P. 1
Hedge Funds Strategy - DnB

Hedge Funds Strategy - DnB

|Views: 14|Likes:
Published by Manuela Mihailiasa

More info:

Published by: Manuela Mihailiasa on Jan 20, 2013
Copyright:Attribution Non-commercial

Availability:

Read on Scribd mobile: iPhone, iPad and Android.
download as PDF, TXT or read online from Scribd
See more
See less

12/02/2013

pdf

text

original

An Analysis of Hedge Fund Strategies

Daniel P.J. CAPOCCI HEC-ULG Management School – University of Liège (Belgium) PhD Thesis in Management

Electronic copy available at: http://ssrn.com/abstract=1008319

A mon petit Louis

Electronic copy available at: http://ssrn.com/abstract=1008319

Electronic copy available at: http://ssrn.com/abstract=1008319

1

An Analysis of Hedge Fund Strategies

by

Daniel P.J. CAPOCCI Senior Portfolio Manager – Kredietbank Luxembourgeoise S.A. HEC-ULG Management School – University of Liège (Belgium) Luxembourg School of Finance – University of Luxembourg EDHEC Risk and Asset Management Research Center

An Analysis of Hedge Fund Strategies – Table of Contents

Acknowledgements .............................................................................................. i

An Analysis of Hedge Fund Strategies - Abstract .................................................. i

Preface................................................................................................................ ii

Introduction and Purpose .................................................................................... 1

Analysis of Hedge Fund Performance ................................................................ 45

Hedge Fund Performance and Persistence in Bull and Bear Markets ................ 120

The Sustainability of Hedge Fund Performance: New Insights ......................... 190

The Neutrality of Market Neutral Funds ........................................................... 273

Diversifying using Hedge Funds: A Utility-Based Approach .............................. 339

General Conclusion ......................................................................................... 405

An Analysis of Hedge Fund Strategies – Table of Contents

References ...................................................................................................... 410

An Analysis of Hedge Fund Strategies – List of Figures ................................... 424

An Analysis of Hedge Fund Strategies – List of Figures ................................... 425

An Analysis of Hedge Fund Strategies – List of Tables ..................................... 426

An Analysis of Hedge Fund Strategies – Detailed Table of Contents ................ 431

An Analysis of Hedge Fund Strategies – Detailed Table of Contents ................ 431

Acknowledgements

I would like to thank sincerely everyone that helped me directly and indirectly during my work on this thesis. I will start by thanking my parents, Christine Wenders and Nazzareno Capocci for their constant support. I would also like to thank my boss, Patrick Vander Eecken, my sweetheart Renata Vitórica as well as all the authors, anonymous referees and conference participants (European Investment Review – 2003 and 2004, Global Finance Conference – 2005, Institutional Fund Management – 2006) for their constructive and helpful comments on my work. In particular, I would like to thank Vikas Agarwal, Naratip Balajiva, Jean-Marc Brisy, David Capocci, Mark Carhart, Gabriel Catherin, Kenneth French, Greg N. Gregoriou, David Hsieh, Bing Liang, Loistl Otto, Narayan Naik, Roger Otten, Magali Thelen and Dee Weber. I would also like to thank Frédéric Duquenne for using his impressive programming capabilities to help me finalise my final study. Last but certainly not least, I would like to thank the members of my doctoral committee and particularly my co-supervisors Georges Hübner, who worked with me on several studies and who showed me the way on several occasions, as well as Albert Corhay who provided constructive remarks on my papers.

i

i

Our aim is to present hedge funds. we develop a multi-factor performance analysis model. but also skewness and kurtosis in order to determine whether hedge funds are really beneficial to investors. This model aims to determine both whether hedge funds create pure alpha over time (alpha over classical markets) and whether there is persistence in hedge fund returns over time. i . Finally. their neutrality relative to equity markets in order to validate hedge fund managers’ claims that they are market neutral. I analyse another specific aspect of hedge funds.Abstract This PhD thesis analyses hedge fund strategies in detail by decomposing hedge fund performance figures.An Analysis of Hedge Fund Strategies . use it over several time periods and improve it over time. to understand what managers expect to do and to understand how they make or destroy value over time. we develop new efficient frontier measures. In order to achieve this objective. Following this. which not only include returns and volatility.

