Upload
others
View
1
Download
0
Embed Size (px)
Citation preview
Duration of Poor Performance, Fund Flows, and
Risk-Shifting by Hedge Fund Managers1
Ying Li
2
A. Steven Holland3
Hossein B. Kazemi4
Abstract
A typical hedge fund manager receives greater compensation when the fund has a strong
absolute or relative performance. Asymmetric performance fees and fund flow-performance
relationship may create incentives for risk-shifting, estimated in our study by the change in fund
return volatility in the middle of the year. However, hedge funds that cannot attract new funds or
have had poor performance for a long period may face different incentives. The combination of
these two observations confronts hedge fund managers with a complex strategic decision
regarding the optimal level of their funds’ return volatility. While an increase in return volatility
generally increases the expected payoff of the compensation contract, excessive volatility is not
sustainable. This paper empirically examines the factors that affect hedge fund managers’
decisions to risk-shifting. We show that (1) if the fund has had prior poor performance, the
magnitude of risk-shifting is larger; (2) as the duration of poor performance increases, risk-
shifting is reduced; (3) if the fund is experiencing capital outflows, the magnitude of risk-shifting
is smaller and (4) funds that have outflows and also use leverage or have short redemption notice
periods display a smaller degree of risk-shifting. (JEL G11, G23, G32)
Keywords: hedge fund, fund flow, performance, incentive, risk
1 We wish to thank George Aragon, Stephen Brown, Peter DeMarzo, Rajib Doogar, Mila Getmansky, Nikunj
Kapadia, Bing Liang, PK Sen, Mark Westerfield, and conference participants at the EFM 2011 Alternative
Investments Symposium at York University for their many helpful comments. We are responsible for any errors. 2 Corresponding author, Assistant Professor of Business, University of Washington Bothell. Email: [email protected],
Tel: 425-352-3413
3 Professor of Business, University of Washington Bothell. Email: [email protected].
4 Professor of Finance, University of Massachusetts, Amherst. Email: [email protected].
1. Introduction
Hedge fund managers receive payoffs that are convex functions of the fund’s
performance, measured in both absolute and relative terms. On one hand, a typical manager
earns 20% performance fee (Liang, 2000), which is based on the fund’s absolute performance,
and is usually not subject to a claw-back policy. On the other hand, a convex fund flow-
performance relation has been observed in the hedge fund industry (Agarwal et al., 2002), which
means fund flows respond more to a high ranking of relative performance than they do to a low
ranking of relative performance. Since both performance fee and relative performance ranking
are usually assessed annually at the end of a calendar year, the payoff structure, which is similar
to that of a call option, may lead the managers to engage in mid-year risk-shifting, i.e., a mid-
year change of return volatility (Carpenter, 2000; Aragon and Nanda, 2012).
While it is generally believed that convex incentive structures lead to risk-shifting, the
questions of why and under what conditions risk-shifting occurs remain unanswered (Ross,
2004). Previous studies provide evidence supporting the presence of risk-shifting due to the
implicit incentive from the convex flow-performance relation, also known as the “tournament
incentive” (Brown et al., 1996). However, there is mixed evidence on how the explicit incentive,
which is based on the convex compensation contract, influences risk-shifting (Brown et al.,
2001; Aragon and Nanda, 2012). Clearly, the impact of the explicit incentive on the risk-shifting
behavior of managers is of general interest because stock options are used extensively in
executive compensation contracts. While executives have the power to influence their
compensation contracts, hedge funds usually set up their compensation structure at inception
with little alteration going forward (Agarwal et al., 2009). Hedge funds, therefore, offer a
laboratory to test the relation between convex payoff and risk-taking behavior with less concern
about endogeneity.
This paper attempts to address the gap in the literature by focusing on the joint impact of
hedge fund managers’ expected payoffs from their compensation contracts, direction of fund
flows into or from their funds, and past performance on their risk-shifting decisions. Since
managers hold a portfolio of investor-specific1 performance fee options, we show that not only
1 Though theoretical models of optimal risk-taking by fund managers refer to a unique fund-level HWM (Carpenter,
2000; Panageas and Westfield, 2009), in practice, in order to avoid the free rider problem, hedge funds issue new
share classes as new investors contribute capital to the fund. This means, the HWM is defined at the investor level
the most recent performance, but cumulative past performance matters for their risk-shifting
decisions. Risk-shifting is not a strategy that can repeatedly be used, as a fund’s return volatility
will become exceedingly unattractive. We show that the duration of a fund’s poor performance, a
factor that, to the best of our knowledge, has not been explored in prior research, influences the
risk-shifting decision of managers. Furthermore, we show that two important fund
characteristics, the length of redemption notice period required by a fund and, especially, the use
of leverage by the fund manager, interact with fund flows to impact a manager’s risk-shifting
policy. Taking the history of a fund’s performance into account also helps reconcile the
previously reported mixed evidence regarding the relation between the moneyness of the
performance fee option and risk-shifting policies of fund managers.
Our main source of data is the Morningstar/CISDM database over the sample period of
January 1994 through December 2012. Similar to Aragon and Nanda (2012), we construct a
proxy for the high-water mark (HWM) of a hedge fund by tracking its net asset value (NAV)
over a three-year window using reported net-of-fees returns. Our empirical framework to identify
the conditions that may influence a hedge fund manager’s risk-shifting decision includes two
variables that are measured after the fund achieves its HWM: performance and fund flow. For
performance, we also consider the duration of underperformance, to assess the long-term
feasibility of risk-shifting.
We use the performance and fund flow estimates along with the HWM proxy to identify
the “state” that describes (1) the moneyness of a hedge fund manager’s compensation option
portfolio and (2) the fund flow condition. Then, we focus on the interactions between (1) the
state variables and the first half-year performance and (2) the state variables and fund
characteristics in a regression setting to investigate hedge funds’ state-dependent risk-shifting
behavior.
We document several new findings. First, a hedge fund manager’s risk-shifting is more
pronounced when the fund is in a state that its performance fee option is partly out-of-the-money.
The magnitude of risk-shifting for a fund in such a state is more than double that of a fund in the
state where the performance fee option is at-the-money. Most hedge funds have performed well
during our sample period with about 75% of managers having at-the-money performance fee
and not at the fund level. The incentive to change the fund’s volatility, therefore, is based on the sensitivity of the
value of a portfolio of call options with different strike prices (i.e., HWMs) to the changes in the volatility of the
fund’s portfolio.
option. We find less risk-shifting among these managers, which may explain the mixed evidence
reported by Brown et al. (2001) and Aragon and Nanda (2012). Second, risk-shifting ceases if a
fund’s NAV is below its HWM proxy for an extended period. We find that a fund reacts to poor
performance by increasing its return volatility in the short-run, but does not engage in risk-
shifting after being underwater for more than one year. Third, when facing outflows, a fund that
employs leverage usage or imposes short redemption notice on its investors engages in a smaller
magnitude of risk-shifting.
Besides performance, fund flows, leverage, and redemption periods, a number of other
manager characteristics such as the manager’s physical location, education, and risk-aversion
could influence a fund’s risk-shifting incentives. A manager is typically part of the ownership
structure of a hedge fund, and, therefore, remains with the fund throughout its life (Liang, 2000).
Consequently, regressions with fund fixed effects greatly alleviate the concern of biased
estimates due to the presence of omitted time-invariant variables.2 Our empirical findings are
robust to using fund fixed effects regressions, and are robust to variations in our methodology,
including varying the definitions of poor performance and fund flows. The findings also appear
to be robust to the impact of potential biases, as our results are similar when we use a sample that
is less subject to backfill bias (the “no backfill bias” sample) and the full sample.
Our study is related to a large literature on hedge funds’ managerial risk-taking incentives
and behavior conditional on fund characteristics. A number of empirical studies have examined
hedge fund managers’ risk-taking by investigating the impacts of survival and career concerns
(Fung and Hsieh, 1997; Brown et al., 2001; Aragon and Nanda, 2012), presence of high-water
mark provisions and managerial stakes (Aragon and Nanda, 2012), usage of derivatives (Chen,
2013), and past performance history (Aragon and Nanda, 2012; Ray, 2012). Past studies also
document a convex relationship between fund flows and performance for mutual funds
(Chevalier and Ellison, 1997; Sirri and Tufano, 1998), and for hedge funds (Agarwal et al.,
2002), how fund flows are related to risk taking decisions in the mutual fund industry (Basak et
al., 2007; Hu et al., 2011) and how a less convex flow-performance relation leads to less risk-
shifting in the pension fund industry (Del Guercio and Tkac, 2002). Additional studies examine
the interaction of leverage policy and risk taking strategies (Brown et al., 2001; Ang et al., 2011),
2 Firm fixed effects regression models are widely used in the corporate finance literature to reduce endogeneity
concerns due to time-invariant omitted variables. For example, Adams and Ferreira (2009), Oikonomou, Brooks,
and Pavelin (2012).
and explore the interaction between fund flows, performance, and managerial incentive in the
hedge fund industry (Agarwal et al., 2002 & 2009). The importance of liquidity provisions in
affecting hedge fund risk-return profiles is examined in Getmansky et al. (2004) and Aragon
(2007). Li and Kazemi (2007) and Cumming and Dai (2010) investigate the conditional
properties of hedge funds’ return distributions.
We add to this vast literature by showing that hedge fund managers use risk-shifting as a
transient strategy to maximize payoffs. We also show that fund outflows play an important role
in hedge fund managers’ risk-shifting decisions. Because we control for fund-level fixed effects,
our results are less prone to endogeneity concerns due to time-invariant non-observable missing
variables that may influence a fund’s risk-shifting decision.
The paper’s structure is as follows. Section 2 presents the methodology, testable
hypotheses, and the data, Section 3 discusses the research design and results, Section 4 presents
tests for robustness, and Section 5 concludes.
2. Methodology, hypotheses and data
2.1 Methodology
Following Aragon and Nanda (2012), our research variable is the change in the return
volatility of each hedge fund in our sample during the second half of the year compared to the
volatility in the first-half year. From the limited data disclosed by hedge funds, we use a rolling
three-year window to estimate the maximum HWM of each fund’s current investments and to
obtain a history of its fund flows. With the help of the HWM proxy, we estimate two states for
each hedge fund: (1) Past performance-related state, which we refer to as the Moneyness of its
performance fee option and (2) fund Flow. We use these estimates to examine the responses of
hedge fund managers to poor performance in the first half of the fourth year. The maximum
value of each fund’s NAV during the three-year window is compared to the fund’s NAV at the
beginning of the fourth year (evaluation year) as the measure of moneyness of the performance
fee option:
,
,
,
i t
i t
i t
NAVMoneyness
HWM (1)
In the above definition of Moneyness, 𝑁𝐴𝑉𝑖,𝑡 is the net asset value of a share of hedge
fund 𝑖 as of the end of the three-year (36-month) window (the beginning of the evaluation year),
and 𝐻𝑊𝑀𝑖,𝑡 = max(𝑁𝐴𝑉𝑖,𝑡−𝜏; 0 ≤ 𝜏 < 36) is the maximum level of NAV over the three-year
window.
