Crises, Liquidity Shocks, and Fire Sales at Hedge Funds

Crises, Liquidity Shocks, andFire Sales at Hedge FundsNicole Boyson, Jean Helwege, and Jan Jindra

This document is a paper presented at the Annual Meeting of theMidwest Finance Association, March 15, 2013. It is also NortheasternUniversity D’Amore-McKim School of Business Research Paper no.2013-03.

Crises, Liquidity Shocks, and Fire Sales at Hedge Funds*

Nicole Boysona

Northeastern University

Jean Helwegeb University of South Carolina

and

Jan Jindrac

Ohio State University

November 29, 2012

ABSTRACT

We investigate hedge fund stock trading from 1998-2010 to test for fire sales. While funds with high capital outflows sell large amounts of stock during crises, these funds also buy stock, rather than using all the proceeds to fulfill redemptions. Further, funds with large outflows rarely sell the same stocks at the same time. For the relatively few stocks that are sold en masse, there is no evidence of price pressure, largely because hedge funds overwhelmingly choose to sell their most liquid, largest, and best-performing stocks. We provide new and compelling evidence that hedge funds neither engage in nor induce fire sales, since their well-diversified portfolios allow them to cherry-pick the most appropriate stocks to sell during crises. JEL Classification: G21, G24, G28, G32, G33, E44, E58, E61 Keywords: Liquidity shocks, fire sales, financial crisis, hedge funds, financial constraints

*We are grateful for helpful comments from Nikunj Kapadia, Rabih Moussawi, Rene Stulz, and seminar participants at Northeastern University. a Nicole Boyson, Northeastern University, Boston, MA 02186; email: [email protected]; phone: 617.373.4775. b Jean Helwege, University of South Carolina, Columbia, SC 29208; email: [email protected]; phone: 803.777.4926. Corresponding author.

c Jan Jindra, Ohio State University, Columbus, OH 43210; email: [email protected]; phone: 650.489.6807.

1

While the subprime crisis is widely viewed as the result of a severe housing shock, it is

less clear how problems in the housing sector were transmitted via financial intermediaries to the

rest of the economy. A number of academic studies focus on the connection between financial

crises, liquidity shocks, and asset fire sales to explain the large dislocations in financial markets.

Because of their capital structures, hedge funds likely suffer significantly from prolonged

pressure on their balance sheets when liquidity dries up; hence, their asset sales could be central

to the amplification of crises (e.g., Khandani and Lo (2007), Brunnermeier (2009), Brunnermeier

and Pedersen (2009), and Aragon and Strahan (2012)).

To date, however, no study has directly tested for the existence of fire sales by hedge

funds. If hedge funds are large sellers of stock during crises at fire sale prices, their sales can lead

to liquidity spirals that exacerbate the crisis, and they deserve some blame for its propagation. If,

however, hedge fund stock sales do not occur at fire sale prices, these sales will not exacerbate

the crisis and hedge funds are less culpable in its propagation. While prior studies document that

hedge funds do indeed sell more stock during crises (Ben-David, Franzoni, and Moussawi (2012)

and He, Khang, and Krishnamurthy (2010)), and thus are suggestive of fire sales, the literature

stops short of performing direct tests.

In this paper we analyze whether hedge fund stock sales in crises can be characterized as

fire sales. We use two fund-specific proxies for liquidity shocks: fund flows and fund-imposed

capital lockup and redemption periods. Controlling for a variety of fund characteristics, we find

that during crises funds with high outflows sell significantly more stock than funds with high

inflows, consistent with liquidity shocks imposing selling pressure. However, on average, funds

with outflows use nearly all their proceeds from stock sales to buy additional stock, inconsistent

2

with forced sales. We also note that the buying is not likely a result of covering existing short

positions, since it is doubtful that hedge funds would continue to hold these stocks at quarter-end.

Next, we investigate whether funds that sell stock engage in fire sales. A necessary

condition for fire sales is that hedge funds experiencing outflows sell stocks that are common

across their portfolios. We find that such sales of stocks are fairly uncommon, accounting on

average for only about 1% of the total stocks held by hedge funds each quarter. We then test for

fire sales using the approach of Coval and Stafford (2007), which examines the price impact of

en masse stock sales by funds with significant capital outflows (hereafter, forced sales) and en

masse stock purchases by funds with large inflows (forced buys). Our results show no evidence

of fire sales during crises. Although funds with large outflows sell a large quantity of stock in

crises, these transactions do not lead to the sort of price pressure that characterizes fire sales.

Finally, we examine characteristics of forced sale and purchase stocks to understand why

sales (purchases) have so little impact on prices. Although we identify few forced sales, these

sales are substantial in size, comprising about 1.5% of the stock’s total market capitalization and

3.1% of the stock’s quarterly volume, and are about twice this high in December 2008. Forced

purchases are larger, on average, comprising 2.5% of market capitalization and 5.9% of trading

volume, but do not change noticeably during the subprime crisis. Importantly, we find that these

large transactions do not involve price pressure because forced sale and purchase stocks are

significantly different from other stocks held by hedge funds. In all periods, and even more so

during crises, forced sale stocks have better prior year performance, larger size, higher average

trading volume, and better liquidity. Therefore, when forced to sell, hedge funds avoid fire sales

by cherry-picking stocks based on high liquidity, large size, and good past performance. Forced

purchase stocks do not involve price pressure for the same reason, as those stocks are similar in

3

characteristics to forced sale stocks. A key inference of this result is that most hedge funds are

quite well-diversified and refrain from selling less liquid, smaller, and poorly performing stocks,

thereby managing to avoid fire sales altogether.

The paper is organized as follows. The next section discusses the literature and develops

our hypotheses. Section II describes the data. Section III examines the factors that drive hedge

funds to sell stocks. Section IV performs the fire sale tests. Section V studies the characteristics

of stocks in hedge fund portfolios, and Section VI concludes.

I. Analytical Framework

The recent financial crisis has focused researchers’ attention on the need for liquidity

among financial institutions that may be forced to sell assets in a crisis if they run short of cash.

For example, in Brunnermeier and Pedersen (2009), hedge funds and other speculators finance

their trades through collateralized borrowing from banks, which could cause a liquidity squeeze

if hedge funds experience a decline in the value of their assets, an increase in margins, or large

investor withdrawals. They posit that a fund will react to such shocks by selling assets, which in

turn will cause downward price pressure on the assets it sells. Their theory is closely related to

the work of Shleifer and Vishny (1992, 2011), who show that fire sales occur when firms are

forced to sell assets to repay debt at a time when other firms in the industry (that could extract

full value from the assets) are unable to purchase them. Research using amplification models,

such as Krishnamurthy (2009), links such fire sales to loss spirals that erode capital and

ultimately cause a financial crisis.1

1See also Gromb and Vayanos (2002), Adrian and Shin (2008, 2010), Diamond and Rajan (2010), Geanakoplos (2010), Allen and Carletti (2006), and Uhlig (2010).Boyson, Helwege, and Jindra (2012a, 2012b) evaluate these theories for commercial banks and investment banks, respectively, and find little evidence of widespread fire sales during financial crises.

4

These theories start with the premise that a crisis arises because financial firms are forced

to sell assets as a result of funding pressures. Thus, we first consider the extent to which hedge

funds are forced to sell assets as a result of liquidity shocks. We investigate hedge fund trading

during crises and consider whether pressure to sell, if it exists, is greater among hedge funds that

are more likely under funding pressure, including funds with capital outflows, short lockup

periods, and/or short redemption periods. We also examine the characteristics of funds that

experience these pressures during crises and investigate whether there are systematic differences

in outflows across funds of different styles. Previous studies of the subprime crisis indicate that

hedge funds were forced into severe asset sales (Ben-David, Franzoni, and Moussawi (2012) and

He, Khang, and Krishnamurthy (2010)) and that these sales led to liquidity problems in the

stocks they held (Aragon and Strahan (2012)). Ben-David, Franzoni and Moussawi (2012)

conclude, however, that some hedge funds sold stock in order to invest in profitable

opportunities elsewhere. Further, Chen, Hanson, Hong and Stein (2008), Zhang (2009) and

Massa, Simonov, and Yan (2012) present evidence that hedge funds and mutual funds are often

the buyers who profit from fire sales by other institutions.

We first consider whether hedge funds are forced to sell stock as a result of a market-

wide breakdown in liquidity. We identify systematic shocks by separating the sample into crisis

and boom periods, and we investigate idiosyncratic shocks using data on fund flows, fund

performance, leverage usage, and lockup and redemption periods. If liquidity shocks force hedge

funds to sell in a crisis, stock sales should be higher during crises than during booms. Further,

sales should be concentrated in hedge funds facing capital outflows or having other financial

constraints. These patterns in gross sales should be matched in the data on net sales (i.e., sales of

existing stock less purchases of new stock). After all, if hedge funds are under pressure to sell

5

stocks due to redemption requests, they should use the proceeds to fund cash withdrawals and

not to buy new stock.

For hedge funds to be a source of stress in a crisis, not only must they sell large quantities

of assets to meet margin calls and redemptions, but also the asset sales must put downward

pressure on prices. Boyson, Stahel, and Stulz (2010) link contagion in hedge fund returns to

liquidity shocks, and Hameed, Kang and Viswanathan (2010) and Maier, Schaub, and Schmid

(2011) argue that funding problems lead to decreases in liquidity and stock prices. Sadka (2010)

and Billio, Getmansky, and Pelizzon (2010) tie aggregate liquidity risk to underperformance

during crises, although the latter also argue that hedge funds’ contributions to systemic risk were

not especially large.

By contrast, using a proprietary database of institutional trades, Anand, Irvine, Puckett,

and Venkataraman (2012) show that during the crisis, institutions tilted their selling activity

towards less liquidity-sensitive stocks suggesting that such stocks may serve as a liquidity hedge

for institutions in a downturn. Ambrose, Cai, and Helwege (2012) argue that selling assets in the

absence of information effects does not cause losses in efficient markets. Brown, Green, and

Hand (2010) note that few hedge funds failed during the crisis, and that hedge fund losses were

small compared those reported by other financial institutions. Further, recent work provides

evidence that hedge funds adjust their trading activities to better withstand crises. For example,

Ang, Gorovvy, and van Inwegan (2012) find that hedge funds decreased leverage before and

during the subprime crisis, and Liu and Mello (2011) show theoretically that hedge funds will

hold more cash during crises due to their fragile capital structures. Finally, Shadab (2009) argues

that hedge funds actually reduced the severity of the financial crisis by acting as net purchasers

of mortgage-backed securities and as lenders to distressed borrowers.

6

Thus, prior evidence is inconclusive regarding whether selling by hedge funds drives

down stock prices in a crisis. We fill this gap by examining the impact of sales during crises on

stock prices. We use a test of forced sales (Coval and Stafford (2007)), which disentangles price

pressure from information effects by analyzing sales that are motivated by necessity in periods of

distress. This approach allows us to consider a symmetric response to price pressure in booms

when hedge funds experience abnormally high inflows. We also examine the characteristics of

stocks sold in crises to determine the extent to which hedge funds can alleviate price pressure

through strategic selling of their stock holdings.

II. Data

Hedge fund data for the period 1998-2010 are obtained from two sources: mandatory 13F

filings for stock holdings for institutions with over $100 million under management (reported by

Thomson-Reuters) and the Lipper TASS hedge fund database. While data limitations require that

we focus only on long stock positions of hedge funds, common stock holdings comprise a

substantial 22% of hedge fund assets reported in Lipper TASS for fund families with over $100

million in assets under management. Further, using Lipper TASS data, Cao, Chen, Liang, and Lo

(2010) show that hedge funds have significant exposure to the stock market.

