THE IMPACT OF HEDGE FUND MANAGERS CAREER CONCERNS ON THEIR RETURNS, RISK-TAKING BEHAVIOR, AND PERFORMANCE PERSISTENCE DISSERTATION

Transcription

1 THE IMPACT OF HEDGE FUND MANAGERS CAREER CONCERNS ON THEIR RETURNS, RISK-TAKING BEHAVIOR, AND PERFORMANCE PERSISTENCE DISSERTATION Presented in Partial Fulfillment of the Requirements for the Degree Doctor of Philosophy in the Graduate School of The Ohio State University By Nicole M. Boyson, M.B.A. The Ohio State University 2003 Dissertation Committee: Professor René M. Stulz, Adviser Professor Jean Helwege Professor G. Andrew Karolyi Professor Karen Hopper Wruck Approved by Adviser Graduate Program in Business Administration

2 ABSTRACT Recent theoretical and empirical research suggests a link between a manager s career concerns the desire to keep his current job or obtain a better job and his risk-taking behavior. As a manager s career concerns change over time, his risk-taking behavior is likely to change as well. This dissertation provides evidence supporting this concept: risk-taking behavior that is related to career concerns does indeed change over time in the hedge fund industry, and this changing behavior explains the underperformance of more experienced hedge fund managers relative to their less-experienced counterparts. This result is used to motivate a test of performance persistence in which funds are selected for investment based on both past performance and manager experience. This more powerful selection process results in a finding of short-term (three month) persistence of less experienced, past good performers over more experienced, past poor performers, which is driven primarily by persistence in poor performance among more experienced managers. The first dissertation essay documents a negative relationship between hedge fund manager experience and performance. Less experienced hedge fund managers have significantly better performance than more experienced managers, even when controlling for style and risk factors. This performance differential is related to career concerns that differ among more and less experienced managers. More experienced managers employ less volatile trading strategies and tend to mimic the strategies of other managers (they herd ). Much of the underperformance of more experienced managers can be explained by these reductions in risk-taking behavior. Overall, the evidence strongly supports the hypothesis that, motivated by an increasing desire to keep their current jobs, hedge fund managers reduce risk as their careers progress, leading to a significant reduction in their performance. ii

3 The second essay examines whether hedge funds exhibit performance persistence. Although there is evidence of persistence among raw returns, when these returns are properly adjusted for common risk and style factors there is no evidence of short term (three-month) or long term (1-year) persistence. The remainder of the paper uses this relationship to design a more powerful analysis of persistence that includes manager tenure as well as past performance in selecting funds for investment. A portfolio that is long good past performing/less experienced managers and short bad past performing/more experienced managers displays persistence at the three-month time horizon. This finding is driven primarily by significantly poor performance that persists among more experienced managers. To explain this result, a conditional survival analysis finds an asymmetry in termination probabilities: conditional upon performance that is in the bottom two-thirds of managers, less experienced managers are significantly more likely to be terminated than more experienced managers. Since more experienced managers have lower returns, this drives the persistence results. This finding is broadly consistent with a number of reputational hypotheses: managers with more experience (reputation) are weeded out less quickly for a poor showing than are managers with less experience (reputation). This result is also consistent with the first dissertation essay, as the high level of performance required for young managers to avoid significantly high termination probabilities induces them to take on more volatile trading strategies and herd less than old managers. iii

4 ACKNOWLEDGMENTS I would like to thank my adviser René Stulz for his patience and guidance through numerous drafts of this dissertation. His comments have been invaluable in this respect. I would also like to thank my committee member, Jean Helwege, for her overall support, guidance, advice, and encouragement through my five years at Ohio State, and specifically for her assistance with this dissertation. Next, I would like to thank my committee member G. Andrew Karolyi for his guidance and support, particularly with the econometric issues I faced, as well as for his careful reading of some of the more preliminary drafts of this dissertation. Finally, I would like to thank my committee member Karen Hopper Wruck for her overall guidance, practical advice, and especially for her encouragement with respect to this dissertation. I also thank Vikas Agarwal, Richard Brealey, Stephen Brown, Steve Buser, Mike Cliff, Mike Cooper, Dave Denis, Diane Denis, Jean Helwege, David Hirshleifer, Kewei Hou, Bing Liang, Bernadette Minton, John McConnell, Narayan Naik, and seminar participants at The Ohio State University, Purdue University, and London Business School's Centre for Hedge Fund Research and Education for helpful comments and suggestions. This research was supported with a grant from the Dice Center of Financial Economics at The Ohio State University. iv

5 VITA July 22, Born Elyria, Ohio Bachelor of Business Administration, Accounting, Kent State University 1998.Master of Business Administration, Finance, Case Western Reserve University Graduate Research and Teaching Associate, The Ohio State University 2003 present..visiting Professor of Finance, Purdue University FIELDS OF STUDY Major Field: Business Administration Concentration: Finance v

6 TABLE OF CONTENTS ABSTRACT...ii ACKNOWLEDGEMENTS...iv VITA....v LIST OF TABLES......ix LIST OF FIGURES.....xi Chapter 1: Introduction... 1 Chapter 2: Why do experienced hedge fund managers have lower returns? Introduction Data Measures of risk-taking behavior Measures of performance The relationship between returns and manager experience The relationship between manager tenure and risk-taking behavior Risk-taking behavior and fund failure Description of proportional hazards model Estimation of proportional hazards model.27 vi

7 2.6. The relationship between risk-taking behavior, returns, and manager tenure Robustness tests Endogeneity tests Fee income Conclusions. 34 Chapter 3: Do hedge funds exhibit performance persistence? A new approach that accounts for manager tenure Introduction Data Performance measures and portfolio formation process Performance measures Portfolio formation process Do hedge funds exhibit risk and style-adjusted persistence? Manager tenure as a predictor of persistence Why do old, past bad returns persist? Relationship of termination and performance persistence to managerial career concerns Conclusions.57 Chapter 4: Conclusions 60 vii

8 LIST OF REFERENCES.. 64 APPENDIX A: Tables..70 APPENDIX B: Figures.94 viii

9 LIST OF TABLES Table 1: Summary statistics: return, manager, and fund characteristics.70 Table 2: Conversion of tenure variable to age estimates 73 Table 3: Descriptions of hedge fund indices..74 Table 4: Summary statistics: index returns...76 Table 5: Relationship between hedge fund performance and manager tenure...77 Table 6: Summary of statistics of risk and return variables by age, size, and age/size interactions 79 Table 7: Investment style categories...80 Table 8: Relationship between risk measures and manager tenure and other fund characteristics 82 Table 9: Relationship between risk measures and manager tenure and other fund characteristics, including interactions with personal capital variable Table 10: Time-varying proportional hazards model Table 11: Regression of annual returns on risk variables Table 12: The relationship between risk measures and fee income and manager tenure/fee income variables Table 13: Summary statistics for sample of 1,659 funds Table 14: Summary statistics for passive and hedge fund indices..90 Table 15: Quarterly persistence analysis when funds are selected based on prior performance...91 ix

10 Table 16: Quarterly persistence analysis when funds are selected based on prior performance and manager tenure.. 92 Table 17: Conditional time-varying proportional hazards models.93 x

11 LIST OF FIGURES Figure 1: Risk measures versus reputational proxies...94 Figure 2: Size/tenure category detail...95 Figure 3: Estimated survivor functions by manager tenure...96 xi

12 CHAPTER 1 INTRODUCTION Recent theoretical and empirical research suggests a link between a manager s career concerns the desire to keep his current job or obtain a better job and his risk-taking behavior. As a manager s career concerns change over time, his risk-taking behavior is likely to change as well. This dissertation provides evidence supporting this theory: risk-taking behavior that is related to career concerns does indeed change over time in the hedge fund industry, and this changing behavior explains the underperformance of more experienced hedge fund managers relative to their less-experienced counterparts. This result is used to motivate a test of performance persistence in which funds are selected for investment based on both past performance and manager experience. This more powerful selection process results in a finding of short-term (three month) persistence of less experienced, past good performers over more experienced, past poor performers, which is driven primarily by persistence in poor performance among more experienced managers. The first dissertation essay documents a negative relationship between hedge fund manager experience and performance. The initial analysis shows that less experienced hedge fund managers have significantly better performance than more experienced managers, even when controlling for style factors and common risk factors (e.g., correlations with market indices). This performance differential is related to career concerns that differ among more and less experienced managers. In the hedge fund industry, since more experienced managers have more to lose in terms of income, job security, and personal wealth should their funds fail, this essay tests (and finds supporting evidence for) the following hypothesis: Career concerns of more experienced managers cause them to reduce certain types of risk-taking behavior that might lead to fund failure, which results in their lower returns. Consistent with this hypothesis, a cross-sectional analysis shows that more experienced managers do 1

13 indeed engage in less risky behavior than more experienced managers. Specifically, more experienced managers employ less volatile trading strategies and tend to mimic the strategies of other managers (they herd ). An analysis of the factors leading to fund failure provides a powerful incentive for this behavior: engaging in less volatile investment strategies and herding with other managers increases the probability of a fund s survival. Thus, more experienced managers who are concerned about keeping their jobs respond to this incentive by reducing risky behavior, resulting in higher survival rates among more experienced managers. Finally, much of the underperformance of more experienced managers can be explained by these reductions in risk-taking behavior. Overall, the evidence strongly supports the hypothesis that, motivated by an increasing desire to keep their current jobs, hedge fund managers reduce risk as their careers progress, leading to a significant reduction in their performance. The second essay examines whether hedge funds exhibit performance persistence. Typically, studies of performance persistence examine whether past good funds continue to outperform and past bad funds continue to underperform; if so, these funds are said to exhibit persistence. In this second essay, the initial test of persistence is performed in the same way by selecting funds based on their past performance. Although there is evidence of persistence among raw returns, when these returns are properly adjusted for common risk and style factors there is no evidence of short term (three-month) or long term (1-year) persistence. However, this test of persistence ignores the first dissertation essay s finding that less experienced managers are better than more experienced managers. Therefore, the remainder of the paper uses this relationship to design a more powerful analysis of persistence. The motivation is as follows: since less experienced managers are better than more experienced managers, a bad return for a less experienced manager is more likely to be due to bad luck than for a more experienced manager, while a good return for a less experienced manager is more likely to be due to manager skill than for a more experienced manager. Given this result, choosing funds for investment based both on past performance and manager experience should result in a finding of persistence. The results of this test support this idea; a portfolio that is long good past performing/less experienced managers and short bad past performing/more experienced managers displays persistence at the threemonth time horizon. This finding is driven primarily by significantly poor performance that persists 2

14 among more experienced managers. The survival analysis from the first dissertation essay provides a clue as to why: less experienced managers are terminated more often (i.e., their funds fail more often) than more experienced managers. A more thorough analysis of this result provides even more evidence: conditional upon performance that is not in the top one-third of all performers, less experienced managers are significantly more likely to be terminated than more experienced managers. In other words, for a less experienced manager to survive, his performance must be far superior to that of a more experienced manager. The result of these asymmetries in the attrition process is that a larger number of more experienced managers survive to the next period, while fewer of the less experienced managers survive. Since more experienced managers have lower returns, on average, this drives the persistence results. This finding is broadly consistent with a number of reputational hypotheses: managers with more experience (reputation) are weeded out less quickly for a poor showing than are managers with less experience (reputation). It is also consistent with the results of the first essay, and provides an additional incentive: for young managers to survive, they must have performance that is in the top one-third of all performers. This performance is not likely to be achieved by herding or taking on less volatile trading strategies, so young managers must incur more risk if they wish to equate their probability of survival with that of older managers. For older managers, this result provides an additional incentive for them to reduce risk and engage in herding behavior: in addition to greater career concerns due to having more at stake than young managers, old managers also enjoy survival probabilities that are not performance sensitive, providing them additional incentive to incur less risk and herd more than their younger counterparts. 3

15 CHAPTER 2 WHY DO EXPERIENCED HEDGE FUND MANAGERS HAVE LOWER RETURNS? 2.1. Introduction A number of theoretical models address the idea that managers may alter their risk-taking behavior as their careers progress. Some models predict that risk-taking behavior will increase with time, while others predict that it will decrease. One model that posits an increase is Avery and Chevalier (1999) (hereafter, the AC model). This model argues that as managers gain experience, they obtain more precise knowledge about their own abilities. Hence, older managers follow their own signals since they are more confident that they are correct, while younger managers, by contrast, place less emphasis on their own signals and tend to mimic the behavior of other managers. Empirically, in studies of mutual fund managers, security analysts, and macroeconomic forecasters respectively, Chevalier and Ellison (1999b), Hong, Kubik, and Solomon (2000), and Lamont (2002) report results consistent with this theory: old herd less than young. By contrast, other models predict that risk-taking behavior will decrease as managers age (see Prendergast and Stole (1996), hereafter PS, and Graham (1999)). The PS model argues that managers wish to quickly acquire a reputation for learning, and thus, when faced with new information each period, young managers initially overreact (to show they have good information) while old managers underreact (so as not to signal that their actions in prior periods were wrong). Other research ascribes the decrease in risk-taking behavior to increasing risk aversion with age. A large labor economics and management literature supports this idea (see Salancik (1977), Morin and Suarez (1983), Holmstrom and Milgrom (1987), Kanodia, Bushman, and Dickhaut (1989), and Bernheim (1994)) as well as some related psychology/sociology literature (see Vroom and Pahl (1971), Kiesler (1971), and Kahneman, 4

16 Slovic, and Tversky (1982)). Empirically, research by Graham (1999) (investment newsletters) and Li (2002) (securities analysts) supports this theory. Finally, additional empirical support appears in the management literature regarding CEO and top management team behavior (see Katz (1982), Hambrick and Mason (1984), Finkelstein and Hambrick (1990), and Jaggia and Thosar (2000)). The opposing theories and empirical findings suggest that managerial risk-taking behavior depends upon the relative importance -- over time -- of managerial career concerns. The career concerns literature argues that principal-agent problems (which occur when the incentives of investors and fund managers are misaligned) can sometimes be mitigated by a manager's desire to keep his current job or obtain a better job (i.e., his career concerns ). See Fama, (1980). However, under certain conditions, career concerns can work too well: resolving some agency problems while causing others (see Holmstrom (1982), and Holmstrom and Ricart i Costa (1986)). An alternative view is that career concerns are a special type of agency problem, which sometimes mitigate and sometimes exacerbate problems caused by other sources. This paper studies the implications of the career concerns theories in a new venue: the hedge fund industry. The hedge fund industry is of interest due to two unique features: low agency costs and career concerns that change over time. Agency costs are likely low in the hedge fund industry for a number of reasons. First, hedge funds are not required to report their holdings, trades, or returns publicly, reducing the incentive to window-dress. 1 Second, they have option-like incentive fees (usually about 20% of profits) which better align shareholder and manager interests relative to the asset-based fees (usually 1-2% of fund assets) used in the mutual fund industry. 2 The empirical literature supports this idea: in an attempt to increase asset size, mutual fund managers tend to increase risk when their funds are down during the first half of a year (see Brown, Harlow, and Starks (1996) and Chevalier and Ellison (1997)). By contrast, hedge fund managers do not substantially increase risk when their funds are down (see 1 Window-dressing refers to altering one's portfolio near a required reporting period in an attempt to attract new customers. See Lakonishok, et al (1991). 2 Since incentive fees for hedge fund managers are asymmetrical in that managers earn fees if their profits are positive but do not have to pay investors back if their returns are negative (similar to the returns from a long call option), there may be concerns that this fee structure encourages managers to take on excessive risk. Theoretical and empirical literature refutes this idea, arguing that these types of fees actually encourage an appropriate amount of risk. See Starks (1987), Gibbons and Murphy (1992), Carpenter (2000), and Das and Sundaram (2002). 5

17 Brown, Goetzmann, and Park (2001) (hereafter, BGP)). Additionally, the hedge fund literature documents a positive relationship between incentive fees and fund returns, which implies that investors are willing to pay the best managers the highest fees. (See Ackerman, McEnally, and Ravenscraft (1999) (hereafter, AMR), Liang (1999), and Edwards and Caglayan (2001)). Third, hedge fund managers are typically thought to have substantial wealth invested in their own funds. For example, the June, 2002 issue of Institutional Investor notes that Bruce Cohen (who manages the SAC Capital Advisors' fund) has about $1 billion of his own capital in the fund. An article in the April, 2002 issue of Fortune notes that Crispin Odey and his team at Odey Asset Management have at least $30 million of their own assets in the fund. In contrast, mutual fund managers rarely have significant personal capital invested in their own funds. Finally, many hedge funds have lock-up periods during which investors are not permitted to withdraw money, which might induce managers to invest in less liquid, but potentially profitable, investment strategies, and will likely reduce the amount of cash that a manager must hold. Career concerns of hedge fund managers are likely to change over a manager's career, for three reasons. First, hedge fund managers earn large salaries which tend to increase as their careers progress, due to good initial performance that attracts new money to their funds. A manager overseeing $115 million would earn about $4 million a year (including both fixed management fees and variable incentive fees), assuming historical average returns and fees. 3 By contrast, the average mutual fund manager made about $440,000 in Second, failed hedge fund managers rarely start new hedge funds (see BGP, (2001)). Finally, for managers investing their own money in their funds, experienced managers have more to lose -- personally -- should their funds fail. Thus an older manager, who presumably has more at stake in terms of personal wealth, fund assets, and opportunity costs related to losing his job, will likely have different career concerns than a young manager. For example, he has 3 This estimate is very conservative, since it does not include earnings from a manager's personal investment in his fund. 4 See Mutual Fund Managers' Pay Stays in Bull Market: Median Pay is Seen Risking 35% Since 1999 Despite Stocks' Woes, The Wall Street Journal, June 7,

18 strong incentives to avoid failure in order to protect his wealth and reputation. 5 Both features low agency costs and career concerns that change over time are consistent with the PS and Graham models: if agency costs are low, career concerns will be relatively more important, and to the extent that career concerns change over time, managers should become more risk-averse as their careers progress. With a large sample of hedge fund managers over the period , this paper tests the opposing theories by examining how the risk-taking behavior of hedge fund managers changes over time. 6 Consistent with the predictions of Prendergast and Stole, I find a strong negative relationship between manager experience and risk-taking behavior; i.e., older managers are more risk-averse. This result is in direct contrast with previous empirical findings for mutual fund managers; in mutual funds, less experienced managers face higher agency costs and have greater reputational concerns, leading to less risk-taking than their more experienced counterparts (see Avery and Chevalier (1999) and Chevalier and Ellison (1999b)). The remainder of the paper explains this negative relationship between manager age and risktaking behavior in light of the unique career concerns of older hedge fund managers (i.e., older managers have more to lose should their funds fail). I hypothesize that this negative relationship is driven by older managers consciously reducing the types of risk that might be associated with fund failure. Consistent with this hypothesis, I show that high levels of risk-taking behavior do indeed lead to fund failure. These results are statistically and economically significant. Finally, this paper uses the finding that older managers reduce risk-taking behavior to explain the empirical result that older managers have lower returns. (See Liang (1999) and Edwards and 5 The opposing argument may also be made. A young manager arguably has more at stake in terms of expected future earnings than an older manager. However, a young manager also will not have had the success and past profitability of an old manager, so the transition from the hedge fund industry, to say, the mutual fund industry, may be less difficult for him. Regardless, the empirical results of this paper support the idea that hedge fund manager career concerns increase, rather than decrease, with age. 6 Consistent with prior literature (e.g., Chevalier and Ellison (1999b)), this paper defines risk-taking behavior of managers in two ways. The first measure of risky behavior is standard deviation of returns. The second measure (for which two proxies are employed) relates to herding behavior, where managers mimic others. This is typically considered a low-risk behavior (for theoretical and empirical justification, see for example, Diamond (1991), Hirshleifer and Thakor (1992), Prendergast and Stole (1996), Chevalier and Ellison (1999b), Graham (1999), and Hong, Kubik, and Solomon (2000).) These definitions of risk-taking behavior are not proxies for market beta risk or correlations with passive or market factors. As noted in Footnote 7, below, market beta risk is controlled for and measured separately. 7

19 Caglayan (2001).) While this difference cannot be explained by systematic differences in market beta risk (measured as coefficients from regressions of fund returns on a number of market factors), it can be explained by differences in volatility and herding behavior documented above. 7 These results are rather compelling. For example, risk-taking behavior explains a large portion of the difference in returns at various levels of manager experience: while controlling for market beta risk and fund characteristics, but without controlling for the risk measures of interest (volatility and herding measures), the difference in returns between a 52 year-old manager and the average (47 year-old) manager is about 4.0% annually. However, when controlling for volatility and herding risk, this difference drops to about 2.1% per year. These results are consistent across all measures of volatility and herding risk, and robustness checks confirm that they are not likely due to survival bias or decreasing manager ability over time. This paper is the first to apply the career concerns models to the hedge fund industry, expanding the literature in this area. Recent papers on hedge funds have focused on constructing assetpricing models to explain hedge fund returns (see Fung and Hsieh (1997), Liang (1999), and Agarwal and Naik (2000, 2003)), using fund characteristics to explain risk and return (see AMR (1999)), and documenting and understanding the risks of hedge funds (see Liang (1999), Fung and Hsieh (2000)). Most closely related to this paper is BGP (2001) who show that managers reduce risk when their funds perform well during the first half of a year and do not increase risk when their funds perform poorly. They hypothesize that career concerns (specifically, concern for survival) cause this behavior. Consistent with this idea, this paper provides direct support that career concerns are indeed important to hedge fund managers. Second, this paper's focus on explaining differences in returns among young and old managers with differences in risk-taking behavior is unique. Most other empirical tests of these models focus on understanding and documenting how risk-taking changes over a career, not on the implications of this behavior on fund performance (see Chevalier and Ellison (1999b), Hong, Kubik, and Solomon (2000), 7 I calculate regression coefficients for a number of proxies for the market. These include various passive indices, hedge fund indices, and returns on dynamic trading strategies. While these factors are often relevant in explaining hedge fund performance, the coefficients on these factors do not systematically differ among young and old managers. 8