If Moneyness is equal to one, the manager’s performance fee option for all assets under
management is at the money. If Moneyness is less than one, the manager’s performance fee
option is out-of-the-money for at least some of the assets under management. Clearly, our
estimate of the moneyness of the performance fee option is subject to errors. We check for the
robustness of our results by changing the cutoff threshold of Moneyness in Section 4.
It is important to note that Moneyness is a conditioning variable that refers to the status of
the performance fee option at the beginning of the evaluation year.3 Furthermore, depending on
the timing of the investment, each share class will have its own specific HWM, and, therefore, its
specific degree of moneyness. This means that the variable defined as Moneyness in Equation
(1) should be interpreted as a proxy for the aggregated moneyness of all share classes of a given
fund at the beginning of the year. In addition, unless there are inflows, a fund that has negative
performance during the first half of the year will not collect a performance fee unless it displays
positive performance during the second half of the year.
We estimate the fund-level aggregate flow by comparing the actual assets under
management as of mid-evaluation year to the expected assets under management calculated
based on the fund’s cumulative return since the time the HWM was achieved. The aggregated
fund-level flow over [𝑡𝐻𝑊𝑀, 𝑡 + 6]is calculated following Sirri and Tufano (1998):
, 6 , , 6 ,
, , 6
,
1HWM HWM
HWM
HWM
i t i t t i t
i t t
i t
AUM AR AUMFlow
AUM
(2)
Where HWMt is the time when the HWM was reached, t is the beginning of the evaluation year,
𝐴𝑈𝑀𝑖,𝑡+6is the amount of assets under management as of June 30 of the evaluation year,
𝐴𝑈𝑀𝑖,𝑡𝐻𝑊𝑀is the amount of assets under management at the time when the HWM was reached,
and 𝐴𝑅𝑖,𝑡𝐻𝑊𝑀,𝑡+6 is the accumulative return over the window [𝑡𝐻𝑊𝑀, 𝑡 + 6].
We define two dummy variables, M and F, to describe the state a fund is in depending on
its moneyness and flow conditions. We set M equal to 1 if Moneyness < 0.9 as of the beginning
of the evaluation year, indicating that the fund has had meaningful negative returns after
3 The results do not change if Moneyness is measured mid-year. The results are available from the authors upon
request.
reaching its HWM. We set F equal to 1 if , indicating that the fund has experienced
outflows after reaching its HWM.4
2.2 Hypotheses
As hedge fund managers react to convex payoffs, whether they are from the embedded
option in their compensation contracts or the asymmetric flow-performance relation, our first
hypothesis is straightforward:
Hypothesis 1: Fund volatility increases following poor performance in the first half of the
year.
Hypothesis 2 relates a hedge fund manager’s risk-shifting to the state the fund may find
itself in depending on Moneyness and Flow. If a large proportion of the manager’s performance
fee option portfolio is out-of-the-money at the beginning of the year, the manager will need to
increase the volatility by a larger amount to obtain the same amount of benefits obtained by the
fund if only a small portion of the performance fee option is out-of-the-money. Thus, the
moneyness determines the magnitude of change in volatility needed to bring about a given
change in the value of the performance fee option.
We hypothesize that fund outflows affect how a fund weighs the cost of risk-shifting.5
That is, if a fund has negative performance but no outflows, the manager may conclude that
investors still have a positive view of the fund, and, therefore, the costs associated with risk-
shifting are not significant. However, if a fund experiences outflows following poor
performance, costs associated with increased risk-shifting will be deemed significant, and the
magnitude of risk-shifting by such a fund is likely to be small. This argument leads to our
second hypothesis:
Hypothesis 2: The increase in return volatility following poor performance in the first
half of the year is greater for a fund that has an out-of-the-money performance fee option and
has experienced no fund outflows.
4 We use dummy variables (M and F) instead of continuous variables for Moneyness and Flow to mitigate some
multicollinearity issue and to allow errors of margin since Moneyness and Flow are based on the estimated HWM
proxy and therefore are not exact measures. We also conduct robustness checks for various M and F definitions in
Section 4. 5 Fund flow has been used as a measure for investors’ sentiment. (For example, Jain and Wu, 2000; Barber, Odean,
and Zheng, 2004; Frazzini and Lamont, 2008).
0Flow
In addition, we explore the roles fund characteristics play in the dynamics of fund flows,
the moneyness of the performance fee option, and the manager’s risk-shifting policy. Past studies
have examined how certain fund characteristics influence hedge fund managers’ risk-shifting
incentive and behavior. For example, using a theoretical model, Hodder and Jackwerth (2007)
argue that managerial stake reduces the incentive for risk-shifting; Aragon and Nanda (2012)
provide empirical evidence supporting this argument. We focus on two other characteristics that
are likely to stand out due to their close relationship with fund flows and risk: the redemption
notice period and the use of leverage. Getmansky et al. (2011) and Aragon et al. (2014) show
that the flow-performance relationship turns concave from convex when the fund imposes share
restrictions. Redemption notice period is a type of share restriction, and short redemption notice
periods allow investors to react quickly to changes in the fund’s investment style, and, therefore,
risk-shifting could lead to increased outflows.
Hypothesis 3a: Conditional on fund outflows, the increase in return volatility is smaller
for hedge funds that have short redemption notice periods.
Just as funds with short redemption notice periods could face increased cost associated
with shifting risk, funds that use leverage may also find such a policy costly. In the absence of
any changes in a fund’s investment and leverage policy, return volatility of the fund will
endogenously increase following a period of poor performance. On the other hand, when the
fund experiences outflows, managers may be more wary of the cost of increasing volatility and
refrain from doing so. Furthermore, even if the manager wanted to increase volatility, lenders
may impose active constraints on risk-shifting policies of a fund (Brown et al., 2001). Both
scenarios would result in a smaller increase in volatility.
Brown et al. (2001) find empirical evidence for non-passive leverage policy, Ang et al.
(2011) show hedge funds are experts at managing their leverages compared to listed financial
intermediaries, and Lan et al. (2013) suggest active leverage policy in a theoretical model; based
on these studies, we argue that levered funds will find it too costly to increase their return
volatility if they have experienced fund outflows.
Hypothesis 3b: Conditional on fund outflows, the increase in return volatility is smaller
for hedge funds that use leverage.
Our last hypothesis is about the impact of the duration of time that a fund’s portfolio of
performance fee options has been partly out-of-the-money on its risk-shifting decision. On one
hand, if a fund has not collected performance fees for a long time, the fund may have attempted
risk-shifting before or the fund’s performance fee option is deeply out-of-the-money so that risk-
shifting may no longer help. Thus, it may not react to poor first half-year performance. On the
other hand, the manager of such a fund may consider closing the fund and walking away from it.
Such a manager may decide to increase the volatility as a last attempt to collect performance fees
and extend the life of the fund (see Hodder and Jackwerth, 2007; Drechsler, 2014). Thus, the
fund may further increase its return volatility after poor first-half year performance.
Hypothesis 4: Duration of poor performance affects the risk-shifting behavior of hedge
fund managers.
2.3 Hedge fund sample
We use the Morningstar-CISDM Hedge Fund Database for this study.6 This database
provides information on performance and various characteristics of thousands of hedge funds,
funds of hedge funds, and CTAs.7 For each fund, we observe monthly net-of-fees returns, NAV,
major trading strategy claimed by the fund, whether the fund is listed on an exchange, and the
regulatory agency for the fund.8 For most funds, we are also able to obtain information on their
initiation months; assets under management (AUM); currency of AUM; performance fee rate;
management fee rate; whether they use leverage; and liquidity constraints they impose including
lockup period, redemption notice period, and redemption frequency. Following Sirri and Tufano
(1998), we use monthly AUM and returns to calculate aggregated fund flows occurring after a
fund has achieved its HWM. We do not have data on manager’s ownership share of the fund, or
the reasons for a fund to appear in the defunct database.
We focus on US-dollar-denominated individual hedge funds and exclude funds of hedge
funds and CTAs.9 Each hedge fund is required to have 48 months of return history
10 and
6 The Morningstar Hedge Fund Database is built on the CISDM database, which used to be Zurich Hedge Fund
Universe, formerly known as the MAR database. 7 Funds of hedge funds are portfolios of individual hedge funds. CTAs are funds that specialize in futures trading.
8 We use net-of-fees returns for this study. We recognize that the HWM and fees are based on returns computed
before all management fees and performance fees are subtracted, hence gross returns will be most appropriate to use.
However, past research demonstrates little difference from using net-of-fee returns versus using gross returns. 9 The incentive to shift risk is less obvious for funds of hedge funds. For example, Agarwal et al.(2002) do not find
funds of hedge funds to exhibit strategic risk-taking behavior and attribute it to the inflexibility of the business.
information on characteristics, including fund size (AUM), and the strategy followed by the
manager (style). We are aware that these requirements may introduce certain biases into our
sample selection process. We perform robustness checks to understand the potential impact of
these biases. After applying the above filters, we are left with 11934 individual observations for
the full sample and 8199 observations for the no backfill bias sample.11
12
3. Empirical results
3.1 Summary statistics
We examine monthly return data from January 1994 to December 2012 to evaluate hedge
funds’ risk-shifting behavior from 1997 to 2012. We include both live and defunct funds to
minimize survivorship bias. To mitigate unwarranted noise from outliers, we also winsorize our
sample at 1% and 99% for the following measures: fund’s return volatility in the first and second
half years, the moneyness of a fund’s performance fee option and fund flow.
The summary statistics of our sample are reported in Table 1. There are 11934 individual
observations appearing in the database between January 1994 and December 2012 with the
number increasing every year through 2007 steadily and declining thereafter. Panel A shows that
the mean level of AUM at the beginning of each evaluation year increased from $131.83 million
in 1997 to $234.92 million in 2008 and then fell to $155.81 million in 2012. Panel B reports
the number of fund-year observations for the eight styles (namely, Convertible Arbitrage, Debt-
related, Equity-related, Event Driven, Global Macro, Merger Arbitrage, Multi-strategy and
Volatility) in our study. It shows that Equity-related is the most prominent style of funds in our
sample, representing about 62% of both the full and no backfill samples.