We begin with the dataset of Griffin and Xu (2009), which contains hedge funds with

13F filings during 1998-2004 and extend it for 2005-2010 by matching fund and manager names

in Lipper TASS with Thomson-Reuters.2 Since we require Lipper TASS data for all analyses, we

drop the Griffin and Xu (2009) funds that do not appear in Lipper TASS. Since we wish to

exclude companies with a substantial part of their business coming from non-hedge fund entities,

we check SEC-required ADV filings for all funds to ascertain that filers have at least 50% of

2 We thank John Griffin for providing this list. Lipper TASS reports data for individual hedge funds while 13F filings are at the fund family (management company) level. Note that this data does not suffer from the survivorship bias that is frequently encountered when using self-reported data.

7

their assets in hedge funds or at least 50% of their assets owned by high net worth individuals,

and that they also charge incentive fees.3 If ADV filings are not available, which happens for

twenty of the funds that appear in our final sample, we perform a web-based search to ensure that

a reliable source, such as the fund’s own website, a reputable news source such as Bloomberg, or

an interview with the hedge fund manager, identifies the entity as a hedge fund. As a final check

for these twenty funds, we search the CRSP mutual fund database to ensure these fund families

are not also managing mutual funds. This process yields 262 hedge fund families that appear in

both Lipper TASS and Thomson Reuters.4

Summary statistics for the hedge fund sample are in Table 1. Since the 13F data are

reported at the fund family level and the Lipper TASS data are at the fund level, we aggregate

the Lipper TASS data to the family level and take size-weighted averages of fund characteristics.

All variables in the paper are winsorized at the 5% and 95% tails. Panel A presents fund family

characteristics. The mean family size is about $743 million, minimum required investment is $2

million, mean family age is about 10 years, lockup period is 6 months, and redemption period is

156 days. The lockup period is the initial time after an investor purchases the fund during which

he is not permitted to withdraw his capital, and the redemption period is the sum of the

redemption notification period (number of days notice an investor most give before requesting to

withdraw his capital) and the redemption frequency period (how frequently an investor may

withdraw his capital during a year; for example, quarterly). Following Agarwal, Daniel, and

3For a brief period in 2006, all hedge funds were required to file ADV forms. In 2009, filing again became mandatory. Hence, in same cases, we are imputing the fund’s classification from an ADV form filed after the sample data. To test that this approach is appropriate, we compared the ADV filings for 2006 with those for 2009 or later for the same fund (when available) and found no differences in classification between the two filings. 4 Our sample is similar to that ofBen-David, Franzoni, and Moussawi (2012), who also investigate stock trading by hedge funds. They select their sample using a proprietary list from Thomson-Reuters of entities that file 13F forms, both in the aggregate and detailed by entity within the aggregate filings, and exclude all filings where more than half the included entities are not hedge funds. They then filter the remaining filings by examining ADV forms to ensure the majority of the filer’s business comes from hedge fund activity.

8

Naik (2009), we sum the redemption notification and frequency periods and report the total.

Turning to the indicator variables, we note that about 62% of funds report that they use leverage

and about 78% of funds have a high water mark. As for style characteristics, 52% of funds

follow a long-short equity style and 19% of funds follow an event-driven style. Less than 1% of

funds report following a dedicated short style, fixed income arbitrage, managed futures or equity

market neutral style.

Panel B reports the breakdown of the sample year by year, including the number of fund

families, the mean fund size as reported in the Lipper TASS database, the average equity

portfolio size reported in the 13F database, the number of stocks per manager, and the mean

change in hedge fund holdings per family. Means are first calculated by quarter and then

averaged across funds. The number of funds in the sample doubles from the beginning to the end,

and the mean fund size increases through 2007, drops in 2008-2009, and grows again in 2010.

The average equity portfolio size and number of stocks held remains steady throughout the

period. The change in hedge fund holdings, which is the value of stock purchases less sales

divided by the value of all stock holdings, measured by prior quarter market value, is highest

during boom periods and lowest in 2008, when the median change in holdings is -1%. Holdings

rebound during 2010. Comparing our results to Ben-David, Franzoni, and Moussawi (2012), the

summary data are quite similar to their sample for all periods of overlap (2004-2009).

Panel C presents information about hedge fund stock portfolios. The total mean (median)

market value of stocks in the portfolio is $726 billion ($329 million), and the mean (median)

number of stocks is 144 (64). The average (median) hedge fund holding as a percentage of the

stock’s market capitalization is about 1% (0.5%) and the average transaction size is about $5.6

million ($3.1 million). Finally, we report the average change in hedge fund holdings, calculated

9

as in Panel B, averaged across all quarters. The mean change is 9.3% with a median of 4.4%,

indicating that hedge funds have net quarterly purchases of about 9.3% of their prior period

assets. This value is in line with the 10.6% reported (for a slightly different time period) by Ben-

David, Franzoni, and Moussawi (2012).5 Finally, despite winsorization, the medians are much

smaller than the means, indicating the presence of some very large portfolios.

Last, Panel D reports characteristics at the stock level. The average (median) prior year

annual return is 8.2% (3.5%) with standard deviation of 13.0% (10.8%). These estimates use

monthly data from CRSP. The mean (median) total assets for each firm is $2 billion ($550

million), market capitalization is $2.4 billion ($745 million), and Tobin’s Q is 2.3 (1.7). Mean

(median) quarterly volume is 43 (15) million shares. This data is from Compustat. We also report

the Amihud (2002) illiquidity measure (multiplied by 108) and the bid-ask spread (multiplied by

102) using daily data, and the means (medians) are 0.972 (0.058) and 0.06 (0.03), respectively.

There is significant heterogeneity in stock returns given the 25th percentile value of -24% and the

75th percentile value of 31%, and evidence that some portfolios contain very large stocks. The

mean and 25th percentile values of the Amihud measure are quite high despite winsorization,

indicating that some hedge funds invest in very illiquid stocks. The Tobin’s Q, volume, and

liquidity measures are reasonable given the average size of the stocks held.

Following Boyson, Helwege, and Jindra (2012a, 2012b), we identify booms and crises

over the sample period of 1998-2010 by examining NBER cycles, bank failures, the TED spread,

Moody’s Aaa-Baa bond spreads, flight to quality (Collin-Dufresne, Goldstein, and Helwege

(2009)), the Long Term Capital Management (LTCM) episode, and credit crunches (Bordo and

Haubrich (2010)). Quartiles of the monthly distributions of bank failures, TED and credit spreads, 5Since this is a key variable in both our paper and in Ben-David, Franzoni, and Moussawi (2012), we compute its mean and median using their sample period (2004-2009), and obtain a mean of 0.084 (0.037). These values are close to their reported mean (median) values of 0.106 (0.033), confirming the general similarity of our samples.

10

stock market quarterly returns, NBER recessions, flight to quality, and LTCM episodes are used

to determine boom periods and crises. Table 2 presents the details on the three crisis periods and

two boom periods.6 For example, the crisis period from March 2007 through June 2009 is

characterized by an NBER contraction, a high number of bank failures, high TED and credit

spreads, high incidence of flight to quality, a credit crunch, and low stock market quarterly

returns. The boom period from January 2003 through February 2004 is characterized by an

NBER expansion, low credit spreads, and high stock market returns. For a particular observation

to be classified as occurring during a crisis (boom) period, the crisis must last for the full quarter.

III. Fund Flows and Stock Trading

A. Fund Flows

Our first set of tests considers whether hedge funds are forced to sell assets in a crisis. We

hypothesize that funds with large capital outflows should be under more pressure to sell, since

the only alternative for a fund facing outflows is to borrow the cash, which may be quite difficult

in a crisis. Likewise, we predict that hedge funds with large capital inflows are more likely to

buy stock. We calculate quarterly flows at the hedge fund family level (net flows/prior period

fund family assets), and divide this variable into quintiles each quarter.

Table 3 reports summary data on hedge fund flows. Panel A reports average fund flows

and provides evidence that fund outflows are larger during crises for all flow quintiles.

Differences between all crises/the subprime crisis and booms are significant for all quintiles,

with the most extreme outflows occurring in the subprime crisis among funds in (the worst)

quintile 1 (with a difference of -3.5% between the subprime crisis and boom periods). Notably,

regardless of time period, average flows are always negative for the worst two quintiles of flows

6 The remaining time periods are “neutral.” In unreported results we shorten the crisis periods by one quarter for both the beginning and the end and find that our conclusions are not sensitive to the change.

11

(quintiles 1 and 2), and average flows are negative for the first three quintiles of flows during all

crises and the subprime crisis.

Panel B provides evidence that outflows follow poor fund performance and inflows

follow good performance regardless of the time period, consistent with Baquero and Verbeek

(2009). Differences in performance between the worst and best flow quintiles average about -3%

for each of the time periods, which is statistically significant at the 1% level. During crises funds

exhibit the worst lagged performance, with negative performance for the first three flow quintiles

during crises. During booms, funds have positive performance even for the worst flow quintile,

and differences in performance between booms and the subprime crisis and for booms and all

crises range from -3.0% to -5.6% and are statistically significant for all quintiles studied.

Finally, Panel C presents the proportion of funds with outflows (negative flows).

Focusing on all funds during all periods, about 45% of funds have outflows; this value increases

to just over 50% for the subprime crisis. By contrast, the proportion is about 42% during booms.

Differences between booms and crisis periods, including the subprime crisis, are statistically

significant at the 1% level. Turning to the style results, there is notable variation in the

proportion of funds that experience outflows during crises. Strikingly, both the global macro and

convertible arbitrage styles have fewer funds with outflows during the subprime crisis than

booms, at 29% vs. 33% for global macro and 36% vs. 42% for convertible arbitrage in crises and

booms, respectively. These results are consistent with the stated strategies of these funds – since

convertible arbitrage funds generally hold short stock positions and global macro funds are

opportunistic in nature, these funds are probably the most likely to have good performance

during crises. However, due to small sample sizes, these results are statistically insignificant.

Both the stock-heavy long/short equity and event driven styles, which together represent over

12

two-thirds of the sample, have significantly more funds with outflows during the subprime crisis

than booms: 54% vs. 45% for long-short and 51% vs. 38% for event driven during the subprime

crisis and booms, respectively. Finally, both managed futures and fixed income arbitrage styles

have more funds with outflows during crises, although the small sample sizes here warrant

caution. Taken together, the results show that hedge funds experience worse outflows during

crises, that these outflows follow poor performance, and that the proportion of funds with

outflows increases in a crisis. In the next section, we analyze the selling (and purchasing) activity

of hedge funds.

B. Stock Sales and Purchases

Hedge funds frequently trade stocks, so any analysis of selling activity among these firms

in a crisis must consider the volume of transactions that ordinarily takes place in this asset class.

Moreover, while the volume of stock sales is best measured by the market value of the

transactions, changes in the market value of holdings are also caused by stock price changes. In

this paper, we use two separate approaches to estimate selling activity: (1) using the market value

of the shares sold, as in Ben-David, Franzoni and Moussawi (2012), and (2) using the number of

shares sold, as in Coval and Stafford (2007). The latter calculates percentage changes for each

stockholding in each portfolio each quarter based on the prior quarter’s holdings, categorizing

activity as shares eliminated, decreased, unchanged or increased, but does not include purchases

of new stocks. Table 4 Panel A reports these proportions for all funds all periods as 38%

eliminated, 22% decreased, 24% unchanged, and 16% increased, which sum to 100%. These

figures do not differ appreciably for the boom, crisis, or the subprime crisis.