20 and Lamont (2002)). Using different samples and time periods, earlier research has documented an inverse relationship between manager experience and performance (see Liang (1999) and Edwards and Caglayan (2001)). This paper confirms the negative relationship and provides an explanation: due to career concerns, experienced managers incur less risk than inexperienced. To the extent that career concerns dominate the behavior of experienced hedge fund managers such that their returns are significantly below average, this may not be good news for investors. However, the higher failure rate of young hedge funds is also not welcome news. In selecting a hedge fund, an investor must weigh the higher probability of failure of a young fund against the lower predicted returns of an old fund. Finally, this paper has implications for the optimal design of incentive contracts. While the typical fee arrangement which pays managers a percentage of profits as well as a percentage of assets should align manager and investor interests, and appears to properly align incentives for younger managers, it may not work as well for experienced managers. As managers age, the desire to survive can outweigh their desire to outperform other managers, which encourages them to reduce risk, which significantly reduces their returns. Recent theoretical work regarding appropriate contract design addresses this point directly (see Dybvig, Farnsworth, and Carpenter (2001)). They argue that the optimal contract will not only pay the manager a percentage of profits over and above a fixed benchmark, but also will give the manager additional compensation for taking a position that deviates from the benchmark. Their model is an important step towards creating incentive contracts that address the temptation for older managers to take on too little risk. This paper is organized as follows. Section 2.2 describes the data. Section 2.3 confirms the negative relationship between hedge fund manager experience and fund returns, which holds for both raw returns and when controlling for correlation with market indices. I hypothesize that career concerns of older managers lead to a decrease in the types of risk that could lead the fund to fail, which causes the lower returns. The remaining sections confirm this hypothesis. First, Section 2.4 shows that experienced managers do indeed reduce certain types of risk-taking behavior. Next, Section 2.5 documents that increases in these types of risk-taking behavior are positively related to fund failure providing an explanation for the observed low risk-taking behavior of older managers and also finds 9

21 that more experienced managers tend to have lower failure rates than less experienced, implying that managers are responding to this incentive. Finally, Section 2.6 shows that this decrease in risky behavior can explain a substantial portion of the fall in performance of experienced managers. Section 2.7 provides a number of robustness checks and Section 2.8 concludes Data Data was provided by Tremont Advisory Shareholders Services (TASS). TASS has been collecting hedge fund data directly from managers since the late 1980's, and currently has over 2,400 funds in their database, both living and dead. 8 The database includes monthly net-of-fee returns, as well as expenses, fees, size, terms, age, and style of the funds. Many of the funds also include a biographical sketch of the manager, providing information such as age, schooling, professional designations, and prior work experience. TASS categorizes funds as dead when they stop reporting their monthly information to the database. Because the fund managers voluntarily report this data, there are a number of reasons why funds stop reporting, including failure, merger, or a more arbitrary reason, such as the fund manager's belief that being included in the database is not helping him raise additional investment dollars. While some of these reasons are clearly associated with poor performance (e.g., failure), others are less obvious. In most cases, TASS includes the reason for leaving the database. Each fund must have 24 months of consecutive returns and at least $5 million in assets during the period January, 1994 to December, 2000 for inclusion in the sample. Obviously, this requirement represents a trade-off between sample size and ensuring that each fund has a long enough time series for meaningful regression results. Additionally, I adjust the sample to account for backfilling (or instant history ) bias (see Edwards and Park (1996)). When TASS adds a new fund to the database, they backfill historical returns. At the beginning of its existence, a hedge fund undergoes an incubation period where the only significant investor is the fund manager. He hopes to compile a good track record before making the fund available to the public. Thus, most funds arrive in the database with a history of 8 TASS has maintained data on dead funds since

22 strong performance which was never available to outside investors, biasing returns upward. This difference is often large using the TASS database, Fung and Hsieh (2000) calculate the bias as about 3.6% per year. TASS provides the incubation period for each fund; thus, to control for this bias I drop the data for the initial incubation period for each fund. The final sample includes 271 funds with at least 24 months of returns, $5 million in assets, and all of the manager and fund characteristic variables. Since the sample with the full set of manager characteristics is fairly small, I construct another sample. The second (larger) sample has 982 funds with at least 24 months of returns and size greater than $5 million, and the full set of fund returns and fund characteristics, including manager tenure, but does not include the other manager characteristics. Since manager tenure is the variable of interest in many of the following tests, the large sample is used throughout most of the paper, with the smaller sample available for robustness tests. Table 1 includes summary statistics of the return, manager and fund characteristics variables. To facilitate interpretation of results from some of the tests performed on the large (982 fund) sample, the manager tenure variable is converted to a manager age variable. Since the manager tenure variable is available for the large (982 fund) sample, and manager age is only available for the small (271 fund) sample, categorical estimates for manager age are estimated by dividing the small sample into 10 percentiles by tenure, and then calculating the average manager age for each of the ten percentiles. The tenure percentile estimates for the small sample are within a few months of the tenure percentile estimates for the large sample, so the age estimates appear to be reasonable. These estimates are reported in Table 2. Of the 982 funds in the sample, 285 funds are categorized by TASS as dead. For many of the funds that stop reporting, their self-reported description indicates that they actually failed (went bankrupt or stated that they closed the fund due to poor performance), while for others, whether or not the fund has actually failed is more difficult to determine. Section 2.5 performs a detailed analysis of fund failure. 9 9 Some managers might leave the sample because they are very successful and, from a marketing perspective, do not need to report their data any longer. However, as a preview of results from the analysis in Section 2.5, most funds that leave the sample are suffering from poor performance, and 11

23 As a brief summary, the average hedge fund manager in the sample is 47 years old, has 22 years of work experience, manages a fund of about $113 million dollars, and comes from an undergraduate institution with SAT scores about 300 points over the national average. About half the managers in the sample have MBAs, and approximately 70% of the managers use leverage and have their own money invested in their funds. Clearly, this is an experienced, well-educated group. In comparison, the average mutual fund manager is about 3 years younger, attended a school with average SAT scores about 200 points lower, and is slightly more likely to have an MBA than the average hedge fund manager Measures of risk-taking behavior This paper focuses on explaining hedge fund manager behavior in the presence of career concerns, and also links this behavior to realized returns. Specifically, it seeks to understand how managers' risk-taking behavior varies over their careers, and whether this variation in risk-taking behavior can explain performance that declines over time. In this respect, risk-taking behavior must first be defined and measured. The first measure of risk-taking behavior that is used is the total standard deviation of a portfolio's returns. Standard deviation is an absolute measure in that it does not compare the risk of a portfolio to the risk of a benchmark. Hence, it may or may not be correlated with the regression coefficients estimated from an asset-pricing model, and as such, may contain additional useful information regarding managerial risk-taking behavior. Additionally, because hedge fund managers can engage in dynamic strategies, using leverage, options and other derivative securities, they may have some control over the standard deviation of their funds' returns. While standard deviation measures total risk, this paper also uses two relative risk (or herding ) measures. The first is tracking error deviation measured relative to a fund-of-funds (FOF) index. A fund-of-funds is a professionally managed hedge fund that invests in other funds. Fung and whether they actually failed at the time of leaving the sample or at some time thereafter (and got out of the sample beforehand to save face), most of them were not successful funds (they were smaller and had lower than average returns). 10 This data is obtained from Chevalier and Ellison (1999a). 12

24 Hsieh (2000) note that a fund-of-funds index is an appropriate benchmark for measuring hedge fund performance, which motivates its use in the estimates that follow. Most FOF managers diversify their portfolios among hedge fund styles, and market their funds as providing broad exposure to the universe of hedge funds. Tracking error measures the volatility in returns not explained by market volatility (specifically, it is the square root of the residual in a time-series regression of a individual fund's returns on the FOF index). Tracking error deviation measures how much a manager's tracking error differs from that of the average manager in the same style category. Since it is measured relative to other managers, it is a type of herding measure. The second relative risk measure is beta deviation, following Chevalier and Ellison (1999b). Beta deviation uses each individual fund's time-series coefficient (beta) from a regression of that fund's returns on the FOF index, and is calculated as the difference between a fund's beta on the FOF index and the average beta on the FOF index for all other funds in the same style category. 11 Like tracking error deviation, beta deviation is also a herding measure. In summary, this paper uses three measures: standard deviation, tracking error deviation, and beta deviation to measure a manager's risk-taking behavior Measures of performance The focus of this paper is to explain why hedge fund performance varies with manager tenure. We measure hedge fund performance in three ways. The first measure uses simple excess-of-risk-freerate returns. The risk-free rate is the 30-day Treasury bill rate. The second performance measure adjusts the returns for exposure to a number of passive indices, the factors used by Carhart (1997), and a put option return as suggested by Agarwal and Naik (2003). The third performance measure controls 11 This measure differs slightly from that used by Chevalier and Ellison (1999b). For their sample of equity mutual funds, they calculate betas for each fund by regressing returns on a broad stock market index. For each fund, beta deviation is the absolute value of the difference between a fund's beta and 1 (since a well-diversified equity mutual fund would have a beta of about 1). For hedge funds, betas vary widely by style, and thus, the measure used in this paper accounts for these style differences by measuring each fund's beta deviation relative to the average beta in its style category. 13

25 for exposure to hedge fund indices, in the spirit of Sharpe's style analysis (see Sharpe (1992, 1994)), and as suggested by Fung and Hsieh (2001) and L'habitant (2001). These indices should proxy for the specific risks faced by hedge funds. The passive indices used in the second measure of performance are obtained from Datastream and include: the US Trade Weighted Dollar index to capture currency risk, gold and commodity indices, the Lehman Brothers Eurobond, 30-year Treasury bond, and U.S. aggregate bond indices, and the S&P 500, Wilshire 5000, and MSCI Eafe stock market indices. Additionally included are the Fama-French (1992,1993) SMB (a zero-investment portfolio constructed by subtracting the returns of large market capitalization firms from the returns of small capitalization firms) and HML (a zero investment portfolio constructed by subtracting the returns of low book to market ratio stocks from the returns of high book to market ratio stocks) factors, as well as Jegadeesh and Titman's (1993) momentum factor (MOM) a zero-investment portfolio constructed as the spread between the performance of stocks which were in the top 30% of returns in the prior twelve months and those which were in the bottom 30%. 12 Finally, the regressions include a simple option trading strategy as suggested by Agarwal and Naik (2003). For each passive index listed above, the return from investing each month in a one-month at-the-money put option, holding the option for a month, and then reinvesting in next month's option is calculated, using the Black-Scholes model to estimate the option prices. Then, each month, an equally weighted average of all the index option returns is calculated and used as a regressor. As mentioned above, the third measure of performance uses hedge fund indices to control for market exposure. These indices are published jointly by Credit Suisse First Boston (CSFB) and TASS, and represent a number of hedge fund trading strategies. They are constructed so as to minimize survivorship bias. For further detail about the indices used, see Table 3. Table 4 reports summary statistics for all indices used in the paper. For the second and third measures of performance (which control for exposures to passive indices and hedge fund indices, respectively), we estimate two-step Fama-Macbeth (1973) regressions. 12 These returns were obtained from the website of Kenneth French. 14

26 The first step performs time-series regressions to estimate each fund's loading on each index. The second step uses these loadings as controls for market exposure in cross-sectional regressions of each fund's return on fund and manager characteristics. A well-known problem caused by two-stage regressions is that to the extent the coefficients from the first-pass (time series) regressions are estimated with error, this error will then be included in the second-pass (cross-sectional) regressions. Fama and Macbeth refer to this as the errors-invariables (EIV) problem. To minimize this error, Fama and Macbeth suggest using portfolios of securities (or, in this case, hedge funds), rather than individual securities (hedge funds), in the first-pass regression. Since the individual security (hedge fund) returns will be less than perfectly positively correlated, combining them into portfolios results in a diversification effect where estimation errors can be averaged out within a portfolio. The method by which securities (hedge funds) are sorted into portfolios is important. If securities (hedge funds) are sorted into portfolios randomly, the portfolios will have betas that are very close to one, resulting in too little dispersion among betas. Thus, Fama and Macbeth suggest forming portfolios based on their betas so as to maximize this dispersion. In related work, Fama and French (1992, 1993) show that firm (hedge fund) size can be used as an additional sorting variable: since firm (hedge fund) size typically has a low correlation with firm (hedge fund) beta, cross-sorts on both size and beta can provide maximum dispersion of betas among portfolios formed on these criterion. 13 However, using betas to sort funds into portfolios can be problematic. If the individual firm (hedge fund) betas by which portfolios are formed are estimated with error, the portfolio betas calculated from the time-series regressions will also have estimation error. Fama and Macbeth refer to this effect as the regression phenomenon. They address this problem by forming portfolios based on individual security (hedge fund) betas that are calculated in one period (the formation period), and then estimating the portfolio betas to be used in the second pass (cross-sectional) regressions in a 13 Consistent with Fama and French's finding of low correlations between firm size and firm betas, I also find low correlations between hedge fund betas and hedge fund size. The correlations between the betas on each index and hedge fund size range from to

27 subsequent period (the estimation period ). The estimation errors are likely to be uncorrelated between the formation period and the estimation periods, which should reduce the severity of the regression phenomenon. This paper closely follows the Fama-Macbeth approach, with one exception. Since the length of the time-series of data available for this study (7 years) is short relative to the length of time-series data available to Fama and Macbeth (38 years), forming portfolios in one period and estimating betas in another period becomes impractical. Fama and Macbeth used seven-year formation, five-year estimation, and four-year testing (second-pass regression) periods. Trying to follow their approach exactly would leave no data for the estimation period (let alone the testing period). Therefore, I form portfolios (based on fund size and fund beta), estimate portfolio betas, and perform cross-sectional regressions using the entire (7 year) time-series of data. While this approach addresses the errors-invariables problem since using portfolios rather than individual funds arguably reduces estimation error, it is still subject to the regression phenomenon criticism. To address this concern, I perform two robustness checks. First, I form portfolios based on size only, as opposed to both size and beta. This approach completely eliminates the regression phenomenon, since as noted above, estimated betas and hedge fund size have very low correlations. While this approach addresses the regression phenomenon, it comes at a cost: less dispersion among portfolio betas. However, when this methodology is used, the main results of the paper remain unchanged, providing support that the results of the paper are not spuriously driven by the regression phenomenon. The second robustness check sorts funds into portfolios based on the first two years of data, and estimates portfolio betas using the next two years of data. Then, cross-sectional regressions are performed using the remaining three years of data. Again, the results from this approach are similar to those obtained when all the data is used in the formation, estimation, and testing periods. The magnitude and direction of the main results of the paper still hold, although the results are statistically significant at lower levels due to the very short formation, estimation, and testing periods. These 16

28 robustness checks provide comfort that the approach of using all the data for the formation, estimation, and testing periods does not cause a significant regression phenomenon problem. The specifics of the process are as follows. First, I perform time-series regressions using all the data, where I calculate sorting coefficients on each index for each fund. 14 A single-factor model estimates the time-series coefficient (beta) for each fund for each index, as follows: i r t α t bs tindex t ε t = + + (1) where i t r is fund i's return in period t (t=1 to 84 months), and (INDEX t) refers to the index used. Next, I sort funds into quartile portfolios, based on this sorting beta (b st ) and fund size. This sort results in sixteen (16) portfolios for each index. 15 Once the portfolios are formed, I calculate equally-weighted averages of excess returns (net of the risk-free rate) for each set of portfolios. Then, time-series regressions of the portfolio returns (r p ) are performed for each index as follows: j pt r = α pj + b pjindex t + ε t (2) where j pt r are monthly portfolio returns in excess of the risk-free rate (number of months, t=84), and INDEX is the return on the INDEX used in the sorting process. Thus, sixteen (j=16) regressions for each INDEX are performed (one for each of the sixteen portfolios). I then assign the resulting coefficients (the b pj s) to each fund in the sample, based on its portfolio classification. This process results in each fund being assigned a beta coefficient for each index. These coefficients are then used as market controls in the cross-sectional regressions in Section 2.3. The coefficients on the passive indices are used as market controls in the second measure of performance, while the coefficients on the hedge fund indices are used as market controls in the third measure of performance. 14 Recall that there are two separate sets of indices used as controls for market risk factors, passive and hedge fund. Descriptive statistics of these indices may be found in Table The results are robust to using nine (9) or twenty-five (25) portfolios. 17

29 As noted earlier, the main appeal of this approach is that the cross-sectional regressions can be performed for every month in the sample (84 months) which allows full utilization of the time-series data available. The next section begins the empirical investigation of the relationship between manager tenure and fund performance The relationship between returns and manager experience Since the primary focus of this paper is to explain the negative relationship between hedge fund manager experience and performance, I begin by confirming that this relationship (which was noted in previous studies by Liang (1999) and Edwards and Caglayan (2001)) actually holds for this paper's data set, since the prior studies use different data sets and time frames than this paper. Liang uses a sample of 385 funds from the Hedge Fund Research (HFR) database with at least three years of monthly returns for the period 1994 to Edwards and Caglayan use a sample of 836 funds with at least 24 months of returns from the MAR/Hedge database for the time period January 1990 to August By contrast, this paper uses a sample of 982 hedge funds, with at least 24 months of returns and $5 million in assets, from the Tremont Advisory Shareholder Services (TASS) database, for the period 1994 to I estimate monthly cross-sectional regressions in which the dependent variable is one of three different performance measures as described in Section 2.2.2, and the independent variables are a number of manager and fund characteristics. The variable of interest in these regressions is manager tenure (and/or manager age). In the first set of regressions, the dependent variable is excess return, and no controls for market exposure are included. The second set of regressions controls for exposure to passive market indices using the passive betas calculated in Section 2.2.2, while the third set of regressions controls for exposure to hedge fund market indices using the hedge fund betas calculated in Section

30 The regression for the first performance measure (excess returns) is as follows: J K i i i = t α + t b + jt, j, t c + kt, k, t ε (3) t j= 1 k = 1 r M F where the i t r s are the monthly returns of fund i in month t, the M, i jt s are the manager characteristic variables of fund i in month t, and the F i kt, s are the fund characteristics of fund i in month t. The second and third cross-sectional regressions also include controls for market exposure. Regression two includes passive market controls, where regression three includes hedge fund market controls, as follows: J K L i i i i (4) r = α + b M + c F + d E + ε t t j, t j, t k, t k, t l, t l, t t j= 1 k= 1 l= 1 where the i t i i r s the M s and the jt, F s are the same as in Equation (3) above, and the kt, E, i lt s are the market control coefficients assigned to each fund i from the time-series regressions on passive (regression 2), or hedge fund (regression 3), indices as described in Section Table 5 reports the results from these regressions. In the table, regressions 1a and 1b use excess return as the dependent variable as in equation (3) above, and do not control for exposure to market indices. Regressions 2a and 2b control for exposure to passive indices, and regressions 3a and 3b control for exposure to hedge fund indices, as in equation (4) above. For all specifications, regression a is performed on the small sample (271 funds), which includes all manager, as well as all fund characteristics, while regression b is performed on the larger sample (982 funds), which includes only fund characteristics and manager tenure. Consistent with previous literature, manager tenure is strongly negatively related to performance. This finding is robust to all specifications tested. The results are significant, both economically and statistically. For each year of experience, annual returns decrease by about 0.8%. (For 19

31 ease of interpretation, the coefficients on manager age and tenure are annualized in Table 5.) Thus, a 52 year-old manager has returns about 4% lower than the average (age 47) manager. Additionally, the coefficient on manager age is significant and negative, the same direction as the manager tenure variable. 16 Both tenure and age are meaningful proxies for reputation, since a manager's reputation may have been developed before he became a hedge fund manager. Including the market control variables in regressions 2a and b and 3a and b provides additional explanatory power; however, the coefficient on manager tenure remains significant and negative. Most importantly, including these market control variables does not change the manager tenure coefficient by very much: it drops from -0.8%/year to -0.7%/year. In other words, although market beta risk explains a fair amount of the cross-sectional variation in returns the R 2 s from the regressions that do not include market control variables (regressions 1a and 1b) average about 0.07, while the R 2 s from the regressions that do include the market risk variables (regressions 2a, 2b, 3a, and 3b) average about 0.30 the negative relationship between manager tenure and returns still holds significantly. This section has established the existence of a negative relationship between manager tenure and hedge fund performance, one that is not explained by other fund characteristics or by exposure to market beta risk. The remainder of this paper investigates a potential explanation for this relationship as follows, which is motivated by the theoretical work of Prendergast and Stole (1996) and Graham (1999). I test the following general hypothesis: Managers have career concerns which lead them to reduce risk as their careers progress. Specifically, older hedge fund managers are particularly concerned about keeping their jobs and maintaining their personal wealth, since they have more to lose than younger managers should their funds fail. These concerns cause them to reduce the types of risk that may lead their funds to fail. In turn, this reduction in risk-taking behavior explains their lower returns. The next section begins testing this hypothesis by examining whether older managers take on less risk than young. 16 The number of years experience variable has coefficients and significance levels nearly identical to those found for manager age, so results are not reported in Table 5. 20

