Firm Characteristics and Long-run Stock Returns after Corporate Events * Hendrik Bessembinder David Eccles School of Business University of Utah 1655 E. Campus Center Drive Salt Lake City, UT 84112 [email protected]Tel: 801-581-8268 Feng Zhang David Eccles School of Business University of Utah 1655 E. Campus Center Drive Salt Lake City, UT 84112 [email protected]Tel: 801-587-9476 This Version: May 2012 Abstract The frequently-documented negative abnormal long run buy-and-hold returns to bidding firms, SEO firms, and IPO firms can be attributed to imperfect control-firm matching. Control firms are most often selected on the basis of firm size and market-to-book ratios. However, event firms differ from control firms in terms of idiosyncratic volatility, illiquidity, beta, and return momentum, each of which is also known to be related to returns. We propose a simple regression-based approach to control for differences between event and control firms in characteristics other than those used to match, and show that long run abnormal returns do not differ significantly from zero for bidding firms, SEO firms or IPO firms in the 1980 to 2005 period. Our method also reconciles results of studies relying on time series factor model regressions with those focusing on abnormal buy-and-hold returns. Keywords: firm characteristics, long-run stock returns, BHARs, calendar-time portfolio, wealth relative, event study JEL classification: G34
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
Firm Characteristics and Long-run Stock Returns after Corporate Events*
Firm Characteristics and Long-Run Stock Returns after Corporate Events
Two approaches have commonly been employed to measure long-run abnormal stock
returns after corporate events. The first approach is based on “buy-and-hold abnormal returns
(BHARs)”, which are assessed as the difference between long run buy-and-hold returns to event
firms compared to control firms that are matched to event firms based on firm characteristics
such as size and book-to-market ratio (B/M). The second, “calendar time portfolio,” approach
focuses on mean time series returns to a portfolio of event firms, after adjusting for portfolio
exposure to various risk factors.
The two approaches produce contrasting results in many studies. For example, Betton,
Eckbo, and Thorburn (2008) study bidding firms in mergers and acquisitions (M&A), and report
statistically significant five-year BHARs of -21.9% during the 1980 to2003 period, but also
report an economically small (0.08% per month) and statistically insignificant “alpha” (intercept
in the regression of calendar portfolio returns on risk factors) estimate for the same firms. Eckbo,
Masulis, and Norli (2008) study industrial firms that issued secondary (SEO) and initial (IPO)
offerings of common stocks during the 1980 to 2000 period and report significant five-year
BHARs of -29.7% and -18.0%, respectively. In contrast, the estimated alphas are for -0.18%
and -0.16%, respectively, and neither is statistically significant. Table 1 provides a summary of
results reported in several additional studies of returns to bidders and to firms issuing common
stock.1
There are at least three possible reasons for the discrepancy in results across the two
methods. First, the risk factors used to adjust calendar time portfolio returns may be imperfect,
1 See Fama (1998), Loughran and Ritter (2000), Betton, Eckbo, and Thorburn (2008), and Eckbo, Masulis, and Norli
(2008) for additional discussion of the contrasting results.
2
which is the “bad-model problem” identified by Fama (1998). Second, returns to equal-
weighted portfolios and coefficient estimated in OLS return regressions are biased due to noise
in security prices, while buy-and-hold returns are largely free of such bias.2 Third, while
matching firms are typically selected on the basis of firm characteristics known to be related to
average stock returns, it is typically not practical to match firms based on more than two or three
characteristics. The literature has identified numerous characteristics capable of explaining
variation in average stock returns, leaving open the possibility that the matched firms
systematically differ in non-matched characteristics that also affect returns. Further, firm
characteristics may change after corporate events, implying that pairs of firms that are well-
matched in terms of selected characteristics at a point in time may not remain so.