In contrast to the Coval and Stafford approach, using the market value of the portfolio in

the previous quarter allows one to measure the amount of new shares purchased (although the

13

fractions no longer sum to 100%).7 Since these proportions are measured relative to prior period

assets, they represent changes in the equity holdings due to actual trades, not just due to changes

in price. We sum quarterly changes in holdings in prior quarter market value dollars for all

existing stocks eliminated, decreased, or increased, and all new purchases, and divide by the sum

of the market value of stockholdings for the fund in dollars in the prior quarter. Since this

calculation allows us to measure the dollar amount of new shares purchased and is relative to

prior period assets, we believe it is the most appropriate measure for our tests and focus on it for

the remainder of the discussion. For all funds in all periods, existing stock eliminations,

decreases, and increases make up 25%, 12%, and 13% of prior quarter holdings, respectively,

and new buys make up 33%, as reported in Table 4. Subtracting all sells from all buys results in

an average net increase (net buys) of about 9%. We also report these proportions separately for

booms, crises, and the subprime crisis. For booms and all crises, net buys are also around 11%

(8%) and drop to 7% during the subprime crisis. Still, buys outweigh sells in all time periods.

We also perform this analysis for various subsets of the data. Panel B presents results for

funds in the bottom quintile based on flows. For all periods, the difference between buys and

sells is about 5%, lower than for Panel A but still positive. During crises, we find more selling

activity, with net buys of about 2% for all crises and for the subprime crisis. In Panels C and D,

we perform this analysis for funds with the shortest lockup and redemption periods. In theory,

funds with shorter lockup and redemption periods are more likely to be forced into asset sales in

a crisis, while funds with longer lockup and redemption periods have more flexibility in their

7We follow Ben-David, Franzoni and Moussawi (2012), whose SAS program (“Institutional Trades, Flows, and Turnover Ratios using Thomson-Reuters 13F data”) is found at http:/wrds.wharton.upenn.edu/. We thank Rabih Moussawi for making this code available. In this method, shares held are adjusted for splits and distributions, and the quarterly holding snapshots are used to derive the trades. Holdings that originate from changes in the universe of 13F filers are filtered out. We require that hedge fund appear in two consecutive quarters, so if a hedge fund does not appear we eliminate the observation.

14

portfolio decisions.8 In Panel C we report data for the funds in the lowest lockup period quintile,

which have an average lockup period of about 2 days (for comparison, the average lockup period

for the highest quintile is about 22 months). Here, there is positive net buying during booms at

14%, but net buying during crises is still high at around 12%. Finally, in Panel D we report data

for funds in the lowest redemption period quintile, which averages about 51 days (for

comparison, the average redemption period for the highest quintile is about 9 months). For this

variable, there is much less net buying during crises, especially during the subprime crisis where

net buys are a negative -0.7% as compared to net buys during booms of about 9%.

Overall, based on univariate analyses of hedge fund portfolio changes, we conclude that

even the most constrained funds (those with low flows, short lockups, and/or short redemption

periods) are net buyers of stock, on average. However, during crises, we find that funds with

high outflows and short redemption periods feel pressure, as their net buying activity is

significantly lower than during all periods or during booms.

C. Buys and Sells at the Fund Level

We next consider measures of selling and buying pressure at the fund level, rather than

among all funds at a given point in time, using a multivariate analysis. Table 5 reports results

from regressions estimated separately for four samples: all periods, all crises, subprime crisis,

and all booms. We focus on funds with high and low flows and create indicator variables for the

lowest and highest flow quintiles (D1 and D5, respectively). The first three columns are for the

entire sample period while the other columns focus on crises (including the subprime crisis) and

booms. The regressions are pooled over time and across fund families, and standard errors are

clustered by time and hedge fund family. The dependent variable in the first column in each time

period is total sales. Sales are measured as the sum of existing stock eliminations and decreases 8See Aragon (2007).

15

as a proportion of prior period portfolio market value. In the second column of each time period,

the dependent variable is total purchases, which is the sum of purchases of existing and new

stocks as a proportion of prior period market value. The dependent variable in the third column is

the difference between the first two dependent variables (net sales). We include a number of fund

characteristics as controls: an indicator variable for leverage, the log of the size of the stock

portfolio, the fund’s quarterly return, the log of fund age, incentive fees measured as a percent of

profits, management fees measured as a percent of assets, the log of minimum required

investment, a high watermark dummy variable, a dummy variable for registered investment

adviser (RIA), and style dummies.

Focusing on the full sample period, we find that on a gross basis, funds both buy and sell

more stock in booms, but neither sell nor buy more stock in crises. However, a subset of funds –

those in the bottom quintile based on outflows – engage in more gross selling and more net

selling during the full sample period, which could be at the heart of fire sales, although the net

sales results are only significant at the 10% level. Further, funds with large inflows engage in

more buying and net buying than funds in the middle three flow quintiles. The inflow results are

larger and more significant than the outflow results. In addition, larger funds and funds with

better returns have lower net sales. Finally, there is weak evidence that funds with high minimum

investments sell more.9

While the regressions for all periods do not show increased net sales in a crisis, there is

evidence that funds with outflows sell more stock. If sales by these funds are high enough, this

selling activity could lead to declining stock prices and fire sales. We examine this hypothesis in

the set of regressions labeled “All Crises.” The results show that funds with the largest outflows

9 We also perform regressions where we interact the flow indicator variables with the crisis indicator variables and find insignificant coefficients on these interaction terms.

16

have significantly more stock sales than other funds during a crisis. And since they have fewer

buys, their net sales are significantly larger as well. Also, funds with large inflows continue to

buy significantly more stock. Older funds tend to sell more, both on a net and gross basis. To

examine economic significance, we set all continuous explanatory variables to their means for

the “all crises” periods, and set the dummy variables to 1 for high watermark, 1 for leverage, 0

for registered investment advisor, and 1 for long-short equity, since these values represent the

majority of the funds in the sample. We then compare the predicted net sales for high outflow

funds and high inflow funds. High outflow funds have predicted net sales of 3% of prior period

assets, while high inflow funds have predicted net purchases of 11% of prior period assets, so

this difference is economically as well as statistically significant (as noted in the last row of the

table, the differences in coefficients on high and low flow funds is statistically significant at the

0% level). We also note that the control variable coefficients and measures of statistical

significance are comparable to the values in the previous set of regressions for all periods.

For the recent subprime crisis of Q1 2007-Q2 2009, the results are similar to all crises in

that funds with large outflows sell more stock. The constant term in the net selling regression is

significantly positive during the most recent financial crisis and all crises, implying higher net

selling across the middle three flow quintiles as well. As with the previous regression, funds with

inflows purchase significantly more stock, and therefore have significantly lower net selling. We

perform the same test of economic significance as in the prior set of regressions, and find

predicted net sales of about 4% for high outflow funds compared to net purchases of about 10%

for high inflow funds, which again has strong economic significance.

The results thus far indicate that hedge funds with large outflows sell more stock during

the entire sample period, during all crises, and during the recent crisis period. The magnitude of

17

the coefficients on the large outflow indicator variable is higher for the subprime crisis (0.047)

than for all crises (0.045), and both these coefficients are higher than for all periods (0.033),

which suggests that sales may be more severe during crises. This finding is consistent with the

results of Table 4 that indicate that funds sell more stock during crises than during normal times.

However, the results of a chi-squared test indicate no significant differences among these three

coefficients, with p-values of 0.60 for the difference between all periods and all crises, 0.59 for

the difference between all crises and the subprime crisis, and 0.46 for the difference between all

periods and the subprime crisis. Further, we perform the same regression analysis for all periods

excluding crisis periods (not reported), and obtain a statistically significant coefficient of 0.029

on the large outflow indicator variable. The p-value from a chi-squared test that compares this

coefficient to all crises (the subprime crisis) is 0.59 (0.48).

Finally, the last set of regressions in Table 5 focuses on boom periods. Contrary to the

estimates for other time periods, net sales by funds in the bottom flow quintile are not larger in

booms. A chi-squared test comparing the coefficient on high outflow funds during booms to all

crises (the subprime crisis) results in an insignificant p-value of 0.49 (0.41). Funds in the high

flow quintile buy more stock, consistent with results for other periods. The size and performance

results that held for all periods and crisis periods are no longer significant, but fund age and

minimum investment are positively related to sales. We perform the same test for economic

significance as in the all crises and subprime crisis regressions, and find predicted net purchases

of 2% for low flow funds and 14% for high flow funds. Again, this difference is economically

significant, but in contrast to crisis periods, low flow funds are actually net buyers during booms.

As for all other regression specifications, the difference in coefficients on flow quintiles is large

and statistically significant for net sales.

18

Table 6 examines the effects of lock-up periods and redemption restrictions on gross

sales, gross purchases and net sales. The correlation between lockup and redemption periods is

positive, but not large at about 0.22. If we re-estimate the regressions separately including either

lockup period dummies or redemption restriction dummies, the results do not change from those

reported. Similar to the specifications in the prior table we report results for all periods, all crises,

subprime crisis, and all booms.10 As noted earlier, data for these variables are not available for

all funds so these regressions are estimated with a slightly reduced sample. As with the flow

variable, we create quintile indicators for redemption periods and lockups, and include the top

and bottom quintile indicators in the regressions. The net sales regression in the “All Periods”

column has coefficients on the crisis period and fund flow quintile variables that are very similar

to Table 5. That is, crises do not involve significantly more selling, ceteris paribus.

As in Table 5, funds with outflows sell more and funds with inflows buy more stock.

However, it does not appear that funds with outflows are forced into sales: the coefficients on the

indicator for short lockup periods implies that these funds actually sell less stock on a net basis,

although on a gross basis they both sell more and buy more stock. Further, while funds with

short redemption periods sell more stock on a net basis, consistent with the hypothesis of forced

sales, their gross sales are no different from other funds. The funds that should have the greatest

ability to avoid fire sales, those with long redemption periods, do have fewer forced sales, as

measured by the significant negative coefficient in the gross sales regression, but they also buy

significantly fewer stocks so that their net sales are comparable to those of other funds.11

10We also perform these analyses including the volatility of past fund flows as in Manconi, Massimo, and Yasuda (2012) and find that this measure does not predict sales, does not add to the explanatory power of the regression, and does not change the main results. For brevity and because its inclusion reduces our sample size, we omit it. 11We also perform regressions where we interact the lockup and redemption indicator variables with the crisis indicator variables and find insignificant coefficients on all these interaction terms.

19

The results in the crisis period regressions provide evidence that lockups and redemption

restrictions have little effect on forced stock sales. Funds that should suffer more in a crisis from

short lockup features tend to sell more stock, but they also buy more stock. Indeed, funds with

short lockups buy so much more stock in the two crisis periods that they have more net buying

than other funds. Nor is there evidence that the funds with short redemption periods are under

pressure to sell more assets than other funds. While they do sell more stock during the subprime

crisis on net, this result is driven by sharply lower buying, not higher selling. Moreover, these

funds also have greater net sales in booms, suggesting that the net selling is not prompted by

liquidity problems due to outflows. Likewise, the coefficients on the long lockup and redemption

period indicators are largely unrelated to gross or net selling during crises. For the control

variables, results are generally consistent with Table 5.12 Further, tests of economic significance

during all periods and all subperiods have very similar results to those performed in Table 5.

In summary, hedge funds with large outflows sell more stock than funds with large

inflows, regardless of the time period studied. While the regression results also suggest that

funds with large outflows sell more stock during crisis periods than boom periods or all periods,

the differences between time periods are not statistically significant. Evidence of differences in

selling behavior is weaker when considering other hedge fund constraints like leverage, lockups

and redemption periods. Leverage is never significantly related to buying or selling. The weak

results for lockup and redemption periods are perhaps not too surprising, as they are not direct

measures of fund liquidity. Also, lockup periods apply to new fund investors only, and

redemption periods can be altered when hedge funds erect gates. Funds with poor performance

12 We also perform a robustness test (not tabulated) using a logit regression in which the dependent variable is set to 1 if the firm has net sales during a quarter and zero otherwise. We find nearly identical results to the regressions presented in Tables 5 and 6.