32 2.4. The relationship between manager tenure and risk-taking behavior To address the hypothesis that, due to career concerns, experienced managers reduce risk over time causing lower returns, the first step is to examine the relationship between manager tenure and risk-taking behavior. This section examines this relationship, directly testing the implications of the theoretical literature regarding career concerns. This literature proposes two main theories, with conflicting predictions: one branch of the literature argues that managerial risk-taking behavior will increase over time, as managers gain confidence in their abilities. This idea is proposed by Avery and Chevalier (1999), and supported empirically by Chevalier and Ellison (1999b) (mutual funds), Hong, Kubik, and Solomon (2000) (analysts), and Lamont (2002) (macroeconomic forecasters). The alternative theories (which are consistent with this paper's hypothesis) argue that the opposite will be true: that is, that managerial risk-taking behavior will decrease over time as managers become more risk-averse due to career concerns, i.e., fear of losing their jobs. This theory is proposed by Prendergast and Stole (1996) and Graham (1999) and supported empirically by Graham (1999) (investment newsletters) and Li (2002) (analysts). In these models, time or experience is used as a proxy for the magnitude or importance of a manager's reputation. Logically, as managers progress in their careers, their reputations become more established. Another reasonable proxy for the importance of a manager's reputation is fund size. As funds grow, their managers become better known and may believe they have more at stake (both with respect to wages and status) than managers of smaller funds. For hedge funds in particular, larger funds typically pay larger wages, and to the extent that managers have their own money invested in their funds, managers with larger funds may have large dollar amounts of their own assets invested in the fund. Finally, the interaction of tenure and size can proxy for reputation: at the extremes, experienced managers with large funds should have higher reputational concerns than inexperienced managers with small funds. Thus, all three measures (tenure, size, and the tenure/size interactions) are used as proxies for reputation in the tests that follow. The first investigation into this relationship involves performing nonparametric tests. For each of the 982 funds, I calculate the mean value for manager tenure and fund size over the period 1994 to 21

33 2000. Funds are then sorted into thirds based on their average manager tenure as either young tenure, middle tenure, or old tenure. Additionally, funds are sorted into three size categories based on the average fund size: small, medium and large. Finally, funds are cross-sorted into nine tenure/size categories based on the prior classifications. Then, means of the risk measures (standard deviation, tracking error deviation, and beta deviation) are calculated for each tenure, size, and tenure/size interaction category. These means are reported in Table 6, and Wilcox rank-sum tests for differences in means are performed. Additionally, Figure 1 graphs the risk measures by reputation proxies. This figure and Table 6 clarify that the general pattern of risk-taking behavior indicates that managers reduce risk as they gain experience (i.e., as reputation becomes more important to them). The results are significant: for all risk measures, the Wilcox tests of the difference in means between the extreme categories (young and old, small and large, and small/young and old/large) are significant: risk shrinks over time. Additionally, Figure 2 plots the risk measures by the nine (9) fund tenure/size categories. The pattern is even more pronounced and shows a nearly monotonic decrease in risk by tenure/size category. While these findings are suggestive, they do not control for heterogeneity in fund characteristics, which could be influencing the results. Additionally, they aggregate the data over each fund's entire existence during , so that the time-series variation in the data is not considered. To address these issues, Table 8 performs a number of annual regressions where the dependent variable is the risk measure: either standard deviation, tracking error deviation or beta deviation, and the independent variable of interest is the measure of reputation (either tenure, size, or the tenure/size interaction). These regressions control for a number of fund characteristics which are listed in Table 2: indicator variables as to whether funds are listed on an exchange, are located onshore, are open to new investment, are open to non-accredited investors, have personal capital invested, and use leverage. Additionally included are the following variables: the number of months that must pass before an investor can redeem her shares (lockup redemption period), the number of months between entrance 22

34 periods, the fund's minimum investment, the management fee as a percentage of assets, the incentive fee as a percentage of profits, and fund style controls (See Table 7 for a description of the 17 fund styles). Panel A of Table 8 regresses standard deviation against manager tenure and all of the fund characteristic variables (specification 1a), fund size and all of the fund characteristic variables (regression 1b), and the manager tenure/fund size interaction variable and all of the fund characteristic variables (regression 1c). Panels B and C (regressions 2a-3c) perform the same three regressions for each of the other risk measures, tracking error deviation and beta deviation, respectively. The results in Table 8 are consistent with the results in Table 6: for all specifications and for all risk measures, tenure, fund size, and tenure/size interactions are strongly negatively related to risk. Managers with more reputation at stake (as measured by tenure, size, or tenure/size interactions) have lower risk. It is worth mentioning two additional results from Table 8. First, for two of the three measures of risk (standard deviation and tracking error deviation), the coefficient on incentive fee is positive and significant: managers with higher incentive fees incur more risk. Coupled with the positive relationship between incentive fees and performance documented in Table 5, this indicates support for the idea that, on average, incentive fees align manager and investor interests. Second, the coefficient on personal capital invested (a 0/1 indicator variable) is positive and significant for two of the three risk measures (standard deviation and beta deviation). This is somewhat puzzling, given this paper's hypothesis that managers with the most personal capital at stake have incentives to reduce risk in order to preserve their assets. One possible reason for this result might simply be that the personal capital invested variable is not very informative. Since it is a 0/1 variable, it says nothing about the magnitude of the manager's investment or about the percentage of personal wealth that the investment might represent. About 67% of managers answered yes to this question, and it is reasonable to think that their personal stakes in their funds might vary. With these caveats in mind, additional regressions are performed and the results are reported in Table 9. These regressions include interaction variables where the reputation variable of interest (either manager tenure, fund size, or tenure/size interactions) is interacted with the personal capital invested 23

35 variable. The focus is on the personal capital interaction variables: personal capital*manager tenure, personal capital*fund size, and personal capital*the size/tenure interaction variables. A negative coefficient on the personal capital*manager tenure, personal capital*fund size, and personal capital*large/old variables indicates that managers with greater reputational concerns (as measured by tenure or size) that have personal capital invested in their funds are taking on less risk than their counterparts with lower reputational concerns. The results from Table 9, regression 1a, indicate that for the standard deviation risk measure, older managers with personal capital invested do indeed take on less risk (this result is significant at the 1% level). Additionally, for the beta deviation risk measure, the results in Table 9, regression 3a indicates that managers having large funds and personal capital invested also reduce risk. Finally, also for the beta deviation risk measure, Table 9, regression 3c indicates that older managers with large funds and personal capital invested reduce risk as well. Thus, although the proxy for personal capital invested is likely quite noisy, there is at least some support that managers with greater career concerns and personal capital at stake do reduce risk. The main finding of this section is that old managers incur less risk than young managers. I hypothesize that due to career concerns the reason for this risk reduction among older managers is that they wish to reduce the likelihood that their funds will fail. Hence, the next section investigates whether these risk-taking behaviors do indeed lead to fund failure Risk-taking behavior and fund failure This paper argues that older managers have greater career concerns than younger managers. They have more at stake, in terms of personal wealth, status, and fee income, and thus, they reduce risk in order to decrease the likelihood of failure. Supporting the first part of this argument, the previous section documents that older managers do indeed reduce risk. This next section provides support for the second part of the argument: that increased risk-taking behavior results in fund failure. To study the relationship between risk-taking behavior and fund failure, I use a time-varying proportional hazards model. 17 The model allows for complete use of the time-series variation in the 17 Much of the following description closely follows Helwege (1996). 24

36 sample of hedge fund data. Additionally, time-varying proportional hazards models (which are a category of the more general hazard functions) have several advantages over the more commonly-used probit and logit models. First, they put fewer distributional assumptions on the data; second, they calculate the conditional rather than the absolute probability of failure (conditional upon not having failed in a prior period); and finally, they do not introduce sample-selection bias into the data. Instead of using annual failure rates, the more flexible proportional-hazards model allows for monthly failure times which reduces bias and adds precision to the estimates. Recent work in finance has used this type of model. Examples include Helwege (1996), who uses a proportional hazards model to examine the determinants of Savings and Loan failures in the 1980's, Lunde, Timmerman, and Blake (1999), who use the model to examine the determinants of mutual fund failures, and BGP (2001) who use this model to examine the determinants of hedge fund failure, notably volatility and manager tenure Description of proportional hazards model The time-varying proportional hazard model estimates the relationship between the hazard rate (the likelihood of fund failure), λ(t), and a number of explanatory variables, z(t), that are permitted to vary over time. The proportional hazard function is specified so that the explanatory variables shift an underlying baseline hazard function up or down. The baseline hazard function, λ o (t), can follow any distribution for which proportionality holds. Examples include the Weibull, exponential, and lognormal distributions. The time-varying proportional hazard function is described by the following equation: () λ( tzt ; ( )) = λ ( t ) 0 e β zt (5) where β is the set of coefficients to be estimated. Cox (1972) describes how β can be estimated by maximizing the partial likelihood function of the probability of failure observed in the sample. β is estimated from inferences on the conditional 25

37 probability of failing in a given time period. Because of proportionality, the estimation ignores the baseline hazard function, which makes specifying a functional form for the baseline unnecessary. Assume that there is a sample of n hedge funds, k of which fail during the sample period with failure times t 1 < t 2 < < t k. The assumption of this model is that each failure occurs in a different time period, and the failures are ordered from 1 to k chronologically. 18 The remaining n-k funds are censored and have no failure times during the sample period. However, these funds could fail some time after the sample period ends. Assign δ t equal to 1 if a fund in period i fails and zero if it does not fail. Let z i (t) be z(t) for the fund with failure time t(i) and let z j (t) be z(t) for each fund at risk at time t(i). R i is the set of funds at risk of failure in period i. The partial likelihood function to be maximized is: n L( β ) = i = 1 β z ( t ) e i i β z j( ) ti Ri e δ i (6) Intuitively, Equation (6) examines each hedge fund that fails (one per time period) and compares its explanatory variables to the explanatory variables on the set of hedge funds that could have failed during the period but did not. If the values of the explanatory variables for those that failed differ from the values of the explanatory variables for those that survived, the coefficients will be significantly different from zero This methodology assumes no ties (that is, no funds failed in the same period). In the sample of interest, there are several tied failure times, which will be addressed later. For simplicity, the case with no ties is described here. 19 Maximum likelihood estimation is used to estimate the partial likelihood function. Since this process is computationally demanding, an approximation (see Breslow (1974)) is usually used to save time and computer resources. However, the Breslow approximation can be less accurate when there are many tied failure times (as is the case in the hedge fund data.) Thus, following estimation uses the exact method (see Kalbfleisch and Prentice (1980)). 26

38 Estimation of proportional hazard model In this section, I estimate a time-varying proportional hazard model, using the time that a fund drops out of the sample as the failure time for that fund. As noted in Section 2.2, there are a number of reasons that funds may be removed from the database. Some of the funds actually fail, while others voluntarily stop reporting their data. Although TASS often provides the reason that the fund has been removed, it is often ambiguous and difficult to interpret. Thus, the analysis below is performed on two samples: one that considers as dead all 285 funds that TASS categorizes as dead (out of the total sample of 982 funds), and another that considers as dead only the funds that clearly state that the fund had failed or gone bankrupt (the rest were dropped from the sample). The results for both samples are nearly identical, implying that most of the funds that left the sample had similar characteristics consistent with actual failure. Therefore, Table 10 reports the results using the larger sample. Table 10 employs the time-varying proportional hazards model to analyze the general relationship between termination and risk-taking behavior. These estimates include controls for fund characteristics (described in Table 1 and Section 2.4, above) and seasoning (year effects). Including seasoning effects reflects the possibility that termination probabilities may vary from year to year due to factors such as evaluations based on absolute performance or changes in the labor market for hedge fund managers. The first column, regression 1, examines the relationship between termination and standard deviation. Columns 2 and 3 examine the relationship between termination and tracking error deviation and beta deviation, respectively. All regressions include the fund characteristic variables as well as a measure of fund performance (excess returns above the risk-free rate). In the table, a negative coefficient indicates a positive likelihood of survival, while a positive coefficient indicates a positive likelihood of failure. Examining the results in Table 10, two patterns emerge. First, as would be expected, good performance is negatively related to failure. Good funds survive. Second, all three risk measures are positively related to failure. Funds that take on more risk have a higher likelihood of failing. This finding confirms the second part of the hypothesis discussed above: certain types of risktaking behavior do increase the likelihood of failure, which provides an incentive for managers with career concerns to reduce these risk-taking behaviors. 27

39 The approach of this section has been to link various risk-taking behaviors to the probability of fund failure. It provides evidence that engaging in certain behaviors significantly increases the probability that a fund will fail. Another way to think about risk-taking is to consider fund failure itself as evidence of risk-taking behavior. To the extent that standard deviation, tracking error deviation, and beta deviation do not capture every type of risk a manager may incur, fund failure itself can proxy for risky behavior. Therefore, the next analysis focuses on the relationship between fund failure and manager tenure, without regard to pre-determined measures of risk. If the relationship between manager tenure and fund failure is negative (that is, if older managers survive more often than young), this relationship provides indirect evidence that older managers have been taking on less risk. In order to capture the effect of manager tenure without regard to risk-taking behavior, I reestimate the model excluding risk and performance measures. Column 4 of Table 10 shows these results. In this estimation, the coefficient on the tenure variable is This indicates that for each month of experience, the average probability of failure decreases by about 6%. Consistent with the results in columns 1 to 3, this relationship suggests that older managers are reducing risk over time to decrease their probability of failure. To examine how this relationship varies by tenure and over time, I estimate the shape of the survivor function (which is the inverse of the hazard function). Holding all other variables constant (at their cross-sectional averages), the function is estimated at nine levels of the tenure variable, representing the 10th through the 90th percentile of this variable. Figure 3 graphs the estimated survival function at three percentiles: 10th, 50th, and 90th. Examining this figure, it is clear that holding all else constant, more experienced managers have a much higher probability of survival. Interpreting fund failure as a proxy for risk, this figure is consistent with earlier results that older managers reduce risk over time. In summary, this section provides evidence that certain types of risk-taking behaviors can lead to fund failure, establishing an implicit incentive for managers who wish to avoid failure to reduce risk. This is consistent with the hypothesis of this paper: Older managers with strong career concerns reduce certain types of risk-taking behaviors that lead to fund failure. This section also provides indirect 28

40 evidence that older managers reduce other types of risk as well: taking fund failure itself as a proxy for risky behavior, the survival function in Figure 3 provides strong evidence that the probability of failing decreases significantly with manager tenure. The previous two sections have established that older managers take on less risk than young, and this behavior has been explained in terms of career concerns -- older managers have more at stake, and thus, have stronger incentives to keep their jobs than young managers. The next section examines whether the reduction in risk among older managers can also explain their reduction in returns The relationship between risk-taking behavior, returns, and manager tenure In Section 2.3, a negative relationship between hedge fund manager tenure and performance was established. Sections 2.4 and 2.5 find that older managers take on systematically less risk than younger managers, and explain this result as being driven by increasing career concerns of older managers. This section investigates the hypothesis that the reduction in risk-taking behavior of older managers can explain their reduced performance. To examine this relationship, Table 11 performs a number of cross-sectional regressions where for all the regressions, the dependent variable is annual return less the risk-free rate, and the independent variables include manager tenure and all fund characteristics and one of the three measures of risk-taking behavior -- standard deviation, tracking error deviation, and beta deviation. The methodology is similar to that in Section 2.3 (and Table 5), but the regressions in this section are performed annually instead of monthly, since the three risk-taking measures are estimated on an annual basis. To establish the baseline relationship between returns and manager tenure (on an annual level), the first set of regressions (Panel A) do not include the risk-taking variables as regressors, so the first column (1a) includes as independent variables manager tenure and fund characteristics only. The second column (1b) includes controls for exposure to passive indices, while the third column of Panel A (regression 1c) also includes controls for exposure to hedge fund indices. The next three sets of regressions (Panels B-D) are identical to regressions 1b and 1c, except they also include either standard deviation (Panel B), tracking error deviation (Panel C), or beta deviation (Panel D). Examining Table 29

41 11, we find a positive relationship between the three risk measures and hedge fund returns, confirming that the risk-return relationship holds for this sample of hedge funds. In other words, risk-taking behavior is rewarded with significantly higher returns. Most importantly, the coefficients on the manager tenure variable decrease significantly when risk-taking behavior is considered. These results are economically significant: without considering the impact of risk-taking behavior or correlations with market beta coefficients on returns, the difference in returns for a 52 year old manager versus the average (47 year-old) manager is about -4% per year. When market beta controls are included, this difference drops to about -3.5% per year. However, when both the market beta controls and the risk-taking measures are included, the difference drops to about -2% per year. Differences in risk-taking behavior explain nearly half of the differences in returns between old and young managers. The remaining difference may be due to other risk factors that are more difficult to measure, or to omitted variables that capture dynamic trading strategies of these managers. Thus, the results of this section confirm the hypothesis that the negative relationship between manager tenure and returns can be explained by reduced risk-taking behavior of older managers Robustness tests This section includes two robustness tests. The first attempts to rule out the following explanation for the above findings: Perhaps older managers are not reducing risk over time at all, but instead, they have taken on lower risk all along, thus increasing their probability of survival. Section refutes this argument. The second robustness test (see Section 2.7.2) relates fee income to risktaking to provide additional support for the idea that as managers' salaries increase, they reduce risk to increase the probability that their funds (and their careers) will survive Endogeneity tests One concern with the above tests is that they do not rule out the following possibility: Perhaps older managers are not, in fact, reducing risk over time, but rather, have taken lower risk all along, thus 30

42 increasing their probability of survival. Examining this issue is difficult, due to data constraints. The first problem is that data prior to 1994 is not free of survivorship bias. Thus, using this data to test this idea (perhaps with a fixed-effects regression) is not appropriate: if dead funds are not included before 1994, this introduces bias. The managers of failed funds would likely have taken on higher risk at the beginning of their lives, and if they did not properly reduce risk, they would not be included in the sample. Thus, I can only use the data since This is also problematic, since some of the funds in the sample have been in existence for many years prior to 1994, and thus, the early years of their existence (when their risk was likely to be the highest) are not included in the sample. Given these caveats, the following tests were performed in an attempt to rule out the idea that hedge fund managers that survived did not reduce risk over time, but rather, took less risk all along. To test this alternative hypothesis, all funds in existence during the year 1994 are followed forward to the year 2000, and I record the following data: each of their risk measures (standard deviation, tracking error deviation, and beta deviation) for each year, and whether the fund failed at some point during the entire sample period. Then, each year, the funds are sorted into deciles based on manager experience. For each year, for each decile, the funds are then sorted into two categories: whether the fund was alive or dead at the end of the sample. The idea is to examine whether the risktaking behavior at the beginning of the sample period of the funds that eventually failed differed from the funds that did not fail. If I find that the funds that eventually failed took on more risk at the beginning of their lives than the funds that did not fail, the alternative hypothesis is supported. However, if there is not a significant difference between risk-taking behavior for extant and failed funds at the early stages of their lives, then this paper's argument that risk-taking changes with age is supported. Using Wilcox rank-sum tests, I find that the risk-taking behavior of young managers (decile 1) at the beginning of the sample does not differ significantly between funds that survived the entire sample and funds that eventually failed. This indicates that at the beginning of their lives, funds have similar risk levels, which is consistent with the theory and empirical results presented: funds reduce risk over time in order to survive. Additionally, I find that for decile 10 managers (those with the most 31

43 experience) at the beginning of the sample, there is a significant difference between levels of risk-taking behavior among surviving and failed funds: failed funds have higher levels of tracking error deviation and standard deviation than surviving funds, indicating that as these funds got older, those that reduced risk were able to survive, while those that did not were not Fee income One argument made earlier in this paper relates fee income to risk-taking behavior. The idea is that successful hedge fund managers are so well-paid that they will go to great lengths to protect their income. The opportunity set in the industry is such that hedge fund managers will almost surely take a very large pay cut if they lose their jobs. However, if they reduce risk and keep their jobs, they are able to maintain this earning potential, even if the reduction in risk leads to lower returns. Since managers earn a percentage of profits, as long as their returns are positive they are still paid. And, if their funds are large (as is likely the case with high fee earners in the past), they can still earn exceptional salaries on mediocre investment returns. To test this hypothesis, Table 12 regresses each of the three risk measures against variables that capture cumulative prior fee income for each manager. This variable is a simple estimate of the fund manager's cumulative income since To estimate this variable, I multiply each year's average fund size by the management fee percentage, and add the incentive fee, which I estimate as the annual return for the fund times the incentive fee percentage times the average total assets for the year. Each year, for each fund, I perform this calculation and sum the results over the life of the fund. The estimates for this variable are then sorted into thirds, and assigned 0/1 indicator variables representing the magnitude (high, medium, or low) of cumulative fee income. A number of caveats are in place: first, the fee income is not the only income managers earn from their funds. The majority (67%) have some of their own assets invested in the fund, which would increase the estimates of fee income. However, as noted above, this personal capital invested variable is likely not very useful, as it does not provide information about the magnitude of personal capital that managers have invested, nor what proportion of each manager's net worth that personal capital comprises. Second, the cumulative 32