In this study, we assess whether the conflicting results obtained in studies of long run
abnormal returns obtained across the matched-firm versus time-series portfolio methods can be
explained on the basis of imperfect matching. We study long-run returns to bidding firms in
mergers and acquisitions (M&As), to firms that issue seasoned equity (SEOs), and to firms
making initial public offerings (IPOs) of common stock. The matched firm approach is most
often implemented on the basis of firm size and book-to-market ratio. However, we show that
event firms differ significantly from their size- and B/M-matched counterparts in terms of
idiosyncratic volatility, illiquidity, market beta, and return momentum.3 In particular, bidding
firms, SEO firms and IPO firms all typically have significantly greater idiosyncratic volatility
2 Lyon, Barber, and Tsai (1999) note that buy-and-hold returns avoid the upward bias that exists in rebalanced (e.g.
equal-weighted) portfolio returns. Such bias was first emphasized by Blume and Stambaugh (1983). Asparouhova,
Bessembinder, and Kalcheva (2012) show that the bias extends also to coefficients estimated in OLS regressions
with security or portfolio returns as the dependent variable. 3Idiosyncratic volatility is computed as the annualized standard deviation of the residual stock returns in the Fama-
French three factor regression, as in Ang, Hodrick, Xing, and Zhang (2006). Illiquidity is computed as the average
of the daily ratio of absolute stock return to dollar trading volume, as in Amihud (2002). Momentum is measured by
the cumulative return from 12 months to 2 months before the match date, as in Jegadeesh and Titman (1993). We
skip one month before the match date when calculating momentum to avoid the short-term reversal documented by
Jegadeesh (1990).
3
than their matching firms. All three types of event firms (M&As, SEOs, and IPOs) are
significantly less illiquid than their matching firm, and all three types are characterized by higher
market betas than their matching firms, while both SEO and M&A merger firms have
significantly higher momentum than their size and B/M matched counterparts.
Differences in idiosyncratic volatility, illiquidity, market beta, and momentum across
event firms and their matched counterparts potentially affect BHARs. Ang, Hodrick, Xing, and
Zhang (2006) among others show that idiosyncratic volatility is negatively associated with
expected stock returns.4 Amihud and Mendelson (1986), Amihud (2002), and Pástor and
Stambaugh (2003), among others, find that both measures of illiquidity and systematic illiquidity
risk are priced in the cross-section of stock returns.5 Numerous studies, commencing with
Jegadeesh and Titman (1993), have documented that firms with relatively high recent returns
tend to continue earning high returns in the intermediate term.
These results imply that matching event and control firms on the basis of firm size and
B/M may not provide a complete control for differences in average returns that are unrelated to
the event being studied, i.e. that the matched firms may not provide an appropriate benchmark to
assess the event firms’ performance. Consistent with this reasoning, when we divide sample
event firms into quintiles based on differences in idiosyncratic volatility and illiquidity between
event firms and matching firms over the 12 months before M&As and SEOs or over the 12
months after IPOs, we observe that BHARs significantly decrease with the difference in
idiosyncratic volatility and increase with the difference in illiquidity.
4 Fu (2009), in contrast, finds that expected idiosyncratic volatility is positively associated with expected stock
returns. Fu (2009) and Ang, Hodrick, Xing, and Zhang (2006, 2009) imply return premia with opposite signs. See
the following research for more evidence on the relation between idiosyncratic volatility and stock returns: Bali and
Cakici (2008), Guo and Savickas (2008), Huang, Liu, Rhee, and Zhang (2010), and Jiang, Xu, and Yao (2009). 5 See Amihud, Mendelson, and Pedersen (2005) for summaries of the literature on illiquidity and stock returns.
4
In light of the evidence that firm size and B/M ratio matching is only partially successful
in allowing for differences in mean returns across event and matched firms, and since matching
on a large number of firm characteristics is likely to be impractical, we propose a simple new
approach to measuring abnormal returns after corporate events. In particular, we propose to
assess abnormal returns as the intercept obtained when regressing differences in monthly log
returns across matched and control stocks on standardized differences in relevant firm
characteristics, including idiosyncratic volatility, illiquidity, momentum, and market beta.