20

also tend to sell more stock, consistent with the positive correlation between poor performance

and outflows reported in Table 3.

Our result that funds with outflows sell more stock is consistent with Ben-David,

Franzoni, and Moussawi (2012), who attribute greater selling pressure on hedge funds to larger

than normal outflows. However, our results do not necessarily imply that these funds are forced

to sell stocks. First, we show that funds with outflows use most of the proceeds of stock sales to

buy more stock. If they were under pressure to sell stocks that they would rather hold, they

would not likely voluntarily sell more than necessary. Furthermore, the coefficients on the

indicator variable for a crisis are not significant in our gross or net sales regressions, suggesting

that the sales in a crisis happen for the same reasons they occur in normal times.

IV. Tests of Fire Sales

Our results on flows and sales suggest that hedge fund sales may put downward pressure

on stock prices in a crisis. We examine price pressure around forced stock sales (and purchases)

on a stock-by-stock basis using the approach of Coval and Stafford (2007). They consider a sale

(purchase) to be forced if the stock is widely sold (bought) by a number of managers

experiencing severe outflows from (inflows to) their funds. We create their variable PRESSURE

by summing the difference between flow-induced purchases and sales each quarter and dividing

the difference by the average trading volume of the stocks from the prior two quarters. Then,

severe flows are those below/above the 10th/90th percentile of quarterly fund flow, where flow is

calculated as the dollar flow divided by prior period assets:

𝑃𝑅𝐸𝑆𝑆𝑈𝑅𝐸!,! =

(!"# !,∆!"#$%&'(!,!,! !"#$!,!!!"#$"%&'(" !"!! ! !"# !,∆!"#$%&'(!,!,! !"#$!,!!!"#$"%&'(" !"!!!!

!"#$%&'()!,!!!":!!! (1)

21

where i is the individual stock, j is the fund, and t is the quarter. We sort stocks based on the

PRESSURE variable, and define those falling in the 10th percentile as pressure sale stocks and

those in the 90th percentile as pressure purchase stocks. Since this definition of PRESSURE relies

on frequent trading of the same stock by different managers, we require a stock be owned by at

least 10 managers during the quarter to calculate the PRESSURE measure.13

While it is difficult to disentangle price pressure from information effects, this approach

focuses on forced selling of the same stock by many funds and should help differentiate the two

effects. If information effects cause widespread sales of stock, returns should drop during the

selling period and not rebound afterwards. By contrast, if price pressure is the cause of selling,

large capital outflows that induce several funds to sell the same stock at the same time should

result in an initial price drop followed by a subsequent increase to compensate the liquidity

providers. Thus, we construct monthly cumulative abnormal returns for stocks that are widely

sold by subtracting from their returns the equally-weighted return of all stocks held by all hedge

funds – except the stock of interest – in a given month. We control for cross-sectional

dependence in monthly returns with Fama and MacBeth (1973) portfolios (i.e., we calculate

average abnormal returns each month and use the time series of mean abnormal returns). Finally,

as in Coval and Stafford (2007), we require the existence of at least 25 hedge funds in each

quarter for a stock to be included (which does not exclude any quarters in our sample).

Using this approach, we find too few pressure sales and purchases for a meaningful

analysis for all periods prior to 2001. Therefore, the pressure sale analysis covers the time period

2001-2010. Figure 1 graphs the number of pressure sale and purchase stocks by quarter for this

period. Stocks that are neither sold nor purchased average 4,189 separate stocks, which is about

13 For the remainder of the paper, we interchangeably use the terms “pressure sale (purchase)” and “forced sale (purchase)”.

22

98% of all stocks held by hedge funds each quarter. By contrast, pressure sale and purchase

stocks average 40 and 41 stocks, respectively – or about 1% each – per quarter. Pressure sale

stocks as a proportion of all stocks is highest in June 2007, at 1.9%, but then drops significantly

afterward, and pressure purchases follow a similar pattern.

Cumulative abnormal returns for pressure sales and purchases are presented in Table 7

and graphed in Figure 2. Figure 2 shows monthly data while Table 7 aggregates the data into

various periods. In Figure 2, we present cumulative abnormal returns for the 12 months prior to

the 18 months following the quarter of sale. Since we have quarterly data on sales and monthly

data on returns, and do not know exactly when during the quarter that the sale occurs, we label

the three months during the quarter of sale as T1, T2, and T3 respectively. All other months are

labeled relative to this quarter. Examining Figure 2 and focusing on pressure sales, there is no

obvious pattern of fire sales in either the whole period or any subperiod, since returns do not

decrease during the period of selling and increase subsequently. In stark contrast to Coval and

Stafford’s (2007) findings for mutual funds, the returns on the stocks that are sold during all

periods and crises are large and significantly positive prior to the sale, but do not change much

after the sale takes place. Further, for all periods, returns never even reach negative territory

before, during, or after the selling activity, hardly an indication of fire sales. The similarity in

patterns of pressure sale and pressure purchase stocks for all periods and the crisis periods is

striking. Only during booms do the returns on pressure sale and pressure purchase stocks diverge,

with pressure sales drifting slightly upward after the sale (although the sale itself does not affect

the price) and pressure purchase stocks drifting downward.

The data behind these graphical results are reported in Table 7, where we aggregate the

returns over various subperiods and perform t-tests for statistical significance. There is no

23

evidence of fire sales during any period: while stock prices do drop during the selling period,

these changes are not statistically significant, and there is not a significant price rebound after the

sale period; in fact, for the crisis subperiods, the performance of pressure sale stocks continues to

drift downward. There is also no consistent evidence of temporary increases in stock prices due

to pressure purchases – while the general pattern of the data does show that prices rise when

stocks are bought and fall afterward, none of these changes is statistically significant.

We perform several additional (untabulated) robustness tests concerning price pressure.

First, we perform tests using Coval and Stafford’s (2007) alternate pressure measures with very

similar results. Next, we sort stocks into quintiles by proxies other than fund flows, and create

the pressure variable based on these proxies. These proxies include lockup and redemption

periods, fund performance, leverage, fund performance interacted with leverage, required

minimum investment, and fund size. In each case, the results are consistent with those in Table 7

in that they provide no evidence of fire sales as a result of hedge fund selling.

V. Characteristics of Stocks Sold (and Purchased) by Hedge Funds

While our results in section III indicate that hedge funds sell more stock in crises in

response to outflows, these sales do not appear to cause downward price pressure on the stocks

as shown in Section IV. This could arise from several factors, including the fact that hedge funds

do not herd to the same extent as mutual funds and thus do not engage in massive selling of the

same stocks. Hedge funds can erect gates when redemptions become overwhelming. Or, hedge

funds may avoid extreme sales of the same stocks by increasing their cash holdings as trouble

appears more likely, as suggested by Liu and Mello (2011). In fact, in September 2008,

Citigroup’s prime brokerage division published a report estimating hedge fund cash holdings at

24

20% prior to the crisis and 30% during the third quarter of 2008.14 And in November, 2008,

Barton Biggs, the manager of Traxis Partners, in an article in Fortune, noted two reasons why he

was not worried about hedge fund withdrawals igniting a market selloff: first, hedge funds have

losses to make up before they can begin to earn their performance fees, so that investors pulling

out during this period would be forgoing a free ride with no incentive fees, and second, he states

that: “It is a fact that hedge fund margin debt has been declining since July 2007, and prime

brokers report massive hedge fund cash holdings.”15 Hence, there appear to be many ways that

hedge funds can avoid fire sales – even during crises – and avoid selling en masse.

Further, it is possible that when hedge funds sell the same stocks they simply do not sell

enough stock to move prices. We examine this idea in Figure 3, which graphs, by quarter, the

percentage of stock market value (Panel A) and quarterly volume (Panel B) that the average

pressure sale, purchase, and other holding comprises. Untabulated results indicate that pressure

sales (purchases) comprise an average of 1.5% (2.5%) of market capitalization and 3.1% (5.9%)

of quarterly volume, which is fairly substantial. Further, pressure sales are even higher during the

crisis at 7.1% of volume in December 2008 and 5% in March 2009. This indicates that, although

there are few pressure sales and purchases, those that do occur represent a fairly substantial

proportion of these stocks’ volume and market capitalization, especially during the recent crisis.

However, despite this result, the results of the prior section provide strong evidence that

these stocks are not sold at fire sale prices. Perhaps these stocks have characteristics that make

them better suited to withstand such pressure. Boyson, Helwege, and Jindra (2012a) note in their

study of commercial banks that firms have several ways to avoid fire sales when faced with

selling pressure. In particular, they note that banks can avoid the sale of toxic assets and instead

14 See “Hedge Fund Cash Holdings at 30%,” FinAlternatives article: http://www.finalternatives.com/node/5510 . 15 See http://money.cnn.com/2008/11/10/news/economy/hedgefunds_Biggs.fortune/index.htm.

25

sell off assets that have retained value in the crisis (cherry picking). Likewise, when faced with

selling pressure, hedge funds can also cherry pick the stocks they do sell. We hypothesize that

hedge funds will sell past good performers and those with strong liquidity and/or large size. We

investigate this hypothesis in Tables 8 and 9.

Table 8 presents medians of stock and firm characteristics of pressure sale (purchase)

stocks during all periods, all crisis periods, the subprime crisis, and booms.16 Column 1 also

presents medians for the entire universe of CRSP stocks as a comparison. Focusing first on all

periods, pressure sale stocks have significantly higher prior period returns than other stocks held

by hedge funds, at 12.6% versus 3.3% respectively. The typical firm size is about five times

higher for pressure sale stocks based on either total assets or market capitalization. Tobin’s Q is

also significantly higher for pressure sales by 0.40. For liquidity, pressure sale stocks trade far

more often (trading volume is 69 million shares versus only 14 million for the comparison

group). Also, the Amihud liquidity measure and the bid-ask spreads are significantly lower for

pressure stocks than the comparison group.

Comparing pressure sales to pressure purchases, we find that while a few characteristics

of pressure sale stocks differ from pressure purchase stocks in the overall sample period, their

characteristics remain much closer to each other than to other stocks owned by hedge funds. For

example, pressure sale stocks have worse performance then pressure purchase stocks by 2% per

year, much smaller than the 9% difference between pressure sale stocks and all other stocks held

by hedge funds. Pressure sale stocks are about the same size as pressure purchase stocks and

have somewhat higher volume. They also have worse bid-ask spreads and Amihud liquidity

measures, but these differences are far from being economically significant.

16We present medians here since, as previously noted, means are often skewed due to outliers. Results using means are similar in direction and significance.

26

Turning to results for crisis periods, the differences in performance and other

characteristics between pressure sale stocks and the comparison group are even more compelling.

One notable difference in crises is the magnitude of the returns in the previous year. Not only do

hedge funds sell stocks with better past performance in a crisis, but the difference in the returns

is sharply higher, at 12.6% and 12.9% more than the other stocks’ prior year returns for all crises

and the subprime crisis, respectively. As with the overall period, the differences between

pressure sale and pressure purchase stocks are sometimes statistically significant but

economically small, and both pressure purchase and pressure sale stocks are larger and more

liquid than hedge funds’ other stockholdings. During boom periods, the differences between

pressure sale and all other stocks remain strong and large, while differences between pressure

purchase and pressure sale stocks are even less significant.