44 measure is only calculated from 1994 onward. Some of the funds in the sample have been in existence since However, the data for these funds is not as complete for these prior periods, so the measure is only calculated from 1994 forward. Both of these issues likely cause the estimate of cumulative fee income to be understated, which biases against finding any relationship in the tests that follow. A final caveat is that some funds have high water marks, meaning that if the net asset value of the fund falls below the level of the prior year, the manager has to makeup the difference before he can begin collecting fees again. This problem would cause the cumulative fee income variable to be overstated. Still, the understatement problem caused by not including years prior to 1994 and not including profits earned on managers' own assets is probably larger than the overstatement problem of not accounting for high water marks. Keeping these caveats in mind, Table 12 regresses each of the risk measures against the fee income indicator variables for high cumulative fee income and low cumulative fee income (the middle fee income variable is excluded.) These regressions are performed in column a of each panel (Panel A, column 1a uses standard deviation as the dependent variable; panel B, column 2a uses tracking error deviation as the dependent variable, and Panel C, column 3a includes beta deviation as the dependent variable). For each of these regressions, manager tenure, fund characteristics, and market beta control variables are also included. Examining the results, for both herding measures of risk (tracking error deviation and beta deviation), the results are consistent with the hypothesis that managers with more to lose in terms of fee income take on the least risk, while those with less to lose take on the most risk. The coefficients for standard deviation are not significant. Next, another variable is created that interacts the fee income indicator variables (low, medium, and high) with the manager tenure variable (young, middle, and old). The idea is that old managers with high fees likely have more at stake than young managers with low fees. If this is true, we would expect to find negative coefficients on the old age/high fees variable. And, consistent with the results for cumulative fee income, for two of the three regressions (tracking error deviation and beta deviation), the coefficients are as expected, indicating again that managers with more at stake reduce 33

45 risk. These findings provide further evidence that manager career concerns are increasing with reputation: managers with high fee income have strong explicit incentives to keep their jobs, and these incentives are more important to more experienced managers Conclusions The key finding of this paper is summarized as follows: As hedge fund managers gain job experience, they reduce risk to increase their likelihood of survival. One effect of this risk-reducing behavior is that their returns decline significantly. As managers progress in their careers, the desire to keep their jobs outweighs their desire to earn above-average returns. These empirical findings differ sharply from evidence from the mutual fund industry: in mutual funds, old managers take on more risk than young. This finding is consistent with the agency costs and career concerns of mutual fund managers. For example, the termination/performance relationship in mutual funds shows that young managers are punished more than old managers for poor performance, but for average to good performance, the termination rates do not vary among young and old managers. Therefore, it is arguably most important for young managers to avoid poor performance (and risk termination) rather than to achieve excellent performance. Thus, young managers herd, so as not to risk a poor performance outcome relative to their peers. As their careers progress and they become more experienced, they are less likely to herd and take the chance of an unfavorable outcome for at least three reasons: first, the probability of failure for a poor outcome is much less than for younger managers, second (in contrast to most experienced hedge fund managers) they do not likely have substantial personal assets invested in their funds so a poor showing will not affect their personal capital substantially, and finally, the flow/performance relationship indicates that the very best funds attract the most new inflows, which are directly related to a manager s compensation. 20 This finding of 20 Chevalier and Ellison (1997) document an asymmetry in the flow/performance relationship. While the very best funds attract significant inflows, the very worst funds do not experience significant outflows. Average funds experience net inflows. See also Gruber (1996), Goetzmann and Peles (1997), and Sirri and Tufano (1998). 34

46 more risk-taking by older managers is also consistent with theoretical models arguing that managers learn more about their abilities over time, and thus, are more confident in making investment decisions that deviate from their peers. Likewise, the hedge fund findings in this paper correspond with the agency costs and career concerns of hedge fund managers. In hedge funds, older managers have more to lose than young should their funds fail they usually have significant personal assets in their funds, they understand that risky behavior increases their probability of termination, they know that it is difficult to start a new hedge fund if their current fund fails which would result in a large pay cut and loss of autonomy all of which induce them to take less risk over time. Additionally, the pay structures in hedge funds encourage risk-taking among younger managers versus old: since hedge fund managers are compensated both by incentive fees (a percentage of profits) and management fees (a percentage of assets), young, smaller funds focus on maximizing profits for two reasons: first, high profits increase their incentive fees substantially, and second, high profits attract more assets to their funds, increasing their future management fees. For older managers with larger funds, they earn significant fees from the management fee component alone, which might induce them to take less risk in an attempt to increase their incentive fees. These findings support the theoretical models arguing that young managers wish to showcase their abilities early in their careers by taking innovative actions, while older managers tend to reduce risk, fearing the ramifications on their careers of deviating from their own prior actions. Finally, these findings are also broadly consistent with corporate finance reputational models (see Hirshleifer and Thakor (1992) and Diamond (1991)) which argue that managers with greater reputational concerns will avoid risk with respect to determining capital structure or debt acquisition. The results of this paper indicate that differences in risk-taking behavior explain about half of the differential in returns between young and old managers. These results have implications for investors in hedge funds: young managers are substantially better than old, which they achieve by using more aggressive trading strategies and avoiding herding behavior. However, the increased probability of failure from young to old funds is striking: for a manager with one year of tenure, the probability of surviving the next six years is only about 35%. For a manager with four years of tenure, the probability 35

47 of surviving the next six years is nearly 90%. Due to the trade-off between returns, risk, and survival, fund selection by investors should take into account this relationship: these results indicate that selecting very young or very old funds are not likely in investors' best interests. Additionally, these results have ramification for the appropriate design of incentive contracts. While it has been thought that the incentive fee structure used by hedge fund managers appropriately aligns the incentives of managers and investors, this paper provides evidence that, at least in some cases, it does not. Particularly among older funds, the desire to survive and remain in the industry can outweigh the desire to take the necessary risks to achieve superior returns. Thus, as suggested by Dybvig, Farnsworth, and Carpenter (2001), perhaps contracts should be designed that encourage managers to take on more risk. For younger managers, these contracts reward them for actions they already are taking, and for older managers, perhaps these contracts could provide incentive to take on more risk. This is a fruitful area for future research. 36

48 CHAPTER 3 DO HEDGE FUNDS EXHIBIT PERFORMANCE PERSISTENCE? A NEW ANALYSIS THAT ACCOUNTS FOR MANAGER TENURE 3.1. Introduction In selecting a hedge fund for investment, is it helpful to consult the manager s prior performance record? If past performance is indicative of future results, there is value to investors in this information. If not, then investors may be better off selecting a manager based on his reputation, investment style, or trading costs. Recent research regarding this issue finds consistent results: there is some evidence of short term (one to three month) persistence among individual hedge funds. (See Agarwal and Naik (1999, 2000), Bares, Gibson, and Gyger (2002), and Baquero, ter Horst, and Verbeek (2002).) This persistence is not driven by the existence of survivorship bias. At longer time horizons (semi-annual or beyond), however, persistence largely disappears; see e.g., Brown, Goetzmann, and Ibbotson (1999), and Brown and Goetzmann (2001). With a more rigorous approach that controls for common risk and style factors in hedge fund returns, this paper finds no evidence of persistence (short or long-term) when funds are selected based on past performance alone. Style factors explain the previous findings of short-term persistence, consistent with the work of Brown and Goetzmann (2001) who show that certain styles perform well in certain periods; in other periods, these same style do not perform as well. Thus, controlling for style is important in an analysis of performance persistence among hedge funds. However, while controlling for style casts doubt upon the previous findings of persistence, there is another important factor that should be considered in constructing a test of performance persistence manager tenure. Boyson (2003) shows that less experienced managers (hereafter referred 37

49 to as young or low tenure managers) significantly outperform more experienced managers (hereafter referred to as old or high tenure managers). Specifically, in a sample of hedge funds for the period 1994 to 2000, she finds that after controlling for common risk and style factors, the annual difference in performance between young and old managers drops by about 0.75% for each year of experience. That is, a manager who is 52 years of age has annual performance about 4% lower than the average (47 year old) manager in the sample. Her results suggest the following: since low tenure managers are better, then a bad return for a low tenure manager is more likely to be due to bad luck than for a high tenure manager. Likewise, a good return for a high tenure manager is more likely to be due to good luck than for a low tenure manager. In other words, good (bad) returns for low tenure managers are likely to be due to superior manager skill (bad luck); good (bad) returns for high tenure managers are likely to be due to good luck (lack of manager skill). Thus, properly accounting for manager tenure when performing a persistence analysis should detect performance persistence among the young versus the old hedge fund managers. The remainder of the paper designs a more powerful test of performance persistence, taking into account manager tenure. I construct a portfolio that takes a long position in low tenure/past good performers and a short position in high tenure/past poor performers, which by design, should maximize the likelihood of finding persistence. And, this portfolio demonstrates quarterly persistence: controlling for risk and style factors, the excess performance is about 9% annually, which is both economically and statistically significant. This result is driven primarily by persistent underperformance among old, past poor performers. Next, we explain the concentration of persistence among old past performers with the following hypothesis: that the termination relationship is more performance-sensitive for young managers. If this is the case, then old, poor performers have a low probability of being terminated and thus are more likely to persist in the next period. This hypothesis is motivated by theoretical literature that suggests that young managers will be punished more severely for poor performance than are old See, for example, Zwiebel (1995) and Holmstrom (1999) who describe the process by which investors find out about managers with a learning model. Each period, investors observe a new performance outcome (in this case, a monthly return) by which they learn about manager ability. Since 38

50 It is also motivated by empirical results for mutual fund managers and security analysts. (See Chevalier and Ellison (1999b) and Hong, Kubik, and Solomon (2000)). A conditional survival analysis documents the following results: conditional on having been a poor past performer (in the bottom third of returns), young managers are significantly more likely to be terminated than old. Also, conditional upon having been a middle performer (in the middle third of performance) young managers are still significantly more likely to be terminated than old. Only when we condition upon having been a past good, performer (top third of returns) is there no difference in survival rates among young and old managers. Thus, being in the bottom two-thirds of performance significantly hurts young managers relative to old. The second survival analysis (this time, conditional upon manager tenure) establishes the following result: conditional on being young, past poor performers are more likely to be terminated than past good performers. However, conditional on being old, there is no difference in survival rates between past poor and past good performers. These findings support each other are broadly consistent with the idea that investors are more likely to tolerate poor performance from managers with moreestablished reputations (i.e., more experienced managers). Thus, this finding helps to explain the continued poor performance among old, past poor performers. This paper makes two contributions to the literature. First, it takes advantage of the empirical result that young managers outperform old to design a test that detects risk- and style-adjusted performance persistence at the quarterly level. While selecting funds based on past performance alone results in a finding of no performance persistence, the more powerful approach of choosing funds based on both past performance and manager tenure does result in a finding of persistence. This persistence is mostly concentrated among the old, past poor performers. To our knowledge, this is the first paper in the literature to test for performance persistence in this manner. Second, this paper explains this finding of persistence (notably, that it is concentrated among the old, past poor performers) as being driven by differences in termination rates among young and old managers. Specifically, there is an interesting asymmetry in the shape of the termination and there are more observations for older managers than young, this implies that the sensitivity of a manager s reputation is less dependent on the most recent observation. Hence, old managers are less likely to be assessed as inferior based on a recent bad outcome than are young. 39

51 age/performance relationship: the termination process is much more performance-sensitive for young managers than for old. At first glance, this relationship appears similar to that in the mutual fund industry: Chevalier and Ellison (1999b) also find that young mutual fund managers are more likely to be terminated than old for poor performance. However, there is a key difference between their results and the results of this paper. They find that for young mutual fund managers, the probability of termination decreases steeply with performance when managers have negative excess returns, but it is fairly insensitive to performance differences at positive excess return levels. 22 As long as a young manager s returns are positive, his probability of failure does not differ from that of an older manager. By contrast, in the hedge fund industry the performance threshold is much higher, and is measured relative to other managers rather than based on absolute performance: unless a young hedge fund manager s returns are in the top third of managers, his probability of failure is significantly higher than that of an older manager. In other words, young hedge fund managers have to beat two-thirds of other managers to reduce their probability of failure to the same as that of older managers. Clearly, this high threshold of performance in the hedge fund industry sets up a different incentive structure for young mutual fund versus young hedge fund managers: while young mutual fund managers concerned about survival need only avoid posting a negative excess return (which gives them an incentive to play it safe and avoid idiosyncratic risk), young hedge fund managers that are concerned with survival need to post returns in the top third of all performers (which gives them an incentive to make bold investment decisions relative to other managers.) The empirical evidence for hedge fund managers is completely consistent with this implied incentive: Boyson (2003) shows that young managers herd less and take on more idiosyncratic risk than old. While this paper contributes to a relatively small and recent literature in the hedge fund industry, researchers have been studying persistence in the mutual fund and pension fund industries for many years, with mixed results. An early study by Jensen (1968) finds no support for persistence. Papers supporting persistence over five to ten year periods include Grinblatt and Titman (1992), Elton, Gruber, Das and Hlavka (1993), and Elton, Gruber, Das and Blake (1996), who attribute this 22 Chevalier and Ellison (1999b), page

52 persistence to manager stock-picking ability. Support for shorter-term (one to three year) persistence comes from Hendricks, Patel, and Zeckhauser (1993), Goetzmann and Ibbotson (1994), Brown and Goetzmann (1995), and Wermers (1999). Carhart (1997) shows that the one-year momentum effect of Jegadeesh and Titman (1993) accounts for much of the performance persistence found by Hendricks, Patel, and Zeckhauser (1993), and that differences in mutual fund expenses and trading costs can explain nearly all of the remaining persistence. Christopherson, Ferson, and Glassman (1998) apply conditional performance evaluation techniques to a sample of pension funds, and show that the conditional approach is better able to detect persistence and predict future performance than unconditional (linear) methods. This persistence is mostly concentrated among the worst performers. More recently, using a Bayesian approach with daily mutual fund data, Bollen and Busse (2002), and Busse and Irvine (2002) find evidence of quarterly performance persistence that is not explained by momentum. This paper is organized as follows. Section 3.2 describes the data. Section 3.3 describes the performance measures and portfolio formation process used in Section 3.4. Section 3.4 performs the first analysis of persistence, and shows that when common risk and style factors are properly accounted for, there is no evidence of quarterly persistence when funds are selected based on past performance only. Section 3.5 motivates and designs a more powerful test of persistence that incorporates manager tenure, and shows evidence of persistence at the quarterly level. Section 3.6 performs a detailed survival analysis to explain the patterns in persistence found in Section 3.5. Section 3.7 discusses the results of the survival analysis in light of the career concerns literature. Section 3.8 concludes. 3.2 Data Data was provided by Tremont Advisory Shareholders Services (TASS). TASS has been collecting hedge fund data directly from managers since the late 1980's, and currently has over 2,400 funds in their database, both living and dead. 23 The database includes monthly net-of-fee returns, as well as expenses, fees, size, terms, age, and style of the funds. For the quarterly persistence tests, we 23 TASS has maintained data on dead funds since

53 require that each fund have at least 6 months of consecutive returns for inclusion in the sample. For the semi-annual persistence tests, each fund must have at least 12 consecutive months of returns for inclusion, and for the annual persistence tests, each fund must have at least 24 months of consecutive returns. For all time frames (quarterly, semi-annual, and annual) each fund must have at least $5 million in assets during the period January, 1994 to December, In constructing the sample, an important issue must be considered: backfilling or instant history bias (see Edwards and Park (1996). On the date that TASS adds a new fund to their database, they backfill historical returns. Typically, a hedge fund manager will start his fund with a limited amount of personal capital before selling shares to the public. He hopes to compile a good track record so as to eventually attract outside investors. Thus, most funds arrive in the database with a history of strong performance which was never available to the public, which biases returns upward. This difference is often large -- using the TASS database, Fung and Hsieh (2000) calculate the bias as about 3.6% per year. The average incubation period in our sample is about one year; thus, to control for this bias we drop the incubation period for each fund. The final sample used to perform the quarterly persistence tests includes 1,659 funds with at least 6 months of returns, $5 million in assets, and all of the fund characteristic variables. Table 13 includes summary statistics of the return and fund characteristic variables. The final sample for the semi-annual persistence tests has 1,503 funds, and the final sample for the annual persistence tests has 982 funds. Summary statistics for these samples do not differ materially from the quarterly sample. As a brief summary, the average fund in the sample is about three years old, has about $80 million in net assets, and annualized excess returns of about 8%. More than half the managers use leverage and have personal capital invested, and the styles of US Equity, Relative Value, Event Driven, and Emerging Markets make up over 60% of the funds in the sample. 24 About 30% of the sample consists of funds that failed at some time during the sample period. 24 See Table 7 for a description of the fund styles. 42

54 With respect to failed funds, a number of researchers have emphasized the importance of including dead as well as extant funds in an analysis of performance. 25 Not including defunct funds in the sample can bias returns upwards. In an earlier study of offshore hedge funds, Brown, Goetzmann, and Ibbotson (1999) find that not including defunct funds in the sample biases returns upwards by 3% per year. Liang (2000) compares two major hedge fund databases (Hedge Fund Research (HFR) and TASS), and finds that the TASS database tends to contain more dead funds, and thus, should be more appropriate for the type of study in this paper. He finds survivorship bias in the TASS data of about 2% per year, which is approximately what we calculate using the sample of 1,659 funds. 3.3 Performance measures and portfolio formation process Performance measures This paper uses a multi-factor model to control for common risk factors in hedge fund performance. Since hedge funds have exposures to a number of markets, and can engage in dynamic trading strategies (such as using options, futures, and leverage), using a broad set of indices is appropriate (see Fung and Hsieh (1997) and Ben Dor, Jagannathan, and Meier (2003) for further discussion). The passive indices are obtained from Datastream and include: the US Trade Weighted Dollar index to capture currency risk, gold and commodity indices, and the Lehman Brothers 30-year Treasury bond and US aggregate bond indices. For stock market risk, the Value-Weighted CRSP index (obtained from WRDS) is used. Additionally included are the Fama-French (1992,1993) SMB (a zero-investment portfolio constructed by subtracting the returns of large market capitalization firms from the returns of small capitalization firms) and HML (a zero investment portfolio constructed by subtracting the returns of low book to market ratio stocks from the returns of high book to market ratio stocks) factors, as well as Jegadeesh and Titman's (1993) momentum factor (MOM) a zero-investment portfolio constructed as 25 In the mutual fund literature, see, for example, Brown, Goetzmann, Ibbotson, and Ross (1992), Wermers (1996), and Carhart (1997). 43

55 the spread between the performance of stocks which were in the top 30% of returns in the prior twelve months and those which were in the bottom 30%. 26 Table 14 reports summary statistics for the passive indices and other risk factors (HML, SMB, and MOM) used in the paper. Since a number of researchers have stressed the importance of considering style in a study of hedge fund performance (see, e.g., Fung and Hsieh (1997), Brown, Goetzmann, and Ibbotson (1999), Ibbotson and Patel (2002), and Ben Dor, Jagannathan, and Meier (2003)), the model includes style as well as common risk factors. These style factors are hedge fund indices are published jointly by Credit Suisse First Boston (CSFB) and TASS, and represent a number of hedge fund trading strategies or styles. They are constructed so as to minimize survivorship bias. For further detail about the indices used, see Table 3. Table 14 reports summary statistics for all hedge fund style indices used in the paper. The model is as follows: K r = α + b F + b H + ε pt pt pkt kt pdt dt t k= 1 d= 1 D (7) where r pt is portfolio p s return in month t in excess of the risk-free rate (t = 1 to T months), the F kt s are each of the passive index returns and HML, SMB, and MOM (k = 1 to K) in month t, and the H dt s are each of the hedge fund index returns (d = 1 to D) in month t Portfolio formation process This section describes the methodology by which hedge funds are sorted into portfolios to evaluate performance persistence. Similar to HPZ (1993), and Carhart (1997), we form portfolios on lagged hedge fund returns and test for both short-term and longer-term persistence. 27 As noted earlier, 26 These returns were obtained from the website of Kenneth French. 27 When we form portfolios based on lagged returns, we do this in two different ways. The first method is to form portfolios based on lagged excess-of-risk-free rate returns. The second method forms portfolios based on their returns in excess of their style average. For both methods, the raw returns of the portfolios are then regressed against a number of common risk factors AND style factors as in 44

56 quarterly (but not longer) persistence in hedge funds has been documented by Agarwal and Naik (2000), Baquero, ter Horst, and Verbeek (2002), and Bares, Gibson, and Gyger (2002). Following Carhart (1997), funds are sorted into decile portfolios based on lagged returns, which are net of fees and expenses and in excess of the risk-free rate. They span the period January, 1994 to December, 2000 for a maximum time series per fund of 84 months. Not all funds have all 84 months of returns available. Some funds fail before the end of the sample period, and other funds do not begin until some time after January, 1994; however, as long as a fund has at least 6 consecutive monthly returns, it is included in the sample. 28 Funds are sorted into portfolios based on three different time frames: three months, six months, and one year. 29 This initial period is called the formation period, and for the persistence analysis, each portfolio is held for a length of time equal to its formation period. For example, at the three month time horizon, funds are first sorted into portfolios using the prior quarter's return. These portfolios are held for three months, and equal-weighted portfolio returns are calculated for each of the three months. Every three months, portfolios are re-formed. This yields a time-series of 81 monthly returns for each portfolio (the first three months are used in the initial formation period). For the sixmonth formation period, there are 78 monthly portfolio returns, and for the one-year formation period, there are 72 monthly portfolio returns. As noted above, the process of using portfolios of funds to study persistence has been used by HPZ (1993) and Carhart (1997). Another common methodology examines persistence using individual funds (rather than portfolios of funds). See, for example, Brown and Goetzmann (1992), Goetzmann and Ibbotson (1994), and Agarwal and Naik (2000). While the individual fund approach is appropriate in certain cases, this paper uses the portfolio approach, since this approach is better-suited to a study of hedge funds. The main reason is that the portfolio approach allows for a risk- and style- adjusted analysis of persistence while using a minimum number of time-series observations. As described Equation (7). We find that the results in the paper are robust to whether the portfolios are formed based on excess-of-risk-free-rate or excess-of-style-category returns. Thus, only the results from the excessof-risk-free rate are reported throughout. 28 We require six months of returns for quarterly tests, 12 months for semi-annual tests, and 24 months for annual tests. 29 Since in all the tests performed, funds never exhibit persistence at the semi-annual or annual level, this paper only reports results for quarterly examination periods. Results for other time horizons are available from the author by request. 45