A substantial prior literature has focused on statistical issues that arise when testing
whether long run BHARs differ from zero.6 Lyon, Barber, and Tsai (1999) note that the matched
firm approach, where buy-and-hold returns to event firms are compared to returns on size and
B/M matched control firms, yields test statistics that are well-behaved in random samples.
These apparently well-behaved test statistics indicate significantly negative long run abnormal
returns to bidding firms, IPO firms, and SEO firms (though the latter only on a value-weighted
basis). We contribute to this literature from a new angle, by investigating whether additional
firm characteristics, including idiosyncratic volatility, illiquidity, and momentum, are able to
explain observed BHARs. We show that the proposed regression approach that allows for
differences in firm characteristics across event and control firms can fully explain observed
BHARs when studying M&A, IPO, and SEO events. The method therefore substantially
reconciles the conflicting evidence in the literature obtained from BHAR versus time series
portfolio return methods.7
6 See Fama (1998), Brav (2000), and Kothari and Warner (2007) for summaries of this literature. For details of the
statistical issues and suggested solutions, see Barber and Lyon (1997), Kothari and Warner (1997), Lyon, Barber,
and Tsai (1999), Mitchell and Stafford (2000), and Jegadeesh and Karceski (2004). 7 Further, since the method is implemented in continuously compounded returns, it is free of biases attributable to
noisy prices identified in Asparouhova, Bessembinder, and Kalcheva (2012), while the calendar time portfolio
method is not.
5
The extant study that is closest to our own in terms of the research question posed is
Eckbo, Masulis, and Norli (2000). They note that the negative BHARs observed for SEO firms
in earlier studies potentially arise because event firms differ from matched firms in terms of
sensitivity to an array of macroeconomic risks. Consistent with this interpretation, they estimate
time series regressions of “zero-investment” portfolio (long event firms and short control firms)
returns on macroeconomic factors and report insignificant intercepts and some significant slope
coefficients. We note, though, that they estimate insignificant intercepts for portfolios of event
firms and portfolios of control firms as well, implying that abnormal returns are zero for both
SEO and control firms. Their study therefore highlights the tension between results obtained by
studying calendar time portfolio alphas versus studying BHARs relative to matched firms. We
show how this tension can be resolved by considering additional firm characteristics.
I. Data
I.A. Our Samples of Event Firms
This paper focuses in particular on the impacts of imperfect matching on long-run
abnormal stock returns after three types of corporate events: mergers and acquisitions, seasoned
equity offerings, and initial public offerings. However, the insights obtained here potentially
apply to studies of other corporate events such as stock splits, dividend initiations, etc.
To form the merger and acquisition sample, we identify completed mergers and
acquisitions by US public companies over the period 1980 to 2005 from Thomson Financial’s
SDC database. The sample ends with deals completed in 2005, to allow a five-year period to
measure bidder firms’ long-run stock returns. We impose the following filters. First, the
acquisition must take the following forms: merger (SDC deal form M), acquisition of majority
6
interest (AM), acquisition of remaining interest (AR), acquisition of partial interest (AP), or
acquisition of assets (AA). Second, the acquisition must be a control bid where the acquirer
owns less than fifty percent of the target and intends to hold more than fifty percent of the target
after the acquisition. These two filters follow Betton, Eckbo, and Thorburn (2008). In addition,
we require the relative size of the deal (transaction size divided by the market value of the bidder
firm before deal completion) to be greater than five percent, and the transaction value to be more
than one million dollars. The last filter excludes small deals that are less likely to have material
impacts on the long-run performance of the acquirer. We are able to identify 5,148 such
transactions.
We identify matching firms for the mergers and acquisitions sample using a procedure
similar to that of Lyon, Barber, and Tsai (1999) and Eckbo, Masulis, and Norli (2000). Each
matched firm is selected, on the basis of data from the end of December preceding the deal, as
the firm with the closest book-to-market ratio among firms with market capitalization between
70% and 130% of the bidder firm. To be included, the matching firm must not have itself
acquired other firms during the ten years around the matching date. If a matching firm delists,
then the candidate matching firm with the second closest book-to-market ratio is added for the
remainder of the five-year period. If the second matching firm delists, the candidate matching
firm with the third closest book-to-market ratio is added, etc. We are able to identify matching
firms for 4,579 of the bidder firms in our sample.