Panels C and D report the summary statistics of our dependent variable, independent
variables and control variables used in base case regressions for both no backfill bias and full
CTAs are a group of managers that trade on highly liquid futures and there is less variation in the liquidity provision
at CTAs. However, robustness checks show that our results are robust to including all funds of hedge funds. These
additional results are available upon request. 10
We relax this requirement in robustness tests and our results are qualitatively the same. 11 The Morningstar-CISDM Hedge Fund Database uses a dummy variable for high-water mark provision, where a
“1” means the fund has a high-water-mark. About 88% of the hedge funds in our sample report that they have a
high-water mark (10002 out of the full sample and 6851 out of the no backfill sample, some funds have missing
information). We report results below by assuming all hedge funds with missing information have high-water mark
provisions. The results from a sample that excludes funds with those funds with missing information on their high-
water mark provisions are largely the same. The results are available upon request. 12
No backfill bias sample excludes returns that are “backfilled”, which occur prior to when the fund was added to
the database, to minimize the backfill bias as discussed in Fung and Hsieh (2000).
samples. Out of 8199 fund-year observations in the no backfill bias sample, 2020 have M = 1,
while 4502 have F = 1, and 1520 funds have both M and F equal to 1, indicating that these funds
have had negative flows and performance decline of more than 10% since reaching HWM during
the previous 36 months. Out of the 11934 fund-year observations in the full sample, 2809 have
had such negative return history (M = 1), 6180 have had negative aggregate fund flows (F = 1),
2037 have had both M and F equal to 1. The summary statistics show that even though poor
performance is likely to result in fund outflows, the distinction between poor performance and
outflows is meaningful and necessary. While M and F are correlated, they capture separate fund
characteristics.
Finally, Panel E reports the distribution of funds with and without leverage usage and
with and without short redemption notice periods under different moneyness and flow conditions
for both no backfill bias and full samples. As we can observe from Panel E, more fund-year
observations are associated with leverage usage, and more fund-year observations are associated
with short redemption notice periods. For example, for the no-backfill-bias sample, there are
4403 and 5163 fund-year observations that report leverage usage and impose short redemption
periods, respectively. There are 1338 and 3036 fund-year observations that do not use leverage
and impose long short redemption periods, respectively.
[Table 1 about here]
We also report the correlation coefficients for our variables in Table 2. The Pearson
correlation coefficient for M and F is 0.23, suggesting that multicollinearity should not be a
major concern.
[Table 2 about here]
3.2 Regression analysis
To investigate the risk-shifting of hedge fund managers when they are in states defined
by M and F, we use multivariate regression to estimate the following equation over the period of
1997-2012, with a focus on the interaction terms 𝑀 × 𝑃𝑒𝑟𝑓, 𝐹 × 𝑃𝑒𝑟𝑓, 𝑇𝑖𝑚𝑒 × 𝑃𝑒𝑟𝑓, 𝐹 ×
𝐿𝑒𝑣𝑒𝑟𝑎𝑔𝑒 and 𝐹 × 𝑆ℎ𝑜𝑟𝑡𝑁𝑜𝑡𝑖𝑐𝑒:
1 2 3 4 5 6 7
8 9 10 11 12
j j j j j
j j
DifStd M Perf M Perf F F Perf TimeUnder TimeUnder Perf
F Leverage F ShortNotice LagStd AuCorr SecondFlow
FundCharacteristic M FundCharacteristic F FundCharacte
j
j
k k
k
ristic
Dummy
(3)
where M = 1 if Moneyness < 0.9 and 0 otherwise,13
and F = 1 if Flow < 0 and 0 otherwise.
The performance of a fund during the first half of the year, which is denoted by Perf, is a
critical variable. We use three different measures of the variable Perf: (1) NegRet, a dummy
variable that equals one if the cumulative return over the first half of an evaluation year is
negative and zero otherwise, (2) LowAbsRetRank, a dummy that equals one if a fund’s
cumulative return over the first half-year is below the median of all funds and zero otherwise,
and (3) LowRelRetRank, a dummy that equals one if the cumulative return over the first-half of
the year is below the median of funds within a given style and zero otherwise. If funds engage in
mid-year risk shifting, the coefficient on Perf will be positive as funds with poor first half-year
performance increase volatility in the second half-year. We initially focus on NegRet to explore
how the explicit incentives arising from out-of-the-money compensation contracts affect hedge
fund risk shifting. We consider the other two variables (LowAbsRetRank and LowRelRetRank)
in additional tests in Section 4.
The interaction terms M Perf and F Perf indicate how the magnitude of volatility
change following poor first half-year performance reacts to changes in M and F, i.e., when the
fund is in a certain M- and F- dependent state.
TimeUnder, an abbreviation for time-underwater, is a continuous integer that ranges
between 0 and 35 and is defined as
, ,t ,
,
,
36 0.9
0 0.9
HWM i i t
i t
i t
t if Moneyness
TimeUnder
if Moneyness
(4)
for fund i as of time t, the beginning of each evaluation year. TimeUnder is an estimate of the
number of months that a fund’s NAV has been less than 0.9 of its HWM, suggesting that the
fund has not collected performance fees for at least some of its assets under management during
13 We conduct robustness checks to evaluate the sensitivity of our results in the next section.
those months. The interaction term 𝑇𝑖𝑚𝑒 × 𝑃𝑒𝑟𝑓 indicates the magnitude of risk-shifting in
response to the duration of past poor performance.
We control for lagged fund volatility (PStd), the percentage of net flow in the second half
year (SecondFlow), and return autocorrelation (AuCorr) because they have previously been
documented to have an impact on fund volatility. PStd controls for mean reversion in risk
changes that may be induced by mis-measurement (Koski and Pontiff, 1999; Aragon and Nanda,
2012). We expect 𝛽10to be negative and greater than -1. As pointed out in Getmansky, et al.
(2004), autocorrelation in fund returns may be a symptom of return smoothing by fund
managers, and may lead to a downward bias in estimating the true return volatility. Since less
liquid strategies usually have higher autocorrelation (Getmansky et al., 2004), higher AuCorr
reflects less liquid asset holdings, which may exhibit lower volatility. Both the downward-
estimation bias and the less-liquid-asset-holding predict a negative coefficient estimate for
AuCorr. We therefore expect 𝛽11 to be negative. SecondFlow controls for a spurious relation
between mid-year performance and changes in fund risk (Ferson and Warther, 1996; Koski and
Pontiff, 1999).
We also control for a number of fund characteristics that are likely to influence a fund’s
risk-shifting decisions. These include fund size, leverage usage, and the length of the redemption
notice period:
Small = 1 if a fund has below-median AUM and 0 otherwise. On one hand, smaller funds
collect less management fees, so the performance fee is more important, suggesting a positive
effect on volatility (more aggressive risk taking). On the other hand, smaller funds cannot afford
to lose assets, and higher volatility could lead to greater losses and loss of funds. This
observation suggests a negative effect on volatility (less aggressive risk taking).
Leverage = 1 if a fund reports to the database that it uses leverage and 0 otherwise. As
suggested before, funds that use leverage should be less aggressive in increasing their return
volatility if they have experienced outflows.
ShortNotice = 1 if a fund has a redemption notice shorter than 30 days and 0 otherwise.
All else constant, funds that offer relatively short redemption notices are expected to be less
aggressive in increasing their return volatility following outflows. We expect the signs of 𝛽8 and
𝛽9,the coefficients for the interaction terms F Leverage and FShortNotice to both be
negative.
In addition to the variables discussed above, we estimate Equation (3) with style fixed
effects and year fixed effects as control variables and separately with fund fixed effects and year
fixed effects. Style fixed effects are dummy variables that represent the different styles of hedge
funds.14
If a fund belongs to a certain style, then the dummy variable corresponding to that style
is set equal to one. The base case used in our regression is equity-related style, which is the
largest category. Year fixed effects are dummy variables that represent each year the fund
manager makes decisions on volatility. There are 16 such evaluation years (1997-2012) for our
sample and the base year is 1997.15
The reported t-statistics are based on robust standard errors
that cluster at fund levels.
Table 3 reports the results from estimating Equation (3) for the no-backfill-bias sample,
with Perf represented by NegRet.16
Columns (1) – (2) report results from a pooled OLS
regression with controls of style and year fixed effects and Columns (3) – (4) report results from
a fixed fund effects regression with year fixed effects.
In Column (1), M, F, and their interactions with NegRet, Small are included. We do not
include the interaction terms between M, F and Leverage and ShortNotice to conserve the
number of observations used for the regression since there are missing values for Leverage and
ShortNotice. The estimated coefficient for NegRet is positive and significant, consistent with
Hypothesis 1: poor performance in the first half of the year leads to increased return volatility.
The estimated coefficients for M and F are both insignificant but the estimate for interaction term
𝑀 ×𝑁𝑒𝑔𝑅𝑒𝑡is positive and significant, with a large coefficient more than double the magnitude
of that on NegRet. This indicates the magnitude of risk-shifting in response to poor performance
is larger for funds that have a larger proportion of their compensation option portfolio out-of-the-
money, consistent with Hypothesis 2. The estimated coefficient for Small is positive and
significant, suggesting that funds with smaller sizes and no outflows have higher risk-shifting.
14
Hedge funds are categorized into various styles to reflect the fact that they may carry very different risk profiles as
they employ different strategies. For our sample, there are 24 categories: Europe HF global emerging markets
equity, HF Asia/pacific long/short equity, HF bear market equity, HF China long/short equity, HF convertible
arbitrage, HF debt arbitrage, HF distressed securities, HF diversified arbitrage, HF emerging markets long-only
equity, HF emerging markets long/short equity, HF equity market neutral, HF Europe long/short equity, HF event
driven, HF global long/short equity, HF global macro, long-only debt, long-only equity, long-only other, long/short
debt, merger arbitrage, multi-strategy, U.S. long/short equity, U.S. small cap long/short equity, volatility. We
summarize these styles into eight based on their assets classes: equity-related, debt-related, convertible arbitrage,
emerging markets, event driven, global macro, multi-strategy and volatility-related. 15
We use 1997 as the base year. Our results do not change with the selection of base year. 16
Results from the full sample are qualitatively the same as those from the no backfill bias sample and are not
reported in this and the following tables to save space. These results are available upon request.