57 above, in the portfolio approach, each period a portfolio of funds is created for which an average return is calculated, resulting in a long time-series of portfolio returns which may then be adjusted for common risk and style factors. In this case, only the returns used in the initial formation period are not analyzed for persistence, and persistence may be examined for a larger number of funds over any time frame (e.g., monthly, quarterly, semiannually, etc.). By contrast, when persistence among individual funds is studied, properly adjusting for risk and style factors is typically accomplished by calculating intercepts (alphas) from time-series regressions over the period being analyzed. These alphas are then compared to each other on a period-over-period basis to test for persistence. If one wishes to use a multi-factor model to control for a number of risk and style factors, this necessarily lengthens the time frame over which persistence may be analyzed. For example, this paper uses an model with eighteen (18) independent variables. A study of persistence among individual funds that uses an 18-factor model will not be able to test for persistence at the quarterly, semiannual, or even annual level, since alphas could not be calculated for periods shorter than 18 months due to the degrees of freedom constraint. Additionally, funds would be required to have a much longer time-series of returns than in the portfolio approach, which would reduce the sample size significantly. Thus, while the individual fund approach is well-suited to mutual fund studies (which have much longer time series of returns and typically have exposures to a smaller number of risk and style factors), it is less suited to a study of hedge funds, with their short time-series of returns and exposures to a large number of risk and style factors. Hence, the portfolio approach is used in the analyses that follow Do hedge funds exhibit risk and style-adjusted persistence? In this section, we test for performance persistence when controlling for fund exposure to common risk and style factors. Controlling for style is accomplished by including hedge fund style indices as independent variables, using Equation (7) from Section A slightly different way to control for the effect of style on performance would be to model the return process for each fund style. Fung and Hsieh (2001) refer to this approach as developing asset-based style factors. For example, Fung and Hsieh (1997) show that the returns of the trend-following style (which is probably most 46

58 closely related to the Global Macro style in this paper) can be modeled as a look-back straddle on the S&P 500 index. Mitchell and Pulvino (2001) show that the returns to merger arbitrage hedge funds closely resemble short positions in put options. Finally, Agarwal and Naik (2003) show that a good deal of variation in hedge fund returns can be explained with simple option buying/writing strategies. While these approaches are very helpful in understanding the return processes for each hedge fund style, there are not yet asset-based factors developed for each fund style. Since we wish to control for as much of the style exposure as possible, we use-reported styles as regressors. 30 As described in Section above, decile portfolios are formed based on a fund's lagged quarterly returns in excess of that fund s style s average return. Then for each portfolio, equallyweighted monthly returns are calculated and regressed against a number of passive indices, the HML, SMB, and MOM factors, and a number of hedge fund indices (style factors) using Equation (7), above. Results from this regression are in Table 15. The first column shows the monthly excess-of-risk-free rate returns and standard deviations for the formation period portfolios. The second column shows the monthly excess-of-risk-free rate returns and standard deviations for the lagged decile portfolios. For the lagged decile portfolios, average monthly returns are fairly monotonic, increasing from -0.66% for decile 1 (worst) to 0.49% for decile 10 (best). Standard deviation is highest among the best and worst deciles and lowest in the middle deciles. Examining the alphas (intercepts) from the regressions, the intercept on the best minus worst (10-1) portfolio is positive, but not statistically significant. In addition to the alphas, the coefficients from the most statistically significant independent variables are shown. This analysis provides evidence that once common risk and style factors are considered, there is no evidence of quarterly persistence. This result is in direct contrast with the results of other hedge fund studies, which find some persistence at the quarterly level. There are at least two reasons why the 30 In unreported results, we include the option-based returns developed by Agarwal and Naik (2003) as additional independent variables. When these returns are included separately (without including the hedge fund style index returns), the portfolios load significantly on these factors, and the results are consistent with Agarwal and Naik (2003) in that the returns of hedge funds are similar to short positions in put options. However, when the option returns are included in addition to the hedge fund style index returns, they never receive significant loadings. This result occurs because the option returns are fairly highly correlated with a number of the hedge fund styles. Regardless of whether the option returns are included, the intercepts from the regressions (in which we are most interested) are quite similar. 47

59 results of this paper contrast with theirs. First, most previous studies examine individual fund performance and define persistence as a fund s being in the top half of returns for two consecutive periods. This paper sets a more difficult standard for persistence, requiring that funds be in the top 10% of performers (rather than the top 50%) for two consecutive periods. Second, while the other studies control for common risk factors, they do not control for style in the same way as in this paper. To control for style effects, other studies compare fund returns in excess of style average, but do not adjust these net returns for exposures to other style indices. To examine the incremental effect that controlling for style indices has on the ability to find persistence, I re-perform the above analysis controlling only for common risk factors (and not style indices). With this approach (in unreported results), I do find evidence of performance persistence at the quarterly (but not longer) time horizon. Thus, this paper provides evidence that style factors account for much of the persistence found in prior studies Manager tenure as a predictor of persistence While the results of the previous section indicate that prior research findings of quarterly performance persistence can be largely explained by omitted style factors, there is another factor systematically linked to performance that has been ignored in tests of performance persistence. This factor is manager tenure the length of time that a manager has been overseeing his fund. Boyson (2003) shows that, controlling for common risk and style factors, manager tenure is related to performance. Specifically, more experienced (e.g., older or higher tenure) managers underperform less experienced (e.g., younger or lower tenure) managers by approximately -0.75% for each year of tenure. This difference is both economically and statistically significant. If young managers are more skilled than old, this result should be of use in designing a more powerful test of performance persistence. The idea is that young managers with good returns likely achieved those returns due to skill, while old managers with good returns have a higher probability of having achieved those returns due to good luck. The converse also should hold: young managers with poor returns likely experienced bad luck, 48

60 while old managers with poor returns are likely to have experienced those returns due to lack of skill. Thus, a persistence test that selects funds based both on past performance and on manager tenure should be able to detect persistence. Thus, this result is used to design the following test of performance persistence. Funds are sorted into thirds based on two factors: first, they are sorted into thirds based on prior period returns, and then these portfolios are again sorted into thirds based on manager tenure. These sorts result in nine portfolios, ranging from old, past poor performers (portfolio 1, the worst portfolio) to young, past good performers (portfolio 9, the best portfolio). As before, a portfolio that is long the best portfolio (young, past good) and short the worst portfolio (old, past bad), is formed (this is referred to as the 9-1 portfolio). Again as before, these portfolios are held for the three months (six months, one year), and equal-weighted portfolio returns are calculated for each of the three months (six months, one year). Every three months, portfolios are re-formed. This yields a time-series of 81 (78, 72) monthly returns for each portfolio (the first three (six, twelve) months are used in the initial formation period). The quarterly persistence results are in Table 16. The intercept from the 9-1 portfolio is positive and significant at the 5% level (t-value = 2.05). The annualized excess return from investing in this portfolio is about 9%/year, which is economically significant as well. To conserve space, this table shows coefficients and related t-statistics only for the dependent variables which are statistically significant in at least one of the portfolio regressions. While investing in the 9-1 portfolio results in significant quarterly persistence, there is no persistence at the semi-annual and annual levels (which is consistent with prior research). An examination of the coefficients on the explanatory variables indicates some interesting patterns in the data. First, the worst portfolios load positively and significantly on the value-weighted CRSP index, while the best portfolios do not have significant exposure to this factor. This could be interpreted as herding behavior by the worst (and oldest) managers, which is consistent with the findings of Boyson (2003). Additionally, the best portfolios have positive exposure to the currency and commodity indices, while the worst portfolios have negative exposure to these indices. Also, the worst portfolios have negative exposure to bond indices, while the best have positive exposure. Finally, style 49

61 plays an important role in explaining the return differences; managed futures appear to have been out of favor during the time frame, while the long-short equity style was very successful during this time. Thus, it appears that the best managers were successful in both short and long positions in the equity market (as evidenced by their negligible exposure to the VW CRSP index) while the worst managers had a more pronounced long exposure (as evidenced by their significant exposure to the VW CRSP index and insignificant exposure to the long/short equity style index). While there is evidence of quarterly performance persistence, it appears that poor performance among old, past bad managers is driving this result. The net annualized return of 9% for the 9-1 portfolio attributes about -5.5% to the poor performance of the old, past bad managers (which is statistically significant at the 10% level) and about +3.5% to the good performance of the young, past good managers (which is not statistically significant at conventional levels). Thus, there appears to be an asymmetry in persistence: old, past bad managers continue to perform quite badly, while young, past good managers continue to perform fairly well, although at levels that are not statistically significant from zero. Due to the lack of statistical significance, it is probably most accurate to say that while young, past good managers may not continue their past good performance, the are at least able to avoid future poor performance. The next section investigates the likely cause of this asymmetry in more detail Why do old, past bad returns persist? The persistence test in Section 3.5 indicates persistence at the quarterly level when funds are selected for investment based on both manager tenure and past performance. This persistence is concentrated among old, past poor performers. This section investigates the likely cause of this continued poor performance. We consider the following hypothesis: young managers are fired more often than old for poor performance. If this is true, then old managers with past poor performance are less likely than young to fail, and thus are more likely to show (poor) performance persistence. This idea comes from models that relate termination to a learning process where investors learn about a manager s ability over time. (See, for example, Jovanovic (1979), Zwiebel (1995), and 50

62 Holmstrom (1999)). Early in a manager s career, when his reputation is not well-established, investors put more weight on his most recent performance (in the case of a hedge fund manager, his most recently reported monthly return). Eventually, as his reputation becomes more established, each subsequent monthly return has less and less impact on his assessed reputation. The implication of this process is that the sensitivity of termination to the most recent performance evaluation should decrease over time as managers gain reputation. Another reason is noted by Chevalier and Ellison (1999b): since more experienced managers have survived a selection process, the market s assessment of their abilities may be further away from the threshold level at which it becomes efficient to fire the manager. Thus, we examine whether the termination process is more performance sensitive for young managers. In studying this relationship, we repeat and augment the analysis of Boyson (2003), who performs an unconditional survival test and shows that age and manager ability are both positively related to the likelihood of a manager s survival. Here, we extend her work by performing a conditional survival analysis. In this analysis, we follow her approach, which uses a time-varying proportional hazards model to study the relationship between manager termination and a number of factors. Intuitively, this model examines each hedge fund that fails (one per time period) and compares its explanatory variables to the explanatory variables on the set of hedge funds that could have failed during the period but did not. If the values of the explanatory variables for those that failed differ from the values of the explanatory variables for those that survived, the coefficients will be significantly different from zero. Time-varying proportional hazards models (which are a category of the more general hazard functions) have several advantages over the more commonly-used probit and logit models. First, they put fewer distributional assumptions on the data; second, they calculate the conditional rather than the absolute probability of failure (conditional upon not having failed in a prior period); and finally, they do not 51

63 introduce sample-selection bias into the data. Instead of using annual failure rates, the more flexible proportional-hazards model allows for more frequent failure times which reduces bias and adds precision to the estimates. 31 Table 17 performs a number of specifications of the model. Panel A reports results from unconditional regression specifications that include as dependent variables a number of lagged quarterly returns and the manager tenure variable. For ease of interpretation, the coefficient on the manager tenure variable is annualized. In the table, a negative coefficient implies a positive probability of survival, while a positive coefficient implies a positive probability of failure. The results are consistent with Boyson (2003) who uses a smaller, though similar, sample: both current and past returns, as well as manager tenure, are strongly negatively related to failure. All else equal, good funds and old managers are more likely to survive than bad funds and young managers. Panel B begins the first conditional analysis; in this case, conditional upon past performance. For certain of the portfolios described in Section 3.5, dummy variables are created as follows. The first regression (Column 1) models the probability of survival by tenure, given that the fund s past performance was in the bottom third of returns. This category corresponds to portfolios 1,2, and 3 from Table 16. If a fund is in portfolio 1 (high tenure managers with poor past performance) it is assigned a value of one (1). If a fund is in portfolios 2 or 3 (middle tenure managers with poor past performance, and short tenure managers with poor past performance, respectively), it is assigned a value of zero (0). Since this analysis is focusing on past poor performers, portfolios 4-9 are excluded. Hence, the proportional hazards model is comparing the probability of failure for managers in portfolio 1 (old and bad) to the probability of failure for managers in portfolios 2 and 3 (middle tenure and bad, and young tenure and bad, respectively). The negative and statistically significant coefficient on the dummy variable indicates that old and bad funds have a 26.5% lower probability of failure due to poor performance than do young and middle tenure bad funds. 31 For details on the model and a thorough description of the estimation process, see Boyson (2003). For technical details, see Cox (1972) and Kalbfleisch and Prentice (1980). Finally, for related finance/economics literature that uses this methodology, see Helwege (1996), Lunde, Timmerman, and Blake (1999), and Brown, Goetzmann, and Park (2001). 52

64 The next column repeats the analysis of the first column, this time modeling the probability of failure for managers in portfolio 4 from Table 16 (middle third of performance with high tenure) against the probability of failure for managers in portfolios 5 and 6 (middle third of performance with middle tenure, and middle third of performance with low tenure, respectively). This time, portfolios 1-3 and 7-9 are excluded from the analysis. The results are similar to those in column 1: the negative and statistically significant coefficient on this variable indicates that old and middle-third performing funds have a 36% lower probability of failure than do young and middle tenure, middle-third performing funds. The final analysis is performed in column 3, which models the probability of failure for managers in portfolio 7 from Table 16 (top third of performance with high tenure) against the probability of failure for managers in portfolios 8 and 9 (top third of performance with middle tenure, and top third of performance with short tenure, respectively). In this case, the coefficient is positive but not statistically significant. This indicates that for good managers, failure probabilities do not vary systematically by manager tenure. To summarize the results of Panel B, termination is much more performance-sensitive for young than for old managers. Young managers must perform in the top third of all managers to significantly reduce their likelihood of being terminated. Next, Panel C of Table 17 estimates another set of survival functions, this time conditional upon manager tenure. From Panel B, it is clear that the termination relationship is highly performancesensitive for young managers. Panel C provides additional evidence that supports this result. Specifically, column 1 examines the relationship between termination and performance, conditional upon being a young manager. In this case, the managers in column 9 (the young, past good performers) are assigned a dummy variable value of one (1), while the managers in columns 3 and 6 (the young, past poor performers and the young, past middle third performers) are assigned a dummy variable value of zero (0). This analysis models the probability of fund failure for young managers in the top third of performance against the probability of failure for young managers in the bottom two-thirds of performance. The negative and significant coefficient indicates that young, past good performers are about 45% more likely to survive than young managers in the bottom two-thirds of performance. 53

65 The last analysis in Panel C examines only the old managers (columns 1, 4, and 7 of Table 16) for differences in termination probability that are related to performance. In this case, managers in column 7 (old tenure and good past performance) are assigned a dummy variable of one (1), while managers in columns 1 and 4 (old tenure and bottom two-thirds of performance) are assigned a dummy variable of zero (0)). The statistically insignificant coefficient on this variable indicates that there are not differences in termination probabilities related to performance for older managers. Thus, the results of Panel C support the results of Panel B above: young managers can increase their probability of survival significantly by being in the top third of performers, while among old managers performance is unrelated to the likelihood of survival. As noted above, these results provide evidence that fund survival is much more performancesensitive for young than for old managers. This asymmetry in the termination-performance relationship drives the result from the previous section that performance persistence is concentrated among old, past poor performers. Old managers survive more often, regardless of performance. Thus, there is a greater likelihood of seeing persistence among old, past poor performers (since they are unlikely to drop out of the sample) than among young, past good performers (since these managers have to continue to have very strong performance in order to survive). Section 3.7. Relationship of termination and performance persistence to managerial career concerns This section relates the results of Sections 3.5 and 3.6 to previous hedge fund and mutual fund literature regarding the effect of a manager s reputational concerns (or his career concerns ) on his behavior and ultimately, on his performance. The career concerns literature discusses the idea that a manager s concern for keeping his current job or obtaining a better job can mitigate potential agency problems which occur as a result of misaligned incentives between managers and investors. See Fama (1980). It is reasonable to think that these concerns will change over a manager s career and affect his behavior, specifically his propensity to take on risk. 54

66 For example, Chevalier and Ellison (1999b) study the behavior of mutual fund managers, and show that these managers increase risk-taking behavior as their careers progress (they tend to increase idiosyncratic risk and mimic other managers less (or herd less)). Their explanation is that there are implicit incentives in the mutual fund industry which relate to a manager s likelihood of losing his job that cause young managers to be more risk-averse than old. As evidence, they model the termination/performance relationship and show that for excess performance below zero, termination is more likely for young than for old managers. However, for excess performance above zero, termination rates do not differ among young and old managers. Additionally, the termination/performance relationship is fairly flat at all levels of return for old managers: termination is much less performance dependent for old managers. Thus, the implicit incentives are clear: a young manager that wishes to avoid termination will avoid risks that could lead to negative excess performance. That is, he will avoid unsystematic risk and herd with other managers. And, since termination is not dependent on performance for old managers, these managers will take on more risk (they will herd less) in order to increase the possibility that their returns will end up in positive territory. By contrast, Boyson (2003) argues the opposing case for hedge fund managers. She shows that hedge fund managers reduce volatility risk and herd more as they age. She attributes this behavior to career concerns that increase over time in the hedge fund industry (by contrast, Chevalier and Ellison (1999b) argue that career concerns decrease over time in the mutual fund industry). Specifically, she argues that hedge fund managers have more to lose in terms of reputation, personal capital, and current income than do hedge fund managers, should their funds fail. One piece of evidence that supports this argument is the empirical regularity cited in Brown, Goetzmann, and Park (2001) that a failed hedge fund manager is unlikely to start a successful hedge fund in the future. By contrast, mutual fund managers that fail often are hired by other fund companies. Other support for her argument is that hedge fund managers have tremendously high salaries (often 10 or more times that of a comparable mutual fund manager), and have their own capital invested in their funds. These factors all imply that 55

67 failure would be much less desirable for old than for young managers: their funds tend to be larger, so their incomes and personal capital invested are higher, and finding another job would result in a much larger pay cut than for a younger manager with a smaller fund. Next, Boyson (2003) shows that, all else equal, risk-taking behavior leads to termination. Both the existence of high career concerns and the higher probability of termination for risk-taking behavior motivate old managers to reduce their risk-taking behavior, which they do. Finally, she links this reduction in risk-taking behavior to lower returns for old hedge fund managers: since old managers take on less volatility risk and herd more, this behavior results in lower returns for these managers. Thus, the different results of Boyson (2003) and Chevalier and Ellison (1999b) can be reconciled by differences in career concerns in the hedge fund and mutual fund industries. These results are also consistent with the finding of Section 3.5 of this essay that old, past poor performers have persistently worse returns than young, past good performers. However, there is one finding of this current essay that is on the surface, puzzling. In Section 3.6 we show that young managers are more likely to be terminated for poor performance than are old. At first glance, this result looks very similar to that of Chevalier and Ellison (1999b): they also find that termination is more performance-sensitive for young than for old managers. But a closer look at these results shows a very important difference: while for young mutual fund managers, termination is only more performance-sensitive for excess returns below zero, for young hedge fund managers, termination is more performance-sensitive at a larger range of returns: specifically, if a young manager s returns are not in the top third of all returns, he is significantly more likely to fail than an old manager. Thus, the bar is set much higher in the hedge fund industry: young managers must strive to be in the top third of returns in order to continue in the industry. This finding suggests the following story, which is completely consistent with Boyson (2003): for young hedge fund managers who wish to continue in the industry, they must achieve returns higher than one-third of all managers. To maximize their probability of doing this, they must take on more volatility risk and herd less than other managers. Boyson (2003) shows that this type of risk-taking behavior is indeed associated with higher returns. Old hedge fund managers who wish to survive in the 56

