ndash salary and total compensation
ltltINSERT TABLE VI gtgt
Table VI examines the effect of institutional influence on the level of salary In model
(1) we use the HS sample and specification to obtain almost identical results The
coefficient on institutional concentration is (negative) $247509 that is almost identical to
(negative) $247967 in HS Both are highly significant In model (2) we make a single
change to the model ndash we use the natural logarithm of market capitalization This single
change reverses the sign of institutional concentration with the coefficient still being
significant at the 1 percent level In models (3) and (4) we measure institutional influence
using Top 5 Ownership In model (3) where we measure firm size using the level of
market capitalization Top 5 Ownership is insignificant When we use the natural
logarithm of market capitalization in model (4) Top 5 Ownership is significant and
positively relates to the salary level In models (5) to (8) we estimate the same models
using our full sample We find similar results for the full sample as for the period 1992ndash
1997
ltltINSERT TABLE VII gtgt
Table VII examines the effect of institutional influence on total compensation In
model (1) we replicate the HS methodology using the HS data to obtain a similar
significant negative relation between institutional concentration and total compensation of
(negative) $1608633 However after changing to a logarithmic specification of market
capitalization in model (2) institutional concentration increases total compensation by a
significant $32400014
Importantly when institutional influence is measured using Top 5
14 = (5688751 times 0101)
20
Ownership an insignificant relation exists with total compensation in model (3) using the
HS size control but it becomes both positive and significant in (4) with the logged size
control In models (5) and (6) where models are estimated using the full sample we
observe a similar reversal in sign as occurs in models (1) and (2) However now the
magnitude of the reversal is almost equal with a difference of about $67m that is very
large indeed In model (5) the coefficient on institutional concentration is (negative)
$342m and then in model (6) the coefficient is $33m Top 5 Ownership is insignificant
in both models (7) and (8) unlike the results for Salary
VI Robustness
Two issues are examined in is this Section ndash reverse causality and the appropriate
transformation of market capitalization A methodological issue inherent in examining the
effect of ownership-based measures of institutional influence on executive compensation
is endogeneity driven by reverse causality We utilize an instrumental variables approach
where orthogonalized institutional influence measures are regressed on each of the
compensation measures lagged option grant pay-for-performance sensitivity salary and
total compensation In addition market capitalization and a CEO dummy are included in
each regression We use the residual from each regression as an instrument for
institutional influence Such a process removes any preferences for compensation from
the institutional influence measures
ltltINSERT TABLES VIII AND IXgtgt
Table VIII presents the results where we instrument the HS measure lagged
institutional concentration with a compensation-orthogonalized equivalent The analysis
indicates that the instrumented institutional concentration measure is positively and
significantly related to the level of salary and total compensation and insignificantly
21
related to option grant pay-for-performance sensitivity (PPSOG) Table IX presents
results where we instrument Top 5 Ownership with a compensation-orthogonalized
equivalent The instrumented Top 5 Ownership measure insignificantly relates to option
grant pay-for-performance sensitivity and barely negatively relates to salary although it
negatively relates to total compensation suggesting reverse causality is driving a
significant portion of our previous results
ltltINSERT TABLE Xgtgt
Another issue examined in this section is the appropriate transformation of market
capitalization Aneuryn-Evans and Deaton (1980) provide an evaluation of methods
comparing linear and log-linear specifications We use the Box-Cox test to compare
logarithm versus linear specifications In applying the Box-Cox test we specify
transformation of only market capitalization ndash the dependent variable and all other
independent variables remain unchanged Table X presents the lambda coefficient for
each model and the log likelihood score In all models in Panel A and B a lambda of zero
receives a higher (closer to zero) log likelihood score suggesting the log specification is
appropriate
VII Conclusion
Hartzell and Starks (2003) appears to document an effective role for institutional
investors in increasing executive incentives and lowering compensation levels Our paper
re-addresses the issue of concentrated institutional monitoring from a perspective
informed by an additional decade of research since their paper appeared We cannot find
evidence that concentrated institutional investors really have any positive (or negative for
that matter) effect on executive incentive levels While superficially we find evidence
that they may raise executive pay levels this is most likely driven by an institutional
22
preference for highly remunerated executives in large companies and thus by reverse
causality Moreover our findings are not consistent with the rent-extraction hypothesis
suggesting that firms systematically overpay executives in the absence of effective
monitoring by outside parties such as institutional investors
23
References
Almazan Andres Jay C Hartzell and Laura T Starks 2005 ldquoActive Institutional
Shareholders and Costs of Monitoring Evidence from Executive Compensationrdquo
Financial Management Winter 5ndash34
Aneuryn-Evans Gwyn and Angus Deaton 1980 ldquoTesting Linear versus Logarithmic
Regression Modelsrdquo Review of Economic Studies 47 (1) Econometrics Issue 275ndash
291
Babenko Ilona 2009 ldquoShare Repurchases and Pay-Performance Sensitivity of Employee
Compensation Contractsrdquo Journal of Finance 64117-150
Baker George P and Brian J Hall 2004 ldquoCEO Incentives and Firm Sizerdquo Journal of
Labor Market Economics 22 767ndash798
Bebchuk Lucian A Jesse M Fried and David Walker 2002 ldquoManagerial power and
rent extraction in the design of executive compensationrdquo University of Chicago Law
Review 69 751ndash846
Bebchuk Lucian A and Jesse M Fried 2003 ldquoExecutive Compensation as an Agency
Problemrdquo Journal of Economic Perspectives 17 71ndash92
Bebchuk Lucian A and Jesse M Fried 2004 Pay without Performance The Unfulfilled
Promise of Executive Compensation Cambridge MA Harvard University Press
Bertrand Marianne and Sendhil Mullainathan 2001 ldquoAre CEOs Rewarded for Luck
The Ones Without Principals Arerdquo Quarterly Journal of Economics 116 901ndash32
Black Fischer and Myron Scholes 1973 ldquoThe Pricing of Options and Corporate
Liabilitiesrdquo Journal of Political Economy 81 637ndash54
Cadman Brian Sandy Klasa and Steve Matsunaga 2010 ldquoDeterminants of CEO Pay A
Comparison of ExecuComp and Non-ExecuComp Firmsrdquo The Accounting Review 85
1511ndash1543
Chhaochharia Vidhi and Yaniv Grinstein 2009 ldquoCEO Compensation and Board
Structurerdquo Journal of Finance 64 231ndash261
Cichello Michael S 2005 ldquoThe impact of firm size on pay-performance sensitivitiesrdquo
Journal of Corporate Finance 11 609ndash627
Davis Gerald F and E Han Kim 2007 ldquoBusiness ties and proxy voting by mutual
fundsrdquo Journal of Financial Economics 85 552ndash570
Dikolli Shane S Susan L Kulp and Karen L Sedatole 2009 ldquoTransient Institutional
Ownership and CEO Contractingrdquo Accounting Review 84 737ndash770
Edmans Alex Xavier Gabaix and Augustin Landier 2009 ldquoA Multiplicative Model of
Optimal CEO Incentives in Market Equilibriumrdquo Review of Financial Studies 22
4881ndash4917
Edmans Alex 2009 ldquoBlockholder Trading Market Efficiency and Managerial Myopiardquo
Journal of Finance 64 2481ndash2513
Ertimur Yonca Fabrizio Ferri and Volkan Muslu 2011 ldquoShareholder Activism and
CEO Payrdquo Review of Financial Studies 24 535ndash592
24
Gabaix Xavier and Augustin Landier 2008 ldquoWhy Has CEO Pay Increased So Muchrdquo
Quarterly Journal of Economics 123 49ndash100
Garen John 1994 ldquoExecutive compensation and principal-agent theoryrdquo Journal of
Political Economy 102 1175ndash1199
Gibbons Robert and Kevin J Murphy 1992 ldquoOptimal Incentive Contracts in the
Presence of Career Concerns Theory and Evidencerdquo Journal of Political Economy
100 468ndash505
Greene William H 2000 (4th
Edition) Econometric Analysis Prentice-Hall NJ
Guthrie Katherine Jan Sokolowsky and Kam-Ming Wan 2012 ldquoCEO Compensation
and Board Structure Revisitedrdquo Journal of Finance Forthcoming
Hadlock Charles R and Gerald B Lumer 1997 ldquoCompensation turnover and top
management incentives historical evidencerdquo Journal of Business 70 153ndash187
Hartzell Jay C and Laura T Starks 2003 ldquoInstitutional Investors and Executive
Compensationrdquo Journal of Finance 58 2351ndash2374
Huson Mark R Robert Parrino and Laura T Starks 2001 ldquoInternal Monitoring
Mechanisms and CEO Turnover A Long-Term Perspectiverdquo Journal of Finance 56
2265ndash2297
Jensen Michael C and Kevin J Murphy 1990 ldquoPerformance Pay and Top Management
Incentivesrdquo Journal of Political Economy 98 225ndash264
Jin Li 2002 ldquoCEO compensation diversification and incentivesrdquo Journal of Financial
Economics 66 29ndash63
Kang Qiang and Qiao Liu 2008 ldquoStock Market Information Production and Executive
Incentivesrdquo Journal of Corporate Finance 14 484ndash498
Merton Robert 1973 ldquoTheory of Rational Option Pricingrdquo Bell Journal of Economics
and Management Science 4 141-183
Morse Adair Vikram Nanda and Amit Seru 2011 ldquoAre Incentive Contracts Rigged by
Powerful CEOsrdquo Journal of Finance 66 1779ndash1821
Murphy Kevin J 1985 ldquoCorporate performance and managerial remuneration An
empirical analysisrdquo Journal of Accounting and Economics 7 11ndash42
Murphy Kevin J 1999 ldquoExecutive compensationrdquo in Orley Ashenfelter and David
Card Handbook of Labor Economics vol 3 Elsevier
Rosen Sherwin 1992 ldquoContracts and the market for executivesrdquo In Contract
Economics ed Lars Werin and Hans Wijkander 181ndash211 Cambridge MA
Blackwell
Rothberg Burton and Steven Lilien 2006 ldquoMutual Funds and Proxy Voting New
Evidence on Corporate Governancerdquo Journal of Business and Technology Law 1
157ndash184
Schaefer Scott 1998 ldquoThe dependence of payndashperformance sensitivity on the size of the
firmrdquo Review of Economics and Statistics 80 436ndash443
Yermack David 1995 ldquoDo Corporations Award CEO Stock Options Effectivelyrdquo
Journal of Financial Economics 39 237ndash69
25
Table I Univariate Descriptive Statistics
This table reports sample statistics for the key variables used in this study based on our comprehensive
ExecuComp dataset Institutional Concentration is the shares owned by the institutions with the five largest
holdings divided by the shares held by all institutions Top 5 Ownership is the proportion of shares
outstanding held by the institutions with the five largest holdings Total Institutional Ownership is the
proportion of shares held by all institutions Panel B shows the measures executive compensation The pay-
for-performance sensitivity (PPS) of option grants shows the change in executive wealth due to a $1000
change in firm in shareholder wealth Expressions for Salary and Total Compensation are in thousands of
dollars while Market Capitalization is in millions of dollars Panel C shows firm characteristics
Variable Mean Median Std Dev P10 P90
Panel A Institutional Investor Holdings
Institutional Concentration 0450 0424 0151 0282 0656
Top 5 Institutional Ownership 0229 0220 0101 0110 0355
Total Institutional Ownership 0546 0561 0206 0258 0801
Panel B Executive Compensation
PPS of Option Grants 1161 0156 3829 0 2822
Salary 36311 29462 25231 15179 67000
Total Compensation 258313 85878 980701 23875 511022
Panel C Firm Characteristics
Tobins Q 1744 1139 2738 0495 3322
Market Capitalization 560699 112094 1988108 19754 1048754
26
Table II Pairwise Correlation Coefficients
This table presents the correlations between the key variables based on our comprehensive
ExecuComp dataset Institutional Concentration is the shares owned by the institutions with the five
largest holdings divided by the shares held by all institutions Top 5 Ownership is the proportion of
shares outstanding held by the institutions with the five largest holdings Total Institutional
Ownership is the proportion of shares held by all institutions The pay-for-performance sensitivity
(PPS) of option grants is the change in executive wealth due to a $1000 change in shareholder
wealth
(1) (2) (3) (4) (5) (6) (7) (8)
(1) Institutional Concentration 1
(2) Top 5 Ownership 0268 1
(3) Total Ownership -0415 0706 1
(4) PPS of Option Grants 0122 0042 -0024 1
(5) Log Salary -0270 -0012 0171 0038 1
(6) Log Market Capitalization -0630 -0157 0261 -0156 0420 1
(7) Level Market Capitalization -0241 -0157 0007 -0058 0219 0534 1
(8) Tobins Q -0008 -0034 0002 0015 -0083 0066 0121 1
27
Table III
Descriptive Statistics by Market Capitalization Deciles
Market
Capitalization
Deciles
Market
Capitalization
($m)
Total
Compensation
($000)
Salary
($)
PPS Option
Grants
($)
Institutional
Concentration
()
Top 5 Ownership
()
Institutional
Total Ownership
()
1 (Small) 1181 76008 234700 2720 611 200 327
2 2668 90206 257955 1796 495 217 438
3 4193 105946 275120 1517 454 220 485
4 6045 129892 296067 1376 420 206 491
5 8742 155391 315864 1216 394 198 502
6 13086 191888 345942 1072 380 202 531
7 20790 239501 377021 0874 368 196 532
8 34920 314955 427808 0687 354 190 538
9 67747 450970 480319 0569 328 173 528
10 (Big) 366540 846066 614354 0297 275 138 502
28
Table IV Pay-for-Performance Sensitivity of Option Grants as a Function of Institutional Influence
Similar in construction to parts of HS (Table II) this table shows the coefficients from Tobit regressions of option grant pay-for-performance sensitivity per $1000 change in
shareholder wealth against the lagged change and contemporary change in shareholder wealth measures of institutional influence Tobinrsquos Q firm market capitalization and
dummy controls for CEO and year effects We estimate models (1) to (4) over the period 1992‒1997 using data supplied by HS and models (5) to (8) are estimated over the
whole sample Models (1) (2) (5) and (6) use Institutional Concentration as a measure of institutional influence which HS define as the shares owned by the institutions
with the five largest holdings divided by the shares held by all institutions These models control for total institutional ownership Models (3) (4) (7) and (8) use Top 5
Ownership as a measure of institutional influence which we define as the proportion of shares outstanding held by the institutions with the five largest holdings Coefficients
for industry and year controls are not presented One two and three asterisks denote significance at 10 5 and 1 levels
Hartzell and Starks Database 1992‒1997 Full Sample from ExecuComp 1992‒2002
(1) (2) (3) (4) (5) (6) (7) (8)
Change in Shareholder Wealth 77721 191781 89874 137368 -446 -218 -664 -203
(171) (429) (194) (317) (319) (168) (453) (158)
Lag1 Change in Shareholder Wealth 14446 99963 88719 8544 -144 0273 -0342 00852
(027) (212) (166) (184) (071) (015) (016) (005)
Lag1 Institutional Concentration 13994 -00692 26405 03304
(767) (032) (1402) (166)
Lag1 Total Institutional Ownership 08055 10434 07106 0937
(593) (781) (666) (898)
Lag1 Top 5 Institutional Ownership 10567 07457 12116 06293
(399) ( 283) (533) (278)
Lag1 Tobinrsquos Q 0068 00877 00647 00813 00498 00727 00508 00695
(337) (434) (319) (401) (504) (734) (51) (703)
Lag1 Market Capitalization -106202 -173515 -776 -117
(474) (742) (1011) (153)
Lag1 Log Market Capitalization -0297 -02319 -04111 -03908
(1576) (1439) (2997) (2841)
CEO Dummy 13745 13682 13765 13674 28197 28128 28155 28143
(1913) (1918) (1914) (1916) (3388) (3396) (3377) (34)
Number of Observations 33878 33878 33878 33878 97679 97679 97679 97679
29
Table V Pay-for-Performance Sensitivity of Cash Compensation (Salary plus Bonus) as a Function of Institutional Influence - Level Market Capitalization
Similar in parts to models (1) and (2) of HS (Table IV) this table displays the coefficients from OLS regressions of the change in cash compensation against the lagged
change and contemporary change in shareholder wealth measures of institutional influence Tobinrsquos Q firm market capitalization and dummy controls for CEO and year
effects We estimate models (1) to (4) over the period 1992‒1997 based on our ExecuComp sample and models (5) to (8) over the whole sample Models (1) (2) (5) and (6)
use Institutional Concentration as a measure of institutional influence which HS define as the shares owned by the institutions with the five largest holdings divided by the
shares held by all institutions These models control for total institutional ownership Models (4) and (8) use Top 5 Ownership as a measure of institutional influence which
we define as the proportion of shares outstanding held by the institutions with the 5 largest holdings Coefficients for industry and year controls are not presented One two
and three asterisks denote significance at 10 5 and 1 levels
Our Database based on ExecuComp 1992‒1997 Full Sample based on ExecuComp 1992‒2002
(1) (2) (3) (4) (5) (6) (7) (8)
Lag1 Change in Shareholder Wealth 00014 00017 00088 00089 -00006 -00007 00072 00086
(051) (062) (129) (103) (027) (029) (131) (16)
Change in Shareholder Wealth -00102 00102 00105 00032 00102 00101
(097) (292) (293) (055) (611) (608)
Change Shareholder Wealth Lag1 Institutional
Concentration
00435 00764 00363 00253
(463) (258) (689) (144)
Lag1 Change Shareholder Wealth Lag1
Institutional Concentration
-00197 -00283
(122) (197)
Lag1 Institutional Concentration -492182 -288861 400149 453623
(09) (056) (098) (114)
Lag1 Total Institutional Ownership 720688 538372 655091 559587 682069 726226
(276) (232) (281) (314) (308) (322)
Lag1 Change Shareholder Wealth Lag1 Top 5
Institutional Ownership
-0046 -00689
(091) (256)
Lag1 Top 5 Institutional Ownership 583044 1191634
(092) (25)
Lag1 Tobinrsquos Q -32761 -36806 -19422 -18023 -54283 -48681 -29534 -24243
(138) (172) (077) (078) (227) (245) (136) (111)
Lag1 Market Capitalization 00004 00004 00002 00006 00007 00008 00007 00007
(042) (044) (022) (07) (086) (092) (082) (09)
CEO Dummy 768035 765474 768334 773634 616315 61729 619028 618535
(354) (355) (356) (355) (417) (418) (419) (418)
Adjusted R-Sq 00057 0006 00052 0005 00097 00097 00098 00102
Number of Observations 41305 41305 41305 41305 85518 85518 85518 85518
30
Table VI Level of Salary as a Function of Institutional Influence
Similar in parts to models (1) and (2) of HS (Table V) this table displays the coefficients from OLS regressions of salary against the lagged change and contemporary change in
shareholder wealth measures of institutional influence Tobinrsquos Q firm market capitalization and dummy controls for CEO industry and year effects We estimate models (1) to
(4) over the period 1992‒1997 using the actual data supplied by HS and models (5) to (8) over the whole sample 1992‒2002 Models (1) (2) (5) and (6) use Institutional
Concentration as a measure of institutional influence which HS define as the shares owned by the institutions with the five largest holdings divided by the shares held by all
institutions These models control for total institutional ownership Models (3) (4) (7) and (8) use Top 5 Ownership as a measure of institutional influence which we define as
the proportion of shares outstanding held by the institutions with the 5 largest holdings Coefficients for industry and year controls are not presented One two and three
asterisks denote significance at 10 5 and 1 levels
Hartzell and Starks Database 1992‒1997 Full Sample based on ExecuComp 1992‒2002
(1) (2) (3) (4) (5) (6) (7) (8)
Change in Shareholder Wealth 04125E-4 0003 -2439E-4 0003 00021 00012 00024 00012
(057) (464) (031) (468) (1004) (72) (1041) (706)
Lag1 Change in Shareholder Wealth -00062 7856E-4 -0007 9003E-4 -00014 00002 -00014 00003
(594) (098) (592) (112) (522) (123) (5) (145)
Lag1 Institutional Concentration -2475087 555134 -2734983 943317
(3245) (710) (4965) (1671)
Lag1 Total Institutional Ownership 1020109 217488 936337 252464
(1904) (432) (2471) (704)
Lag1 Top 5 Institutional Ownership -148355 31354 -87926 423166
(139) (332) (118) (643)
Lag1 Tobinrsquos Q -195348 -252191 -192344 -25158 -119837 -146373 -122132 -143821
(2812) (3762) (2721) (3752) (248) (274) (2583) (2727)
Lag1 Market Capitalization 844E-06 103E-5 00034 00039
(2781) (3051) (2874) (3104)
Lag1 Log Market Capitalization 750021 723625 776657 732279
(7511) (9285) (11224) (14039)
CEO Dummy 2683978 2702238 2668244 2702417 3281133 3285103 3292175 3283145
(8625) (9277) (8246) (9271) (12739) (13616) (12336) (13605)
Adjusted R-Sq 04604 05302 04187 05295 03856 04588 03476 04574
Number of Observations 36280 36280 36280 36280 97679 97679 97679 97679
31
Table VII Level of Total Direct Compensation as a Function of Institutional Influence
Similar in parts to models (3) and (5) of HS (Table V) this table displays the coefficients from OLS regressions of total direct compensation against the lagged change and
contemporary change in shareholder wealth measures of institutional influence Tobinrsquos Q firm market capitalization and dummy controls for CEO industry and year effects
We estimate models (1) to (4) over the period 1992‒1997 and models (5) to (8) over the whole sample Models (1) (2) (5) and (6) use Institutional Concentration as a measure
of institutional influence which HS define as the shares owned by the institutions with the five largest holdings divided by the shares held by all institutions These models
control for total institutional ownership Models (3) (4) (7) and (8) use Top 5 Ownership as a measure of institutional influence which we define as the proportion of shares
outstanding held by the institutions with the five largest holdings Coefficients for industry and year controls are not presented One two and three asterisks denote significance
at 10 5 and 1 levels
Hartzell and Starks Database 1992‒1997 Full Sample from ExecuComp 1992‒2002
(1) (2) (3) (4) (5) (6) (7) (8)
Change in Shareholder Wealth 005779 007006 005282 006791 00423 00148 0046 00138
(415) (581) (369) (565) (203) (07) (22) (066)
Lag1 Change in Shareholder Wealth 003182 00714 00272 00715 00613 0126 00614 01278
(126) (298) (107) (299) (275) (488) (276) (494)
Lag1 Institutional Concentration -1608633 3248998 -3424892 3334186
(151) (283) (1172) (118)
Lag1 Total Institutional Ownership 104013 5473128 1531013 -2926764
(1645) (831) (876) (158)
Lag1 Top 5 Institutional Ownership 188585 3927986 -2159177 -2640879
(016) (343) (067) (092)
Lag1 Tobinrsquos Q -445692 -803612 -454859 -833566 3257027 2971201 3213936 3086205
(427) (791) (432) (831) (83) (77) (815) (793)
Lag1 Market Capitalization 49E-5 615E-5 01031 01088
(113) (1365) (1555) (168)
Lag1 Log Market Capitalization 4683497 4759377 1645156 1427856
(2775) (3469) (3679) (4239)
CEO Dummy 1501307 1508886 1493688 1508736 4713079 4713629 47278 4703908
(2648) (2684) (2618) (2683) (282) (2828) (2819) (2824)
Adjusted R-Sq 01861 02060 01695 02051 01071 01113 01033 011
Number of Observations 33229 33229 33229 33229 97679 97679 97679 97679
32
Table VIII Two Stage Least Squares (Reverse Causality) Tests - Executive Compensation Measures as a Function of Institutional Concentration
This table shows the coefficients from the Two-Stage Least Squares (2SLS) regressions of executive compensation measures against the lagged change
and contemporary change in shareholder wealth instrumented institutional concentration and total institutional ownership Tobinrsquos Q the natural
logarithm of firm market capitalization and dummy variable controls for CEO and year effects We instrument measures of institutional concentration
and total institutional ownership by orthogonalizing with respect to option grant pay-for-performance sensitivity salary and total compensation
Coefficients for industry and year controls are not presented First stage results are not presented One two and three asterisks denote significance at
10 5 and 1 levels
ExecuComp Sample 1992‒1997 Full Sample from ExecuComp 1992‒2002
PPSOG Salary TC PPSOG Salary TC
Change in Shareholder Wealth 144E-05 0004 02133 -409E-06 00011 00157
(308) (775) (596) (691) (638) (068)
Lag1 Change in Shareholder Wealth -240E-05 -00004 00866 118E-06 00002 01422
(131) (087) (346) (097) (116) (51)
Lag1 Institutional Concentration -00898 668681 3263143 02024 916656 2856417
(052) (801) (176) (163) (1545) (1004)
Lag1 Total Institutional Ownership 01021 -43997 2512587 02983 -104171 -1028347
(09) (084) (153) (4) (269) (487)
Lag1 Tobinrsquos Q 00754 -225606 119297 00655 -158657 2997205
(596) (1649) (053) (834) (2535) (727)
Lag1 Log Market Capitalization -04234 828713 7864046 -04002 825106 1685638
(2195) (8132) (2977) (4106) (11133) (373)
CEO Dummy 15848 3214117 2950274 19149 3205862 4529101
(2223) (9389) (2255) (3802) (1331) (2758)
Overall R-Sq -- 04981 01206 -- 04738 01178
Number of Observations 41305 41305 41305 85518 85518 85518
33
Table IX Two Stage Least Squares (Reverse Causality) Tests - Executive Compensation Measures as a Function of Top 5 Ownership
This table shows the coefficients from 2SLS regressions of executive compensation measures against the lagged change and contemporary change
in shareholder wealth instrumented Top 5 institutional ownership Tobinrsquos Q natural logarithm of firm market capitalization and dummy variable
controls for CEO and year effects We instrument Top 5 institutional ownership by orthogonalizing with respect to option grant pay-for-
performance sensitivity salary and total compensation Coefficients for industry and year controls are not presented First stage results are not
presented One two and three asterisks denote significance at 10 5 and 1 levels
ExecuComp sample 1992‒1997 Full Sample 1992‒2002
PPSOG Salary TC PPSOG Salary TC
Change in Shareholder Wealth 136E-05 00041 02128 -407E-06 00011 00146
(293) (79) (596) (692) (619) (063)
Lag1 Change in Shareholder Wealth -241E-05 -00003 00868 116E-06 00003 01439
(133) (07) (347) (096) (142) (514)
Lag1 Top 5 Institutional Ownership -03774 -187512 -3848531 00179 -104261 -126659
(163) (186) (166) (011) (149) (436)
Lag1 Tobinrsquos Q 00731 -223886 10074 00642 -155467 3115073
(573) (1631) (045) (819) (2509) (75)
Lag1 Log Market Capitalization -0416 775637 7696302 -04015 764632 1462536
(2345) (10084) (3766) (3688) (13708) (4072)
CEO Dummy 15854 3211991 2950158 19152 3203152 4519385
(2223) (9381) (2254) (3804) (133) (2752)
Overall R-Sq -- 04967 01202 -- 04722 01168
Number of Observations 41305 41305 41305 85518 85518 85518
34
Table X Specification Tests utilizing Box-Cox Transformation of Market Capitalization
This table presents results from a Box-Cox transformation of market capitalization In
each model we specify transformation of market capitalization only Panel A shows
results for the period 1992‒1997 and Panel B shows results for the entire period
1992‒2002 The first part of each panel shows the Lambda coefficient that applies to
the transformed market capitalization The second part of each panel shows log
likelihood scores with LR Chi2 Statistics in parentheses ‒ these test for the
appropriateness of a linear or log model specification
Panel A Our ExecuComp sample 1992‒1997
(1) (2) (3)
PPSOG Salary TC
Lambda -05143 01881 02876
(-1914) (2413) (1329)
Test H0 (Log Likelihood Score)
Lambda = -1 -1279748 -3178575 -4795326
(23866) (85952) (99688)
Lambda = 0 -1280152 -3138456 -4791159
(31934) (57147) (16339)
Lambda = 1 -1283157 -3163807 -4793506
(92039) (564156) (63292)
Panel B Full ExecuComp sample 1992‒2002
(1) (2) (4)
PPSOG Salary TC
Lambda -044 01853 04036
(-3015) (40) (3666)
Test H0 (Log Likelihood Score)
Lambda = 0 -2688334 -6476494 -10384490
(66599) (160356) (126236)
Lambda = 1 -2694886 -6538506 -10386795
(197644) (1400605) (172345)
35
Figure 2 Institutional Ownership Measures (LHA) vs Total
Compensation (RHA) by Market Capitalization
0
10
20
30
40
50
60
70
1 2 3 4 5 6 7 8 9 10
$00
$10
$20
$30
$40
$50
$60
$70
$80
$90
Millio
ns
InstitutionalConcentration (LHS)Top5 Ownership (LHS)
Institutional TotalOwnership (LHS)Total Compensation(RHS)
Page 16
16
logarithmic form A dummy variable is included to identify whether the executive is the
CEO Dummy variables are also included to control for industry effects This variable
takes a value of 1 for the two-digit Standard Industrial Classification (SIC) in which the
firm operates A year fixed effects (dummy) variable is included taking a value of 1 if the
observation was from the given year
IV Executive Incentives as a Function of Institutional Investor Influence
This section presents the results for the effect of institutional influence on option grant
pay-for-performance sensitivity and changes in compensation
A Option Grant Pay-for-Performance Sensitivity
Table IV presents the results for the effect of institutional influence on option grant
pay-for-performance sensitivity
ltltINSERT TABLE IVgtgt
Model (1) is a direct replication of HS using the HS database as supplied to us by HS
with all the variables computed in the same way and in the same functional form as with
HS The coefficient and t-statistic on lagged institutional concentration interacted with the
change in shareholder wealth is 13994 and 767 that compares to 1358 and 842 in HS
We are unable to account for these minor discrepancies for values that should be
identical In model (2) the only change to the model is that we use the natural logarithm
of market capitalization instead of the level of market capitalization This single change
has a sizeable effect on the coefficient on institutional concentration The coefficient on
institutional concentration is now -00692 and is insignificant The fact that an order
preserving transformation to a control variable can change the results is of interest The
log-transformation to market capitalization also changes the coefficient on Top 5
Ownership in models (3) and (4) The coefficient on Top 5 Ownership is significant in
17
both model (3) and in model (4) but the better size control reduces the coefficient by over
25 percent The better size control does not affect Top 5 Ownership by as much as its
design ensures less sensitivity to what could be inadequate size controls
We also test the effect of institutional influence on option grant pay-for-performance
sensitivity over the period 1992‒2002 This sample is then approximately three times as
large as the sample used by HS Institutional concentration is highly positively significant
in model (5) similar to the results for the period 1992‒1997 When we use the log of
market capitalization in model (6) the coefficient on institutional concentration remains
positive and significant at the 10 percent level with a t-statistic of 166 Of importance
though is that the coefficient on institutional concentration is one-eighth of the coefficient
in model (5) Even for a measure of institutional influence which we believe is less
sensitive to firm size the use of log market capitalization reduces the coefficient on Top 5
Ownership by approximately 50 percent from model (7) to model (8)
B Pay-for-Performance Sensitivity due to Changes in Cash and Total
Compensation
For robustness HS examine the effect that institutional influence conditioned on firm
performance has on a changes in cash and total executive compensation using the general
method of Jensen and Murphy (1990) to compute pay-for-performance sensitivity
ltltINSERT TABLE V gtgt
Table V presents the effect of institutional influence on the change in cash
compensation using the HS size control measure throughout12
Model (1) replicates the
specification in HS for the period of 1992‒1997 and we produce a similar result ndash the
interaction of lagged institutional concentration and contemporaneous change in
12 This test uses our data for the two periods 1992‒1997 and 1992‒2002 We do not use the HS data for
cash compensation as data for bonus is missing in the data supplied to us
18
shareholder wealth positively and significantly relates to a change in executive cash
compensation In model (2) we include the contemporaneous change in shareholder
wealth and the lagged institutional concentration as stand-alone controls as indicated by
Greenersquos (2000) specification of equation (1) above and the interaction term remains
significant In model (3) we interact lagged institutional concentration with the lagged
change in shareholder wealth as in equation (2) above that tries to take account of what
participants could have observed at the time The interaction term is then negative but
insignificant altering the HS result Model (4) provides a similar result where our
preferred proxy Top 5 Ownership captures concentrated institutional influence In
models (5) to (8) we use our full sample In model (5) we find that the lagged
institutional concentration interaction with the contemporaneous change in shareholder
wealth is positive and significant in HSrsquos specification However including change in
contemporaneous shareholder wealth and lagged institutional concentration as separate
control variables as suggested by equation (1)rsquos specification removes the significance
from the interaction term in model (6) Moreover in model (7) when lagged institutional
concentration is interaction with the lagged change in shareholder wealth in accord with
equation (2) this interaction term is now negative and significant reversing the HS
finding We obtain a similar result in model (8) when we measure institutional
concentration influence using Top 5 Ownership13
V The Levels of Executive Fixed and Total Compensation as Functions of
Institutional Investor Influence
13 In unreported results we use the log of market capitalization as a size control These changes to the size
control do not dramatically change the findings from Table V The reason why this size issue is not as
important as in the findings for option grants pay-for-performance is that the change in cash compensation
is not so strongly influenced by firm size In unreported results available on request we also examine the
effect of institutional influence on changes in total compensation with similar results to the change in cash
compensation
19
Two measures of compensation are examined ndash salary and total compensation
ltltINSERT TABLE VI gtgt
Table VI examines the effect of institutional influence on the level of salary In model
(1) we use the HS sample and specification to obtain almost identical results The
coefficient on institutional concentration is (negative) $247509 that is almost identical to
(negative) $247967 in HS Both are highly significant In model (2) we make a single
change to the model ndash we use the natural logarithm of market capitalization This single
change reverses the sign of institutional concentration with the coefficient still being
significant at the 1 percent level In models (3) and (4) we measure institutional influence
using Top 5 Ownership In model (3) where we measure firm size using the level of
market capitalization Top 5 Ownership is insignificant When we use the natural
logarithm of market capitalization in model (4) Top 5 Ownership is significant and
positively relates to the salary level In models (5) to (8) we estimate the same models
using our full sample We find similar results for the full sample as for the period 1992ndash
1997
ltltINSERT TABLE VII gtgt
Table VII examines the effect of institutional influence on total compensation In
model (1) we replicate the HS methodology using the HS data to obtain a similar
significant negative relation between institutional concentration and total compensation of
(negative) $1608633 However after changing to a logarithmic specification of market
capitalization in model (2) institutional concentration increases total compensation by a
significant $32400014
Importantly when institutional influence is measured using Top 5
14 = (5688751 times 0101)
20
Ownership an insignificant relation exists with total compensation in model (3) using the
HS size control but it becomes both positive and significant in (4) with the logged size
control In models (5) and (6) where models are estimated using the full sample we
observe a similar reversal in sign as occurs in models (1) and (2) However now the
magnitude of the reversal is almost equal with a difference of about $67m that is very
large indeed In model (5) the coefficient on institutional concentration is (negative)
$342m and then in model (6) the coefficient is $33m Top 5 Ownership is insignificant
in both models (7) and (8) unlike the results for Salary
VI Robustness
Two issues are examined in is this Section ndash reverse causality and the appropriate
transformation of market capitalization A methodological issue inherent in examining the
effect of ownership-based measures of institutional influence on executive compensation
is endogeneity driven by reverse causality We utilize an instrumental variables approach
where orthogonalized institutional influence measures are regressed on each of the
compensation measures lagged option grant pay-for-performance sensitivity salary and
total compensation In addition market capitalization and a CEO dummy are included in
each regression We use the residual from each regression as an instrument for
institutional influence Such a process removes any preferences for compensation from
the institutional influence measures
ltltINSERT TABLES VIII AND IXgtgt
Table VIII presents the results where we instrument the HS measure lagged
institutional concentration with a compensation-orthogonalized equivalent The analysis
indicates that the instrumented institutional concentration measure is positively and
significantly related to the level of salary and total compensation and insignificantly
21
related to option grant pay-for-performance sensitivity (PPSOG) Table IX presents
results where we instrument Top 5 Ownership with a compensation-orthogonalized
equivalent The instrumented Top 5 Ownership measure insignificantly relates to option
grant pay-for-performance sensitivity and barely negatively relates to salary although it
negatively relates to total compensation suggesting reverse causality is driving a
significant portion of our previous results
ltltINSERT TABLE Xgtgt
Another issue examined in this section is the appropriate transformation of market
capitalization Aneuryn-Evans and Deaton (1980) provide an evaluation of methods
comparing linear and log-linear specifications We use the Box-Cox test to compare
logarithm versus linear specifications In applying the Box-Cox test we specify
transformation of only market capitalization ndash the dependent variable and all other
independent variables remain unchanged Table X presents the lambda coefficient for
each model and the log likelihood score In all models in Panel A and B a lambda of zero
receives a higher (closer to zero) log likelihood score suggesting the log specification is
appropriate
VII Conclusion
Hartzell and Starks (2003) appears to document an effective role for institutional
investors in increasing executive incentives and lowering compensation levels Our paper
re-addresses the issue of concentrated institutional monitoring from a perspective
informed by an additional decade of research since their paper appeared We cannot find
evidence that concentrated institutional investors really have any positive (or negative for
that matter) effect on executive incentive levels While superficially we find evidence
that they may raise executive pay levels this is most likely driven by an institutional
22
preference for highly remunerated executives in large companies and thus by reverse
causality Moreover our findings are not consistent with the rent-extraction hypothesis
suggesting that firms systematically overpay executives in the absence of effective
monitoring by outside parties such as institutional investors
23
References
Almazan Andres Jay C Hartzell and Laura T Starks 2005 ldquoActive Institutional
Shareholders and Costs of Monitoring Evidence from Executive Compensationrdquo
Financial Management Winter 5ndash34
Aneuryn-Evans Gwyn and Angus Deaton 1980 ldquoTesting Linear versus Logarithmic
Regression Modelsrdquo Review of Economic Studies 47 (1) Econometrics Issue 275ndash
291
Babenko Ilona 2009 ldquoShare Repurchases and Pay-Performance Sensitivity of Employee
Compensation Contractsrdquo Journal of Finance 64117-150
Baker George P and Brian J Hall 2004 ldquoCEO Incentives and Firm Sizerdquo Journal of
Labor Market Economics 22 767ndash798
Bebchuk Lucian A Jesse M Fried and David Walker 2002 ldquoManagerial power and
rent extraction in the design of executive compensationrdquo University of Chicago Law
Review 69 751ndash846
Bebchuk Lucian A and Jesse M Fried 2003 ldquoExecutive Compensation as an Agency
Problemrdquo Journal of Economic Perspectives 17 71ndash92
Bebchuk Lucian A and Jesse M Fried 2004 Pay without Performance The Unfulfilled
Promise of Executive Compensation Cambridge MA Harvard University Press
Bertrand Marianne and Sendhil Mullainathan 2001 ldquoAre CEOs Rewarded for Luck
The Ones Without Principals Arerdquo Quarterly Journal of Economics 116 901ndash32
Black Fischer and Myron Scholes 1973 ldquoThe Pricing of Options and Corporate
Liabilitiesrdquo Journal of Political Economy 81 637ndash54
Cadman Brian Sandy Klasa and Steve Matsunaga 2010 ldquoDeterminants of CEO Pay A
Comparison of ExecuComp and Non-ExecuComp Firmsrdquo The Accounting Review 85
1511ndash1543
Chhaochharia Vidhi and Yaniv Grinstein 2009 ldquoCEO Compensation and Board
Structurerdquo Journal of Finance 64 231ndash261
Cichello Michael S 2005 ldquoThe impact of firm size on pay-performance sensitivitiesrdquo
Journal of Corporate Finance 11 609ndash627
Davis Gerald F and E Han Kim 2007 ldquoBusiness ties and proxy voting by mutual
fundsrdquo Journal of Financial Economics 85 552ndash570
Dikolli Shane S Susan L Kulp and Karen L Sedatole 2009 ldquoTransient Institutional
Ownership and CEO Contractingrdquo Accounting Review 84 737ndash770
Edmans Alex Xavier Gabaix and Augustin Landier 2009 ldquoA Multiplicative Model of
Optimal CEO Incentives in Market Equilibriumrdquo Review of Financial Studies 22
4881ndash4917
Edmans Alex 2009 ldquoBlockholder Trading Market Efficiency and Managerial Myopiardquo
Journal of Finance 64 2481ndash2513
Ertimur Yonca Fabrizio Ferri and Volkan Muslu 2011 ldquoShareholder Activism and
CEO Payrdquo Review of Financial Studies 24 535ndash592
24
Gabaix Xavier and Augustin Landier 2008 ldquoWhy Has CEO Pay Increased So Muchrdquo
Quarterly Journal of Economics 123 49ndash100
Garen John 1994 ldquoExecutive compensation and principal-agent theoryrdquo Journal of
Political Economy 102 1175ndash1199
Gibbons Robert and Kevin J Murphy 1992 ldquoOptimal Incentive Contracts in the
Presence of Career Concerns Theory and Evidencerdquo Journal of Political Economy
100 468ndash505
Greene William H 2000 (4th
Edition) Econometric Analysis Prentice-Hall NJ
Guthrie Katherine Jan Sokolowsky and Kam-Ming Wan 2012 ldquoCEO Compensation
and Board Structure Revisitedrdquo Journal of Finance Forthcoming
Hadlock Charles R and Gerald B Lumer 1997 ldquoCompensation turnover and top
management incentives historical evidencerdquo Journal of Business 70 153ndash187
Hartzell Jay C and Laura T Starks 2003 ldquoInstitutional Investors and Executive
Compensationrdquo Journal of Finance 58 2351ndash2374
Huson Mark R Robert Parrino and Laura T Starks 2001 ldquoInternal Monitoring
Mechanisms and CEO Turnover A Long-Term Perspectiverdquo Journal of Finance 56
2265ndash2297
Jensen Michael C and Kevin J Murphy 1990 ldquoPerformance Pay and Top Management
Incentivesrdquo Journal of Political Economy 98 225ndash264
Jin Li 2002 ldquoCEO compensation diversification and incentivesrdquo Journal of Financial
Economics 66 29ndash63
Kang Qiang and Qiao Liu 2008 ldquoStock Market Information Production and Executive
Incentivesrdquo Journal of Corporate Finance 14 484ndash498
Merton Robert 1973 ldquoTheory of Rational Option Pricingrdquo Bell Journal of Economics
and Management Science 4 141-183
Morse Adair Vikram Nanda and Amit Seru 2011 ldquoAre Incentive Contracts Rigged by
Powerful CEOsrdquo Journal of Finance 66 1779ndash1821
Murphy Kevin J 1985 ldquoCorporate performance and managerial remuneration An
empirical analysisrdquo Journal of Accounting and Economics 7 11ndash42
Murphy Kevin J 1999 ldquoExecutive compensationrdquo in Orley Ashenfelter and David
Card Handbook of Labor Economics vol 3 Elsevier
Rosen Sherwin 1992 ldquoContracts and the market for executivesrdquo In Contract
Economics ed Lars Werin and Hans Wijkander 181ndash211 Cambridge MA
Blackwell
Rothberg Burton and Steven Lilien 2006 ldquoMutual Funds and Proxy Voting New
Evidence on Corporate Governancerdquo Journal of Business and Technology Law 1
157ndash184
Schaefer Scott 1998 ldquoThe dependence of payndashperformance sensitivity on the size of the
firmrdquo Review of Economics and Statistics 80 436ndash443
Yermack David 1995 ldquoDo Corporations Award CEO Stock Options Effectivelyrdquo
Journal of Financial Economics 39 237ndash69
25
Table I Univariate Descriptive Statistics
This table reports sample statistics for the key variables used in this study based on our comprehensive
ExecuComp dataset Institutional Concentration is the shares owned by the institutions with the five largest
holdings divided by the shares held by all institutions Top 5 Ownership is the proportion of shares
outstanding held by the institutions with the five largest holdings Total Institutional Ownership is the
proportion of shares held by all institutions Panel B shows the measures executive compensation The pay-
for-performance sensitivity (PPS) of option grants shows the change in executive wealth due to a $1000
change in firm in shareholder wealth Expressions for Salary and Total Compensation are in thousands of
dollars while Market Capitalization is in millions of dollars Panel C shows firm characteristics
Variable Mean Median Std Dev P10 P90
Panel A Institutional Investor Holdings
Institutional Concentration 0450 0424 0151 0282 0656
Top 5 Institutional Ownership 0229 0220 0101 0110 0355
Total Institutional Ownership 0546 0561 0206 0258 0801
Panel B Executive Compensation
PPS of Option Grants 1161 0156 3829 0 2822
Salary 36311 29462 25231 15179 67000
Total Compensation 258313 85878 980701 23875 511022
Panel C Firm Characteristics
Tobins Q 1744 1139 2738 0495 3322
Market Capitalization 560699 112094 1988108 19754 1048754
26
Table II Pairwise Correlation Coefficients
This table presents the correlations between the key variables based on our comprehensive
ExecuComp dataset Institutional Concentration is the shares owned by the institutions with the five