The estimated coefficient for 𝐹 × 𝑆𝑚𝑎𝑙𝑙 is negative and significant and of nearly equal absolute
value to the coefficient for Small, indicating that small funds with outflows have little or no risk-
shifting. The estimated coefficient for TimeUnder is positive and significant, indicating that
higher risk-taking follows longer duration of underperformance. The estimated coefficient for
𝑇𝑖𝑚𝑒𝑈𝑛𝑑𝑒𝑟 × 𝑁𝑒𝑔𝑅𝑒𝑡, however, is negative and significant, indicating less risk-shifting for
hedge funds as the duration of underperformance increases. These findings are consistent with
Hypothesis 4 that the duration of underperformance influences hedge fund managers’ risk-
shifting behavior. The estimated coefficients for AuCorr and PStd are both negative and
significant (and greater than -1 for PStd), consistent with our expectation.
We report results in Column (2) for the regression that includes the additional interaction
terms 𝐹 × 𝐿𝑒𝑣𝑒𝑟𝑎𝑔𝑒 and 𝐹 × 𝑆ℎ𝑜𝑟𝑡𝑁𝑜𝑡𝑖𝑐𝑒. The number of observations goes down from 7765
to 5497 and the estimated coefficients for NegRet, Small, 𝑀 ×𝑁𝑒𝑔𝑅𝑒𝑡, 𝐹 × 𝑆𝑚𝑎𝑙𝑙, TimeUnder,
𝑇𝑖𝑚𝑒𝑈𝑛𝑑𝑒𝑟 × 𝑁𝑒𝑔𝑅𝑒𝑡 and AuCorr have the same sign and statistical significance as those in
Column (1). The coefficient estimate for F becomes highly significant with a magnitude of
0.621. The coefficient estimates for Leverage and ShortNotice are both positive and significant
while those for F Leverage and F ShortNotice , are negative and significant (-0.504 and -
0.289, with t-values of -3.243 and -2.280, respectively). The results suggest that (a) the presence
of leverage and short redemption notice period is associated with larger mid-year volatility
change when funds have not experienced outflows and (b) the presence of leverage and short
notice periods does not increase volatility for hedge funds that experience outflows and may
even decrease it.
The findings point to fund flows as a critical variable in the decision making of fund
managers: a fund that experiences outflow but does not use leverage or have short redemption
notice period has higher fund volatility while a fund that experiences outflow and uses leverage
and has short redemption notice period has lower fund volatility. These results are consistent
with Hypotheses 2 and 3. Furthermore, none of the coefficients on M, 𝑀 × 𝐿𝑒𝑣𝑒𝑟𝑎𝑔𝑒 and
𝑀 × 𝑆ℎ𝑜𝑟𝑡𝑁𝑜𝑡𝑖𝑐𝑒 is significant, suggesting no difference in risk-shifting for hedge funds with
either leverage usage or short redemption notice period when part of the performance fee option
portfolio is out-of-the-money. The estimated coefficient for 𝐹 × 𝑁𝑒𝑔𝑅𝑒𝑡is negative but
insignificant, suggesting that risk-shifting is mitigated for hedge fund managers who experience
outflows. The estimated coefficient for SecondFlow is positive and significant, suggesting
higher fund flow accompanies higher volatility.
Results in Column (3) and (4) are based on estimating Equation (3) using fund-level fixed
effects rather than style-level fixed effects which alleviates the endogeneity issues that could
arise from unobservable time-invariant variables. The estimated coefficients for NegRet,
𝑀 ×𝑁𝑒𝑔𝑅𝑒𝑡, TimeUnder and 𝑇𝑖𝑚𝑒𝑈𝑛𝑑𝑒𝑟 × 𝑁𝑒𝑔𝑅𝑒𝑡 have the same sign and statistical
significance as those using pooled OLS regressions in Columns (1) and (2). The coefficient
estimate for F Leverage remains negative and significant and the estimate for F ShortNotice
is negative but no longer significant. A number of control variables, including Small, Leverage,
and ShortNotice, AuCorr also lose their significance. Since fixed effects regression removes
cross-sectional variation and relies on within-fund variation only, it does not have much power
for variables that have little variation at fund level (Zhou, 2001), which may explain the loss of
significance of F ShortNotice and the control variables.
In summary, results from the regression of Equation (3) using the no backfill bias sample
are consistent and robust for our Hypothesis 1, 2, 3b, and Hypothesis 4. We also find supporting
evidence for Hypothesis 3a, even though it is not as robust as the other findings.
[Table 3 about here]
3.3 Time spent underwater
Next, we explore in more detail how the duration of underperformance is associated with
changes in a hedge fund’s return volatility during the second half of the year. With a mean of
7.02 months, and a range between 0 and 35 months, TimeUnder is highly skewed with a median
of 2 months, and a 90th
percentile of 24 months. It is therefore possible that the relationship
between risk-shifting and TimeUnder is non-linear. To explore that possibility, we construct
three dummies for funds with partly out-of-the-money performance fee option portfolio and
various length of time being under-water: M_ST is set to one if a fund has M = 1 for less than 12
months and zero otherwise, M_MT is set to one if a fund has M = 1 for between 12 and 24
months and zero otherwise, M_LT is set to one if a fund has M = 1 for over 24 months and zero
otherwise. M_ST, M_MT, and M_LT are more evenly distributed with a mean of 9.98%, 7.90%,
and 6.76%, respectively. We include both M_ST and M_LT, re-estimate Equation (3), and report
estimation results from both fund fixed effects and pooled OLS regressions in Columns (1) and
(2) (M_MT serves as the base case) of Table 4, respectively. The coefficient for M_ST is negative
(and statistically significant for the regression estimates using fund fixed effects) and the
coefficient for M_LT is positive and significant generally. The difference between the
coefficient for M_ST and M_LT suggests that for funds experiencing positive returns, a longer
duration of poor performance leads to greater volatility of returns. The estimate for 𝑀_𝑆𝑇 ×
𝑁𝑒𝑔𝑅𝑒𝑡 is positive and significant, with a large magnitude, more than two times that for NegRet
(1.435 vs 0.447 in Column (1)). Therefore, a fund manager tends to be more aggressive in risk-
shifting after poor first half-year performance if the fund’s NAV has been less than the HWM for
a year or less. When the duration of underperformance extends over to more than one year,
hedge fund managers have less risk-shifting, supporting Hypothesis 4.17
[Table 4 about here]
4. Additional tests
4.1 Alternative performance measures
In this section we consider measures of the performance of the hedge fund relative to
other funds. Specifically, we investigate whether and how M and F influence hedge fund risk-
shifting decisions when poor first half-year performance (𝑃𝑒𝑟𝑓) is measured either by the rank
of absolute performance among all hedge funds (LowAbsRetRank) or by the rank of relative
within-style performance (LowRelRetRank) and focus on the coefficient estimates on the
interaction terms 𝑀 × 𝑃𝑒𝑟𝑓, 𝐹 × 𝑃𝑒𝑟𝑓, 𝑇𝑖𝑚𝑒 × 𝑃𝑒𝑟𝑓.
Panel A of Table 5 reports the results from the estimation of Equation (3) with the two
alternative performance measures. The findings are qualitatively similar to those in Table 3,
suggesting larger magnitudes of tournament-like risk-shifting due to already out-of-the-money
performance fee option. These findings, again, show that Moneyness and Flow are important
factors for risk-shifting related to tournament incentives as well.
17
An alternative explanation is related to different behavior for defunct funds. We explore whether the higher
volatility change at funds with long-time being under-water is related to the chaos before a fund ceases operation by
investigating the risk-shifting behavior in the final year of funds that are in the defunct funds database. We do not
find differences from the funds that are in the live funds database. Since defunct fund database includes both funds
that liquidate and cease operation and those stop reporting to the database, it is highly likely that we do not have
return data for the actual final year of a fund that ceases operation.
Panel B of Table 5 reports the impact of duration of underperformance on risk-shifting
using the two alternative performance measures, LowAbsRetRank, and LowRelRetRank and the
results are similar: the estimates for 𝑀_𝑆𝑇 × 𝐿𝑜𝑤𝐴𝑏𝑠𝑅𝑒𝑡𝑅𝑎𝑛𝑘 and 𝑀_𝑆𝑇 × 𝐿𝑜𝑤𝑅𝑒𝑙𝑅𝑒𝑡𝑅𝑎𝑛𝑘
are positive and significant yet insignificant for the interaction terms between 𝑀_𝐿𝑇 and the two
alternative performance measures. Our findings indicate as before that a fund shifts risk by more
in response to poor first half-year performance if the duration of poor performance is less than
one year.
[Table 5 about here]
4.2 Subsample analysis and alternative cutoffs for M and F
To examine further details for the impact of outflows on the propensity to risk-shifting
for a fund with out-of-the-money performance fee option, we analyze a subsample of hedge
funds with 1M and free of backfill bias, and report the results in Table 6. We conduct
robustness checks using different thresholds for setting the values of M and F equal to one.18
There are three thresholds (Moneyness <1, or 0.9, or 0.8) that are used to set M = 1 and another
three thresholds (Flow<0, or <‒0.1%, or <‒0.2%) that are used to define F = 1.19
For example,
the results reported in Column 2 are obtained by examining only those funds where moneyness is
less than or equal to 0.9 and when Flow is less than or equal to 0.
The results in Table 6 show that, first, the coefficients for NegRet are significant and
positive for all cases. Therefore, regardless of the definition for M = 1 and F = 1, a fund tends to
increase its return volatility following poor first half-year performance. Second, the coefficients
for the interaction term 𝐹 × 𝑁𝑒𝑔𝑅𝑒𝑡 are negative and significant except for the cases in which
Moneyness is defined to be a ratio of NAV to HWM less than 1. Therefore, for the subsample
funds with M=1, fund outflows mitigate risk-shifting for hedge funds unless the performance fee
option is relatively close to be at-the-money. Table 6 provides further support for our Hypotheses
2.
[Table 6 about here]
18
In robustness checks with various definitions of M using changing thresholds for Moneyness, we set TimeUnder to
0 whenever M=0. 19
Note that if Moneyness is equal to 1, it means that HWM is achieved at the end of the 36-month window and the
performance fee option is at-the-money.
We also use the entire no-backfill bias sample to conduct additional tests for robustness
of our findings to the thresholds assumed for Moneyness and Flow, which define M and F. We
vary these thresholds for M and F and report results based on the same nine combinations as
those in Table 6. The results of the interactions of these nine cutoffs with poor performance in
the first half-year are qualitatively the same.20
4.3 Discussion
Brown et al. (1996) and Brown et al. (2001) both use contingency table tests to
investigate how the number of mutual/hedge fund managers with above- and below-median
style-adjusted variance ratios in the second half-year vary with their performance in the first half-
year. They both find supporting evidence for risk-shifting due to tournament-like behavior.