These univariate results suggest that the absence of fire sales for hedge funds in crises

owes to their selective sales of highly liquid assets. We examine this hypothesis in greater detail

with a logit regression in Table 9. The dependent variable is set to 1 if the stock in question is a

pressure purchase (sale) stock and zero for stocks that are neither pressure purchase nor pressure

sale stocks. Hence, pressure sale stocks are compared to stocks that are not pressure purchases

and the pressure purchase stocks are compared to the other stocks that are not pressure sales. The

logit regressions include time indicators, and standard errors are clustered at the time and stock

level. We exclude the log of the market value and Amihud measure from the regressions since

they are highly correlated with the log of assets and bid-ask spread, respectively.

For all periods and all subperiods, the logit results indicate that pressure sale stocks are

larger with better past performance, higher volume and better liquidity than other stocks. The

results are similar when comparing pressure purchase stocks to other stocks and are consistent

27

with the univariate results in Table 8. These results are highly economically significant as well as

significantly significant. For example, for all periods, the marginal probability of a stock being

pressure sold increases from 0.8% to 1% (a change of 25%) if the stock’s prior period annual

return moves from the 25th percentile to the 75th percentile. Similarly, the marginal probability

goes up by 33% (from 0.6% to 0.9%) as size moves from the 25th to the 75th percentile, up by

233% (from 0.3% to 1%) as volume moves from the 25th to the 75th percentile, and down by 30%

(from 1% to 0.7%) as the bid-ask spread goes from the 25th to the 75th percentile. Results are

even stronger for the recent crisis, with changes in marginal probabilities of 25%, 100%, 500%,

and -20% for moves from the 25th to 75th percentile for annual return, total assets, volume, and

bid-ask spreads, respectively. Panel B of Table 9 compares pressure buy stocks to pressure sell

stocks and omits the stocks that are not heavily traded. The only consistent statistically

significant result in these regressions is that pressure sale stocks have higher volume than

pressure purchase stocks. This result is also economically significant: for example, during the

recent crisis, the probability of a pressure sale goes from 44% to 58% (a 32% increase) for a

volume move from the 25th to the 75th percentile.

Taken together, the results in Tables 8 and 9 provide strong evidence that when hedge

funds are forced to sell or purchase stocks they largely avoid fire sales. The stocks they do sell

when under outflow pressure are similar to stock they buy when under inflow pressure, and both

types of trades reflect a propensity to cherry pick the stocks likely to have the smallest price

impact upon purchase or sale. These results are consistent with the earlier finding that

redemptions in crises do not force hedge funds into fire sales. Additionally, even funds with

significant selling pressure during the recent financial crisis had well-diversified stock portfolios

that allowed them to strategically sell off positions. While some hedge funds likely did not have

28

this ability to cherry pick assets, the variety of stocks sold in crises and their generally high

liquidity leads us to conclude that hedge fund pressures did not result in detectable fire sale

activity at any time in our sample period.

VI. Conclusion

A commonly held view of hedge funds in the subprime crisis describes them as victims

of severe outflows and margin calls that forced them to sell assets into a declining market.

Consistent with this portrait of a fragile industry, Ben-David, Franzoni and Moussawi (2012)

find that hedge funds had significant stock sales during the worst quarters of the recent crisis.

With such a shortage of buyers, their sales alone might explain the precipitous drop in the stock

market from 2008 to 2009. By contrast, Liu and Mello (2011) argue that hedge funds’ fragile

capital structures lead hedge funds to better plan for expected crises, by reducing leverage,

increasing cash positions, and holding more liquid assets.

While many of the stylized facts suggest that the financial crisis was exacerbated by

forced sales of stocks by hedge funds, our analysis suggests otherwise. While hedge funds

experiencing outflows sell more stock than funds experiencing inflows, and outflows are larger

during crises, we find no evidence that these stock sales occur at fire sale prices. Furthermore,

hedge funds with high flows, short lockup and short redemption periods, which should be the

most vulnerable funds, do sell stock in response to outflows, but they use most of the proceeds to

buy more stock. We also show that hedge funds hold well-diversified portfolios and rarely sell

the same stocks at the same time. This evidence calls into question the extent to which hedge

funds are forced to sell into a falling market.

We explicitly test for fire sales among hedge funds and find that stocks under the most

pressure from widespread hedge fund sales do not lose value in the quarter in which they are sold.

29

Nor are there price reversals one would expect when excess selling of stocks drives their prices

below their fundamental values. In sum, we find no evidence that hedge funds engage in fire

sales during the subprime crisis or any other time during our 1998-2010 sample period. Our

results also hold for pressure buys, which show no evidence of price increases when hedge funds

are forced to buy them.

We attribute the lack of price pressure from hedge fund trading to two factors. First,

hedge funds rarely sell the same stocks during the same quarters. Furthermore, when hedge funds

do sell large quantities of stock, they sell the best performing, largest, and most liquid stocks in

their portfolios. This cherry picking strategy minimizes the downward price pressure on stock

prices when they are forced to sell. We infer that these funds own sufficiently well-diversified

portfolios in that their trading activity does not lead to a significant impact on the stock price,

and therefore the pressure sales and purchases and not followed by price reversals when the

pressure ends.

30

References

Adrian, T. and H. S. Shin. 2008. Liquidity and Financial Cycles. Working paper.

Adrian, T. and H.S. Shin. 2010. Liquidity and Leverage. Journal of Financial Intermediation 19:418-437.

Agarwal, V., N.D. Daniel, and N.Y. Naik. 2009. Role of Managerial Incentives and Discretion in Hedge Fund Performance. Journal of Finance, 64:2221-2256.

Allen, F. and E. Carletti. 2006. Credit Risk Transfer and Contagion. Journal of Monetary Economics 53:89-111.

Ambrose, B. W., K. Cai, and J. Helwege. 2012. Fallen Angels and Price Pressure. Journal of Fixed Income 21:74-86.

Amihud, Y. 2002. Illiquidity and Stock Returns: Cross-Section and Time-Series Effects. Journal of Financial Markets 5:31-56.

Anand, A., P. J. Irvine, A. Puckett, and K. Venkataraman. 2012. Institutional Trading and Stock Resiliency: Evidence from the 2007-2009 Financial Crisis. forthcoming Journal of Financial Economics.

Ang, A., S. Gorovvy, and G. van Inwegan. 2011. Hedge Fund Leverage. Journal of Financial Economics 102:102-126.

Aragon, G. O. 2007. Share Restrictions and Asset Pricing: Evidence from the Hedge Fund Industry. Journal of Financial Economics 83:33-58.

Aragon, G. O. and P. E. Strahan. 2012. Hedge Funds as Liquidity Providers: Evidence from the Lehman Bankruptcy. Journal of Financial Economics 103:570-587.

Baquero, G. and M. Verbeek. 2009. A Portrait of Hedge Fund Investors: Flows, Performance, and Smart Money. Working paper, Erasmus University.

Ben-David, I., F. Franzoni, and R. Moussawi. 2012. Hedge Fund Stock Trading in the Financial Crisis of 2007-2009. Review of Financial Studies 25:1-54.

Billio, M., M. Getmansky, and L. Pelizzon. 2010. Crises and Hedge Fund Risk. Working paper.

Boyson, N. M., J. Helwege, and J. Jindra. 2012a. Liquidity Shocks, Crises and Fire Sales at Commercial Banks. Working paper.

Boyson, N. M., J. Helwege, and J. Jindra. 2012b. Liquidity Shocks, Crises and Fire Sales at Investment Banks. Working paper.

Boyson, N. M., C. W. Stahel, and R. M. Stulz. 2010. Hedge Fund Contagion and Liquidity Shocks. Journal of Finance 65:1789-1816.

Brown, G., J. Green, and J. Hand. 2010. Are Hedge Funds Systemically Important. Working Paper.

31

Brunnermeier, M. K. 2009. Deciphering the 2007-2008 Liquidity and Credit Crunch. Journal of Economic Perspectives 23:77-100.

Brunnermeier, M. K. and L. H. Pedersen. 2009. Market Liquidity and Funding Liquidity. Review of Financial Studies 22:2201-2238.

Cao, C., Y. Chen, B. Liang, and A. Lo. 2010. Can Hedge Funds Time Market Liquidity? Working paper.

Chen, J., S. Hanson, H. Hong, and J. Stein. 2008. Do Hedge Funds Profit from Mutual Fund Distress? Working paper.

Coval, J. and E. Stafford. 2007. Asset Fire Sales (and Purchases) in Equity Markets. Journal of Financial Economics 86:479-512.

Diamond, D. and R. Rajan. 2010. Fear of Fire Sales, Illiquidity and Credit Freezes. Quarterly Journal of Economics 126:557-591.

Ding, B., M. Getmansky, B. Liang, and R. Wermers. 2009. Share Restrictions and Investors Flows in the Hedge Fund Industry. Working paper.

Fama, E. and J. MacBeth. 1973. Risk, Return, and Equilibrium: Empirical Tests. Journal of Political Economy 81:607-636.

Geanakoplos, J. 2010. The Leverage Cycle. In D. Acemoglu, K. Rogoff, and M. Woodford, eds., NBER Macroeconomics Annual 2009, 24, 1-65.

Griffin, J. M. and J. Xu. 2009. How Smart are the Smart Guys? A Unique View from Hedge Fund Stock Holdings. Review of Financial Studies 22:2531–2570.

Gromb, D. and D. Vayanos. 2002. Equilibrium and Welfare in Markets with Financially Constrained Arbitrageurs. Journal of Financial Economics 66:361-407.

Hameed, A., W. Kang, and S. Viswanathan. 2010. Stock Market Declines and Liquidity. Journal of Finance 65:257-293..

He, Z., I. G. Khang, and A. Krishnamurthy. 2010. Balance Sheet Adjustments in the 2008 Crisis. IMF Review 58:118-156.

Khandani, A.E. and A.W. Lo. 2007. What Happened to the Quants in August 2007? Journal of Investment Management 5:5-54.

Krishnamurthy, A. 2009. Amplification Mechanisms in Liquidity Crises. American Economic Journal 2:1-33.

Liu, X. and A. Mello. 2011. The Fragile Capital Structure of Hedge Funds and the Limits to Arbitrage. Journal of Financial Economics 102(3): 491-506.

Maier, T., N. Schaub, and M. Schmid. 2011. Hedge Fund Liquidity and Performance: Evidence from Financial Crises. Working paper.

32

Manconi, A., M. Massimo, and A.Yasuda 2012. The Role of Institutional Investors in Propagating the Crisis of 2007-2008. forthcoming Journal of Financial Economics.

Massa, M., A. Simonov, and S. Yan. 2012. Contrarian Hedge Funds and Momentum Mutual Funds. Working paper.

Sadka, R. 2010. Liquidity Risk and the Cross-Section of Hedge-Fund Returns. Journal of Financial Economics 98:54-71.

Shadab, Houman D. 2009. Hedge Funds and the Financial Crisis. Mercatus on Policy, No. 24.

Shleifer, A. and R. W. Vishny. 1992. Liquidation Values and Debt Capacity: A Market Equilibrium Approach. Journal of Finance 47:1343-1366.

Shleifer, A. and R. W. Vishny. 2011. Fire Sales and Finance and Macroeconomics. Journal of Economic Perspectives 25:29-48.

Uhlig, H. 2010. A Model of a Systemic Bank Run. Journal of Monetary Economics 57:78-96.

Zhang, Hanjiang. 2009. Asset Fire Sales, Liquidity Provision, and Mutual Fund Performance. Working paper, University of Texas at Austin.

33

Table 1: Summary Statistics Panel A presents summary statistics at the hedge fund family level for funds that owned stocks during the period 1998 to 2010. Panel B presents additional portfolio, by year, at the hedge fund family level. Panel C presents stock portfolio statistics at the hedge fund family level. Panel D presents individual stock characteristics at the individual stock level for stocks traded by hedge funds.