largest holdings divided by the shares held by all institutions Top 5 Ownership is the proportion of
shares outstanding held by the institutions with the five largest holdings Total Institutional
Ownership is the proportion of shares held by all institutions The pay-for-performance sensitivity
(PPS) of option grants is the change in executive wealth due to a $1000 change in shareholder
wealth
(1) (2) (3) (4) (5) (6) (7) (8)
(1) Institutional Concentration 1
(2) Top 5 Ownership 0268 1
(3) Total Ownership -0415 0706 1
(4) PPS of Option Grants 0122 0042 -0024 1
(5) Log Salary -0270 -0012 0171 0038 1
(6) Log Market Capitalization -0630 -0157 0261 -0156 0420 1
(7) Level Market Capitalization -0241 -0157 0007 -0058 0219 0534 1
(8) Tobins Q -0008 -0034 0002 0015 -0083 0066 0121 1
27
Table III
Descriptive Statistics by Market Capitalization Deciles
Market
Capitalization
Deciles
Market
Capitalization
($m)
Total
Compensation
($000)
Salary
($)
PPS Option
Grants
($)
Institutional
Concentration
()
Top 5 Ownership
()
Institutional
Total Ownership
()
1 (Small) 1181 76008 234700 2720 611 200 327
2 2668 90206 257955 1796 495 217 438
3 4193 105946 275120 1517 454 220 485
4 6045 129892 296067 1376 420 206 491
5 8742 155391 315864 1216 394 198 502
6 13086 191888 345942 1072 380 202 531
7 20790 239501 377021 0874 368 196 532
8 34920 314955 427808 0687 354 190 538
9 67747 450970 480319 0569 328 173 528
10 (Big) 366540 846066 614354 0297 275 138 502
28
Table IV Pay-for-Performance Sensitivity of Option Grants as a Function of Institutional Influence
Similar in construction to parts of HS (Table II) this table shows the coefficients from Tobit regressions of option grant pay-for-performance sensitivity per $1000 change in
shareholder wealth against the lagged change and contemporary change in shareholder wealth measures of institutional influence Tobinrsquos Q firm market capitalization and
dummy controls for CEO and year effects We estimate models (1) to (4) over the period 1992‒1997 using data supplied by HS and models (5) to (8) are estimated over the
whole sample Models (1) (2) (5) and (6) use Institutional Concentration as a measure of institutional influence which HS define as the shares owned by the institutions
with the five largest holdings divided by the shares held by all institutions These models control for total institutional ownership Models (3) (4) (7) and (8) use Top 5
Ownership as a measure of institutional influence which we define as the proportion of shares outstanding held by the institutions with the five largest holdings Coefficients
for industry and year controls are not presented One two and three asterisks denote significance at 10 5 and 1 levels
Hartzell and Starks Database 1992‒1997 Full Sample from ExecuComp 1992‒2002
(1) (2) (3) (4) (5) (6) (7) (8)
Change in Shareholder Wealth 77721 191781 89874 137368 -446 -218 -664 -203
(171) (429) (194) (317) (319) (168) (453) (158)
Lag1 Change in Shareholder Wealth 14446 99963 88719 8544 -144 0273 -0342 00852
(027) (212) (166) (184) (071) (015) (016) (005)
Lag1 Institutional Concentration 13994 -00692 26405 03304
(767) (032) (1402) (166)
Lag1 Total Institutional Ownership 08055 10434 07106 0937
(593) (781) (666) (898)
Lag1 Top 5 Institutional Ownership 10567 07457 12116 06293
(399) ( 283) (533) (278)
Lag1 Tobinrsquos Q 0068 00877 00647 00813 00498 00727 00508 00695
(337) (434) (319) (401) (504) (734) (51) (703)
Lag1 Market Capitalization -106202 -173515 -776 -117
(474) (742) (1011) (153)
Lag1 Log Market Capitalization -0297 -02319 -04111 -03908
(1576) (1439) (2997) (2841)
CEO Dummy 13745 13682 13765 13674 28197 28128 28155 28143
(1913) (1918) (1914) (1916) (3388) (3396) (3377) (34)
Number of Observations 33878 33878 33878 33878 97679 97679 97679 97679
29
Table V Pay-for-Performance Sensitivity of Cash Compensation (Salary plus Bonus) as a Function of Institutional Influence - Level Market Capitalization
Similar in parts to models (1) and (2) of HS (Table IV) this table displays the coefficients from OLS regressions of the change in cash compensation against the lagged
change and contemporary change in shareholder wealth measures of institutional influence Tobinrsquos Q firm market capitalization and dummy controls for CEO and year
effects We estimate models (1) to (4) over the period 1992‒1997 based on our ExecuComp sample and models (5) to (8) over the whole sample Models (1) (2) (5) and (6)
use Institutional Concentration as a measure of institutional influence which HS define as the shares owned by the institutions with the five largest holdings divided by the
shares held by all institutions These models control for total institutional ownership Models (4) and (8) use Top 5 Ownership as a measure of institutional influence which
we define as the proportion of shares outstanding held by the institutions with the 5 largest holdings Coefficients for industry and year controls are not presented One two
and three asterisks denote significance at 10 5 and 1 levels
Our Database based on ExecuComp 1992‒1997 Full Sample based on ExecuComp 1992‒2002
(1) (2) (3) (4) (5) (6) (7) (8)
Lag1 Change in Shareholder Wealth 00014 00017 00088 00089 -00006 -00007 00072 00086
(051) (062) (129) (103) (027) (029) (131) (16)
Change in Shareholder Wealth -00102 00102 00105 00032 00102 00101
(097) (292) (293) (055) (611) (608)
Change Shareholder Wealth Lag1 Institutional
Concentration
00435 00764 00363 00253
(463) (258) (689) (144)
Lag1 Change Shareholder Wealth Lag1
Institutional Concentration
-00197 -00283
(122) (197)
Lag1 Institutional Concentration -492182 -288861 400149 453623
(09) (056) (098) (114)
Lag1 Total Institutional Ownership 720688 538372 655091 559587 682069 726226
(276) (232) (281) (314) (308) (322)
Lag1 Change Shareholder Wealth Lag1 Top 5
Institutional Ownership
-0046 -00689
(091) (256)
Lag1 Top 5 Institutional Ownership 583044 1191634
(092) (25)
Lag1 Tobinrsquos Q -32761 -36806 -19422 -18023 -54283 -48681 -29534 -24243
(138) (172) (077) (078) (227) (245) (136) (111)
Lag1 Market Capitalization 00004 00004 00002 00006 00007 00008 00007 00007
(042) (044) (022) (07) (086) (092) (082) (09)
CEO Dummy 768035 765474 768334 773634 616315 61729 619028 618535
(354) (355) (356) (355) (417) (418) (419) (418)
Adjusted R-Sq 00057 0006 00052 0005 00097 00097 00098 00102
Number of Observations 41305 41305 41305 41305 85518 85518 85518 85518
30
Table VI Level of Salary as a Function of Institutional Influence
Similar in parts to models (1) and (2) of HS (Table V) this table displays the coefficients from OLS regressions of salary against the lagged change and contemporary change in
shareholder wealth measures of institutional influence Tobinrsquos Q firm market capitalization and dummy controls for CEO industry and year effects We estimate models (1) to
(4) over the period 1992‒1997 using the actual data supplied by HS and models (5) to (8) over the whole sample 1992‒2002 Models (1) (2) (5) and (6) use Institutional
Concentration as a measure of institutional influence which HS define as the shares owned by the institutions with the five largest holdings divided by the shares held by all
institutions These models control for total institutional ownership Models (3) (4) (7) and (8) use Top 5 Ownership as a measure of institutional influence which we define as
the proportion of shares outstanding held by the institutions with the 5 largest holdings Coefficients for industry and year controls are not presented One two and three
asterisks denote significance at 10 5 and 1 levels
Hartzell and Starks Database 1992‒1997 Full Sample based on ExecuComp 1992‒2002
(1) (2) (3) (4) (5) (6) (7) (8)
Change in Shareholder Wealth 04125E-4 0003 -2439E-4 0003 00021 00012 00024 00012
(057) (464) (031) (468) (1004) (72) (1041) (706)
Lag1 Change in Shareholder Wealth -00062 7856E-4 -0007 9003E-4 -00014 00002 -00014 00003
(594) (098) (592) (112) (522) (123) (5) (145)
Lag1 Institutional Concentration -2475087 555134 -2734983 943317
(3245) (710) (4965) (1671)
Lag1 Total Institutional Ownership 1020109 217488 936337 252464
(1904) (432) (2471) (704)
Lag1 Top 5 Institutional Ownership -148355 31354 -87926 423166
(139) (332) (118) (643)
Lag1 Tobinrsquos Q -195348 -252191 -192344 -25158 -119837 -146373 -122132 -143821
(2812) (3762) (2721) (3752) (248) (274) (2583) (2727)
Lag1 Market Capitalization 844E-06 103E-5 00034 00039
(2781) (3051) (2874) (3104)
Lag1 Log Market Capitalization 750021 723625 776657 732279
(7511) (9285) (11224) (14039)
CEO Dummy 2683978 2702238 2668244 2702417 3281133 3285103 3292175 3283145
(8625) (9277) (8246) (9271) (12739) (13616) (12336) (13605)
Adjusted R-Sq 04604 05302 04187 05295 03856 04588 03476 04574
Number of Observations 36280 36280 36280 36280 97679 97679 97679 97679
31
Table VII Level of Total Direct Compensation as a Function of Institutional Influence
Similar in parts to models (3) and (5) of HS (Table V) this table displays the coefficients from OLS regressions of total direct compensation against the lagged change and
contemporary change in shareholder wealth measures of institutional influence Tobinrsquos Q firm market capitalization and dummy controls for CEO industry and year effects
We estimate models (1) to (4) over the period 1992‒1997 and models (5) to (8) over the whole sample Models (1) (2) (5) and (6) use Institutional Concentration as a measure
of institutional influence which HS define as the shares owned by the institutions with the five largest holdings divided by the shares held by all institutions These models
control for total institutional ownership Models (3) (4) (7) and (8) use Top 5 Ownership as a measure of institutional influence which we define as the proportion of shares
outstanding held by the institutions with the five largest holdings Coefficients for industry and year controls are not presented One two and three asterisks denote significance
at 10 5 and 1 levels
Hartzell and Starks Database 1992‒1997 Full Sample from ExecuComp 1992‒2002
(1) (2) (3) (4) (5) (6) (7) (8)
Change in Shareholder Wealth 005779 007006 005282 006791 00423 00148 0046 00138
(415) (581) (369) (565) (203) (07) (22) (066)
Lag1 Change in Shareholder Wealth 003182 00714 00272 00715 00613 0126 00614 01278
(126) (298) (107) (299) (275) (488) (276) (494)
Lag1 Institutional Concentration -1608633 3248998 -3424892 3334186
(151) (283) (1172) (118)
Lag1 Total Institutional Ownership 104013 5473128 1531013 -2926764
(1645) (831) (876) (158)
Lag1 Top 5 Institutional Ownership 188585 3927986 -2159177 -2640879
(016) (343) (067) (092)
Lag1 Tobinrsquos Q -445692 -803612 -454859 -833566 3257027 2971201 3213936 3086205
(427) (791) (432) (831) (83) (77) (815) (793)
Lag1 Market Capitalization 49E-5 615E-5 01031 01088
(113) (1365) (1555) (168)
Lag1 Log Market Capitalization 4683497 4759377 1645156 1427856
(2775) (3469) (3679) (4239)
CEO Dummy 1501307 1508886 1493688 1508736 4713079 4713629 47278 4703908
(2648) (2684) (2618) (2683) (282) (2828) (2819) (2824)
Adjusted R-Sq 01861 02060 01695 02051 01071 01113 01033 011
Number of Observations 33229 33229 33229 33229 97679 97679 97679 97679
32
Table VIII Two Stage Least Squares (Reverse Causality) Tests - Executive Compensation Measures as a Function of Institutional Concentration
This table shows the coefficients from the Two-Stage Least Squares (2SLS) regressions of executive compensation measures against the lagged change
and contemporary change in shareholder wealth instrumented institutional concentration and total institutional ownership Tobinrsquos Q the natural
logarithm of firm market capitalization and dummy variable controls for CEO and year effects We instrument measures of institutional concentration
and total institutional ownership by orthogonalizing with respect to option grant pay-for-performance sensitivity salary and total compensation
Coefficients for industry and year controls are not presented First stage results are not presented One two and three asterisks denote significance at
10 5 and 1 levels
ExecuComp Sample 1992‒1997 Full Sample from ExecuComp 1992‒2002
PPSOG Salary TC PPSOG Salary TC
Change in Shareholder Wealth 144E-05 0004 02133 -409E-06 00011 00157
(308) (775) (596) (691) (638) (068)
Lag1 Change in Shareholder Wealth -240E-05 -00004 00866 118E-06 00002 01422
(131) (087) (346) (097) (116) (51)
Lag1 Institutional Concentration -00898 668681 3263143 02024 916656 2856417
(052) (801) (176) (163) (1545) (1004)
Lag1 Total Institutional Ownership 01021 -43997 2512587 02983 -104171 -1028347
(09) (084) (153) (4) (269) (487)
Lag1 Tobinrsquos Q 00754 -225606 119297 00655 -158657 2997205
(596) (1649) (053) (834) (2535) (727)
Lag1 Log Market Capitalization -04234 828713 7864046 -04002 825106 1685638
(2195) (8132) (2977) (4106) (11133) (373)
CEO Dummy 15848 3214117 2950274 19149 3205862 4529101
(2223) (9389) (2255) (3802) (1331) (2758)
Overall R-Sq -- 04981 01206 -- 04738 01178
Number of Observations 41305 41305 41305 85518 85518 85518
33
Table IX Two Stage Least Squares (Reverse Causality) Tests - Executive Compensation Measures as a Function of Top 5 Ownership
This table shows the coefficients from 2SLS regressions of executive compensation measures against the lagged change and contemporary change
in shareholder wealth instrumented Top 5 institutional ownership Tobinrsquos Q natural logarithm of firm market capitalization and dummy variable
controls for CEO and year effects We instrument Top 5 institutional ownership by orthogonalizing with respect to option grant pay-for-
performance sensitivity salary and total compensation Coefficients for industry and year controls are not presented First stage results are not
presented One two and three asterisks denote significance at 10 5 and 1 levels
ExecuComp sample 1992‒1997 Full Sample 1992‒2002
PPSOG Salary TC PPSOG Salary TC
Change in Shareholder Wealth 136E-05 00041 02128 -407E-06 00011 00146
(293) (79) (596) (692) (619) (063)
Lag1 Change in Shareholder Wealth -241E-05 -00003 00868 116E-06 00003 01439
(133) (07) (347) (096) (142) (514)
Lag1 Top 5 Institutional Ownership -03774 -187512 -3848531 00179 -104261 -126659
(163) (186) (166) (011) (149) (436)
Lag1 Tobinrsquos Q 00731 -223886 10074 00642 -155467 3115073
(573) (1631) (045) (819) (2509) (75)
Lag1 Log Market Capitalization -0416 775637 7696302 -04015 764632 1462536
(2345) (10084) (3766) (3688) (13708) (4072)
CEO Dummy 15854 3211991 2950158 19152 3203152 4519385
(2223) (9381) (2254) (3804) (133) (2752)
Overall R-Sq -- 04967 01202 -- 04722 01168
Number of Observations 41305 41305 41305 85518 85518 85518
34
Table X Specification Tests utilizing Box-Cox Transformation of Market Capitalization
This table presents results from a Box-Cox transformation of market capitalization In
each model we specify transformation of market capitalization only Panel A shows
results for the period 1992‒1997 and Panel B shows results for the entire period
1992‒2002 The first part of each panel shows the Lambda coefficient that applies to
the transformed market capitalization The second part of each panel shows log
likelihood scores with LR Chi2 Statistics in parentheses ‒ these test for the
appropriateness of a linear or log model specification
Panel A Our ExecuComp sample 1992‒1997
(1) (2) (3)
PPSOG Salary TC
Lambda -05143 01881 02876
(-1914) (2413) (1329)
Test H0 (Log Likelihood Score)
Lambda = -1 -1279748 -3178575 -4795326
(23866) (85952) (99688)
Lambda = 0 -1280152 -3138456 -4791159
(31934) (57147) (16339)
Lambda = 1 -1283157 -3163807 -4793506
(92039) (564156) (63292)
Panel B Full ExecuComp sample 1992‒2002
(1) (2) (4)
PPSOG Salary TC
Lambda -044 01853 04036
(-3015) (40) (3666)
Test H0 (Log Likelihood Score)
Lambda = 0 -2688334 -6476494 -10384490
(66599) (160356) (126236)
Lambda = 1 -2694886 -6538506 -10386795
(197644) (1400605) (172345)
35
Figure 2 Institutional Ownership Measures (LHA) vs Total
Compensation (RHA) by Market Capitalization
0
10
20
30
40
50
60
70
1 2 3 4 5 6 7 8 9 10
$00
$10
$20
$30
$40
$50
$60
$70
$80
$90
Millio
ns
InstitutionalConcentration (LHS)Top5 Ownership (LHS)
Institutional TotalOwnership (LHS)Total Compensation(RHS)
Page 17
17
both model (3) and in model (4) but the better size control reduces the coefficient by over
25 percent The better size control does not affect Top 5 Ownership by as much as its
design ensures less sensitivity to what could be inadequate size controls
We also test the effect of institutional influence on option grant pay-for-performance
sensitivity over the period 1992‒2002 This sample is then approximately three times as
large as the sample used by HS Institutional concentration is highly positively significant
in model (5) similar to the results for the period 1992‒1997 When we use the log of
market capitalization in model (6) the coefficient on institutional concentration remains
positive and significant at the 10 percent level with a t-statistic of 166 Of importance
though is that the coefficient on institutional concentration is one-eighth of the coefficient
in model (5) Even for a measure of institutional influence which we believe is less
sensitive to firm size the use of log market capitalization reduces the coefficient on Top 5
Ownership by approximately 50 percent from model (7) to model (8)
B Pay-for-Performance Sensitivity due to Changes in Cash and Total
Compensation
For robustness HS examine the effect that institutional influence conditioned on firm
performance has on a changes in cash and total executive compensation using the general
method of Jensen and Murphy (1990) to compute pay-for-performance sensitivity
ltltINSERT TABLE V gtgt
Table V presents the effect of institutional influence on the change in cash
compensation using the HS size control measure throughout12
Model (1) replicates the
specification in HS for the period of 1992‒1997 and we produce a similar result ndash the
interaction of lagged institutional concentration and contemporaneous change in
12 This test uses our data for the two periods 1992‒1997 and 1992‒2002 We do not use the HS data for
cash compensation as data for bonus is missing in the data supplied to us
18
shareholder wealth positively and significantly relates to a change in executive cash
compensation In model (2) we include the contemporaneous change in shareholder
wealth and the lagged institutional concentration as stand-alone controls as indicated by
Greenersquos (2000) specification of equation (1) above and the interaction term remains
significant In model (3) we interact lagged institutional concentration with the lagged
change in shareholder wealth as in equation (2) above that tries to take account of what
participants could have observed at the time The interaction term is then negative but
insignificant altering the HS result Model (4) provides a similar result where our
preferred proxy Top 5 Ownership captures concentrated institutional influence In
models (5) to (8) we use our full sample In model (5) we find that the lagged
institutional concentration interaction with the contemporaneous change in shareholder
wealth is positive and significant in HSrsquos specification However including change in
contemporaneous shareholder wealth and lagged institutional concentration as separate
control variables as suggested by equation (1)rsquos specification removes the significance
from the interaction term in model (6) Moreover in model (7) when lagged institutional
concentration is interaction with the lagged change in shareholder wealth in accord with
equation (2) this interaction term is now negative and significant reversing the HS
finding We obtain a similar result in model (8) when we measure institutional
concentration influence using Top 5 Ownership13
V The Levels of Executive Fixed and Total Compensation as Functions of
Institutional Investor Influence
13 In unreported results we use the log of market capitalization as a size control These changes to the size
control do not dramatically change the findings from Table V The reason why this size issue is not as
important as in the findings for option grants pay-for-performance is that the change in cash compensation
is not so strongly influenced by firm size In unreported results available on request we also examine the
effect of institutional influence on changes in total compensation with similar results to the change in cash
compensation
19
Two measures of compensation are examined ndash salary and total compensation
ltltINSERT TABLE VI gtgt
Table VI examines the effect of institutional influence on the level of salary In model
(1) we use the HS sample and specification to obtain almost identical results The
coefficient on institutional concentration is (negative) $247509 that is almost identical to
(negative) $247967 in HS Both are highly significant In model (2) we make a single
change to the model ndash we use the natural logarithm of market capitalization This single
change reverses the sign of institutional concentration with the coefficient still being
significant at the 1 percent level In models (3) and (4) we measure institutional influence
using Top 5 Ownership In model (3) where we measure firm size using the level of
market capitalization Top 5 Ownership is insignificant When we use the natural
logarithm of market capitalization in model (4) Top 5 Ownership is significant and
positively relates to the salary level In models (5) to (8) we estimate the same models
using our full sample We find similar results for the full sample as for the period 1992ndash
1997
ltltINSERT TABLE VII gtgt
Table VII examines the effect of institutional influence on total compensation In
model (1) we replicate the HS methodology using the HS data to obtain a similar
significant negative relation between institutional concentration and total compensation of
(negative) $1608633 However after changing to a logarithmic specification of market
capitalization in model (2) institutional concentration increases total compensation by a
significant $32400014
Importantly when institutional influence is measured using Top 5
14 = (5688751 times 0101)
20
Ownership an insignificant relation exists with total compensation in model (3) using the
HS size control but it becomes both positive and significant in (4) with the logged size
control In models (5) and (6) where models are estimated using the full sample we
observe a similar reversal in sign as occurs in models (1) and (2) However now the
magnitude of the reversal is almost equal with a difference of about $67m that is very
large indeed In model (5) the coefficient on institutional concentration is (negative)
$342m and then in model (6) the coefficient is $33m Top 5 Ownership is insignificant
in both models (7) and (8) unlike the results for Salary
VI Robustness
Two issues are examined in is this Section ndash reverse causality and the appropriate
transformation of market capitalization A methodological issue inherent in examining the
effect of ownership-based measures of institutional influence on executive compensation
is endogeneity driven by reverse causality We utilize an instrumental variables approach
where orthogonalized institutional influence measures are regressed on each of the
compensation measures lagged option grant pay-for-performance sensitivity salary and
total compensation In addition market capitalization and a CEO dummy are included in
each regression We use the residual from each regression as an instrument for
institutional influence Such a process removes any preferences for compensation from
the institutional influence measures
ltltINSERT TABLES VIII AND IXgtgt
Table VIII presents the results where we instrument the HS measure lagged
institutional concentration with a compensation-orthogonalized equivalent The analysis
indicates that the instrumented institutional concentration measure is positively and
significantly related to the level of salary and total compensation and insignificantly
21
related to option grant pay-for-performance sensitivity (PPSOG) Table IX presents
results where we instrument Top 5 Ownership with a compensation-orthogonalized
equivalent The instrumented Top 5 Ownership measure insignificantly relates to option
grant pay-for-performance sensitivity and barely negatively relates to salary although it
negatively relates to total compensation suggesting reverse causality is driving a
significant portion of our previous results
ltltINSERT TABLE Xgtgt
Another issue examined in this section is the appropriate transformation of market
capitalization Aneuryn-Evans and Deaton (1980) provide an evaluation of methods
comparing linear and log-linear specifications We use the Box-Cox test to compare
logarithm versus linear specifications In applying the Box-Cox test we specify
transformation of only market capitalization ndash the dependent variable and all other
independent variables remain unchanged Table X presents the lambda coefficient for
each model and the log likelihood score In all models in Panel A and B a lambda of zero
receives a higher (closer to zero) log likelihood score suggesting the log specification is
appropriate
VII Conclusion
Hartzell and Starks (2003) appears to document an effective role for institutional
investors in increasing executive incentives and lowering compensation levels Our paper
re-addresses the issue of concentrated institutional monitoring from a perspective
informed by an additional decade of research since their paper appeared We cannot find
evidence that concentrated institutional investors really have any positive (or negative for
that matter) effect on executive incentive levels While superficially we find evidence
that they may raise executive pay levels this is most likely driven by an institutional
22
preference for highly remunerated executives in large companies and thus by reverse
causality Moreover our findings are not consistent with the rent-extraction hypothesis
suggesting that firms systematically overpay executives in the absence of effective
monitoring by outside parties such as institutional investors
23
References
Almazan Andres Jay C Hartzell and Laura T Starks 2005 ldquoActive Institutional
Shareholders and Costs of Monitoring Evidence from Executive Compensationrdquo
Financial Management Winter 5ndash34
Aneuryn-Evans Gwyn and Angus Deaton 1980 ldquoTesting Linear versus Logarithmic
Regression Modelsrdquo Review of Economic Studies 47 (1) Econometrics Issue 275ndash
291
Babenko Ilona 2009 ldquoShare Repurchases and Pay-Performance Sensitivity of Employee
Compensation Contractsrdquo Journal of Finance 64117-150
Baker George P and Brian J Hall 2004 ldquoCEO Incentives and Firm Sizerdquo Journal of
Labor Market Economics 22 767ndash798
Bebchuk Lucian A Jesse M Fried and David Walker 2002 ldquoManagerial power and
rent extraction in the design of executive compensationrdquo University of Chicago Law
Review 69 751ndash846
Bebchuk Lucian A and Jesse M Fried 2003 ldquoExecutive Compensation as an Agency
Problemrdquo Journal of Economic Perspectives 17 71ndash92
Bebchuk Lucian A and Jesse M Fried 2004 Pay without Performance The Unfulfilled
Promise of Executive Compensation Cambridge MA Harvard University Press
Bertrand Marianne and Sendhil Mullainathan 2001 ldquoAre CEOs Rewarded for Luck
The Ones Without Principals Arerdquo Quarterly Journal of Economics 116 901ndash32
Black Fischer and Myron Scholes 1973 ldquoThe Pricing of Options and Corporate
Liabilitiesrdquo Journal of Political Economy 81 637ndash54
Cadman Brian Sandy Klasa and Steve Matsunaga 2010 ldquoDeterminants of CEO Pay A
Comparison of ExecuComp and Non-ExecuComp Firmsrdquo The Accounting Review 85
1511ndash1543
Chhaochharia Vidhi and Yaniv Grinstein 2009 ldquoCEO Compensation and Board
Structurerdquo Journal of Finance 64 231ndash261
Cichello Michael S 2005 ldquoThe impact of firm size on pay-performance sensitivitiesrdquo
Journal of Corporate Finance 11 609ndash627
Davis Gerald F and E Han Kim 2007 ldquoBusiness ties and proxy voting by mutual
fundsrdquo Journal of Financial Economics 85 552ndash570
Dikolli Shane S Susan L Kulp and Karen L Sedatole 2009 ldquoTransient Institutional
Ownership and CEO Contractingrdquo Accounting Review 84 737ndash770
Edmans Alex Xavier Gabaix and Augustin Landier 2009 ldquoA Multiplicative Model of
Optimal CEO Incentives in Market Equilibriumrdquo Review of Financial Studies 22
4881ndash4917
Edmans Alex 2009 ldquoBlockholder Trading Market Efficiency and Managerial Myopiardquo
Journal of Finance 64 2481ndash2513
Ertimur Yonca Fabrizio Ferri and Volkan Muslu 2011 ldquoShareholder Activism and
CEO Payrdquo Review of Financial Studies 24 535ndash592
24
Gabaix Xavier and Augustin Landier 2008 ldquoWhy Has CEO Pay Increased So Muchrdquo
Quarterly Journal of Economics 123 49ndash100
Garen John 1994 ldquoExecutive compensation and principal-agent theoryrdquo Journal of
Political Economy 102 1175ndash1199
Gibbons Robert and Kevin J Murphy 1992 ldquoOptimal Incentive Contracts in the
Presence of Career Concerns Theory and Evidencerdquo Journal of Political Economy
100 468ndash505
Greene William H 2000 (4th
Edition) Econometric Analysis Prentice-Hall NJ
Guthrie Katherine Jan Sokolowsky and Kam-Ming Wan 2012 ldquoCEO Compensation
and Board Structure Revisitedrdquo Journal of Finance Forthcoming
Hadlock Charles R and Gerald B Lumer 1997 ldquoCompensation turnover and top
management incentives historical evidencerdquo Journal of Business 70 153ndash187
Hartzell Jay C and Laura T Starks 2003 ldquoInstitutional Investors and Executive
Compensationrdquo Journal of Finance 58 2351ndash2374
Huson Mark R Robert Parrino and Laura T Starks 2001 ldquoInternal Monitoring
Mechanisms and CEO Turnover A Long-Term Perspectiverdquo Journal of Finance 56
2265ndash2297
Jensen Michael C and Kevin J Murphy 1990 ldquoPerformance Pay and Top Management
Incentivesrdquo Journal of Political Economy 98 225ndash264
Jin Li 2002 ldquoCEO compensation diversification and incentivesrdquo Journal of Financial
Economics 66 29ndash63
Kang Qiang and Qiao Liu 2008 ldquoStock Market Information Production and Executive
Incentivesrdquo Journal of Corporate Finance 14 484ndash498
Merton Robert 1973 ldquoTheory of Rational Option Pricingrdquo Bell Journal of Economics
and Management Science 4 141-183
Morse Adair Vikram Nanda and Amit Seru 2011 ldquoAre Incentive Contracts Rigged by
Powerful CEOsrdquo Journal of Finance 66 1779ndash1821
Murphy Kevin J 1985 ldquoCorporate performance and managerial remuneration An
empirical analysisrdquo Journal of Accounting and Economics 7 11ndash42
Murphy Kevin J 1999 ldquoExecutive compensationrdquo in Orley Ashenfelter and David
Card Handbook of Labor Economics vol 3 Elsevier
Rosen Sherwin 1992 ldquoContracts and the market for executivesrdquo In Contract
Economics ed Lars Werin and Hans Wijkander 181ndash211 Cambridge MA
Blackwell
Rothberg Burton and Steven Lilien 2006 ldquoMutual Funds and Proxy Voting New
Evidence on Corporate Governancerdquo Journal of Business and Technology Law 1
157ndash184
Schaefer Scott 1998 ldquoThe dependence of payndashperformance sensitivity on the size of the
firmrdquo Review of Economics and Statistics 80 436ndash443
Yermack David 1995 ldquoDo Corporations Award CEO Stock Options Effectivelyrdquo
Journal of Financial Economics 39 237ndash69
25
Table I Univariate Descriptive Statistics
This table reports sample statistics for the key variables used in this study based on our comprehensive
ExecuComp dataset Institutional Concentration is the shares owned by the institutions with the five largest
holdings divided by the shares held by all institutions Top 5 Ownership is the proportion of shares
outstanding held by the institutions with the five largest holdings Total Institutional Ownership is the
proportion of shares held by all institutions Panel B shows the measures executive compensation The pay-
for-performance sensitivity (PPS) of option grants shows the change in executive wealth due to a $1000
change in firm in shareholder wealth Expressions for Salary and Total Compensation are in thousands of
dollars while Market Capitalization is in millions of dollars Panel C shows firm characteristics
Variable Mean Median Std Dev P10 P90
Panel A Institutional Investor Holdings
Institutional Concentration 0450 0424 0151 0282 0656
Top 5 Institutional Ownership 0229 0220 0101 0110 0355
Total Institutional Ownership 0546 0561 0206 0258 0801
Panel B Executive Compensation
PPS of Option Grants 1161 0156 3829 0 2822
Salary 36311 29462 25231 15179 67000
Total Compensation 258313 85878 980701 23875 511022
Panel C Firm Characteristics
Tobins Q 1744 1139 2738 0495 3322
Market Capitalization 560699 112094 1988108 19754 1048754
26
Table II Pairwise Correlation Coefficients
This table presents the correlations between the key variables based on our comprehensive
ExecuComp dataset Institutional Concentration is the shares owned by the institutions with the five
largest holdings divided by the shares held by all institutions Top 5 Ownership is the proportion of
shares outstanding held by the institutions with the five largest holdings Total Institutional
Ownership is the proportion of shares held by all institutions The pay-for-performance sensitivity
(PPS) of option grants is the change in executive wealth due to a $1000 change in shareholder
wealth
(1) (2) (3) (4) (5) (6) (7) (8)
(1) Institutional Concentration 1
(2) Top 5 Ownership 0268 1
(3) Total Ownership -0415 0706 1
(4) PPS of Option Grants 0122 0042 -0024 1
(5) Log Salary -0270 -0012 0171 0038 1
(6) Log Market Capitalization -0630 -0157 0261 -0156 0420 1
(7) Level Market Capitalization -0241 -0157 0007 -0058 0219 0534 1
(8) Tobins Q -0008 -0034 0002 0015 -0083 0066 0121 1
27
Table III
Descriptive Statistics by Market Capitalization Deciles
Market
Capitalization
Deciles
Market
Capitalization
($m)
Total
Compensation
($000)
Salary
($)
PPS Option
Grants
($)
Institutional
Concentration
()
Top 5 Ownership
()
Institutional
Total Ownership
()
1 (Small) 1181 76008 234700 2720 611 200 327
2 2668 90206 257955 1796 495 217 438
3 4193 105946 275120 1517 454 220 485
4 6045 129892 296067 1376 420 206 491
5 8742 155391 315864 1216 394 198 502
6 13086 191888 345942 1072 380 202 531
7 20790 239501 377021 0874 368 196 532
8 34920 314955 427808 0687 354 190 538
9 67747 450970 480319 0569 328 173 528
10 (Big) 366540 846066 614354 0297 275 138 502
28
Table IV Pay-for-Performance Sensitivity of Option Grants as a Function of Institutional Influence
Similar in construction to parts of HS (Table II) this table shows the coefficients from Tobit regressions of option grant pay-for-performance sensitivity per $1000 change in
shareholder wealth against the lagged change and contemporary change in shareholder wealth measures of institutional influence Tobinrsquos Q firm market capitalization and
dummy controls for CEO and year effects We estimate models (1) to (4) over the period 1992‒1997 using data supplied by HS and models (5) to (8) are estimated over the
whole sample Models (1) (2) (5) and (6) use Institutional Concentration as a measure of institutional influence which HS define as the shares owned by the institutions
with the five largest holdings divided by the shares held by all institutions These models control for total institutional ownership Models (3) (4) (7) and (8) use Top 5
Ownership as a measure of institutional influence which we define as the proportion of shares outstanding held by the institutions with the five largest holdings Coefficients
for industry and year controls are not presented One two and three asterisks denote significance at 10 5 and 1 levels
Hartzell and Starks Database 1992‒1997 Full Sample from ExecuComp 1992‒2002
(1) (2) (3) (4) (5) (6) (7) (8)
Change in Shareholder Wealth 77721 191781 89874 137368 -446 -218 -664 -203
(171) (429) (194) (317) (319) (168) (453) (158)
Lag1 Change in Shareholder Wealth 14446 99963 88719 8544 -144 0273 -0342 00852
(027) (212) (166) (184) (071) (015) (016) (005)
Lag1 Institutional Concentration 13994 -00692 26405 03304
(767) (032) (1402) (166)
Lag1 Total Institutional Ownership 08055 10434 07106 0937
(593) (781) (666) (898)
Lag1 Top 5 Institutional Ownership 10567 07457 12116 06293
(399) ( 283) (533) (278)
Lag1 Tobinrsquos Q 0068 00877 00647 00813 00498 00727 00508 00695
(337) (434) (319) (401) (504) (734) (51) (703)
Lag1 Market Capitalization -106202 -173515 -776 -117
(474) (742) (1011) (153)
Lag1 Log Market Capitalization -0297 -02319 -04111 -03908
(1576) (1439) (2997) (2841)
CEO Dummy 13745 13682 13765 13674 28197 28128 28155 28143
(1913) (1918) (1914) (1916) (3388) (3396) (3377) (34)
Number of Observations 33878 33878 33878 33878 97679 97679 97679 97679
29
Table V Pay-for-Performance Sensitivity of Cash Compensation (Salary plus Bonus) as a Function of Institutional Influence - Level Market Capitalization
Similar in parts to models (1) and (2) of HS (Table IV) this table displays the coefficients from OLS regressions of the change in cash compensation against the lagged
change and contemporary change in shareholder wealth measures of institutional influence Tobinrsquos Q firm market capitalization and dummy controls for CEO and year
effects We estimate models (1) to (4) over the period 1992‒1997 based on our ExecuComp sample and models (5) to (8) over the whole sample Models (1) (2) (5) and (6)
use Institutional Concentration as a measure of institutional influence which HS define as the shares owned by the institutions with the five largest holdings divided by the
shares held by all institutions These models control for total institutional ownership Models (4) and (8) use Top 5 Ownership as a measure of institutional influence which
we define as the proportion of shares outstanding held by the institutions with the 5 largest holdings Coefficients for industry and year controls are not presented One two
and three asterisks denote significance at 10 5 and 1 levels
Our Database based on ExecuComp 1992‒1997 Full Sample based on ExecuComp 1992‒2002
(1) (2) (3) (4) (5) (6) (7) (8)
Lag1 Change in Shareholder Wealth 00014 00017 00088 00089 -00006 -00007 00072 00086
(051) (062) (129) (103) (027) (029) (131) (16)
Change in Shareholder Wealth -00102 00102 00105 00032 00102 00101
(097) (292) (293) (055) (611) (608)
Change Shareholder Wealth Lag1 Institutional
Concentration
00435 00764 00363 00253
(463) (258) (689) (144)
Lag1 Change Shareholder Wealth Lag1
Institutional Concentration
-00197 -00283
(122) (197)
Lag1 Institutional Concentration -492182 -288861 400149 453623
(09) (056) (098) (114)
Lag1 Total Institutional Ownership 720688 538372 655091 559587 682069 726226
(276) (232) (281) (314) (308) (322)
Lag1 Change Shareholder Wealth Lag1 Top 5
Institutional Ownership
-0046 -00689
(091) (256)
Lag1 Top 5 Institutional Ownership 583044 1191634
(092) (25)
Lag1 Tobinrsquos Q -32761 -36806 -19422 -18023 -54283 -48681 -29534 -24243
(138) (172) (077) (078) (227) (245) (136) (111)
Lag1 Market Capitalization 00004 00004 00002 00006 00007 00008 00007 00007
(042) (044) (022) (07) (086) (092) (082) (09)
CEO Dummy 768035 765474 768334 773634 616315 61729 619028 618535
(354) (355) (356) (355) (417) (418) (419) (418)
Adjusted R-Sq 00057 0006 00052 0005 00097 00097 00098 00102
Number of Observations 41305 41305 41305 41305 85518 85518 85518 85518
30
Table VI Level of Salary as a Function of Institutional Influence
Similar in parts to models (1) and (2) of HS (Table V) this table displays the coefficients from OLS regressions of salary against the lagged change and contemporary change in
shareholder wealth measures of institutional influence Tobinrsquos Q firm market capitalization and dummy controls for CEO industry and year effects We estimate models (1) to
(4) over the period 1992‒1997 using the actual data supplied by HS and models (5) to (8) over the whole sample 1992‒2002 Models (1) (2) (5) and (6) use Institutional
Concentration as a measure of institutional influence which HS define as the shares owned by the institutions with the five largest holdings divided by the shares held by all
institutions These models control for total institutional ownership Models (3) (4) (7) and (8) use Top 5 Ownership as a measure of institutional influence which we define as
the proportion of shares outstanding held by the institutions with the 5 largest holdings Coefficients for industry and year controls are not presented One two and three
asterisks denote significance at 10 5 and 1 levels
Hartzell and Starks Database 1992‒1997 Full Sample based on ExecuComp 1992‒2002
(1) (2) (3) (4) (5) (6) (7) (8)
Change in Shareholder Wealth 04125E-4 0003 -2439E-4 0003 00021 00012 00024 00012
(057) (464) (031) (468) (1004) (72) (1041) (706)
Lag1 Change in Shareholder Wealth -00062 7856E-4 -0007 9003E-4 -00014 00002 -00014 00003
(594) (098) (592) (112) (522) (123) (5) (145)
Lag1 Institutional Concentration -2475087 555134 -2734983 943317
(3245) (710) (4965) (1671)
Lag1 Total Institutional Ownership 1020109 217488 936337 252464
(1904) (432) (2471) (704)
Lag1 Top 5 Institutional Ownership -148355 31354 -87926 423166
(139) (332) (118) (643)
Lag1 Tobinrsquos Q -195348 -252191 -192344 -25158 -119837 -146373 -122132 -143821
(2812) (3762) (2721) (3752) (248) (274) (2583) (2727)
Lag1 Market Capitalization 844E-06 103E-5 00034 00039
(2781) (3051) (2874) (3104)
Lag1 Log Market Capitalization 750021 723625 776657 732279
(7511) (9285) (11224) (14039)
CEO Dummy 2683978 2702238 2668244 2702417 3281133 3285103 3292175 3283145
(8625) (9277) (8246) (9271) (12739) (13616) (12336) (13605)
Adjusted R-Sq 04604 05302 04187 05295 03856 04588 03476 04574
Number of Observations 36280 36280 36280 36280 97679 97679 97679 97679
31
Table VII Level of Total Direct Compensation as a Function of Institutional Influence
Similar in parts to models (3) and (5) of HS (Table V) this table displays the coefficients from OLS regressions of total direct compensation against the lagged change and
contemporary change in shareholder wealth measures of institutional influence Tobinrsquos Q firm market capitalization and dummy controls for CEO industry and year effects
We estimate models (1) to (4) over the period 1992‒1997 and models (5) to (8) over the whole sample Models (1) (2) (5) and (6) use Institutional Concentration as a measure
of institutional influence which HS define as the shares owned by the institutions with the five largest holdings divided by the shares held by all institutions These models
control for total institutional ownership Models (3) (4) (7) and (8) use Top 5 Ownership as a measure of institutional influence which we define as the proportion of shares
outstanding held by the institutions with the five largest holdings Coefficients for industry and year controls are not presented One two and three asterisks denote significance
at 10 5 and 1 levels
Hartzell and Starks Database 1992‒1997 Full Sample from ExecuComp 1992‒2002
(1) (2) (3) (4) (5) (6) (7) (8)
Change in Shareholder Wealth 005779 007006 005282 006791 00423 00148 0046 00138
(415) (581) (369) (565) (203) (07) (22) (066)
Lag1 Change in Shareholder Wealth 003182 00714 00272 00715 00613 0126 00614 01278
(126) (298) (107) (299) (275) (488) (276) (494)
Lag1 Institutional Concentration -1608633 3248998 -3424892 3334186
(151) (283) (1172) (118)
Lag1 Total Institutional Ownership 104013 5473128 1531013 -2926764
(1645) (831) (876) (158)
Lag1 Top 5 Institutional Ownership 188585 3927986 -2159177 -2640879
(016) (343) (067) (092)
Lag1 Tobinrsquos Q -445692 -803612 -454859 -833566 3257027 2971201 3213936 3086205
(427) (791) (432) (831) (83) (77) (815) (793)
Lag1 Market Capitalization 49E-5 615E-5 01031 01088
(113) (1365) (1555) (168)
Lag1 Log Market Capitalization 4683497 4759377 1645156 1427856
(2775) (3469) (3679) (4239)
CEO Dummy 1501307 1508886 1493688 1508736 4713079 4713629 47278 4703908
(2648) (2684) (2618) (2683) (282) (2828) (2819) (2824)
Adjusted R-Sq 01861 02060 01695 02051 01071 01113 01033 011
Number of Observations 33229 33229 33229 33229 97679 97679 97679 97679
32
Table VIII Two Stage Least Squares (Reverse Causality) Tests - Executive Compensation Measures as a Function of Institutional Concentration
This table shows the coefficients from the Two-Stage Least Squares (2SLS) regressions of executive compensation measures against the lagged change
and contemporary change in shareholder wealth instrumented institutional concentration and total institutional ownership Tobinrsquos Q the natural
logarithm of firm market capitalization and dummy variable controls for CEO and year effects We instrument measures of institutional concentration
and total institutional ownership by orthogonalizing with respect to option grant pay-for-performance sensitivity salary and total compensation
Coefficients for industry and year controls are not presented First stage results are not presented One two and three asterisks denote significance at
10 5 and 1 levels
ExecuComp Sample 1992‒1997 Full Sample from ExecuComp 1992‒2002
PPSOG Salary TC PPSOG Salary TC
Change in Shareholder Wealth 144E-05 0004 02133 -409E-06 00011 00157
(308) (775) (596) (691) (638) (068)
Lag1 Change in Shareholder Wealth -240E-05 -00004 00866 118E-06 00002 01422
(131) (087) (346) (097) (116) (51)
Lag1 Institutional Concentration -00898 668681 3263143 02024 916656 2856417
(052) (801) (176) (163) (1545) (1004)
Lag1 Total Institutional Ownership 01021 -43997 2512587 02983 -104171 -1028347
(09) (084) (153) (4) (269) (487)
Lag1 Tobinrsquos Q 00754 -225606 119297 00655 -158657 2997205
(596) (1649) (053) (834) (2535) (727)
Lag1 Log Market Capitalization -04234 828713 7864046 -04002 825106 1685638
(2195) (8132) (2977) (4106) (11133) (373)
CEO Dummy 15848 3214117 2950274 19149 3205862 4529101
(2223) (9389) (2255) (3802) (1331) (2758)
Overall R-Sq -- 04981 01206 -- 04738 01178
Number of Observations 41305 41305 41305 85518 85518 85518
33
Table IX Two Stage Least Squares (Reverse Causality) Tests - Executive Compensation Measures as a Function of Top 5 Ownership
This table shows the coefficients from 2SLS regressions of executive compensation measures against the lagged change and contemporary change
in shareholder wealth instrumented Top 5 institutional ownership Tobinrsquos Q natural logarithm of firm market capitalization and dummy variable
controls for CEO and year effects We instrument Top 5 institutional ownership by orthogonalizing with respect to option grant pay-for-
performance sensitivity salary and total compensation Coefficients for industry and year controls are not presented First stage results are not
presented One two and three asterisks denote significance at 10 5 and 1 levels
ExecuComp sample 1992‒1997 Full Sample 1992‒2002
PPSOG Salary TC PPSOG Salary TC
Change in Shareholder Wealth 136E-05 00041 02128 -407E-06 00011 00146
(293) (79) (596) (692) (619) (063)
Lag1 Change in Shareholder Wealth -241E-05 -00003 00868 116E-06 00003 01439
(133) (07) (347) (096) (142) (514)
Lag1 Top 5 Institutional Ownership -03774 -187512 -3848531 00179 -104261 -126659
(163) (186) (166) (011) (149) (436)
Lag1 Tobinrsquos Q 00731 -223886 10074 00642 -155467 3115073
(573) (1631) (045) (819) (2509) (75)
Lag1 Log Market Capitalization -0416 775637 7696302 -04015 764632 1462536
(2345) (10084) (3766) (3688) (13708) (4072)