Brown et al. (2001) do not find evidence that hedge fund managers increase volatility in the
second half-year after negative return in the first half-year for their sample over the period of
1989-1995. We reproduce the contingency table test using our sample of data over the period of
1997-2012 and present results for all funds, funds with M = 1 (funds with more than 10%
performance decline since achieving HWM), and funds with M = 0 (funds with less than 10%
performance decline since achieving HWM), respectively, in Table 7. The Chi-square statistic
is statistically significant for the group of all funds across the three performance measures
(NegRet, LowAbsRetRank, and LowRelRetRank), providing support for risk-shifting in hedge
funds with respect to convex payoff due to both asymmetric performance fee option and
asymmetric flow-performance relation. Despite the smaller number of funds in the group of all
funds with M = 1, this group has the strongest tendency to engage in risk-shifting after poor first
half-year performance. However, when the performance measure is NegRet, which suggests the
fund has a negative return in the first half-year, there is no evidence of risk-shifting for the group
of funds with M = 0. This is consistent with the findings in Brown et al. (2001) which show that,
despite of their asymmetric compensation contracts, hedge funds do not react by increasing
volatility after negative first half-year returns. Our results show that this is indeed the case for
hedge funds with close to at-the-money compensation option at the beginning of the year.
However, when hedge fund managers have a large amount of out-of-the-money compensation
options at the beginning of the year, they are more aggressive in risk-shifting. Hedge funds
20
These results are available upon request from the authors.
generally performed well during the 1990s (Ackermann et al., 1999) and, thus, we expect that
most of the funds in the Brown et al. (2001) sample to have M = 0.
[Table 7 about here]
Chen (2013) shows that hedge funds that use derivatives have smaller risk-shifting
compared to those that do not. Since when hedge funds report to the database that they use
leverage, they could mean the usage of futures, swaps and other derivatives besides bank or
primary brokerage credit, it is plausible that our results on the lower volatility change for funds
that use leverage conditional on outflow may be due to their usage of derivatives instead. To
examine that possibility, we exclude hedge funds that use derivatives and re-estimate Equation
(3). Our results remain the same, and the coefficient for 𝐹 × 𝐿𝑒𝑣𝑒𝑟𝑎𝑔𝑒is negative and
significant, with a similar magnitude (t-stat= -2.69).21
4.4 More robustness tests
The results reported so far are based on the HWM observed over a rolling three-year
window starting from Jan 1, 1994, where we treat the maximum NAV over such a window as a
proxy for HWM of a typical investor. We relax the requirement of a three-year window and use
the maximum NAV as a proxy for HWM for a fund with at least four years of return history over
a period which starts from Jan 1, 1994 and ends right before the beginning of the evaluation year.
This maximum NAV can be viewed as the maximum HWM of a hedge fund’s investors starting
after Jan 1, 1994 at the fund. Using this alternative definition of HWM, we re-calculate summary
statistics and re-estimate Equation (3) with the three performance measures (NegRet,
LowAbsRetRank, and LowRelRetRank); we find our main results remain unchanged. Our
findings are also robust to using the full sample of hedge funds. 22
We explore whether the fund’s moneyness and flow conditions are associated with the
flow-performance relationship examined in Chevalier and Ellison (1997) and Sirri and Tufano
(1998). We find that the flow-performance relation documented by previous papers is weaker for
hedge funds with partly out-of-the-money performance fee option portfolio.23
We may
hypothesize that managers are aware of the weaker relation when they decide whether to
21
These results are available from the authors upon request. 22
These results are available upon request. 23
These results are available upon request.
implement a risk-shifting policy. If the performance-flow relationship is indeed weaker, the cost
associated with potential outflows will be lower, and, therefore, the manager may choose a more
aggressive risk-shifting policy.24
5. Conclusion
In response to Ross (2004)’s call, we show that time-varying performance- and flow-
related conditions influence hedge funds’ risk-shifting behavior. We also show that hedge fund
managers adjust their risk-taking in a way that is consistent with their fund characteristics. We
find that the most recent half-year performance and cumulative past performance influences risk-
shifting as does the direction of fund flows. We also show that risk-shifting behavior is driven
to a great extent by funds that whose performance fee option portfolio has been partly out-of-the-
money for a short period. The use of leverage or short redemption notice periods reduces risk
shifting for funds that are experiencing outflows.
Our results help depict how hedge funds use volatility strategies that are based on
performance, flow history, and fund characteristics over time. Our findings highlight the
importance of the explicit incentive for risk-shifting due to the convex payoff structure of
compensation contracts for hedge fund managers, an issue of interest to the investment
community and to scholars of corporate finance. More research is needed, however, especially to
fully understand the relations between risk-taking and various combinations of flow and
performance, as well as how leverage usage is related to risk-taking.
24
In a related study but a different context, Del Guercio and Tkac (2002) show that pension fund managers have
much less tournament-like risk-shifting incentives due to much less convex flow-performance relation.
References
Ackermann, C., R. McEnally, and D. Ravenscraft, 1999. The performance of hedge funds:
risk, return, and incentives. Journal of Finance 54, 833-874.
Adams, R.B. and D. Ferreira, 2009. Women in the boardroom and their impact on corporate
governance and performance. Journal of Financial Economics 94, 291-309.
Agarwal, V., N. Daniel, and N. Y. Naik, 2002. Determinants of money-flow and risk-taking
in the hedge fund industry. Unpublished working paper, Georgia State University.
Agarwal, V., N. Daniel, and N. Y. Naik, 2009, Role of managerial incentives and discretion in
hedge fund performance. Journal of Finance 64, 2221-2256.
Ang, A., S. Gorovyy, G.B. van Inwegen, 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., B. Liang, H. Park, 2014, Onshore and offshore hedge funds: are they twins?
Management Science 60, 74-91
Aragon, G. O. and V. Nanda, 2012, Tournament behavior in hedge funds: High water marks,
fund liquidation, and managerial stake. Review of Financial Studies 25, 937-974
Basak, S., A. Pavlova, and A. Shapiro, 2007, Optimal Asset Allocation and Risk Shifting in
Money Management. Review of Financial Studies 20, 1583-1621
Barber, B., T. Odean, and L. Zheng, 2004, Out of sight, out of mind: the effects of expenses
on mutual fund flows. Journal of Business 78, 2095-2120.
Brown, S., W. Goetzmann, and J. Park, 2001, Careers and survival: Competition and risk in
the hedge fund and CTA industry. Journal of Finance 56, 1869-1886
Brown, K. C., W. V. Harlow, and L. T. Starks, 1996, Of tournaments and temptations: an
analysis of managerial incentives in the mutual fund industry. Journal of Finance 51, 85-110
Carpenter, J., Does option compensation increase managerial risk appetite. 2000, Journal of
Finance 55, 2311-2331
Chevalier, J., and G. Ellison, 1997, Risk taking by mutual funds as a response to incentives.
Journal of Political Economy 105, 1167-2000
Cumming, D. and N. Dai, 2010, Hedge fund regulation and misreported returns. European
Financial Management 16, 829-857
Del Guercio, D. and P. Tkac, 2002, The determinants of the flow of funds of managed
portfolios: mutual funds versus pension funds, Journal of Financial and Quantitative Analysis 37,
523-557.
Drechsler, I. 2014, Risk choice under high-water marks. forthcoming, Review of Financial
Studies, doi: 10.1093/rfs/hht081.
Ferson, W. and V. Warther, 1996, Evaluating Fund Performance in a Dynamic Market.
Financial Analyst Journal 52, 20-28.
Frazzini, A. and O. A. Lamont, 2008, Dumb money: mutual fund flows and the cross-section
of stock returns, Journal of Financial Economics 88, 299-322.
Fung, W. and D. A. Hsieh, 1997, Empirical characteristics of dynamic trading strategies: The
case of hedge funds, Review of Financial Studies 10, 275-302
Fung, W. and D. A. Hsieh, 2000, Performance Characteristics of Hedge Funds and CTA
Funds: Natural versus Spurious Biases, Journal of Financial and Quantitative Analysis 35, 291-
307
Getmansky, M., A. Lo, and I. Makarov, 2004, An econometric model of serial correlation and
illiquidity in hedge fund returns, Journal of Financial Economics 74, 29-610
Getmansky, M., B. Liang, C. Schwarz, and R. Wermers, 2011, Share restrictions and investor
flows in the hedge fund industry. Working paper, University of Massachusetts, Amherst
Goetzmann, W. N., J. E. Jr. Ingersoll, and S. A. Ross, 2003, High-water marks and hedge
fund management contracts, The Journal of Finance 58, 1685-1717
Hodder, J.E. and J. C. Jackwerth, 2007, Incentive contracts and hedge fund management,
Journal of Financial and Quantitative Analysis 42, 811-26
Hu, P., J. R. Kale, M. Pagani, and A. Subramanian, 2011, Fund flows, performance,
managerial career concerns, and risk taking, Management Science 57, 628-646
Jain, P. and J. Wu, 2000, Truth in mutual fund advertising: evidence on future performance
and fund flows. Journal of Finance 55, 937-958.
Koski, J. L. and J. Pontiff, 1999, How are derivatives used? Evidence from the mutual fund
industry. Journal of Finance 54, 791-816.
Lan, Y., N. Wang, and J. Yang, 2013, The economics of hedge funds, Journal of Financial
Economics 110, 300-323
Li, Y. and H. B. Kazemi, 2007, Conditional properties of hedge funds: Evidence from daily
return, European Financial Management 13(2), pp.211-238
Liang, B., Hedge funds: The living and the dead, 2000, Journal of Financial and Quantitative
Analysis 35, pp.309-326
Oikonomou, I., C. Brooks, and S. Pavelin, 2012.The Impact of Corporate Social Performance on
Financial Risk and Utility: A Longitudinal Analysis. Financial Management 41, 483-515.
Panageas, S. and M. Westerfield, 2009, High water marks: High risk appetites? Convex
compensation, long horizons, and portfolio choice, Journal of Finance 64, 1-36
Ray, S., 2012, The downside risk of high water marks: an empirical study, Journal of
Investment Management, Second Quarter
Ross, S. A., 2004, Compensation, incentives, and the duality of risk aversion and riskiness,
Journal of Finance 59, 207-225.
Sirri, E. and P. Tufano, 1998, Costly search and mutual fund flows, Journal of Finance 53,
1589-1622.