hedge funds are present almost everywhere. At this time. This paper was published in the Review of Financial Studies and was called Empirical Characteristics of Dynamic Trading Strategies: the Case of Hedge Funds. hedge funds were somewhat unknown investment vehicles providing very limited information. the real risks underlying alternative strategies. who have tried to understand the funds themselves. At the heart of my thesis are the performance figures of hedge fund managers. Today. how hedge fund managers make money. The thesis is classified in three parts: Part 1: The Persistence in Hedge Fund Performance Analysis of Hedge Fund Performance Hedge Fund Performance and Persistence in Bull and Bear Markets Sustainability in Hedge Fund Performance: New Insights ii . Since then. They have been studied for years by dozens of academicians and professionals. the use of hedge funds as diversification tools and how they are really different from other investment products. My entire thesis is based on performance figures and their use as a way to understand hedge fund strategies. the finance industry and the world of hedge funds in particular have changed dramatically.Preface I discovered the term hedge fund at the library of the University of Liège in September 1999 while reading the abstract of a paper by William Fung and David Hsieh.

looks at this prominent hedge fund strategy. as measured by the standard deviation. thereby allowing them to reach a higher efficient frontier. The main difference between this and my first research on the subject is that both are based on the same underlying idea.Part 2: An Analysis of Hedge Fund’s Market Exposure The Neutrality of Market Neutral Funds Part 3: Hedge Funds as Diversification Tools Diversifying Using Hedge Funds: A Utility-Based Approach The thesis starts with an extended multi-factor model and a new way of classifying funds that enables us to identify alpha creators following a persistence analysis (part one). iii . the last study of my doctoral thesis attempts to determine whether an investor can reach a higher level of utility by including hedge funds in his portfolio of stock and bond mutual funds. on market neutral funds. My first study on hedge funds performed prior to my PhD uses “spanning” as a way of determining whether hedge funds enable investors to improve significantly the risk-return profile of their portfolio. The difference is that in the final study I define and estimate an adapted efficient frontier that takes into consideration not only volatility. The third and final study is on the inclusion of hedge funds in a portfolio. but also higher moments into account when determining the impact of hedge fund insertion in a classical portfolio (the hypothesis of normality is no longer needed). Part two. Interestingly.

assets under management as well as the strategies employed. I have really enjoyed exploring the world of hedge funds and this thesis is only a start.In coming to the end of this thesis. iv . I believe that I have not only learned a great deal. but have also developed new ideas encouraging me to continue to work on the subject. since the world of hedge funds is continuously evolving in terms of the number of funds.

.

.

when the NASDAQ Composite Index reached an all-time high of 5.000 hedge funds were managing assets totalling ca. and finished three years later with a floor level of 1. This part of the definition is much more open and allows almost anyone to classify his fund as a hedge fund as long as it is long and short… Since the early 1990s. but since then has enjoyed a renewed vitality with an estimated total of 10. Capocci. using various investment strategies and these funds are accessible to large investors only. 1 . this definition is precise.000 funds managing more than a trillion US dollars by the end of 2006.132. From another point of view. The industry only suffered from a relative slowdown in 1998. The growing trend of the sector remained remarkably sustained during the stock market collapse that started in March 2000. $60 billion.000 HF managing around $1trillion. Capocci & Hübner (2004) estimated that there were 6. the subsequent growth in the number and asset base of hedge funds has never really been refuted. On the one hand. it is very broad. More recently. on the other. 2003). the funds may use various kinds of securities on various markets. 2002). In the meantime. contrasting with a decrease of 2. This is clear and focused.7% in the worldwide mutual fund industry (Investment Company Institute. the global net asset value (NAV) of hedge funds continued to grow at a steady rate of 10.253.000 HF managing around $400b. when around 2. This is a growth of 11% in the number of funds and 26% in assets over six years.Introduction and Purpose Hedge funds are private investment vehicles that can take long and short positions in various markets. In 2007.6% (Van Hedge Funds Advisors International. Duquenne & Hübner (2007) estimate that there are 10. in 2001.

market timers or funds with a more directional bias (see Part two: An Analysis of Hedge Fund’s Market Exposure). Hedge Fund Performance and Persistence in Bull and Bear Markets and Sustainability in Hedge Fund Performance: New Insights).The purpose of this doctoral thesis is clearly established: to understand hedge fund strategies by looking at the performance numbers produced. The second objective of the thesis is clearly linked to the first. I check for this neutrality and analyse what kind of funds consistently outperform over time: the pure market neutral funds. testing and improving a performance analysis model to understand hedge fund performance. market neutral funds must a limited exposure to the market. I achieve this objective in three complementary studies grouped in Part 1 (Analysis of Hedge Fund Performance. By definition. market neutral funds. I perform a specific analysis on the most represented and the most interesting. Our first objective of the studies is to understand clearly hedge fund managers and to explain how they create alpha over time. Since the purpose is to understand hedge fund strategies in detail. This involves developing. while developing and adapting a methodology to determine whether there is any persistence in hedge fund returns on the other. 2 .