CEO Dummy 15854 3211991 2950158 19152 3203152 4519385
(2223) (9381) (2254) (3804) (133) (2752)
Overall R-Sq -- 04967 01202 -- 04722 01168
Number of Observations 41305 41305 41305 85518 85518 85518
34
Table X Specification Tests utilizing Box-Cox Transformation of Market Capitalization
This table presents results from a Box-Cox transformation of market capitalization In
each model we specify transformation of market capitalization only Panel A shows
results for the period 1992‒1997 and Panel B shows results for the entire period
1992‒2002 The first part of each panel shows the Lambda coefficient that applies to
the transformed market capitalization The second part of each panel shows log
likelihood scores with LR Chi2 Statistics in parentheses ‒ these test for the
appropriateness of a linear or log model specification
Panel A Our ExecuComp sample 1992‒1997
(1) (2) (3)
PPSOG Salary TC
Lambda -05143 01881 02876
(-1914) (2413) (1329)
Test H0 (Log Likelihood Score)
Lambda = -1 -1279748 -3178575 -4795326
(23866) (85952) (99688)
Lambda = 0 -1280152 -3138456 -4791159
(31934) (57147) (16339)
Lambda = 1 -1283157 -3163807 -4793506
(92039) (564156) (63292)
Panel B Full ExecuComp sample 1992‒2002
(1) (2) (4)
PPSOG Salary TC
Lambda -044 01853 04036
(-3015) (40) (3666)
Test H0 (Log Likelihood Score)
Lambda = 0 -2688334 -6476494 -10384490
(66599) (160356) (126236)
Lambda = 1 -2694886 -6538506 -10386795
(197644) (1400605) (172345)
35
Figure 2 Institutional Ownership Measures (LHA) vs Total
Compensation (RHA) by Market Capitalization
0
10
20
30
40
50
60
70
1 2 3 4 5 6 7 8 9 10
$00
$10
$20
$30
$40
$50
$60
$70
$80
$90
Millio
ns
InstitutionalConcentration (LHS)Top5 Ownership (LHS)
Institutional TotalOwnership (LHS)Total Compensation(RHS)
Page 18
18
shareholder wealth positively and significantly relates to a change in executive cash
compensation In model (2) we include the contemporaneous change in shareholder
wealth and the lagged institutional concentration as stand-alone controls as indicated by
Greenersquos (2000) specification of equation (1) above and the interaction term remains
significant In model (3) we interact lagged institutional concentration with the lagged
change in shareholder wealth as in equation (2) above that tries to take account of what
participants could have observed at the time The interaction term is then negative but
insignificant altering the HS result Model (4) provides a similar result where our
preferred proxy Top 5 Ownership captures concentrated institutional influence In
models (5) to (8) we use our full sample In model (5) we find that the lagged
institutional concentration interaction with the contemporaneous change in shareholder
wealth is positive and significant in HSrsquos specification However including change in
contemporaneous shareholder wealth and lagged institutional concentration as separate
control variables as suggested by equation (1)rsquos specification removes the significance
from the interaction term in model (6) Moreover in model (7) when lagged institutional
concentration is interaction with the lagged change in shareholder wealth in accord with
equation (2) this interaction term is now negative and significant reversing the HS
finding We obtain a similar result in model (8) when we measure institutional
concentration influence using Top 5 Ownership13
V The Levels of Executive Fixed and Total Compensation as Functions of
Institutional Investor Influence
13 In unreported results we use the log of market capitalization as a size control These changes to the size
control do not dramatically change the findings from Table V The reason why this size issue is not as
important as in the findings for option grants pay-for-performance is that the change in cash compensation
is not so strongly influenced by firm size In unreported results available on request we also examine the
effect of institutional influence on changes in total compensation with similar results to the change in cash
compensation
19
Two measures of compensation are examined ndash salary and total compensation
ltltINSERT TABLE VI gtgt
Table VI examines the effect of institutional influence on the level of salary In model
(1) we use the HS sample and specification to obtain almost identical results The
coefficient on institutional concentration is (negative) $247509 that is almost identical to
(negative) $247967 in HS Both are highly significant In model (2) we make a single
change to the model ndash we use the natural logarithm of market capitalization This single
change reverses the sign of institutional concentration with the coefficient still being
significant at the 1 percent level In models (3) and (4) we measure institutional influence
using Top 5 Ownership In model (3) where we measure firm size using the level of
market capitalization Top 5 Ownership is insignificant When we use the natural
logarithm of market capitalization in model (4) Top 5 Ownership is significant and
positively relates to the salary level In models (5) to (8) we estimate the same models
using our full sample We find similar results for the full sample as for the period 1992ndash
1997
ltltINSERT TABLE VII gtgt
Table VII examines the effect of institutional influence on total compensation In
model (1) we replicate the HS methodology using the HS data to obtain a similar
significant negative relation between institutional concentration and total compensation of
(negative) $1608633 However after changing to a logarithmic specification of market
capitalization in model (2) institutional concentration increases total compensation by a
significant $32400014
Importantly when institutional influence is measured using Top 5
14 = (5688751 times 0101)
20
Ownership an insignificant relation exists with total compensation in model (3) using the
HS size control but it becomes both positive and significant in (4) with the logged size
control In models (5) and (6) where models are estimated using the full sample we
observe a similar reversal in sign as occurs in models (1) and (2) However now the
magnitude of the reversal is almost equal with a difference of about $67m that is very
large indeed In model (5) the coefficient on institutional concentration is (negative)
$342m and then in model (6) the coefficient is $33m Top 5 Ownership is insignificant
in both models (7) and (8) unlike the results for Salary
VI Robustness
Two issues are examined in is this Section ndash reverse causality and the appropriate
transformation of market capitalization A methodological issue inherent in examining the
effect of ownership-based measures of institutional influence on executive compensation
is endogeneity driven by reverse causality We utilize an instrumental variables approach
where orthogonalized institutional influence measures are regressed on each of the
compensation measures lagged option grant pay-for-performance sensitivity salary and
total compensation In addition market capitalization and a CEO dummy are included in
each regression We use the residual from each regression as an instrument for
institutional influence Such a process removes any preferences for compensation from
the institutional influence measures
ltltINSERT TABLES VIII AND IXgtgt
Table VIII presents the results where we instrument the HS measure lagged
institutional concentration with a compensation-orthogonalized equivalent The analysis
indicates that the instrumented institutional concentration measure is positively and
significantly related to the level of salary and total compensation and insignificantly
21
related to option grant pay-for-performance sensitivity (PPSOG) Table IX presents
results where we instrument Top 5 Ownership with a compensation-orthogonalized
equivalent The instrumented Top 5 Ownership measure insignificantly relates to option
grant pay-for-performance sensitivity and barely negatively relates to salary although it
negatively relates to total compensation suggesting reverse causality is driving a
significant portion of our previous results
ltltINSERT TABLE Xgtgt
Another issue examined in this section is the appropriate transformation of market
capitalization Aneuryn-Evans and Deaton (1980) provide an evaluation of methods
comparing linear and log-linear specifications We use the Box-Cox test to compare
logarithm versus linear specifications In applying the Box-Cox test we specify
transformation of only market capitalization ndash the dependent variable and all other
independent variables remain unchanged Table X presents the lambda coefficient for
each model and the log likelihood score In all models in Panel A and B a lambda of zero
receives a higher (closer to zero) log likelihood score suggesting the log specification is
appropriate
VII Conclusion
Hartzell and Starks (2003) appears to document an effective role for institutional
investors in increasing executive incentives and lowering compensation levels Our paper
re-addresses the issue of concentrated institutional monitoring from a perspective
informed by an additional decade of research since their paper appeared We cannot find
evidence that concentrated institutional investors really have any positive (or negative for
that matter) effect on executive incentive levels While superficially we find evidence
that they may raise executive pay levels this is most likely driven by an institutional
22
preference for highly remunerated executives in large companies and thus by reverse
causality Moreover our findings are not consistent with the rent-extraction hypothesis
suggesting that firms systematically overpay executives in the absence of effective
monitoring by outside parties such as institutional investors
23
References
Almazan Andres Jay C Hartzell and Laura T Starks 2005 ldquoActive Institutional
Shareholders and Costs of Monitoring Evidence from Executive Compensationrdquo
Financial Management Winter 5ndash34
Aneuryn-Evans Gwyn and Angus Deaton 1980 ldquoTesting Linear versus Logarithmic
Regression Modelsrdquo Review of Economic Studies 47 (1) Econometrics Issue 275ndash
291
Babenko Ilona 2009 ldquoShare Repurchases and Pay-Performance Sensitivity of Employee
Compensation Contractsrdquo Journal of Finance 64117-150
Baker George P and Brian J Hall 2004 ldquoCEO Incentives and Firm Sizerdquo Journal of
Labor Market Economics 22 767ndash798
Bebchuk Lucian A Jesse M Fried and David Walker 2002 ldquoManagerial power and
rent extraction in the design of executive compensationrdquo University of Chicago Law
Review 69 751ndash846
Bebchuk Lucian A and Jesse M Fried 2003 ldquoExecutive Compensation as an Agency
Problemrdquo Journal of Economic Perspectives 17 71ndash92
Bebchuk Lucian A and Jesse M Fried 2004 Pay without Performance The Unfulfilled
Promise of Executive Compensation Cambridge MA Harvard University Press
Bertrand Marianne and Sendhil Mullainathan 2001 ldquoAre CEOs Rewarded for Luck
The Ones Without Principals Arerdquo Quarterly Journal of Economics 116 901ndash32
Black Fischer and Myron Scholes 1973 ldquoThe Pricing of Options and Corporate
Liabilitiesrdquo Journal of Political Economy 81 637ndash54
Cadman Brian Sandy Klasa and Steve Matsunaga 2010 ldquoDeterminants of CEO Pay A
Comparison of ExecuComp and Non-ExecuComp Firmsrdquo The Accounting Review 85
1511ndash1543
Chhaochharia Vidhi and Yaniv Grinstein 2009 ldquoCEO Compensation and Board
Structurerdquo Journal of Finance 64 231ndash261
Cichello Michael S 2005 ldquoThe impact of firm size on pay-performance sensitivitiesrdquo
Journal of Corporate Finance 11 609ndash627
Davis Gerald F and E Han Kim 2007 ldquoBusiness ties and proxy voting by mutual
fundsrdquo Journal of Financial Economics 85 552ndash570
Dikolli Shane S Susan L Kulp and Karen L Sedatole 2009 ldquoTransient Institutional
Ownership and CEO Contractingrdquo Accounting Review 84 737ndash770
Edmans Alex Xavier Gabaix and Augustin Landier 2009 ldquoA Multiplicative Model of
Optimal CEO Incentives in Market Equilibriumrdquo Review of Financial Studies 22
4881ndash4917
Edmans Alex 2009 ldquoBlockholder Trading Market Efficiency and Managerial Myopiardquo
Journal of Finance 64 2481ndash2513
Ertimur Yonca Fabrizio Ferri and Volkan Muslu 2011 ldquoShareholder Activism and
CEO Payrdquo Review of Financial Studies 24 535ndash592
24
Gabaix Xavier and Augustin Landier 2008 ldquoWhy Has CEO Pay Increased So Muchrdquo
Quarterly Journal of Economics 123 49ndash100
Garen John 1994 ldquoExecutive compensation and principal-agent theoryrdquo Journal of
Political Economy 102 1175ndash1199
Gibbons Robert and Kevin J Murphy 1992 ldquoOptimal Incentive Contracts in the
Presence of Career Concerns Theory and Evidencerdquo Journal of Political Economy
100 468ndash505
Greene William H 2000 (4th
Edition) Econometric Analysis Prentice-Hall NJ
Guthrie Katherine Jan Sokolowsky and Kam-Ming Wan 2012 ldquoCEO Compensation
and Board Structure Revisitedrdquo Journal of Finance Forthcoming
Hadlock Charles R and Gerald B Lumer 1997 ldquoCompensation turnover and top
management incentives historical evidencerdquo Journal of Business 70 153ndash187
Hartzell Jay C and Laura T Starks 2003 ldquoInstitutional Investors and Executive
Compensationrdquo Journal of Finance 58 2351ndash2374
Huson Mark R Robert Parrino and Laura T Starks 2001 ldquoInternal Monitoring
Mechanisms and CEO Turnover A Long-Term Perspectiverdquo Journal of Finance 56
2265ndash2297
Jensen Michael C and Kevin J Murphy 1990 ldquoPerformance Pay and Top Management
Incentivesrdquo Journal of Political Economy 98 225ndash264
Jin Li 2002 ldquoCEO compensation diversification and incentivesrdquo Journal of Financial
Economics 66 29ndash63
Kang Qiang and Qiao Liu 2008 ldquoStock Market Information Production and Executive
Incentivesrdquo Journal of Corporate Finance 14 484ndash498
Merton Robert 1973 ldquoTheory of Rational Option Pricingrdquo Bell Journal of Economics
and Management Science 4 141-183
Morse Adair Vikram Nanda and Amit Seru 2011 ldquoAre Incentive Contracts Rigged by
Powerful CEOsrdquo Journal of Finance 66 1779ndash1821
Murphy Kevin J 1985 ldquoCorporate performance and managerial remuneration An
empirical analysisrdquo Journal of Accounting and Economics 7 11ndash42
Murphy Kevin J 1999 ldquoExecutive compensationrdquo in Orley Ashenfelter and David
Card Handbook of Labor Economics vol 3 Elsevier
Rosen Sherwin 1992 ldquoContracts and the market for executivesrdquo In Contract
Economics ed Lars Werin and Hans Wijkander 181ndash211 Cambridge MA
Blackwell
Rothberg Burton and Steven Lilien 2006 ldquoMutual Funds and Proxy Voting New
Evidence on Corporate Governancerdquo Journal of Business and Technology Law 1
157ndash184
Schaefer Scott 1998 ldquoThe dependence of payndashperformance sensitivity on the size of the
firmrdquo Review of Economics and Statistics 80 436ndash443
Yermack David 1995 ldquoDo Corporations Award CEO Stock Options Effectivelyrdquo
Journal of Financial Economics 39 237ndash69
25
Table I Univariate Descriptive Statistics
This table reports sample statistics for the key variables used in this study based on our comprehensive
ExecuComp dataset Institutional Concentration is the shares owned by the institutions with the five largest
holdings divided by the shares held by all institutions Top 5 Ownership is the proportion of shares
outstanding held by the institutions with the five largest holdings Total Institutional Ownership is the
proportion of shares held by all institutions Panel B shows the measures executive compensation The pay-
for-performance sensitivity (PPS) of option grants shows the change in executive wealth due to a $1000
change in firm in shareholder wealth Expressions for Salary and Total Compensation are in thousands of
dollars while Market Capitalization is in millions of dollars Panel C shows firm characteristics
Variable Mean Median Std Dev P10 P90
Panel A Institutional Investor Holdings
Institutional Concentration 0450 0424 0151 0282 0656
Top 5 Institutional Ownership 0229 0220 0101 0110 0355
Total Institutional Ownership 0546 0561 0206 0258 0801
Panel B Executive Compensation
PPS of Option Grants 1161 0156 3829 0 2822
Salary 36311 29462 25231 15179 67000
Total Compensation 258313 85878 980701 23875 511022
Panel C Firm Characteristics
Tobins Q 1744 1139 2738 0495 3322
Market Capitalization 560699 112094 1988108 19754 1048754
26
Table II Pairwise Correlation Coefficients
This table presents the correlations between the key variables based on our comprehensive
ExecuComp dataset Institutional Concentration is the shares owned by the institutions with the five
largest holdings divided by the shares held by all institutions Top 5 Ownership is the proportion of
shares outstanding held by the institutions with the five largest holdings Total Institutional
Ownership is the proportion of shares held by all institutions The pay-for-performance sensitivity
(PPS) of option grants is the change in executive wealth due to a $1000 change in shareholder
wealth
(1) (2) (3) (4) (5) (6) (7) (8)
(1) Institutional Concentration 1
(2) Top 5 Ownership 0268 1
(3) Total Ownership -0415 0706 1
(4) PPS of Option Grants 0122 0042 -0024 1
(5) Log Salary -0270 -0012 0171 0038 1
(6) Log Market Capitalization -0630 -0157 0261 -0156 0420 1
(7) Level Market Capitalization -0241 -0157 0007 -0058 0219 0534 1
(8) Tobins Q -0008 -0034 0002 0015 -0083 0066 0121 1
27
Table III
Descriptive Statistics by Market Capitalization Deciles
Market
Capitalization
Deciles
Market
Capitalization
($m)
Total
Compensation
($000)
Salary
($)
PPS Option
Grants
($)
Institutional
Concentration
()
Top 5 Ownership
()
Institutional
Total Ownership
()
1 (Small) 1181 76008 234700 2720 611 200 327
2 2668 90206 257955 1796 495 217 438
3 4193 105946 275120 1517 454 220 485
4 6045 129892 296067 1376 420 206 491
5 8742 155391 315864 1216 394 198 502
6 13086 191888 345942 1072 380 202 531
7 20790 239501 377021 0874 368 196 532
8 34920 314955 427808 0687 354 190 538
9 67747 450970 480319 0569 328 173 528
10 (Big) 366540 846066 614354 0297 275 138 502
28
Table IV Pay-for-Performance Sensitivity of Option Grants as a Function of Institutional Influence
Similar in construction to parts of HS (Table II) this table shows the coefficients from Tobit regressions of option grant pay-for-performance sensitivity per $1000 change in
shareholder wealth against the lagged change and contemporary change in shareholder wealth measures of institutional influence Tobinrsquos Q firm market capitalization and
dummy controls for CEO and year effects We estimate models (1) to (4) over the period 1992‒1997 using data supplied by HS and models (5) to (8) are estimated over the
whole sample Models (1) (2) (5) and (6) use Institutional Concentration as a measure of institutional influence which HS define as the shares owned by the institutions
with the five largest holdings divided by the shares held by all institutions These models control for total institutional ownership Models (3) (4) (7) and (8) use Top 5
Ownership as a measure of institutional influence which we define as the proportion of shares outstanding held by the institutions with the five largest holdings Coefficients
for industry and year controls are not presented One two and three asterisks denote significance at 10 5 and 1 levels
Hartzell and Starks Database 1992‒1997 Full Sample from ExecuComp 1992‒2002
(1) (2) (3) (4) (5) (6) (7) (8)
Change in Shareholder Wealth 77721 191781 89874 137368 -446 -218 -664 -203
(171) (429) (194) (317) (319) (168) (453) (158)
Lag1 Change in Shareholder Wealth 14446 99963 88719 8544 -144 0273 -0342 00852
(027) (212) (166) (184) (071) (015) (016) (005)
Lag1 Institutional Concentration 13994 -00692 26405 03304
(767) (032) (1402) (166)
Lag1 Total Institutional Ownership 08055 10434 07106 0937
(593) (781) (666) (898)
Lag1 Top 5 Institutional Ownership 10567 07457 12116 06293
(399) ( 283) (533) (278)
Lag1 Tobinrsquos Q 0068 00877 00647 00813 00498 00727 00508 00695
(337) (434) (319) (401) (504) (734) (51) (703)
Lag1 Market Capitalization -106202 -173515 -776 -117
(474) (742) (1011) (153)
Lag1 Log Market Capitalization -0297 -02319 -04111 -03908
(1576) (1439) (2997) (2841)
CEO Dummy 13745 13682 13765 13674 28197 28128 28155 28143
(1913) (1918) (1914) (1916) (3388) (3396) (3377) (34)
Number of Observations 33878 33878 33878 33878 97679 97679 97679 97679
29
Table V Pay-for-Performance Sensitivity of Cash Compensation (Salary plus Bonus) as a Function of Institutional Influence - Level Market Capitalization
Similar in parts to models (1) and (2) of HS (Table IV) this table displays the coefficients from OLS regressions of the change in cash compensation against the lagged
change and contemporary change in shareholder wealth measures of institutional influence Tobinrsquos Q firm market capitalization and dummy controls for CEO and year
effects We estimate models (1) to (4) over the period 1992‒1997 based on our ExecuComp sample and models (5) to (8) over the whole sample Models (1) (2) (5) and (6)
use Institutional Concentration as a measure of institutional influence which HS define as the shares owned by the institutions with the five largest holdings divided by the
shares held by all institutions These models control for total institutional ownership Models (4) and (8) use Top 5 Ownership as a measure of institutional influence which
we define as the proportion of shares outstanding held by the institutions with the 5 largest holdings Coefficients for industry and year controls are not presented One two
and three asterisks denote significance at 10 5 and 1 levels
Our Database based on ExecuComp 1992‒1997 Full Sample based on ExecuComp 1992‒2002
(1) (2) (3) (4) (5) (6) (7) (8)
Lag1 Change in Shareholder Wealth 00014 00017 00088 00089 -00006 -00007 00072 00086
(051) (062) (129) (103) (027) (029) (131) (16)
Change in Shareholder Wealth -00102 00102 00105 00032 00102 00101
(097) (292) (293) (055) (611) (608)
Change Shareholder Wealth Lag1 Institutional
Concentration
00435 00764 00363 00253
(463) (258) (689) (144)
Lag1 Change Shareholder Wealth Lag1
Institutional Concentration
-00197 -00283
(122) (197)
Lag1 Institutional Concentration -492182 -288861 400149 453623
(09) (056) (098) (114)
Lag1 Total Institutional Ownership 720688 538372 655091 559587 682069 726226
(276) (232) (281) (314) (308) (322)
Lag1 Change Shareholder Wealth Lag1 Top 5
Institutional Ownership
-0046 -00689
(091) (256)
Lag1 Top 5 Institutional Ownership 583044 1191634
(092) (25)
Lag1 Tobinrsquos Q -32761 -36806 -19422 -18023 -54283 -48681 -29534 -24243
(138) (172) (077) (078) (227) (245) (136) (111)
Lag1 Market Capitalization 00004 00004 00002 00006 00007 00008 00007 00007
(042) (044) (022) (07) (086) (092) (082) (09)
CEO Dummy 768035 765474 768334 773634 616315 61729 619028 618535
(354) (355) (356) (355) (417) (418) (419) (418)
Adjusted R-Sq 00057 0006 00052 0005 00097 00097 00098 00102
Number of Observations 41305 41305 41305 41305 85518 85518 85518 85518
30
Table VI Level of Salary as a Function of Institutional Influence
Similar in parts to models (1) and (2) of HS (Table V) this table displays the coefficients from OLS regressions of salary against the lagged change and contemporary change in
shareholder wealth measures of institutional influence Tobinrsquos Q firm market capitalization and dummy controls for CEO industry and year effects We estimate models (1) to
(4) over the period 1992‒1997 using the actual data supplied by HS and models (5) to (8) over the whole sample 1992‒2002 Models (1) (2) (5) and (6) use Institutional
Concentration as a measure of institutional influence which HS define as the shares owned by the institutions with the five largest holdings divided by the shares held by all
institutions These models control for total institutional ownership Models (3) (4) (7) and (8) use Top 5 Ownership as a measure of institutional influence which we define as
the proportion of shares outstanding held by the institutions with the 5 largest holdings Coefficients for industry and year controls are not presented One two and three
asterisks denote significance at 10 5 and 1 levels
Hartzell and Starks Database 1992‒1997 Full Sample based on ExecuComp 1992‒2002
(1) (2) (3) (4) (5) (6) (7) (8)
Change in Shareholder Wealth 04125E-4 0003 -2439E-4 0003 00021 00012 00024 00012
(057) (464) (031) (468) (1004) (72) (1041) (706)
Lag1 Change in Shareholder Wealth -00062 7856E-4 -0007 9003E-4 -00014 00002 -00014 00003
(594) (098) (592) (112) (522) (123) (5) (145)
Lag1 Institutional Concentration -2475087 555134 -2734983 943317
(3245) (710) (4965) (1671)
Lag1 Total Institutional Ownership 1020109 217488 936337 252464
(1904) (432) (2471) (704)
Lag1 Top 5 Institutional Ownership -148355 31354 -87926 423166
(139) (332) (118) (643)
Lag1 Tobinrsquos Q -195348 -252191 -192344 -25158 -119837 -146373 -122132 -143821
(2812) (3762) (2721) (3752) (248) (274) (2583) (2727)
Lag1 Market Capitalization 844E-06 103E-5 00034 00039
(2781) (3051) (2874) (3104)
Lag1 Log Market Capitalization 750021 723625 776657 732279
(7511) (9285) (11224) (14039)
CEO Dummy 2683978 2702238 2668244 2702417 3281133 3285103 3292175 3283145
(8625) (9277) (8246) (9271) (12739) (13616) (12336) (13605)
Adjusted R-Sq 04604 05302 04187 05295 03856 04588 03476 04574
Number of Observations 36280 36280 36280 36280 97679 97679 97679 97679
31
Table VII Level of Total Direct Compensation as a Function of Institutional Influence
Similar in parts to models (3) and (5) of HS (Table V) this table displays the coefficients from OLS regressions of total direct compensation against the lagged change and
contemporary change in shareholder wealth measures of institutional influence Tobinrsquos Q firm market capitalization and dummy controls for CEO industry and year effects
We estimate models (1) to (4) over the period 1992‒1997 and models (5) to (8) over the whole sample Models (1) (2) (5) and (6) use Institutional Concentration as a measure
of institutional influence which HS define as the shares owned by the institutions with the five largest holdings divided by the shares held by all institutions These models
control for total institutional ownership Models (3) (4) (7) and (8) use Top 5 Ownership as a measure of institutional influence which we define as the proportion of shares
outstanding held by the institutions with the five largest holdings Coefficients for industry and year controls are not presented One two and three asterisks denote significance
at 10 5 and 1 levels
Hartzell and Starks Database 1992‒1997 Full Sample from ExecuComp 1992‒2002
(1) (2) (3) (4) (5) (6) (7) (8)
Change in Shareholder Wealth 005779 007006 005282 006791 00423 00148 0046 00138
(415) (581) (369) (565) (203) (07) (22) (066)
Lag1 Change in Shareholder Wealth 003182 00714 00272 00715 00613 0126 00614 01278
(126) (298) (107) (299) (275) (488) (276) (494)
Lag1 Institutional Concentration -1608633 3248998 -3424892 3334186
(151) (283) (1172) (118)
Lag1 Total Institutional Ownership 104013 5473128 1531013 -2926764
(1645) (831) (876) (158)
Lag1 Top 5 Institutional Ownership 188585 3927986 -2159177 -2640879
(016) (343) (067) (092)
Lag1 Tobinrsquos Q -445692 -803612 -454859 -833566 3257027 2971201 3213936 3086205
(427) (791) (432) (831) (83) (77) (815) (793)
Lag1 Market Capitalization 49E-5 615E-5 01031 01088
(113) (1365) (1555) (168)
Lag1 Log Market Capitalization 4683497 4759377 1645156 1427856
(2775) (3469) (3679) (4239)
CEO Dummy 1501307 1508886 1493688 1508736 4713079 4713629 47278 4703908
(2648) (2684) (2618) (2683) (282) (2828) (2819) (2824)
Adjusted R-Sq 01861 02060 01695 02051 01071 01113 01033 011
Number of Observations 33229 33229 33229 33229 97679 97679 97679 97679
32
Table VIII Two Stage Least Squares (Reverse Causality) Tests - Executive Compensation Measures as a Function of Institutional Concentration
This table shows the coefficients from the Two-Stage Least Squares (2SLS) regressions of executive compensation measures against the lagged change
and contemporary change in shareholder wealth instrumented institutional concentration and total institutional ownership Tobinrsquos Q the natural
logarithm of firm market capitalization and dummy variable controls for CEO and year effects We instrument measures of institutional concentration
and total institutional ownership by orthogonalizing with respect to option grant pay-for-performance sensitivity salary and total compensation
Coefficients for industry and year controls are not presented First stage results are not presented One two and three asterisks denote significance at
10 5 and 1 levels
ExecuComp Sample 1992‒1997 Full Sample from ExecuComp 1992‒2002
PPSOG Salary TC PPSOG Salary TC
Change in Shareholder Wealth 144E-05 0004 02133 -409E-06 00011 00157
(308) (775) (596) (691) (638) (068)
Lag1 Change in Shareholder Wealth -240E-05 -00004 00866 118E-06 00002 01422
(131) (087) (346) (097) (116) (51)
Lag1 Institutional Concentration -00898 668681 3263143 02024 916656 2856417
(052) (801) (176) (163) (1545) (1004)
Lag1 Total Institutional Ownership 01021 -43997 2512587 02983 -104171 -1028347
(09) (084) (153) (4) (269) (487)
Lag1 Tobinrsquos Q 00754 -225606 119297 00655 -158657 2997205
(596) (1649) (053) (834) (2535) (727)
Lag1 Log Market Capitalization -04234 828713 7864046 -04002 825106 1685638
(2195) (8132) (2977) (4106) (11133) (373)
CEO Dummy 15848 3214117 2950274 19149 3205862 4529101
(2223) (9389) (2255) (3802) (1331) (2758)
Overall R-Sq -- 04981 01206 -- 04738 01178
Number of Observations 41305 41305 41305 85518 85518 85518
33
Table IX Two Stage Least Squares (Reverse Causality) Tests - Executive Compensation Measures as a Function of Top 5 Ownership
This table shows the coefficients from 2SLS regressions of executive compensation measures against the lagged change and contemporary change
in shareholder wealth instrumented Top 5 institutional ownership Tobinrsquos Q natural logarithm of firm market capitalization and dummy variable
controls for CEO and year effects We instrument Top 5 institutional ownership by orthogonalizing with respect to option grant pay-for-
performance sensitivity salary and total compensation Coefficients for industry and year controls are not presented First stage results are not
presented One two and three asterisks denote significance at 10 5 and 1 levels
ExecuComp sample 1992‒1997 Full Sample 1992‒2002
PPSOG Salary TC PPSOG Salary TC
Change in Shareholder Wealth 136E-05 00041 02128 -407E-06 00011 00146
(293) (79) (596) (692) (619) (063)
Lag1 Change in Shareholder Wealth -241E-05 -00003 00868 116E-06 00003 01439
(133) (07) (347) (096) (142) (514)
Lag1 Top 5 Institutional Ownership -03774 -187512 -3848531 00179 -104261 -126659
(163) (186) (166) (011) (149) (436)
Lag1 Tobinrsquos Q 00731 -223886 10074 00642 -155467 3115073
(573) (1631) (045) (819) (2509) (75)
Lag1 Log Market Capitalization -0416 775637 7696302 -04015 764632 1462536
(2345) (10084) (3766) (3688) (13708) (4072)
CEO Dummy 15854 3211991 2950158 19152 3203152 4519385
(2223) (9381) (2254) (3804) (133) (2752)
Overall R-Sq -- 04967 01202 -- 04722 01168
Number of Observations 41305 41305 41305 85518 85518 85518
34
Table X Specification Tests utilizing Box-Cox Transformation of Market Capitalization
This table presents results from a Box-Cox transformation of market capitalization In
each model we specify transformation of market capitalization only Panel A shows
results for the period 1992‒1997 and Panel B shows results for the entire period
1992‒2002 The first part of each panel shows the Lambda coefficient that applies to
the transformed market capitalization The second part of each panel shows log
likelihood scores with LR Chi2 Statistics in parentheses ‒ these test for the
appropriateness of a linear or log model specification
Panel A Our ExecuComp sample 1992‒1997
(1) (2) (3)
PPSOG Salary TC
Lambda -05143 01881 02876
(-1914) (2413) (1329)
Test H0 (Log Likelihood Score)
Lambda = -1 -1279748 -3178575 -4795326
(23866) (85952) (99688)
Lambda = 0 -1280152 -3138456 -4791159
(31934) (57147) (16339)
Lambda = 1 -1283157 -3163807 -4793506
(92039) (564156) (63292)
Panel B Full ExecuComp sample 1992‒2002
(1) (2) (4)
PPSOG Salary TC
Lambda -044 01853 04036
(-3015) (40) (3666)
Test H0 (Log Likelihood Score)
Lambda = 0 -2688334 -6476494 -10384490
(66599) (160356) (126236)
Lambda = 1 -2694886 -6538506 -10386795
(197644) (1400605) (172345)
35
Figure 2 Institutional Ownership Measures (LHA) vs Total
Compensation (RHA) by Market Capitalization
0
10
20
30
40
50
60
70
1 2 3 4 5 6 7 8 9 10
$00
$10
$20
$30
$40
$50
$60
$70
$80
$90
Millio
ns
InstitutionalConcentration (LHS)Top5 Ownership (LHS)
Institutional TotalOwnership (LHS)Total Compensation(RHS)
Page 19
19
Two measures of compensation are examined ndash salary and total compensation
ltltINSERT TABLE VI gtgt
Table VI examines the effect of institutional influence on the level of salary In model
(1) we use the HS sample and specification to obtain almost identical results The
coefficient on institutional concentration is (negative) $247509 that is almost identical to
(negative) $247967 in HS Both are highly significant In model (2) we make a single
change to the model ndash we use the natural logarithm of market capitalization This single
change reverses the sign of institutional concentration with the coefficient still being
significant at the 1 percent level In models (3) and (4) we measure institutional influence
using Top 5 Ownership In model (3) where we measure firm size using the level of
market capitalization Top 5 Ownership is insignificant When we use the natural
logarithm of market capitalization in model (4) Top 5 Ownership is significant and
positively relates to the salary level In models (5) to (8) we estimate the same models
using our full sample We find similar results for the full sample as for the period 1992ndash
1997
ltltINSERT TABLE VII gtgt
Table VII examines the effect of institutional influence on total compensation In
model (1) we replicate the HS methodology using the HS data to obtain a similar
significant negative relation between institutional concentration and total compensation of
(negative) $1608633 However after changing to a logarithmic specification of market
capitalization in model (2) institutional concentration increases total compensation by a
significant $32400014
Importantly when institutional influence is measured using Top 5
14 = (5688751 times 0101)
20
Ownership an insignificant relation exists with total compensation in model (3) using the
HS size control but it becomes both positive and significant in (4) with the logged size
control In models (5) and (6) where models are estimated using the full sample we
observe a similar reversal in sign as occurs in models (1) and (2) However now the
magnitude of the reversal is almost equal with a difference of about $67m that is very
large indeed In model (5) the coefficient on institutional concentration is (negative)
$342m and then in model (6) the coefficient is $33m Top 5 Ownership is insignificant
in both models (7) and (8) unlike the results for Salary
VI Robustness
Two issues are examined in is this Section ndash reverse causality and the appropriate
transformation of market capitalization A methodological issue inherent in examining the
effect of ownership-based measures of institutional influence on executive compensation
is endogeneity driven by reverse causality We utilize an instrumental variables approach
where orthogonalized institutional influence measures are regressed on each of the
compensation measures lagged option grant pay-for-performance sensitivity salary and
total compensation In addition market capitalization and a CEO dummy are included in
each regression We use the residual from each regression as an instrument for
institutional influence Such a process removes any preferences for compensation from
the institutional influence measures
ltltINSERT TABLES VIII AND IXgtgt
Table VIII presents the results where we instrument the HS measure lagged
institutional concentration with a compensation-orthogonalized equivalent The analysis
indicates that the instrumented institutional concentration measure is positively and
significantly related to the level of salary and total compensation and insignificantly
21
related to option grant pay-for-performance sensitivity (PPSOG) Table IX presents
results where we instrument Top 5 Ownership with a compensation-orthogonalized
equivalent The instrumented Top 5 Ownership measure insignificantly relates to option
grant pay-for-performance sensitivity and barely negatively relates to salary although it
negatively relates to total compensation suggesting reverse causality is driving a
significant portion of our previous results
ltltINSERT TABLE Xgtgt
Another issue examined in this section is the appropriate transformation of market
capitalization Aneuryn-Evans and Deaton (1980) provide an evaluation of methods
comparing linear and log-linear specifications We use the Box-Cox test to compare
logarithm versus linear specifications In applying the Box-Cox test we specify
transformation of only market capitalization ndash the dependent variable and all other
independent variables remain unchanged Table X presents the lambda coefficient for
each model and the log likelihood score In all models in Panel A and B a lambda of zero
receives a higher (closer to zero) log likelihood score suggesting the log specification is
appropriate
VII Conclusion
Hartzell and Starks (2003) appears to document an effective role for institutional
investors in increasing executive incentives and lowering compensation levels Our paper
re-addresses the issue of concentrated institutional monitoring from a perspective
informed by an additional decade of research since their paper appeared We cannot find
evidence that concentrated institutional investors really have any positive (or negative for
that matter) effect on executive incentive levels While superficially we find evidence
that they may raise executive pay levels this is most likely driven by an institutional
22
preference for highly remunerated executives in large companies and thus by reverse
causality Moreover our findings are not consistent with the rent-extraction hypothesis
suggesting that firms systematically overpay executives in the absence of effective
monitoring by outside parties such as institutional investors
23
References
Almazan Andres Jay C Hartzell and Laura T Starks 2005 ldquoActive Institutional
Shareholders and Costs of Monitoring Evidence from Executive Compensationrdquo
Financial Management Winter 5ndash34
Aneuryn-Evans Gwyn and Angus Deaton 1980 ldquoTesting Linear versus Logarithmic
Regression Modelsrdquo Review of Economic Studies 47 (1) Econometrics Issue 275ndash
291
Babenko Ilona 2009 ldquoShare Repurchases and Pay-Performance Sensitivity of Employee
Compensation Contractsrdquo Journal of Finance 64117-150
Baker George P and Brian J Hall 2004 ldquoCEO Incentives and Firm Sizerdquo Journal of
Labor Market Economics 22 767ndash798
Bebchuk Lucian A Jesse M Fried and David Walker 2002 ldquoManagerial power and
rent extraction in the design of executive compensationrdquo University of Chicago Law
Review 69 751ndash846
Bebchuk Lucian A and Jesse M Fried 2003 ldquoExecutive Compensation as an Agency
Problemrdquo Journal of Economic Perspectives 17 71ndash92
Bebchuk Lucian A and Jesse M Fried 2004 Pay without Performance The Unfulfilled
Promise of Executive Compensation Cambridge MA Harvard University Press
Bertrand Marianne and Sendhil Mullainathan 2001 ldquoAre CEOs Rewarded for Luck
The Ones Without Principals Arerdquo Quarterly Journal of Economics 116 901ndash32
Black Fischer and Myron Scholes 1973 ldquoThe Pricing of Options and Corporate
Liabilitiesrdquo Journal of Political Economy 81 637ndash54
Cadman Brian Sandy Klasa and Steve Matsunaga 2010 ldquoDeterminants of CEO Pay A
Comparison of ExecuComp and Non-ExecuComp Firmsrdquo The Accounting Review 85
1511ndash1543
Chhaochharia Vidhi and Yaniv Grinstein 2009 ldquoCEO Compensation and Board
Structurerdquo Journal of Finance 64 231ndash261
Cichello Michael S 2005 ldquoThe impact of firm size on pay-performance sensitivitiesrdquo
Journal of Corporate Finance 11 609ndash627
Davis Gerald F and E Han Kim 2007 ldquoBusiness ties and proxy voting by mutual
fundsrdquo Journal of Financial Economics 85 552ndash570
Dikolli Shane S Susan L Kulp and Karen L Sedatole 2009 ldquoTransient Institutional
Ownership and CEO Contractingrdquo Accounting Review 84 737ndash770
Edmans Alex Xavier Gabaix and Augustin Landier 2009 ldquoA Multiplicative Model of
Optimal CEO Incentives in Market Equilibriumrdquo Review of Financial Studies 22
4881ndash4917
Edmans Alex 2009 ldquoBlockholder Trading Market Efficiency and Managerial Myopiardquo
Journal of Finance 64 2481ndash2513
Ertimur Yonca Fabrizio Ferri and Volkan Muslu 2011 ldquoShareholder Activism and
CEO Payrdquo Review of Financial Studies 24 535ndash592
24
Gabaix Xavier and Augustin Landier 2008 ldquoWhy Has CEO Pay Increased So Muchrdquo
Quarterly Journal of Economics 123 49ndash100
Garen John 1994 ldquoExecutive compensation and principal-agent theoryrdquo Journal of
Political Economy 102 1175ndash1199
Gibbons Robert and Kevin J Murphy 1992 ldquoOptimal Incentive Contracts in the
Presence of Career Concerns Theory and Evidencerdquo Journal of Political Economy
100 468ndash505
Greene William H 2000 (4th
Edition) Econometric Analysis Prentice-Hall NJ
Guthrie Katherine Jan Sokolowsky and Kam-Ming Wan 2012 ldquoCEO Compensation
and Board Structure Revisitedrdquo Journal of Finance Forthcoming
Hadlock Charles R and Gerald B Lumer 1997 ldquoCompensation turnover and top
management incentives historical evidencerdquo Journal of Business 70 153ndash187
Hartzell Jay C and Laura T Starks 2003 ldquoInstitutional Investors and Executive
Compensationrdquo Journal of Finance 58 2351ndash2374
Huson Mark R Robert Parrino and Laura T Starks 2001 ldquoInternal Monitoring
Mechanisms and CEO Turnover A Long-Term Perspectiverdquo Journal of Finance 56
2265ndash2297
Jensen Michael C and Kevin J Murphy 1990 ldquoPerformance Pay and Top Management
Incentivesrdquo Journal of Political Economy 98 225ndash264
Jin Li 2002 ldquoCEO compensation diversification and incentivesrdquo Journal of Financial
Economics 66 29ndash63
Kang Qiang and Qiao Liu 2008 ldquoStock Market Information Production and Executive
Incentivesrdquo Journal of Corporate Finance 14 484ndash498
Merton Robert 1973 ldquoTheory of Rational Option Pricingrdquo Bell Journal of Economics
and Management Science 4 141-183
Morse Adair Vikram Nanda and Amit Seru 2011 ldquoAre Incentive Contracts Rigged by
Powerful CEOsrdquo Journal of Finance 66 1779ndash1821
Murphy Kevin J 1985 ldquoCorporate performance and managerial remuneration An
empirical analysisrdquo Journal of Accounting and Economics 7 11ndash42
Murphy Kevin J 1999 ldquoExecutive compensationrdquo in Orley Ashenfelter and David
Card Handbook of Labor Economics vol 3 Elsevier
Rosen Sherwin 1992 ldquoContracts and the market for executivesrdquo In Contract
Economics ed Lars Werin and Hans Wijkander 181ndash211 Cambridge MA
Blackwell
Rothberg Burton and Steven Lilien 2006 ldquoMutual Funds and Proxy Voting New
Evidence on Corporate Governancerdquo Journal of Business and Technology Law 1
157ndash184
Schaefer Scott 1998 ldquoThe dependence of payndashperformance sensitivity on the size of the
firmrdquo Review of Economics and Statistics 80 436ndash443
Yermack David 1995 ldquoDo Corporations Award CEO Stock Options Effectivelyrdquo
Journal of Financial Economics 39 237ndash69
25
Table I Univariate Descriptive Statistics
This table reports sample statistics for the key variables used in this study based on our comprehensive
ExecuComp dataset Institutional Concentration is the shares owned by the institutions with the five largest
holdings divided by the shares held by all institutions Top 5 Ownership is the proportion of shares
outstanding held by the institutions with the five largest holdings Total Institutional Ownership is the
proportion of shares held by all institutions Panel B shows the measures executive compensation The pay-
for-performance sensitivity (PPS) of option grants shows the change in executive wealth due to a $1000
change in firm in shareholder wealth Expressions for Salary and Total Compensation are in thousands of
dollars while Market Capitalization is in millions of dollars Panel C shows firm characteristics
Variable Mean Median Std Dev P10 P90
Panel A Institutional Investor Holdings
Institutional Concentration 0450 0424 0151 0282 0656
Top 5 Institutional Ownership 0229 0220 0101 0110 0355
Total Institutional Ownership 0546 0561 0206 0258 0801
Panel B Executive Compensation
PPS of Option Grants 1161 0156 3829 0 2822
Salary 36311 29462 25231 15179 67000
Total Compensation 258313 85878 980701 23875 511022
Panel C Firm Characteristics
Tobins Q 1744 1139 2738 0495 3322
Market Capitalization 560699 112094 1988108 19754 1048754
26
Table II Pairwise Correlation Coefficients
This table presents the correlations between the key variables based on our comprehensive
ExecuComp dataset Institutional Concentration is the shares owned by the institutions with the five
largest holdings divided by the shares held by all institutions Top 5 Ownership is the proportion of
shares outstanding held by the institutions with the five largest holdings Total Institutional
Ownership is the proportion of shares held by all institutions The pay-for-performance sensitivity
(PPS) of option grants is the change in executive wealth due to a $1000 change in shareholder
wealth
(1) (2) (3) (4) (5) (6) (7) (8)
(1) Institutional Concentration 1
(2) Top 5 Ownership 0268 1
(3) Total Ownership -0415 0706 1
(4) PPS of Option Grants 0122 0042 -0024 1
(5) Log Salary -0270 -0012 0171 0038 1
(6) Log Market Capitalization -0630 -0157 0261 -0156 0420 1
(7) Level Market Capitalization -0241 -0157 0007 -0058 0219 0534 1
(8) Tobins Q -0008 -0034 0002 0015 -0083 0066 0121 1
27
Table III
Descriptive Statistics by Market Capitalization Deciles
Market
Capitalization
Deciles
Market
Capitalization
($m)
Total
Compensation
($000)
Salary
($)
PPS Option
Grants
($)
Institutional
Concentration
()
Top 5 Ownership
()
Institutional
Total Ownership
()
1 (Small) 1181 76008 234700 2720 611 200 327
2 2668 90206 257955 1796 495 217 438
3 4193 105946 275120 1517 454 220 485
4 6045 129892 296067 1376 420 206 491
5 8742 155391 315864 1216 394 198 502
6 13086 191888 345942 1072 380 202 531
7 20790 239501 377021 0874 368 196 532
8 34920 314955 427808 0687 354 190 538
9 67747 450970 480319 0569 328 173 528
10 (Big) 366540 846066 614354 0297 275 138 502
28
Table IV Pay-for-Performance Sensitivity of Option Grants as a Function of Institutional Influence
Similar in construction to parts of HS (Table II) this table shows the coefficients from Tobit regressions of option grant pay-for-performance sensitivity per $1000 change in
shareholder wealth against the lagged change and contemporary change in shareholder wealth measures of institutional influence Tobinrsquos Q firm market capitalization and
dummy controls for CEO and year effects We estimate models (1) to (4) over the period 1992‒1997 using data supplied by HS and models (5) to (8) are estimated over the