68 industry face a much lower hurdle. Since survival rates for old managers do not vary much with performance (see Section 3.6), they have strong incentives to herd and reduce risk so as to avoid the possibility of a very bad return that would increase their probability of failure (and would likely result in a loss of personal assets as well). Thus, young managers strive for high returns by taking on more risk, while old managers strive for average (or even below average) returns by taking on less risk Conclusions This paper designs a more powerful test for performance persistence that is able to detect quarterly persistence among hedge fund managers. This test is motivated by the results of Boyson (2003), who shows that young managers outperform old, on average, which implies that for young managers, good results are driven by skill, while for old managers, good results are driven by luck. If young managers are more skilled, they should show persistence. And this is indeed true: at the quarterly time horizon, young, past good managers outperform old, past poor managers by about 9%/year. This result is driven primarily by the propensity of old, past poor managers to continue underperforming. Additionally, we perform a survival analysis to investigate the result that old, past poor managers have persistence underperformance, and find the following asymmetric relationship. Young managers must perform in the top third of all managers to have survival probabilities that are the same as those of old managers. That is, young managers are punished (by fund failure) significantly more often than old if their returns are in the bottom two-thirds of managers. Thus, a larger number of old, past poor performers survive from one period to the next, which leads to this persistence. Finally, these results are linked to the career concerns study of Boyson (2003). Boyson finds that old managers take on less risk than young. She argues that this results from greater career concerns of older managers: since older managers have more to lose in terms of reputation, personal capital, and current income than young, they reduce volatility risk and herd more to increase their probability of survival (or more precisely, to decrease their probability of an extremely poor performance which would decrease their probability of survival). By contrast, young managers take on more risk and herd 57

69 less than old, which increases their returns. The evidence from this essay provides additional support: the higher bar that younger managers face in terms of survival (they must achieve returns in the top third of all managers to have the same survival probabilities of old managers) provides an additional incentive for young managers to take more risk than old. And, the relatively flat relationship between termination and performance for old managers, coupled with their desire to protect their current job and personal assets invested in their funds, leads old managers to take on less volatility risk and herd more than their younger counterparts. These results have broad implications for hedge fund investors. Specifically, the result of quarterly persistence among young, past good performers over old, past poor performers might provide a way for investors to achieve higher returns. However, for many hedge fund investors, it is not possible to trade hedge funds as often as quarterly, due to lockup periods (most hedge fund managers do not allow very frequent entrance and redemption of assets). There is one class of investors, however, that could theoretically benefit from this finding: fund of funds (FOF) investors. A fund of funds is a hedge fund that invests in other hedge funds, and hedge fund managers will often waive lockup periods for FOF investors. However, there is another problem with implementing this investment strategy. Specifically, the persistence is concentrated among the worst, rather than the best, performers. Currently, there is no way to short a hedge fund. While investing in a long-only portfolio of the best performers (young, past good managers) is possible, the excess return from this strategy would have been 3.5%/year, which is not statistically significant, and arguably, not very economically significant either. The very high failure rates of young hedge funds lead to this result: even good young hedge funds fail at higher rates that old funds, so selecting a portfolio of young, past good performers is no guarantee of performance persistence since many of these funds are likely to fail in the next period, due simply to their age. These results are broadly similar to many studies of mutual funds which find persistence among the worst performers. Evidence for why this is true can be found in Agarwal, Daniel, and Naik (2003) who show that, similar to mutual funds, good hedge funds attract significant inflows while bad 58

70 past performers do not experience as significant outflows. Thus, the bad performers persist while the good performers may not. A more detailed look into this relationship is an interesting area for future research. Finally, an additional area for future research is to model the termination/performance relationship in much more detail. We show that there is a large asymmetry: for young managers, their returns must be in the top third of all managers to have the same survival probabilities as for old funds. However, this conclusion is based on a simple sort of managers into thirds based on age and performance. A closer look at the shape of the relationship can provide better evidence regarding incentives and the subsequent risk-taking behavior of hedge fund managers. 59

71 CHAPTER 4 CONCLUSIONS This dissertation essay contributes to two areas of research in finance. The first is the career concerns literature, theoretically modeled by Fama (1980), Holmstrom and i Costa (1986), Scharfstein and Stein (1990), Zwiebel (1995), Prendergast and Stole (1996), and Avery and Chevalier (1999), and tested empirically by Graham (1999), Chevalier and Ellison (1999b), Hong, Kubik, and Solomon (2000), and Lamont (2002). The second area of research is the performance persistence literature, which has been studied extensively in the mutual fund literature (see e.g., Jensen (1968), Grinblatt and Titman (1992), Elton, Gruber, Das and Hlavka (1993), Hendricks, Patel, and Zeckhauser (1993), Goetzmann and Ibbotson (1994), Brown and Goetzmann (1995), Elton, Gruber, Das and Blake (1996), Carhart (1997), Christopherson, Ferson, and Glassman (1998), Wermers (1999), Bollen and Busse (2002), and Busse and Irvine (2002)). Persistence has been studied less extensively in the hedge fund literature, by Agarwal and Naik (1999, 2000), Bares, Gibson, and Gyger (2002), and Baquero, ter Horst, and Verbeek (2002). The first essay shows that more experienced (old or high tenure) hedge fund managers have lower returns than less experienced (young or low tenure) managers, and explains this difference in performance with differences in risk-taking behavior. Specifically, young managers engage in more volatile trading strategies and herd less than old managers. This difference in risk taking behavior is explained with differing career concerns: since old managers have more to lose (in terms of reputation, current income, and personal capital) than young managers should their funds fail, they avoid risky behavior that might be positively related to fund failure. This avoidance of risky behavior increases their probability of survival, but decreases their returns relative to younger managers. 60

72 There are at least two implications of this result. First, for investors in hedge funds, it is important to consider a manager s tenure before investing. A manager that is too young has a much higher probability of failure but also has a higher probability of achieving better returns, while a manager that is too old has a lower probability of failure, but this is accompanied by returns that are significantly below average. Thus, very young or very old managers should be avoided. Second, there are implications for incentive contracts design. Hedge fund contracts typically pay a manager a percentage of profits (usually about 20%) in addition to a percentage of assets (usually about 1-2%). While this type of option-like pay structure might induce managers to take on too much risk (since their incentive compensation would be zero if their fund posts a loss and positive if their fund posts a gain), researchers (see e.g., Carpenter (2000)) has shown that these types of contracts perform well at aligning the interests of managers and investors. It appears that this type of contract works well in aligning incentives for young managers, but not as well for older managers: as managers get older, their desire to avoid failure outweighs their desire to earn large incentive fees causing them to incur too little risk. A recent paper by Dybvig, Farnsworth, and Carpenter (2001) suggests a contract that would encourage managers to avoid herding behavior. This sort of contract is a promising option to better align the incentives of old managers with those of their investors. The second dissertation essay studies performance persistence of hedge funds. Prior studies of performance persistence among hedge funds have examined persistence at the quarterly, semiannual, and annual levels, with the result that hedge funds demonstrate short-term (one to three-month) persistence when funds are selected based on prior period performance. However, these studies do not rigorously account for style effects in determining performance. Brown and Goetzmann (2001) show that in certain periods, certain hedge fund styles cluster in terms of performance, and thus, controlling for style in a study of persistence is important. When we control for both common risk factors and style factors we show that style factors can explain the previous findings of short term persistence. Thus, when funds are selected based on past performance alone and adjusted for common risk and style factors, there is no evidence of performance persistence at any time horizon examined. 61

73 However, selecting funds based on past performance alone ignores the results of the first essay. The first essay shows that young hedge funds outperform old hedge funds, even when adjusting for common risk and style factors. Thus, we use this result to design a more powerful test of persistence that takes into account both prior performance and manager tenure. Specifically, we find that a portfolio that is long young, good past performers and short old, poor past performers shows evidence of persistence of about 9% (annualized) at the quarterly level. This result is mostly concentrated among old, past poor performers. An examination into why the persistence is found mostly among old, past poor performers identifies an asymmetry in the termination/performance relationship that can explain this result. A conditional survival analysis indicates that termination is more performance-sensitive for young managers than for old. The bar is set quite high for young managers: their returns must be in the top 1/3 of all returns during a period to avoid having a significantly higher probability of failure than an old manager with the same level of returns. The relationship of performance to termination for old managers is much flatter: conditional upon being an old manager, performance is not significantly related to termination. Thus, many more old managers survive in a given period than do young, explaining the persistence among old, poor past performers. This result that termination is more performance-sensitive for young than for old managers is also consistent with the results of the first dissertation essay, which shows that young managers take on more risk (they herd less and engage in more volatile trading strategies) than do old. Since young managers need to achieve returns in at least the top third of all returns in order to have the same survival probability as old managers, it is not optimal for them to engage in herding behavior or low volatility strategies, since on average, these strategies will result in an average (mean) level of performance. And, at this level of performance, their probability of termination is significantly higher than it is for old managers. Thus, young managers will optimally engage in riskier strategies to increase the probability that their funds will achieve top one-third performance. By contrast, older managers, who enjoy a much less performance-sensitive termination probability, can herd and take on less risk consistent with their 62

74 greater career concerns described in the first essay without fear of termination related to average or below average performance. These results are completely consistent with and provide further support for the first essay s finding: young managers take on more risk while old managers take on less. 63

75 LIST OF REFERENCES Ackerman, Carl, Richard McEnally, and David Ravenscraft, 1999, The performance of hedge funds: Risk, return, and incentives, The Journal of Finance 54(3), Agarwal, Vikas and Narayan Y. Naik, 1999, On taking the alternative route: Risks, rewards, style, and performance persistence of hedge funds," The Journal of Alternative Investments 2(4), Agarwal, Vikas and Narayan Y. Naik, 2003, Risks and portfolio decisions involving hedge funds, forthcoming, The Review of Financial Studies. Asness, Clifford, Krail, Robert, and John Liew, 2001, Do hedge funds hedge?, The Journal of Portfolio Management, Fall 2001, Avery, Christopher N. and Judith A. Chevalier, 1999, Herding over the career, Economics Letters 53, Baquero, Guillermo, ter Horst, Jenke, and Mario Verbeek, 2002, Survival, look-ahead bias, and the persistence in hedge fund performance, Working paper, ERIM Report Series Research in Management, Bares, P.A., R. Gibson, and S. Gyger, 2002, Performance in the hedge funds industry: An analysis of short and long-term persistence, Working paper, Institute of Theoretical Physics Swiss Federal Institute of Technology Lausanne EPFL, Switzerland. Bikhchandani, Sushil, Hirshleifer, David, and Ivo Welch, 1992, A theory of fads, fashion, custom, and cultural change as information cascades, The Journal of Political Economy 100(5), Breslow, N., 1974, `Covariance analysis of censored survival data, Biometrics 20, Brown, Keith C., and V.W. Harlow, 2002, Staying the course: The impact of investment style consistency on mutual fund performance, Working paper, University of Texas. Brown, Keith, Harlow W.V., and Laura T. Starks, 1996, `Of tournaments and temptations: An analysis of managerial incentives in the mutual fund industry, The Journal of Finance 51(1), Brown, Stephen and William Goetzmann, 1995, Performance persistence, The Journal of Finance 50(2), Brown, Stephen and William Goetzmann, 2001, Hedge funds with style, NBER Working Paper 8173, Brown, Stephen, William Goetzmann and Roger Ibbotson, 1999, Offshore hedge funds: Survival and performance , The Journal of Business 72,

76 Brown, Stephen, William Goetzmann, and James Park, 2001, Conditions for survival: Changing risk and the performance of hedge fund managers and CTA's, The Journal of Finance 56(5), Brown, Stephen, William Goetzmann, Roger G. Ibbotson, and Stephen A. Ross, 1992, Survivorship bias in performance studies, The Review of Financial Studies 5(4), Carhart, Mark M., 1997, On persistence in mutual fund performance, The Journal of Finance 52(1), Carhart, Mark M., Jennifer N. Carpenter, Anthony W. Lynch, and David K. Musto., 2002, Mutual fund survivorship, The Review of Financial Studies 15(5), Carpenter, Jennifer, ``Does option compensation increase managerial risk appetite?, 2000, The Journal of Finance 55, Chevalier, Judith and Glenn Ellison, 1997, Risk-taking by mutual funds as a response to incentives, The Journal of Political Economy 105(6), Chevalier, Judith and Glenn Ellison, 1999a, Are some mutual fund managers better than others?, The Journal of Finance 54(3), Chevalier, Judith and Glenn Ellison, 1999b, Career concerns of mutual fund managers, Quarterly Journal of Economics 114(2), Christopherson, Jon A., Ferson, Wayne E., and Debra A. Glassman, 1998, Conditional measures of performance and persistence for pension funds, Research in Finance 16, Cox, D, 1972, Regression models and life tables (with discussion.), Journal of the Royal Statistical Society 34, Das, Sanjiv and Rangarajan Sundaram, 2003, Fee speech: Signaling, risk-sharing, and the impact of fee structure on investor welfare, forthcoming The Review of Financial Studies. Diamond, Douglas, 1991, Reputation acquisition in debt markets, The Journal of Political Economy 97, Edwards, Franklin R. and James M. Park, 1996, Do managed futures make good investments?, Journal of Futures Markets 16(5), Edwards, Franklin R. and Mustafa Caglayan, 2001, Hedge fund performance and manager skill, Journal of Futures Markets 21(11), Elton, Edwin, Gruber, Martin J., and Christopher R. Blake, 1996, Survivorship bias and mutual fund performance, The Review of Financial Studies 9(4), Elton, Edwin, Gruber, Martin J., and Christopher R. Blake, 1996, The persistence of risk-adjusted mutual fund performance, The Journal of Business 69(2), Fama, Eugene, 1980, Agency problems and the theory of the firm, The Journal of Political Economy 88(2), Fama, Eugene F. and James D. MacBeth, 1973, Risk, return, and equilibrium: Empirical tests, The Journal of Political Economy 81,

77 Fama, Eugene and Kenneth R. French, 1992, The cross-section of expected stock returns, The Journal of Finance 47, Fama, Eugene and Kenneth R. French, 1993, Common risk factors in the returns on bonds and stocks," Journal of Financial Economics 33, Finkelstein, Sydney and Donald Hambrick, 1990, Top-management team tenure and organizational outcomes: The moderating role of managerial Discretion, Administrative Sciences Quarterly 35(3), Fung, William and David A. Hsieh, 1997, Empirical characteristics of dynamic trading strategies: The case of hedge funds, The Review of Financial Studies 10(2), Fung, William and David A. Hsieh, 1999, A primer on hedge funds, Journal of Empirical Finance 6, Fung, William and David A. Hsieh, 2000, Performance characteristics of hedge funds and commodity funds: Natural versus spurious biases, Journal of Financial and Quantitative Analysis 35, Fung, William and David A. Hsieh, 2001, The risk in hedge fund strategies: Theory and evidence from trend followers, The Review of Financial Studies 14(1), Fung, William and David A. Hsieh, 2003, Asset-based hedge fund styles and portfolio diversification, forthcoming Financial Analysts Journal. Gibbons, Robert and Kevin J. Murphy, 1992, Optimal incentive contracts in the presence of career concerns: Theory and evidence, The Journal of Political Economy 100(3), Glosten, Lawrence and Ravi Jagannathan, 1994, A contingent claim approach to performance evaluation, Journal of Empirical Finance 1, Goetzmann, William N. and Roger G. Ibbotson, 1994, Do winners repeat?, The Journal of Portfolio Management, Winter 1994, Goetzmann, William N. and Nadev Peles, 1997, Cognitive dissonance and mutual fund investors, The Journal of Financial Research 20(2), Graham, John R., 1999, Herding among investment newsletters: Theory and evidence, The Journal of Finance 54(1), Grinblatt, Mark and Sheridan Titman, 1989, Adverse risk incentives and the Design of performancebased contracts, Management Science 35(7), Grinblatt, Mark and Sheridan Titman, 1989, Portfolio performance evaluation: Old issues and new insights, The Review of Financial Studies 2(3), Grinblatt, Mark and Sheridan Titman, 1992, The persistence of mutual fund performance, The Journal of Finance 47(5), Grinblatt, Mark and Sheridan Titman, 1994, A study of monthly mutual fund returns and performance evaluation techniques, Journal of Financial and Quantitative Analysis 29(3),

78 Grinblatt, Mark, Titman, Sheridan, and Russ Wermers, 1995, Momentum investment strategies, portfolio performance, and herding: A study of mutual fund behavior, The American Economic Review 85, Gruber, Martin J., 1996, Another puzzle: The growth in actively managed mutual funds, The Journal of Finance 51(30), Hambrick, Donald and Phyllis Mason, 1984, Upper echelons: The organization as a reflection of its top managers, The Academy of Management Review 9(2), Helwege, Jean, 1996, Determinants of savings and loan failures: Estimates of a time-varying proportional hazard function, Journal of Financial Services Research 10, Hendricks, Darryll, Patel, Jayendu, and Richard Zeckhauser, 1993, Hot hands in mutual funds: Shortrun persistence of relative performance , The Journal of Finance 48(1), Hendricks, Darryll, Patel, Jayendu, and Richard Zeckhauser, 1997, The j-shape of performance persistence given survivorship bias, The Review of Economics and Statistics 74(2), Hirshleifer, David and Anjan V. Thakor, 1992, Managerial conservatism, project choice, and debt, The Review of Financial Studies 5(3), Holmstrom, Bengt, 1982/1999, Managerial incentive problems -- A dynamic perspective, NBER Working Paper #6875. Holmstrom, Bengt and Paul R. Milgrom, 1987, Aggregation and linearity in the provision of intertemporal incentives, Econometrica 55, Holmstrom, Bengt and Joan Ricart I Costa, 1986, Managerial incentives and Capital management, Quarterly Journal of Economics 101(4), Hong, Harrison, Kubik, Jeffrey D., and Amit Solomon, 2000, Security analysts' career concerns and herding of earnings forecasts, RAND Journal of Economics 31(1), Huddart, Steven, 1999, Reputation and performance fee effects on portfolio choice by investment advisers, Journal of Financial Markets 2, Ineichen, Alexander, 2000, In search of alpha, UBS Warburg, Ippolito, R., 1989, Efficiency with costly information: A study of mutual fund performance , Quarterly Journal of Economics 104, Jaggia, S. and Satish Thosar, 2000, Risk aversion and the investment horizon: A new perspective on the time diversification debate, Journal of Psychology and Financial Markets 1(3), Jegadeesh, Narisimhan and Sheridan Titman, 1993, Returns to buying winners and selling losers: Implications for stock market efficiency, The Journal of Finance 48, Jensen, Michael C. and Kevin J. Murphy, 1990, Performance pay and top-management incentives, The Journal of Political Economy 98(2), Jovanovic, Boyan, 1979, Job matching and the theory of turnover, Journal of Political Economy 87,

79 Kahneman, D., Slovic, P., and Tversky, A., Judgment under Uncertainty: Heuristics and Biases. Cambridge, NY: Cambridge University Press. Kalbfleisch, J. and R. Prentice, The Statistical Analysis of Failure Time Data. New York: John Wiley & Sons. Kanodia, Chandra, Bushman, Robert, and John Dickhaut, 1989, Escalation errors and the sunk cost effect: An explanation based on reputation and information asymmetries, Journal of Accounting Research 27, Katz, Ralph, 1982, The effects of group longevity on project communication and performance, Administrative Science Quarterly 27, Kiesler, C., The Psychology of Commitment. New-York: Academic Press. Lakonishok, J., A. Shliefer, R. Thaler, and R. Vishny 1991, Window dressing by pension fund managers, America Economic Review, Papers and Proceedings 81, Lamont, Owen, 2002, Macroeconomic forecasts and microeconomic forecasters, Journal of Economic Behavior and Organization 48, Lehmann, Bruce N. and David M. Modest, 1987, Mutual fund performance evaluation: A comparison of benchmarks and benchmark comparisons, The Journal of Finance 42(2), L'habitant, François-Serge, 2001, Assessing market risk for hedge funds and hedge fund portfolios, Working paper, HEC University of Lausanne. Li, Xi, 2002, Performance, herding, and career concerns of individual financial analysts, Working paper, Vanderbilt University. Liang, Bing, 1999, On the performance of hedge funds, Financial Analysts Journal 55(4), Liang, Bing, 2000, Hedge funds: The living and the dead, Journal of Financial and Quantitative Analysis 35(3), Lo, Andrew, 2001, Risk management for hedge funds: Introduction and overview, Financial Analysts Journal 57(6), Lunde, Asger, Allan Timmermann, and David Blake, 1999, The hazards of underperformance: A Cox regression analysis, Journal of Empirical Finance 6, Malkiel, Burton G., 1995, Returns from investing in equity mutual funds 1971 to 1991, The Journal of Finance 50(2), Mitchell, Mark and Todd Pulvino, 2001, Characteristics of risk and return in risk arbitrage, The Journal of Finance 56(6), Morin, Roger A. and A. Fernandez Suarez, 1983, Risk aversion revisited, The Journal of Finance 38(4), Prendergast, Canice and Lars Stole, 1996, Impetuous youngsters and jaded old-timers: Acquiring a reputation for learning, The Journal of Political Economy 194(6),

80 Salancik G.R., Commitment and the control of organizational behavior and belief, in B. Staw and G.R. Salancik (eds.), New Directions in Organizational Behavior, St. Clair Press, Chicago, 1977, Scharfstein, David S. and Jeremy C. Stein, 1990, Herd behavior and investment, The American Economic Review 80(3), Shumway, Tyler, 2001, Forecasting bankruptcy more accurately: A simple hazard model, Journal of Business 74(1), Sirri, Erik R. and Peter Tufano, 1998, Costly search and mutual fund flows, The Journal of Finance 53(5), Starks, Laura T., 1987, Performance incentive fees: An agency theoretic approach, Journal of Financial and Quantitative Analysis 22, Vroom, Victor and Bernd Pahl, 1971, The relationship between age and risk-taking among managers, Journal of Applied Psychology 35, Weismann, Andrew, 2001, Conservation of volatility and the interpretation of hedge fund performance, Alternative Investment Management Association, June/July Welch, Ivo, 2000, Herding among security analysts, Journal of Financial Economics 58, Wermers, Russ, 1997, Momentum investment strategies of mutual funds, performance persistence, and survivorship bias, Working paper, University of Maryland. Wermers, Russ, 2000, Mutual fund performance: An empirical decomposition into stock-picking talent, style, transaction costs, and expenses, The Journal of Finance 55(4), Zitzewitz, Eric, 2001, Opinion-producing agents: Career concerns and exaggeration, Working paper, Stanford Graduate School of Business. Zwiebel, Jeffrey, 1995, Corporate conservatism and relative compensation. The Journal of Political Economy 103(1),