To form the SEO sample we first identify all completed SEOs contained in the SDC
database during the 1980 to 2005 period. Following Eckbo, Masulis, and Norli (2000, 2008) we
exclude ADRs, GDRs, unit offerings, financial companies, and public utilities. There are 7,204
such deals. For each, we select as a match the firm with the closest book-to-market ratio among
7
firms whose market capitalization is between 70% and 130% of the SEO firm at the end of
December preceding the SEO. To be included, the matching firm must not have conducted any
SEO during the ten years around the matching date. If a matching firm delists, then the
candidate matching firm with the second closest book-to-market ratio is added for the remainder
of the five-year period. We are able to identify matching firms for 5,573 of the SEO offerings in
our sample.
To form the IPO sample we first identify all completed IPOs in the SDC database over
the period 1980-2005, excluding REITs, closed-end funds, and ADRs, of which there are 9,035.
To form the matched firm sample we follow Loughran and Ritter (1995). Each IPO firm is
matched, based on data at the end of December after the IPO, with the firm with the closest but
greater market capitalization.8 To be included, the matching firm must have been publicly traded
for more than five years. If a matching firm delists, then the candidate matching firm with the
second closest market capitalization is substituted for the remainder of the five-year period We
are able to identify matching firms for 8,987 of the IPO firms in our sample.
Table 2 reports the number of event firms in each sample on an annual basis. The
number of M&A deals is small before 1984 due to the limited coverage of the SDC database.
The number of event firms also shows substantial variation across sample years. In particular,
the number of IPOs jumps substantially in the 1990s to a peak of 797 in 1996, before declining
to less than 100 per year in years 2001 to 2003.
I.B. Characteristics of Event and Control Firms
The matching of event with control firms on the basis of firm size and B/M ratio is
standard in the literature on long run returns. However, as we demonstrate, firms matched on
8 Loughran and Ritter did not match on B/M ratio.
8
these characteristics are not necessarily well-matched on other characteristics, including
idiosyncratic volatility, illiquidity, return momentum, and market beta. In addition, firms that are
well-matched in terms of size and market capitalization as of a given date do not typically remain
well-matched in these dimensions as time passes.
We construct a measure of idiosyncratic volatility following Ang, Hodrick, Xing, and
Zhang (2006). In particular, for each event and control stock, idiosyncratic volatility is
computed as the annualized standard deviation of the residuals in a regression of daily stock
returns on the three Fama-French (1993) factors. We compute separate estimates for each of the
60 months before and after the corporate event. We measure illiquidity using the metric
introduced by Amihud (2002). In particular, for each stock and for each of the 60 months before
and after the corporate event, we compute the Amihud illiquidity measure as the average of the
daily ratio of absolute stock return to dollar trading volume. For each of the 120 months around
corporate events, we calculate return momentum for event and matching firms as their respective
cumulative returns from the 12th
month to the 2nd
month prior to that month. Market beta is
estimated for each firm in each of the 120 months around the corporate events by implementing
the market model in daily stock returns.
Figure 1 displays cross-sectional median idiosyncratic volatility, illiquidity, return
momentum, market beta, firms size, and B/M ratio for bidding firms and for matched firms, on a
monthly basis from 60 months before to 60 months after the bid. Notably, sample bidding firms
have greater median idiosyncratic volatility than their matching firms. Bidding firms also differ
from control firms in terms of illiquidity, as the median illiquidity of bidding firms is always
smaller than that of their matching firms over the 120 months around the M&A. Though not
displayed, bidding firms also have greater mean idiosyncratic volatility and smaller mean
9
illiquidity than their matching firms. We also observe that bidding firms have larger market
betas throughout the 120 month interval, and higher return momentum during approximately the
24 months surrounding the bid.