Panel A: Quarterly Fund Family Characteristics at the Fund Family Level N Mean Median Q1 Q3 Continuous Variables Size ($ million) 4,973 743 270 91 748 Min. required investment ($ million) 4,953 2.0 1.0 0.55 1.50 Fund family age in years 4,973 10 9 7 12 Lockup period in months 4,973 6 4 0 12 Total redemption period in days 4,857 156 135 111 163 Quarterly flow as a % of AUM 4,973 0.009 0.008 -0.048 0.066 Quarterly flow volatility 4,507 0.108 0.050 0.030 0.088 Quarterly return 4,973 0.014 0.013 -0.015 0.045 Incentive fee (% of assets) 4,973 0.19 0.20 0.20 0.20 Management fee (% of profits) 4,973 0.014 0.013 0.010 0.015 Indicator Variables Manager is a registered inv. advisor? 4,973 0.14 Fund uses leverage? 4,973 0.62 Fund has high water mark? 4,973 0.78 Fund is convertible arbitrage style? 4,973 0.06 Fund is dedicated short bias style? 4,973 0.00 Fund is event driven style? 4,973 0.18 Fund is equity market neutral style? 4,973 0.00 Fund is long-short equity style? 4,973 0.52 Fund is global macro style? 4,973 0.02 Fund is managed futures style? 4,973 0.01 Fund is fixed income arbitrage style? 4,973 0.01

Panel B: Fund Family Level, by Year

Year

Average Number of

Funds

Mean Fund Size

Equity Portfolio

(13F) $ Million

Number of Stocks Per

Manager

Avg. Quarterly Change in HF Holdings, (% Share of Equity

Holdings) Number Mean Mean Mean Median Mean Median

1998 36 603 683 133 72 0.091 0.032 1999 46 461 757 118 63 0.118 0.055 2000 60 488 797 143 76 0.070 0.026 2001 80 500 672 137 68 0.133 0.076 2002 91 569 557 138 68 0.086 0.054 2003 105 588 655 143 73 0.127 0.067 2004 124 773 792 155 69 0.108 0.055 2005 140 810 771 150 65 0.094 0.046 2006 147 939 796 152 60 0.076 0.029 2007 141 967 905 157 60 0.096 0.048 2008 116 878 621 132 49 0.017 -0.010 2009 86 687 553 136 50 0.114 0.045 2010 72 752 726 134 63 0.104 0.053

34

Table 1: Summary Statistics, Continued

Panel C: Stock Portfolio Characteristics at the Fund Family Level

N Mean Median Q1 Q3 Market value of stock portfolio ($MM) 4,973 726 329 137 769 Number of stocks per filing 4,973 144 64 34 120 Hedge fund holding/Stock market cap 4,973 0.009 0.005 0.002 0.010 Transaction size ($ million) 4,964 5.62 3.11 1.34 6.40 Quarterly change in HF holdings 4,973 0.093 0.044 -0.085 0.225

Panel D: Stock Characteristics at the Stock Level

N Mean Median Q1 Q3 Prior year annual return 155,069 0.082 0.035 -0.238 0.318 Return standard deviation 155,069 0.130 0.108 0.070 0.168 Total assets ($ million) 155,069 3,433 633 125 2,683 Market Capitalization ($ billion) 155,069 1,980 551 185 1,790 Tobin’s Q 155,069 2.30 1.63 1.05 2,96 Quarterly volume (millions of shares) 155,069 43 15 4 46 Amihud Illiquidity Measure * 108 155,069 0.972 0.058 0.011 0.387 (Bid-Ask)/Bid (Daily) * 102 155,069 0.063 0.027 0.013 0.071

35

Table 2: Crisis and Boom Periods

This table reports the periods of Booms and Crisis as determined by in-sample distribution of the following indicators: NBER recessions and expansions; bank failures (normalized by the contemporaneous number of banks); TED spread; Moody's AAA-Baa credit spreads; flight to quality indicator (Collin-Dufresne, Goldstein, and Helwege (2009)); the Long Term Capital Management (“LTCM”) episode; stock market declines and increases; and credit crunch periods (Bordo and Haubrich (2009)).

Periods Dates Notes Crisis 8/1998 - 1/1999 - High TED spread

- Flight to quality - LTCM episode - Low stock market returns

3/2001 - 11/2001 - NBER contraction (3/2001 - 11/2001) - Flight to quality - Low stock market returns

3/2007 - 06/2009 - NBER contraction - High bank failures - High TED and credit spreads - Flight to quality - Credit crunch - Low stock market returns

Boom 6/1999 - 5/2000 - NBER expansion

- High stock market returns

1/2003 - 2/2004 - NBER expansion - Low credit spreads - High stock market returns

36

Table 3: Fund Flows

Funds are divided into quintiles each quarter, based on flows. Panel A reports average fund flows, and Panel B reports average lagged quarterly performance by flow quintile. Panel C reports the proportion of funds that have outflows (e.g. negative flows), both in the aggregate and by fund style. Both panels report flows for all periods, boom, crises, and the subprime crisis. t-tests for differences in means between the subprime crisis and booms and for all crises and booms are performed for all analyses, and Panels A and B include tests for differences between flow quintiles.

Panel A: Flows by Fund Flow Quintile

Quintile N Overall Boom Crisis Subprime

Crisis Diff: Sub Crisis - Boom

Diff: Crisis-Boom

1 (worst) 1,000 -0.149 -0.135 -0.158 -0.170 -0.035*** -0.012***

2 1,019 -0.038 -0.024 -0.054 -0.061 -0.037*** -0.007*** 3 994 0.008 0.013 0.000 -0.008 -0.021*** -0.008*** 4 1,006 0.056 0.058 0.048 0.039 -0.019*** -0.009***

5 (best) 954 0.177 0.184 0.156 0.148 -0.036*** -0.008*** Diff (1-5) NA -0.326*** -0.320*** 0.315*** -0.318*** NA NA

Panel B: Quarterly Lagged Fund Performance by Flow Quintile

Quintile N Overall Boom Crisis Subprime

Crisis Diff: Sub Crisis - Boom

Diff: Crisis-Boom

1 (worst) 958 -0.003 0.019 -0.018 -0.012 -0.031*** -0.037*** 2 981 0.012 0.042 -0.011 -0.014 -0.056*** -0.053*** 3 953 0.016 0.036 -0.007 -0.006 -0.042*** -0.043*** 4 958 0.020 0.040 -0.001 0.003 -0.037*** -0.041***

5 (best) 867 0.030 0.041 0.011 0.010 -0.031*** -0.030*** Diff (1-5) NA -0.033*** 0.022*** 0.030*** 0.023*** NA NA

Panel C: Proportion of Funds with Outflows

Quintile N Overall Boom Crisis Subprime

Crisis Diff: Sub Crisis - Boom

Diff: Crisis-Boom

All Funds 4,973 0.445 0.424 0.479 0.506 0.082*** 0.055*** By Style Global Macro 63 0.317 0.333 0.368 0.294 -0.039 0.035 Conv. Arbitrage 248 0.440 0.415 0.370 0.364 -0.051 -0.045 Emerging Markets 104 0.394 0.278 0.426 0.452 0.174* 0.148 Event Driven 849 0.393 0.378 0.444 0.505 0.127*** 0.066* Multi-Strategy 902 0.453 0.439 0.449 0.469 0.030 0.010 Fund of Funds 182 0.500 0.360 0.491 0.477 0.117 0.131 Long-Short 2,528 0.456 0.446 0.508 0.539 0.093*** 0.062** Managed Futures 62 0.597 0.364 0.727 0.696 0.332** 0.363** Fixed Inc. Arbitrage 43 0.512 0.500 0.857 1.000 0.500** 0.357*

37

Table 4: Portfolio Changes by Period and Measures of Pressure

This table reports mean portfolio changes as measured using the Coval and Stafford (2007) approach and the Ben-David et. al. (2012) approach as detailed in Section III. Panel A reports data for all funds, Panel B reports data for funds in the bottom quintile of flows, Panel C reports data for funds in the bottom tercile of lockup period, and Panel D reports data for funds in the bottom tercile of redemption period. All Panels report data for all periods, booms, all crises, and the subprime crisis separately.

Panel A: All Funds

All Periods

Booms

Crises

Subprime Crisis

Portfolio Changes (Coval and Stafford) Eliminate as % of prior period shares 0.378 0.371 0.392 0.390 Decrease as % of prior period shares 0.226 0.233 0.228 0.234 Increase as % of prior period shares 0.239 0.233 0.236 0.239 Maintain as % of prior period shares 0.157 0.163 0.144 0.137 Portfolio Changes (Ben-David et. al.) Eliminate as a % of prior period MV 0.249 0.257 0.243 0.242 Decrease as a % of prior period MV 0.117 0.126 0.117 0.120 Increase existing % of prior period MV 0.129 0.131 0.125 0.126 Buy new as % of prior period MV 0.330 0.365 0.310 0.304 Net Buys 0.093 0.113 0.075 0.068

Panel B – Only Funds in Bottom Flow Quintile

All Periods

Booms

Crises Subprime

Crisis Portfolio Changes (Coval and Stafford)

Eliminate as % of prior period shares 0.420 0.398 0.435 0.443 Decrease as % of prior period shares 0.236 0.261 0.238 0.240 Increase as % of prior period shares 0.211 0.208 0.210 0.208 Maintain as % of prior period shares 0.134 0.133 0.117 0.109 Portfolio Changes (Ben-David et. al.) Eliminate as a % of prior period MV 0.282 0.270 0.274 0.276 Decrease as a % of prior period MV 0.129 0.132 0.133 0.136 Increase existing % of prior period MV 0.116 0.115 0.118 0.118 Buy new as % of prior period MV 0.340 0.366 0.312 0.314 Net Buys 0.045 0.079 0.023 0.020

38

Table 4: Portfolio Changes by Period and Measures of Pressure, continued

Panel C – Only Funds in Bottom Lockup Period Quintile

All Periods

Booms

Crises

Subprime Crisis

Portfolio Changes (Coval and Stafford)

Eliminate as % of prior period shares 0.414 0.389 0.442 0.449 Decrease as % of prior period shares 0.215 0.231 0.208 0.211 Increase as % of prior period shares 0.226 0.227 0.222 0.223 Maintain as % of prior period shares 0.145 0.153 0.128 0.117 Portfolio Changes (Ben-David et. al.) Eliminate as a % of prior period MV 0.271 0.266 0.270 0.276 Decrease as a % of prior period MV 0.118 0.126 0.118 0.122 Increase existing % of prior period MV 0.136 0.140 0.134 0.134 Buy new as % of prior period MV 0.369 0.394 0.374 0.371 Net Buys 0.116 0.143 0.119 0.107

Panel D – Only Funds in Bottom Redemption Period Quintile

All Periods

Booms

Crises Subprime

Crisis Portfolio Changes (Coval and Stafford)

Eliminate as % of prior period shares 0.371 0.356 0.375 0.378 Decrease as % of prior period shares 0.235 0.211 0.241 0.244 Increase as % of prior period shares 0.257 0.251 0.263 0.266 Maintain as % of prior period shares 0.138 0.181 0.121 0.111 Portfolio Changes (Ben-David et. al.) Eliminate as a % of prior period MV 0.247 0.262 0.242 0.253 Decrease as a % of prior period MV 0.120 0.120 0.121 0.121 Increase existing % of prior period MV 0.128 0.128 0.125 0.124 Buy new as % of prior period MV 0.312 0.340 0.270 0.243 Net Buys 0.073 0.086 0.032 -0.007

39

Table 5 Buying and Selling Activity as a Function of Flows, Leverage and Performance

This table performs OLS regression analyses examining quarterly sells, buys, and net sells (sells-buys) of stock at the hedge fund family level. The independent variables are sales as a proportion of market value, buys as a proportion of market value, and net sells as a proportion of market value. Regressions are performed for the entire sample period (1998-2010), all crisis periods, the recent crisis period, and all boom periods. All regressions include style dummies. Independent variables are described in Section III. Standard errors are clustered at the family and time levels. Statistical significance is denoted with * ,** , and *** at the 1%, 5%, and 10% levels, respectively.