whole sample Models (1) (2) (5) and (6) use Institutional Concentration as a measure of institutional influence which HS define as the shares owned by the institutions
with the five largest holdings divided by the shares held by all institutions These models control for total institutional ownership Models (3) (4) (7) and (8) use Top 5
Ownership as a measure of institutional influence which we define as the proportion of shares outstanding held by the institutions with the five largest holdings Coefficients
for industry and year controls are not presented One two and three asterisks denote significance at 10 5 and 1 levels
Hartzell and Starks Database 1992‒1997 Full Sample from ExecuComp 1992‒2002
(1) (2) (3) (4) (5) (6) (7) (8)
Change in Shareholder Wealth 77721 191781 89874 137368 -446 -218 -664 -203
(171) (429) (194) (317) (319) (168) (453) (158)
Lag1 Change in Shareholder Wealth 14446 99963 88719 8544 -144 0273 -0342 00852
(027) (212) (166) (184) (071) (015) (016) (005)
Lag1 Institutional Concentration 13994 -00692 26405 03304
(767) (032) (1402) (166)
Lag1 Total Institutional Ownership 08055 10434 07106 0937
(593) (781) (666) (898)
Lag1 Top 5 Institutional Ownership 10567 07457 12116 06293
(399) ( 283) (533) (278)
Lag1 Tobinrsquos Q 0068 00877 00647 00813 00498 00727 00508 00695
(337) (434) (319) (401) (504) (734) (51) (703)
Lag1 Market Capitalization -106202 -173515 -776 -117
(474) (742) (1011) (153)
Lag1 Log Market Capitalization -0297 -02319 -04111 -03908
(1576) (1439) (2997) (2841)
CEO Dummy 13745 13682 13765 13674 28197 28128 28155 28143
(1913) (1918) (1914) (1916) (3388) (3396) (3377) (34)
Number of Observations 33878 33878 33878 33878 97679 97679 97679 97679
29
Table V Pay-for-Performance Sensitivity of Cash Compensation (Salary plus Bonus) as a Function of Institutional Influence - Level Market Capitalization
Similar in parts to models (1) and (2) of HS (Table IV) this table displays the coefficients from OLS regressions of the change in cash compensation against the lagged
change and contemporary change in shareholder wealth measures of institutional influence Tobinrsquos Q firm market capitalization and dummy controls for CEO and year
effects We estimate models (1) to (4) over the period 1992‒1997 based on our ExecuComp sample and models (5) to (8) over the whole sample Models (1) (2) (5) and (6)
use Institutional Concentration as a measure of institutional influence which HS define as the shares owned by the institutions with the five largest holdings divided by the
shares held by all institutions These models control for total institutional ownership Models (4) and (8) use Top 5 Ownership as a measure of institutional influence which
we define as the proportion of shares outstanding held by the institutions with the 5 largest holdings Coefficients for industry and year controls are not presented One two
and three asterisks denote significance at 10 5 and 1 levels
Our Database based on ExecuComp 1992‒1997 Full Sample based on ExecuComp 1992‒2002
(1) (2) (3) (4) (5) (6) (7) (8)
Lag1 Change in Shareholder Wealth 00014 00017 00088 00089 -00006 -00007 00072 00086
(051) (062) (129) (103) (027) (029) (131) (16)
Change in Shareholder Wealth -00102 00102 00105 00032 00102 00101
(097) (292) (293) (055) (611) (608)
Change Shareholder Wealth Lag1 Institutional
Concentration
00435 00764 00363 00253
(463) (258) (689) (144)
Lag1 Change Shareholder Wealth Lag1
Institutional Concentration
-00197 -00283
(122) (197)
Lag1 Institutional Concentration -492182 -288861 400149 453623
(09) (056) (098) (114)
Lag1 Total Institutional Ownership 720688 538372 655091 559587 682069 726226
(276) (232) (281) (314) (308) (322)
Lag1 Change Shareholder Wealth Lag1 Top 5
Institutional Ownership
-0046 -00689
(091) (256)
Lag1 Top 5 Institutional Ownership 583044 1191634
(092) (25)
Lag1 Tobinrsquos Q -32761 -36806 -19422 -18023 -54283 -48681 -29534 -24243
(138) (172) (077) (078) (227) (245) (136) (111)
Lag1 Market Capitalization 00004 00004 00002 00006 00007 00008 00007 00007
(042) (044) (022) (07) (086) (092) (082) (09)
CEO Dummy 768035 765474 768334 773634 616315 61729 619028 618535
(354) (355) (356) (355) (417) (418) (419) (418)
Adjusted R-Sq 00057 0006 00052 0005 00097 00097 00098 00102
Number of Observations 41305 41305 41305 41305 85518 85518 85518 85518
30
Table VI Level of Salary as a Function of Institutional Influence
Similar in parts to models (1) and (2) of HS (Table V) this table displays the coefficients from OLS regressions of salary against the lagged change and contemporary change in
shareholder wealth measures of institutional influence Tobinrsquos Q firm market capitalization and dummy controls for CEO industry and year effects We estimate models (1) to
(4) over the period 1992‒1997 using the actual data supplied by HS and models (5) to (8) over the whole sample 1992‒2002 Models (1) (2) (5) and (6) use Institutional
Concentration as a measure of institutional influence which HS define as the shares owned by the institutions with the five largest holdings divided by the shares held by all
institutions These models control for total institutional ownership Models (3) (4) (7) and (8) use Top 5 Ownership as a measure of institutional influence which we define as
the proportion of shares outstanding held by the institutions with the 5 largest holdings Coefficients for industry and year controls are not presented One two and three
asterisks denote significance at 10 5 and 1 levels
Hartzell and Starks Database 1992‒1997 Full Sample based on ExecuComp 1992‒2002
(1) (2) (3) (4) (5) (6) (7) (8)
Change in Shareholder Wealth 04125E-4 0003 -2439E-4 0003 00021 00012 00024 00012
(057) (464) (031) (468) (1004) (72) (1041) (706)
Lag1 Change in Shareholder Wealth -00062 7856E-4 -0007 9003E-4 -00014 00002 -00014 00003
(594) (098) (592) (112) (522) (123) (5) (145)
Lag1 Institutional Concentration -2475087 555134 -2734983 943317
(3245) (710) (4965) (1671)
Lag1 Total Institutional Ownership 1020109 217488 936337 252464
(1904) (432) (2471) (704)
Lag1 Top 5 Institutional Ownership -148355 31354 -87926 423166
(139) (332) (118) (643)
Lag1 Tobinrsquos Q -195348 -252191 -192344 -25158 -119837 -146373 -122132 -143821
(2812) (3762) (2721) (3752) (248) (274) (2583) (2727)
Lag1 Market Capitalization 844E-06 103E-5 00034 00039
(2781) (3051) (2874) (3104)
Lag1 Log Market Capitalization 750021 723625 776657 732279
(7511) (9285) (11224) (14039)
CEO Dummy 2683978 2702238 2668244 2702417 3281133 3285103 3292175 3283145
(8625) (9277) (8246) (9271) (12739) (13616) (12336) (13605)
Adjusted R-Sq 04604 05302 04187 05295 03856 04588 03476 04574
Number of Observations 36280 36280 36280 36280 97679 97679 97679 97679
31
Table VII Level of Total Direct Compensation as a Function of Institutional Influence
Similar in parts to models (3) and (5) of HS (Table V) this table displays the coefficients from OLS regressions of total direct compensation against the lagged change and
contemporary change in shareholder wealth measures of institutional influence Tobinrsquos Q firm market capitalization and dummy controls for CEO industry and year effects
We estimate models (1) to (4) over the period 1992‒1997 and models (5) to (8) over the whole sample Models (1) (2) (5) and (6) use Institutional Concentration as a measure
of institutional influence which HS define as the shares owned by the institutions with the five largest holdings divided by the shares held by all institutions These models
control for total institutional ownership Models (3) (4) (7) and (8) use Top 5 Ownership as a measure of institutional influence which we define as the proportion of shares
outstanding held by the institutions with the five largest holdings Coefficients for industry and year controls are not presented One two and three asterisks denote significance
at 10 5 and 1 levels
Hartzell and Starks Database 1992‒1997 Full Sample from ExecuComp 1992‒2002
(1) (2) (3) (4) (5) (6) (7) (8)
Change in Shareholder Wealth 005779 007006 005282 006791 00423 00148 0046 00138
(415) (581) (369) (565) (203) (07) (22) (066)
Lag1 Change in Shareholder Wealth 003182 00714 00272 00715 00613 0126 00614 01278
(126) (298) (107) (299) (275) (488) (276) (494)
Lag1 Institutional Concentration -1608633 3248998 -3424892 3334186
(151) (283) (1172) (118)
Lag1 Total Institutional Ownership 104013 5473128 1531013 -2926764
(1645) (831) (876) (158)
Lag1 Top 5 Institutional Ownership 188585 3927986 -2159177 -2640879
(016) (343) (067) (092)
Lag1 Tobinrsquos Q -445692 -803612 -454859 -833566 3257027 2971201 3213936 3086205
(427) (791) (432) (831) (83) (77) (815) (793)
Lag1 Market Capitalization 49E-5 615E-5 01031 01088
(113) (1365) (1555) (168)
Lag1 Log Market Capitalization 4683497 4759377 1645156 1427856
(2775) (3469) (3679) (4239)
CEO Dummy 1501307 1508886 1493688 1508736 4713079 4713629 47278 4703908
(2648) (2684) (2618) (2683) (282) (2828) (2819) (2824)
Adjusted R-Sq 01861 02060 01695 02051 01071 01113 01033 011
Number of Observations 33229 33229 33229 33229 97679 97679 97679 97679
32
Table VIII Two Stage Least Squares (Reverse Causality) Tests - Executive Compensation Measures as a Function of Institutional Concentration
This table shows the coefficients from the Two-Stage Least Squares (2SLS) regressions of executive compensation measures against the lagged change
and contemporary change in shareholder wealth instrumented institutional concentration and total institutional ownership Tobinrsquos Q the natural
logarithm of firm market capitalization and dummy variable controls for CEO and year effects We instrument measures of institutional concentration
and total institutional ownership by orthogonalizing with respect to option grant pay-for-performance sensitivity salary and total compensation
Coefficients for industry and year controls are not presented First stage results are not presented One two and three asterisks denote significance at
10 5 and 1 levels
ExecuComp Sample 1992‒1997 Full Sample from ExecuComp 1992‒2002
PPSOG Salary TC PPSOG Salary TC
Change in Shareholder Wealth 144E-05 0004 02133 -409E-06 00011 00157
(308) (775) (596) (691) (638) (068)
Lag1 Change in Shareholder Wealth -240E-05 -00004 00866 118E-06 00002 01422
(131) (087) (346) (097) (116) (51)
Lag1 Institutional Concentration -00898 668681 3263143 02024 916656 2856417
(052) (801) (176) (163) (1545) (1004)
Lag1 Total Institutional Ownership 01021 -43997 2512587 02983 -104171 -1028347
(09) (084) (153) (4) (269) (487)
Lag1 Tobinrsquos Q 00754 -225606 119297 00655 -158657 2997205
(596) (1649) (053) (834) (2535) (727)
Lag1 Log Market Capitalization -04234 828713 7864046 -04002 825106 1685638
(2195) (8132) (2977) (4106) (11133) (373)
CEO Dummy 15848 3214117 2950274 19149 3205862 4529101
(2223) (9389) (2255) (3802) (1331) (2758)
Overall R-Sq -- 04981 01206 -- 04738 01178
Number of Observations 41305 41305 41305 85518 85518 85518
33
Table IX Two Stage Least Squares (Reverse Causality) Tests - Executive Compensation Measures as a Function of Top 5 Ownership
This table shows the coefficients from 2SLS regressions of executive compensation measures against the lagged change and contemporary change
in shareholder wealth instrumented Top 5 institutional ownership Tobinrsquos Q natural logarithm of firm market capitalization and dummy variable
controls for CEO and year effects We instrument Top 5 institutional ownership by orthogonalizing with respect to option grant pay-for-
performance sensitivity salary and total compensation Coefficients for industry and year controls are not presented First stage results are not
presented One two and three asterisks denote significance at 10 5 and 1 levels
ExecuComp sample 1992‒1997 Full Sample 1992‒2002
PPSOG Salary TC PPSOG Salary TC
Change in Shareholder Wealth 136E-05 00041 02128 -407E-06 00011 00146
(293) (79) (596) (692) (619) (063)
Lag1 Change in Shareholder Wealth -241E-05 -00003 00868 116E-06 00003 01439
(133) (07) (347) (096) (142) (514)
Lag1 Top 5 Institutional Ownership -03774 -187512 -3848531 00179 -104261 -126659
(163) (186) (166) (011) (149) (436)
Lag1 Tobinrsquos Q 00731 -223886 10074 00642 -155467 3115073
(573) (1631) (045) (819) (2509) (75)
Lag1 Log Market Capitalization -0416 775637 7696302 -04015 764632 1462536
(2345) (10084) (3766) (3688) (13708) (4072)
CEO Dummy 15854 3211991 2950158 19152 3203152 4519385
(2223) (9381) (2254) (3804) (133) (2752)
Overall R-Sq -- 04967 01202 -- 04722 01168
Number of Observations 41305 41305 41305 85518 85518 85518
34
Table X Specification Tests utilizing Box-Cox Transformation of Market Capitalization
This table presents results from a Box-Cox transformation of market capitalization In
each model we specify transformation of market capitalization only Panel A shows
results for the period 1992‒1997 and Panel B shows results for the entire period
1992‒2002 The first part of each panel shows the Lambda coefficient that applies to
the transformed market capitalization The second part of each panel shows log
likelihood scores with LR Chi2 Statistics in parentheses ‒ these test for the
appropriateness of a linear or log model specification
Panel A Our ExecuComp sample 1992‒1997
(1) (2) (3)
PPSOG Salary TC
Lambda -05143 01881 02876
(-1914) (2413) (1329)
Test H0 (Log Likelihood Score)
Lambda = -1 -1279748 -3178575 -4795326
(23866) (85952) (99688)
Lambda = 0 -1280152 -3138456 -4791159
(31934) (57147) (16339)
Lambda = 1 -1283157 -3163807 -4793506
(92039) (564156) (63292)
Panel B Full ExecuComp sample 1992‒2002
(1) (2) (4)
PPSOG Salary TC
Lambda -044 01853 04036
(-3015) (40) (3666)
Test H0 (Log Likelihood Score)
Lambda = 0 -2688334 -6476494 -10384490
(66599) (160356) (126236)
Lambda = 1 -2694886 -6538506 -10386795
(197644) (1400605) (172345)
35
Figure 2 Institutional Ownership Measures (LHA) vs Total
Compensation (RHA) by Market Capitalization
0
10
20
30
40
50
60
70
1 2 3 4 5 6 7 8 9 10
$00
$10
$20
$30
$40
$50
$60
$70
$80
$90
Millio
ns
InstitutionalConcentration (LHS)Top5 Ownership (LHS)
Institutional TotalOwnership (LHS)Total Compensation(RHS)
Page 20
20
Ownership an insignificant relation exists with total compensation in model (3) using the
HS size control but it becomes both positive and significant in (4) with the logged size
control In models (5) and (6) where models are estimated using the full sample we
observe a similar reversal in sign as occurs in models (1) and (2) However now the
magnitude of the reversal is almost equal with a difference of about $67m that is very
large indeed In model (5) the coefficient on institutional concentration is (negative)
$342m and then in model (6) the coefficient is $33m Top 5 Ownership is insignificant
in both models (7) and (8) unlike the results for Salary
VI Robustness
Two issues are examined in is this Section ndash reverse causality and the appropriate
transformation of market capitalization A methodological issue inherent in examining the
effect of ownership-based measures of institutional influence on executive compensation
is endogeneity driven by reverse causality We utilize an instrumental variables approach
where orthogonalized institutional influence measures are regressed on each of the
compensation measures lagged option grant pay-for-performance sensitivity salary and
total compensation In addition market capitalization and a CEO dummy are included in
each regression We use the residual from each regression as an instrument for
institutional influence Such a process removes any preferences for compensation from
the institutional influence measures
ltltINSERT TABLES VIII AND IXgtgt
Table VIII presents the results where we instrument the HS measure lagged
institutional concentration with a compensation-orthogonalized equivalent The analysis
indicates that the instrumented institutional concentration measure is positively and
significantly related to the level of salary and total compensation and insignificantly
21
related to option grant pay-for-performance sensitivity (PPSOG) Table IX presents
results where we instrument Top 5 Ownership with a compensation-orthogonalized
equivalent The instrumented Top 5 Ownership measure insignificantly relates to option
grant pay-for-performance sensitivity and barely negatively relates to salary although it
negatively relates to total compensation suggesting reverse causality is driving a
significant portion of our previous results
ltltINSERT TABLE Xgtgt
Another issue examined in this section is the appropriate transformation of market
capitalization Aneuryn-Evans and Deaton (1980) provide an evaluation of methods
comparing linear and log-linear specifications We use the Box-Cox test to compare
logarithm versus linear specifications In applying the Box-Cox test we specify
transformation of only market capitalization ndash the dependent variable and all other
independent variables remain unchanged Table X presents the lambda coefficient for
each model and the log likelihood score In all models in Panel A and B a lambda of zero
receives a higher (closer to zero) log likelihood score suggesting the log specification is
appropriate
VII Conclusion
Hartzell and Starks (2003) appears to document an effective role for institutional
investors in increasing executive incentives and lowering compensation levels Our paper
re-addresses the issue of concentrated institutional monitoring from a perspective
informed by an additional decade of research since their paper appeared We cannot find
evidence that concentrated institutional investors really have any positive (or negative for
that matter) effect on executive incentive levels While superficially we find evidence
that they may raise executive pay levels this is most likely driven by an institutional
22
preference for highly remunerated executives in large companies and thus by reverse
causality Moreover our findings are not consistent with the rent-extraction hypothesis
suggesting that firms systematically overpay executives in the absence of effective
monitoring by outside parties such as institutional investors
23
References
Almazan Andres Jay C Hartzell and Laura T Starks 2005 ldquoActive Institutional
Shareholders and Costs of Monitoring Evidence from Executive Compensationrdquo
Financial Management Winter 5ndash34
Aneuryn-Evans Gwyn and Angus Deaton 1980 ldquoTesting Linear versus Logarithmic
Regression Modelsrdquo Review of Economic Studies 47 (1) Econometrics Issue 275ndash
291
Babenko Ilona 2009 ldquoShare Repurchases and Pay-Performance Sensitivity of Employee
Compensation Contractsrdquo Journal of Finance 64117-150
Baker George P and Brian J Hall 2004 ldquoCEO Incentives and Firm Sizerdquo Journal of
Labor Market Economics 22 767ndash798
Bebchuk Lucian A Jesse M Fried and David Walker 2002 ldquoManagerial power and
rent extraction in the design of executive compensationrdquo University of Chicago Law
Review 69 751ndash846
Bebchuk Lucian A and Jesse M Fried 2003 ldquoExecutive Compensation as an Agency
Problemrdquo Journal of Economic Perspectives 17 71ndash92
Bebchuk Lucian A and Jesse M Fried 2004 Pay without Performance The Unfulfilled
Promise of Executive Compensation Cambridge MA Harvard University Press
Bertrand Marianne and Sendhil Mullainathan 2001 ldquoAre CEOs Rewarded for Luck
The Ones Without Principals Arerdquo Quarterly Journal of Economics 116 901ndash32
Black Fischer and Myron Scholes 1973 ldquoThe Pricing of Options and Corporate
Liabilitiesrdquo Journal of Political Economy 81 637ndash54
Cadman Brian Sandy Klasa and Steve Matsunaga 2010 ldquoDeterminants of CEO Pay A
Comparison of ExecuComp and Non-ExecuComp Firmsrdquo The Accounting Review 85
1511ndash1543
Chhaochharia Vidhi and Yaniv Grinstein 2009 ldquoCEO Compensation and Board
Structurerdquo Journal of Finance 64 231ndash261
Cichello Michael S 2005 ldquoThe impact of firm size on pay-performance sensitivitiesrdquo
Journal of Corporate Finance 11 609ndash627
Davis Gerald F and E Han Kim 2007 ldquoBusiness ties and proxy voting by mutual
fundsrdquo Journal of Financial Economics 85 552ndash570
Dikolli Shane S Susan L Kulp and Karen L Sedatole 2009 ldquoTransient Institutional
Ownership and CEO Contractingrdquo Accounting Review 84 737ndash770
Edmans Alex Xavier Gabaix and Augustin Landier 2009 ldquoA Multiplicative Model of
Optimal CEO Incentives in Market Equilibriumrdquo Review of Financial Studies 22
4881ndash4917
Edmans Alex 2009 ldquoBlockholder Trading Market Efficiency and Managerial Myopiardquo
Journal of Finance 64 2481ndash2513
Ertimur Yonca Fabrizio Ferri and Volkan Muslu 2011 ldquoShareholder Activism and
CEO Payrdquo Review of Financial Studies 24 535ndash592
24
Gabaix Xavier and Augustin Landier 2008 ldquoWhy Has CEO Pay Increased So Muchrdquo
Quarterly Journal of Economics 123 49ndash100
Garen John 1994 ldquoExecutive compensation and principal-agent theoryrdquo Journal of
Political Economy 102 1175ndash1199
Gibbons Robert and Kevin J Murphy 1992 ldquoOptimal Incentive Contracts in the
Presence of Career Concerns Theory and Evidencerdquo Journal of Political Economy
100 468ndash505
Greene William H 2000 (4th
Edition) Econometric Analysis Prentice-Hall NJ
Guthrie Katherine Jan Sokolowsky and Kam-Ming Wan 2012 ldquoCEO Compensation
and Board Structure Revisitedrdquo Journal of Finance Forthcoming
Hadlock Charles R and Gerald B Lumer 1997 ldquoCompensation turnover and top
management incentives historical evidencerdquo Journal of Business 70 153ndash187
Hartzell Jay C and Laura T Starks 2003 ldquoInstitutional Investors and Executive
Compensationrdquo Journal of Finance 58 2351ndash2374
Huson Mark R Robert Parrino and Laura T Starks 2001 ldquoInternal Monitoring
Mechanisms and CEO Turnover A Long-Term Perspectiverdquo Journal of Finance 56
2265ndash2297
Jensen Michael C and Kevin J Murphy 1990 ldquoPerformance Pay and Top Management
Incentivesrdquo Journal of Political Economy 98 225ndash264
Jin Li 2002 ldquoCEO compensation diversification and incentivesrdquo Journal of Financial
Economics 66 29ndash63
Kang Qiang and Qiao Liu 2008 ldquoStock Market Information Production and Executive
Incentivesrdquo Journal of Corporate Finance 14 484ndash498
Merton Robert 1973 ldquoTheory of Rational Option Pricingrdquo Bell Journal of Economics
and Management Science 4 141-183
Morse Adair Vikram Nanda and Amit Seru 2011 ldquoAre Incentive Contracts Rigged by
Powerful CEOsrdquo Journal of Finance 66 1779ndash1821
Murphy Kevin J 1985 ldquoCorporate performance and managerial remuneration An
empirical analysisrdquo Journal of Accounting and Economics 7 11ndash42
Murphy Kevin J 1999 ldquoExecutive compensationrdquo in Orley Ashenfelter and David
Card Handbook of Labor Economics vol 3 Elsevier
Rosen Sherwin 1992 ldquoContracts and the market for executivesrdquo In Contract
Economics ed Lars Werin and Hans Wijkander 181ndash211 Cambridge MA
Blackwell
Rothberg Burton and Steven Lilien 2006 ldquoMutual Funds and Proxy Voting New
Evidence on Corporate Governancerdquo Journal of Business and Technology Law 1
157ndash184
Schaefer Scott 1998 ldquoThe dependence of payndashperformance sensitivity on the size of the
firmrdquo Review of Economics and Statistics 80 436ndash443
Yermack David 1995 ldquoDo Corporations Award CEO Stock Options Effectivelyrdquo
Journal of Financial Economics 39 237ndash69
25
Table I Univariate Descriptive Statistics
This table reports sample statistics for the key variables used in this study based on our comprehensive
ExecuComp dataset Institutional Concentration is the shares owned by the institutions with the five largest
holdings divided by the shares held by all institutions Top 5 Ownership is the proportion of shares
outstanding held by the institutions with the five largest holdings Total Institutional Ownership is the
proportion of shares held by all institutions Panel B shows the measures executive compensation The pay-
for-performance sensitivity (PPS) of option grants shows the change in executive wealth due to a $1000
change in firm in shareholder wealth Expressions for Salary and Total Compensation are in thousands of
dollars while Market Capitalization is in millions of dollars Panel C shows firm characteristics
Variable Mean Median Std Dev P10 P90
Panel A Institutional Investor Holdings
Institutional Concentration 0450 0424 0151 0282 0656
Top 5 Institutional Ownership 0229 0220 0101 0110 0355
Total Institutional Ownership 0546 0561 0206 0258 0801
Panel B Executive Compensation
PPS of Option Grants 1161 0156 3829 0 2822
Salary 36311 29462 25231 15179 67000
Total Compensation 258313 85878 980701 23875 511022
Panel C Firm Characteristics
Tobins Q 1744 1139 2738 0495 3322
Market Capitalization 560699 112094 1988108 19754 1048754
26
Table II Pairwise Correlation Coefficients
This table presents the correlations between the key variables based on our comprehensive
ExecuComp dataset Institutional Concentration is the shares owned by the institutions with the five
largest holdings divided by the shares held by all institutions Top 5 Ownership is the proportion of
shares outstanding held by the institutions with the five largest holdings Total Institutional
Ownership is the proportion of shares held by all institutions The pay-for-performance sensitivity
(PPS) of option grants is the change in executive wealth due to a $1000 change in shareholder
wealth
(1) (2) (3) (4) (5) (6) (7) (8)
(1) Institutional Concentration 1
(2) Top 5 Ownership 0268 1
(3) Total Ownership -0415 0706 1
(4) PPS of Option Grants 0122 0042 -0024 1
(5) Log Salary -0270 -0012 0171 0038 1
(6) Log Market Capitalization -0630 -0157 0261 -0156 0420 1
(7) Level Market Capitalization -0241 -0157 0007 -0058 0219 0534 1
(8) Tobins Q -0008 -0034 0002 0015 -0083 0066 0121 1
27
Table III
Descriptive Statistics by Market Capitalization Deciles
Market
Capitalization
Deciles
Market
Capitalization
($m)
Total
Compensation
($000)
Salary
($)
PPS Option
Grants
($)
Institutional
Concentration
()
Top 5 Ownership
()
Institutional
Total Ownership
()
1 (Small) 1181 76008 234700 2720 611 200 327
2 2668 90206 257955 1796 495 217 438
3 4193 105946 275120 1517 454 220 485
4 6045 129892 296067 1376 420 206 491
5 8742 155391 315864 1216 394 198 502
6 13086 191888 345942 1072 380 202 531
7 20790 239501 377021 0874 368 196 532
8 34920 314955 427808 0687 354 190 538
9 67747 450970 480319 0569 328 173 528
10 (Big) 366540 846066 614354 0297 275 138 502
28
Table IV Pay-for-Performance Sensitivity of Option Grants as a Function of Institutional Influence
Similar in construction to parts of HS (Table II) this table shows the coefficients from Tobit regressions of option grant pay-for-performance sensitivity per $1000 change in
shareholder wealth against the lagged change and contemporary change in shareholder wealth measures of institutional influence Tobinrsquos Q firm market capitalization and
dummy controls for CEO and year effects We estimate models (1) to (4) over the period 1992‒1997 using data supplied by HS and models (5) to (8) are estimated over the
whole sample Models (1) (2) (5) and (6) use Institutional Concentration as a measure of institutional influence which HS define as the shares owned by the institutions
with the five largest holdings divided by the shares held by all institutions These models control for total institutional ownership Models (3) (4) (7) and (8) use Top 5
Ownership as a measure of institutional influence which we define as the proportion of shares outstanding held by the institutions with the five largest holdings Coefficients
for industry and year controls are not presented One two and three asterisks denote significance at 10 5 and 1 levels
Hartzell and Starks Database 1992‒1997 Full Sample from ExecuComp 1992‒2002
(1) (2) (3) (4) (5) (6) (7) (8)
Change in Shareholder Wealth 77721 191781 89874 137368 -446 -218 -664 -203
(171) (429) (194) (317) (319) (168) (453) (158)
Lag1 Change in Shareholder Wealth 14446 99963 88719 8544 -144 0273 -0342 00852
(027) (212) (166) (184) (071) (015) (016) (005)
Lag1 Institutional Concentration 13994 -00692 26405 03304
(767) (032) (1402) (166)
Lag1 Total Institutional Ownership 08055 10434 07106 0937
(593) (781) (666) (898)
Lag1 Top 5 Institutional Ownership 10567 07457 12116 06293
(399) ( 283) (533) (278)
Lag1 Tobinrsquos Q 0068 00877 00647 00813 00498 00727 00508 00695
(337) (434) (319) (401) (504) (734) (51) (703)
Lag1 Market Capitalization -106202 -173515 -776 -117
(474) (742) (1011) (153)
Lag1 Log Market Capitalization -0297 -02319 -04111 -03908
(1576) (1439) (2997) (2841)
CEO Dummy 13745 13682 13765 13674 28197 28128 28155 28143
(1913) (1918) (1914) (1916) (3388) (3396) (3377) (34)
Number of Observations 33878 33878 33878 33878 97679 97679 97679 97679
29
Table V Pay-for-Performance Sensitivity of Cash Compensation (Salary plus Bonus) as a Function of Institutional Influence - Level Market Capitalization
Similar in parts to models (1) and (2) of HS (Table IV) this table displays the coefficients from OLS regressions of the change in cash compensation against the lagged
change and contemporary change in shareholder wealth measures of institutional influence Tobinrsquos Q firm market capitalization and dummy controls for CEO and year
effects We estimate models (1) to (4) over the period 1992‒1997 based on our ExecuComp sample and models (5) to (8) over the whole sample Models (1) (2) (5) and (6)
use Institutional Concentration as a measure of institutional influence which HS define as the shares owned by the institutions with the five largest holdings divided by the
shares held by all institutions These models control for total institutional ownership Models (4) and (8) use Top 5 Ownership as a measure of institutional influence which
we define as the proportion of shares outstanding held by the institutions with the 5 largest holdings Coefficients for industry and year controls are not presented One two
and three asterisks denote significance at 10 5 and 1 levels
Our Database based on ExecuComp 1992‒1997 Full Sample based on ExecuComp 1992‒2002
(1) (2) (3) (4) (5) (6) (7) (8)
Lag1 Change in Shareholder Wealth 00014 00017 00088 00089 -00006 -00007 00072 00086
(051) (062) (129) (103) (027) (029) (131) (16)
Change in Shareholder Wealth -00102 00102 00105 00032 00102 00101
(097) (292) (293) (055) (611) (608)
Change Shareholder Wealth Lag1 Institutional
Concentration
00435 00764 00363 00253
(463) (258) (689) (144)
Lag1 Change Shareholder Wealth Lag1
Institutional Concentration
-00197 -00283
(122) (197)
Lag1 Institutional Concentration -492182 -288861 400149 453623
(09) (056) (098) (114)
Lag1 Total Institutional Ownership 720688 538372 655091 559587 682069 726226
(276) (232) (281) (314) (308) (322)
Lag1 Change Shareholder Wealth Lag1 Top 5
Institutional Ownership
-0046 -00689
(091) (256)
Lag1 Top 5 Institutional Ownership 583044 1191634
(092) (25)
Lag1 Tobinrsquos Q -32761 -36806 -19422 -18023 -54283 -48681 -29534 -24243
(138) (172) (077) (078) (227) (245) (136) (111)
Lag1 Market Capitalization 00004 00004 00002 00006 00007 00008 00007 00007
(042) (044) (022) (07) (086) (092) (082) (09)
CEO Dummy 768035 765474 768334 773634 616315 61729 619028 618535
(354) (355) (356) (355) (417) (418) (419) (418)
Adjusted R-Sq 00057 0006 00052 0005 00097 00097 00098 00102
Number of Observations 41305 41305 41305 41305 85518 85518 85518 85518
30
Table VI Level of Salary as a Function of Institutional Influence
Similar in parts to models (1) and (2) of HS (Table V) this table displays the coefficients from OLS regressions of salary against the lagged change and contemporary change in
shareholder wealth measures of institutional influence Tobinrsquos Q firm market capitalization and dummy controls for CEO industry and year effects We estimate models (1) to
(4) over the period 1992‒1997 using the actual data supplied by HS and models (5) to (8) over the whole sample 1992‒2002 Models (1) (2) (5) and (6) use Institutional
Concentration as a measure of institutional influence which HS define as the shares owned by the institutions with the five largest holdings divided by the shares held by all
institutions These models control for total institutional ownership Models (3) (4) (7) and (8) use Top 5 Ownership as a measure of institutional influence which we define as
the proportion of shares outstanding held by the institutions with the 5 largest holdings Coefficients for industry and year controls are not presented One two and three
asterisks denote significance at 10 5 and 1 levels
Hartzell and Starks Database 1992‒1997 Full Sample based on ExecuComp 1992‒2002
(1) (2) (3) (4) (5) (6) (7) (8)
Change in Shareholder Wealth 04125E-4 0003 -2439E-4 0003 00021 00012 00024 00012
(057) (464) (031) (468) (1004) (72) (1041) (706)
Lag1 Change in Shareholder Wealth -00062 7856E-4 -0007 9003E-4 -00014 00002 -00014 00003
(594) (098) (592) (112) (522) (123) (5) (145)
Lag1 Institutional Concentration -2475087 555134 -2734983 943317
(3245) (710) (4965) (1671)
Lag1 Total Institutional Ownership 1020109 217488 936337 252464
(1904) (432) (2471) (704)
Lag1 Top 5 Institutional Ownership -148355 31354 -87926 423166
(139) (332) (118) (643)
Lag1 Tobinrsquos Q -195348 -252191 -192344 -25158 -119837 -146373 -122132 -143821
(2812) (3762) (2721) (3752) (248) (274) (2583) (2727)
Lag1 Market Capitalization 844E-06 103E-5 00034 00039
(2781) (3051) (2874) (3104)
Lag1 Log Market Capitalization 750021 723625 776657 732279
(7511) (9285) (11224) (14039)
CEO Dummy 2683978 2702238 2668244 2702417 3281133 3285103 3292175 3283145
(8625) (9277) (8246) (9271) (12739) (13616) (12336) (13605)
Adjusted R-Sq 04604 05302 04187 05295 03856 04588 03476 04574
Number of Observations 36280 36280 36280 36280 97679 97679 97679 97679
31
Table VII Level of Total Direct Compensation as a Function of Institutional Influence
Similar in parts to models (3) and (5) of HS (Table V) this table displays the coefficients from OLS regressions of total direct compensation against the lagged change and
contemporary change in shareholder wealth measures of institutional influence Tobinrsquos Q firm market capitalization and dummy controls for CEO industry and year effects
We estimate models (1) to (4) over the period 1992‒1997 and models (5) to (8) over the whole sample Models (1) (2) (5) and (6) use Institutional Concentration as a measure
of institutional influence which HS define as the shares owned by the institutions with the five largest holdings divided by the shares held by all institutions These models
control for total institutional ownership Models (3) (4) (7) and (8) use Top 5 Ownership as a measure of institutional influence which we define as the proportion of shares
outstanding held by the institutions with the five largest holdings Coefficients for industry and year controls are not presented One two and three asterisks denote significance
at 10 5 and 1 levels
Hartzell and Starks Database 1992‒1997 Full Sample from ExecuComp 1992‒2002
(1) (2) (3) (4) (5) (6) (7) (8)
Change in Shareholder Wealth 005779 007006 005282 006791 00423 00148 0046 00138
(415) (581) (369) (565) (203) (07) (22) (066)
Lag1 Change in Shareholder Wealth 003182 00714 00272 00715 00613 0126 00614 01278
(126) (298) (107) (299) (275) (488) (276) (494)
Lag1 Institutional Concentration -1608633 3248998 -3424892 3334186
(151) (283) (1172) (118)
Lag1 Total Institutional Ownership 104013 5473128 1531013 -2926764
(1645) (831) (876) (158)
Lag1 Top 5 Institutional Ownership 188585 3927986 -2159177 -2640879
(016) (343) (067) (092)
Lag1 Tobinrsquos Q -445692 -803612 -454859 -833566 3257027 2971201 3213936 3086205
(427) (791) (432) (831) (83) (77) (815) (793)
Lag1 Market Capitalization 49E-5 615E-5 01031 01088
(113) (1365) (1555) (168)
Lag1 Log Market Capitalization 4683497 4759377 1645156 1427856
(2775) (3469) (3679) (4239)
CEO Dummy 1501307 1508886 1493688 1508736 4713079 4713629 47278 4703908
(2648) (2684) (2618) (2683) (282) (2828) (2819) (2824)
Adjusted R-Sq 01861 02060 01695 02051 01071 01113 01033 011
Number of Observations 33229 33229 33229 33229 97679 97679 97679 97679
32
Table VIII Two Stage Least Squares (Reverse Causality) Tests - Executive Compensation Measures as a Function of Institutional Concentration
This table shows the coefficients from the Two-Stage Least Squares (2SLS) regressions of executive compensation measures against the lagged change
and contemporary change in shareholder wealth instrumented institutional concentration and total institutional ownership Tobinrsquos Q the natural
logarithm of firm market capitalization and dummy variable controls for CEO and year effects We instrument measures of institutional concentration
and total institutional ownership by orthogonalizing with respect to option grant pay-for-performance sensitivity salary and total compensation
Coefficients for industry and year controls are not presented First stage results are not presented One two and three asterisks denote significance at
10 5 and 1 levels
ExecuComp Sample 1992‒1997 Full Sample from ExecuComp 1992‒2002
PPSOG Salary TC PPSOG Salary TC
Change in Shareholder Wealth 144E-05 0004 02133 -409E-06 00011 00157
(308) (775) (596) (691) (638) (068)
Lag1 Change in Shareholder Wealth -240E-05 -00004 00866 118E-06 00002 01422
(131) (087) (346) (097) (116) (51)
Lag1 Institutional Concentration -00898 668681 3263143 02024 916656 2856417
(052) (801) (176) (163) (1545) (1004)
Lag1 Total Institutional Ownership 01021 -43997 2512587 02983 -104171 -1028347
(09) (084) (153) (4) (269) (487)
Lag1 Tobinrsquos Q 00754 -225606 119297 00655 -158657 2997205
(596) (1649) (053) (834) (2535) (727)
Lag1 Log Market Capitalization -04234 828713 7864046 -04002 825106 1685638
(2195) (8132) (2977) (4106) (11133) (373)
CEO Dummy 15848 3214117 2950274 19149 3205862 4529101
(2223) (9389) (2255) (3802) (1331) (2758)
Overall R-Sq -- 04981 01206 -- 04738 01178
Number of Observations 41305 41305 41305 85518 85518 85518
33
Table IX Two Stage Least Squares (Reverse Causality) Tests - Executive Compensation Measures as a Function of Top 5 Ownership
This table shows the coefficients from 2SLS regressions of executive compensation measures against the lagged change and contemporary change
in shareholder wealth instrumented Top 5 institutional ownership Tobinrsquos Q natural logarithm of firm market capitalization and dummy variable
controls for CEO and year effects We instrument Top 5 institutional ownership by orthogonalizing with respect to option grant pay-for-
performance sensitivity salary and total compensation Coefficients for industry and year controls are not presented First stage results are not
presented One two and three asterisks denote significance at 10 5 and 1 levels
ExecuComp sample 1992‒1997 Full Sample 1992‒2002
PPSOG Salary TC PPSOG Salary TC
Change in Shareholder Wealth 136E-05 00041 02128 -407E-06 00011 00146
(293) (79) (596) (692) (619) (063)
Lag1 Change in Shareholder Wealth -241E-05 -00003 00868 116E-06 00003 01439
(133) (07) (347) (096) (142) (514)
Lag1 Top 5 Institutional Ownership -03774 -187512 -3848531 00179 -104261 -126659
(163) (186) (166) (011) (149) (436)
Lag1 Tobinrsquos Q 00731 -223886 10074 00642 -155467 3115073
(573) (1631) (045) (819) (2509) (75)
Lag1 Log Market Capitalization -0416 775637 7696302 -04015 764632 1462536
(2345) (10084) (3766) (3688) (13708) (4072)
CEO Dummy 15854 3211991 2950158 19152 3203152 4519385
(2223) (9381) (2254) (3804) (133) (2752)
Overall R-Sq -- 04967 01202 -- 04722 01168
Number of Observations 41305 41305 41305 85518 85518 85518
34
Table X Specification Tests utilizing Box-Cox Transformation of Market Capitalization
This table presents results from a Box-Cox transformation of market capitalization In
each model we specify transformation of market capitalization only Panel A shows
results for the period 1992‒1997 and Panel B shows results for the entire period
1992‒2002 The first part of each panel shows the Lambda coefficient that applies to
the transformed market capitalization The second part of each panel shows log
likelihood scores with LR Chi2 Statistics in parentheses ‒ these test for the
appropriateness of a linear or log model specification
Panel A Our ExecuComp sample 1992‒1997
(1) (2) (3)
PPSOG Salary TC
Lambda -05143 01881 02876
(-1914) (2413) (1329)
Test H0 (Log Likelihood Score)
Lambda = -1 -1279748 -3178575 -4795326
(23866) (85952) (99688)
Lambda = 0 -1280152 -3138456 -4791159
(31934) (57147) (16339)
Lambda = 1 -1283157 -3163807 -4793506
(92039) (564156) (63292)
Panel B Full ExecuComp sample 1992‒2002
(1) (2) (4)
PPSOG Salary TC
Lambda -044 01853 04036
(-3015) (40) (3666)
Test H0 (Log Likelihood Score)
Lambda = 0 -2688334 -6476494 -10384490
(66599) (160356) (126236)
Lambda = 1 -2694886 -6538506 -10386795
(197644) (1400605) (172345)
35
Figure 2 Institutional Ownership Measures (LHA) vs Total
Compensation (RHA) by Market Capitalization
0
10
20
30
40
50
60
70
1 2 3 4 5 6 7 8 9 10
$00
$10
$20
$30
$40
$50
$60
$70
$80
$90
Millio
ns
InstitutionalConcentration (LHS)Top5 Ownership (LHS)
Institutional TotalOwnership (LHS)Total Compensation(RHS)
Page 21
21
related to option grant pay-for-performance sensitivity (PPSOG) Table IX presents
results where we instrument Top 5 Ownership with a compensation-orthogonalized
equivalent The instrumented Top 5 Ownership measure insignificantly relates to option
grant pay-for-performance sensitivity and barely negatively relates to salary although it
negatively relates to total compensation suggesting reverse causality is driving a
significant portion of our previous results
ltltINSERT TABLE Xgtgt
Another issue examined in this section is the appropriate transformation of market
capitalization Aneuryn-Evans and Deaton (1980) provide an evaluation of methods
comparing linear and log-linear specifications We use the Box-Cox test to compare
logarithm versus linear specifications In applying the Box-Cox test we specify
transformation of only market capitalization ndash the dependent variable and all other
independent variables remain unchanged Table X presents the lambda coefficient for
each model and the log likelihood score In all models in Panel A and B a lambda of zero
receives a higher (closer to zero) log likelihood score suggesting the log specification is
appropriate
VII Conclusion
Hartzell and Starks (2003) appears to document an effective role for institutional
investors in increasing executive incentives and lowering compensation levels Our paper
re-addresses the issue of concentrated institutional monitoring from a perspective
informed by an additional decade of research since their paper appeared We cannot find
evidence that concentrated institutional investors really have any positive (or negative for
that matter) effect on executive incentive levels While superficially we find evidence
that they may raise executive pay levels this is most likely driven by an institutional
22
preference for highly remunerated executives in large companies and thus by reverse
causality Moreover our findings are not consistent with the rent-extraction hypothesis
suggesting that firms systematically overpay executives in the absence of effective
monitoring by outside parties such as institutional investors
23
References
Almazan Andres Jay C Hartzell and Laura T Starks 2005 ldquoActive Institutional
Shareholders and Costs of Monitoring Evidence from Executive Compensationrdquo
Financial Management Winter 5ndash34
Aneuryn-Evans Gwyn and Angus Deaton 1980 ldquoTesting Linear versus Logarithmic
Regression Modelsrdquo Review of Economic Studies 47 (1) Econometrics Issue 275ndash
291
Babenko Ilona 2009 ldquoShare Repurchases and Pay-Performance Sensitivity of Employee
Compensation Contractsrdquo Journal of Finance 64117-150
Baker George P and Brian J Hall 2004 ldquoCEO Incentives and Firm Sizerdquo Journal of
Labor Market Economics 22 767ndash798
Bebchuk Lucian A Jesse M Fried and David Walker 2002 ldquoManagerial power and
rent extraction in the design of executive compensationrdquo University of Chicago Law
Review 69 751ndash846
Bebchuk Lucian A and Jesse M Fried 2003 ldquoExecutive Compensation as an Agency
Problemrdquo Journal of Economic Perspectives 17 71ndash92
Bebchuk Lucian A and Jesse M Fried 2004 Pay without Performance The Unfulfilled
Promise of Executive Compensation Cambridge MA Harvard University Press
Bertrand Marianne and Sendhil Mullainathan 2001 ldquoAre CEOs Rewarded for Luck
The Ones Without Principals Arerdquo Quarterly Journal of Economics 116 901ndash32
Black Fischer and Myron Scholes 1973 ldquoThe Pricing of Options and Corporate
Liabilitiesrdquo Journal of Political Economy 81 637ndash54
Cadman Brian Sandy Klasa and Steve Matsunaga 2010 ldquoDeterminants of CEO Pay A
Comparison of ExecuComp and Non-ExecuComp Firmsrdquo The Accounting Review 85
1511ndash1543
Chhaochharia Vidhi and Yaniv Grinstein 2009 ldquoCEO Compensation and Board
Structurerdquo Journal of Finance 64 231ndash261
Cichello Michael S 2005 ldquoThe impact of firm size on pay-performance sensitivitiesrdquo
Journal of Corporate Finance 11 609ndash627
Davis Gerald F and E Han Kim 2007 ldquoBusiness ties and proxy voting by mutual
fundsrdquo Journal of Financial Economics 85 552ndash570
Dikolli Shane S Susan L Kulp and Karen L Sedatole 2009 ldquoTransient Institutional
Ownership and CEO Contractingrdquo Accounting Review 84 737ndash770
Edmans Alex Xavier Gabaix and Augustin Landier 2009 ldquoA Multiplicative Model of
Optimal CEO Incentives in Market Equilibriumrdquo Review of Financial Studies 22
4881ndash4917
Edmans Alex 2009 ldquoBlockholder Trading Market Efficiency and Managerial Myopiardquo
Journal of Finance 64 2481ndash2513
Ertimur Yonca Fabrizio Ferri and Volkan Muslu 2011 ldquoShareholder Activism and
CEO Payrdquo Review of Financial Studies 24 535ndash592
24
Gabaix Xavier and Augustin Landier 2008 ldquoWhy Has CEO Pay Increased So Muchrdquo
Quarterly Journal of Economics 123 49ndash100
Garen John 1994 ldquoExecutive compensation and principal-agent theoryrdquo Journal of
Political Economy 102 1175ndash1199
Gibbons Robert and Kevin J Murphy 1992 ldquoOptimal Incentive Contracts in the
Presence of Career Concerns Theory and Evidencerdquo Journal of Political Economy
100 468ndash505
Greene William H 2000 (4th
Edition) Econometric Analysis Prentice-Hall NJ
Guthrie Katherine Jan Sokolowsky and Kam-Ming Wan 2012 ldquoCEO Compensation
and Board Structure Revisitedrdquo Journal of Finance Forthcoming
Hadlock Charles R and Gerald B Lumer 1997 ldquoCompensation turnover and top
management incentives historical evidencerdquo Journal of Business 70 153ndash187
Hartzell Jay C and Laura T Starks 2003 ldquoInstitutional Investors and Executive
Compensationrdquo Journal of Finance 58 2351ndash2374
Huson Mark R Robert Parrino and Laura T Starks 2001 ldquoInternal Monitoring
Mechanisms and CEO Turnover A Long-Term Perspectiverdquo Journal of Finance 56
2265ndash2297
Jensen Michael C and Kevin J Murphy 1990 ldquoPerformance Pay and Top Management
Incentivesrdquo Journal of Political Economy 98 225ndash264
Jin Li 2002 ldquoCEO compensation diversification and incentivesrdquo Journal of Financial
Economics 66 29ndash63
Kang Qiang and Qiao Liu 2008 ldquoStock Market Information Production and Executive
Incentivesrdquo Journal of Corporate Finance 14 484ndash498