Figure 2 plots the same information for SEO sample firms and their size-and-book-to-
market-matched comparable firms. The SEO firms have greater median idiosyncratic volatility
than their matching firms over the 120 months around SEO. SEO firms have much lower levels
of illiquidity and larger betas as compared to their matching firms, particularly in the months
after the SEO. Further, SEO firms have greater return momentum in approximately the twenty
four months around the issue.
Figure 3 displays monthly medians on the same variables for the sample IPO firms and
their size-matched comparable firms over the 60 months after going public. The IPO firms are
characterized by substantially greater median idiosyncratic volatility and market beta, and
moderately less median illiquidity as compared to their matching firms.
Figures 1 to 3 also provide evidence regarding the extent to which event and control
firms are well-matched on the basis of firm size and book-to-market through time. While each
sample is indeed well matched on average at a point in time (the December prior to the event for
the merger and SEO samples and the December after the event for the IPO sample), the
closeness of the match degrades as time passes. The median size of bidding firms exceeds that
of control firms by the event date, and thereafter. The median size of SEO firms increases
substantially in the months before the SEO, and substantially exceeds that of the matched sample
throughout the post event period. In contrast the median size (as well as the median book-to-
market ratio) of IPO firms is considerably less than that of control firms during most of the post-
event period.
10
To summarize, Figures 1-3 show that the event firms differ on average from their size-
and-book-to-market-matched comparable firms in terms of idiosyncratic volatility, illiquidity,
market beta, and return momentum. All three types of event firms have greater idiosyncratic
volatility, smaller median illiquidity, and larger market betas than their matching firms. Since a
number of studies have shown mean security returns to be related to idiosyncratic volatility,
momentum, beta, and illiquidity, these results imply that divergences in buy-and-hold abnormal
returns (BHARs) after the corporate events returns across event and control firms in the months
following the events may be attributable in whole or part to differences in idiosyncratic volatility
and illiquidity. In addition, while event and control firms are well matched in terms of size and
book-to-market ratio at a point in time, the closeness of the match degrades in subsequent
months.
We assess in Section II whether BHARs are indeed related to divergences in firm
characteristics across event and control firms. In Section III we assess whether differences in
idiosyncratic volatility, illiquidity, momentum, and market beta, in combination with “drift” in
size and book-to-market ratios, can fully explain average BHARs for sample firms.
II. Firm Characteristics and BHARs: Univariate Analysis
To assess the potential effects of idiosyncratic volatility, illiquidity, return momentum,
and beta on computed BHARs, we divide the sample firms into quintiles based on the differences
in these characteristics between the event firms and their matching firms. For the M&A and
SEO samples we compute idiosyncratic volatility, illiquidity, return momentum, and beta over
the 12 months before the event, while for the IPO sample, these four firm characteristics are
computed over the 12 months after the IPO. We then average these estimates across the 12
11
months, and compute the difference in the average across event and matching firms. The
difference in return momentum between an event firm and its matching firm is calculated using
the cumulative returns over months -12 to -2 before the event (for M&As and SEOs) or over
months 1 to 11 after the event (for IPOs).
Table 3 reports the 10th
through the 90th
percentiles of the distribution of differences in
idiosyncratic volatility, illiquidity, return momentum, and beta between the event firms and their
matching firms. The results indicate substantial variation across firms in the extent of the
mismatch between event and control firm idiosyncratic volatility and illiquidity. For example,
more than forty percent of the SEO sample have idiosyncratic volatility that differs (in absolute
value) by more than 12% per year from that of their matching firms. The idiosyncratic volatility
of the 10th
percentile IPO firm differs from that of the 90th
percentile IPO firm by over 75% per
year. Sixty percent of bidding firms have estimated market betas that differ by more than 0.8, etc.
In Table 4 we report equal- and value-weighted BHARs for event firms over the 60
months after the corporate events. Following Eckbo, Masulis, and Norli (2008), the buy-and-
hold return of each event firm is calculated as its cumulative stock return from the first month
after the event to the earliest of the 60th
month after the event, the delisting date of the event firm,
or the next corporate event of the same type. Each BHAR is computed as the buy-and-hold
return for the event firm less the buy-and-hold return for the matching control firm. We divide
event firms into quintiles based on differences in idiosyncratic volatility, illiquidity, return
momentum, and beta across event and control firms, and report average BHARs by quintile in
Table 4.