81 APPENDIX A TABLES Table 1: Summary statistics: Return, manager and fund characteristics Below are summary statistics for the return, manager, and fund variables used in the paper. Each statistic is first calculated by fund, and then across funds. CFA, CPA, MBA, Ph.D., Other Advanced Degree, Law Degree, Listed on Exchange, Onshore, Open to New Investment, Open to Non-Accredited Investors, Uses Leverage, and Personal Capital Invested are (0/1) indicator variables set to one if the value is "yes" and zero if the value is "no". The first column in each category shows statistics for the small sample (for which all manager and fund characteristic variables are available), and the second shows statistics for the large sample (for which only manager tenure and all fund characteristics are available.) HF market-adjusted. return is calculated by estimating the market exposure for each fund for each hedge fund index (see Section 2.2 for details), weighting each of these exposures by the performance of the hedge fund index, and subtracting these weighted exposures from the simple excess return of the fund. Passive market-adjusted return is calculated in the same manner as HF marketadjusted return, but passive indices are used instead of hedge fund indices. See Section 2.2 for detail on the hedge fund and passive indices. 70

82 Returns Mean Median Maximum Minimum n=271 n=982 n=271 n=982 n=271 n=982 n=271 n=982 (Return - risk free rate) (monthly) 1.01% 0.85% 0.75% 0.66% % % % % (Return - style average) (monthly) 0.09% 0.10% -0.02% 0.02% % % % % HF market-adjusted return (monthly) -0.25% -0.25% -0.22% 0.03% 0.92% 0.98% -2.00% -2.38% Pass. market-adjusted (monthly) 0.08% 0.01% 0.05% 0.00% 1.87% 1.87% -1.50% -3.02% Manager education, experience CFA 0.10 n/a - n/a - n/a - n/a CPA 0.02 n/a - n/a - n/a - n/a MBA 0.49 n/a - n/a - n/a - n/a Ph.D n/a - n/a - n/a - n/a 71 Manager education, experience (continued) Other advanced degree 0.08 n/a - n/a - n/a - n/a Law degree 0.10 n/a - n/a - n/a - n/a Average undergraduate institution SAT score 1323 n/a 1361 n/a 1475 n/a 958 n/a Number of years experience 22 n/a 20 n/a 47 n/a 6 n/a Manager age 47 n/a 44 n/a 72 n/a 31 n/a Manager tenure (years) 7.79 n/a Fund features: location, risk Listed on exchange Onshore Table 1 (continued): Summary statistics: Return, manager and fund characteristics

83 Mean Median Maximum Minimum n=271 n=982 n=271 n=982 n=271 n=982 n=271 n=982 Fund features: location, risk (cont.) Open to new investment Open to non-accredited investors Personal capital invested Uses leverage Fund policies and size Lockup redemption period (months) Lockup entrance period (months) Minimum investment (millions) $.91 $.78 $.50 $.50 $25 $50 $.002 $.001 Size as of 12/31/00 (millions) $161 $113 $52 $30 $1,800 $1,893 $5 $5 72 Fund fees Management fee (% of assets) 1.20% 1.30% 1.00% 1.00% 3.00% 6.00% 0.00% 0.00% Incentive fee (% of profits) 18.85% 17.79% 20.00% 20.00% 33.00% 50.00% 0.00% 0.00% n/a = not available Table 1 (continued) : Summary statistics: Return, manager and fund characteristics

84 Percentile Manager Tenure in Years Estimated Manager Age Table 2: Conversion of tenure values to age estimates Manager tenure is divided into percentiles based upon the large sample of 982 funds. Manager age is divided into percentiles based on the smaller sample of 271 funds. The values shown are the means for each percentile. 73

85 CONVERTIBLE ARBITRAGE This strategy is identified by hedge investing in the convertible securities of a company. A typical investment is to be long the convertible bond and short the common stock of the same company. Positions are designed to generate profits from the fixed income security as well as the short sale of stock, while protecting principal from market moves. DEDICATED SHORT BIAS Dedicated short sellers were once a robust category of hedge funds before the long bull market rendered the strategy difficult to implement. A new category, short biased, has emerged. The strategy is to maintain net short as opposed to pure short exposure. Short biased managers take short positions in mostly equities and derivatives. The short bias of a manager s portfolio must be constantly greater than zero to be classified in this category. EMERGING MARKETS This strategy involved equity or fixed income investing in emerging markets around the world. Because many emerging markets do not allow short selling, nor offer viable futures or other derivative products with which to hedge, emerging market investing often employs a long-only strategy. EQUITY MARKET NEUTRAL This investment strategy is designed to exploit equity market inefficiencies and usually involved being simultaneously long and short matched equity portfolios of the same size within a country. Market neutral portfolios are designed to be either beta or currency neutral, or both. Well-designed portfolios typically control for industry, sector, market capitalization, and other exposures. Leverage is often applied to enhance returns. EVENT DRIVEN This strategy is defined as special situations investing designed to capture price movement generated by a significant pending corporate event such as a merger, corporate restructuring, liquidation, bankruptcy or reorganization. There are three popular sub-categories in even-driven strategies: risk (merger) arbitrage, distressed/high yield securities, and Regulation D. FIXED INCOME ARBITRAGE The fixed income arbitrageur aims to profit from price anomalies between related interest rate securities. Most managers trade globally with categories including interest rate swap arbitrage, US and non-us government bond arbitrage, forward yield curve arbitrage, and mortgage-backed securities arbitrage. The mortgage-backed market is primarily US-based, over-the-counter and particularly complex. Table 3: Description of hedge fund indices (Source: The methodology utilized in the CSFB/Tremont Hedge Fund Index starts by defining the universe it is measuring. Credit Suisse First Boston Tremont Index LLC uses the TASS database, which tracks over 2,600 funds. The universe consists only of funds with a minimum of US $10 million under management and a current audited financial statement. Funds are separated into primary sub-categories based on their investment style. The Index in all cases represents at least 85% of the assets under management in the universe. CSFB/Tremont analyzes the percentage of assets invested in each subcategory and selects funds for the Index based on those percentages, matching the shape of the Index to the shape of the universe. The Index is re-balanced monthly. Funds are re-selected on a quarterly basis as necessary. Funds must meet the Credit Suisse First Boston Tremont Index LLC reporting requirements. Funds are not removed from the Index until they are liquidated or fail to meet the financial reporting requirements. The objective is to minimize survivorship bias. 74

86 GLOBAL MACRO Global macro managers carry long and short positions in any of the world s major capital or derivative markets. These positions reflect their views on overall market direction as influenced by major economic trends and/or events. The portfolios of these funds can include stocks, bonds, currencies, and commodities in the form of cash or derivatives instruments. Most funds invest globally in both developed and emerging markets. LONG-SHORT EQUITY This directional strategy involves equity-oriented investing on both the long and short sides of the market. The objective is not to be market neutral. Managers have the ability to shift from value to growth, fro small to medium to large capitalization stocks, and from a net long position to a new short position. Managers may use futures and options to hedge. The focus may be regional, such as long/short US or European equity, or sector specific, such as long and short technology or healthcare stocks. Long/short equity funds tend to build and hold portfolios that are substantially more concentrated than those of traditional stock funds. MANAGED FUTURES This strategy invests in listed financial and commodity futures markets and currency markets around the world. The managers are usually referred to as Commodity Trading Advisors, or CTA s. Trading disciplines are generally systematic or discretionary. Systematic traders tend to use price and market specific information (often technical) to make trading decisions, while discretionary managers use a judgmental approach. Table 3 (continued): Description of hedge fund indices 75

87 Panel A: Passive Indices Mean Median Max Min Std Dev Skew Kurtosis US Dollar Weighted Index -0.19% -0.24% 3.78% -5.18% 1.73% Gold -0.73% -0.94% 20.51% -7.11% 3.69% Commodities 0.42% 0.47% 16.81% % 5.40% S&P % 1.56% 9.23% % 4.16% LB Aggregate Bond -0.43% -0.26% 3.60% -3.66% 1.26% LB 30 Yr. US T-bond 0.15% 0.10% 8.61% -7.78% 2.81% SMB -0.36% -0.46% 14.23% % 4.93% HML -0.28% -5.35% 15.40% % 3.83% Momentum 1.56% 0.81% 18.23% 8.98% 4.43% Panel B: Hedge Fund Indices Mean Median Max Min Std Dev Skew Kurtosis Managed Futures 0.08% -0.25% 9.56% -9.82% 3.32% Convertible Arbitrage 0.45% 0.73% 3.13% -5.06% 1.44% Emerging Markets 0.14% 0.43% 16.14% % 5.93% Distressed Securities -0.33% -0.53% 22.32% -9.08% 5.52% Market Neutral 0.54% 0.57% 2.89% -1.53% 0.98% Event Driven 0.56% 0.67% 3.43% % 1.91% Fixed Income Arbitrage 0.14% 0.40% 1.61% -7.35% 1.25% Global Macro 0.74% 0.83% 10.14% % 4.12% Long/Short Equity 0.91% 0.95% 12.63% % 3.65% Fund of Funds Index 0.24% -0.10% 4.92% -3.73% 1.83% Table 4: Summary statistics: Index returns Below are mean buy and hold monthly returns and other summary statistics for the "market" indices used in the paper. The period is January, 1994 to December, All returns are in excess of the riskfree rate. The sources for the indices are Datastream (passive) and CSFB/Tremont (hedge fund). The returns on HML (a high book value minus low book value stock portfolio), SMB (a small-capitalization minus large-capitalization stock portfolio), and MOM (a momentum portfolio) were obtained from Kenneth French's website. Panel A shows statistics for the passive indices, while panel B shows statistics for the hedge fund indices. For a detailed description of the hedge fund indices, see Table 3. 76

88 Specification Independent variables Intercept (-1.16) Mgr. Education, experience, age MBA (-1.03) Manager age (-2.60) Manager tenure (-5.26) Fund location, risk Listed on exchange (-1.89) Onshore (2.40) Open to new investment (-2.22) Open to nonaccredited investors (-0.29) Personal capital invested (-0.51) Uses leverage (0.19) Fund Policies Log redemption frequency (-0.19) Log entrance frequency Log minimum investment Dependent Variable: Excess Monthly Return Includes no controls for market exposure Includes controls for passive indices Includes controls for hedge fund indices 1a 1b 2a 2b 3a 3b (small (large (small (large (small (large sample) sample) sample) sample) sample) sample) (0.23) (0.46) (-0.38) (-5.44) (-0.57) (2.13) (-3.89) (-4.77) (0.09) (1.66) (0.03) (0.75) (2.96) (-0.59) (-2.33) (-3.60) (-4.23) (-2.47) (3.76) (-0.64) (-0.83) (-1.81) (1.14) (0.76) (0.71) (0.63) (-1.47) (-6.66) (-2.12) (2.44) (-2.44) (-4.49) (-0.37) (1.15) (0.25) (1.14) (4.72) Table 5: Relationship between hedge fund performance and manager tenure (-0.04) (-2.50) (-2.90) (-4.69) (-1.82) (3.31) (-1.54) (-0.36) (-2.46) (1.07) (0.48) (-0.20) (0.19) (-2.10) (-7.35) (-1.11) (2.87) (-3.97) (-4.24) (-0.24) (1.79) (0.56) (1.22) (4.85) Cross-sectional monthly regressions of excess returns against fund and manager characteristics are performed below. The time frame is January, 1994 to December, 2000, for a total of 84 monthly regressions. The coefficients are calculated by averaging across cross-sectional regressions, as in Fama-Macbeth (1973). The standard errors are adjusted for autocorrelation and heteroskedasticity. For ease of interpretation, all coefficients are annualized. The dependent variable is monthly return less the risk-free rate. To account for each fund's correlation with passive or hedge fund indices, a number of "market" control variables are included in specifications 2a-3b, while specifications 1a and 1b do not control for market exposure. Finally, each regression is performed on two samples: the first (regression a in each set) is performed on a smaller sample of 271 funds that includes all manager and fund characteristics, while the second (regression b in each set) is performed on a larger sample of 982 funds that includes only manager tenure and all fund characteristics. t-statistics are in parentheses. Number of funds and adjusted R 2 are shown below.

89 Includes no controls for market exposure Dependent Variable: Excess Monthly Return Includes controls for passive indices Includes controls for hedge fund indices Specification Independent variables 1a (small sample) 1b (large sample) 2a (small sample) 2b (large sample) 3a (small sample) 3b (large sample) Fund fees and style Mgmt. fee (% of assets) (1.51) (0.03) (0.72) (-.015) (1.25) (0.52) Incentive fee (% of profits) (4.07) (1.59) (6.20) (2.78) (4.98) (2.59) U.S. equity style (2.54) (3.02) (3.65) (3.62) (3.23) (3.93) Europe equity style (1.04) (3.14) (1.29) (3.31) (1.16) (3.01) Relative value style (0.00) (0.33) (0.73) (2.53) (1.16) (2.46) Event driven style (0.00) (1.63) (1.87) (2.70) (1.89) (2.89) Adjusted R Table 5 (continued): Relationship between hedge fund performance and manager tenure 78

90 Mean Mean Mean Wilcox p-value Wilcox p-value Wilcox p-value Age Category Young Middle Old Young vs. Middle Young vs. Old Middle vs. Old Standard Deviation 17.62% 17.14% 15.87% Tracking Error Deviation 2.21% 2.55% -6.44% Beta Deviation Mean Mean Mean Wilcox p-value Wilcox p-value Wilcox p-value Size Category Small Medium Large Small vs. Medium Small vs. Large Medium vs. Large Standard Deviation 20.35% 16.87% 13.88% <.0001 <.0001 <.0001 Tracking Error Deviation 20.05% 0.57% % <.0001 <.0001 <.0001 Beta Deviation < Mean Mean Mean Wilcox p-value Wilcox p-value Wilcox p-value Age/Size Category Sm/Young Med/Mid Lg/Old Sm/Young vs. Med/Mid Sm/Young vs. Lg/Old Med/Mid vs. Lg/Old 79 Standard Deviation 20.47% 17.07% 12.52% <.0001 <.0001 Tracking Error Deviation 17.34% 1.99% % <.0001 <.0001 Beta Deviation < Table 6: Summary statistics of risk and return variables by age, size, and age/size interactions Below are summary statistics and Wilcox rank-sum tests for differences in means for a number of risk variables, categorized by age, size, and age/size categories. For each fund, averages of manager tenure and fund size are calculated for the period 1994 and This results in an average value for manager tenure and fund size for each of the 982 funds. Based on these average values, funds are equally sorted into three tenure categories: young, middle, and old. Additionally, funds are equally sorted into three size categories: small, medium, and large. Nine size-tenure categories are then created from the size and tenure sorts. Then, means of three risk variables are calculated for each tenure, size, and tenure/size category. The risk variables include: annual standard deviation, tracking error deviation, and beta deviation. Wilcox rank-sum tests for differences in means among the categories are performed, and results are reported below. Of interest in this table are the funds with the most and least reputation at stake: funds with the most reputation at stake are large and old, while funds with the least reputation at stake are small and young. For comparison, the middle group (medium and middle) is also included

91 US Long/Short Equity: The investment manager takes long and short positions in U.S. equities. European Equity Hedge: Same as above, with the European equities as the major focus. Global/International Equity Hedge: Same as above, with an international focus. Event Driven: The investment manager typically takes long or short positions in equities or debt instruments in anticipation of an even (i.e., corporate restructuring, planned joint venture, etc.) expected cause substantial price movement. Distressed Securities: The investment manager invests in the securities of bankrupt companies in Chapters 11 Status in the U.S. Risk Arbitrage/Deal Arbitrage: Strategy involves the simultaneous purchase of stock in a company being acquired a sale of stock in the acquiring company. Also called takeover arbitrage and merger arbitrage. Special Situations: The investment focus is on takeover situations, as well as distressed or financially troubled securities. The manager looks for events that characteristically happen very rarely in the case of a company/issuer. Relative Value: The manager looks to establish offsetting long and short positions in related primary or derivative markets based on the belief that one instrument or security is undervalued in terms of risk, liquidity and/or return relative to another. Market Neutral: Strategies that, in theory, do not depend on directional movement in markets traded. Investment managers take offsetting long and short positions in related primary and derivative markets with the intention of capturing pricing inequities. While resulting profits can be impacted by market direction, positions should generate positive returns in either up or down markets. Convertible Arbitrage: The investment manager simultaneously establishes long and short positions in different forms of convertible securities from the same corporate issuer, and in so doing, captures pricing inefficiencies between the different securities. Statistical Arbitrage: The investment manager establishes long and short positions in related securities based on quantitative models that identify pricing inequities. Fixed Income Arbitrage: The investment manager establishes long and short positions in related debt securities or derivative instruments. Global Macro Discretionary: The investment manager utilizes fundamental and/or technical analysis to establish directional positions in any publicly traded market around the world. Typically, managers follow a top down analysis that attempts to identify the largest economic forces within the global economy and position accordingly through the debt, equity, currency or commodity markets. Table 7: Investment style categories (Source: TASS) 80

92 Global Macro Systematic: The investment manager uses technical systems to establish directional positions in major primary and derivative markets around the world. Typically, investment decisions are generated by proprietary computer programs that dictate the specific buy and sell strategies. Dedicated Short Seller: The investment manager attempts to identify securities that are overprice or which it is believed will decrease in value in the near future and establishes short positions. Pure Currency Fund: The strategy is dedicated to trading currencies only. Different currency funds will employ various investment approaches, either fundamental/discretionary or technical/systematic. The strategy can be directional or arbitrage or both. Pure Futures Fund: The strategy is implemented primarily in futures markets, though many managers will carry foreign exchange positions in the interbank market. Pure Emerging Markets: The fund invests exclusively in the emerging market debt or equity markets. Emerging Market funds are the only long only funds listed in the TASS Database. Table 7 (continued): Investment style categories (Source: TASS) 81

93 Panel A Panel B Panel C Dependent Variable: Standard Deviation Dependent Variable: Tracking Error Deviation Dependent Variable: Beta Deviation Specification 1a 1b 1c 2a 2b 2c 3a 3b 3c Reputation Variable Tenure Size Tenure /Size Tenure Size Tenure /Size Tenure Size Tenure /Size Intercept Reputation Variables Mgr. Tenure Fund Size Young/Small (Tenure/Size) Old/Large (Tenure/Size) Location Listed on Exchange Onshore Open to New Investment Open to Non- Accredited Pers. Capital Invested Uses Leverage Fund Policies Log Red. Freqeuncy Log Entr. Frequency Log Minimum Investment (7.79) (-7.61) (3.44) (-0.88) (0.25) (-3.88) (2.41) (0.35) (-3.41) (1.50) (-9.02) (8.00) (-5.35) (-7.30) (-1.75) (-0.90) (-3.81) (2.59) (0.60) (-3.65) (2.65) (-3.12) (7.66) (3.02) (-7.25) (4.54) (-1.18) (-0.10) (-3.46) (2.39) 0.00 (0.24) (-3.48) (1.79) (-7.62) (3.02) (-6.05) (0.79) (-0.24) (1.32) (-1.28) (-0.29) (-0.41) (-1.56) (-0.36) (-5.15) (9.45) (-10.97) (1.98) (-1.18) (0.26) (-0.93) (0.20) (-0.61) (-2.12) (1.41) (1.31) (1.30) (-3.89) (-8.15) (1.05) (-0.52) (1.11) (-1.04) (-0.01) (-0.74) (-1.89) (0.21) (-3.22) (4.38) (-3.49) (-0.27) (0.53) (-2.17) (0.50) (2.46) (1.51) (-2.34) (3.13) (-1.25) (5.37) (-5.70) (1.72) (-0.34) (-3.03) (0.60) (2.43) (1.64) (-2.68) (4.21) (1.22) Table 8: Relationship between risk measures and manager tenure and other fund characteristics (4.01) (3.52) (-10.59) (0.42) (0.19) (-1.97) (0.63) (2.44) (1.30) (-2.50) (3.58) (-0.27) Cross-sectional annual regressions of risk measures against manager tenure and fund characteristics are performed below. The time frame is January, 1994 to December, 2000, for a total of 7 annual regressions. The coefficients are calculated by averaging across cross-sectional regressions, as in Fama-Macbeth (1973). The standard errors are adjusted for autocorrelation and heteroskedasticity. For Panel A, the dependent variable is annual standard deviation. For Panel B, the dependent variable is tracking error deviation, and for Panel C, the dependent variable is beta deviation. For each panel, regression specification (a) uses manager tenure as a proxy for manager reputation, regression specification (b) uses fund size as a proxy for manager reputation, and regression specification (c) uses the manager tenure/fund size interaction variables as proxies for manager reputation. t-values are in parentheses. Number of funds is 982. Adjusted R 2 are reported below.