Zhou, X., 2001, Understanding the determinants of managerial ownership and the link
between ownership and performance: comment, Journal of Financial Economics 62, 559-571.
Appendix. Definition of Variables
PStd Fund volatility calculated using monthly returns in the first half of an evaluation year
𝑃𝑆𝑡𝑑𝑖 = 𝜎𝑡,𝑡+6 ∗ 100
NStd Fund volatility calculated using monthly returns in the second half of an evaluation
year. 𝑁𝑆𝑡𝑑𝑖 = 𝜎𝑡+7,𝑡+12 ∗ 100
DifStd (NStd-PStd) *100, dependent variable
HWM
𝑡𝑖,𝐻𝑊𝑀
High-water mark proxy as of the beginning of the evaluation year based on the
previous 36-month window. 𝐻𝑊𝑀𝑖,𝑡 = max(𝑁𝐴𝑉𝑖,𝑡−𝜏; 0 ≤ 𝜏 < 36) The month when hedge fund 𝑖 reaches its HWM.
Moneyness
, , ,/
i t i t i tMoneyness NAV HWM , where 𝑁𝐴𝑉𝑖,𝑡 is net asset value per hedge fund share
at the beginning of the evaluation year.
Flow 𝐹𝑙𝑜𝑤𝑖,𝑡+6 =𝐴𝑈𝑀𝑖,𝑡+6−(1+𝐴𝑅𝑖,𝑡𝐻𝑊𝑀,𝑡+6)×𝐴𝑈𝑀𝑖,𝑡𝐻𝑊𝑀
𝐴𝑈𝑀𝑖,𝑡𝐻𝑊𝑀
where 𝐴𝑈𝑀 is the asset under
management and 𝐴𝑅𝑖,𝑡𝐻𝑊𝑀,𝑡+6 is the accumulative return between 𝑡𝐻𝑊𝑀 and 𝑡 + 6,
(June 30th of the evaluation year). Flow is a proxy for the net flow from the month a
fund achieves its HWM to Jun 30 of the evaluation year. M A dummy variable that equals 1 if Moneyness < 0.9 and 0 otherwise.
M_ST A dummy variable that equals 1 if M=1 and TimeUnder is less than 1 year and 0
otherwise.
M_MT A dummy variable that equals 1 if M=1 and TimeUnder is more than 1 year and less
than 2 years. It is 0 otherwise.
M_LT A dummy variable that equals 1 if M=1 and TimeUnder is longer than 2 years and 0
otherwise.
F A dummy variable that equals 1 if Flow < 0 and 0 otherwise.
NegRet A dummy variable that equals 1 if cumulative return over the first half of the
evaluation year is negative and 0 otherwise
LowAbsRetRank
LowRelRetRank
Leverage
Small
A dummy variable that equals 1 if accumulative return over the first half of the
evaluation year is ranked below median among all funds and 0 otherwise
A dummy variable that equals 1 if accumulative return over the first half of the
evaluation year is ranked below median among funds within a given style and 0
otherwise
A dummy variable that equals 1 if the fund reports that it uses leverage and 0
otherwise
A dummy variable that equals 1 if the fund’s asset under management is below-
median and 0 otherwise.
ShortNotice A dummy variable that equals 1 if a fund has a short redemption notice that is less
than or equal to 30 days and 0 otherwise.
TimeUnder
SecondFlow
The number of months a fund has been under its maximum NAV in a three-year
window
Net flow to the fund from Jul 1 to December 31 of the evaluation year.
AuCorr First-order autocorrelation coefficient on a fund's return in a 3-year historical return
window.
Table 1. Summary Statistics of Hedge Funds Sample, 1994 - 2012
Summary statistics are reported for the hedge funds sample from the Morningstar-CISDM
database. We include only funds that have returns in Morningstar-CISDM after January 1994 to
minimize survivorship bias. To be included in the sample, each fund must have at least four
years of return history, information on assets under management, management fee and
performance fee rates, strategy, as well as returns for the entire four-year window. The sample is
winsorized at 1% and 99% values of DifStd, Moneyness and Flow to minimize the impact from
outliers. The Appendix provides detailed definition for each variable.
Panel A. Hedge fund counts and fund size by evaluation year.
Window Evaluation
Year
Frequency Percent Min 25th
Pctl
50th
Pctl
75th
Pctl
Max Mean
94-96 1997 74 0.58 0.20 18.00 43.94 90.40 2818.00 131.83
95-97 1998 122 0.99 0.20 20.05 58.60 120.37 3054.00 153.59
96-98 1999 185 1.68 0.20 15.00 45.93 111.40 3054.00 131.07
97-99 2000 289 2.43 0.20 16.00 48.00 132.10 3054.00 146.74
98-00 2001 406 3.37 0.20 17.36 45.57 139.00 3054.00 153.74
99-01 2002 537 4.64 0.20 17.30 49.20 140.03 2580.27 150.35
00-02 2003 730 5.96 0.20 14.80 47.10 120.00 2829.00 131.05
01-03 2004 924 7.48 0.20 19.08 60.42 166.70 2877.00 167.64
02-04 2005 1132 9.01 0.20 19.42 65.00 198.05 3054.00 193.01
03-05 2006 1267 10.3 0.20 18.77 66.67 199.80 3054.00 196.45
04-06 2007 1294 10.62 0.20 18.80 65.00 210.39 3054.00 209.95
05-07 2008 1155 10.19 0.20 21.00 72.49 237.50 3054.00 234.92
06-08 2009 1071 9.25 0.20 13.00 41.33 131.54 3054.00 166.30
07-09 2010 1004 8.55 0.20 15.60 46.00 141.00 3054.00 162.05
08-10 2011 941 8.02 0.28 17.81 50.76 140.56 3054.00 163.93
09-11 2012 805 6.92 0.28 15.25 45.39 136.79 3054.00 155.81
94-12 11934 100% 0.20 17.40 54.19 164.80 3054.00 177.39
Panel B. Style Distribution
No Backfill-Bias sample Full sample
Style Counts
Convertible Arbitrage 275 398
Debt-related 989 1524
Equity-related 5072 7387
Event Driven 558 765
Global Macro 544 795
Merger Arbitrage 126 154
Multi-strategy 604 861
Volatility 31 50
Panel C. Summary Statistics for No backfill Bias Sample.
No back Fill Sample M = 1 F = 1 1M F
Variable N Mean Std N Mean Std N Mean Std N Mean Std
PStd 8193 3.34 3.03 2014 5.40 3.94 4496 3.51 3.15 1514 5.29 3.91
NStd 8194 3.42 3.60 2015 4.75 4.62 4497 3.47 3.51 1515 4.60 4.27
𝐷𝑖𝑓𝑆𝑡𝑑 8193 0.08 3.12 2014 -0.64 4.42 4496 -0.05 3.08 1514 -0.69 4.14
𝑀 8199 0.25 0.43 2020 1.00 0.00 4502 0.34 0.47 1520 1.00 0.00
𝐹 8199 0.55 0.50 2020 0.75 0.43 4502 1.00 0.00 1520 1.00 0.00
M F � 8199 0.19 0.39 2020 0.75 0.43 4502 0.34 0.47 1520 1.00 0.00
𝑇𝑖𝑚𝑒𝑈𝑛𝑑𝑒𝑟 8199 7.02 9.80 2020 17.26 9.81 4502 9.54 10.75 1520 17.84 9.79
𝑆𝑚𝑎𝑙𝑙 8199 0.50 0.50 2020 0.62 0.48 4502 0.55 0.50 1520 0.61 0.49
𝐿𝑒𝑣𝑒𝑟𝑎𝑔𝑒 5741 0.77 0.42 1341 0.75 0.43 3175 0.78 0.42 1019 0.77 0.42
𝑆ℎ𝑜𝑟𝑡𝑁𝑜𝑡𝑖𝑐𝑒 8199 0.63 0.48 2020 0.67 0.47 4502 0.64 0.48 1520 0.66 0.47
𝐴𝑢𝐶𝑜𝑟𝑟 8193 0.14 0.22 2014 0.19 0.22 4496 0.16 0.22 1514 0.20 0.22
Panel D. Summary Statistics for Full Sample. Full Sample M = 1 F = 1 1M F
Variable N Mean Std N Mean Std N Mean Std N Mean Std
PStd 11925 3.34 3.13 2800 5.45 4.18 6171 3.49 3.20 2028 5.25 3.91
NStd 11926 3.41 3.62 2801 4.84 4.66 6172 3.47 3.56 2029 4.63 4.28
𝐷𝑖𝑓𝑆𝑡𝑑 11925 0.08 3.16 2800 -0.61 4.45 6171 -0.03 3.12 2028 -0.62 4.10
𝑀 11934 0.24 0.42 2809 1.00 0.00 6180 0.33 0.47 2037 1.00 0.00
𝐹 11934 0.52 0.50 2809 0.73 0.45 6180 1.00 0.00 2037 1.00 0.00
M F � 11934 0.17 0.38 2809 0.73 0.45 6180 0.33 0.47 2037 1.00 0.00
𝑇𝑖𝑚𝑒𝑈𝑛𝑑𝑒𝑟 11934 6.61 9.51 2809 17.03 9.83 6180 9.00 10.42 2037 17.39 9.75
𝑆𝑚𝑎𝑙𝑙 11934 0.50 0.50 2809 0.62 0.49 6180 0.55 0.50 2037 0.61 0.49
𝐿𝑒𝑣𝑒𝑟𝑎𝑔𝑒 8012 0.78 0.41 1761 0.76 0.43 4191 0.79 0.41 1294 0.78 0.42
𝑆ℎ𝑜𝑟𝑡𝑁𝑜𝑡𝑖𝑐𝑒 11934 0.63 0.48 2809 0.67 0.47 6180 0.63 0.48 2037 0.66 0.48
𝐴𝑢𝐶𝑜𝑟𝑟 11925 0.14 0.22 2800 0.18 0.22 6171 0.15 0.22 2028 0.20 0.22
Panel E. Distribution of Leverage Usage and Short Redemption Period.
Full Sample No Backfill Sample
Leverage ShortNotice Leverage ShortNotice
0 1 0 1 0 1 0 1
All 1747 6265 4470 7464 1338 4403 3036 5163
1M 428 1333 925 1884 330 1011 658 1362
0M 1319 4932 3545 5580 1008 3392 2378 3801
1F 898 3293 2266 3914 714 2461 1625 2877
0F 849 2972 2204 3550 624 1942 1411 2286
M F =1 289 1005 700 1337 235 784 519 1001
0M F 1458 5260 3770 6127 1103 3619 2517 4162
Table 2. Correlation Matrix with P-values in Parentheses.