Panel A presents the BHARs of the bidding firms. Consistent with Betton, Eckbo, and
Thorburn (2008) we find that, for the full sample (reported on the bottom row), bidding firms
12
suffer negative BHARs of -12.04% on an equal- weighted basis and -17.16% when value-
weighted. However, BHARs differ substantially across subsamples with differing firm
characteristics. The BHARs of bidding firms significantly decrease as the difference in
idiosyncratic volatility increases. Bidding firms in the first two quintiles have modestly negative
or positive BHARs while the bidding firms in the last two quintiles have large negative BHARs.
For example, the bidding firms in the first idiosyncratic volatility quintile have a 60-month
equal-weighted BHAR of -2.60%, while those in the fifth quintile have a BHAR of -46.49%.
The BHARs of bidding firms significantly increase as the difference in illiquidity
increases. For example, the bidding firms in the first illiquidity difference quintile have a 60-
month equal-weighted BHAR of -29.49%, while those in the fifth quintile have a BHAR of
+13.09%.
In addition, we find that the 60-month BHARs weighted by market capitalization of the
bidding firms significantly decrease as the differences in beta and return momentum increase.
For example, the bidding firms in the first momentum difference quintile have 60-month value-
weighted BHARs of -9.61%, while those in the fifth quintile have a 60-month value-weighted
BHARs of -50.05%. Figure 4 depicts the BHARs of the bidding firms across the quintiles based
on idiosyncratic volatility, illiquidity, return momentum, and beta.
Panel B of Table 4 reports on BHARs for sample SEO firms. Consistent with Eckbo,
Masulis, and Norli (2000), Loughran and Ritter (1995), and Spiess and Affleck-Graves (1995),
we find that SEO firms earn negative BHARs: the equal-weighted 60-month BHARs are -8.65%,
while the value-weighted 60-month BHARs are -19.55%. However, we again observe cross-
sectional variation in BHARs related to firm characteristics. The BHARs for SEO firms
significantly decrease as the difference in idiosyncratic volatility increases, a similar pattern as
13
the bidding firms. The SEO firms in the first quintile based on difference in idiosyncratic
volatility have positive five-year BHARs, while those in the last quintiles have negative BHARs.
For example, the SEO firms in the first idiosyncratic volatility difference quintile have a 60-
month equal-weighted BHAR of +18.75%, while those in the fifth quintile have a BHAR of -
26.68%.
The BHARs of the SEO firms also significantly increase as the difference in illiquidity
increases. The pattern, again, is similar to that of the bidding firms. For example, the SEO firms
in the first quintile based on difference in illiquidity have a 60-month equal-weighted BHAR of -
13.07%, while those in the fifth quintile have a BHAR of +13.08%.
The value-weighted BHARs of the SEO firms also significantly decrease as the
differences in beta and return momentum increase. For example, the SEO firms in the first
momentum difference quintile have a60-month value-weighted BHARs of -6.74%, while those
in the fifth quintile have 60-month value-weighted BHARs of -35.19%. Figure 5 depicts the
BHARs of the SEO firms across the quintiles based on idiosyncratic volatility, illiquidity, return
momentum, and beta.
Panel C of Table 4 presents BHARs for IPO firms. Consistent with Loughran and Ritter
(1995) and Eckbo, Masulis, and Norli (2008), IPO firms suffer significant negative BHARs for
the full sample, of -47.90% on an equal-weighted basis and -55.06% on a value-weighted basis.