Finally. I present the data issue before disserting on investing in hedge funds for the final investors. 3 .Finally. Then. The main originality of this study centres upon the development of a new efficient frontier. The specificities and main objectives of each study are reported in Table 1. In the remainder of this introduction I present a global literature review. we present the three parts of the Thesis in detail. based not only on volatility but also on higher moments (skewness and kurtosis) and on a utility function that more closely corresponds to that of the investor without normality or other strong assumption. the third complementary objective for the thesis is to determine whether hedge fund strategies should be included in a classical portfolio of stocks and bonds. we analyse the inclusion of hedge funds in a portfolio of stocks and bonds. Diversifying Using Hedge Funds: A Utility-Based Approach. In Part three.

Determine if HF strategies sign.Find a systematic way of buying HF in order to sign. and persis. Hedge Fund Performance and Persistence in Bull and Bear Markets and the Sustainability of Hedge Fund Performance: new insights).Determine if HF strategies sign.Determine if bond and/or equity classical portfolio taking abnormality a portfolio can be diversified with Diversifying using investors should include HF in their into account securities displaying abnormal return Hedge Funds portfolio . For each Part we report the main specificity of the study and its first and second objective.Analyse the impact of inserting HF in a .Determine if high/low beta market neutral funds outperform their hig/low beta peers . outperform classical markets .Determine the market exposure of market neutral funds . measures The Neutrality of Market Neutral Funds . charact. outperform classical markets in bull and/or bear market conditions .Focus on market neutral funds (28% of the industry) . undir HF and dist. based on performance Hedge Fund Performance & other risk-adj. Part 2: An Analysis of Hedge Fund’s Market Exposure contains one study (The Neutrality of Market Neutral Funds).Introduction of the extended multifactor model .Determine if HF strategies sign. outperform classical markets .Develop a methodology to determine if .Determine if HF strategies sign. Part 1: The Persistence in Hedge Fund Performance contains three studies (An Analysis of Hedge Fund Performance.Adapted model (option factors) The Sustainability of .Distinguish between dir. and persis.Consider various market conditions & adpated model (high yield & mortgage factors) .Analysis based on performance .Adapted meth. Part 3: Hedge Funds as Diversification Tools contains one study (Diversifying Using Hedge Funds: A Utility-Based Approach). outperform classical markets in bull and/or bear market conditions . outperform classical markets .Focused on LT and ST periods and various market conditions . and persis. FoF This Table reports the specificities and the objectives of the studies grouped in this PhD Thesis. .Determine if HF strategies sign.Table 1: Studies Specificities and Objectives Specificity .Analysis based on performance Objective 1 Objective 2 An Analysis of Hedge Fund Performance Hedge Fund Performance and Persistence in Bull and Bear Markets . outperform classical markets using several risk-adjusted measures .

2004) are more cautious in their conclusion. In the first global category. there is statistical evidence of negative persistence for directional portfolios in both the bullish and the bearish periods. 2. hedge fund academic studies can be classified into four global categories: 1. Hübner and Papageorgiou (2006) find that there are three kinds of persistence in hedge fund returns. 1999. while others (Ackermann et al. Hedge fund performance. Liang. Agarwal and Naik. 1999. The first of these fields includes studies that compare the performance of hedge funds with equity and other indices (see for example Ackermann.Global Literature Review There have been many studies on hedge funds covering many different aspects of the industry. Brown. 3. 1999.. As illustrated in Figure 1.. Firstly. 1999. Agarwal and Naik. Gibson and Gyger. Finally. Liang. 2002. we perform literature reviews including studies whose results are directly linked to the subject under analysis. There are three fields within this first category of hedge fund performance analysis. 2001. Results of such studies are mitigated. Hedge fund investment style. Secondly. we report studies that are focused on hedge fund performance. 2001. Capocci et al. the authors find statistical evidence of progressive positive 5 . Goetzmann and Ibbotson. 1999. Correlation analysis and diversification power and 4. Some authors (Brown et al. Amin and Kat. 2004). McEnally and Ravenscraft. Other studies. In each of the specific studies reported in the heart of the thesis. that there is statistical evidence of positive persistence based on alphas for non-directional portfolios in the bullish period. 2005) conclude that hedge funds have been able to outperform these indices. 1999. we provide a global literature review on hedge fund studies. In this introduction. Liang. Liang. Barès. 2003.