All Periods (1998-2010) All Crises Subprime Crisis (Q107-Q209) All Booms Sells Buys Net Sells Sells Buys Net Sells Sells Buys Net Sells Sells Buys Net Sells Constant 0.918*** 0.788*** 0.131 0.868*** 0.607*** 0.261* 0.904*** 0.587*** 0.317* 0.909*** 0.977*** -0.068 Boom Indicator 0.013* 0.030* -0.016 Crisis Indicator -0.001 -0.015 0.013 Flow Quintile Indicators Bottom Flow Quintile (D1) 0.036*** 0.002 0.033* 0.037*** -0.009 0.045*** 0.041*** -0.006 0.047*** 0.009 -0.001 0.010 Top Flow Quintile (D5) -0.004 0.082*** -0.086*** -0.011 0.084*** -0.095*** -0.006 0.085*** -0.091*** 0.005 0.115*** -0.110*** Control Variables Leverage Dummy -0.032 -0.024 -0.008 -0.031 -0.033 0.002 -0.025 -0.036 0.012 -0.027 -0.036 0.009 Log Stock Portfolio Size -0.037*** -0.017** -0.020*** -0.036*** -0.006 -0.030*** -0.038*** -0.009 -0.029*** -0.036*** -0.023 -0.013 Fund Quarterly Return 0.195*** 0.419*** -0.223** 0.207*** 0.534*** -0.327*** 0.161*** 0.529*** -0.367*** 0.200 0.241 -0.041 Log of Fund Age 0.036** 0.016 0.020 0.030** -0.005 0.035** 0.028* -0.008 0.036** 0.055*** 0.014 0.041*** Incentive Fee (% Profits) -0.036 0.248 -0.284 0.458 0.479 -0.021 0.803** 1.034*** -0.231 -0.496* -0.103 -0.393 Mgmt. Fee (% AUM) 7.248*** 6.473** 0.775 7.919*** 7.018*** 0.901 7.917*** 8.480*** -0.564 8.968** 13.315*** -4.347* Log of Min. Investment 0.007 -0.007 0.014** 0.003 -0.007 0.010 0.000 -0.007 0.008 0.006 -0.020 0.026** High Watermark Dummy -0.032 -0.036 0.004 -0.045 -0.069* 0.024 -0.063*** -0.112** 0.049 -0.014 -0.012 -0.002 RIA Dummy -0.026 -0.011 -0.015 -0.030 -0.028 -0.002 -0.034* -0.035 0.001 0.010 0.058* -0.048** p-value: D1=D5? 0.00*** 0.00*** 0.00*** 0.00*** 0.00*** 0.00*** 0.00*** 0.00*** 0.00*** 0.39 0.00*** 0.00*** Adjusted R2 0.149 0.098 0.080 0.167 0.111 0.091 0.182 0.140 0.095 0.137 0.112 0.111 Includes Style Dummies? Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes N 4,953 4,953 4,953 1,506 1,506 1,506 1,197 1,197 1,197 899 899 899

40

Table 6 Buying and Selling Activity as a Function of Hedge Fund Constraints

This table performs OLS regression analyses examining quarterly sells, buys, and net sells (sells-buys) of stock at the hedge fund family level. The dependent variables are sales as a proportion of market value, buys as a proportion of market value, and net buys as a proportion of market value. Regressions are performed for the entire sample period (1998-2010), all crisis periods, the recent crisis period, and all boom periods. Independent variables are described in Section III. Standard errors are clustered at the family and time levels. Statistical significance is denoted with * ,** , and *** at the 1%, 5%, and 10% levels, respectively.

All Periods (1998-2010) All Crises Subprime Crisis (Q107-Q209) All Booms Sells Buys Net Sells Sells Buys Net Sells Sells Buys Net Sells Sells Buys Net Sells Constant 0.877*** 0.802*** 0.075 0.858*** 0.718*** 0.140 0.820*** 0.640*** 0.180 1.064*** 1.142*** -0.078 Boom Indicator 0.013 0.031** -0.018** Crisis Indicator -0.003 -0.013 0.010 Flow Quintile Indicators Bottom Flow Quintile (D1) 0.035*** 0.002 0.033* 0.032*** -0.022 0.054*** 0.032*** -0.020* 0.052*** 0.008 0.017 -0.008 Top Flow Quintile (D5) -0.003 0.088*** -0.091*** -0.020** 0.085*** -0.106*** -0.014* 0.089*** -0.103*** -0.008 0.138*** -0.146*** Lockup and Red. Period Short Lockup Pd. Quintile 0.027 0.056** -0.029** 0.048* 0.116*** -0.068*** 0.066*** 0.124*** -0.058** 0.013 0.039 -0.026 Long Lockup Pd. Quintile -0.010 -0.039 0.029 0.002 -0.039 0.041 0.001 -0.043 0.044 0.013 -0.055 0.068** Short Red. Pd. Quintile -0.014 -0.052 0.038* -0.031 -0.091* 0.059 -0.017 -0.134*** 0.117*** -0.020 -0.095*** 0.075*** Long Red. Pd. Quintile -0.038* -0.055** 0.018 -0.011 -0.011 0.000 -0.004 -0.022 0.018 -0.059*** -0.057 -0.002 Control Variables Leverage Dummy -0.031 -0.020 -0.011 -0.032 -0.026 -0.006 -0.023 -0.028 0.006 -0.035 -0.027 -0.008 Log Stock Portfolio Size -0.038*** -0.019*** -0.019*** -0.037*** -0.009 -0.028** -0.039*** -0.012 -0.027 -0.040*** -0.030* -0.011 Fund Quarterly Return 0.221*** 0.438*** -0.217*** 0.218*** 0.544*** -0.326*** 0.176*** 0.508*** -0.332*** 0.268** 0.238 0.030 Log of Fund Age 0.038** 0.022 0.017 0.028** -0.006 0.035** 0.027 0.000 0.027 0.057*** 0.029** 0.028** Incentive Fee (% Profits) -0.038 0.157 -0.195 0.422 0.212 0.210 0.797* 0.472 0.324 -0.524* -0.041 -0.483 Mgmt. Fee (% AUM) 7.678*** 7.705** -0.028 8.029*** 8.591** -0.562 8.839*** 12.198*** -3.359* 10.442*** 11.420*** -0.978 Log of Min. Investment 0.009 -0.009 0.018*** 0.004 -0.014 0.018** 0.004 -0.011 0.015 0.000 -0.024 0.024* High Watermark Dummy -0.022 -0.015 -0.007 -0.028 -0.020 -0.008 -0.034 -0.042 0.008 -0.010 -0.028 0.019 RIA Dummy -0.036 -0.024 -0.013 -0.031 -0.020 -0.011 -0.037 -0.022 -0.015 0.005 0.043 -0.038 p-value: D1-D5? 0.00*** 0.00*** 0.00*** 0.00*** 0.00*** 0.00*** 0.00*** 0.00*** 0.00*** 0.94 0.01*** 0.00*** Pseudo R2 0.167 0.119 0.086 0.183 0.153 0.112 0.202 0.190 0.119 0.165 0.137 0.131 Includes Style Dummies? Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes N 4,546 4,546 4,546 1,380 1,380 1,380 1,095 1,095 1,095 791 791 791

41

Table 7: Monthly Cumulative Average Abnormal Return for Stocks Around Hedge Fund Sales This table reports cumulative abnormal returns in excess of the equally-weighted average of all stocks held by hedge funds at the beginning of the month. Results are reported for “pressure sale” stocks, which are stocks with PRESSURE below the 20th percentile. Results are also reported for “pressure buy” stocks, which are stocks with PRESSURE above the 20th percentile. PRESSURE is stock-level variable calculated as the quantity of shares bought by mutual funds with inflows above the 20th percentile less the quantity of shares sold by mutual funds with outflows below the 20th percentile scaled by the average monthly trading volume for the stock, requiring at least 10 owners for each stock and at least 25 hedge funds during the event quarter. Statistical significance is denoted with * ,** , and *** at the 1%, 5%, and 10% levels, respectively. These results are graphed in Figure 1.

Panel A: All Periods

Panel B: All Crises

Panel C: Subprime Crisis

Panel D: Booms

Pressure Sales Pressure Buys Time CAAR t-statistic CAAR t-statistic

Event Period [t-12, t-1] 0.121*** 2.81 0.127*** 3.26 Event Period [t1, t3] -0.017 -1.12 0.015 1.34

Event Period [t+1, t+6] -0.004 -0.12 -0.035* -1.88 Event Period [t+1, t+12] 0.018 0.50 -0.036 -1.53 Event Period [t+1, t+18] 0.005 0.12 -0.062 -1.63

Pressure Sales Pressure Buys Time CAAR t-statistic CAAR t-statistic

Event Period [t-12, t-1] 0.166** 2.06 0.241*** 2.79 Event Period [t1, t3] -0.034 -1.28 0.016 0.69

Event Period [t+1, t+6] -0.061 -1.34 -0.035 -1.09 Event Period [t+1, t+12] -0.053 -1.19 -0.045 -1.00 Event Period [t+1, t+18] -0.069* -1.73 -0.083 -1.40

Pressure Sales Pressure Buys Time CAAR t-statistic CAAR t-statistic

Event Period [t-12, t-1] 0.134** 1.95 0.185*** 2.48 Event Period [t1, t3] -0.017 -0.74 0.034 1.36

Event Period [t+1, t+6] -0.013 -0.36 -0.030 -0.85 Event Period [t+1, t+12] -0.001 -0.02 -0.052 -1.34 Event Period [t+1, t+18] -0.026 -0.84 -0.078 -1.43

Pressure Sales Pressure Buys Time CAAR t-statistic CAAR t-statistic

Event Period [t-12, t-1] 0.165 1.26 0.064 0.40 Event Period [t1, t3] -0.051 -1.50 0.011 0.22

Event Period [t+1, t+6] 0.047 0.59 -0.169 -1.52 Event Period [t+1, t+12] 0.151 0.82 -0.152 -1.15 Event Period [t+1, t+18] 0.096 0.59 -0.221 -0.98

42

Table 8 Characteristics of Pressure Buys and Pressure Sales at the Individual Stock Level

This table reports medians of stock-level characteristics of pressure stocks and all other holdings for the period 2001-2010. Pressure buys are stocks with heavy buying pressure and pressure sells are stocks with heavy selling pressure, as defined in Section IV.