The BHARs of the IPO firms significantly decrease as the difference in idiosyncratic volatility
increases, a similar pattern as the bidder firms and the SEO firms. The IPO firms in the first two
quintiles based on difference in idiosyncratic volatility have positive BHARs, while those in the
last two quintiles have significantly negative BHARs. For example, the IPO firms in the first
idiosyncratic volatility difference quintile have a 60-month equal-weighted BHAR of -8.59%,
14
while those in the fifth quintile have a BHAR of -113.44%. In contrast, the BHARs of the IPO
firms are not monotone across the quintiles based on difference in illiquidity. The value-
weighted BHARs of the IPO firms decrease significantly as the beta difference increases. The
IPO firms in the first beta difference quintile have value-weighted BHARs of -24.49%, much
greater than the value-weighted BHARs of the IPO firms in the fifth beta difference quintile
which are -104.81%. Figure 6 depicts the BHARs of the IPO firms across the quintiles based on
idiosyncratic volatility and illiquidity.
In summary, we find that long-run BHARs for M&A bidder firms and firms issuing
initial and secondary equity offerings are systematically related to differences in idiosyncratic
volatility and illiquidity between the event firms and their matching control firms. In general,
the effects on BHARs are as would be anticipated based on the extant literature. In particular,
Ang, Hodrick, Xing, and Zhang (2006) document lower average returns for firms with greater
idiosyncratic volatility. Consistent with this insight, we document larger BHARs for firms with
lower idiosyncratic volatility as compared to their matched counterparts, and vice versa, for the
bidder, SEO, and IPO samples. Amihud (2002) and others document higher average returns for
firms with greater illiquidity. Consistent with this finding, we document smaller BHARs for
firms with smaller illiquidity, as compared to their matched counterparts, and vice-versa, for the
bidder and SEO samples. For the IPO sample, in contrast, the relation between BHARs and the
extent of the illiquidity mismatch is not monotone. We also find weaker evidence that the
BHARs of the event firms are associated the differences in beta and return momentum between
the event firms and their matching firms.
The analysis in this section was based on univariate comparisons. Of course, illiquidity,
idiosyncratic volatility, return momentum, and beta may be correlated, and may all matter for
15
BHARs. We next turn to a multivariate analysis of BHARs that simultaneously considers
illiquidity, idiosyncratic volatility, firm size, market to book ratios, return momentum, and
market beta.
III. Firm Characteristics and BHARs
In Section II, we show that firm characteristics, and in particular illiquidity and
idiosyncratic volatility, are significantly related to the BHARs of the event firms. In this section,
we propose a general framework to investigate the effects of time series and cross-sectional
variation in firm characteristics on long-run abnormal stock returns after corporate events.
III.A Model for Long-run Stock Returns after Corporate Events
The BHAR of event firm e over T months after a corporate event at date 0 is:
})1ln(exp{})1ln(exp{ )1()1(1111
T
t
mt
T
t
et
T
t
mt
T
t
eteT rrrrBHAR , (1)
where ret and rmt are the monthly stock returns of the event firm and its firm-characteristics-
matched comparable firm, respectively. Consider also the wealth relative (WR) as defined by
Ritter (1991) and Loughran and Ritter (1995):
.
)1(
)1(
)}1ln()1{ln(exp{
1
1
1
T
t
mt
T
t
et
mt
T
t
eteT
r
r
rrWR
(2)
Wealth relative measures the T-period gross return to a $1 investment in the event firm relative
to the T-period gross return to the same investment in the matching firm. Testing whether
0eTBHAR is equivalent to testing whether 1eTWR , as both equations hold if the time series
mean log return is equal across event and control firms.
16
Guided by the evidence reported in Section II and the equivalence between testing
expressions (1) and (2), we propose to explain the differences in monthly stock returns between
the event firm and its matching firm based on differences in firm characteristics. In light of the
evidence displayed on Figures 1 through 3, we measure firm characteristics on a monthly basis,
allowing time variation in the closeness of the match between event and control firms to
contribute to a potential explanation of abnormal returns.