Merton Robert 1973 ldquoTheory of Rational Option Pricingrdquo Bell Journal of Economics
and Management Science 4 141-183
Morse Adair Vikram Nanda and Amit Seru 2011 ldquoAre Incentive Contracts Rigged by
Powerful CEOsrdquo Journal of Finance 66 1779ndash1821
Murphy Kevin J 1985 ldquoCorporate performance and managerial remuneration An
empirical analysisrdquo Journal of Accounting and Economics 7 11ndash42
Murphy Kevin J 1999 ldquoExecutive compensationrdquo in Orley Ashenfelter and David
Card Handbook of Labor Economics vol 3 Elsevier
Rosen Sherwin 1992 ldquoContracts and the market for executivesrdquo In Contract
Economics ed Lars Werin and Hans Wijkander 181ndash211 Cambridge MA
Blackwell
Rothberg Burton and Steven Lilien 2006 ldquoMutual Funds and Proxy Voting New
Evidence on Corporate Governancerdquo Journal of Business and Technology Law 1
157ndash184
Schaefer Scott 1998 ldquoThe dependence of payndashperformance sensitivity on the size of the
firmrdquo Review of Economics and Statistics 80 436ndash443
Yermack David 1995 ldquoDo Corporations Award CEO Stock Options Effectivelyrdquo
Journal of Financial Economics 39 237ndash69
25
Table I Univariate Descriptive Statistics
This table reports sample statistics for the key variables used in this study based on our comprehensive
ExecuComp dataset Institutional Concentration is the shares owned by the institutions with the five largest
holdings divided by the shares held by all institutions Top 5 Ownership is the proportion of shares
outstanding held by the institutions with the five largest holdings Total Institutional Ownership is the
proportion of shares held by all institutions Panel B shows the measures executive compensation The pay-
for-performance sensitivity (PPS) of option grants shows the change in executive wealth due to a $1000
change in firm in shareholder wealth Expressions for Salary and Total Compensation are in thousands of
dollars while Market Capitalization is in millions of dollars Panel C shows firm characteristics
Variable Mean Median Std Dev P10 P90
Panel A Institutional Investor Holdings
Institutional Concentration 0450 0424 0151 0282 0656
Top 5 Institutional Ownership 0229 0220 0101 0110 0355
Total Institutional Ownership 0546 0561 0206 0258 0801
Panel B Executive Compensation
PPS of Option Grants 1161 0156 3829 0 2822
Salary 36311 29462 25231 15179 67000
Total Compensation 258313 85878 980701 23875 511022
Panel C Firm Characteristics
Tobins Q 1744 1139 2738 0495 3322
Market Capitalization 560699 112094 1988108 19754 1048754
26
Table II Pairwise Correlation Coefficients
This table presents the correlations between the key variables based on our comprehensive
ExecuComp dataset Institutional Concentration is the shares owned by the institutions with the five
largest holdings divided by the shares held by all institutions Top 5 Ownership is the proportion of
shares outstanding held by the institutions with the five largest holdings Total Institutional
Ownership is the proportion of shares held by all institutions The pay-for-performance sensitivity
(PPS) of option grants is the change in executive wealth due to a $1000 change in shareholder
wealth
(1) (2) (3) (4) (5) (6) (7) (8)
(1) Institutional Concentration 1
(2) Top 5 Ownership 0268 1
(3) Total Ownership -0415 0706 1
(4) PPS of Option Grants 0122 0042 -0024 1
(5) Log Salary -0270 -0012 0171 0038 1
(6) Log Market Capitalization -0630 -0157 0261 -0156 0420 1
(7) Level Market Capitalization -0241 -0157 0007 -0058 0219 0534 1
(8) Tobins Q -0008 -0034 0002 0015 -0083 0066 0121 1
27
Table III
Descriptive Statistics by Market Capitalization Deciles
Market
Capitalization
Deciles
Market
Capitalization
($m)
Total
Compensation
($000)
Salary
($)
PPS Option
Grants
($)
Institutional
Concentration
()
Top 5 Ownership
()
Institutional
Total Ownership
()
1 (Small) 1181 76008 234700 2720 611 200 327
2 2668 90206 257955 1796 495 217 438
3 4193 105946 275120 1517 454 220 485
4 6045 129892 296067 1376 420 206 491
5 8742 155391 315864 1216 394 198 502
6 13086 191888 345942 1072 380 202 531
7 20790 239501 377021 0874 368 196 532
8 34920 314955 427808 0687 354 190 538
9 67747 450970 480319 0569 328 173 528
10 (Big) 366540 846066 614354 0297 275 138 502
28
Table IV Pay-for-Performance Sensitivity of Option Grants as a Function of Institutional Influence
Similar in construction to parts of HS (Table II) this table shows the coefficients from Tobit regressions of option grant pay-for-performance sensitivity per $1000 change in
shareholder wealth against the lagged change and contemporary change in shareholder wealth measures of institutional influence Tobinrsquos Q firm market capitalization and
dummy controls for CEO and year effects We estimate models (1) to (4) over the period 1992‒1997 using data supplied by HS and models (5) to (8) are estimated over the
whole sample Models (1) (2) (5) and (6) use Institutional Concentration as a measure of institutional influence which HS define as the shares owned by the institutions
with the five largest holdings divided by the shares held by all institutions These models control for total institutional ownership Models (3) (4) (7) and (8) use Top 5
Ownership as a measure of institutional influence which we define as the proportion of shares outstanding held by the institutions with the five largest holdings Coefficients
for industry and year controls are not presented One two and three asterisks denote significance at 10 5 and 1 levels
Hartzell and Starks Database 1992‒1997 Full Sample from ExecuComp 1992‒2002
(1) (2) (3) (4) (5) (6) (7) (8)
Change in Shareholder Wealth 77721 191781 89874 137368 -446 -218 -664 -203
(171) (429) (194) (317) (319) (168) (453) (158)
Lag1 Change in Shareholder Wealth 14446 99963 88719 8544 -144 0273 -0342 00852
(027) (212) (166) (184) (071) (015) (016) (005)
Lag1 Institutional Concentration 13994 -00692 26405 03304
(767) (032) (1402) (166)
Lag1 Total Institutional Ownership 08055 10434 07106 0937
(593) (781) (666) (898)
Lag1 Top 5 Institutional Ownership 10567 07457 12116 06293
(399) ( 283) (533) (278)
Lag1 Tobinrsquos Q 0068 00877 00647 00813 00498 00727 00508 00695
(337) (434) (319) (401) (504) (734) (51) (703)
Lag1 Market Capitalization -106202 -173515 -776 -117
(474) (742) (1011) (153)
Lag1 Log Market Capitalization -0297 -02319 -04111 -03908
(1576) (1439) (2997) (2841)
CEO Dummy 13745 13682 13765 13674 28197 28128 28155 28143
(1913) (1918) (1914) (1916) (3388) (3396) (3377) (34)
Number of Observations 33878 33878 33878 33878 97679 97679 97679 97679
29
Table V Pay-for-Performance Sensitivity of Cash Compensation (Salary plus Bonus) as a Function of Institutional Influence - Level Market Capitalization
Similar in parts to models (1) and (2) of HS (Table IV) this table displays the coefficients from OLS regressions of the change in cash compensation against the lagged
change and contemporary change in shareholder wealth measures of institutional influence Tobinrsquos Q firm market capitalization and dummy controls for CEO and year
effects We estimate models (1) to (4) over the period 1992‒1997 based on our ExecuComp sample and models (5) to (8) over the whole sample Models (1) (2) (5) and (6)
use Institutional Concentration as a measure of institutional influence which HS define as the shares owned by the institutions with the five largest holdings divided by the
shares held by all institutions These models control for total institutional ownership Models (4) and (8) use Top 5 Ownership as a measure of institutional influence which
we define as the proportion of shares outstanding held by the institutions with the 5 largest holdings Coefficients for industry and year controls are not presented One two
and three asterisks denote significance at 10 5 and 1 levels
Our Database based on ExecuComp 1992‒1997 Full Sample based on ExecuComp 1992‒2002
(1) (2) (3) (4) (5) (6) (7) (8)
Lag1 Change in Shareholder Wealth 00014 00017 00088 00089 -00006 -00007 00072 00086
(051) (062) (129) (103) (027) (029) (131) (16)
Change in Shareholder Wealth -00102 00102 00105 00032 00102 00101
(097) (292) (293) (055) (611) (608)
Change Shareholder Wealth Lag1 Institutional
Concentration
00435 00764 00363 00253
(463) (258) (689) (144)
Lag1 Change Shareholder Wealth Lag1
Institutional Concentration
-00197 -00283
(122) (197)
Lag1 Institutional Concentration -492182 -288861 400149 453623
(09) (056) (098) (114)
Lag1 Total Institutional Ownership 720688 538372 655091 559587 682069 726226
(276) (232) (281) (314) (308) (322)
Lag1 Change Shareholder Wealth Lag1 Top 5
Institutional Ownership
-0046 -00689
(091) (256)
Lag1 Top 5 Institutional Ownership 583044 1191634
(092) (25)
Lag1 Tobinrsquos Q -32761 -36806 -19422 -18023 -54283 -48681 -29534 -24243
(138) (172) (077) (078) (227) (245) (136) (111)
Lag1 Market Capitalization 00004 00004 00002 00006 00007 00008 00007 00007
(042) (044) (022) (07) (086) (092) (082) (09)
CEO Dummy 768035 765474 768334 773634 616315 61729 619028 618535
(354) (355) (356) (355) (417) (418) (419) (418)
Adjusted R-Sq 00057 0006 00052 0005 00097 00097 00098 00102
Number of Observations 41305 41305 41305 41305 85518 85518 85518 85518
30
Table VI Level of Salary as a Function of Institutional Influence
Similar in parts to models (1) and (2) of HS (Table V) this table displays the coefficients from OLS regressions of salary against the lagged change and contemporary change in
shareholder wealth measures of institutional influence Tobinrsquos Q firm market capitalization and dummy controls for CEO industry and year effects We estimate models (1) to
(4) over the period 1992‒1997 using the actual data supplied by HS and models (5) to (8) over the whole sample 1992‒2002 Models (1) (2) (5) and (6) use Institutional
Concentration as a measure of institutional influence which HS define as the shares owned by the institutions with the five largest holdings divided by the shares held by all
institutions These models control for total institutional ownership Models (3) (4) (7) and (8) use Top 5 Ownership as a measure of institutional influence which we define as
the proportion of shares outstanding held by the institutions with the 5 largest holdings Coefficients for industry and year controls are not presented One two and three
asterisks denote significance at 10 5 and 1 levels
Hartzell and Starks Database 1992‒1997 Full Sample based on ExecuComp 1992‒2002
(1) (2) (3) (4) (5) (6) (7) (8)
Change in Shareholder Wealth 04125E-4 0003 -2439E-4 0003 00021 00012 00024 00012
(057) (464) (031) (468) (1004) (72) (1041) (706)
Lag1 Change in Shareholder Wealth -00062 7856E-4 -0007 9003E-4 -00014 00002 -00014 00003
(594) (098) (592) (112) (522) (123) (5) (145)
Lag1 Institutional Concentration -2475087 555134 -2734983 943317
(3245) (710) (4965) (1671)
Lag1 Total Institutional Ownership 1020109 217488 936337 252464
(1904) (432) (2471) (704)
Lag1 Top 5 Institutional Ownership -148355 31354 -87926 423166
(139) (332) (118) (643)
Lag1 Tobinrsquos Q -195348 -252191 -192344 -25158 -119837 -146373 -122132 -143821
(2812) (3762) (2721) (3752) (248) (274) (2583) (2727)
Lag1 Market Capitalization 844E-06 103E-5 00034 00039
(2781) (3051) (2874) (3104)
Lag1 Log Market Capitalization 750021 723625 776657 732279
(7511) (9285) (11224) (14039)
CEO Dummy 2683978 2702238 2668244 2702417 3281133 3285103 3292175 3283145
(8625) (9277) (8246) (9271) (12739) (13616) (12336) (13605)
Adjusted R-Sq 04604 05302 04187 05295 03856 04588 03476 04574
Number of Observations 36280 36280 36280 36280 97679 97679 97679 97679
31
Table VII Level of Total Direct Compensation as a Function of Institutional Influence
Similar in parts to models (3) and (5) of HS (Table V) this table displays the coefficients from OLS regressions of total direct compensation against the lagged change and
contemporary change in shareholder wealth measures of institutional influence Tobinrsquos Q firm market capitalization and dummy controls for CEO industry and year effects
We estimate models (1) to (4) over the period 1992‒1997 and models (5) to (8) over the whole sample Models (1) (2) (5) and (6) use Institutional Concentration as a measure
of institutional influence which HS define as the shares owned by the institutions with the five largest holdings divided by the shares held by all institutions These models
control for total institutional ownership Models (3) (4) (7) and (8) use Top 5 Ownership as a measure of institutional influence which we define as the proportion of shares
outstanding held by the institutions with the five largest holdings Coefficients for industry and year controls are not presented One two and three asterisks denote significance
at 10 5 and 1 levels
Hartzell and Starks Database 1992‒1997 Full Sample from ExecuComp 1992‒2002
(1) (2) (3) (4) (5) (6) (7) (8)
Change in Shareholder Wealth 005779 007006 005282 006791 00423 00148 0046 00138
(415) (581) (369) (565) (203) (07) (22) (066)
Lag1 Change in Shareholder Wealth 003182 00714 00272 00715 00613 0126 00614 01278
(126) (298) (107) (299) (275) (488) (276) (494)
Lag1 Institutional Concentration -1608633 3248998 -3424892 3334186
(151) (283) (1172) (118)
Lag1 Total Institutional Ownership 104013 5473128 1531013 -2926764
(1645) (831) (876) (158)
Lag1 Top 5 Institutional Ownership 188585 3927986 -2159177 -2640879
(016) (343) (067) (092)
Lag1 Tobinrsquos Q -445692 -803612 -454859 -833566 3257027 2971201 3213936 3086205
(427) (791) (432) (831) (83) (77) (815) (793)
Lag1 Market Capitalization 49E-5 615E-5 01031 01088
(113) (1365) (1555) (168)
Lag1 Log Market Capitalization 4683497 4759377 1645156 1427856
(2775) (3469) (3679) (4239)
CEO Dummy 1501307 1508886 1493688 1508736 4713079 4713629 47278 4703908
(2648) (2684) (2618) (2683) (282) (2828) (2819) (2824)
Adjusted R-Sq 01861 02060 01695 02051 01071 01113 01033 011
Number of Observations 33229 33229 33229 33229 97679 97679 97679 97679
32
Table VIII Two Stage Least Squares (Reverse Causality) Tests - Executive Compensation Measures as a Function of Institutional Concentration
This table shows the coefficients from the Two-Stage Least Squares (2SLS) regressions of executive compensation measures against the lagged change
and contemporary change in shareholder wealth instrumented institutional concentration and total institutional ownership Tobinrsquos Q the natural
logarithm of firm market capitalization and dummy variable controls for CEO and year effects We instrument measures of institutional concentration
and total institutional ownership by orthogonalizing with respect to option grant pay-for-performance sensitivity salary and total compensation
Coefficients for industry and year controls are not presented First stage results are not presented One two and three asterisks denote significance at
10 5 and 1 levels
ExecuComp Sample 1992‒1997 Full Sample from ExecuComp 1992‒2002
PPSOG Salary TC PPSOG Salary TC
Change in Shareholder Wealth 144E-05 0004 02133 -409E-06 00011 00157
(308) (775) (596) (691) (638) (068)
Lag1 Change in Shareholder Wealth -240E-05 -00004 00866 118E-06 00002 01422
(131) (087) (346) (097) (116) (51)
Lag1 Institutional Concentration -00898 668681 3263143 02024 916656 2856417
(052) (801) (176) (163) (1545) (1004)
Lag1 Total Institutional Ownership 01021 -43997 2512587 02983 -104171 -1028347
(09) (084) (153) (4) (269) (487)
Lag1 Tobinrsquos Q 00754 -225606 119297 00655 -158657 2997205
(596) (1649) (053) (834) (2535) (727)
Lag1 Log Market Capitalization -04234 828713 7864046 -04002 825106 1685638
(2195) (8132) (2977) (4106) (11133) (373)
CEO Dummy 15848 3214117 2950274 19149 3205862 4529101
(2223) (9389) (2255) (3802) (1331) (2758)
Overall R-Sq -- 04981 01206 -- 04738 01178
Number of Observations 41305 41305 41305 85518 85518 85518
33
Table IX Two Stage Least Squares (Reverse Causality) Tests - Executive Compensation Measures as a Function of Top 5 Ownership
This table shows the coefficients from 2SLS regressions of executive compensation measures against the lagged change and contemporary change
in shareholder wealth instrumented Top 5 institutional ownership Tobinrsquos Q natural logarithm of firm market capitalization and dummy variable
controls for CEO and year effects We instrument Top 5 institutional ownership by orthogonalizing with respect to option grant pay-for-
performance sensitivity salary and total compensation Coefficients for industry and year controls are not presented First stage results are not
presented One two and three asterisks denote significance at 10 5 and 1 levels
ExecuComp sample 1992‒1997 Full Sample 1992‒2002
PPSOG Salary TC PPSOG Salary TC
Change in Shareholder Wealth 136E-05 00041 02128 -407E-06 00011 00146
(293) (79) (596) (692) (619) (063)
Lag1 Change in Shareholder Wealth -241E-05 -00003 00868 116E-06 00003 01439
(133) (07) (347) (096) (142) (514)
Lag1 Top 5 Institutional Ownership -03774 -187512 -3848531 00179 -104261 -126659
(163) (186) (166) (011) (149) (436)
Lag1 Tobinrsquos Q 00731 -223886 10074 00642 -155467 3115073
(573) (1631) (045) (819) (2509) (75)
Lag1 Log Market Capitalization -0416 775637 7696302 -04015 764632 1462536
(2345) (10084) (3766) (3688) (13708) (4072)
CEO Dummy 15854 3211991 2950158 19152 3203152 4519385
(2223) (9381) (2254) (3804) (133) (2752)
Overall R-Sq -- 04967 01202 -- 04722 01168
Number of Observations 41305 41305 41305 85518 85518 85518
34
Table X Specification Tests utilizing Box-Cox Transformation of Market Capitalization
This table presents results from a Box-Cox transformation of market capitalization In
each model we specify transformation of market capitalization only Panel A shows
results for the period 1992‒1997 and Panel B shows results for the entire period
1992‒2002 The first part of each panel shows the Lambda coefficient that applies to
the transformed market capitalization The second part of each panel shows log
likelihood scores with LR Chi2 Statistics in parentheses ‒ these test for the
appropriateness of a linear or log model specification
Panel A Our ExecuComp sample 1992‒1997
(1) (2) (3)
PPSOG Salary TC
Lambda -05143 01881 02876
(-1914) (2413) (1329)
Test H0 (Log Likelihood Score)
Lambda = -1 -1279748 -3178575 -4795326
(23866) (85952) (99688)
Lambda = 0 -1280152 -3138456 -4791159
(31934) (57147) (16339)
Lambda = 1 -1283157 -3163807 -4793506
(92039) (564156) (63292)
Panel B Full ExecuComp sample 1992‒2002
(1) (2) (4)
PPSOG Salary TC
Lambda -044 01853 04036
(-3015) (40) (3666)
Test H0 (Log Likelihood Score)
Lambda = 0 -2688334 -6476494 -10384490
(66599) (160356) (126236)
Lambda = 1 -2694886 -6538506 -10386795
(197644) (1400605) (172345)
35
Figure 2 Institutional Ownership Measures (LHA) vs Total
Compensation (RHA) by Market Capitalization
0
10
20
30
40
50
60
70
1 2 3 4 5 6 7 8 9 10
$00
$10
$20
$30
$40
$50
$60
$70
$80
$90
Millio
ns
InstitutionalConcentration (LHS)Top5 Ownership (LHS)
Institutional TotalOwnership (LHS)Total Compensation(RHS)
Page 22
22
preference for highly remunerated executives in large companies and thus by reverse
causality Moreover our findings are not consistent with the rent-extraction hypothesis
suggesting that firms systematically overpay executives in the absence of effective
monitoring by outside parties such as institutional investors
23
References
Almazan Andres Jay C Hartzell and Laura T Starks 2005 ldquoActive Institutional
Shareholders and Costs of Monitoring Evidence from Executive Compensationrdquo
Financial Management Winter 5ndash34
Aneuryn-Evans Gwyn and Angus Deaton 1980 ldquoTesting Linear versus Logarithmic
Regression Modelsrdquo Review of Economic Studies 47 (1) Econometrics Issue 275ndash
291
Babenko Ilona 2009 ldquoShare Repurchases and Pay-Performance Sensitivity of Employee
Compensation Contractsrdquo Journal of Finance 64117-150
Baker George P and Brian J Hall 2004 ldquoCEO Incentives and Firm Sizerdquo Journal of
Labor Market Economics 22 767ndash798
Bebchuk Lucian A Jesse M Fried and David Walker 2002 ldquoManagerial power and
rent extraction in the design of executive compensationrdquo University of Chicago Law
Review 69 751ndash846
Bebchuk Lucian A and Jesse M Fried 2003 ldquoExecutive Compensation as an Agency
Problemrdquo Journal of Economic Perspectives 17 71ndash92
Bebchuk Lucian A and Jesse M Fried 2004 Pay without Performance The Unfulfilled
Promise of Executive Compensation Cambridge MA Harvard University Press
Bertrand Marianne and Sendhil Mullainathan 2001 ldquoAre CEOs Rewarded for Luck
The Ones Without Principals Arerdquo Quarterly Journal of Economics 116 901ndash32
Black Fischer and Myron Scholes 1973 ldquoThe Pricing of Options and Corporate
Liabilitiesrdquo Journal of Political Economy 81 637ndash54
Cadman Brian Sandy Klasa and Steve Matsunaga 2010 ldquoDeterminants of CEO Pay A
Comparison of ExecuComp and Non-ExecuComp Firmsrdquo The Accounting Review 85
1511ndash1543
Chhaochharia Vidhi and Yaniv Grinstein 2009 ldquoCEO Compensation and Board
Structurerdquo Journal of Finance 64 231ndash261
Cichello Michael S 2005 ldquoThe impact of firm size on pay-performance sensitivitiesrdquo
Journal of Corporate Finance 11 609ndash627
Davis Gerald F and E Han Kim 2007 ldquoBusiness ties and proxy voting by mutual
fundsrdquo Journal of Financial Economics 85 552ndash570
Dikolli Shane S Susan L Kulp and Karen L Sedatole 2009 ldquoTransient Institutional
Ownership and CEO Contractingrdquo Accounting Review 84 737ndash770
Edmans Alex Xavier Gabaix and Augustin Landier 2009 ldquoA Multiplicative Model of
Optimal CEO Incentives in Market Equilibriumrdquo Review of Financial Studies 22
4881ndash4917
Edmans Alex 2009 ldquoBlockholder Trading Market Efficiency and Managerial Myopiardquo
Journal of Finance 64 2481ndash2513
Ertimur Yonca Fabrizio Ferri and Volkan Muslu 2011 ldquoShareholder Activism and
CEO Payrdquo Review of Financial Studies 24 535ndash592
24
Gabaix Xavier and Augustin Landier 2008 ldquoWhy Has CEO Pay Increased So Muchrdquo
Quarterly Journal of Economics 123 49ndash100
Garen John 1994 ldquoExecutive compensation and principal-agent theoryrdquo Journal of
Political Economy 102 1175ndash1199
Gibbons Robert and Kevin J Murphy 1992 ldquoOptimal Incentive Contracts in the
Presence of Career Concerns Theory and Evidencerdquo Journal of Political Economy
100 468ndash505
Greene William H 2000 (4th
Edition) Econometric Analysis Prentice-Hall NJ
Guthrie Katherine Jan Sokolowsky and Kam-Ming Wan 2012 ldquoCEO Compensation
and Board Structure Revisitedrdquo Journal of Finance Forthcoming
Hadlock Charles R and Gerald B Lumer 1997 ldquoCompensation turnover and top
management incentives historical evidencerdquo Journal of Business 70 153ndash187
Hartzell Jay C and Laura T Starks 2003 ldquoInstitutional Investors and Executive
Compensationrdquo Journal of Finance 58 2351ndash2374
Huson Mark R Robert Parrino and Laura T Starks 2001 ldquoInternal Monitoring
Mechanisms and CEO Turnover A Long-Term Perspectiverdquo Journal of Finance 56
2265ndash2297
Jensen Michael C and Kevin J Murphy 1990 ldquoPerformance Pay and Top Management
Incentivesrdquo Journal of Political Economy 98 225ndash264
Jin Li 2002 ldquoCEO compensation diversification and incentivesrdquo Journal of Financial
Economics 66 29ndash63
Kang Qiang and Qiao Liu 2008 ldquoStock Market Information Production and Executive
Incentivesrdquo Journal of Corporate Finance 14 484ndash498
Merton Robert 1973 ldquoTheory of Rational Option Pricingrdquo Bell Journal of Economics
and Management Science 4 141-183
Morse Adair Vikram Nanda and Amit Seru 2011 ldquoAre Incentive Contracts Rigged by
Powerful CEOsrdquo Journal of Finance 66 1779ndash1821
Murphy Kevin J 1985 ldquoCorporate performance and managerial remuneration An
empirical analysisrdquo Journal of Accounting and Economics 7 11ndash42
Murphy Kevin J 1999 ldquoExecutive compensationrdquo in Orley Ashenfelter and David
Card Handbook of Labor Economics vol 3 Elsevier
Rosen Sherwin 1992 ldquoContracts and the market for executivesrdquo In Contract
Economics ed Lars Werin and Hans Wijkander 181ndash211 Cambridge MA
Blackwell
Rothberg Burton and Steven Lilien 2006 ldquoMutual Funds and Proxy Voting New
Evidence on Corporate Governancerdquo Journal of Business and Technology Law 1
157ndash184
Schaefer Scott 1998 ldquoThe dependence of payndashperformance sensitivity on the size of the
firmrdquo Review of Economics and Statistics 80 436ndash443
Yermack David 1995 ldquoDo Corporations Award CEO Stock Options Effectivelyrdquo
Journal of Financial Economics 39 237ndash69
25
Table I Univariate Descriptive Statistics
This table reports sample statistics for the key variables used in this study based on our comprehensive
ExecuComp dataset Institutional Concentration is the shares owned by the institutions with the five largest
holdings divided by the shares held by all institutions Top 5 Ownership is the proportion of shares
outstanding held by the institutions with the five largest holdings Total Institutional Ownership is the
proportion of shares held by all institutions Panel B shows the measures executive compensation The pay-
for-performance sensitivity (PPS) of option grants shows the change in executive wealth due to a $1000
change in firm in shareholder wealth Expressions for Salary and Total Compensation are in thousands of
dollars while Market Capitalization is in millions of dollars Panel C shows firm characteristics
Variable Mean Median Std Dev P10 P90
Panel A Institutional Investor Holdings
Institutional Concentration 0450 0424 0151 0282 0656
Top 5 Institutional Ownership 0229 0220 0101 0110 0355
Total Institutional Ownership 0546 0561 0206 0258 0801
Panel B Executive Compensation
PPS of Option Grants 1161 0156 3829 0 2822
Salary 36311 29462 25231 15179 67000
Total Compensation 258313 85878 980701 23875 511022
Panel C Firm Characteristics
Tobins Q 1744 1139 2738 0495 3322
Market Capitalization 560699 112094 1988108 19754 1048754
26
Table II Pairwise Correlation Coefficients
This table presents the correlations between the key variables based on our comprehensive
ExecuComp dataset Institutional Concentration is the shares owned by the institutions with the five
largest holdings divided by the shares held by all institutions Top 5 Ownership is the proportion of
shares outstanding held by the institutions with the five largest holdings Total Institutional
Ownership is the proportion of shares held by all institutions The pay-for-performance sensitivity
(PPS) of option grants is the change in executive wealth due to a $1000 change in shareholder
wealth
(1) (2) (3) (4) (5) (6) (7) (8)
(1) Institutional Concentration 1
(2) Top 5 Ownership 0268 1
(3) Total Ownership -0415 0706 1
(4) PPS of Option Grants 0122 0042 -0024 1
(5) Log Salary -0270 -0012 0171 0038 1
(6) Log Market Capitalization -0630 -0157 0261 -0156 0420 1
(7) Level Market Capitalization -0241 -0157 0007 -0058 0219 0534 1
(8) Tobins Q -0008 -0034 0002 0015 -0083 0066 0121 1
27
Table III
Descriptive Statistics by Market Capitalization Deciles
Market
Capitalization
Deciles
Market
Capitalization
($m)
Total
Compensation
($000)
Salary
($)
PPS Option
Grants
($)
Institutional
Concentration
()
Top 5 Ownership
()
Institutional
Total Ownership
()
1 (Small) 1181 76008 234700 2720 611 200 327
2 2668 90206 257955 1796 495 217 438
3 4193 105946 275120 1517 454 220 485
4 6045 129892 296067 1376 420 206 491
5 8742 155391 315864 1216 394 198 502
6 13086 191888 345942 1072 380 202 531
7 20790 239501 377021 0874 368 196 532
8 34920 314955 427808 0687 354 190 538
9 67747 450970 480319 0569 328 173 528
10 (Big) 366540 846066 614354 0297 275 138 502
28
Table IV Pay-for-Performance Sensitivity of Option Grants as a Function of Institutional Influence
Similar in construction to parts of HS (Table II) this table shows the coefficients from Tobit regressions of option grant pay-for-performance sensitivity per $1000 change in
shareholder wealth against the lagged change and contemporary change in shareholder wealth measures of institutional influence Tobinrsquos Q firm market capitalization and
dummy controls for CEO and year effects We estimate models (1) to (4) over the period 1992‒1997 using data supplied by HS and models (5) to (8) are estimated over the
whole sample Models (1) (2) (5) and (6) use Institutional Concentration as a measure of institutional influence which HS define as the shares owned by the institutions
with the five largest holdings divided by the shares held by all institutions These models control for total institutional ownership Models (3) (4) (7) and (8) use Top 5
Ownership as a measure of institutional influence which we define as the proportion of shares outstanding held by the institutions with the five largest holdings Coefficients
for industry and year controls are not presented One two and three asterisks denote significance at 10 5 and 1 levels
Hartzell and Starks Database 1992‒1997 Full Sample from ExecuComp 1992‒2002
(1) (2) (3) (4) (5) (6) (7) (8)
Change in Shareholder Wealth 77721 191781 89874 137368 -446 -218 -664 -203
(171) (429) (194) (317) (319) (168) (453) (158)
Lag1 Change in Shareholder Wealth 14446 99963 88719 8544 -144 0273 -0342 00852
(027) (212) (166) (184) (071) (015) (016) (005)
Lag1 Institutional Concentration 13994 -00692 26405 03304
(767) (032) (1402) (166)
Lag1 Total Institutional Ownership 08055 10434 07106 0937
(593) (781) (666) (898)
Lag1 Top 5 Institutional Ownership 10567 07457 12116 06293
(399) ( 283) (533) (278)
Lag1 Tobinrsquos Q 0068 00877 00647 00813 00498 00727 00508 00695
(337) (434) (319) (401) (504) (734) (51) (703)
Lag1 Market Capitalization -106202 -173515 -776 -117
(474) (742) (1011) (153)
Lag1 Log Market Capitalization -0297 -02319 -04111 -03908
(1576) (1439) (2997) (2841)
CEO Dummy 13745 13682 13765 13674 28197 28128 28155 28143
(1913) (1918) (1914) (1916) (3388) (3396) (3377) (34)
Number of Observations 33878 33878 33878 33878 97679 97679 97679 97679
29
Table V Pay-for-Performance Sensitivity of Cash Compensation (Salary plus Bonus) as a Function of Institutional Influence - Level Market Capitalization
Similar in parts to models (1) and (2) of HS (Table IV) this table displays the coefficients from OLS regressions of the change in cash compensation against the lagged
change and contemporary change in shareholder wealth measures of institutional influence Tobinrsquos Q firm market capitalization and dummy controls for CEO and year
effects We estimate models (1) to (4) over the period 1992‒1997 based on our ExecuComp sample and models (5) to (8) over the whole sample Models (1) (2) (5) and (6)
use Institutional Concentration as a measure of institutional influence which HS define as the shares owned by the institutions with the five largest holdings divided by the
shares held by all institutions These models control for total institutional ownership Models (4) and (8) use Top 5 Ownership as a measure of institutional influence which
we define as the proportion of shares outstanding held by the institutions with the 5 largest holdings Coefficients for industry and year controls are not presented One two
and three asterisks denote significance at 10 5 and 1 levels
Our Database based on ExecuComp 1992‒1997 Full Sample based on ExecuComp 1992‒2002
(1) (2) (3) (4) (5) (6) (7) (8)
Lag1 Change in Shareholder Wealth 00014 00017 00088 00089 -00006 -00007 00072 00086
(051) (062) (129) (103) (027) (029) (131) (16)
Change in Shareholder Wealth -00102 00102 00105 00032 00102 00101
(097) (292) (293) (055) (611) (608)
Change Shareholder Wealth Lag1 Institutional
Concentration
00435 00764 00363 00253
(463) (258) (689) (144)
Lag1 Change Shareholder Wealth Lag1
Institutional Concentration
-00197 -00283
(122) (197)
Lag1 Institutional Concentration -492182 -288861 400149 453623
(09) (056) (098) (114)
Lag1 Total Institutional Ownership 720688 538372 655091 559587 682069 726226
(276) (232) (281) (314) (308) (322)
Lag1 Change Shareholder Wealth Lag1 Top 5
Institutional Ownership
-0046 -00689
(091) (256)
Lag1 Top 5 Institutional Ownership 583044 1191634
(092) (25)
Lag1 Tobinrsquos Q -32761 -36806 -19422 -18023 -54283 -48681 -29534 -24243
(138) (172) (077) (078) (227) (245) (136) (111)
Lag1 Market Capitalization 00004 00004 00002 00006 00007 00008 00007 00007
(042) (044) (022) (07) (086) (092) (082) (09)
CEO Dummy 768035 765474 768334 773634 616315 61729 619028 618535
(354) (355) (356) (355) (417) (418) (419) (418)
Adjusted R-Sq 00057 0006 00052 0005 00097 00097 00098 00102
Number of Observations 41305 41305 41305 41305 85518 85518 85518 85518
30
Table VI Level of Salary as a Function of Institutional Influence
Similar in parts to models (1) and (2) of HS (Table V) this table displays the coefficients from OLS regressions of salary against the lagged change and contemporary change in
shareholder wealth measures of institutional influence Tobinrsquos Q firm market capitalization and dummy controls for CEO industry and year effects We estimate models (1) to
(4) over the period 1992‒1997 using the actual data supplied by HS and models (5) to (8) over the whole sample 1992‒2002 Models (1) (2) (5) and (6) use Institutional
Concentration as a measure of institutional influence which HS define as the shares owned by the institutions with the five largest holdings divided by the shares held by all
institutions These models control for total institutional ownership Models (3) (4) (7) and (8) use Top 5 Ownership as a measure of institutional influence which we define as
the proportion of shares outstanding held by the institutions with the 5 largest holdings Coefficients for industry and year controls are not presented One two and three
asterisks denote significance at 10 5 and 1 levels
Hartzell and Starks Database 1992‒1997 Full Sample based on ExecuComp 1992‒2002
(1) (2) (3) (4) (5) (6) (7) (8)
Change in Shareholder Wealth 04125E-4 0003 -2439E-4 0003 00021 00012 00024 00012
(057) (464) (031) (468) (1004) (72) (1041) (706)
Lag1 Change in Shareholder Wealth -00062 7856E-4 -0007 9003E-4 -00014 00002 -00014 00003
(594) (098) (592) (112) (522) (123) (5) (145)
Lag1 Institutional Concentration -2475087 555134 -2734983 943317
(3245) (710) (4965) (1671)
Lag1 Total Institutional Ownership 1020109 217488 936337 252464
(1904) (432) (2471) (704)
Lag1 Top 5 Institutional Ownership -148355 31354 -87926 423166
(139) (332) (118) (643)
Lag1 Tobinrsquos Q -195348 -252191 -192344 -25158 -119837 -146373 -122132 -143821
(2812) (3762) (2721) (3752) (248) (274) (2583) (2727)
Lag1 Market Capitalization 844E-06 103E-5 00034 00039
(2781) (3051) (2874) (3104)
Lag1 Log Market Capitalization 750021 723625 776657 732279
(7511) (9285) (11224) (14039)
CEO Dummy 2683978 2702238 2668244 2702417 3281133 3285103 3292175 3283145
(8625) (9277) (8246) (9271) (12739) (13616) (12336) (13605)
Adjusted R-Sq 04604 05302 04187 05295 03856 04588 03476 04574
Number of Observations 36280 36280 36280 36280 97679 97679 97679 97679
31
Table VII Level of Total Direct Compensation as a Function of Institutional Influence
Similar in parts to models (3) and (5) of HS (Table V) this table displays the coefficients from OLS regressions of total direct compensation against the lagged change and
contemporary change in shareholder wealth measures of institutional influence Tobinrsquos Q firm market capitalization and dummy controls for CEO industry and year effects
We estimate models (1) to (4) over the period 1992‒1997 and models (5) to (8) over the whole sample Models (1) (2) (5) and (6) use Institutional Concentration as a measure
of institutional influence which HS define as the shares owned by the institutions with the five largest holdings divided by the shares held by all institutions These models
control for total institutional ownership Models (3) (4) (7) and (8) use Top 5 Ownership as a measure of institutional influence which we define as the proportion of shares
outstanding held by the institutions with the five largest holdings Coefficients for industry and year controls are not presented One two and three asterisks denote significance
at 10 5 and 1 levels
Hartzell and Starks Database 1992‒1997 Full Sample from ExecuComp 1992‒2002
(1) (2) (3) (4) (5) (6) (7) (8)
Change in Shareholder Wealth 005779 007006 005282 006791 00423 00148 0046 00138
(415) (581) (369) (565) (203) (07) (22) (066)
Lag1 Change in Shareholder Wealth 003182 00714 00272 00715 00613 0126 00614 01278
(126) (298) (107) (299) (275) (488) (276) (494)
Lag1 Institutional Concentration -1608633 3248998 -3424892 3334186
(151) (283) (1172) (118)
Lag1 Total Institutional Ownership 104013 5473128 1531013 -2926764
(1645) (831) (876) (158)
Lag1 Top 5 Institutional Ownership 188585 3927986 -2159177 -2640879
(016) (343) (067) (092)
Lag1 Tobinrsquos Q -445692 -803612 -454859 -833566 3257027 2971201 3213936 3086205
(427) (791) (432) (831) (83) (77) (815) (793)
Lag1 Market Capitalization 49E-5 615E-5 01031 01088
(113) (1365) (1555) (168)
Lag1 Log Market Capitalization 4683497 4759377 1645156 1427856
(2775) (3469) (3679) (4239)
CEO Dummy 1501307 1508886 1493688 1508736 4713079 4713629 47278 4703908
(2648) (2684) (2618) (2683) (282) (2828) (2819) (2824)
Adjusted R-Sq 01861 02060 01695 02051 01071 01113 01033 011
Number of Observations 33229 33229 33229 33229 97679 97679 97679 97679
32
Table VIII Two Stage Least Squares (Reverse Causality) Tests - Executive Compensation Measures as a Function of Institutional Concentration
This table shows the coefficients from the Two-Stage Least Squares (2SLS) regressions of executive compensation measures against the lagged change
and contemporary change in shareholder wealth instrumented institutional concentration and total institutional ownership Tobinrsquos Q the natural
logarithm of firm market capitalization and dummy variable controls for CEO and year effects We instrument measures of institutional concentration
and total institutional ownership by orthogonalizing with respect to option grant pay-for-performance sensitivity salary and total compensation
Coefficients for industry and year controls are not presented First stage results are not presented One two and three asterisks denote significance at
10 5 and 1 levels
ExecuComp Sample 1992‒1997 Full Sample from ExecuComp 1992‒2002
PPSOG Salary TC PPSOG Salary TC
Change in Shareholder Wealth 144E-05 0004 02133 -409E-06 00011 00157
(308) (775) (596) (691) (638) (068)
Lag1 Change in Shareholder Wealth -240E-05 -00004 00866 118E-06 00002 01422
(131) (087) (346) (097) (116) (51)
Lag1 Institutional Concentration -00898 668681 3263143 02024 916656 2856417
(052) (801) (176) (163) (1545) (1004)
Lag1 Total Institutional Ownership 01021 -43997 2512587 02983 -104171 -1028347
(09) (084) (153) (4) (269) (487)
Lag1 Tobinrsquos Q 00754 -225606 119297 00655 -158657 2997205
(596) (1649) (053) (834) (2535) (727)
Lag1 Log Market Capitalization -04234 828713 7864046 -04002 825106 1685638
(2195) (8132) (2977) (4106) (11133) (373)
CEO Dummy 15848 3214117 2950274 19149 3205862 4529101
(2223) (9389) (2255) (3802) (1331) (2758)
Overall R-Sq -- 04981 01206 -- 04738 01178
Number of Observations 41305 41305 41305 85518 85518 85518
33
Table IX Two Stage Least Squares (Reverse Causality) Tests - Executive Compensation Measures as a Function of Top 5 Ownership
This table shows the coefficients from 2SLS regressions of executive compensation measures against the lagged change and contemporary change
in shareholder wealth instrumented Top 5 institutional ownership Tobinrsquos Q natural logarithm of firm market capitalization and dummy variable
controls for CEO and year effects We instrument Top 5 institutional ownership by orthogonalizing with respect to option grant pay-for-
performance sensitivity salary and total compensation Coefficients for industry and year controls are not presented First stage results are not
presented One two and three asterisks denote significance at 10 5 and 1 levels
ExecuComp sample 1992‒1997 Full Sample 1992‒2002
PPSOG Salary TC PPSOG Salary TC
Change in Shareholder Wealth 136E-05 00041 02128 -407E-06 00011 00146
(293) (79) (596) (692) (619) (063)
Lag1 Change in Shareholder Wealth -241E-05 -00003 00868 116E-06 00003 01439
(133) (07) (347) (096) (142) (514)
Lag1 Top 5 Institutional Ownership -03774 -187512 -3848531 00179 -104261 -126659
(163) (186) (166) (011) (149) (436)
Lag1 Tobinrsquos Q 00731 -223886 10074 00642 -155467 3115073
(573) (1631) (045) (819) (2509) (75)
Lag1 Log Market Capitalization -0416 775637 7696302 -04015 764632 1462536
(2345) (10084) (3766) (3688) (13708) (4072)
CEO Dummy 15854 3211991 2950158 19152 3203152 4519385
(2223) (9381) (2254) (3804) (133) (2752)
Overall R-Sq -- 04967 01202 -- 04722 01168
Number of Observations 41305 41305 41305 85518 85518 85518
34
Table X Specification Tests utilizing Box-Cox Transformation of Market Capitalization
This table presents results from a Box-Cox transformation of market capitalization In
each model we specify transformation of market capitalization only Panel A shows
results for the period 1992‒1997 and Panel B shows results for the entire period
1992‒2002 The first part of each panel shows the Lambda coefficient that applies to
the transformed market capitalization The second part of each panel shows log
likelihood scores with LR Chi2 Statistics in parentheses ‒ these test for the
appropriateness of a linear or log model specification
Panel A Our ExecuComp sample 1992‒1997
(1) (2) (3)
PPSOG Salary TC
Lambda -05143 01881 02876
(-1914) (2413) (1329)
Test H0 (Log Likelihood Score)
Lambda = -1 -1279748 -3178575 -4795326
(23866) (85952) (99688)
Lambda = 0 -1280152 -3138456 -4791159
(31934) (57147) (16339)
Lambda = 1 -1283157 -3163807 -4793506
(92039) (564156) (63292)
Panel B Full ExecuComp sample 1992‒2002
(1) (2) (4)
PPSOG Salary TC
Lambda -044 01853 04036
(-3015) (40) (3666)
Test H0 (Log Likelihood Score)
Lambda = 0 -2688334 -6476494 -10384490
(66599) (160356) (126236)
Lambda = 1 -2694886 -6538506 -10386795
(197644) (1400605) (172345)
35
Figure 2 Institutional Ownership Measures (LHA) vs Total
Compensation (RHA) by Market Capitalization
0
10
20
30
40
50
60
70
1 2 3 4 5 6 7 8 9 10
$00
$10
$20
$30
$40
$50
$60
$70
$80
$90
Millio
ns
InstitutionalConcentration (LHS)Top5 Ownership (LHS)
Institutional TotalOwnership (LHS)Total Compensation(RHS)
Page 23
23
References
Almazan Andres Jay C Hartzell and Laura T Starks 2005 ldquoActive Institutional
Shareholders and Costs of Monitoring Evidence from Executive Compensationrdquo
Financial Management Winter 5ndash34
Aneuryn-Evans Gwyn and Angus Deaton 1980 ldquoTesting Linear versus Logarithmic
Regression Modelsrdquo Review of Economic Studies 47 (1) Econometrics Issue 275ndash
291
Babenko Ilona 2009 ldquoShare Repurchases and Pay-Performance Sensitivity of Employee
Compensation Contractsrdquo Journal of Finance 64117-150
Baker George P and Brian J Hall 2004 ldquoCEO Incentives and Firm Sizerdquo Journal of
Labor Market Economics 22 767ndash798
Bebchuk Lucian A Jesse M Fried and David Walker 2002 ldquoManagerial power and
rent extraction in the design of executive compensationrdquo University of Chicago Law
Review 69 751ndash846
Bebchuk Lucian A and Jesse M Fried 2003 ldquoExecutive Compensation as an Agency
Problemrdquo Journal of Economic Perspectives 17 71ndash92
Bebchuk Lucian A and Jesse M Fried 2004 Pay without Performance The Unfulfilled
Promise of Executive Compensation Cambridge MA Harvard University Press
Bertrand Marianne and Sendhil Mullainathan 2001 ldquoAre CEOs Rewarded for Luck
The Ones Without Principals Arerdquo Quarterly Journal of Economics 116 901ndash32
Black Fischer and Myron Scholes 1973 ldquoThe Pricing of Options and Corporate
Liabilitiesrdquo Journal of Political Economy 81 637ndash54
Cadman Brian Sandy Klasa and Steve Matsunaga 2010 ldquoDeterminants of CEO Pay A
Comparison of ExecuComp and Non-ExecuComp Firmsrdquo The Accounting Review 85
1511ndash1543
Chhaochharia Vidhi and Yaniv Grinstein 2009 ldquoCEO Compensation and Board
Structurerdquo Journal of Finance 64 231ndash261
Cichello Michael S 2005 ldquoThe impact of firm size on pay-performance sensitivitiesrdquo
Journal of Corporate Finance 11 609ndash627
Davis Gerald F and E Han Kim 2007 ldquoBusiness ties and proxy voting by mutual
fundsrdquo Journal of Financial Economics 85 552ndash570
Dikolli Shane S Susan L Kulp and Karen L Sedatole 2009 ldquoTransient Institutional
Ownership and CEO Contractingrdquo Accounting Review 84 737ndash770
Edmans Alex Xavier Gabaix and Augustin Landier 2009 ldquoA Multiplicative Model of
Optimal CEO Incentives in Market Equilibriumrdquo Review of Financial Studies 22
4881ndash4917
Edmans Alex 2009 ldquoBlockholder Trading Market Efficiency and Managerial Myopiardquo
Journal of Finance 64 2481ndash2513
Ertimur Yonca Fabrizio Ferri and Volkan Muslu 2011 ldquoShareholder Activism and
CEO Payrdquo Review of Financial Studies 24 535ndash592
24
Gabaix Xavier and Augustin Landier 2008 ldquoWhy Has CEO Pay Increased So Muchrdquo
Quarterly Journal of Economics 123 49ndash100
Garen John 1994 ldquoExecutive compensation and principal-agent theoryrdquo Journal of
Political Economy 102 1175ndash1199
Gibbons Robert and Kevin J Murphy 1992 ldquoOptimal Incentive Contracts in the
Presence of Career Concerns Theory and Evidencerdquo Journal of Political Economy
100 468ndash505
Greene William H 2000 (4th
Edition) Econometric Analysis Prentice-Hall NJ
Guthrie Katherine Jan Sokolowsky and Kam-Ming Wan 2012 ldquoCEO Compensation
and Board Structure Revisitedrdquo Journal of Finance Forthcoming
Hadlock Charles R and Gerald B Lumer 1997 ldquoCompensation turnover and top
management incentives historical evidencerdquo Journal of Business 70 153ndash187
Hartzell Jay C and Laura T Starks 2003 ldquoInstitutional Investors and Executive
Compensationrdquo Journal of Finance 58 2351ndash2374
Huson Mark R Robert Parrino and Laura T Starks 2001 ldquoInternal Monitoring
Mechanisms and CEO Turnover A Long-Term Perspectiverdquo Journal of Finance 56
2265ndash2297
Jensen Michael C and Kevin J Murphy 1990 ldquoPerformance Pay and Top Management
Incentivesrdquo Journal of Political Economy 98 225ndash264
Jin Li 2002 ldquoCEO compensation diversification and incentivesrdquo Journal of Financial
Economics 66 29ndash63
Kang Qiang and Qiao Liu 2008 ldquoStock Market Information Production and Executive
Incentivesrdquo Journal of Corporate Finance 14 484ndash498
Merton Robert 1973 ldquoTheory of Rational Option Pricingrdquo Bell Journal of Economics
and Management Science 4 141-183
Morse Adair Vikram Nanda and Amit Seru 2011 ldquoAre Incentive Contracts Rigged by
Powerful CEOsrdquo Journal of Finance 66 1779ndash1821
Murphy Kevin J 1985 ldquoCorporate performance and managerial remuneration An
empirical analysisrdquo Journal of Accounting and Economics 7 11ndash42
Murphy Kevin J 1999 ldquoExecutive compensationrdquo in Orley Ashenfelter and David
Card Handbook of Labor Economics vol 3 Elsevier
Rosen Sherwin 1992 ldquoContracts and the market for executivesrdquo In Contract
Economics ed Lars Werin and Hans Wijkander 181ndash211 Cambridge MA
Blackwell
Rothberg Burton and Steven Lilien 2006 ldquoMutual Funds and Proxy Voting New
Evidence on Corporate Governancerdquo Journal of Business and Technology Law 1
157ndash184
Schaefer Scott 1998 ldquoThe dependence of payndashperformance sensitivity on the size of the