Pairwise correlation coefficients are reported for our main variables in regression analyses for the no backfill bias sample. There are 8199 individual hedge fund-
year observations over the period of 1997–2012 in the no backfill bias sample. The Appendix provides detailed definition for each variable. MF, M×NegRet,
and F×NegRet represent interaction between Moneyness and Flow, Moneyness and NegRet, as well as Flow and NegRet, respectively.
1 2 3 4 5 6 7 8 9 10 11 12 13
1 DifStd 1.00
2 M -0.13 1.00
[<.00]
3 F -0.04 0.23 1.00
[<.00] [<.00]
4 MF -0.12 0.83 0.43 1.00
[<.00] [<.00] [<.00]
5 NegRet 0.10 0.09 0.08 0.07 1.00
[<.00] [<.00] [<.00] [<.00]
6 M×NegRet 0.06 0.54 0.12 0.44 0.49 1.00
[<.00] [<.00] [<.00] [<.00] [<.00]
7 F×NegRet 0.07 0.13 0.41 0.22 0.73 0.46 1.00
[<.00] [<.00] [<.00] [<.00] [<.00] [<.00]
8 TimeUnder -0.03 0.60 0.28 0.53 0.13 0.35 0.18 1.00
[0.00] [<.00] [<.00] [<.00] [<.00] [<.00] [<.00]
9 Small 0.01 0.14 0.10 0.11 0.07 0.12 0.08 0.16 1.00
[0.19] [<.00] [<.00] [<.00] [<.00] [<.00] [<.00] [<.00]
10 Leverage 0.01 -0.02 0.02 0.00 0.02 -0.02 0.02 -0.02 -0.07 1.00
[0.67] [0.20] [0.10] [0.84] [0.21] [0.13] [0.07] [0.12] [<.00]
11 ShortNotice 0.00 0.05 0.02 0.03 0.07 0.05 0.06 0.06 0.13 -0.07 1.00
[0.65] [<.00] [0.05] [0.01] [<.00] [<.00] [<.00] [<.00] [<.00] [<.00]
12 PStd -0.32 0.39 0.06 0.31 0.15 0.22 0.11 0.21 0.15 -0.04 0.10 1.00
[<.00] [<.00] [<.00] [<.00] [<.00] [<.00] [<.00] [<.00] [<.00] [0.01] [<.00]
13 AuCorr -0.04 0.11 0.07 0.13 -0.05 0.02 -0.01 0.06 -0.12 0.01 -0.15 -0.04 1.00
[0.00] [<.00] [<.00] [<.00] [<.00] [0.03] [0.27] [<.00] [<.00] [0.28] [<.00] [<.00]
Table 3. Regressions with Moneyness and Fund Flows: No Backfill Bias Sample
This table reports pooled OLS and fund fixed effects (FE) estimates from Equation (3), the regression of mid-year
change in fund volatility on the fund’s negative return in the first half-year conditional on the past performance and
fund flow and fund characteristics using the no backfill bias sample. The no backfill bias sample includes a total of
8199 individual hedge fund-year observations (7765 observations when we include SecondFlow as a control,
representing 1625 individual funds) over the period of 1997–2012 and 5741 of the observations (5497 observations
when we include SecondFlow as a control, representing 1058 individual funds) have information on leverage usage.
The Appendix provides detailed definition for each variable. Numbers in parentheses are t-statistics based on robust
standard errors clustered at fund levels. Columns (1) – (2) report results from pooled OLS regressions and Columns
(3) – (4) report results from a fund fixed effects regression. The symbols ***, **, and * denote significance levels of
1%, 5% and 10%, respectively.
(1) (2) (3) (4)
VARIABLES OLS OLS Fund FE Fund FE
M -0.198 -0.279 -0.484*** -0.398
(-1.289) (-0.901) (-3.239) (-1.318)
F 0.036 0.621*** 0.078 0.335*
(0.480) (3.355) (1.044) (1.908)
NegRet 0.483*** 0.664*** 0.503*** 0.535***
(3.682) (4.182) (3.699) (3.148)
𝑀 × 𝑁𝑒𝑔𝑅𝑒𝑡 1.287*** 1.340*** 1.360*** 1.447***
(4.703) (3.872) (5.041) (4.314)
𝐹 × 𝑁𝑒𝑔𝑅𝑒𝑡 -0.123 -0.287 -0.161 -0.205
(-0.657) (-1.211) (-0.864) (-0.881)
𝑀 × 𝑆𝑚𝑎𝑙𝑙 0.357** 0.316 -0.009 -0.098
(2.044) (1.598) (-0.050) (-0.454)
𝑀 × 𝐿𝑒𝑣𝑒𝑟𝑎𝑔𝑒 0.069 -0.041
(0.265) (-0.155)
𝑀 × 𝑆ℎ𝑜𝑟𝑡𝑁𝑜𝑡𝑖𝑐𝑒 0.014 -0.162
(0.070) (-0.766)
𝐹 × 𝑆𝑚𝑎𝑙𝑙 -0.330*** -0.241* -0.133 -0.103
(-2.639) (-1.692) (-1.047) (-0.718)
𝐹 × 𝐿𝑒𝑣𝑒𝑟𝑎𝑔𝑒 -0.504*** -0.343**
(-3.243) (-2.052)
𝐹 × 𝑆ℎ𝑜𝑟𝑡𝑁𝑜𝑡𝑖𝑐𝑒 -0.289** -0.036
(-2.280) (-0.281)
TimeUnder 0.020*** 0.023*** 0.021*** 0.026***
(4.118) (4.514) (3.949) (4.223)
𝑇𝑖𝑚𝑒𝑈𝑛𝑑𝑒𝑟 × 𝑁𝑒𝑔𝑅𝑒𝑡 -0.023** -0.025* -0.023** -0.021
(-2.146) (-1.852) (-1.990) (-1.473)
Small 0.354*** 0.296*** 0.145 0.039
(3.598) (2.611) (1.066) (0.239)
Leverage 0.229* 0.311
(1.922) (0.328)
ShortNotice 0.232** -1.825
(2.186) (-1.507)
AuCorr -0.250* -0.386** -0.199 -0.343
(-1.687) (-2.345) (-0.919) (-1.487)
SecondFlow 0.208 0.443*** 0.237* 0.306*
(1.442) (2.721) (1.670) (1.854)
PStd -0.367*** -0.351*** -0.725*** -0.731***
(-18.984) (-15.592) (-24.296) (-20.129)
Constant 1.325*** 0.942*** 2.963*** 3.670***
(4.809) (2.669) (9.050) (3.108)
Observations 7,765 5,497 7,765 5,497
R-squared 0.299 0.281 0.439 0.423
Style FE Yes Yes
Year FE Yes Yes Yes Yes
Fund FE Yes Yes
Number of fundID 1,625 1,058
Table 4. The Impact of Duration of Underperformance: No Backfill Bias Sample
This table focuses on the impact of duration of underperformance and reports pooled OLS and fund fixed effects
(FE) estimates from Equation (3), the regression of mid-year change in fund volatility on the fund’s poor
performance in the first half-year for the no backfill sample. The performance measure is NegRet. The sample
includes a total of 8199 individual hedge fund-year observations over the period of 1997–2012 and 7765 of the
observations have information on SecondFlow, representing 1625 individual funds. Instead of M and TimeUnder,
which is the number of months a fund has been below its HWM, three dummy variables, M_ST, M_MT and M_LT
are used to capture the possible nonlinear relationship between DifStd and NegRet conditional on M and different
amount of TimeUnder. M_ST is set to 1 if M=1 and TimeUnder is less than 12 months and 0 otherwise; M_MT
equals 1 if M=1 and TimeUnder is between12 months and 24 months and 0 otherwise; M_LT equals 1 if M=1 and
TimeUnder is over 24 months and 0 otherwise. The Appendix provides definitions for each variable. Numbers in
parentheses are t-statistics based on robust standard errors clustered at fund levels. The symbols ***, **, and *
denote significance levels of 1%, 5% and 10%, respectively.
(1) (2)
VARIABLES OLS Fund FE
M_ST -0.024 -0.417**
(-0.160) (-2.532)
M_MT
M_LT 0.958*** 0.534**
(4.534) (2.189)
F -0.101 0.026
(-1.614) (0.426)
NegRet 0.470*** 0.477***
(3.340) (3.398)
𝑀_𝑆𝑇 × 𝑁𝑒𝑔𝑅𝑒𝑡 1.175*** 1.435***
(3.491) (3.802)
𝑀_𝑀𝑇 × 𝑁𝑒𝑔𝑅𝑒𝑡
𝑀_𝐿𝑇 × 𝑁𝑒𝑔𝑅𝑒𝑡 0.663 0.544
(1.534) (1.241)
𝐹 × 𝑁𝑒𝑔𝑅𝑒𝑡 -0.143 -0.144
(-0.810) (-0.819)
Small 0.267*** 0.057
(4.351) (0.474)
AuCorr -0.220 -0.175
(-1.475) (-0.812)
SecondFlow 0.163 0.222
(1.133) (1.550)
PStd -0.363*** -0.730***
(-19.121) (-24.323)
Constant 1.350*** 3.032***
(4.886) (9.215)
Observations 7,765 7,765
R-squared 0.299 0.440
Number of fundID 1,625
Fund FE Yes
Year FE Yes Yes
Style FE Yes
Table 5. Regression Results with Alternative Performance Measures: No Backfill Bias Sample
This table reports results using alternative performance measures in estimation of Equation (3), the regression of
mid-year change in fund volatility on the fund’s alternative poor performance measures in the first half-year
conditional on the performance and fund flow and fund characteristics for the no backfill sample. The sample
includes a total of 8199 individual hedge fund-year observations over the period of 1997–2012 yet only 5497 of the
observations have information on both Leverage and SecondFlow. The two alternative poor performance measures
are LowAbsRetRank and LowRelRetRank. Panel A focuses on the impact of M and F and Panel B focuses on the
impact of duration of underperformance. The Appendix provides detailed definition for each variable. Numbers in
parentheses are t-statistics based on robust standard errors clustered at fund levels. The symbols ***, **, and *
denote significance levels of 1%, 5% and 10%, respectively.