We consider six characteristics that have been shown to be associated with expected
Table 6: Explaining Differences in Stock Returns between Event Firms and Their Matching Firms
This table presents the OLS/Fama-MacBeth regression results for the difference in the log monthly returns of the event firm and its size- and B/M-matched
comparable firms. Each event firm is matched with a comparable firm based on size and B/M. For mergers and acquisitions and SEOs, each event firm is
matched with a firm whose market capitalization is between 70% and 130% of the event firm and has the closest book-to-market ratio at the end of the latest
December before the event. At the end of December after IPO, each IPO firm is matched with a firm with the closest but greater market capitalization. Market
beta for July of year t to June of year t+1 is estimated with the monthly stock returns in years t-5 to t-1 using the market model. Size is the market capitalization
at the end of the latest June. B/M is defined as the ratio of the book value of common equity at the end of fiscal year t-1 to the market value of common equity at
the end of the latest June. Momentum is the cumulative return over months -12 to -2. Idiosyncratic risk is the annualized standard deviation of the residual daily
stock returns in the Fama-French three factor regression in month -2. Illiquidity for July of year t to June of year t+1 is computed as the average daily ratio of
absolute stock return to dollar trading volume from July of year t-1 to June of year t divided by the market average illiquidity over the same period, as defined by
Amihud (2002). For each of the six firm characteristics (beta, size, B/M, momentum, illiquidity, and idiosyncratic volatility), we compute the difference between
the event firm and its size-and B/M-matched comparable firm. In every month over our sample period, the positive differences in each firm characteristic are
ranked and normalized to be its percentile ranking. All negative differences are ranked and normalized to be one minus its percentile ranking. Consequently, the
normalized differences in each firm characteristics take a value from -1 to 1, with 0 corresponds to the difference in firm characteristic that is the closest to 0.
Panels A1, B1, and C1 report the pooled-OLS regression results for our M&A, SEO, and IPO samples respectively; Panels A2, B2, and C2 report the Fama-
MacBeth regression results. All model specifications employ robust standard errors. The associated t-statistics are reported in the parentheses below each
coefficient. Superscripts ***, **, and * correspond to statistical significance at the one, five, and ten percent levels, respectively.
49
Panel A1: Mergers and Acquisitions – Pooled OLS Regressions
Table 7: Explaining Differences in Stock Returns between Event Firms and Their Matching Firms: Nonlinear Relations
This table presents the OLS/Fama-MacBeth regression results for the difference in the log monthly returns of the event firm and its size- and B/M-matched
comparable firms. Each event firm is matched with a comparable firm based on size and B/M. For mergers and acquisitions and SEOs, each event firm is
matched with a firm whose market capitalization is between 70% and 130% of the event firm and has the closest book-to-market ratio at the end of the latest
December before the event. At the end of December after IPO, each IPO firm is matched with a firm with the closest but greater market capitalization. Market
beta for July of year t to June of year t+1 is estimated with the monthly stock returns in years t-5 to t-1 using the market model. Size is the market capitalization
at the end of the latest June. B/M is defined as the ratio of the book value of common equity at the end of fiscal year t-1 to the market value of common equity at
the end of the latest June. Momentum is the cumulative return over months -12 to -2. Idiosyncratic risk is the annualized standard deviation of the residual daily
stock returns in the Fama-French three factor regression in month -2. Illiquidity for July of year t to June of year t+1 is computed as the average daily ratio of
absolute stock return to dollar trading volume from July of year t-1 to June of year t divided by the market average illiquidity over the same period, as defined by
Amihud (2002). For each of the six firm characteristics (beta, size, B/M, momentum, illiquidity, and idiosyncratic volatility), we compute the difference between
the event firm and its size-and B/M-matched comparable firm. In every month over our sample period, the positive differences in each firm characteristic are
ranked and normalized to be its percentile ranking. All negative differences are ranked and normalized to be one minus its percentile ranking. Consequently, the
normalized differences in each firm characteristics take a value from -1 to 1, with 0 corresponds to the difference in firm characteristic that is the closest to 0. All
model specifications employ robust standard errors. The associated t-statistics are reported in the parentheses below each coefficient. Superscripts ***, **, and *
correspond to statistical significance at the one, five, and ten percent levels, respectively.