firmrdquo Review of Economics and Statistics 80 436ndash443
Yermack David 1995 ldquoDo Corporations Award CEO Stock Options Effectivelyrdquo
Journal of Financial Economics 39 237ndash69
25
Table I Univariate Descriptive Statistics
This table reports sample statistics for the key variables used in this study based on our comprehensive
ExecuComp dataset Institutional Concentration is the shares owned by the institutions with the five largest
holdings divided by the shares held by all institutions Top 5 Ownership is the proportion of shares
outstanding held by the institutions with the five largest holdings Total Institutional Ownership is the
proportion of shares held by all institutions Panel B shows the measures executive compensation The pay-
for-performance sensitivity (PPS) of option grants shows the change in executive wealth due to a $1000
change in firm in shareholder wealth Expressions for Salary and Total Compensation are in thousands of
dollars while Market Capitalization is in millions of dollars Panel C shows firm characteristics
Variable Mean Median Std Dev P10 P90
Panel A Institutional Investor Holdings
Institutional Concentration 0450 0424 0151 0282 0656
Top 5 Institutional Ownership 0229 0220 0101 0110 0355
Total Institutional Ownership 0546 0561 0206 0258 0801
Panel B Executive Compensation
PPS of Option Grants 1161 0156 3829 0 2822
Salary 36311 29462 25231 15179 67000
Total Compensation 258313 85878 980701 23875 511022
Panel C Firm Characteristics
Tobins Q 1744 1139 2738 0495 3322
Market Capitalization 560699 112094 1988108 19754 1048754
26
Table II Pairwise Correlation Coefficients
This table presents the correlations between the key variables based on our comprehensive
ExecuComp dataset Institutional Concentration is the shares owned by the institutions with the five
largest holdings divided by the shares held by all institutions Top 5 Ownership is the proportion of
shares outstanding held by the institutions with the five largest holdings Total Institutional
Ownership is the proportion of shares held by all institutions The pay-for-performance sensitivity
(PPS) of option grants is the change in executive wealth due to a $1000 change in shareholder
wealth
(1) (2) (3) (4) (5) (6) (7) (8)
(1) Institutional Concentration 1
(2) Top 5 Ownership 0268 1
(3) Total Ownership -0415 0706 1
(4) PPS of Option Grants 0122 0042 -0024 1
(5) Log Salary -0270 -0012 0171 0038 1
(6) Log Market Capitalization -0630 -0157 0261 -0156 0420 1
(7) Level Market Capitalization -0241 -0157 0007 -0058 0219 0534 1
(8) Tobins Q -0008 -0034 0002 0015 -0083 0066 0121 1
27
Table III
Descriptive Statistics by Market Capitalization Deciles
Market
Capitalization
Deciles
Market
Capitalization
($m)
Total
Compensation
($000)
Salary
($)
PPS Option
Grants
($)
Institutional
Concentration
()
Top 5 Ownership
()
Institutional
Total Ownership
()
1 (Small) 1181 76008 234700 2720 611 200 327
2 2668 90206 257955 1796 495 217 438
3 4193 105946 275120 1517 454 220 485
4 6045 129892 296067 1376 420 206 491
5 8742 155391 315864 1216 394 198 502
6 13086 191888 345942 1072 380 202 531
7 20790 239501 377021 0874 368 196 532
8 34920 314955 427808 0687 354 190 538
9 67747 450970 480319 0569 328 173 528
10 (Big) 366540 846066 614354 0297 275 138 502
28
Table IV Pay-for-Performance Sensitivity of Option Grants as a Function of Institutional Influence
Similar in construction to parts of HS (Table II) this table shows the coefficients from Tobit regressions of option grant pay-for-performance sensitivity per $1000 change in
shareholder wealth against the lagged change and contemporary change in shareholder wealth measures of institutional influence Tobinrsquos Q firm market capitalization and
dummy controls for CEO and year effects We estimate models (1) to (4) over the period 1992‒1997 using data supplied by HS and models (5) to (8) are estimated over the
whole sample Models (1) (2) (5) and (6) use Institutional Concentration as a measure of institutional influence which HS define as the shares owned by the institutions
with the five largest holdings divided by the shares held by all institutions These models control for total institutional ownership Models (3) (4) (7) and (8) use Top 5
Ownership as a measure of institutional influence which we define as the proportion of shares outstanding held by the institutions with the five largest holdings Coefficients
for industry and year controls are not presented One two and three asterisks denote significance at 10 5 and 1 levels
Hartzell and Starks Database 1992‒1997 Full Sample from ExecuComp 1992‒2002
(1) (2) (3) (4) (5) (6) (7) (8)
Change in Shareholder Wealth 77721 191781 89874 137368 -446 -218 -664 -203
(171) (429) (194) (317) (319) (168) (453) (158)
Lag1 Change in Shareholder Wealth 14446 99963 88719 8544 -144 0273 -0342 00852
(027) (212) (166) (184) (071) (015) (016) (005)
Lag1 Institutional Concentration 13994 -00692 26405 03304
(767) (032) (1402) (166)
Lag1 Total Institutional Ownership 08055 10434 07106 0937
(593) (781) (666) (898)
Lag1 Top 5 Institutional Ownership 10567 07457 12116 06293
(399) ( 283) (533) (278)
Lag1 Tobinrsquos Q 0068 00877 00647 00813 00498 00727 00508 00695
(337) (434) (319) (401) (504) (734) (51) (703)
Lag1 Market Capitalization -106202 -173515 -776 -117
(474) (742) (1011) (153)
Lag1 Log Market Capitalization -0297 -02319 -04111 -03908
(1576) (1439) (2997) (2841)
CEO Dummy 13745 13682 13765 13674 28197 28128 28155 28143
(1913) (1918) (1914) (1916) (3388) (3396) (3377) (34)
Number of Observations 33878 33878 33878 33878 97679 97679 97679 97679
29
Table V Pay-for-Performance Sensitivity of Cash Compensation (Salary plus Bonus) as a Function of Institutional Influence - Level Market Capitalization
Similar in parts to models (1) and (2) of HS (Table IV) this table displays the coefficients from OLS regressions of the change in cash compensation against the lagged
change and contemporary change in shareholder wealth measures of institutional influence Tobinrsquos Q firm market capitalization and dummy controls for CEO and year
effects We estimate models (1) to (4) over the period 1992‒1997 based on our ExecuComp sample and models (5) to (8) over the whole sample Models (1) (2) (5) and (6)
use Institutional Concentration as a measure of institutional influence which HS define as the shares owned by the institutions with the five largest holdings divided by the
shares held by all institutions These models control for total institutional ownership Models (4) and (8) use Top 5 Ownership as a measure of institutional influence which
we define as the proportion of shares outstanding held by the institutions with the 5 largest holdings Coefficients for industry and year controls are not presented One two
and three asterisks denote significance at 10 5 and 1 levels
Our Database based on ExecuComp 1992‒1997 Full Sample based on ExecuComp 1992‒2002
(1) (2) (3) (4) (5) (6) (7) (8)
Lag1 Change in Shareholder Wealth 00014 00017 00088 00089 -00006 -00007 00072 00086
(051) (062) (129) (103) (027) (029) (131) (16)
Change in Shareholder Wealth -00102 00102 00105 00032 00102 00101
(097) (292) (293) (055) (611) (608)
Change Shareholder Wealth Lag1 Institutional
Concentration
00435 00764 00363 00253
(463) (258) (689) (144)
Lag1 Change Shareholder Wealth Lag1
Institutional Concentration
-00197 -00283
(122) (197)
Lag1 Institutional Concentration -492182 -288861 400149 453623
(09) (056) (098) (114)
Lag1 Total Institutional Ownership 720688 538372 655091 559587 682069 726226
(276) (232) (281) (314) (308) (322)
Lag1 Change Shareholder Wealth Lag1 Top 5
Institutional Ownership
-0046 -00689
(091) (256)
Lag1 Top 5 Institutional Ownership 583044 1191634
(092) (25)
Lag1 Tobinrsquos Q -32761 -36806 -19422 -18023 -54283 -48681 -29534 -24243
(138) (172) (077) (078) (227) (245) (136) (111)
Lag1 Market Capitalization 00004 00004 00002 00006 00007 00008 00007 00007
(042) (044) (022) (07) (086) (092) (082) (09)
CEO Dummy 768035 765474 768334 773634 616315 61729 619028 618535
(354) (355) (356) (355) (417) (418) (419) (418)
Adjusted R-Sq 00057 0006 00052 0005 00097 00097 00098 00102
Number of Observations 41305 41305 41305 41305 85518 85518 85518 85518
30
Table VI Level of Salary as a Function of Institutional Influence
Similar in parts to models (1) and (2) of HS (Table V) this table displays the coefficients from OLS regressions of salary against the lagged change and contemporary change in
shareholder wealth measures of institutional influence Tobinrsquos Q firm market capitalization and dummy controls for CEO industry and year effects We estimate models (1) to
(4) over the period 1992‒1997 using the actual data supplied by HS and models (5) to (8) over the whole sample 1992‒2002 Models (1) (2) (5) and (6) use Institutional
Concentration as a measure of institutional influence which HS define as the shares owned by the institutions with the five largest holdings divided by the shares held by all
institutions These models control for total institutional ownership Models (3) (4) (7) and (8) use Top 5 Ownership as a measure of institutional influence which we define as
the proportion of shares outstanding held by the institutions with the 5 largest holdings Coefficients for industry and year controls are not presented One two and three
asterisks denote significance at 10 5 and 1 levels
Hartzell and Starks Database 1992‒1997 Full Sample based on ExecuComp 1992‒2002
(1) (2) (3) (4) (5) (6) (7) (8)
Change in Shareholder Wealth 04125E-4 0003 -2439E-4 0003 00021 00012 00024 00012
(057) (464) (031) (468) (1004) (72) (1041) (706)
Lag1 Change in Shareholder Wealth -00062 7856E-4 -0007 9003E-4 -00014 00002 -00014 00003
(594) (098) (592) (112) (522) (123) (5) (145)
Lag1 Institutional Concentration -2475087 555134 -2734983 943317
(3245) (710) (4965) (1671)
Lag1 Total Institutional Ownership 1020109 217488 936337 252464
(1904) (432) (2471) (704)
Lag1 Top 5 Institutional Ownership -148355 31354 -87926 423166
(139) (332) (118) (643)
Lag1 Tobinrsquos Q -195348 -252191 -192344 -25158 -119837 -146373 -122132 -143821
(2812) (3762) (2721) (3752) (248) (274) (2583) (2727)
Lag1 Market Capitalization 844E-06 103E-5 00034 00039
(2781) (3051) (2874) (3104)
Lag1 Log Market Capitalization 750021 723625 776657 732279
(7511) (9285) (11224) (14039)
CEO Dummy 2683978 2702238 2668244 2702417 3281133 3285103 3292175 3283145
(8625) (9277) (8246) (9271) (12739) (13616) (12336) (13605)
Adjusted R-Sq 04604 05302 04187 05295 03856 04588 03476 04574
Number of Observations 36280 36280 36280 36280 97679 97679 97679 97679
31
Table VII Level of Total Direct Compensation as a Function of Institutional Influence
Similar in parts to models (3) and (5) of HS (Table V) this table displays the coefficients from OLS regressions of total direct compensation against the lagged change and
contemporary change in shareholder wealth measures of institutional influence Tobinrsquos Q firm market capitalization and dummy controls for CEO industry and year effects
We estimate models (1) to (4) over the period 1992‒1997 and models (5) to (8) over the whole sample Models (1) (2) (5) and (6) use Institutional Concentration as a measure
of institutional influence which HS define as the shares owned by the institutions with the five largest holdings divided by the shares held by all institutions These models
control for total institutional ownership Models (3) (4) (7) and (8) use Top 5 Ownership as a measure of institutional influence which we define as the proportion of shares
outstanding held by the institutions with the five largest holdings Coefficients for industry and year controls are not presented One two and three asterisks denote significance
at 10 5 and 1 levels
Hartzell and Starks Database 1992‒1997 Full Sample from ExecuComp 1992‒2002
(1) (2) (3) (4) (5) (6) (7) (8)
Change in Shareholder Wealth 005779 007006 005282 006791 00423 00148 0046 00138
(415) (581) (369) (565) (203) (07) (22) (066)
Lag1 Change in Shareholder Wealth 003182 00714 00272 00715 00613 0126 00614 01278
(126) (298) (107) (299) (275) (488) (276) (494)
Lag1 Institutional Concentration -1608633 3248998 -3424892 3334186
(151) (283) (1172) (118)
Lag1 Total Institutional Ownership 104013 5473128 1531013 -2926764
(1645) (831) (876) (158)
Lag1 Top 5 Institutional Ownership 188585 3927986 -2159177 -2640879
(016) (343) (067) (092)
Lag1 Tobinrsquos Q -445692 -803612 -454859 -833566 3257027 2971201 3213936 3086205
(427) (791) (432) (831) (83) (77) (815) (793)
Lag1 Market Capitalization 49E-5 615E-5 01031 01088
(113) (1365) (1555) (168)
Lag1 Log Market Capitalization 4683497 4759377 1645156 1427856
(2775) (3469) (3679) (4239)
CEO Dummy 1501307 1508886 1493688 1508736 4713079 4713629 47278 4703908
(2648) (2684) (2618) (2683) (282) (2828) (2819) (2824)
Adjusted R-Sq 01861 02060 01695 02051 01071 01113 01033 011
Number of Observations 33229 33229 33229 33229 97679 97679 97679 97679
32
Table VIII Two Stage Least Squares (Reverse Causality) Tests - Executive Compensation Measures as a Function of Institutional Concentration
This table shows the coefficients from the Two-Stage Least Squares (2SLS) regressions of executive compensation measures against the lagged change
and contemporary change in shareholder wealth instrumented institutional concentration and total institutional ownership Tobinrsquos Q the natural
logarithm of firm market capitalization and dummy variable controls for CEO and year effects We instrument measures of institutional concentration
and total institutional ownership by orthogonalizing with respect to option grant pay-for-performance sensitivity salary and total compensation
Coefficients for industry and year controls are not presented First stage results are not presented One two and three asterisks denote significance at
10 5 and 1 levels
ExecuComp Sample 1992‒1997 Full Sample from ExecuComp 1992‒2002
PPSOG Salary TC PPSOG Salary TC
Change in Shareholder Wealth 144E-05 0004 02133 -409E-06 00011 00157
(308) (775) (596) (691) (638) (068)
Lag1 Change in Shareholder Wealth -240E-05 -00004 00866 118E-06 00002 01422
(131) (087) (346) (097) (116) (51)
Lag1 Institutional Concentration -00898 668681 3263143 02024 916656 2856417
(052) (801) (176) (163) (1545) (1004)
Lag1 Total Institutional Ownership 01021 -43997 2512587 02983 -104171 -1028347
(09) (084) (153) (4) (269) (487)
Lag1 Tobinrsquos Q 00754 -225606 119297 00655 -158657 2997205
(596) (1649) (053) (834) (2535) (727)
Lag1 Log Market Capitalization -04234 828713 7864046 -04002 825106 1685638
(2195) (8132) (2977) (4106) (11133) (373)
CEO Dummy 15848 3214117 2950274 19149 3205862 4529101
(2223) (9389) (2255) (3802) (1331) (2758)
Overall R-Sq -- 04981 01206 -- 04738 01178
Number of Observations 41305 41305 41305 85518 85518 85518
33
Table IX Two Stage Least Squares (Reverse Causality) Tests - Executive Compensation Measures as a Function of Top 5 Ownership
This table shows the coefficients from 2SLS regressions of executive compensation measures against the lagged change and contemporary change
in shareholder wealth instrumented Top 5 institutional ownership Tobinrsquos Q natural logarithm of firm market capitalization and dummy variable
controls for CEO and year effects We instrument Top 5 institutional ownership by orthogonalizing with respect to option grant pay-for-
performance sensitivity salary and total compensation Coefficients for industry and year controls are not presented First stage results are not
presented One two and three asterisks denote significance at 10 5 and 1 levels
ExecuComp sample 1992‒1997 Full Sample 1992‒2002
PPSOG Salary TC PPSOG Salary TC
Change in Shareholder Wealth 136E-05 00041 02128 -407E-06 00011 00146
(293) (79) (596) (692) (619) (063)
Lag1 Change in Shareholder Wealth -241E-05 -00003 00868 116E-06 00003 01439
(133) (07) (347) (096) (142) (514)
Lag1 Top 5 Institutional Ownership -03774 -187512 -3848531 00179 -104261 -126659
(163) (186) (166) (011) (149) (436)
Lag1 Tobinrsquos Q 00731 -223886 10074 00642 -155467 3115073
(573) (1631) (045) (819) (2509) (75)
Lag1 Log Market Capitalization -0416 775637 7696302 -04015 764632 1462536
(2345) (10084) (3766) (3688) (13708) (4072)
CEO Dummy 15854 3211991 2950158 19152 3203152 4519385
(2223) (9381) (2254) (3804) (133) (2752)
Overall R-Sq -- 04967 01202 -- 04722 01168
Number of Observations 41305 41305 41305 85518 85518 85518
34
Table X Specification Tests utilizing Box-Cox Transformation of Market Capitalization
This table presents results from a Box-Cox transformation of market capitalization In
each model we specify transformation of market capitalization only Panel A shows
results for the period 1992‒1997 and Panel B shows results for the entire period
1992‒2002 The first part of each panel shows the Lambda coefficient that applies to
the transformed market capitalization The second part of each panel shows log
likelihood scores with LR Chi2 Statistics in parentheses ‒ these test for the
appropriateness of a linear or log model specification
Panel A Our ExecuComp sample 1992‒1997
(1) (2) (3)
PPSOG Salary TC
Lambda -05143 01881 02876
(-1914) (2413) (1329)
Test H0 (Log Likelihood Score)
Lambda = -1 -1279748 -3178575 -4795326
(23866) (85952) (99688)
Lambda = 0 -1280152 -3138456 -4791159
(31934) (57147) (16339)
Lambda = 1 -1283157 -3163807 -4793506
(92039) (564156) (63292)
Panel B Full ExecuComp sample 1992‒2002
(1) (2) (4)
PPSOG Salary TC
Lambda -044 01853 04036
(-3015) (40) (3666)
Test H0 (Log Likelihood Score)
Lambda = 0 -2688334 -6476494 -10384490
(66599) (160356) (126236)
Lambda = 1 -2694886 -6538506 -10386795
(197644) (1400605) (172345)
35
Figure 2 Institutional Ownership Measures (LHA) vs Total
Compensation (RHA) by Market Capitalization
0
10
20
30
40
50
60
70
1 2 3 4 5 6 7 8 9 10
$00
$10
$20
$30
$40
$50
$60
$70
$80
$90
Millio
ns
InstitutionalConcentration (LHS)Top5 Ownership (LHS)
Institutional TotalOwnership (LHS)Total Compensation(RHS)
Page 24
24
Gabaix Xavier and Augustin Landier 2008 ldquoWhy Has CEO Pay Increased So Muchrdquo
Quarterly Journal of Economics 123 49ndash100
Garen John 1994 ldquoExecutive compensation and principal-agent theoryrdquo Journal of
Political Economy 102 1175ndash1199
Gibbons Robert and Kevin J Murphy 1992 ldquoOptimal Incentive Contracts in the
Presence of Career Concerns Theory and Evidencerdquo Journal of Political Economy
100 468ndash505
Greene William H 2000 (4th
Edition) Econometric Analysis Prentice-Hall NJ
Guthrie Katherine Jan Sokolowsky and Kam-Ming Wan 2012 ldquoCEO Compensation
and Board Structure Revisitedrdquo Journal of Finance Forthcoming
Hadlock Charles R and Gerald B Lumer 1997 ldquoCompensation turnover and top
management incentives historical evidencerdquo Journal of Business 70 153ndash187
Hartzell Jay C and Laura T Starks 2003 ldquoInstitutional Investors and Executive
Compensationrdquo Journal of Finance 58 2351ndash2374
Huson Mark R Robert Parrino and Laura T Starks 2001 ldquoInternal Monitoring
Mechanisms and CEO Turnover A Long-Term Perspectiverdquo Journal of Finance 56
2265ndash2297
Jensen Michael C and Kevin J Murphy 1990 ldquoPerformance Pay and Top Management
Incentivesrdquo Journal of Political Economy 98 225ndash264
Jin Li 2002 ldquoCEO compensation diversification and incentivesrdquo Journal of Financial
Economics 66 29ndash63
Kang Qiang and Qiao Liu 2008 ldquoStock Market Information Production and Executive
Incentivesrdquo Journal of Corporate Finance 14 484ndash498
Merton Robert 1973 ldquoTheory of Rational Option Pricingrdquo Bell Journal of Economics
and Management Science 4 141-183
Morse Adair Vikram Nanda and Amit Seru 2011 ldquoAre Incentive Contracts Rigged by
Powerful CEOsrdquo Journal of Finance 66 1779ndash1821
Murphy Kevin J 1985 ldquoCorporate performance and managerial remuneration An
empirical analysisrdquo Journal of Accounting and Economics 7 11ndash42
Murphy Kevin J 1999 ldquoExecutive compensationrdquo in Orley Ashenfelter and David
Card Handbook of Labor Economics vol 3 Elsevier
Rosen Sherwin 1992 ldquoContracts and the market for executivesrdquo In Contract
Economics ed Lars Werin and Hans Wijkander 181ndash211 Cambridge MA
Blackwell
Rothberg Burton and Steven Lilien 2006 ldquoMutual Funds and Proxy Voting New
Evidence on Corporate Governancerdquo Journal of Business and Technology Law 1
157ndash184
Schaefer Scott 1998 ldquoThe dependence of payndashperformance sensitivity on the size of the
firmrdquo Review of Economics and Statistics 80 436ndash443
Yermack David 1995 ldquoDo Corporations Award CEO Stock Options Effectivelyrdquo
Journal of Financial Economics 39 237ndash69
25
Table I Univariate Descriptive Statistics
This table reports sample statistics for the key variables used in this study based on our comprehensive
ExecuComp dataset Institutional Concentration is the shares owned by the institutions with the five largest
holdings divided by the shares held by all institutions Top 5 Ownership is the proportion of shares
outstanding held by the institutions with the five largest holdings Total Institutional Ownership is the
proportion of shares held by all institutions Panel B shows the measures executive compensation The pay-
for-performance sensitivity (PPS) of option grants shows the change in executive wealth due to a $1000
change in firm in shareholder wealth Expressions for Salary and Total Compensation are in thousands of
dollars while Market Capitalization is in millions of dollars Panel C shows firm characteristics
Variable Mean Median Std Dev P10 P90
Panel A Institutional Investor Holdings
Institutional Concentration 0450 0424 0151 0282 0656
Top 5 Institutional Ownership 0229 0220 0101 0110 0355
Total Institutional Ownership 0546 0561 0206 0258 0801
Panel B Executive Compensation
PPS of Option Grants 1161 0156 3829 0 2822
Salary 36311 29462 25231 15179 67000
Total Compensation 258313 85878 980701 23875 511022
Panel C Firm Characteristics
Tobins Q 1744 1139 2738 0495 3322
Market Capitalization 560699 112094 1988108 19754 1048754
26
Table II Pairwise Correlation Coefficients
This table presents the correlations between the key variables based on our comprehensive
ExecuComp dataset Institutional Concentration is the shares owned by the institutions with the five
largest holdings divided by the shares held by all institutions Top 5 Ownership is the proportion of
shares outstanding held by the institutions with the five largest holdings Total Institutional
Ownership is the proportion of shares held by all institutions The pay-for-performance sensitivity
(PPS) of option grants is the change in executive wealth due to a $1000 change in shareholder
wealth
(1) (2) (3) (4) (5) (6) (7) (8)
(1) Institutional Concentration 1
(2) Top 5 Ownership 0268 1
(3) Total Ownership -0415 0706 1
(4) PPS of Option Grants 0122 0042 -0024 1
(5) Log Salary -0270 -0012 0171 0038 1
(6) Log Market Capitalization -0630 -0157 0261 -0156 0420 1
(7) Level Market Capitalization -0241 -0157 0007 -0058 0219 0534 1
(8) Tobins Q -0008 -0034 0002 0015 -0083 0066 0121 1
27
Table III
Descriptive Statistics by Market Capitalization Deciles
Market
Capitalization
Deciles
Market
Capitalization
($m)
Total
Compensation
($000)
Salary
($)
PPS Option
Grants
($)
Institutional
Concentration
()
Top 5 Ownership
()
Institutional
Total Ownership
()
1 (Small) 1181 76008 234700 2720 611 200 327
2 2668 90206 257955 1796 495 217 438
3 4193 105946 275120 1517 454 220 485
4 6045 129892 296067 1376 420 206 491
5 8742 155391 315864 1216 394 198 502
6 13086 191888 345942 1072 380 202 531
7 20790 239501 377021 0874 368 196 532
8 34920 314955 427808 0687 354 190 538
9 67747 450970 480319 0569 328 173 528
10 (Big) 366540 846066 614354 0297 275 138 502
28
Table IV Pay-for-Performance Sensitivity of Option Grants as a Function of Institutional Influence
Similar in construction to parts of HS (Table II) this table shows the coefficients from Tobit regressions of option grant pay-for-performance sensitivity per $1000 change in
shareholder wealth against the lagged change and contemporary change in shareholder wealth measures of institutional influence Tobinrsquos Q firm market capitalization and
dummy controls for CEO and year effects We estimate models (1) to (4) over the period 1992‒1997 using data supplied by HS and models (5) to (8) are estimated over the
whole sample Models (1) (2) (5) and (6) use Institutional Concentration as a measure of institutional influence which HS define as the shares owned by the institutions
with the five largest holdings divided by the shares held by all institutions These models control for total institutional ownership Models (3) (4) (7) and (8) use Top 5
Ownership as a measure of institutional influence which we define as the proportion of shares outstanding held by the institutions with the five largest holdings Coefficients
for industry and year controls are not presented One two and three asterisks denote significance at 10 5 and 1 levels
Hartzell and Starks Database 1992‒1997 Full Sample from ExecuComp 1992‒2002
(1) (2) (3) (4) (5) (6) (7) (8)
Change in Shareholder Wealth 77721 191781 89874 137368 -446 -218 -664 -203
(171) (429) (194) (317) (319) (168) (453) (158)
Lag1 Change in Shareholder Wealth 14446 99963 88719 8544 -144 0273 -0342 00852
(027) (212) (166) (184) (071) (015) (016) (005)
Lag1 Institutional Concentration 13994 -00692 26405 03304
(767) (032) (1402) (166)
Lag1 Total Institutional Ownership 08055 10434 07106 0937
(593) (781) (666) (898)
Lag1 Top 5 Institutional Ownership 10567 07457 12116 06293
(399) ( 283) (533) (278)
Lag1 Tobinrsquos Q 0068 00877 00647 00813 00498 00727 00508 00695
(337) (434) (319) (401) (504) (734) (51) (703)
Lag1 Market Capitalization -106202 -173515 -776 -117
(474) (742) (1011) (153)
Lag1 Log Market Capitalization -0297 -02319 -04111 -03908
(1576) (1439) (2997) (2841)
CEO Dummy 13745 13682 13765 13674 28197 28128 28155 28143
(1913) (1918) (1914) (1916) (3388) (3396) (3377) (34)
Number of Observations 33878 33878 33878 33878 97679 97679 97679 97679
29
Table V Pay-for-Performance Sensitivity of Cash Compensation (Salary plus Bonus) as a Function of Institutional Influence - Level Market Capitalization
Similar in parts to models (1) and (2) of HS (Table IV) this table displays the coefficients from OLS regressions of the change in cash compensation against the lagged
change and contemporary change in shareholder wealth measures of institutional influence Tobinrsquos Q firm market capitalization and dummy controls for CEO and year
effects We estimate models (1) to (4) over the period 1992‒1997 based on our ExecuComp sample and models (5) to (8) over the whole sample Models (1) (2) (5) and (6)
use Institutional Concentration as a measure of institutional influence which HS define as the shares owned by the institutions with the five largest holdings divided by the
shares held by all institutions These models control for total institutional ownership Models (4) and (8) use Top 5 Ownership as a measure of institutional influence which
we define as the proportion of shares outstanding held by the institutions with the 5 largest holdings Coefficients for industry and year controls are not presented One two
and three asterisks denote significance at 10 5 and 1 levels
Our Database based on ExecuComp 1992‒1997 Full Sample based on ExecuComp 1992‒2002
(1) (2) (3) (4) (5) (6) (7) (8)
Lag1 Change in Shareholder Wealth 00014 00017 00088 00089 -00006 -00007 00072 00086
(051) (062) (129) (103) (027) (029) (131) (16)
Change in Shareholder Wealth -00102 00102 00105 00032 00102 00101
(097) (292) (293) (055) (611) (608)
Change Shareholder Wealth Lag1 Institutional
Concentration
00435 00764 00363 00253
(463) (258) (689) (144)
Lag1 Change Shareholder Wealth Lag1
Institutional Concentration
-00197 -00283
(122) (197)
Lag1 Institutional Concentration -492182 -288861 400149 453623
(09) (056) (098) (114)
Lag1 Total Institutional Ownership 720688 538372 655091 559587 682069 726226
(276) (232) (281) (314) (308) (322)
Lag1 Change Shareholder Wealth Lag1 Top 5
Institutional Ownership
-0046 -00689
(091) (256)
Lag1 Top 5 Institutional Ownership 583044 1191634
(092) (25)
Lag1 Tobinrsquos Q -32761 -36806 -19422 -18023 -54283 -48681 -29534 -24243
(138) (172) (077) (078) (227) (245) (136) (111)
Lag1 Market Capitalization 00004 00004 00002 00006 00007 00008 00007 00007
(042) (044) (022) (07) (086) (092) (082) (09)
CEO Dummy 768035 765474 768334 773634 616315 61729 619028 618535
(354) (355) (356) (355) (417) (418) (419) (418)
Adjusted R-Sq 00057 0006 00052 0005 00097 00097 00098 00102
Number of Observations 41305 41305 41305 41305 85518 85518 85518 85518
30
Table VI Level of Salary as a Function of Institutional Influence
Similar in parts to models (1) and (2) of HS (Table V) this table displays the coefficients from OLS regressions of salary against the lagged change and contemporary change in
shareholder wealth measures of institutional influence Tobinrsquos Q firm market capitalization and dummy controls for CEO industry and year effects We estimate models (1) to
(4) over the period 1992‒1997 using the actual data supplied by HS and models (5) to (8) over the whole sample 1992‒2002 Models (1) (2) (5) and (6) use Institutional
Concentration as a measure of institutional influence which HS define as the shares owned by the institutions with the five largest holdings divided by the shares held by all
institutions These models control for total institutional ownership Models (3) (4) (7) and (8) use Top 5 Ownership as a measure of institutional influence which we define as
the proportion of shares outstanding held by the institutions with the 5 largest holdings Coefficients for industry and year controls are not presented One two and three
asterisks denote significance at 10 5 and 1 levels
Hartzell and Starks Database 1992‒1997 Full Sample based on ExecuComp 1992‒2002
(1) (2) (3) (4) (5) (6) (7) (8)
Change in Shareholder Wealth 04125E-4 0003 -2439E-4 0003 00021 00012 00024 00012
(057) (464) (031) (468) (1004) (72) (1041) (706)
Lag1 Change in Shareholder Wealth -00062 7856E-4 -0007 9003E-4 -00014 00002 -00014 00003
(594) (098) (592) (112) (522) (123) (5) (145)
Lag1 Institutional Concentration -2475087 555134 -2734983 943317
(3245) (710) (4965) (1671)
Lag1 Total Institutional Ownership 1020109 217488 936337 252464
(1904) (432) (2471) (704)
Lag1 Top 5 Institutional Ownership -148355 31354 -87926 423166
(139) (332) (118) (643)
Lag1 Tobinrsquos Q -195348 -252191 -192344 -25158 -119837 -146373 -122132 -143821
(2812) (3762) (2721) (3752) (248) (274) (2583) (2727)
Lag1 Market Capitalization 844E-06 103E-5 00034 00039
(2781) (3051) (2874) (3104)
Lag1 Log Market Capitalization 750021 723625 776657 732279
(7511) (9285) (11224) (14039)
CEO Dummy 2683978 2702238 2668244 2702417 3281133 3285103 3292175 3283145
(8625) (9277) (8246) (9271) (12739) (13616) (12336) (13605)
Adjusted R-Sq 04604 05302 04187 05295 03856 04588 03476 04574
Number of Observations 36280 36280 36280 36280 97679 97679 97679 97679
31
Table VII Level of Total Direct Compensation as a Function of Institutional Influence
Similar in parts to models (3) and (5) of HS (Table V) this table displays the coefficients from OLS regressions of total direct compensation against the lagged change and
contemporary change in shareholder wealth measures of institutional influence Tobinrsquos Q firm market capitalization and dummy controls for CEO industry and year effects
We estimate models (1) to (4) over the period 1992‒1997 and models (5) to (8) over the whole sample Models (1) (2) (5) and (6) use Institutional Concentration as a measure
of institutional influence which HS define as the shares owned by the institutions with the five largest holdings divided by the shares held by all institutions These models
control for total institutional ownership Models (3) (4) (7) and (8) use Top 5 Ownership as a measure of institutional influence which we define as the proportion of shares
outstanding held by the institutions with the five largest holdings Coefficients for industry and year controls are not presented One two and three asterisks denote significance
at 10 5 and 1 levels
Hartzell and Starks Database 1992‒1997 Full Sample from ExecuComp 1992‒2002
(1) (2) (3) (4) (5) (6) (7) (8)
Change in Shareholder Wealth 005779 007006 005282 006791 00423 00148 0046 00138
(415) (581) (369) (565) (203) (07) (22) (066)
Lag1 Change in Shareholder Wealth 003182 00714 00272 00715 00613 0126 00614 01278
(126) (298) (107) (299) (275) (488) (276) (494)
Lag1 Institutional Concentration -1608633 3248998 -3424892 3334186
(151) (283) (1172) (118)
Lag1 Total Institutional Ownership 104013 5473128 1531013 -2926764
(1645) (831) (876) (158)
Lag1 Top 5 Institutional Ownership 188585 3927986 -2159177 -2640879
(016) (343) (067) (092)
Lag1 Tobinrsquos Q -445692 -803612 -454859 -833566 3257027 2971201 3213936 3086205
(427) (791) (432) (831) (83) (77) (815) (793)
Lag1 Market Capitalization 49E-5 615E-5 01031 01088
(113) (1365) (1555) (168)
Lag1 Log Market Capitalization 4683497 4759377 1645156 1427856
(2775) (3469) (3679) (4239)
CEO Dummy 1501307 1508886 1493688 1508736 4713079 4713629 47278 4703908
(2648) (2684) (2618) (2683) (282) (2828) (2819) (2824)
Adjusted R-Sq 01861 02060 01695 02051 01071 01113 01033 011
Number of Observations 33229 33229 33229 33229 97679 97679 97679 97679
32
Table VIII Two Stage Least Squares (Reverse Causality) Tests - Executive Compensation Measures as a Function of Institutional Concentration
This table shows the coefficients from the Two-Stage Least Squares (2SLS) regressions of executive compensation measures against the lagged change
and contemporary change in shareholder wealth instrumented institutional concentration and total institutional ownership Tobinrsquos Q the natural
logarithm of firm market capitalization and dummy variable controls for CEO and year effects We instrument measures of institutional concentration
and total institutional ownership by orthogonalizing with respect to option grant pay-for-performance sensitivity salary and total compensation
Coefficients for industry and year controls are not presented First stage results are not presented One two and three asterisks denote significance at
10 5 and 1 levels
ExecuComp Sample 1992‒1997 Full Sample from ExecuComp 1992‒2002
PPSOG Salary TC PPSOG Salary TC
Change in Shareholder Wealth 144E-05 0004 02133 -409E-06 00011 00157
(308) (775) (596) (691) (638) (068)
Lag1 Change in Shareholder Wealth -240E-05 -00004 00866 118E-06 00002 01422
(131) (087) (346) (097) (116) (51)
Lag1 Institutional Concentration -00898 668681 3263143 02024 916656 2856417
(052) (801) (176) (163) (1545) (1004)
Lag1 Total Institutional Ownership 01021 -43997 2512587 02983 -104171 -1028347
(09) (084) (153) (4) (269) (487)
Lag1 Tobinrsquos Q 00754 -225606 119297 00655 -158657 2997205
(596) (1649) (053) (834) (2535) (727)
Lag1 Log Market Capitalization -04234 828713 7864046 -04002 825106 1685638
(2195) (8132) (2977) (4106) (11133) (373)
CEO Dummy 15848 3214117 2950274 19149 3205862 4529101
(2223) (9389) (2255) (3802) (1331) (2758)
Overall R-Sq -- 04981 01206 -- 04738 01178
Number of Observations 41305 41305 41305 85518 85518 85518
33
Table IX Two Stage Least Squares (Reverse Causality) Tests - Executive Compensation Measures as a Function of Top 5 Ownership
This table shows the coefficients from 2SLS regressions of executive compensation measures against the lagged change and contemporary change
in shareholder wealth instrumented Top 5 institutional ownership Tobinrsquos Q natural logarithm of firm market capitalization and dummy variable
controls for CEO and year effects We instrument Top 5 institutional ownership by orthogonalizing with respect to option grant pay-for-
performance sensitivity salary and total compensation Coefficients for industry and year controls are not presented First stage results are not
presented One two and three asterisks denote significance at 10 5 and 1 levels
ExecuComp sample 1992‒1997 Full Sample 1992‒2002
PPSOG Salary TC PPSOG Salary TC
Change in Shareholder Wealth 136E-05 00041 02128 -407E-06 00011 00146
(293) (79) (596) (692) (619) (063)
Lag1 Change in Shareholder Wealth -241E-05 -00003 00868 116E-06 00003 01439
(133) (07) (347) (096) (142) (514)
Lag1 Top 5 Institutional Ownership -03774 -187512 -3848531 00179 -104261 -126659
(163) (186) (166) (011) (149) (436)
Lag1 Tobinrsquos Q 00731 -223886 10074 00642 -155467 3115073
(573) (1631) (045) (819) (2509) (75)
Lag1 Log Market Capitalization -0416 775637 7696302 -04015 764632 1462536
(2345) (10084) (3766) (3688) (13708) (4072)
CEO Dummy 15854 3211991 2950158 19152 3203152 4519385
(2223) (9381) (2254) (3804) (133) (2752)
Overall R-Sq -- 04967 01202 -- 04722 01168
Number of Observations 41305 41305 41305 85518 85518 85518
34
Table X Specification Tests utilizing Box-Cox Transformation of Market Capitalization
This table presents results from a Box-Cox transformation of market capitalization In
each model we specify transformation of market capitalization only Panel A shows
results for the period 1992‒1997 and Panel B shows results for the entire period
1992‒2002 The first part of each panel shows the Lambda coefficient that applies to
the transformed market capitalization The second part of each panel shows log
likelihood scores with LR Chi2 Statistics in parentheses ‒ these test for the
appropriateness of a linear or log model specification
Panel A Our ExecuComp sample 1992‒1997
(1) (2) (3)
PPSOG Salary TC
Lambda -05143 01881 02876
(-1914) (2413) (1329)
Test H0 (Log Likelihood Score)
Lambda = -1 -1279748 -3178575 -4795326
(23866) (85952) (99688)
Lambda = 0 -1280152 -3138456 -4791159
(31934) (57147) (16339)
Lambda = 1 -1283157 -3163807 -4793506
(92039) (564156) (63292)
Panel B Full ExecuComp sample 1992‒2002
(1) (2) (4)
PPSOG Salary TC
Lambda -044 01853 04036
(-3015) (40) (3666)
Test H0 (Log Likelihood Score)
Lambda = 0 -2688334 -6476494 -10384490
(66599) (160356) (126236)
Lambda = 1 -2694886 -6538506 -10386795
(197644) (1400605) (172345)
35
Figure 2 Institutional Ownership Measures (LHA) vs Total
Compensation (RHA) by Market Capitalization
0
10
20
30
40
50
60
70
1 2 3 4 5 6 7 8 9 10
$00
$10
$20
$30
$40
$50
$60
$70
$80
$90
Millio
ns
InstitutionalConcentration (LHS)Top5 Ownership (LHS)
Institutional TotalOwnership (LHS)Total Compensation(RHS)
Page 25
25
Table I Univariate Descriptive Statistics
This table reports sample statistics for the key variables used in this study based on our comprehensive
ExecuComp dataset Institutional Concentration is the shares owned by the institutions with the five largest
holdings divided by the shares held by all institutions Top 5 Ownership is the proportion of shares
outstanding held by the institutions with the five largest holdings Total Institutional Ownership is the
proportion of shares held by all institutions Panel B shows the measures executive compensation The pay-
for-performance sensitivity (PPS) of option grants shows the change in executive wealth due to a $1000
change in firm in shareholder wealth Expressions for Salary and Total Compensation are in thousands of
dollars while Market Capitalization is in millions of dollars Panel C shows firm characteristics
Variable Mean Median Std Dev P10 P90
Panel A Institutional Investor Holdings
Institutional Concentration 0450 0424 0151 0282 0656
Top 5 Institutional Ownership 0229 0220 0101 0110 0355
Total Institutional Ownership 0546 0561 0206 0258 0801
Panel B Executive Compensation
PPS of Option Grants 1161 0156 3829 0 2822
Salary 36311 29462 25231 15179 67000
Total Compensation 258313 85878 980701 23875 511022
Panel C Firm Characteristics
Tobins Q 1744 1139 2738 0495 3322
Market Capitalization 560699 112094 1988108 19754 1048754
26
Table II Pairwise Correlation Coefficients
This table presents the correlations between the key variables based on our comprehensive
ExecuComp dataset Institutional Concentration is the shares owned by the institutions with the five
largest holdings divided by the shares held by all institutions Top 5 Ownership is the proportion of
shares outstanding held by the institutions with the five largest holdings Total Institutional
Ownership is the proportion of shares held by all institutions The pay-for-performance sensitivity
(PPS) of option grants is the change in executive wealth due to a $1000 change in shareholder
wealth
(1) (2) (3) (4) (5) (6) (7) (8)
(1) Institutional Concentration 1
(2) Top 5 Ownership 0268 1
(3) Total Ownership -0415 0706 1
(4) PPS of Option Grants 0122 0042 -0024 1
(5) Log Salary -0270 -0012 0171 0038 1
(6) Log Market Capitalization -0630 -0157 0261 -0156 0420 1
(7) Level Market Capitalization -0241 -0157 0007 -0058 0219 0534 1
(8) Tobins Q -0008 -0034 0002 0015 -0083 0066 0121 1
27
Table III
Descriptive Statistics by Market Capitalization Deciles
Market
Capitalization
Deciles
Market
Capitalization
($m)
Total
Compensation
($000)
Salary
($)
PPS Option
Grants
($)
Institutional
Concentration
()
Top 5 Ownership
()
Institutional
Total Ownership
()
1 (Small) 1181 76008 234700 2720 611 200 327
2 2668 90206 257955 1796 495 217 438
3 4193 105946 275120 1517 454 220 485
4 6045 129892 296067 1376 420 206 491
5 8742 155391 315864 1216 394 198 502
6 13086 191888 345942 1072 380 202 531
7 20790 239501 377021 0874 368 196 532
8 34920 314955 427808 0687 354 190 538
9 67747 450970 480319 0569 328 173 528
10 (Big) 366540 846066 614354 0297 275 138 502
28
Table IV Pay-for-Performance Sensitivity of Option Grants as a Function of Institutional Influence
Similar in construction to parts of HS (Table II) this table shows the coefficients from Tobit regressions of option grant pay-for-performance sensitivity per $1000 change in
shareholder wealth against the lagged change and contemporary change in shareholder wealth measures of institutional influence Tobinrsquos Q firm market capitalization and
dummy controls for CEO and year effects We estimate models (1) to (4) over the period 1992‒1997 using data supplied by HS and models (5) to (8) are estimated over the
whole sample Models (1) (2) (5) and (6) use Institutional Concentration as a measure of institutional influence which HS define as the shares owned by the institutions
with the five largest holdings divided by the shares held by all institutions These models control for total institutional ownership Models (3) (4) (7) and (8) use Top 5
Ownership as a measure of institutional influence which we define as the proportion of shares outstanding held by the institutions with the five largest holdings Coefficients
for industry and year controls are not presented One two and three asterisks denote significance at 10 5 and 1 levels
Hartzell and Starks Database 1992‒1997 Full Sample from ExecuComp 1992‒2002
(1) (2) (3) (4) (5) (6) (7) (8)
Change in Shareholder Wealth 77721 191781 89874 137368 -446 -218 -664 -203
(171) (429) (194) (317) (319) (168) (453) (158)
Lag1 Change in Shareholder Wealth 14446 99963 88719 8544 -144 0273 -0342 00852
(027) (212) (166) (184) (071) (015) (016) (005)
Lag1 Institutional Concentration 13994 -00692 26405 03304
(767) (032) (1402) (166)
Lag1 Total Institutional Ownership 08055 10434 07106 0937
(593) (781) (666) (898)
Lag1 Top 5 Institutional Ownership 10567 07457 12116 06293
(399) ( 283) (533) (278)
Lag1 Tobinrsquos Q 0068 00877 00647 00813 00498 00727 00508 00695
(337) (434) (319) (401) (504) (734) (51) (703)
Lag1 Market Capitalization -106202 -173515 -776 -117
(474) (742) (1011) (153)
Lag1 Log Market Capitalization -0297 -02319 -04111 -03908
(1576) (1439) (2997) (2841)
CEO Dummy 13745 13682 13765 13674 28197 28128 28155 28143
(1913) (1918) (1914) (1916) (3388) (3396) (3377) (34)
Number of Observations 33878 33878 33878 33878 97679 97679 97679 97679
29
Table V Pay-for-Performance Sensitivity of Cash Compensation (Salary plus Bonus) as a Function of Institutional Influence - Level Market Capitalization
Similar in parts to models (1) and (2) of HS (Table IV) this table displays the coefficients from OLS regressions of the change in cash compensation against the lagged
change and contemporary change in shareholder wealth measures of institutional influence Tobinrsquos Q firm market capitalization and dummy controls for CEO and year
effects We estimate models (1) to (4) over the period 1992‒1997 based on our ExecuComp sample and models (5) to (8) over the whole sample Models (1) (2) (5) and (6)
use Institutional Concentration as a measure of institutional influence which HS define as the shares owned by the institutions with the five largest holdings divided by the
shares held by all institutions These models control for total institutional ownership Models (4) and (8) use Top 5 Ownership as a measure of institutional influence which
we define as the proportion of shares outstanding held by the institutions with the 5 largest holdings Coefficients for industry and year controls are not presented One two
and three asterisks denote significance at 10 5 and 1 levels
Our Database based on ExecuComp 1992‒1997 Full Sample based on ExecuComp 1992‒2002
(1) (2) (3) (4) (5) (6) (7) (8)
Lag1 Change in Shareholder Wealth 00014 00017 00088 00089 -00006 -00007 00072 00086
(051) (062) (129) (103) (027) (029) (131) (16)
Change in Shareholder Wealth -00102 00102 00105 00032 00102 00101
(097) (292) (293) (055) (611) (608)
Change Shareholder Wealth Lag1 Institutional
Concentration
00435 00764 00363 00253
(463) (258) (689) (144)
Lag1 Change Shareholder Wealth Lag1
Institutional Concentration
-00197 -00283
(122) (197)
Lag1 Institutional Concentration -492182 -288861 400149 453623
(09) (056) (098) (114)
Lag1 Total Institutional Ownership 720688 538372 655091 559587 682069 726226
(276) (232) (281) (314) (308) (322)
Lag1 Change Shareholder Wealth Lag1 Top 5
Institutional Ownership
-0046 -00689
(091) (256)
Lag1 Top 5 Institutional Ownership 583044 1191634
(092) (25)
Lag1 Tobinrsquos Q -32761 -36806 -19422 -18023 -54283 -48681 -29534 -24243
(138) (172) (077) (078) (227) (245) (136) (111)
Lag1 Market Capitalization 00004 00004 00002 00006 00007 00008 00007 00007
(042) (044) (022) (07) (086) (092) (082) (09)
CEO Dummy 768035 765474 768334 773634 616315 61729 619028 618535
(354) (355) (356) (355) (417) (418) (419) (418)
Adjusted R-Sq 00057 0006 00052 0005 00097 00097 00098 00102
Number of Observations 41305 41305 41305 41305 85518 85518 85518 85518
30
Table VI Level of Salary as a Function of Institutional Influence
Similar in parts to models (1) and (2) of HS (Table V) this table displays the coefficients from OLS regressions of salary against the lagged change and contemporary change in
shareholder wealth measures of institutional influence Tobinrsquos Q firm market capitalization and dummy controls for CEO industry and year effects We estimate models (1) to
(4) over the period 1992‒1997 using the actual data supplied by HS and models (5) to (8) over the whole sample 1992‒2002 Models (1) (2) (5) and (6) use Institutional
Concentration as a measure of institutional influence which HS define as the shares owned by the institutions with the five largest holdings divided by the shares held by all
institutions These models control for total institutional ownership Models (3) (4) (7) and (8) use Top 5 Ownership as a measure of institutional influence which we define as