94 Panel A Panel B Panel C Dependent Variable: Standard Deviation Dependent Variable: Tracking Error Deviation Dependent Variable: Beta Deviation Specification 1a 1b 1c 2a 2b 2c 3a 3b 3c Reputation Variable Tenure Size Tenure /Size Tenure Size Tenure /Size Tenure Size Tenure /Size Fund Fees and Style Mgmt.Fee (% of assets) (0.02) (-0.66) (0.01) (-0.18) (-0.45) (-0.23) (-2.10) (0.98) (2.46) Incent. Fee (% of profits) (4.93) (3.95) 0.29 (4.51) (4.14) (3.54) (3.57) (0.68) (0.51) (0.41) U.S. Style (0.76) (0.82) (0.77) (-0.63) (-0.23) (-0.30) (0.80) (0.82) (0.80) Europe Equity Style (-6.12) (-5.13) (-5.92) (3.06) (4.83) (4.46) (-9.69) (-8.91) (-8.75) Relative Value Style (-10.07) (-10.60) (-10.36) (-0.83) (-0.18) (-0.67) (-11.05) (-12.26) (-12.03) Event Driven Style (-13.89) (-13.67) (-13.96) (0.57) (4.14) (2.04) (-15.22) (-13.92) (-14.84) Adjusted R Table 8 (continued): Relationship between risk measures and manager tenure and other fund characteristics 83

95 84 Panel A Panel B Panel C Dependent Variable: Tracking Error Deviation Dependent Variable: Standard Deviation Dependent Variable: Beta Deviation Specification 1a 1b 1c 2a 2b 1c 2a 2b 3a Reputation Variable Tenure Size Tenure/Size Tenure Size Tenure/Size Tenure Size Tenure/Size Intercept (7.57) Manager tenure (annualized) (1.12) (13.45) (7.65) Fund Size (-8.96) Young age/small size (1.88) Old age/ Large Size (-1.59) Pers. Cap*Manager Tenure (-3.77) Pers. Cap * Fund Size (-1.13) Pers. Cap * Small/Young (-0.32) Pers. Cap * Large/Old (-1.41) (3.32) (-0.98) (-1.08) (18.89) (-14.23) (2.63) (1.14) (2.27) (-3.32) (-0.89) (1.01) (4.14) (-1.09) (-0.94) (4.38) (-4.19) (-5.08) (3.90) (3.29) (-0.62) (-0.95) (-2.06) Adjusted R Table 9: The relationship between risk measures and manager tenure and other fund characteristics, including interactions with personal capital variable Cross-sectional annual regressions of risk measures against all fund and manager characteristics are performed below. The time frame is January, 1994 to December, 2000, for a total of 7 annual regressions. The coefficients are calculated by averaging across cross-sectional regressions, as in Fama-Macbeth (1973). The standard errors are adjusted for autocorrelation and heteroskedasticity. For each panel, regression specification (a) uses manager tenure as a proxy for manager reputation, regression specification (b) uses fund size as a proxy for manager reputation, and regression specification (c) uses the manager tenure/fund size interaction variables as proxies for manager reputation. Additionally, another interaction variable is included, in which the reputation variable is interacted with the personal capital invested variable. To conserve space, only the coefficients on the variables of interest are included. Coefficients on all other variables are consistent with Table 8. t-statistics are in parentheses. Number of funds is 982. Adjusted R 2 are shown below.

96 Dependent Variable: Fund Failure =1 if failed Specification Return and Risk Variables Annual Excess Return (<.001) (0.011) (0.001) Standard Deviation (0.022) Tracking Error Deviation (0.064) Beta Deviation (0.024) Fund & Mgr. Characteristics Manager Tenure (<.001) (<.001) (<.001) (<.001) Fund Location, Status, Risk Fund Size (<0001) (<.001) (<.001) Listed on Exchange (0.747) (0.505) (0.933) (0.256) Onshore (0.068) (0.821) (0.033) (0.542) Open to New Investment (0.002) (0.118) (0.001) (<.001) Open to Non-Accredited (0.185) (0.664) (0.144) (0.029) Personal Capital Invested (0.191) (0.444) (0.533 (0.971) Uses Leverage (0.009 (0.056) (0.09) (0.052 Table 10: Time-varying proportional hazards model: The relationship between fund failure and returns, risk, and manager tenure A time-varying proportional hazards model is estimated below. This model estimates the relationship between the hazard rate (probability of fund failure) and explanatory variables that are permitted to vary over time. The proportional hazard function is specified so that the explanatory variables shift an underlying baseline hazard function up or down. See Section 3.6 for a complete description of the model. Maximum likelihood estimation is used to estimate the model. In the following table, a negative coefficient indicates a positive likelihood of survival, while a positive coefficient indicates a positive likelihood of failure. The variables are most interest are the return, risk-taking, and manager tenure variables. P-values from a chi-squared test are shown in parentheses. Number of total funds is 1659, while number of defunct funds is 632. Only coefficients on the variables of interest are reported. 85

97 Dependent Variable: Fund Failure =1 if failed Specification Fund Policies Log (Redemption Frequency (0.810) (0.752) (0.655 (0.342) Log (Entrance Frequency) (0.877 (0.605 (0.864 (0.151) Log (Minimum Investment) (0.213) (0.594) (0.1723) (0.0557) Fund Fees and Style Management Fee (% of assets) (0.001) (0.084) (0.001) (0.001) Incentive Fee (% of profits) (0.702) (0.161) (0.446) (0.206) US Style (0.279) (0.451) (0.107) (0.001) Europe Equity Style (0.199) (0.986) (0.236) (0.002) Relative Value Style (0.612) (0.313) (0.829) (0.006) Event Driven Style (0.001) (0.466) (0.025) (<.001) Table 10 (continued): Time-varying proportional hazards model: The relationship between fund failure and returns, risk, and manager tenure 86

98 Panel A Panel B Panel C Panel D Specification 1a 1b 1c 2a 2b 3a 3b 4a 4b Independent Variable: Annual Return Annual Return Annual Return Annual Return Description Without Risk-Taking Variables With Standard Deviation With Tracking Error Deviation With Beta Deviation Intercept (-0.20) (-4.09) (-4.68) (-3.46) (-5.12) (-2.33) (-1.35) (-3.32) (-4.75) Manager Tenure (-3.18) (-6.16) (-9.63) (-2.11) (-4.64) (-0.80) (-3.95) (-2.53) (-5.26) 87 Standard Deviation (1.91) (1.93) Tracking Error Deviation (3.31) (2.73) Beta Deviation (1.91) (2.41) Adj. R H.F. Market Controls Used? No Passive H. Fund Passive H. Fund Passive H. Fund Passive H. Fund Table 11: Regression of annual returns on risk variables Cross-sectional annual regressions of excess returns against risk measures, manager tenure, and fund characteristics are performed below. The time characteristics are performed below. The time frame is January, 1994 to December, 2000, for a total of 7 annual regressions. The coefficients are calculated by averaging across cross-sectional regressions, as in Fama-Macbeth (1973). The standard errors are adjusted for autocorrelation and heteroskedasticity. The dependent variable is the fund s annual return less the risk-free rate. To account for each fund's correlation with passive or hedge fund indices, a number of "market" control variables are included in specifications 1b-4b, while specification 1a does not control for market exposure. Coefficients on fund characteristics are not shown to conserve space, and are consistent with the coefficients from the monthly regressions in Table 5. Number of funds is 982. t-statistics are shown in parentheses. Adjusted R 2 are shown below.

99 Panel A Panel B Panel C Dependent Variable: Tracking Error Deviation Dependent Variable: Standard Deviation Dependent Variable: Beta Deviation Specification 1a 1b 2a 2b 3a 3b Reputation Variable of Interest Fee Income Fee Income* Mgr. Tenure Fee Income Fee Income* Mgr. Tenure Fee Income Fee Income* Mgr. Tenure INDEPENDENT VARIABLES (below) Intercept (7.39) (7.04) (1.26) (2.11) (4.29) (4.13) Manager tenure (annualized) Young manager tenure indicator (bottom 1/3) Old manager tenure indicator (top1/3) Low fee indicator (bottom 1/3) High fee indicator (top 1/3) Young tenure*low fee interaction (indicator) Old tenure*high fee interaction (indicator) (-5.55) (0.03) (1.03) (-1.04) (-2.70) (-0.17) (0.50) (-4.63) (2.60) (-3.44) (0.04) (-1.16) (0.39) (-2.15) (-3.26) (2.34) (-2.67) (-0.93) (0.02) (1.60) (-2.71) Adjusted R Table 12: Relationship between risk measures and fee income and manager tenure/fee income variable Cross-sectional annual regressions of risk measures against fee income, manager tenure, and manager tenure/fee income interaction variables are performed below. Fund characteristics are included as control variables in all the regressions, although their coefficients are not reported. See Table 1 for a list of fund characteristics. The time frame is January, 1994 to December, 2000, for a total of 7 annual regressions. The coefficients are calculated by averaging across cross-sectional regressions as in Fama- Macbeth (1973). The standard errors are adjusted for autocorrelation and heteroskedasticity. For Panel A, the dependent variable is annual standard deviation. For Panel B, the dependent variable is tracking error deviation, and for Panel C, the dependent variable is beta deviation. For each panel, the dependent variables of interest for specification (a) are the indicator variables for low or high estimated cumulative fee income, and the dependent variables of interest for specification (b) are the interactions of manager tenure and fee income. Number of funds is 982. t-statistics are reported in parentheses. 88

100 Variable All Funds Live Funds Dead Funds Monthly Net Return 0.64% 1.11% -0.14% Manager Tenure (months) Fund Size (millions) $82.2 $119.7 $21.5 Listed on Exchange? 16% 15% 17% Onshore 44% 46% 40% Open to New Investment 84% 75% 97% Open to Non-Accredited Investors 13% 8% 21% Personal Capital Invested 59% 57% 63% Uses Leverage 74% 70% 80% Redemption Frequency (months) Entrance Frequency (months) Minimum Investment (millions) $0.83 $1.03 $0.48 Management Fee (% of assets) 1.38% 1.21% 1.65% Incentive Fee (% of profits) 18.17% 18.28% 17.99% U.S. Equity Style 22% 27% 12% European Equity Style 5% 7% 2% Relative Value Style 14% 15% 12% Event Driven Style 10% 14% 4% Number of Funds Table 13: Summary statistics for sample of 1659 funds Above are averages of fund and manager characteristics over the period 1994 to Funds must have at least six months of consecutive returns to be included in the sample. Monthly Net Return is net of all expenses, fees, and is in excess of the 1-month Treasury Bill rate. Manager Tenure is the number of months that the manager has been overseeing the fund. Size is in millions of dollars. Listed on Exchange is a 0/1 indicator variable set to one (1) if the fund is listed on a stock exchange. Onshore is a 0/1 indicator variable set to one (1) if the fund is headquartered in the United States. Open to New Investment is a 0/1 indicator variable set to one (1) if the fund is accepting new investors. Open to Non-Accredited Investors is a 0/1 indicator variable set to one (1) if the fund is open to non-accredited investors. Personal Capital Invested is a 0/1 indicator variable set to one (1) if the manager reports that she has invested her own money in the fund. Uses Leverage is a 0/1 indicator variable that is set to one (1) if the manager reports that she uses leverage. Redemption frequency measures the number of months that an investor must keep his money in the fund before withdrawing it. Entrance frequency measures how often new investors may enter the fund. Minimum investment is the minimum initial investment required and is reported in millions of dollars. Management fee is the annual fee charged by the fund, measured as a percentage of assets. Incentive fee is the annual incentive fee that may be charged by the fund, and is measured as a percentage of annual (positive) profits. US Equity Style, Europe Equity Style, Relative Value Style, and Event Driven Style are 0/1 indicator variables that are set to one (1) if the self-reported fund style is that style. Live funds are those in operation at the end of the sample period. Dead funds are those that ceased operations before the end of the sample period. 89

101 Panel A: Passive Indices Mean Median Max. Min. Std. Dev. Skew. Kurtosis US Dollar Weighted Index -0.19% -0.24% 3.78% -5.18% 1.73% Gold -0.73% -0.94% 20.51% -7.11% 3.69% Commodities 0.42% 0.47% 16.81% % 5.40% CRSP Value Weighted 0.95% 1.56% 9.23% % 4.16% LB Aggregate Bond -0.43% -0.26% 3.60% -3.66% 1.26% LB 30 Yr. U.S. Treasury Bond 0.15% 0.10% 8.61% -7.78% 2.81% SMB -0.36% -0.46% 14.23% % 4.93% HML -0.28% -5.35% 15.40% % 3.83% Momentum 1.56% 0.81% 18.23% 8.98% 4.43% Panel B: Hedge Fund Style Indices Mean Median Max. Min. Std. Dev. Skew. Kurtosis Managed Futures 0.08% -0.25% 9.56% -9.82% 3.32% Convertible Arbitrage 0.45% 0.73% 3.13% -5.06% 1.44% Emerging Markets 0.14% 0.43% 16.14% % 5.93% Distressed Securities -0.33% -0.53% 22.32% -9.08% 5.52% Market Neutral 0.54% 0.57% 2.89% -1.53% 0.98% Event Driven 0.56% 0.67% 3.43% % 1.91% Fixed Income Arbitrage 0.14% 0.40% 1.61% -7.35% 1.25% Global Macro 0.74% 0.83% 10.14% % 4.12% Long/Short Equity 0.91% 0.95% 12.63% % 3.65% Fund of Funds Index 0.24% -0.10% 4.92% -3.73% 1.83% Table 14: Summary statistics for passive and hedge fund indices Above are mean buy-and-hold monthly returns and other summary statistics for the passive indices and hedge fund style indices used in the paper. The period is January, 1994 to December, All returns are in excess of the risk-free rate. The sources for the indices are Datastream and WRDS (passive) and CSFB/Tremont (hedge fund style). The returns on HML (a high book value minus low book value stock portfolio), SMB (a small-capitalization minus large-capitalization stock portfolio), and MOM (a momentum portfolio) were obtained from Kenneth French's website. Panel A shows statistics for the passive indices, while panel B shows statistics for the hedge fund style indices. For a detailed description of the hedge fund indices, see Table 3. 90

102 91 Portfolio Formation Period Testing Period INDEPENDENT VARIABLES Monthly Monthly Monthly Standard Excess Standard Int cept VW US $ Agg. Deviation Return Deviation (Alpha) CRSP Trade Bond Monthly Excess Return 1 (worst) -7.87% 6.19% -0.66% 4.79% -0.90% (-2.48) % 3.70% -0.80% 3.46% -0.46% (-2.18) % 2.20% -0.29% 2.65% 0.05% (0.36) % 1.71% -0.16% 3.07% -0.01% (-0.05) % 1.32% -0.17% 1.40% -0.02% (-0.17) % 1.82% -0.07% 2.06% 0.06% (0.60) % 1.20% -0.31% 1.92% 0.16% (1.30) % 1.75% 0.05% 2.41% 0.24% (1.60) % 2.24% 0.30% 2.41% 0.20% (0.81) 10 (best) 7.14% 5.24% 0.49% 4.11% -0.28% (-0.62) 10-1 (bestworst) 15.01% 8.55% 1.14% 6.25% 0.62% (1.05) (1.79) (3.51) (3.98) (2.94) (2.76) (-0.27) (0.90) (-0.22) (-0.87) (-0.34) (-1.50) (-3.98) (-0.86) (-0.46) (-0.55) (0.44) (1.27) (1.53) (1.76) (1.96) (1.59) (2.61) (-0.24) (-0.97) (0.55) (-1.27) (-1.15) (-3.04) (-1.80) (-2.49) (-1.85) (-1.14) (-0.61) Long Bond EM MF LS (-2.70) (-3.39) (-2.10) (-1.49) (-0.69) (1.26) (1.00) (1.84) (1.10) (0.91) (2.50) (4.43) (2.91) (2.22) (0.82) (0.62) (2.59) (2.21) (1.92) (1.05) (0.87) (-2.25) (0.98) (2.62) (3.73) (1.83) (0.90) (2.44) (2.02) (1.11) (0.99) (1.14) (0.04) (1.01) (1.56) (0.91) (0.68) (1.28) (2.26) (2.37) (5.28) (6.75) (5.08) (3.36) Adj. R Table 15: Quarterly persistence analysis when funds are selected based on prior performance Hedge funds are sorted at the beginning of each quarterly period from the second quarter in 1994 to the final quarter in 2000 into decile portfolios based on their previous quarter's return less the risk-free rate. The portfolios are equally weighted quarterly so the weights are readjusted whenever a fund disappears. Funds with the highest past quarterly returns in excess of the risk-free rate comprise decile 10 and funds with the lowest past quarterly returns comprise decile 1. The dependent variable is the portfolio's excess monthly return. The dependent variables are the passive indices and the hedge fund indices described in Table 14. To save space, only the indices with the most statistically significant coefficients are shown. Alpha is the intercept of the model. t-statistics are in parentheses. Number of funds is 1659.

103 92 Portfolio 1 (poor perf./ old age) 2 (poor perf./ middle age) 3 (poor perf./ young age) 4 (middle perf./old age) 5 (mid. perf./ middle age) 6 (mid. perf./ young age) 7 (good perf./ old age) 8 (good perf./ middle age) 9 (good perf./ young age) Formation Period Testing Period INDEPENDENT VARIABLES Monthly Monthly Monthly Monthly Excess Standard Excess Standard Int cept VW US $ Agg. Long Return Deviation Return Deviation (Alpha) CRSP Trade Bond Bond EM MF LS -2.54% 2.98% 0.27% 2.47% -0.41% (-1.86) -2.70% 3.22% 0.25% 3.09% 0.51% (-2.48) -2.67% 3.18% 0.38% 2.85% -0.30% (-1.24) 0.53% 1.32% 0.50% 1.42% 0.10% (0.90) 0.52% 1.26% 0.49% 1.56% -0.13% (-1.08) 0.56% 1.35% 0.65% 1.47% 0.11% (1.47) 3.76% 3.20% 0.71% 3.10% -0.13% (-0.57) 3.92% 3.06% 0.97% 2.85% 0.10% (0.46) 4.21% 3.28% 1.48% 3.34% 0.29% (1.03) (2.51) (2.94) (2.34) (1.06) (2.09) (1.86) (-0.71) (-0.23) (-0.48) (-1.82) (-1.79) (-2.49) (1.78) (1.07) (-1.26) (2.17) (1.85) (1.62) (0.88) (-0.43) (-1.44) (-0.73) (-2.72) (-3.89) (-2.30) (-1.79) (-1.05) (-2.65) (-3.39) (-2.30) (0.89) (-0.34) (-1.57) (2.13) (0.83) (0.83) % 3.62% 1.21% 2.90% 0.69% (2.05) (-2.09) (2.12) (-1.33) (2.58) Table 16: Quarterly persistence analysis when funds are selected based on prior performance and manager tenure (1.82) (4.44) (4.77) (1.04) (0.77) (2.32) (-0.11) (1.64) (2.23) (0.36) (2.84) (1.36) (1.62) (1.93) (1.90) (1.26) (2.01) (2.05) (-0.03) (-1.98) (1.82) (0.62) (1.50) (0.49) (1.73) (3.78) (6.13) (5.07) (6.47) (4.45) Adj. R Hedge funds are sorted at the beginning of each quarter from the second quarter in 1994 to the final quarter in 2000 into thirds portfolios based on their previous quarterly return less the risk-free rate. The portfolios are equally weighted quarterly so the weights are readjusted whenever a fund disappears. These portfolios are then cross-sorted based on the quarter-end value of manager tenure into three additional portfolios: young, middle, and old. There are nine portfolios, ranging from poor and old to good and young. A tenth portfolio is created which is long the good and young managers and short the poor and old managers. The dependent variable is the portfolio's monthly return in excess of the risk-free rate. The independent variables are the passive and hedge fund indices described in Table 14. To save space, only the indices with the most frequent statistically significant coefficients are shown. Alpha is the intercept of the model. t-statistics are in parentheses. Number of funds is 1659.

104 Specification Current qtr. excess return (<0.001) One qtr. lagged excess return Two qtr. lagged excess return Three qtr. lagged excess return Manager tenure (<0.001) Panel A: Unconditional Estimation (<0.001) (<0.001) (<0001) (<0.001) (<0.001) (<0.001) (<0.001) (<0.001) (<0.001) (<0.001) (<0.001) (<0.001) (<0.001) (<0.001) Bottom third past performance and old tenure Middle third past performance and old tenure Top third past performance and old tenure Young tenure and good past performance Old tenure and good past performance (0.05) (0.01) Panel B: Conditional on Past Performance (0.05) (0.35) Panel C: Conditional on Manager Tenure (0.42) Table 17: Conditional time-varying proportional hazards models A time-varying proportional hazards model is estimated below. This model estimates the relationship between the hazard rate and certain explanatory variables that are permitted to vary over time. The proportional hazard function is specified so that the explanatory variables shift an underlying baseline hazard function up or down. See Section 3.6 for a complete description of the model. Maximum likelihood estimation is used to estimate the model. In the following table, a negative coefficient indicates a positive likelihood of survival, while a positive coefficient indicates a positive likelihood of failure. P-values from a chi-squared test are shown in parentheses. Number of total funds is 1659, while number of defunct funds is 632. Only coefficients on the variables of interest are reported. 93

105 APPENDIX B FIGURES 94

106 95

107 96 Figure 3: Estimated survivor functions by manager tenure