Panel A The Impact of Moneyness and Flow with Alternative Performance Measures
(1) (2) (3) (4)
Alt-Perf= Alt-Perf= Alt-Perf= Alt-Perf=
LowAbsRetRank LowAbsRetRank LowRelRetRank LowRelRetRank
VARIABLES Fund FE OLS Fund FE OLS
M -0.447 -0.476 -0.437 -0.489
(-1.465) (-1.477) (-1.346) (-1.451)
F 0.325* 0.551*** 0.320 0.621***
(1.712) (2.831) (1.623) (3.025)
Perf 0.394*** 0.300*** 0.315*** 0.275**
(3.444) (2.850) (2.706) (2.579)
𝑀 × 𝑃𝑒𝑟𝑓 1.136*** 1.328*** 1.059*** 1.274***
(4.265) (5.045) (3.577) (4.729)
𝐹 × 𝑃𝑒𝑟𝑓 -0.078 0.003 -0.061 -0.125
(-0.480) (0.020) (-0.379) (-0.798)
TimeUnder 0.031*** 0.034*** 0.024*** 0.027***
(3.954) (4.971) (3.180) (4.126)
𝑇𝑖𝑚𝑒𝑈𝑛𝑑𝑒𝑟 × 𝑃𝑒𝑟𝑓 -0.025** -0.036*** -0.009 -0.021**
(-2.064) (-3.341) (-0.760) (-1.994)
𝑀 × 𝑆𝑚𝑎𝑙𝑙 -0.108 0.315 -0.111 0.324
(-0.503) (1.583) (-0.511) (1.629)
𝑀 × 𝐿𝑒𝑣𝑒𝑟𝑎𝑔𝑒 -0.037 0.072 0.017 0.117
(-0.136) (0.274) (0.062) (0.442)
𝑀 × 𝑆ℎ𝑜𝑟𝑡𝑁𝑜𝑡𝑖𝑐𝑒 -0.170 0.008 -0.171 0.029
(-0.798) (0.041) (-0.796) (0.142)
𝐹 × 𝑆𝑚𝑎𝑙𝑙 -0.110 -0.275* -0.104 -0.268*
(-0.749) (-1.886) (-0.712) (-1.839)
𝐹 × 𝐿𝑒𝑣𝑒𝑟𝑎𝑔𝑒 -0.322* -0.482*** -0.329* -0.484***
(-1.899) (-3.077) (-1.928) (-3.062)
𝐹 × 𝑆ℎ𝑜𝑟𝑡𝑁𝑜𝑡𝑖𝑐𝑒 -0.053 -0.292** -0.042 -0.296**
(-0.406) (-2.255) (-0.319) (-2.284)
Small 0.063 0.318*** 0.047 0.311***
(0.386) (2.769) (0.289) (2.732)
Leverage 0.391 0.221* 0.393 0.231*
(0.405) (1.864) (0.408) (1.932)
ShortNotice -1.752 0.229** -1.805 0.229**
(-1.435) (2.131) (-1.453) (2.149)
AuCorr -0.377 -0.418** -0.405* -0.434***
(-1.632) (-2.534) (-1.733) (-2.611)
SecondFlow 0.315* 0.403** 0.325* 0.402**
(1.882) (2.444) (1.895) (2.393)
PStd -0.718*** -0.335*** -0.722*** -0.338***
(-19.902) (-14.836) (-20.145) (-15.150)
Constant 3.360*** 0.829** 3.474*** 0.873**
(2.824) (2.260) (2.868) (2.389)
Observations 5,497 5,497 5,497 5,497
R-squared 0.419 0.277 0.420 0.277
Number of fundID 1,058 1,058
Fund FE Yes Yes
Year FE Yes Yes Yes Yes
Style FE Yes Yes
Panel B The Impact of Duration of Underperformance with Alternative Performance Measures
(1) (2) (3) (4)
Alt-Perf= Alt-Perf= Alt-Perf= Alt-Perf=
LowAbsRetRank LowAbsRetRank LowRelRetRank LowRelRetRank
VARIABLES Fund FE OLS Fund FE OLS
M_ST -0.506*** -0.236 -0.319 -0.056
(-2.661) (-1.282) (-1.384) (-0.269)
M_LT 0.633** 1.166*** 0.449 0.958***
(2.082) (4.294) (1.605) (3.926)
F -0.016 -0.148* 0.005 -0.091
(-0.205) (-1.879) (0.056) (-1.174)
Perf 0.331*** 0.199** 0.349*** 0.265***
(3.498) (2.137) (3.686) (2.858)
𝑀_𝑆𝑇 × 𝑃𝑒𝑟𝑓 1.207*** 1.255*** 0.803** 0.896***
(4.153) (4.838) (2.566) (3.284)
𝑀_𝐿𝑇 × 𝑃𝑒𝑟𝑓 0.169 0.139 0.600 0.586
(0.406) (0.362) (1.478) (1.527)
𝐹 × 𝑃𝑒𝑟𝑓 0.010 0.020 -0.016 -0.086
(0.081) (0.157) (-0.119) (-0.675)
Small 0.071 0.268*** 0.060 0.268***
(0.592) (4.367) (0.498) (4.373)
AuCorr -0.198 -0.229 -0.224 -0.251*
(-0.922) (-1.545) (-1.038) (-1.694)
SecondFlow 0.229 0.132 0.223 0.132
(1.585) (0.927) (1.526) (0.922)
PStd -0.719*** -0.349*** -0.722*** -0.352***
(-24.060) (-18.280) (-24.179) (-18.594)
Constant 2.885*** 1.265*** 2.927*** 1.273***
(8.442) (4.403) (8.538) (4.428)
Observations 7,765 7,765 7,765 7,765
R-squared 0.438 0.297 0.437 0.296
Number of fundID 1,625 1,625
Fund FE Yes Yes
Year FE Yes Yes Yes Yes
Style FE Yes Yes
Table 6. The Effects of Varying M and F Cutoffs: No Backfill Bias Subsample with M=1
This table reports estimates from a subsample analysis using Equation (3) to examine when using different cutoffs, how flow condition, F , influences mid-year change in
fund volatility in response to the fund’s negative return in the first half-year for the no backfill sample. The subsample includes funds that have M=1 only, based on
different cutoffs for Moneyness and Flow. There are three cutoffs for Moneyness and Flow, respectively, resulting in nine combinations of Moneyness and Flow. The
three cutoffs for Moneyness: equal to 1, 0.9, and 0.8; the three cutoffs for Flow: equal to 1, -0.1, -0.2. The Appendix provides detailed definition for each variable
Numbers in parentheses are t-statistics based on robust standard errors clustered at fund levels. The symbols ***, **, and * denote significance levels of 1%, 5% and
10%, respectively.
Variable
1 2 3 4 5 6 7 8 9
Moneyness<1 Moneyness<0.9 Moneyness<0.8 Moneyness<1 Moneyness<0.9 Moneyness<0.8 Moneyness<1 Moneyness<0.9 Moneyness<0.8
Flow<0 Flow<0 Flow<0 Flow<-0.1 Flow<-0.1 Flow<-0.1 Flow<-0.2 Flow<-0.2 Flow<-0.2
Intercept 1.322*** 1.031 -1.299 1.53*** 1.224 -0.165 1.594*** 1.448* 0.173
[3.488] [0.934] [-0.509] [4.641] [1.267] [-0.077] [5.442] [1.724] [0.093]
F 0.699** 1.127 2.03 0.49* 1.068* 0.95 0.578* 1.133** 0.807
[2.213] [1.628] [1.61] [1.663] [1.755] [1.039] [1.797] [1.969] [1.062]
NegRet 0.695** 2.03*** 3.17*** 0.769*** 1.809*** 2.658*** 0.671*** 1.532*** 2.359***
[2.573] [3.287] [3.003] [3.848] [3.983] [3.401] [3.907] [4.166] [3.631]
F NegRet -0.17 -1.434** -2.150* -0.37 -1.399*** -1.743** -0.257 -1.278** -1.62**
[-0.564] [-2.193] [-1.961] [-1.457] [-2.644] [-2.035] [-0.996] [-2.555] [-2.083]
F Small -0.257 -0.292 -0.837 0.02 0.17 0.063 0.187 0.494 0.49
[-1.116] [-0.714] [-1.262] [0.096] [0.455] [0.112] [0.901] [1.417] [1.031]
F Leverage -0.608** -0.804 -1.423 -0.531** -0.892** -0.805 -0.653** -1.046** -0.739
[-2.379] [-1.497] [-1.636] [-2.227] [-1.998] [-1.229] [-2.539] [-2.372] [-1.245]
F ShortNotice -0.394* -0.275 -0.06 -0.376* -0.619 -0.028 -0.384* -0.596* -0.037
[-1.792] [-0.609] [-0.088] [-1.911] [-1.602] [-0.054] [-1.877] [-1.666] [-0.077]
All controls: Yes Yes Yes Yes Yes Yes Yes Yes Yes
Fixed Style
Effects Yes Yes Yes Yes Yes Yes Yes Yes Yes
Fixed Year
Effects Yes Yes Yes Yes Yes Yes Yes Yes Yes
Nobs 3186 1341 801 3186 1341 801 3186 1341 801
Adj. R2 0.277 0.324 0.358 0.278 0.327 0.353 0.277 0.326 352
Table 7. Contingency Table Test with Variance Ratios
This table reports the proportion of hedge funds falling in each classification for three groups of the no backfill sample: all funds,
funds with M = 1, and funds with M = 0. There are 8199 individual hedge fund-year observations over the period of 1997-2012 in the
no backfill sample. The Appendix provides detailed definition for the performance variables. Each classification is a combination of a
particular performance measure (Perf) and the value of variance ratio. Following Brown et al. (2001), the variance ratio is defined as
the ratio of variance of return in excess of style benchmark for the second six-month period to the variance of the first six-month
excess return. Variance ratio low is defined as a variance ratio less than the median for all funds in the calendar year, and variance
ratio high is defined as a variance ratio greater than or equal to the median for all funds. The Chi-square numbers represent the χ2(1)
statistics from the 2 x 2 contingency tables. The symbols ***, and ** denote significance levels of 1% and 5%, respectively.
Perf All Funds
Variance Ratio
Funds with M=1
Variance Ratio
Funds with M=0
Variance Ratio
Low High Chi-
square
Low High Chi-square Low High Chi-square
NegRet 0 50.36 49.64 51.36 48.64 49.99 50.01
1 47.44 52.56 4.79** 44.27 55.73 7.61*** 48.69 51.31 0.64
LowAbsRetRank 0 52.88 47.12 55.97 44.03 51.73 48.27
1 45.84 54.16 34.28*** 43.53 56.47 29.50*** 46.74 53.26 12.51***
LowRelRetRank 0 52.52 47.48 55.73 44.27 51.32 48.68
1 46.38 53.62 26.14*** 44.03 55.97 26.09*** 47.28 52.72 8.21***