the proportion of shares outstanding held by the institutions with the 5 largest holdings Coefficients for industry and year controls are not presented One two and three
asterisks denote significance at 10 5 and 1 levels
Hartzell and Starks Database 1992‒1997 Full Sample based on ExecuComp 1992‒2002
(1) (2) (3) (4) (5) (6) (7) (8)
Change in Shareholder Wealth 04125E-4 0003 -2439E-4 0003 00021 00012 00024 00012
(057) (464) (031) (468) (1004) (72) (1041) (706)
Lag1 Change in Shareholder Wealth -00062 7856E-4 -0007 9003E-4 -00014 00002 -00014 00003
(594) (098) (592) (112) (522) (123) (5) (145)
Lag1 Institutional Concentration -2475087 555134 -2734983 943317
(3245) (710) (4965) (1671)
Lag1 Total Institutional Ownership 1020109 217488 936337 252464
(1904) (432) (2471) (704)
Lag1 Top 5 Institutional Ownership -148355 31354 -87926 423166
(139) (332) (118) (643)
Lag1 Tobinrsquos Q -195348 -252191 -192344 -25158 -119837 -146373 -122132 -143821
(2812) (3762) (2721) (3752) (248) (274) (2583) (2727)
Lag1 Market Capitalization 844E-06 103E-5 00034 00039
(2781) (3051) (2874) (3104)
Lag1 Log Market Capitalization 750021 723625 776657 732279
(7511) (9285) (11224) (14039)
CEO Dummy 2683978 2702238 2668244 2702417 3281133 3285103 3292175 3283145
(8625) (9277) (8246) (9271) (12739) (13616) (12336) (13605)
Adjusted R-Sq 04604 05302 04187 05295 03856 04588 03476 04574
Number of Observations 36280 36280 36280 36280 97679 97679 97679 97679
31
Table VII Level of Total Direct Compensation as a Function of Institutional Influence
Similar in parts to models (3) and (5) of HS (Table V) this table displays the coefficients from OLS regressions of total direct compensation against the lagged change and
contemporary change in shareholder wealth measures of institutional influence Tobinrsquos Q firm market capitalization and dummy controls for CEO industry and year effects
We estimate models (1) to (4) over the period 1992‒1997 and models (5) to (8) over the whole sample Models (1) (2) (5) and (6) use Institutional Concentration as a measure
of institutional influence which HS define as the shares owned by the institutions with the five largest holdings divided by the shares held by all institutions These models
control for total institutional ownership Models (3) (4) (7) and (8) use Top 5 Ownership as a measure of institutional influence which we define as the proportion of shares
outstanding held by the institutions with the five largest holdings Coefficients for industry and year controls are not presented One two and three asterisks denote significance
at 10 5 and 1 levels
Hartzell and Starks Database 1992‒1997 Full Sample from ExecuComp 1992‒2002
(1) (2) (3) (4) (5) (6) (7) (8)
Change in Shareholder Wealth 005779 007006 005282 006791 00423 00148 0046 00138
(415) (581) (369) (565) (203) (07) (22) (066)
Lag1 Change in Shareholder Wealth 003182 00714 00272 00715 00613 0126 00614 01278
(126) (298) (107) (299) (275) (488) (276) (494)
Lag1 Institutional Concentration -1608633 3248998 -3424892 3334186
(151) (283) (1172) (118)
Lag1 Total Institutional Ownership 104013 5473128 1531013 -2926764
(1645) (831) (876) (158)
Lag1 Top 5 Institutional Ownership 188585 3927986 -2159177 -2640879
(016) (343) (067) (092)
Lag1 Tobinrsquos Q -445692 -803612 -454859 -833566 3257027 2971201 3213936 3086205
(427) (791) (432) (831) (83) (77) (815) (793)
Lag1 Market Capitalization 49E-5 615E-5 01031 01088
(113) (1365) (1555) (168)
Lag1 Log Market Capitalization 4683497 4759377 1645156 1427856
(2775) (3469) (3679) (4239)
CEO Dummy 1501307 1508886 1493688 1508736 4713079 4713629 47278 4703908
(2648) (2684) (2618) (2683) (282) (2828) (2819) (2824)
Adjusted R-Sq 01861 02060 01695 02051 01071 01113 01033 011
Number of Observations 33229 33229 33229 33229 97679 97679 97679 97679
32
Table VIII Two Stage Least Squares (Reverse Causality) Tests - Executive Compensation Measures as a Function of Institutional Concentration
This table shows the coefficients from the Two-Stage Least Squares (2SLS) regressions of executive compensation measures against the lagged change
and contemporary change in shareholder wealth instrumented institutional concentration and total institutional ownership Tobinrsquos Q the natural
logarithm of firm market capitalization and dummy variable controls for CEO and year effects We instrument measures of institutional concentration
and total institutional ownership by orthogonalizing with respect to option grant pay-for-performance sensitivity salary and total compensation
Coefficients for industry and year controls are not presented First stage results are not presented One two and three asterisks denote significance at
10 5 and 1 levels
ExecuComp Sample 1992‒1997 Full Sample from ExecuComp 1992‒2002
PPSOG Salary TC PPSOG Salary TC
Change in Shareholder Wealth 144E-05 0004 02133 -409E-06 00011 00157
(308) (775) (596) (691) (638) (068)
Lag1 Change in Shareholder Wealth -240E-05 -00004 00866 118E-06 00002 01422
(131) (087) (346) (097) (116) (51)
Lag1 Institutional Concentration -00898 668681 3263143 02024 916656 2856417
(052) (801) (176) (163) (1545) (1004)
Lag1 Total Institutional Ownership 01021 -43997 2512587 02983 -104171 -1028347
(09) (084) (153) (4) (269) (487)
Lag1 Tobinrsquos Q 00754 -225606 119297 00655 -158657 2997205
(596) (1649) (053) (834) (2535) (727)
Lag1 Log Market Capitalization -04234 828713 7864046 -04002 825106 1685638
(2195) (8132) (2977) (4106) (11133) (373)
CEO Dummy 15848 3214117 2950274 19149 3205862 4529101
(2223) (9389) (2255) (3802) (1331) (2758)
Overall R-Sq -- 04981 01206 -- 04738 01178
Number of Observations 41305 41305 41305 85518 85518 85518
33
Table IX Two Stage Least Squares (Reverse Causality) Tests - Executive Compensation Measures as a Function of Top 5 Ownership
This table shows the coefficients from 2SLS regressions of executive compensation measures against the lagged change and contemporary change
in shareholder wealth instrumented Top 5 institutional ownership Tobinrsquos Q natural logarithm of firm market capitalization and dummy variable
controls for CEO and year effects We instrument Top 5 institutional ownership by orthogonalizing with respect to option grant pay-for-
performance sensitivity salary and total compensation Coefficients for industry and year controls are not presented First stage results are not
presented One two and three asterisks denote significance at 10 5 and 1 levels
ExecuComp sample 1992‒1997 Full Sample 1992‒2002
PPSOG Salary TC PPSOG Salary TC
Change in Shareholder Wealth 136E-05 00041 02128 -407E-06 00011 00146
(293) (79) (596) (692) (619) (063)
Lag1 Change in Shareholder Wealth -241E-05 -00003 00868 116E-06 00003 01439
(133) (07) (347) (096) (142) (514)
Lag1 Top 5 Institutional Ownership -03774 -187512 -3848531 00179 -104261 -126659
(163) (186) (166) (011) (149) (436)
Lag1 Tobinrsquos Q 00731 -223886 10074 00642 -155467 3115073
(573) (1631) (045) (819) (2509) (75)
Lag1 Log Market Capitalization -0416 775637 7696302 -04015 764632 1462536
(2345) (10084) (3766) (3688) (13708) (4072)
CEO Dummy 15854 3211991 2950158 19152 3203152 4519385
(2223) (9381) (2254) (3804) (133) (2752)
Overall R-Sq -- 04967 01202 -- 04722 01168
Number of Observations 41305 41305 41305 85518 85518 85518
34
Table X Specification Tests utilizing Box-Cox Transformation of Market Capitalization
This table presents results from a Box-Cox transformation of market capitalization In
each model we specify transformation of market capitalization only Panel A shows
results for the period 1992‒1997 and Panel B shows results for the entire period
1992‒2002 The first part of each panel shows the Lambda coefficient that applies to
the transformed market capitalization The second part of each panel shows log
likelihood scores with LR Chi2 Statistics in parentheses ‒ these test for the
appropriateness of a linear or log model specification
Panel A Our ExecuComp sample 1992‒1997
(1) (2) (3)
PPSOG Salary TC
Lambda -05143 01881 02876
(-1914) (2413) (1329)
Test H0 (Log Likelihood Score)
Lambda = -1 -1279748 -3178575 -4795326
(23866) (85952) (99688)
Lambda = 0 -1280152 -3138456 -4791159
(31934) (57147) (16339)
Lambda = 1 -1283157 -3163807 -4793506
(92039) (564156) (63292)
Panel B Full ExecuComp sample 1992‒2002
(1) (2) (4)
PPSOG Salary TC
Lambda -044 01853 04036
(-3015) (40) (3666)
Test H0 (Log Likelihood Score)
Lambda = 0 -2688334 -6476494 -10384490
(66599) (160356) (126236)
Lambda = 1 -2694886 -6538506 -10386795
(197644) (1400605) (172345)
35
Figure 2 Institutional Ownership Measures (LHA) vs Total
Compensation (RHA) by Market Capitalization
0
10
20
30
40
50
60
70
1 2 3 4 5 6 7 8 9 10
$00
$10
$20
$30
$40
$50
$60
$70
$80
$90
Millio
ns
InstitutionalConcentration (LHS)Top5 Ownership (LHS)
Institutional TotalOwnership (LHS)Total Compensation(RHS)
Page 26
26
Table II Pairwise Correlation Coefficients
This table presents the correlations between the key variables based on our comprehensive
ExecuComp dataset Institutional Concentration is the shares owned by the institutions with the five
largest holdings divided by the shares held by all institutions Top 5 Ownership is the proportion of
shares outstanding held by the institutions with the five largest holdings Total Institutional
Ownership is the proportion of shares held by all institutions The pay-for-performance sensitivity
(PPS) of option grants is the change in executive wealth due to a $1000 change in shareholder
wealth
(1) (2) (3) (4) (5) (6) (7) (8)
(1) Institutional Concentration 1
(2) Top 5 Ownership 0268 1
(3) Total Ownership -0415 0706 1
(4) PPS of Option Grants 0122 0042 -0024 1
(5) Log Salary -0270 -0012 0171 0038 1
(6) Log Market Capitalization -0630 -0157 0261 -0156 0420 1
(7) Level Market Capitalization -0241 -0157 0007 -0058 0219 0534 1
(8) Tobins Q -0008 -0034 0002 0015 -0083 0066 0121 1
27
Table III
Descriptive Statistics by Market Capitalization Deciles
Market
Capitalization
Deciles
Market
Capitalization
($m)
Total
Compensation
($000)
Salary
($)
PPS Option
Grants
($)
Institutional
Concentration
()
Top 5 Ownership
()
Institutional
Total Ownership
()
1 (Small) 1181 76008 234700 2720 611 200 327
2 2668 90206 257955 1796 495 217 438
3 4193 105946 275120 1517 454 220 485
4 6045 129892 296067 1376 420 206 491
5 8742 155391 315864 1216 394 198 502
6 13086 191888 345942 1072 380 202 531
7 20790 239501 377021 0874 368 196 532
8 34920 314955 427808 0687 354 190 538
9 67747 450970 480319 0569 328 173 528
10 (Big) 366540 846066 614354 0297 275 138 502
28
Table IV Pay-for-Performance Sensitivity of Option Grants as a Function of Institutional Influence
Similar in construction to parts of HS (Table II) this table shows the coefficients from Tobit regressions of option grant pay-for-performance sensitivity per $1000 change in
shareholder wealth against the lagged change and contemporary change in shareholder wealth measures of institutional influence Tobinrsquos Q firm market capitalization and
dummy controls for CEO and year effects We estimate models (1) to (4) over the period 1992‒1997 using data supplied by HS and models (5) to (8) are estimated over the
whole sample Models (1) (2) (5) and (6) use Institutional Concentration as a measure of institutional influence which HS define as the shares owned by the institutions
with the five largest holdings divided by the shares held by all institutions These models control for total institutional ownership Models (3) (4) (7) and (8) use Top 5
Ownership as a measure of institutional influence which we define as the proportion of shares outstanding held by the institutions with the five largest holdings Coefficients
for industry and year controls are not presented One two and three asterisks denote significance at 10 5 and 1 levels
Hartzell and Starks Database 1992‒1997 Full Sample from ExecuComp 1992‒2002
(1) (2) (3) (4) (5) (6) (7) (8)
Change in Shareholder Wealth 77721 191781 89874 137368 -446 -218 -664 -203
(171) (429) (194) (317) (319) (168) (453) (158)
Lag1 Change in Shareholder Wealth 14446 99963 88719 8544 -144 0273 -0342 00852
(027) (212) (166) (184) (071) (015) (016) (005)
Lag1 Institutional Concentration 13994 -00692 26405 03304
(767) (032) (1402) (166)
Lag1 Total Institutional Ownership 08055 10434 07106 0937
(593) (781) (666) (898)
Lag1 Top 5 Institutional Ownership 10567 07457 12116 06293
(399) ( 283) (533) (278)
Lag1 Tobinrsquos Q 0068 00877 00647 00813 00498 00727 00508 00695
(337) (434) (319) (401) (504) (734) (51) (703)
Lag1 Market Capitalization -106202 -173515 -776 -117
(474) (742) (1011) (153)
Lag1 Log Market Capitalization -0297 -02319 -04111 -03908
(1576) (1439) (2997) (2841)
CEO Dummy 13745 13682 13765 13674 28197 28128 28155 28143
(1913) (1918) (1914) (1916) (3388) (3396) (3377) (34)
Number of Observations 33878 33878 33878 33878 97679 97679 97679 97679
29
Table V Pay-for-Performance Sensitivity of Cash Compensation (Salary plus Bonus) as a Function of Institutional Influence - Level Market Capitalization
Similar in parts to models (1) and (2) of HS (Table IV) this table displays the coefficients from OLS regressions of the change in cash compensation against the lagged
change and contemporary change in shareholder wealth measures of institutional influence Tobinrsquos Q firm market capitalization and dummy controls for CEO and year
effects We estimate models (1) to (4) over the period 1992‒1997 based on our ExecuComp sample and models (5) to (8) over the whole sample Models (1) (2) (5) and (6)
use Institutional Concentration as a measure of institutional influence which HS define as the shares owned by the institutions with the five largest holdings divided by the
shares held by all institutions These models control for total institutional ownership Models (4) and (8) use Top 5 Ownership as a measure of institutional influence which
we define as the proportion of shares outstanding held by the institutions with the 5 largest holdings Coefficients for industry and year controls are not presented One two
and three asterisks denote significance at 10 5 and 1 levels
Our Database based on ExecuComp 1992‒1997 Full Sample based on ExecuComp 1992‒2002
(1) (2) (3) (4) (5) (6) (7) (8)
Lag1 Change in Shareholder Wealth 00014 00017 00088 00089 -00006 -00007 00072 00086
(051) (062) (129) (103) (027) (029) (131) (16)
Change in Shareholder Wealth -00102 00102 00105 00032 00102 00101
(097) (292) (293) (055) (611) (608)
Change Shareholder Wealth Lag1 Institutional
Concentration
00435 00764 00363 00253
(463) (258) (689) (144)
Lag1 Change Shareholder Wealth Lag1
Institutional Concentration
-00197 -00283
(122) (197)
Lag1 Institutional Concentration -492182 -288861 400149 453623
(09) (056) (098) (114)
Lag1 Total Institutional Ownership 720688 538372 655091 559587 682069 726226
(276) (232) (281) (314) (308) (322)
Lag1 Change Shareholder Wealth Lag1 Top 5
Institutional Ownership
-0046 -00689
(091) (256)
Lag1 Top 5 Institutional Ownership 583044 1191634
(092) (25)
Lag1 Tobinrsquos Q -32761 -36806 -19422 -18023 -54283 -48681 -29534 -24243
(138) (172) (077) (078) (227) (245) (136) (111)
Lag1 Market Capitalization 00004 00004 00002 00006 00007 00008 00007 00007
(042) (044) (022) (07) (086) (092) (082) (09)
CEO Dummy 768035 765474 768334 773634 616315 61729 619028 618535
(354) (355) (356) (355) (417) (418) (419) (418)
Adjusted R-Sq 00057 0006 00052 0005 00097 00097 00098 00102
Number of Observations 41305 41305 41305 41305 85518 85518 85518 85518
30
Table VI Level of Salary as a Function of Institutional Influence
Similar in parts to models (1) and (2) of HS (Table V) this table displays the coefficients from OLS regressions of salary against the lagged change and contemporary change in
shareholder wealth measures of institutional influence Tobinrsquos Q firm market capitalization and dummy controls for CEO industry and year effects We estimate models (1) to
(4) over the period 1992‒1997 using the actual data supplied by HS and models (5) to (8) over the whole sample 1992‒2002 Models (1) (2) (5) and (6) use Institutional
Concentration as a measure of institutional influence which HS define as the shares owned by the institutions with the five largest holdings divided by the shares held by all
institutions These models control for total institutional ownership Models (3) (4) (7) and (8) use Top 5 Ownership as a measure of institutional influence which we define as
the proportion of shares outstanding held by the institutions with the 5 largest holdings Coefficients for industry and year controls are not presented One two and three
asterisks denote significance at 10 5 and 1 levels
Hartzell and Starks Database 1992‒1997 Full Sample based on ExecuComp 1992‒2002
(1) (2) (3) (4) (5) (6) (7) (8)
Change in Shareholder Wealth 04125E-4 0003 -2439E-4 0003 00021 00012 00024 00012
(057) (464) (031) (468) (1004) (72) (1041) (706)
Lag1 Change in Shareholder Wealth -00062 7856E-4 -0007 9003E-4 -00014 00002 -00014 00003
(594) (098) (592) (112) (522) (123) (5) (145)
Lag1 Institutional Concentration -2475087 555134 -2734983 943317
(3245) (710) (4965) (1671)
Lag1 Total Institutional Ownership 1020109 217488 936337 252464
(1904) (432) (2471) (704)
Lag1 Top 5 Institutional Ownership -148355 31354 -87926 423166
(139) (332) (118) (643)
Lag1 Tobinrsquos Q -195348 -252191 -192344 -25158 -119837 -146373 -122132 -143821
(2812) (3762) (2721) (3752) (248) (274) (2583) (2727)
Lag1 Market Capitalization 844E-06 103E-5 00034 00039
(2781) (3051) (2874) (3104)
Lag1 Log Market Capitalization 750021 723625 776657 732279
(7511) (9285) (11224) (14039)
CEO Dummy 2683978 2702238 2668244 2702417 3281133 3285103 3292175 3283145
(8625) (9277) (8246) (9271) (12739) (13616) (12336) (13605)
Adjusted R-Sq 04604 05302 04187 05295 03856 04588 03476 04574
Number of Observations 36280 36280 36280 36280 97679 97679 97679 97679
31
Table VII Level of Total Direct Compensation as a Function of Institutional Influence
Similar in parts to models (3) and (5) of HS (Table V) this table displays the coefficients from OLS regressions of total direct compensation against the lagged change and
contemporary change in shareholder wealth measures of institutional influence Tobinrsquos Q firm market capitalization and dummy controls for CEO industry and year effects
We estimate models (1) to (4) over the period 1992‒1997 and models (5) to (8) over the whole sample Models (1) (2) (5) and (6) use Institutional Concentration as a measure
of institutional influence which HS define as the shares owned by the institutions with the five largest holdings divided by the shares held by all institutions These models
control for total institutional ownership Models (3) (4) (7) and (8) use Top 5 Ownership as a measure of institutional influence which we define as the proportion of shares
outstanding held by the institutions with the five largest holdings Coefficients for industry and year controls are not presented One two and three asterisks denote significance
at 10 5 and 1 levels
Hartzell and Starks Database 1992‒1997 Full Sample from ExecuComp 1992‒2002
(1) (2) (3) (4) (5) (6) (7) (8)
Change in Shareholder Wealth 005779 007006 005282 006791 00423 00148 0046 00138
(415) (581) (369) (565) (203) (07) (22) (066)
Lag1 Change in Shareholder Wealth 003182 00714 00272 00715 00613 0126 00614 01278
(126) (298) (107) (299) (275) (488) (276) (494)
Lag1 Institutional Concentration -1608633 3248998 -3424892 3334186
(151) (283) (1172) (118)
Lag1 Total Institutional Ownership 104013 5473128 1531013 -2926764
(1645) (831) (876) (158)
Lag1 Top 5 Institutional Ownership 188585 3927986 -2159177 -2640879
(016) (343) (067) (092)
Lag1 Tobinrsquos Q -445692 -803612 -454859 -833566 3257027 2971201 3213936 3086205
(427) (791) (432) (831) (83) (77) (815) (793)
Lag1 Market Capitalization 49E-5 615E-5 01031 01088
(113) (1365) (1555) (168)
Lag1 Log Market Capitalization 4683497 4759377 1645156 1427856
(2775) (3469) (3679) (4239)
CEO Dummy 1501307 1508886 1493688 1508736 4713079 4713629 47278 4703908
(2648) (2684) (2618) (2683) (282) (2828) (2819) (2824)
Adjusted R-Sq 01861 02060 01695 02051 01071 01113 01033 011
Number of Observations 33229 33229 33229 33229 97679 97679 97679 97679
32
Table VIII Two Stage Least Squares (Reverse Causality) Tests - Executive Compensation Measures as a Function of Institutional Concentration
This table shows the coefficients from the Two-Stage Least Squares (2SLS) regressions of executive compensation measures against the lagged change
and contemporary change in shareholder wealth instrumented institutional concentration and total institutional ownership Tobinrsquos Q the natural
logarithm of firm market capitalization and dummy variable controls for CEO and year effects We instrument measures of institutional concentration
and total institutional ownership by orthogonalizing with respect to option grant pay-for-performance sensitivity salary and total compensation
Coefficients for industry and year controls are not presented First stage results are not presented One two and three asterisks denote significance at
10 5 and 1 levels
ExecuComp Sample 1992‒1997 Full Sample from ExecuComp 1992‒2002
PPSOG Salary TC PPSOG Salary TC
Change in Shareholder Wealth 144E-05 0004 02133 -409E-06 00011 00157
(308) (775) (596) (691) (638) (068)
Lag1 Change in Shareholder Wealth -240E-05 -00004 00866 118E-06 00002 01422
(131) (087) (346) (097) (116) (51)
Lag1 Institutional Concentration -00898 668681 3263143 02024 916656 2856417
(052) (801) (176) (163) (1545) (1004)
Lag1 Total Institutional Ownership 01021 -43997 2512587 02983 -104171 -1028347
(09) (084) (153) (4) (269) (487)
Lag1 Tobinrsquos Q 00754 -225606 119297 00655 -158657 2997205
(596) (1649) (053) (834) (2535) (727)
Lag1 Log Market Capitalization -04234 828713 7864046 -04002 825106 1685638
(2195) (8132) (2977) (4106) (11133) (373)
CEO Dummy 15848 3214117 2950274 19149 3205862 4529101
(2223) (9389) (2255) (3802) (1331) (2758)
Overall R-Sq -- 04981 01206 -- 04738 01178
Number of Observations 41305 41305 41305 85518 85518 85518
33
Table IX Two Stage Least Squares (Reverse Causality) Tests - Executive Compensation Measures as a Function of Top 5 Ownership
This table shows the coefficients from 2SLS regressions of executive compensation measures against the lagged change and contemporary change
in shareholder wealth instrumented Top 5 institutional ownership Tobinrsquos Q natural logarithm of firm market capitalization and dummy variable
controls for CEO and year effects We instrument Top 5 institutional ownership by orthogonalizing with respect to option grant pay-for-
performance sensitivity salary and total compensation Coefficients for industry and year controls are not presented First stage results are not
presented One two and three asterisks denote significance at 10 5 and 1 levels
ExecuComp sample 1992‒1997 Full Sample 1992‒2002
PPSOG Salary TC PPSOG Salary TC
Change in Shareholder Wealth 136E-05 00041 02128 -407E-06 00011 00146
(293) (79) (596) (692) (619) (063)
Lag1 Change in Shareholder Wealth -241E-05 -00003 00868 116E-06 00003 01439
(133) (07) (347) (096) (142) (514)
Lag1 Top 5 Institutional Ownership -03774 -187512 -3848531 00179 -104261 -126659
(163) (186) (166) (011) (149) (436)
Lag1 Tobinrsquos Q 00731 -223886 10074 00642 -155467 3115073
(573) (1631) (045) (819) (2509) (75)
Lag1 Log Market Capitalization -0416 775637 7696302 -04015 764632 1462536
(2345) (10084) (3766) (3688) (13708) (4072)
CEO Dummy 15854 3211991 2950158 19152 3203152 4519385
(2223) (9381) (2254) (3804) (133) (2752)
Overall R-Sq -- 04967 01202 -- 04722 01168
Number of Observations 41305 41305 41305 85518 85518 85518
34
Table X Specification Tests utilizing Box-Cox Transformation of Market Capitalization
This table presents results from a Box-Cox transformation of market capitalization In
each model we specify transformation of market capitalization only Panel A shows
results for the period 1992‒1997 and Panel B shows results for the entire period
1992‒2002 The first part of each panel shows the Lambda coefficient that applies to
the transformed market capitalization The second part of each panel shows log
likelihood scores with LR Chi2 Statistics in parentheses ‒ these test for the
appropriateness of a linear or log model specification
Panel A Our ExecuComp sample 1992‒1997
(1) (2) (3)
PPSOG Salary TC
Lambda -05143 01881 02876
(-1914) (2413) (1329)
Test H0 (Log Likelihood Score)
Lambda = -1 -1279748 -3178575 -4795326
(23866) (85952) (99688)
Lambda = 0 -1280152 -3138456 -4791159
(31934) (57147) (16339)
Lambda = 1 -1283157 -3163807 -4793506
(92039) (564156) (63292)
Panel B Full ExecuComp sample 1992‒2002
(1) (2) (4)
PPSOG Salary TC
Lambda -044 01853 04036
(-3015) (40) (3666)
Test H0 (Log Likelihood Score)
Lambda = 0 -2688334 -6476494 -10384490
(66599) (160356) (126236)
Lambda = 1 -2694886 -6538506 -10386795
(197644) (1400605) (172345)
35
Figure 2 Institutional Ownership Measures (LHA) vs Total
Compensation (RHA) by Market Capitalization
0
10
20
30
40
50
60
70
1 2 3 4 5 6 7 8 9 10
$00
$10
$20
$30
$40
$50
$60
$70
$80
$90
Millio
ns
InstitutionalConcentration (LHS)Top5 Ownership (LHS)
Institutional TotalOwnership (LHS)Total Compensation(RHS)
Page 27
27
Table III
Descriptive Statistics by Market Capitalization Deciles
Market
Capitalization
Deciles
Market
Capitalization
($m)
Total
Compensation
($000)
Salary
($)
PPS Option
Grants
($)
Institutional
Concentration
()
Top 5 Ownership
()
Institutional
Total Ownership
()
1 (Small) 1181 76008 234700 2720 611 200 327
2 2668 90206 257955 1796 495 217 438
3 4193 105946 275120 1517 454 220 485
4 6045 129892 296067 1376 420 206 491
5 8742 155391 315864 1216 394 198 502
6 13086 191888 345942 1072 380 202 531
7 20790 239501 377021 0874 368 196 532
8 34920 314955 427808 0687 354 190 538
9 67747 450970 480319 0569 328 173 528
10 (Big) 366540 846066 614354 0297 275 138 502
28
Table IV Pay-for-Performance Sensitivity of Option Grants as a Function of Institutional Influence
Similar in construction to parts of HS (Table II) this table shows the coefficients from Tobit regressions of option grant pay-for-performance sensitivity per $1000 change in
shareholder wealth against the lagged change and contemporary change in shareholder wealth measures of institutional influence Tobinrsquos Q firm market capitalization and
dummy controls for CEO and year effects We estimate models (1) to (4) over the period 1992‒1997 using data supplied by HS and models (5) to (8) are estimated over the
whole sample Models (1) (2) (5) and (6) use Institutional Concentration as a measure of institutional influence which HS define as the shares owned by the institutions
with the five largest holdings divided by the shares held by all institutions These models control for total institutional ownership Models (3) (4) (7) and (8) use Top 5
Ownership as a measure of institutional influence which we define as the proportion of shares outstanding held by the institutions with the five largest holdings Coefficients
for industry and year controls are not presented One two and three asterisks denote significance at 10 5 and 1 levels
Hartzell and Starks Database 1992‒1997 Full Sample from ExecuComp 1992‒2002
(1) (2) (3) (4) (5) (6) (7) (8)
Change in Shareholder Wealth 77721 191781 89874 137368 -446 -218 -664 -203
(171) (429) (194) (317) (319) (168) (453) (158)
Lag1 Change in Shareholder Wealth 14446 99963 88719 8544 -144 0273 -0342 00852
(027) (212) (166) (184) (071) (015) (016) (005)
Lag1 Institutional Concentration 13994 -00692 26405 03304
(767) (032) (1402) (166)
Lag1 Total Institutional Ownership 08055 10434 07106 0937
(593) (781) (666) (898)
Lag1 Top 5 Institutional Ownership 10567 07457 12116 06293
(399) ( 283) (533) (278)
Lag1 Tobinrsquos Q 0068 00877 00647 00813 00498 00727 00508 00695
(337) (434) (319) (401) (504) (734) (51) (703)
Lag1 Market Capitalization -106202 -173515 -776 -117
(474) (742) (1011) (153)
Lag1 Log Market Capitalization -0297 -02319 -04111 -03908
(1576) (1439) (2997) (2841)
CEO Dummy 13745 13682 13765 13674 28197 28128 28155 28143
(1913) (1918) (1914) (1916) (3388) (3396) (3377) (34)
Number of Observations 33878 33878 33878 33878 97679 97679 97679 97679
29
Table V Pay-for-Performance Sensitivity of Cash Compensation (Salary plus Bonus) as a Function of Institutional Influence - Level Market Capitalization
Similar in parts to models (1) and (2) of HS (Table IV) this table displays the coefficients from OLS regressions of the change in cash compensation against the lagged
change and contemporary change in shareholder wealth measures of institutional influence Tobinrsquos Q firm market capitalization and dummy controls for CEO and year
effects We estimate models (1) to (4) over the period 1992‒1997 based on our ExecuComp sample and models (5) to (8) over the whole sample Models (1) (2) (5) and (6)
use Institutional Concentration as a measure of institutional influence which HS define as the shares owned by the institutions with the five largest holdings divided by the
shares held by all institutions These models control for total institutional ownership Models (4) and (8) use Top 5 Ownership as a measure of institutional influence which
we define as the proportion of shares outstanding held by the institutions with the 5 largest holdings Coefficients for industry and year controls are not presented One two
and three asterisks denote significance at 10 5 and 1 levels
Our Database based on ExecuComp 1992‒1997 Full Sample based on ExecuComp 1992‒2002
(1) (2) (3) (4) (5) (6) (7) (8)
Lag1 Change in Shareholder Wealth 00014 00017 00088 00089 -00006 -00007 00072 00086
(051) (062) (129) (103) (027) (029) (131) (16)
Change in Shareholder Wealth -00102 00102 00105 00032 00102 00101
(097) (292) (293) (055) (611) (608)
Change Shareholder Wealth Lag1 Institutional
Concentration
00435 00764 00363 00253
(463) (258) (689) (144)
Lag1 Change Shareholder Wealth Lag1
Institutional Concentration
-00197 -00283
(122) (197)
Lag1 Institutional Concentration -492182 -288861 400149 453623
(09) (056) (098) (114)
Lag1 Total Institutional Ownership 720688 538372 655091 559587 682069 726226
(276) (232) (281) (314) (308) (322)
Lag1 Change Shareholder Wealth Lag1 Top 5
Institutional Ownership
-0046 -00689
(091) (256)
Lag1 Top 5 Institutional Ownership 583044 1191634
(092) (25)
Lag1 Tobinrsquos Q -32761 -36806 -19422 -18023 -54283 -48681 -29534 -24243
(138) (172) (077) (078) (227) (245) (136) (111)
Lag1 Market Capitalization 00004 00004 00002 00006 00007 00008 00007 00007
(042) (044) (022) (07) (086) (092) (082) (09)
CEO Dummy 768035 765474 768334 773634 616315 61729 619028 618535
(354) (355) (356) (355) (417) (418) (419) (418)
Adjusted R-Sq 00057 0006 00052 0005 00097 00097 00098 00102
Number of Observations 41305 41305 41305 41305 85518 85518 85518 85518
30
Table VI Level of Salary as a Function of Institutional Influence
Similar in parts to models (1) and (2) of HS (Table V) this table displays the coefficients from OLS regressions of salary against the lagged change and contemporary change in
shareholder wealth measures of institutional influence Tobinrsquos Q firm market capitalization and dummy controls for CEO industry and year effects We estimate models (1) to
(4) over the period 1992‒1997 using the actual data supplied by HS and models (5) to (8) over the whole sample 1992‒2002 Models (1) (2) (5) and (6) use Institutional
Concentration as a measure of institutional influence which HS define as the shares owned by the institutions with the five largest holdings divided by the shares held by all
institutions These models control for total institutional ownership Models (3) (4) (7) and (8) use Top 5 Ownership as a measure of institutional influence which we define as
the proportion of shares outstanding held by the institutions with the 5 largest holdings Coefficients for industry and year controls are not presented One two and three
asterisks denote significance at 10 5 and 1 levels
Hartzell and Starks Database 1992‒1997 Full Sample based on ExecuComp 1992‒2002
(1) (2) (3) (4) (5) (6) (7) (8)
Change in Shareholder Wealth 04125E-4 0003 -2439E-4 0003 00021 00012 00024 00012
(057) (464) (031) (468) (1004) (72) (1041) (706)
Lag1 Change in Shareholder Wealth -00062 7856E-4 -0007 9003E-4 -00014 00002 -00014 00003
(594) (098) (592) (112) (522) (123) (5) (145)
Lag1 Institutional Concentration -2475087 555134 -2734983 943317
(3245) (710) (4965) (1671)
Lag1 Total Institutional Ownership 1020109 217488 936337 252464
(1904) (432) (2471) (704)
Lag1 Top 5 Institutional Ownership -148355 31354 -87926 423166
(139) (332) (118) (643)
Lag1 Tobinrsquos Q -195348 -252191 -192344 -25158 -119837 -146373 -122132 -143821
(2812) (3762) (2721) (3752) (248) (274) (2583) (2727)
Lag1 Market Capitalization 844E-06 103E-5 00034 00039
(2781) (3051) (2874) (3104)
Lag1 Log Market Capitalization 750021 723625 776657 732279
(7511) (9285) (11224) (14039)
CEO Dummy 2683978 2702238 2668244 2702417 3281133 3285103 3292175 3283145
(8625) (9277) (8246) (9271) (12739) (13616) (12336) (13605)
Adjusted R-Sq 04604 05302 04187 05295 03856 04588 03476 04574
Number of Observations 36280 36280 36280 36280 97679 97679 97679 97679
31
Table VII Level of Total Direct Compensation as a Function of Institutional Influence
Similar in parts to models (3) and (5) of HS (Table V) this table displays the coefficients from OLS regressions of total direct compensation against the lagged change and
contemporary change in shareholder wealth measures of institutional influence Tobinrsquos Q firm market capitalization and dummy controls for CEO industry and year effects
We estimate models (1) to (4) over the period 1992‒1997 and models (5) to (8) over the whole sample Models (1) (2) (5) and (6) use Institutional Concentration as a measure
of institutional influence which HS define as the shares owned by the institutions with the five largest holdings divided by the shares held by all institutions These models
control for total institutional ownership Models (3) (4) (7) and (8) use Top 5 Ownership as a measure of institutional influence which we define as the proportion of shares
outstanding held by the institutions with the five largest holdings Coefficients for industry and year controls are not presented One two and three asterisks denote significance
at 10 5 and 1 levels
Hartzell and Starks Database 1992‒1997 Full Sample from ExecuComp 1992‒2002
(1) (2) (3) (4) (5) (6) (7) (8)
Change in Shareholder Wealth 005779 007006 005282 006791 00423 00148 0046 00138
(415) (581) (369) (565) (203) (07) (22) (066)
Lag1 Change in Shareholder Wealth 003182 00714 00272 00715 00613 0126 00614 01278
(126) (298) (107) (299) (275) (488) (276) (494)
Lag1 Institutional Concentration -1608633 3248998 -3424892 3334186
(151) (283) (1172) (118)
Lag1 Total Institutional Ownership 104013 5473128 1531013 -2926764
(1645) (831) (876) (158)
Lag1 Top 5 Institutional Ownership 188585 3927986 -2159177 -2640879
(016) (343) (067) (092)
Lag1 Tobinrsquos Q -445692 -803612 -454859 -833566 3257027 2971201 3213936 3086205
(427) (791) (432) (831) (83) (77) (815) (793)
Lag1 Market Capitalization 49E-5 615E-5 01031 01088
(113) (1365) (1555) (168)
Lag1 Log Market Capitalization 4683497 4759377 1645156 1427856
(2775) (3469) (3679) (4239)
CEO Dummy 1501307 1508886 1493688 1508736 4713079 4713629 47278 4703908
(2648) (2684) (2618) (2683) (282) (2828) (2819) (2824)
Adjusted R-Sq 01861 02060 01695 02051 01071 01113 01033 011
Number of Observations 33229 33229 33229 33229 97679 97679 97679 97679
32
Table VIII Two Stage Least Squares (Reverse Causality) Tests - Executive Compensation Measures as a Function of Institutional Concentration
This table shows the coefficients from the Two-Stage Least Squares (2SLS) regressions of executive compensation measures against the lagged change
and contemporary change in shareholder wealth instrumented institutional concentration and total institutional ownership Tobinrsquos Q the natural
logarithm of firm market capitalization and dummy variable controls for CEO and year effects We instrument measures of institutional concentration
and total institutional ownership by orthogonalizing with respect to option grant pay-for-performance sensitivity salary and total compensation
Coefficients for industry and year controls are not presented First stage results are not presented One two and three asterisks denote significance at
10 5 and 1 levels
ExecuComp Sample 1992‒1997 Full Sample from ExecuComp 1992‒2002
PPSOG Salary TC PPSOG Salary TC
Change in Shareholder Wealth 144E-05 0004 02133 -409E-06 00011 00157
(308) (775) (596) (691) (638) (068)
Lag1 Change in Shareholder Wealth -240E-05 -00004 00866 118E-06 00002 01422
(131) (087) (346) (097) (116) (51)
Lag1 Institutional Concentration -00898 668681 3263143 02024 916656 2856417
(052) (801) (176) (163) (1545) (1004)
Lag1 Total Institutional Ownership 01021 -43997 2512587 02983 -104171 -1028347
(09) (084) (153) (4) (269) (487)
Lag1 Tobinrsquos Q 00754 -225606 119297 00655 -158657 2997205
(596) (1649) (053) (834) (2535) (727)
Lag1 Log Market Capitalization -04234 828713 7864046 -04002 825106 1685638
(2195) (8132) (2977) (4106) (11133) (373)
CEO Dummy 15848 3214117 2950274 19149 3205862 4529101
(2223) (9389) (2255) (3802) (1331) (2758)
Overall R-Sq -- 04981 01206 -- 04738 01178
Number of Observations 41305 41305 41305 85518 85518 85518
33
Table IX Two Stage Least Squares (Reverse Causality) Tests - Executive Compensation Measures as a Function of Top 5 Ownership
This table shows the coefficients from 2SLS regressions of executive compensation measures against the lagged change and contemporary change
in shareholder wealth instrumented Top 5 institutional ownership Tobinrsquos Q natural logarithm of firm market capitalization and dummy variable
controls for CEO and year effects We instrument Top 5 institutional ownership by orthogonalizing with respect to option grant pay-for-
performance sensitivity salary and total compensation Coefficients for industry and year controls are not presented First stage results are not
presented One two and three asterisks denote significance at 10 5 and 1 levels
ExecuComp sample 1992‒1997 Full Sample 1992‒2002
PPSOG Salary TC PPSOG Salary TC
Change in Shareholder Wealth 136E-05 00041 02128 -407E-06 00011 00146
(293) (79) (596) (692) (619) (063)
Lag1 Change in Shareholder Wealth -241E-05 -00003 00868 116E-06 00003 01439
(133) (07) (347) (096) (142) (514)
Lag1 Top 5 Institutional Ownership -03774 -187512 -3848531 00179 -104261 -126659
(163) (186) (166) (011) (149) (436)
Lag1 Tobinrsquos Q 00731 -223886 10074 00642 -155467 3115073
(573) (1631) (045) (819) (2509) (75)
Lag1 Log Market Capitalization -0416 775637 7696302 -04015 764632 1462536
(2345) (10084) (3766) (3688) (13708) (4072)
CEO Dummy 15854 3211991 2950158 19152 3203152 4519385
(2223) (9381) (2254) (3804) (133) (2752)
Overall R-Sq -- 04967 01202 -- 04722 01168
Number of Observations 41305 41305 41305 85518 85518 85518
34
Table X Specification Tests utilizing Box-Cox Transformation of Market Capitalization
This table presents results from a Box-Cox transformation of market capitalization In
each model we specify transformation of market capitalization only Panel A shows
results for the period 1992‒1997 and Panel B shows results for the entire period
1992‒2002 The first part of each panel shows the Lambda coefficient that applies to
the transformed market capitalization The second part of each panel shows log
likelihood scores with LR Chi2 Statistics in parentheses ‒ these test for the
appropriateness of a linear or log model specification
Panel A Our ExecuComp sample 1992‒1997
(1) (2) (3)
PPSOG Salary TC
Lambda -05143 01881 02876
(-1914) (2413) (1329)
Test H0 (Log Likelihood Score)
Lambda = -1 -1279748 -3178575 -4795326
(23866) (85952) (99688)
Lambda = 0 -1280152 -3138456 -4791159
(31934) (57147) (16339)
Lambda = 1 -1283157 -3163807 -4793506
(92039) (564156) (63292)
Panel B Full ExecuComp sample 1992‒2002
(1) (2) (4)
PPSOG Salary TC
Lambda -044 01853 04036
(-3015) (40) (3666)
Test H0 (Log Likelihood Score)
Lambda = 0 -2688334 -6476494 -10384490
(66599) (160356) (126236)
Lambda = 1 -2694886 -6538506 -10386795
(197644) (1400605) (172345)
35
Figure 2 Institutional Ownership Measures (LHA) vs Total
Compensation (RHA) by Market Capitalization
0
10
20
30
40
50
60
70
1 2 3 4 5 6 7 8 9 10
$00
$10
$20
$30
$40
$50
$60
$70
$80
$90
Millio
ns
InstitutionalConcentration (LHS)Top5 Ownership (LHS)
Institutional TotalOwnership (LHS)Total Compensation(RHS)
Page 28
28
Table IV Pay-for-Performance Sensitivity of Option Grants as a Function of Institutional Influence
Similar in construction to parts of HS (Table II) this table shows the coefficients from Tobit regressions of option grant pay-for-performance sensitivity per $1000 change in
shareholder wealth against the lagged change and contemporary change in shareholder wealth measures of institutional influence Tobinrsquos Q firm market capitalization and
dummy controls for CEO and year effects We estimate models (1) to (4) over the period 1992‒1997 using data supplied by HS and models (5) to (8) are estimated over the
whole sample Models (1) (2) (5) and (6) use Institutional Concentration as a measure of institutional influence which HS define as the shares owned by the institutions
with the five largest holdings divided by the shares held by all institutions These models control for total institutional ownership Models (3) (4) (7) and (8) use Top 5
Ownership as a measure of institutional influence which we define as the proportion of shares outstanding held by the institutions with the five largest holdings Coefficients
for industry and year controls are not presented One two and three asterisks denote significance at 10 5 and 1 levels
Hartzell and Starks Database 1992‒1997 Full Sample from ExecuComp 1992‒2002
(1) (2) (3) (4) (5) (6) (7) (8)
Change in Shareholder Wealth 77721 191781 89874 137368 -446 -218 -664 -203
(171) (429) (194) (317) (319) (168) (453) (158)
Lag1 Change in Shareholder Wealth 14446 99963 88719 8544 -144 0273 -0342 00852
(027) (212) (166) (184) (071) (015) (016) (005)
Lag1 Institutional Concentration 13994 -00692 26405 03304
(767) (032) (1402) (166)
Lag1 Total Institutional Ownership 08055 10434 07106 0937
(593) (781) (666) (898)
Lag1 Top 5 Institutional Ownership 10567 07457 12116 06293
(399) ( 283) (533) (278)
Lag1 Tobinrsquos Q 0068 00877 00647 00813 00498 00727 00508 00695
(337) (434) (319) (401) (504) (734) (51) (703)
Lag1 Market Capitalization -106202 -173515 -776 -117
(474) (742) (1011) (153)
Lag1 Log Market Capitalization -0297 -02319 -04111 -03908
(1576) (1439) (2997) (2841)
CEO Dummy 13745 13682 13765 13674 28197 28128 28155 28143
(1913) (1918) (1914) (1916) (3388) (3396) (3377) (34)
Number of Observations 33878 33878 33878 33878 97679 97679 97679 97679
29
Table V Pay-for-Performance Sensitivity of Cash Compensation (Salary plus Bonus) as a Function of Institutional Influence - Level Market Capitalization
Similar in parts to models (1) and (2) of HS (Table IV) this table displays the coefficients from OLS regressions of the change in cash compensation against the lagged
change and contemporary change in shareholder wealth measures of institutional influence Tobinrsquos Q firm market capitalization and dummy controls for CEO and year
effects We estimate models (1) to (4) over the period 1992‒1997 based on our ExecuComp sample and models (5) to (8) over the whole sample Models (1) (2) (5) and (6)
use Institutional Concentration as a measure of institutional influence which HS define as the shares owned by the institutions with the five largest holdings divided by the
shares held by all institutions These models control for total institutional ownership Models (4) and (8) use Top 5 Ownership as a measure of institutional influence which
we define as the proportion of shares outstanding held by the institutions with the 5 largest holdings Coefficients for industry and year controls are not presented One two
and three asterisks denote significance at 10 5 and 1 levels
Our Database based on ExecuComp 1992‒1997 Full Sample based on ExecuComp 1992‒2002
(1) (2) (3) (4) (5) (6) (7) (8)
Lag1 Change in Shareholder Wealth 00014 00017 00088 00089 -00006 -00007 00072 00086
(051) (062) (129) (103) (027) (029) (131) (16)
Change in Shareholder Wealth -00102 00102 00105 00032 00102 00101
(097) (292) (293) (055) (611) (608)
Change Shareholder Wealth Lag1 Institutional
Concentration
00435 00764 00363 00253
(463) (258) (689) (144)
Lag1 Change Shareholder Wealth Lag1
Institutional Concentration
-00197 -00283
(122) (197)
Lag1 Institutional Concentration -492182 -288861 400149 453623
(09) (056) (098) (114)
Lag1 Total Institutional Ownership 720688 538372 655091 559587 682069 726226
(276) (232) (281) (314) (308) (322)
Lag1 Change Shareholder Wealth Lag1 Top 5
Institutional Ownership
-0046 -00689
(091) (256)
Lag1 Top 5 Institutional Ownership 583044 1191634
(092) (25)
Lag1 Tobinrsquos Q -32761 -36806 -19422 -18023 -54283 -48681 -29534 -24243
(138) (172) (077) (078) (227) (245) (136) (111)
Lag1 Market Capitalization 00004 00004 00002 00006 00007 00008 00007 00007
(042) (044) (022) (07) (086) (092) (082) (09)
CEO Dummy 768035 765474 768334 773634 616315 61729 619028 618535
(354) (355) (356) (355) (417) (418) (419) (418)
Adjusted R-Sq 00057 0006 00052 0005 00097 00097 00098 00102
Number of Observations 41305 41305 41305 41305 85518 85518 85518 85518
30
Table VI Level of Salary as a Function of Institutional Influence
Similar in parts to models (1) and (2) of HS (Table V) this table displays the coefficients from OLS regressions of salary against the lagged change and contemporary change in
shareholder wealth measures of institutional influence Tobinrsquos Q firm market capitalization and dummy controls for CEO industry and year effects We estimate models (1) to
(4) over the period 1992‒1997 using the actual data supplied by HS and models (5) to (8) over the whole sample 1992‒2002 Models (1) (2) (5) and (6) use Institutional
Concentration as a measure of institutional influence which HS define as the shares owned by the institutions with the five largest holdings divided by the shares held by all
institutions These models control for total institutional ownership Models (3) (4) (7) and (8) use Top 5 Ownership as a measure of institutional influence which we define as
the proportion of shares outstanding held by the institutions with the 5 largest holdings Coefficients for industry and year controls are not presented One two and three
asterisks denote significance at 10 5 and 1 levels
Hartzell and Starks Database 1992‒1997 Full Sample based on ExecuComp 1992‒2002
(1) (2) (3) (4) (5) (6) (7) (8)
Change in Shareholder Wealth 04125E-4 0003 -2439E-4 0003 00021 00012 00024 00012
(057) (464) (031) (468) (1004) (72) (1041) (706)
Lag1 Change in Shareholder Wealth -00062 7856E-4 -0007 9003E-4 -00014 00002 -00014 00003
(594) (098) (592) (112) (522) (123) (5) (145)
Lag1 Institutional Concentration -2475087 555134 -2734983 943317
(3245) (710) (4965) (1671)
Lag1 Total Institutional Ownership 1020109 217488 936337 252464
(1904) (432) (2471) (704)
Lag1 Top 5 Institutional Ownership -148355 31354 -87926 423166
(139) (332) (118) (643)
Lag1 Tobinrsquos Q -195348 -252191 -192344 -25158 -119837 -146373 -122132 -143821