Panel A: All Periods All CRSP

Stocks Pressure

Sales (PS) Pressure Buys

(PB) All Other HF

Holdings PS-All Other HF Holdings

PS-PB

Prior Year Annual Return 0.068 0.126 0.147 0.033 0.093*** -0.021**

Prior Year Standard Dev. 0.094 0.095 0.094 0.108 -0.013*** 0.001 Total Assets ($MM) 567 2,851 2,906 609 2,242*** -55 Tobin’s Q 1.54 2.03 1.93 1.63 0.40*** 0.10 Market Cap ($MM) 466 2,873 2,826 532 2,341*** 47 Avg. Qrtly. Vol. (MM) 10 69 59 14 55*** 10*** Amihud Illiquidity * 108 0.073 0.005 0.004 0.061 -0.056*** 0.001* B-A: (B-A)/B*102 0.029 0.011 0.011 0.028 -0.017*** 0.000**

Panel B: All Crises All CRSP

Stocks Pressure

Sales (PS) Pressure Buys

(PB) All Other HF

Holdings PS-All Other HF Holdings

PS-PB

Prior Year Annual Return -0.046 0.041 0.097 -0.085 0.126*** -0.056***

Prior Year Standard Dev. 0.102 0.090 0.091 0.116 -0.026*** -0.001 Total Assets ($MM) 595 4,574 3,832 623 3,951*** 742 Tobin’s Q 1.49 2.12 1.83 1.56 0.56*** 0.29 Market Cap ($MM) 490 3,912 3,565 523 3,389*** 347 Avg. Qrtly. Vol. (MM) 12 96 65 16 80*** 31*** Amihud Illiquidity * 108 0.066 0.003 0.003 0.062 -0.059*** 0 B-A: (B-A)/B*102 0.027 0.012 0.011 0.027 -0.015*** 0.001*

Panel C: Subprime Crisis

All CRSP Stocks

Pressure Sales (PS)

Pressure Buys (PB)

All Other HF Holdings

PS-All Other HF Holdings

PS-PB

Prior Year Annual Return -0.052 0.052 0.112 -0.077 0.129*** -0.060***

Prior Year Standard Dev. 0.093 0.085 0.088 0.104 -0.019*** -0.003 Total Assets ($MM) 647 4,491 3,716 635 3,856*** 775 Tobin’s Q 1.51 2.18 1.84 1.59 0.59*** 0.34 Market Cap ($MM) 557 3,953 3,556 556 3,397*** 397 Avg. Qrtly. Vol. (MM) 14 95 67 17 78*** 28*** Amihud Illiquidity * 108 0.045 0.003 0.003 0.047 -0.044*** 0.000 B-A: (B-A)/B*102 0.02 0.011 0.010 0.021 -0.010*** 0.001**

Panel D: All Booms All CRSP

Stocks Pressure

Sales (PS) Pressure

Buys (PB) All Other HF

Holdings PS-All Other HF Holdings

PS-PB

Prior Year Annual Return 0.045 0.058 -0.006 -0.013 0.071** 0.064

Prior Year Standard Dev. 0.099 0.147 0.127 0.121 0.026*** 0.020** Total Assets ($MM) 551 1,871 2,226 629 1,242*** -355** Tobin’s Q 1.46 1.99 1.81 1.56 0.43*** 0.18 Market Cap ($MM) 353 1,970 2,028 442 1,528*** -58 Avg. Qrtly. Vol. (MM) 6 84 70 10 74*** 14 Amihud Illiquidity * 108 0.154 0.010 0.009 0.107 -0.097*** 0.001 B-A: (B-A)/B*102 0.056 0.028 0.030 0.054 -0.026*** -0.002

43

Table 9 Logistic Regressions for Pressure Sales and Purchases at the Individual Stock Level

This table reports results from logit regressions with the following dependent variables: a dummy set to 1 if the stock is “pressure sold” during the quarter and zero otherwise, and a dummy set to 1 if the stock is “pressure purchased” during the quarter and zero otherwise. In Panel A, all stocks otherwise held, sold, or purchased (e.g., not “pressure sold” or “pressure purchased”) are included in the analysis, with the appropriate dummy variable set to zero. In the “Pressure Buy” analysis, pressure sells are excluded, and in the “Pressure Sell” analysis, pressure buys are excluded. In Panel B, only pressure buys and pressure sells are included, with the dependent variable set to 1 if the transaction is a pressure sale, and 0 if the transaction is a pressure buy. As in Table 8, since there are only a few pressure buys and sells prior to 2001, the analysis includes the periods from 2001-2010. All regressions include time dummies. Independent variables are described in Section V. Standard errors are clustered at the time and stock level.

Panel A: All Stocks Held by Hedge Funds Included

All Periods All Crises Subprime Crisis All Booms Pressure

Sells Pressure

Buys Pressure

Sells Pressure

Buys Pressure

Sells Pressure

Buys Pressure

Sells Pressure

Buys Constant -16.263*** -14.205*** -18.778*** -14.641*** -19.621*** -14.508*** -16.736*** -15.474*** (0.00) (0.00) (0.00) (0.00) (0.00) (0.00) (0.00) (0.00) Prior Year Annual Return 0.409*** 0.631*** 0.527*** 0.957*** 0.576*** 0.894*** 0.512*** 0.182 (0.00) (0.00) (0.00) (0.00) (0.00) (0.00) (0.00) (0.13) Prior Year Return Std. Dev. 1.674** 2.557*** 0.005 2.209 0.587 5.096*** 2.985*** 1.661 (0.03) (0.00) (1.00) (0.16) (0.63) (0.01) (0.00) (0.25) Log Total Assets ($MM) 0.167*** 0.215*** 0.214*** 0.293*** 0.216*** 0.276*** 0.008 0.153*** (0.00) (0.00) (0.00) (0.00) (0.00) (0.00) (0.52) (0.00) Log Avg. Quarterly Volume 0.605*** 0.465*** 0.718*** 0.423*** 0.681*** 0.335*** 0.688*** 0.573*** (0.00) (0.00) (0.00) (0.00) (0.00) (0.01) (0.00) (0.00) Tobin’s Q -0.058*** -0.054** -0.018 0.000 0.008 0.006 -0.077 -0.084*** (0.01) (0.04) (0.59) (1.00) (0.79) (0.93) (0.16) (0.00) Bid-Ask Spread * 102 -7.217*** -9.226 -3.965*** -4.747 -7.078*** -28.744 -7.816** -8.830*** (0.01) (0.14) (0.00) (0.44) (0.01) (0.54) (0.02) (0.00) Pseudo R2 0.14 0.13 0.17 0.14 0.17 0.15 0.13 0.12 N 153,656 153,618 57,594 57,587 43,080 43,076 14,077 14,070 Includes Year Dummies? Yes Yes Yes Yes Yes Yes Yes Yes

44

Table 9, Logistic Regressions for Pressure Sales and Purchases, Continued

Panel B: Only Pressure Sales and Buys Included (Dependent Variable = 1 if Pressure Sale and 0 if Pressure Buy)

All Periods All Crises Subprime Crisis All Booms Constant -2.695* -5.773** -5.880* -2.011 (0.07) (0.02) (0.08) (0.70) Prior Year Annual Return -0.179 -0.346** -0.304 0.212 (0.14) (0.02) (0.11) (0.62) Prior Year Return Standard Dev. -1.039 -2.452 -4.030 0.964 (0.32) (0.37) (0.13) (0.52) Log Total Assets ($MM) -0.053 -0.055 -0.059 -0.422*** (0.17) (0.21) (0.30) (0.01) Average Quarterly Volume 0.183*** 0.378*** 0.385*** 0.281 (0.01) (0.00) (0.01) (0.20) Tobin’s Q -0.002 -0.008 -0.001 -0.068 (0.87) (0.70) (0.96) (0.36) Bid-Ask Spread * 102 0.581 -0.240 5.155 5.026 (0.78) (0.95) (0.53) (0.11) Pseudo R2 0.01 0.03 0.03 0.05 N 2,786 949 830 195 Includes Time Dummies? Yes Yes Yes Yes

45

0

500

1000

1500

2000

2500

3000

3500

4000

4500

5000Mar‐01

Jun‐01

Sep‐01

Dec‐01

Mar‐02

Jun‐02

Sep‐02

Dec‐02

Mar‐03

Jun‐03

Sep‐03

Dec‐03

Mar‐04

Jun‐04

Sep‐04

Dec‐04

Mar‐05

Jun‐05

Sep‐05

Dec‐05

Mar‐06

Jun‐06

Sep‐06

Dec‐06

Mar‐07

Jun‐07

Sep‐07

Dec‐07

Mar‐08

Jun‐08

Sep‐08

Dec‐08

Mar‐09

Jun‐09

Sep‐09

Dec‐09

Mar‐10

Jun‐10

Sep‐10

Figure1:QuarterlyCountsofStocks,byPressureCategory

Non‐PressureStocks PressureSaleStocks PressurePurchaseStocks

46

Figure 2: Performance of Pressure Sales and Pressure Buys

Panel A graphs cumulative monthly abnormal returns around the “pressure sales” and “pressure buys” of stocks for all periods, Panel B graphs crisis periods only, Panel C graphs the recent crisis only, and Panel D graphs all booms. Pressure sales and pressure purchases are defined in Section IV.

Panel A

Panel B

‐0.050

0.000

0.050

0.100

0.150

0.200

0.250

CumulativeAverageAbnormalReturn

TimeRelativetoEventDate

AllPeriods

PressureSales PressureBuys

‐0.050

0.000

0.050

0.100

0.150

0.200

0.250

CumulativeAverageAbnormalReturn

TimeRelativetoEventDate

AllCrises

PressureSales PressureBuys

47

Figure 2, Continued

Panel C

Panel D

‐0.050

0.000

0.050

0.100

0.150

0.200

0.250

CumulativeAverageAbnormalReturn

TimeRelativetoEventDate

SubprimeCrisis

PressureSales PressureBuys

‐0.200

‐0.100

0.000

0.100

0.200

0.300

CumulativeAverageAbnormalReturn

TimeRelativetoEventDate

AllBooms

PressureSales PressureBuys

48

Figure 3: Pressure Sales, Purchases, and All Other Stocks Relative to Market Capitalization and Volume

0.0%

0.5%

1.0%

1.5%

2.0%

2.5%

3.0%

3.5%

4.0%

4.5%

5.0%M

ar-0

1 Ju

n-01

Se

p-01

D

ec-0

1 M

ar-0

2 Ju

n-02

Se

p-02

D

ec-0

2 M

ar-0

3 Ju

n-03

Se

p-03

D

ec-0

3 M

ar-0

4 Ju

n-04

Se

p-04

D

ec-0

4 M

ar-0

5 Ju

n-05

Se

p-05

D

ec-0

5 M

ar-0

6 Ju

n-06

Se

p-06

D

ec-0

6 M

ar-0

7 Ju

n-07

Se

p-07

D

ec-0

7 M

ar-0

8 Ju

n-08

Se

p-08

D

ec-0

8 M

ar-0

9 Ju

n-09

Se

p-09

D

ec-0

9 M

ar-1

0 Ju

n-10

Se

p-10

Prop

ortio

n of

Sto

ck's

Mar

ket C

ap

Panel A: Average Trade Dollar Value/Stock Market Capitalization

All Other Stocks Pressure Sales Pressure Buys

0.0%

2.0%

4.0%

6.0%

8.0%

10.0%

12.0%

14.0%

16.0%

Mar

-01

Jun-

01

Sep-

01

Dec

-01

Mar

-02

Jun-

02

Sep-

02

Dec

-02

Mar

-03

Jun-

03

Sep-

03

Dec

-03

Mar

-04

Jun-

04

Sep-

04

Dec

-04

Mar

-05

Jun-

05

Sep-

05

Dec

-05

Mar

-06

Jun-

06

Sep-

06

Dec

-06

Mar

-07

Jun-

07

Sep-

07

Dec

-07

Mar

-08

Jun-

08

Sep-

08

Dec

-08

Mar

-09

Jun-

09

Sep-

09

Dec

-09

Mar

-10

Jun-

10

Sep-

10

Prop

ortio

n of

Sto

ck's

Qua

rter

ly V

olum

e

Panel B: Number of Shares Traded/Quarterly Volume

All Other Stocks Pressure Sales Pressure Buys