(2812) (3762) (2721) (3752) (248) (274) (2583) (2727)
Lag1 Market Capitalization 844E-06 103E-5 00034 00039
(2781) (3051) (2874) (3104)
Lag1 Log Market Capitalization 750021 723625 776657 732279
(7511) (9285) (11224) (14039)
CEO Dummy 2683978 2702238 2668244 2702417 3281133 3285103 3292175 3283145
(8625) (9277) (8246) (9271) (12739) (13616) (12336) (13605)
Adjusted R-Sq 04604 05302 04187 05295 03856 04588 03476 04574
Number of Observations 36280 36280 36280 36280 97679 97679 97679 97679
31
Table VII Level of Total Direct Compensation as a Function of Institutional Influence
Similar in parts to models (3) and (5) of HS (Table V) this table displays the coefficients from OLS regressions of total direct compensation against the lagged change and
contemporary change in shareholder wealth measures of institutional influence Tobinrsquos Q firm market capitalization and dummy controls for CEO industry and year effects
We estimate models (1) to (4) over the period 1992‒1997 and models (5) to (8) over the whole sample Models (1) (2) (5) and (6) use Institutional Concentration as a measure
of institutional influence which HS define as the shares owned by the institutions with the five largest holdings divided by the shares held by all institutions These models
control for total institutional ownership Models (3) (4) (7) and (8) use Top 5 Ownership as a measure of institutional influence which we define as the proportion of shares
outstanding held by the institutions with the five largest holdings Coefficients for industry and year controls are not presented One two and three asterisks denote significance
at 10 5 and 1 levels
Hartzell and Starks Database 1992‒1997 Full Sample from ExecuComp 1992‒2002
(1) (2) (3) (4) (5) (6) (7) (8)
Change in Shareholder Wealth 005779 007006 005282 006791 00423 00148 0046 00138
(415) (581) (369) (565) (203) (07) (22) (066)
Lag1 Change in Shareholder Wealth 003182 00714 00272 00715 00613 0126 00614 01278
(126) (298) (107) (299) (275) (488) (276) (494)
Lag1 Institutional Concentration -1608633 3248998 -3424892 3334186
(151) (283) (1172) (118)
Lag1 Total Institutional Ownership 104013 5473128 1531013 -2926764
(1645) (831) (876) (158)
Lag1 Top 5 Institutional Ownership 188585 3927986 -2159177 -2640879
(016) (343) (067) (092)
Lag1 Tobinrsquos Q -445692 -803612 -454859 -833566 3257027 2971201 3213936 3086205
(427) (791) (432) (831) (83) (77) (815) (793)
Lag1 Market Capitalization 49E-5 615E-5 01031 01088
(113) (1365) (1555) (168)
Lag1 Log Market Capitalization 4683497 4759377 1645156 1427856
(2775) (3469) (3679) (4239)
CEO Dummy 1501307 1508886 1493688 1508736 4713079 4713629 47278 4703908
(2648) (2684) (2618) (2683) (282) (2828) (2819) (2824)
Adjusted R-Sq 01861 02060 01695 02051 01071 01113 01033 011
Number of Observations 33229 33229 33229 33229 97679 97679 97679 97679
32
Table VIII Two Stage Least Squares (Reverse Causality) Tests - Executive Compensation Measures as a Function of Institutional Concentration
This table shows the coefficients from the Two-Stage Least Squares (2SLS) regressions of executive compensation measures against the lagged change
and contemporary change in shareholder wealth instrumented institutional concentration and total institutional ownership Tobinrsquos Q the natural
logarithm of firm market capitalization and dummy variable controls for CEO and year effects We instrument measures of institutional concentration
and total institutional ownership by orthogonalizing with respect to option grant pay-for-performance sensitivity salary and total compensation
Coefficients for industry and year controls are not presented First stage results are not presented One two and three asterisks denote significance at
10 5 and 1 levels
ExecuComp Sample 1992‒1997 Full Sample from ExecuComp 1992‒2002
PPSOG Salary TC PPSOG Salary TC
Change in Shareholder Wealth 144E-05 0004 02133 -409E-06 00011 00157
(308) (775) (596) (691) (638) (068)
Lag1 Change in Shareholder Wealth -240E-05 -00004 00866 118E-06 00002 01422
(131) (087) (346) (097) (116) (51)
Lag1 Institutional Concentration -00898 668681 3263143 02024 916656 2856417
(052) (801) (176) (163) (1545) (1004)
Lag1 Total Institutional Ownership 01021 -43997 2512587 02983 -104171 -1028347
(09) (084) (153) (4) (269) (487)
Lag1 Tobinrsquos Q 00754 -225606 119297 00655 -158657 2997205
(596) (1649) (053) (834) (2535) (727)
Lag1 Log Market Capitalization -04234 828713 7864046 -04002 825106 1685638
(2195) (8132) (2977) (4106) (11133) (373)
CEO Dummy 15848 3214117 2950274 19149 3205862 4529101
(2223) (9389) (2255) (3802) (1331) (2758)
Overall R-Sq -- 04981 01206 -- 04738 01178
Number of Observations 41305 41305 41305 85518 85518 85518
33
Table IX Two Stage Least Squares (Reverse Causality) Tests - Executive Compensation Measures as a Function of Top 5 Ownership
This table shows the coefficients from 2SLS regressions of executive compensation measures against the lagged change and contemporary change
in shareholder wealth instrumented Top 5 institutional ownership Tobinrsquos Q natural logarithm of firm market capitalization and dummy variable
controls for CEO and year effects We instrument Top 5 institutional ownership by orthogonalizing with respect to option grant pay-for-
performance sensitivity salary and total compensation Coefficients for industry and year controls are not presented First stage results are not
presented One two and three asterisks denote significance at 10 5 and 1 levels
ExecuComp sample 1992‒1997 Full Sample 1992‒2002
PPSOG Salary TC PPSOG Salary TC
Change in Shareholder Wealth 136E-05 00041 02128 -407E-06 00011 00146
(293) (79) (596) (692) (619) (063)
Lag1 Change in Shareholder Wealth -241E-05 -00003 00868 116E-06 00003 01439
(133) (07) (347) (096) (142) (514)
Lag1 Top 5 Institutional Ownership -03774 -187512 -3848531 00179 -104261 -126659
(163) (186) (166) (011) (149) (436)
Lag1 Tobinrsquos Q 00731 -223886 10074 00642 -155467 3115073
(573) (1631) (045) (819) (2509) (75)
Lag1 Log Market Capitalization -0416 775637 7696302 -04015 764632 1462536
(2345) (10084) (3766) (3688) (13708) (4072)
CEO Dummy 15854 3211991 2950158 19152 3203152 4519385
(2223) (9381) (2254) (3804) (133) (2752)
Overall R-Sq -- 04967 01202 -- 04722 01168
Number of Observations 41305 41305 41305 85518 85518 85518
34
Table X Specification Tests utilizing Box-Cox Transformation of Market Capitalization
This table presents results from a Box-Cox transformation of market capitalization In
each model we specify transformation of market capitalization only Panel A shows
results for the period 1992‒1997 and Panel B shows results for the entire period
1992‒2002 The first part of each panel shows the Lambda coefficient that applies to
the transformed market capitalization The second part of each panel shows log
likelihood scores with LR Chi2 Statistics in parentheses ‒ these test for the
appropriateness of a linear or log model specification
Panel A Our ExecuComp sample 1992‒1997
(1) (2) (3)
PPSOG Salary TC
Lambda -05143 01881 02876
(-1914) (2413) (1329)
Test H0 (Log Likelihood Score)
Lambda = -1 -1279748 -3178575 -4795326
(23866) (85952) (99688)
Lambda = 0 -1280152 -3138456 -4791159
(31934) (57147) (16339)
Lambda = 1 -1283157 -3163807 -4793506
(92039) (564156) (63292)
Panel B Full ExecuComp sample 1992‒2002
(1) (2) (4)
PPSOG Salary TC
Lambda -044 01853 04036
(-3015) (40) (3666)
Test H0 (Log Likelihood Score)
Lambda = 0 -2688334 -6476494 -10384490
(66599) (160356) (126236)
Lambda = 1 -2694886 -6538506 -10386795
(197644) (1400605) (172345)
35
Figure 2 Institutional Ownership Measures (LHA) vs Total
Compensation (RHA) by Market Capitalization
0
10
20
30
40
50
60
70
1 2 3 4 5 6 7 8 9 10
$00
$10
$20
$30
$40
$50
$60
$70
$80
$90
Millio
ns
InstitutionalConcentration (LHS)Top5 Ownership (LHS)
Institutional TotalOwnership (LHS)Total Compensation(RHS)
Page 29
29
Table V Pay-for-Performance Sensitivity of Cash Compensation (Salary plus Bonus) as a Function of Institutional Influence - Level Market Capitalization
Similar in parts to models (1) and (2) of HS (Table IV) this table displays the coefficients from OLS regressions of the change in cash compensation against the lagged
change and contemporary change in shareholder wealth measures of institutional influence Tobinrsquos Q firm market capitalization and dummy controls for CEO and year
effects We estimate models (1) to (4) over the period 1992‒1997 based on our ExecuComp sample and models (5) to (8) over the whole sample Models (1) (2) (5) and (6)
use Institutional Concentration as a measure of institutional influence which HS define as the shares owned by the institutions with the five largest holdings divided by the
shares held by all institutions These models control for total institutional ownership Models (4) and (8) use Top 5 Ownership as a measure of institutional influence which
we define as the proportion of shares outstanding held by the institutions with the 5 largest holdings Coefficients for industry and year controls are not presented One two
and three asterisks denote significance at 10 5 and 1 levels
Our Database based on ExecuComp 1992‒1997 Full Sample based on ExecuComp 1992‒2002
(1) (2) (3) (4) (5) (6) (7) (8)
Lag1 Change in Shareholder Wealth 00014 00017 00088 00089 -00006 -00007 00072 00086
(051) (062) (129) (103) (027) (029) (131) (16)
Change in Shareholder Wealth -00102 00102 00105 00032 00102 00101
(097) (292) (293) (055) (611) (608)
Change Shareholder Wealth Lag1 Institutional
Concentration
00435 00764 00363 00253
(463) (258) (689) (144)
Lag1 Change Shareholder Wealth Lag1
Institutional Concentration
-00197 -00283
(122) (197)
Lag1 Institutional Concentration -492182 -288861 400149 453623
(09) (056) (098) (114)
Lag1 Total Institutional Ownership 720688 538372 655091 559587 682069 726226
(276) (232) (281) (314) (308) (322)
Lag1 Change Shareholder Wealth Lag1 Top 5
Institutional Ownership
-0046 -00689
(091) (256)
Lag1 Top 5 Institutional Ownership 583044 1191634
(092) (25)
Lag1 Tobinrsquos Q -32761 -36806 -19422 -18023 -54283 -48681 -29534 -24243
(138) (172) (077) (078) (227) (245) (136) (111)
Lag1 Market Capitalization 00004 00004 00002 00006 00007 00008 00007 00007
(042) (044) (022) (07) (086) (092) (082) (09)
CEO Dummy 768035 765474 768334 773634 616315 61729 619028 618535
(354) (355) (356) (355) (417) (418) (419) (418)
Adjusted R-Sq 00057 0006 00052 0005 00097 00097 00098 00102
Number of Observations 41305 41305 41305 41305 85518 85518 85518 85518
30
Table VI Level of Salary as a Function of Institutional Influence
Similar in parts to models (1) and (2) of HS (Table V) this table displays the coefficients from OLS regressions of salary against the lagged change and contemporary change in
shareholder wealth measures of institutional influence Tobinrsquos Q firm market capitalization and dummy controls for CEO industry and year effects We estimate models (1) to
(4) over the period 1992‒1997 using the actual data supplied by HS and models (5) to (8) over the whole sample 1992‒2002 Models (1) (2) (5) and (6) use Institutional
Concentration as a measure of institutional influence which HS define as the shares owned by the institutions with the five largest holdings divided by the shares held by all
institutions These models control for total institutional ownership Models (3) (4) (7) and (8) use Top 5 Ownership as a measure of institutional influence which we define as
the proportion of shares outstanding held by the institutions with the 5 largest holdings Coefficients for industry and year controls are not presented One two and three
asterisks denote significance at 10 5 and 1 levels
Hartzell and Starks Database 1992‒1997 Full Sample based on ExecuComp 1992‒2002
(1) (2) (3) (4) (5) (6) (7) (8)
Change in Shareholder Wealth 04125E-4 0003 -2439E-4 0003 00021 00012 00024 00012
(057) (464) (031) (468) (1004) (72) (1041) (706)
Lag1 Change in Shareholder Wealth -00062 7856E-4 -0007 9003E-4 -00014 00002 -00014 00003
(594) (098) (592) (112) (522) (123) (5) (145)
Lag1 Institutional Concentration -2475087 555134 -2734983 943317
(3245) (710) (4965) (1671)
Lag1 Total Institutional Ownership 1020109 217488 936337 252464
(1904) (432) (2471) (704)
Lag1 Top 5 Institutional Ownership -148355 31354 -87926 423166
(139) (332) (118) (643)
Lag1 Tobinrsquos Q -195348 -252191 -192344 -25158 -119837 -146373 -122132 -143821
(2812) (3762) (2721) (3752) (248) (274) (2583) (2727)
Lag1 Market Capitalization 844E-06 103E-5 00034 00039
(2781) (3051) (2874) (3104)
Lag1 Log Market Capitalization 750021 723625 776657 732279
(7511) (9285) (11224) (14039)
CEO Dummy 2683978 2702238 2668244 2702417 3281133 3285103 3292175 3283145
(8625) (9277) (8246) (9271) (12739) (13616) (12336) (13605)
Adjusted R-Sq 04604 05302 04187 05295 03856 04588 03476 04574
Number of Observations 36280 36280 36280 36280 97679 97679 97679 97679
31
Table VII Level of Total Direct Compensation as a Function of Institutional Influence
Similar in parts to models (3) and (5) of HS (Table V) this table displays the coefficients from OLS regressions of total direct compensation against the lagged change and
contemporary change in shareholder wealth measures of institutional influence Tobinrsquos Q firm market capitalization and dummy controls for CEO industry and year effects
We estimate models (1) to (4) over the period 1992‒1997 and models (5) to (8) over the whole sample Models (1) (2) (5) and (6) use Institutional Concentration as a measure
of institutional influence which HS define as the shares owned by the institutions with the five largest holdings divided by the shares held by all institutions These models
control for total institutional ownership Models (3) (4) (7) and (8) use Top 5 Ownership as a measure of institutional influence which we define as the proportion of shares
outstanding held by the institutions with the five largest holdings Coefficients for industry and year controls are not presented One two and three asterisks denote significance
at 10 5 and 1 levels
Hartzell and Starks Database 1992‒1997 Full Sample from ExecuComp 1992‒2002
(1) (2) (3) (4) (5) (6) (7) (8)
Change in Shareholder Wealth 005779 007006 005282 006791 00423 00148 0046 00138
(415) (581) (369) (565) (203) (07) (22) (066)
Lag1 Change in Shareholder Wealth 003182 00714 00272 00715 00613 0126 00614 01278
(126) (298) (107) (299) (275) (488) (276) (494)
Lag1 Institutional Concentration -1608633 3248998 -3424892 3334186
(151) (283) (1172) (118)
Lag1 Total Institutional Ownership 104013 5473128 1531013 -2926764
(1645) (831) (876) (158)
Lag1 Top 5 Institutional Ownership 188585 3927986 -2159177 -2640879
(016) (343) (067) (092)
Lag1 Tobinrsquos Q -445692 -803612 -454859 -833566 3257027 2971201 3213936 3086205
(427) (791) (432) (831) (83) (77) (815) (793)
Lag1 Market Capitalization 49E-5 615E-5 01031 01088
(113) (1365) (1555) (168)
Lag1 Log Market Capitalization 4683497 4759377 1645156 1427856
(2775) (3469) (3679) (4239)
CEO Dummy 1501307 1508886 1493688 1508736 4713079 4713629 47278 4703908
(2648) (2684) (2618) (2683) (282) (2828) (2819) (2824)
Adjusted R-Sq 01861 02060 01695 02051 01071 01113 01033 011
Number of Observations 33229 33229 33229 33229 97679 97679 97679 97679
32
Table VIII Two Stage Least Squares (Reverse Causality) Tests - Executive Compensation Measures as a Function of Institutional Concentration
This table shows the coefficients from the Two-Stage Least Squares (2SLS) regressions of executive compensation measures against the lagged change
and contemporary change in shareholder wealth instrumented institutional concentration and total institutional ownership Tobinrsquos Q the natural
logarithm of firm market capitalization and dummy variable controls for CEO and year effects We instrument measures of institutional concentration
and total institutional ownership by orthogonalizing with respect to option grant pay-for-performance sensitivity salary and total compensation
Coefficients for industry and year controls are not presented First stage results are not presented One two and three asterisks denote significance at
10 5 and 1 levels
ExecuComp Sample 1992‒1997 Full Sample from ExecuComp 1992‒2002
PPSOG Salary TC PPSOG Salary TC
Change in Shareholder Wealth 144E-05 0004 02133 -409E-06 00011 00157
(308) (775) (596) (691) (638) (068)
Lag1 Change in Shareholder Wealth -240E-05 -00004 00866 118E-06 00002 01422
(131) (087) (346) (097) (116) (51)
Lag1 Institutional Concentration -00898 668681 3263143 02024 916656 2856417
(052) (801) (176) (163) (1545) (1004)
Lag1 Total Institutional Ownership 01021 -43997 2512587 02983 -104171 -1028347
(09) (084) (153) (4) (269) (487)
Lag1 Tobinrsquos Q 00754 -225606 119297 00655 -158657 2997205
(596) (1649) (053) (834) (2535) (727)
Lag1 Log Market Capitalization -04234 828713 7864046 -04002 825106 1685638
(2195) (8132) (2977) (4106) (11133) (373)
CEO Dummy 15848 3214117 2950274 19149 3205862 4529101
(2223) (9389) (2255) (3802) (1331) (2758)
Overall R-Sq -- 04981 01206 -- 04738 01178
Number of Observations 41305 41305 41305 85518 85518 85518
33
Table IX Two Stage Least Squares (Reverse Causality) Tests - Executive Compensation Measures as a Function of Top 5 Ownership
This table shows the coefficients from 2SLS regressions of executive compensation measures against the lagged change and contemporary change
in shareholder wealth instrumented Top 5 institutional ownership Tobinrsquos Q natural logarithm of firm market capitalization and dummy variable
controls for CEO and year effects We instrument Top 5 institutional ownership by orthogonalizing with respect to option grant pay-for-
performance sensitivity salary and total compensation Coefficients for industry and year controls are not presented First stage results are not
presented One two and three asterisks denote significance at 10 5 and 1 levels
ExecuComp sample 1992‒1997 Full Sample 1992‒2002
PPSOG Salary TC PPSOG Salary TC
Change in Shareholder Wealth 136E-05 00041 02128 -407E-06 00011 00146
(293) (79) (596) (692) (619) (063)
Lag1 Change in Shareholder Wealth -241E-05 -00003 00868 116E-06 00003 01439
(133) (07) (347) (096) (142) (514)
Lag1 Top 5 Institutional Ownership -03774 -187512 -3848531 00179 -104261 -126659
(163) (186) (166) (011) (149) (436)
Lag1 Tobinrsquos Q 00731 -223886 10074 00642 -155467 3115073
(573) (1631) (045) (819) (2509) (75)
Lag1 Log Market Capitalization -0416 775637 7696302 -04015 764632 1462536
(2345) (10084) (3766) (3688) (13708) (4072)
CEO Dummy 15854 3211991 2950158 19152 3203152 4519385
(2223) (9381) (2254) (3804) (133) (2752)
Overall R-Sq -- 04967 01202 -- 04722 01168
Number of Observations 41305 41305 41305 85518 85518 85518
34
Table X Specification Tests utilizing Box-Cox Transformation of Market Capitalization
This table presents results from a Box-Cox transformation of market capitalization In
each model we specify transformation of market capitalization only Panel A shows
results for the period 1992‒1997 and Panel B shows results for the entire period
1992‒2002 The first part of each panel shows the Lambda coefficient that applies to
the transformed market capitalization The second part of each panel shows log
likelihood scores with LR Chi2 Statistics in parentheses ‒ these test for the
appropriateness of a linear or log model specification
Panel A Our ExecuComp sample 1992‒1997
(1) (2) (3)
PPSOG Salary TC
Lambda -05143 01881 02876
(-1914) (2413) (1329)
Test H0 (Log Likelihood Score)
Lambda = -1 -1279748 -3178575 -4795326
(23866) (85952) (99688)
Lambda = 0 -1280152 -3138456 -4791159
(31934) (57147) (16339)
Lambda = 1 -1283157 -3163807 -4793506
(92039) (564156) (63292)
Panel B Full ExecuComp sample 1992‒2002
(1) (2) (4)
PPSOG Salary TC
Lambda -044 01853 04036
(-3015) (40) (3666)
Test H0 (Log Likelihood Score)
Lambda = 0 -2688334 -6476494 -10384490
(66599) (160356) (126236)
Lambda = 1 -2694886 -6538506 -10386795
(197644) (1400605) (172345)
35
Figure 2 Institutional Ownership Measures (LHA) vs Total
Compensation (RHA) by Market Capitalization
0
10
20
30
40
50
60
70
1 2 3 4 5 6 7 8 9 10
$00
$10
$20
$30
$40
$50
$60
$70
$80
$90
Millio
ns
InstitutionalConcentration (LHS)Top5 Ownership (LHS)
Institutional TotalOwnership (LHS)Total Compensation(RHS)
Page 30
30
Table VI Level of Salary as a Function of Institutional Influence
Similar in parts to models (1) and (2) of HS (Table V) this table displays the coefficients from OLS regressions of salary against the lagged change and contemporary change in
shareholder wealth measures of institutional influence Tobinrsquos Q firm market capitalization and dummy controls for CEO industry and year effects We estimate models (1) to
(4) over the period 1992‒1997 using the actual data supplied by HS and models (5) to (8) over the whole sample 1992‒2002 Models (1) (2) (5) and (6) use Institutional
Concentration as a measure of institutional influence which HS define as the shares owned by the institutions with the five largest holdings divided by the shares held by all
institutions These models control for total institutional ownership Models (3) (4) (7) and (8) use Top 5 Ownership as a measure of institutional influence which we define as
the proportion of shares outstanding held by the institutions with the 5 largest holdings Coefficients for industry and year controls are not presented One two and three
asterisks denote significance at 10 5 and 1 levels
Hartzell and Starks Database 1992‒1997 Full Sample based on ExecuComp 1992‒2002
(1) (2) (3) (4) (5) (6) (7) (8)
Change in Shareholder Wealth 04125E-4 0003 -2439E-4 0003 00021 00012 00024 00012
(057) (464) (031) (468) (1004) (72) (1041) (706)
Lag1 Change in Shareholder Wealth -00062 7856E-4 -0007 9003E-4 -00014 00002 -00014 00003
(594) (098) (592) (112) (522) (123) (5) (145)
Lag1 Institutional Concentration -2475087 555134 -2734983 943317
(3245) (710) (4965) (1671)
Lag1 Total Institutional Ownership 1020109 217488 936337 252464
(1904) (432) (2471) (704)
Lag1 Top 5 Institutional Ownership -148355 31354 -87926 423166
(139) (332) (118) (643)
Lag1 Tobinrsquos Q -195348 -252191 -192344 -25158 -119837 -146373 -122132 -143821
(2812) (3762) (2721) (3752) (248) (274) (2583) (2727)
Lag1 Market Capitalization 844E-06 103E-5 00034 00039
(2781) (3051) (2874) (3104)
Lag1 Log Market Capitalization 750021 723625 776657 732279
(7511) (9285) (11224) (14039)
CEO Dummy 2683978 2702238 2668244 2702417 3281133 3285103 3292175 3283145
(8625) (9277) (8246) (9271) (12739) (13616) (12336) (13605)
Adjusted R-Sq 04604 05302 04187 05295 03856 04588 03476 04574
Number of Observations 36280 36280 36280 36280 97679 97679 97679 97679
31
Table VII Level of Total Direct Compensation as a Function of Institutional Influence
Similar in parts to models (3) and (5) of HS (Table V) this table displays the coefficients from OLS regressions of total direct compensation against the lagged change and
contemporary change in shareholder wealth measures of institutional influence Tobinrsquos Q firm market capitalization and dummy controls for CEO industry and year effects
We estimate models (1) to (4) over the period 1992‒1997 and models (5) to (8) over the whole sample Models (1) (2) (5) and (6) use Institutional Concentration as a measure
of institutional influence which HS define as the shares owned by the institutions with the five largest holdings divided by the shares held by all institutions These models
control for total institutional ownership Models (3) (4) (7) and (8) use Top 5 Ownership as a measure of institutional influence which we define as the proportion of shares
outstanding held by the institutions with the five largest holdings Coefficients for industry and year controls are not presented One two and three asterisks denote significance
at 10 5 and 1 levels
Hartzell and Starks Database 1992‒1997 Full Sample from ExecuComp 1992‒2002
(1) (2) (3) (4) (5) (6) (7) (8)
Change in Shareholder Wealth 005779 007006 005282 006791 00423 00148 0046 00138
(415) (581) (369) (565) (203) (07) (22) (066)
Lag1 Change in Shareholder Wealth 003182 00714 00272 00715 00613 0126 00614 01278
(126) (298) (107) (299) (275) (488) (276) (494)
Lag1 Institutional Concentration -1608633 3248998 -3424892 3334186
(151) (283) (1172) (118)
Lag1 Total Institutional Ownership 104013 5473128 1531013 -2926764
(1645) (831) (876) (158)
Lag1 Top 5 Institutional Ownership 188585 3927986 -2159177 -2640879
(016) (343) (067) (092)
Lag1 Tobinrsquos Q -445692 -803612 -454859 -833566 3257027 2971201 3213936 3086205
(427) (791) (432) (831) (83) (77) (815) (793)
Lag1 Market Capitalization 49E-5 615E-5 01031 01088
(113) (1365) (1555) (168)
Lag1 Log Market Capitalization 4683497 4759377 1645156 1427856
(2775) (3469) (3679) (4239)
CEO Dummy 1501307 1508886 1493688 1508736 4713079 4713629 47278 4703908
(2648) (2684) (2618) (2683) (282) (2828) (2819) (2824)
Adjusted R-Sq 01861 02060 01695 02051 01071 01113 01033 011
Number of Observations 33229 33229 33229 33229 97679 97679 97679 97679
32
Table VIII Two Stage Least Squares (Reverse Causality) Tests - Executive Compensation Measures as a Function of Institutional Concentration
This table shows the coefficients from the Two-Stage Least Squares (2SLS) regressions of executive compensation measures against the lagged change
and contemporary change in shareholder wealth instrumented institutional concentration and total institutional ownership Tobinrsquos Q the natural
logarithm of firm market capitalization and dummy variable controls for CEO and year effects We instrument measures of institutional concentration
and total institutional ownership by orthogonalizing with respect to option grant pay-for-performance sensitivity salary and total compensation
Coefficients for industry and year controls are not presented First stage results are not presented One two and three asterisks denote significance at
10 5 and 1 levels
ExecuComp Sample 1992‒1997 Full Sample from ExecuComp 1992‒2002
PPSOG Salary TC PPSOG Salary TC
Change in Shareholder Wealth 144E-05 0004 02133 -409E-06 00011 00157
(308) (775) (596) (691) (638) (068)
Lag1 Change in Shareholder Wealth -240E-05 -00004 00866 118E-06 00002 01422
(131) (087) (346) (097) (116) (51)
Lag1 Institutional Concentration -00898 668681 3263143 02024 916656 2856417
(052) (801) (176) (163) (1545) (1004)
Lag1 Total Institutional Ownership 01021 -43997 2512587 02983 -104171 -1028347
(09) (084) (153) (4) (269) (487)
Lag1 Tobinrsquos Q 00754 -225606 119297 00655 -158657 2997205
(596) (1649) (053) (834) (2535) (727)
Lag1 Log Market Capitalization -04234 828713 7864046 -04002 825106 1685638
(2195) (8132) (2977) (4106) (11133) (373)
CEO Dummy 15848 3214117 2950274 19149 3205862 4529101
(2223) (9389) (2255) (3802) (1331) (2758)
Overall R-Sq -- 04981 01206 -- 04738 01178
Number of Observations 41305 41305 41305 85518 85518 85518
33
Table IX Two Stage Least Squares (Reverse Causality) Tests - Executive Compensation Measures as a Function of Top 5 Ownership
This table shows the coefficients from 2SLS regressions of executive compensation measures against the lagged change and contemporary change
in shareholder wealth instrumented Top 5 institutional ownership Tobinrsquos Q natural logarithm of firm market capitalization and dummy variable
controls for CEO and year effects We instrument Top 5 institutional ownership by orthogonalizing with respect to option grant pay-for-
performance sensitivity salary and total compensation Coefficients for industry and year controls are not presented First stage results are not
presented One two and three asterisks denote significance at 10 5 and 1 levels
ExecuComp sample 1992‒1997 Full Sample 1992‒2002
PPSOG Salary TC PPSOG Salary TC
Change in Shareholder Wealth 136E-05 00041 02128 -407E-06 00011 00146
(293) (79) (596) (692) (619) (063)
Lag1 Change in Shareholder Wealth -241E-05 -00003 00868 116E-06 00003 01439
(133) (07) (347) (096) (142) (514)
Lag1 Top 5 Institutional Ownership -03774 -187512 -3848531 00179 -104261 -126659
(163) (186) (166) (011) (149) (436)
Lag1 Tobinrsquos Q 00731 -223886 10074 00642 -155467 3115073
(573) (1631) (045) (819) (2509) (75)
Lag1 Log Market Capitalization -0416 775637 7696302 -04015 764632 1462536
(2345) (10084) (3766) (3688) (13708) (4072)
CEO Dummy 15854 3211991 2950158 19152 3203152 4519385
(2223) (9381) (2254) (3804) (133) (2752)
Overall R-Sq -- 04967 01202 -- 04722 01168
Number of Observations 41305 41305 41305 85518 85518 85518
34
Table X Specification Tests utilizing Box-Cox Transformation of Market Capitalization
This table presents results from a Box-Cox transformation of market capitalization In
each model we specify transformation of market capitalization only Panel A shows
results for the period 1992‒1997 and Panel B shows results for the entire period
1992‒2002 The first part of each panel shows the Lambda coefficient that applies to
the transformed market capitalization The second part of each panel shows log
likelihood scores with LR Chi2 Statistics in parentheses ‒ these test for the
appropriateness of a linear or log model specification
Panel A Our ExecuComp sample 1992‒1997
(1) (2) (3)
PPSOG Salary TC
Lambda -05143 01881 02876
(-1914) (2413) (1329)
Test H0 (Log Likelihood Score)
Lambda = -1 -1279748 -3178575 -4795326
(23866) (85952) (99688)
Lambda = 0 -1280152 -3138456 -4791159
(31934) (57147) (16339)
Lambda = 1 -1283157 -3163807 -4793506
(92039) (564156) (63292)
Panel B Full ExecuComp sample 1992‒2002
(1) (2) (4)
PPSOG Salary TC
Lambda -044 01853 04036
(-3015) (40) (3666)
Test H0 (Log Likelihood Score)
Lambda = 0 -2688334 -6476494 -10384490
(66599) (160356) (126236)
Lambda = 1 -2694886 -6538506 -10386795
(197644) (1400605) (172345)
35
Figure 2 Institutional Ownership Measures (LHA) vs Total
Compensation (RHA) by Market Capitalization
0
10
20
30
40
50
60
70
1 2 3 4 5 6 7 8 9 10
$00
$10
$20
$30
$40
$50
$60
$70
$80
$90
Millio
ns
InstitutionalConcentration (LHS)Top5 Ownership (LHS)
Institutional TotalOwnership (LHS)Total Compensation(RHS)
Page 31
31
Table VII Level of Total Direct Compensation as a Function of Institutional Influence
Similar in parts to models (3) and (5) of HS (Table V) this table displays the coefficients from OLS regressions of total direct compensation against the lagged change and
contemporary change in shareholder wealth measures of institutional influence Tobinrsquos Q firm market capitalization and dummy controls for CEO industry and year effects
We estimate models (1) to (4) over the period 1992‒1997 and models (5) to (8) over the whole sample Models (1) (2) (5) and (6) use Institutional Concentration as a measure
of institutional influence which HS define as the shares owned by the institutions with the five largest holdings divided by the shares held by all institutions These models
control for total institutional ownership Models (3) (4) (7) and (8) use Top 5 Ownership as a measure of institutional influence which we define as the proportion of shares
outstanding held by the institutions with the five largest holdings Coefficients for industry and year controls are not presented One two and three asterisks denote significance
at 10 5 and 1 levels
Hartzell and Starks Database 1992‒1997 Full Sample from ExecuComp 1992‒2002
(1) (2) (3) (4) (5) (6) (7) (8)
Change in Shareholder Wealth 005779 007006 005282 006791 00423 00148 0046 00138
(415) (581) (369) (565) (203) (07) (22) (066)
Lag1 Change in Shareholder Wealth 003182 00714 00272 00715 00613 0126 00614 01278
(126) (298) (107) (299) (275) (488) (276) (494)
Lag1 Institutional Concentration -1608633 3248998 -3424892 3334186
(151) (283) (1172) (118)
Lag1 Total Institutional Ownership 104013 5473128 1531013 -2926764
(1645) (831) (876) (158)
Lag1 Top 5 Institutional Ownership 188585 3927986 -2159177 -2640879
(016) (343) (067) (092)
Lag1 Tobinrsquos Q -445692 -803612 -454859 -833566 3257027 2971201 3213936 3086205
(427) (791) (432) (831) (83) (77) (815) (793)
Lag1 Market Capitalization 49E-5 615E-5 01031 01088
(113) (1365) (1555) (168)
Lag1 Log Market Capitalization 4683497 4759377 1645156 1427856
(2775) (3469) (3679) (4239)
CEO Dummy 1501307 1508886 1493688 1508736 4713079 4713629 47278 4703908
(2648) (2684) (2618) (2683) (282) (2828) (2819) (2824)
Adjusted R-Sq 01861 02060 01695 02051 01071 01113 01033 011
Number of Observations 33229 33229 33229 33229 97679 97679 97679 97679
32
Table VIII Two Stage Least Squares (Reverse Causality) Tests - Executive Compensation Measures as a Function of Institutional Concentration
This table shows the coefficients from the Two-Stage Least Squares (2SLS) regressions of executive compensation measures against the lagged change
and contemporary change in shareholder wealth instrumented institutional concentration and total institutional ownership Tobinrsquos Q the natural
logarithm of firm market capitalization and dummy variable controls for CEO and year effects We instrument measures of institutional concentration
and total institutional ownership by orthogonalizing with respect to option grant pay-for-performance sensitivity salary and total compensation
Coefficients for industry and year controls are not presented First stage results are not presented One two and three asterisks denote significance at
10 5 and 1 levels
ExecuComp Sample 1992‒1997 Full Sample from ExecuComp 1992‒2002
PPSOG Salary TC PPSOG Salary TC
Change in Shareholder Wealth 144E-05 0004 02133 -409E-06 00011 00157
(308) (775) (596) (691) (638) (068)
Lag1 Change in Shareholder Wealth -240E-05 -00004 00866 118E-06 00002 01422
(131) (087) (346) (097) (116) (51)
Lag1 Institutional Concentration -00898 668681 3263143 02024 916656 2856417
(052) (801) (176) (163) (1545) (1004)
Lag1 Total Institutional Ownership 01021 -43997 2512587 02983 -104171 -1028347
(09) (084) (153) (4) (269) (487)
Lag1 Tobinrsquos Q 00754 -225606 119297 00655 -158657 2997205
(596) (1649) (053) (834) (2535) (727)
Lag1 Log Market Capitalization -04234 828713 7864046 -04002 825106 1685638
(2195) (8132) (2977) (4106) (11133) (373)
CEO Dummy 15848 3214117 2950274 19149 3205862 4529101
(2223) (9389) (2255) (3802) (1331) (2758)
Overall R-Sq -- 04981 01206 -- 04738 01178
Number of Observations 41305 41305 41305 85518 85518 85518
33
Table IX Two Stage Least Squares (Reverse Causality) Tests - Executive Compensation Measures as a Function of Top 5 Ownership
This table shows the coefficients from 2SLS regressions of executive compensation measures against the lagged change and contemporary change
in shareholder wealth instrumented Top 5 institutional ownership Tobinrsquos Q natural logarithm of firm market capitalization and dummy variable
controls for CEO and year effects We instrument Top 5 institutional ownership by orthogonalizing with respect to option grant pay-for-
performance sensitivity salary and total compensation Coefficients for industry and year controls are not presented First stage results are not
presented One two and three asterisks denote significance at 10 5 and 1 levels
ExecuComp sample 1992‒1997 Full Sample 1992‒2002
PPSOG Salary TC PPSOG Salary TC
Change in Shareholder Wealth 136E-05 00041 02128 -407E-06 00011 00146
(293) (79) (596) (692) (619) (063)
Lag1 Change in Shareholder Wealth -241E-05 -00003 00868 116E-06 00003 01439
(133) (07) (347) (096) (142) (514)
Lag1 Top 5 Institutional Ownership -03774 -187512 -3848531 00179 -104261 -126659
(163) (186) (166) (011) (149) (436)
Lag1 Tobinrsquos Q 00731 -223886 10074 00642 -155467 3115073
(573) (1631) (045) (819) (2509) (75)
Lag1 Log Market Capitalization -0416 775637 7696302 -04015 764632 1462536
(2345) (10084) (3766) (3688) (13708) (4072)
CEO Dummy 15854 3211991 2950158 19152 3203152 4519385
(2223) (9381) (2254) (3804) (133) (2752)
Overall R-Sq -- 04967 01202 -- 04722 01168
Number of Observations 41305 41305 41305 85518 85518 85518
34
Table X Specification Tests utilizing Box-Cox Transformation of Market Capitalization
This table presents results from a Box-Cox transformation of market capitalization In
each model we specify transformation of market capitalization only Panel A shows
results for the period 1992‒1997 and Panel B shows results for the entire period
1992‒2002 The first part of each panel shows the Lambda coefficient that applies to
the transformed market capitalization The second part of each panel shows log
likelihood scores with LR Chi2 Statistics in parentheses ‒ these test for the
appropriateness of a linear or log model specification
Panel A Our ExecuComp sample 1992‒1997
(1) (2) (3)
PPSOG Salary TC
Lambda -05143 01881 02876
(-1914) (2413) (1329)
Test H0 (Log Likelihood Score)
Lambda = -1 -1279748 -3178575 -4795326
(23866) (85952) (99688)
Lambda = 0 -1280152 -3138456 -4791159
(31934) (57147) (16339)
Lambda = 1 -1283157 -3163807 -4793506
(92039) (564156) (63292)
Panel B Full ExecuComp sample 1992‒2002
(1) (2) (4)
PPSOG Salary TC
Lambda -044 01853 04036
(-3015) (40) (3666)
Test H0 (Log Likelihood Score)
Lambda = 0 -2688334 -6476494 -10384490
(66599) (160356) (126236)
Lambda = 1 -2694886 -6538506 -10386795
(197644) (1400605) (172345)
35
Figure 2 Institutional Ownership Measures (LHA) vs Total
Compensation (RHA) by Market Capitalization
0
10
20
30
40
50
60
70
1 2 3 4 5 6 7 8 9 10
$00
$10
$20
$30
$40
$50
$60
$70
$80
$90
Millio
ns
InstitutionalConcentration (LHS)Top5 Ownership (LHS)
Institutional TotalOwnership (LHS)Total Compensation(RHS)
Page 32
32
Table VIII Two Stage Least Squares (Reverse Causality) Tests - Executive Compensation Measures as a Function of Institutional Concentration
This table shows the coefficients from the Two-Stage Least Squares (2SLS) regressions of executive compensation measures against the lagged change
and contemporary change in shareholder wealth instrumented institutional concentration and total institutional ownership Tobinrsquos Q the natural
logarithm of firm market capitalization and dummy variable controls for CEO and year effects We instrument measures of institutional concentration
and total institutional ownership by orthogonalizing with respect to option grant pay-for-performance sensitivity salary and total compensation
Coefficients for industry and year controls are not presented First stage results are not presented One two and three asterisks denote significance at
10 5 and 1 levels
ExecuComp Sample 1992‒1997 Full Sample from ExecuComp 1992‒2002
PPSOG Salary TC PPSOG Salary TC
Change in Shareholder Wealth 144E-05 0004 02133 -409E-06 00011 00157
(308) (775) (596) (691) (638) (068)
Lag1 Change in Shareholder Wealth -240E-05 -00004 00866 118E-06 00002 01422
(131) (087) (346) (097) (116) (51)
Lag1 Institutional Concentration -00898 668681 3263143 02024 916656 2856417
(052) (801) (176) (163) (1545) (1004)
Lag1 Total Institutional Ownership 01021 -43997 2512587 02983 -104171 -1028347
(09) (084) (153) (4) (269) (487)
Lag1 Tobinrsquos Q 00754 -225606 119297 00655 -158657 2997205
(596) (1649) (053) (834) (2535) (727)
Lag1 Log Market Capitalization -04234 828713 7864046 -04002 825106 1685638
(2195) (8132) (2977) (4106) (11133) (373)
CEO Dummy 15848 3214117 2950274 19149 3205862 4529101
(2223) (9389) (2255) (3802) (1331) (2758)
Overall R-Sq -- 04981 01206 -- 04738 01178
Number of Observations 41305 41305 41305 85518 85518 85518
33
Table IX Two Stage Least Squares (Reverse Causality) Tests - Executive Compensation Measures as a Function of Top 5 Ownership
This table shows the coefficients from 2SLS regressions of executive compensation measures against the lagged change and contemporary change
in shareholder wealth instrumented Top 5 institutional ownership Tobinrsquos Q natural logarithm of firm market capitalization and dummy variable
controls for CEO and year effects We instrument Top 5 institutional ownership by orthogonalizing with respect to option grant pay-for-
performance sensitivity salary and total compensation Coefficients for industry and year controls are not presented First stage results are not
presented One two and three asterisks denote significance at 10 5 and 1 levels
ExecuComp sample 1992‒1997 Full Sample 1992‒2002
PPSOG Salary TC PPSOG Salary TC
Change in Shareholder Wealth 136E-05 00041 02128 -407E-06 00011 00146
(293) (79) (596) (692) (619) (063)
Lag1 Change in Shareholder Wealth -241E-05 -00003 00868 116E-06 00003 01439
(133) (07) (347) (096) (142) (514)
Lag1 Top 5 Institutional Ownership -03774 -187512 -3848531 00179 -104261 -126659
(163) (186) (166) (011) (149) (436)
Lag1 Tobinrsquos Q 00731 -223886 10074 00642 -155467 3115073
(573) (1631) (045) (819) (2509) (75)
Lag1 Log Market Capitalization -0416 775637 7696302 -04015 764632 1462536
(2345) (10084) (3766) (3688) (13708) (4072)
CEO Dummy 15854 3211991 2950158 19152 3203152 4519385
(2223) (9381) (2254) (3804) (133) (2752)
Overall R-Sq -- 04967 01202 -- 04722 01168
Number of Observations 41305 41305 41305 85518 85518 85518
34
Table X Specification Tests utilizing Box-Cox Transformation of Market Capitalization
This table presents results from a Box-Cox transformation of market capitalization In
each model we specify transformation of market capitalization only Panel A shows
results for the period 1992‒1997 and Panel B shows results for the entire period
1992‒2002 The first part of each panel shows the Lambda coefficient that applies to
the transformed market capitalization The second part of each panel shows log
likelihood scores with LR Chi2 Statistics in parentheses ‒ these test for the
appropriateness of a linear or log model specification
Panel A Our ExecuComp sample 1992‒1997
(1) (2) (3)
PPSOG Salary TC
Lambda -05143 01881 02876
(-1914) (2413) (1329)
Test H0 (Log Likelihood Score)
Lambda = -1 -1279748 -3178575 -4795326
(23866) (85952) (99688)
Lambda = 0 -1280152 -3138456 -4791159
(31934) (57147) (16339)
Lambda = 1 -1283157 -3163807 -4793506
(92039) (564156) (63292)
Panel B Full ExecuComp sample 1992‒2002
(1) (2) (4)
PPSOG Salary TC
Lambda -044 01853 04036
(-3015) (40) (3666)
Test H0 (Log Likelihood Score)
Lambda = 0 -2688334 -6476494 -10384490
(66599) (160356) (126236)
Lambda = 1 -2694886 -6538506 -10386795
(197644) (1400605) (172345)
35
Figure 2 Institutional Ownership Measures (LHA) vs Total
Compensation (RHA) by Market Capitalization
0
10
20
30
40
50
60
70
1 2 3 4 5 6 7 8 9 10
$00
$10
$20
$30
$40
$50
$60
$70
$80
$90
Millio
ns
InstitutionalConcentration (LHS)Top5 Ownership (LHS)
Institutional TotalOwnership (LHS)Total Compensation(RHS)
Page 33
33
Table IX Two Stage Least Squares (Reverse Causality) Tests - Executive Compensation Measures as a Function of Top 5 Ownership
This table shows the coefficients from 2SLS regressions of executive compensation measures against the lagged change and contemporary change
in shareholder wealth instrumented Top 5 institutional ownership Tobinrsquos Q natural logarithm of firm market capitalization and dummy variable
controls for CEO and year effects We instrument Top 5 institutional ownership by orthogonalizing with respect to option grant pay-for-
performance sensitivity salary and total compensation Coefficients for industry and year controls are not presented First stage results are not
presented One two and three asterisks denote significance at 10 5 and 1 levels
ExecuComp sample 1992‒1997 Full Sample 1992‒2002
PPSOG Salary TC PPSOG Salary TC
Change in Shareholder Wealth 136E-05 00041 02128 -407E-06 00011 00146
(293) (79) (596) (692) (619) (063)
Lag1 Change in Shareholder Wealth -241E-05 -00003 00868 116E-06 00003 01439
(133) (07) (347) (096) (142) (514)
Lag1 Top 5 Institutional Ownership -03774 -187512 -3848531 00179 -104261 -126659
(163) (186) (166) (011) (149) (436)
Lag1 Tobinrsquos Q 00731 -223886 10074 00642 -155467 3115073
(573) (1631) (045) (819) (2509) (75)
Lag1 Log Market Capitalization -0416 775637 7696302 -04015 764632 1462536
(2345) (10084) (3766) (3688) (13708) (4072)
CEO Dummy 15854 3211991 2950158 19152 3203152 4519385
(2223) (9381) (2254) (3804) (133) (2752)
Overall R-Sq -- 04967 01202 -- 04722 01168
Number of Observations 41305 41305 41305 85518 85518 85518
34
Table X Specification Tests utilizing Box-Cox Transformation of Market Capitalization
This table presents results from a Box-Cox transformation of market capitalization In
each model we specify transformation of market capitalization only Panel A shows
results for the period 1992‒1997 and Panel B shows results for the entire period
1992‒2002 The first part of each panel shows the Lambda coefficient that applies to
the transformed market capitalization The second part of each panel shows log
likelihood scores with LR Chi2 Statistics in parentheses ‒ these test for the
appropriateness of a linear or log model specification
Panel A Our ExecuComp sample 1992‒1997
(1) (2) (3)
PPSOG Salary TC
Lambda -05143 01881 02876
(-1914) (2413) (1329)
Test H0 (Log Likelihood Score)
Lambda = -1 -1279748 -3178575 -4795326
(23866) (85952) (99688)
Lambda = 0 -1280152 -3138456 -4791159
(31934) (57147) (16339)
Lambda = 1 -1283157 -3163807 -4793506
(92039) (564156) (63292)
Panel B Full ExecuComp sample 1992‒2002
(1) (2) (4)
PPSOG Salary TC
Lambda -044 01853 04036
(-3015) (40) (3666)
Test H0 (Log Likelihood Score)
Lambda = 0 -2688334 -6476494 -10384490
(66599) (160356) (126236)
Lambda = 1 -2694886 -6538506 -10386795
(197644) (1400605) (172345)
35
Figure 2 Institutional Ownership Measures (LHA) vs Total
Compensation (RHA) by Market Capitalization
0
10
20
30
40
50
60
70
1 2 3 4 5 6 7 8 9 10
$00
$10
$20
$30
$40
$50
$60
$70
$80
$90
Millio
ns
InstitutionalConcentration (LHS)Top5 Ownership (LHS)
Institutional TotalOwnership (LHS)Total Compensation(RHS)
Page 34
34
Table X Specification Tests utilizing Box-Cox Transformation of Market Capitalization
This table presents results from a Box-Cox transformation of market capitalization In
each model we specify transformation of market capitalization only Panel A shows
results for the period 1992‒1997 and Panel B shows results for the entire period
1992‒2002 The first part of each panel shows the Lambda coefficient that applies to
the transformed market capitalization The second part of each panel shows log
likelihood scores with LR Chi2 Statistics in parentheses ‒ these test for the
appropriateness of a linear or log model specification
Panel A Our ExecuComp sample 1992‒1997
(1) (2) (3)
PPSOG Salary TC
Lambda -05143 01881 02876
(-1914) (2413) (1329)
Test H0 (Log Likelihood Score)
Lambda = -1 -1279748 -3178575 -4795326
(23866) (85952) (99688)
Lambda = 0 -1280152 -3138456 -4791159
(31934) (57147) (16339)
Lambda = 1 -1283157 -3163807 -4793506
(92039) (564156) (63292)
Panel B Full ExecuComp sample 1992‒2002
(1) (2) (4)
PPSOG Salary TC
Lambda -044 01853 04036
(-3015) (40) (3666)
Test H0 (Log Likelihood Score)
Lambda = 0 -2688334 -6476494 -10384490
(66599) (160356) (126236)
Lambda = 1 -2694886 -6538506 -10386795
(197644) (1400605) (172345)
35
Figure 2 Institutional Ownership Measures (LHA) vs Total
Compensation (RHA) by Market Capitalization
0
10
20
30
40
50
60
70
1 2 3 4 5 6 7 8 9 10
$00
$10
$20
$30
$40
$50
$60
$70
$80
$90
Millio
ns
InstitutionalConcentration (LHS)Top5 Ownership (LHS)
Institutional TotalOwnership (LHS)Total Compensation(RHS)
Page 35
35
Figure 2 Institutional Ownership Measures (LHA) vs Total
Compensation (RHA) by Market Capitalization
0
10
20
30
40
50
60
70
1 2 3 4 5 6 7 8 9 10
$00
$10
$20
$30
$40
$50
$60
$70
$80
$90
Millio
ns
InstitutionalConcentration (LHS)Top5 Ownership (LHS)
Institutional TotalOwnership (LHS)Total Compensation(RHS)