They prove that offshore hedge funds have positive risk adjusted returns. as suggested by Brown. particularly for poorly performing funds that continue to underperform. Ackermann. the hedge fund industry has a higher attrition rate than is the case in mutual funds (see Brown. Goetzmann and Ibbotson. 6 . McEnally and Ravenscraft (1999) and Liang (1999) find that hedge funds consistently achieve better performance than mutual funds. The third field of hedge fund performance analysis includes the study of the persistence of hedge fund returns. In this context. Agarwal and Naik (2000) analyse the presence of persistence in hedge fund returns using a one-year moving average period. Goetzmann and Ibbotson (1999) and Liang (2000.persistence based on alphas for funds of funds in both the bullish and the bearish periods. The second field of hedge fund performance analysis compares the performance of hedge funds with mutual funds. although they are lower and more volatile than the reference market indices considered. Persistence is particularly important in the case of hedge funds because. 1999). 2001). but they attribute this result to style effect and conclude that there is no proof of any particular alpha-generating capacity of some fund managers. They find that there is proof of persistence in hedge fund performance.

. hedge fund investment style. Posthuma and Van der Sluis) . 1999. 2000). 2003. 2002. 2002a. Liang 2003)..Risks (Schneeweis and Spurgin. 1999. . Schneeweis and Spurgin. Amenc and Martellini. Gregoriou and Rouah. 1996. Jorion. 2002. 2001. 1997.Sharpe style analysis (Fung and Hsieh. . 2002). Liang. Barès et al. Hedge fund indices (Brooks and Kat. Ben Dor and Jagannathan. 1997.. 2003. Fung and Hsieh. Figure 1 reports four global categories of hedge fund academic studies. Amenc et al. .. 2004).Dynamic model (Swinkels and Van der Sluis. 2001. 2001.. 1999). Brown and Goetzmann. 1999 and Liang. Fung and Hsieh. diversification power and finally the other studies.CTAs (see for example Edwards and Park. Amin and Kat. Remolona and Tsatsaronis. 2001). Liang 2001. 2000. . Spurgin and Georgiev.Comparison with mutual funds (Ackermann et al. .Persistence in performance (Agarwal and Naik. 1999. 1998. 2000. Liang. Liang. Brown et al. 2002b. Brown et al. 1999. Amenc and Martellini.Bias analysis (Liang. Fung and Hsieh.. 2000. 1999. . We group studies on hedge fund performance. Hübner and Papagergiou. 2001. Agarwal and Naik. Brown et al. Brealy and Kaplanis.Diversification power (Amin and Kat. 2003. Liang 2001. correlation analysis. 1997. Liang. 2000.Rolling regression (McGuire.. . Liang. 2002). 2001. 2002..Comparison with classical markets (Ackermann et al. 2001. 1999). 2005) .Correlation analysis (Fung and Hsieh.Figure 1: Four global categories of hedge fund academic studies HEDGE FUND PERFORMANCE HEDGE FUND INVESTMENT STYLE CORRELATION ANALYSIS and DIVERSIFICATION POWER OTHER STUDIES . 2001. . Liang. 2001. 1999. 2006. 2002b).. Amenc et al. Agarwal and Naik. Berényi.

They find that the regressions had little explanatory power and consequently suggest that the resulting low adjusted r-square is due to the funds’ trading strategy. Goetzmann and Park (1998) analyse hedge fund returns during the 1997-98 8 . in general.. 1997. The second global category of hedge fund academic studies includes authors that try to analyse and describe hedge fund investment style and who explain these features with style models (see for example. More recently. Brealy and Kaplanis. Fung and Hsieh. Event Driven and Macro) provided protection to investors when stock markets decline. rather. They assume that fund returns are linearly related to the returns through a number of factors and measure those factors through eight mimicking portfolios. Brown. Goetzmann and Park. Brown et al. In this context. Brown. Fung and Hsieh (1997) apply Sharpe's style analysis (see Sharpe. This effect comes from the fact that hedge fund managers are opportunity driven and therefore change style over time. Liang 2001. hedge funds did not provide investor protection after the market downturn of March 2000. 2001. A particular aspect that has been taken into account more recenthly is the style drift in hedge fund returns. their superior performance was mostly due to the good market timing of their managers. Ennis and Sebastian (2003) contend that. The periods under review do not favour this exercise. 1992) to a large sample of hedge funds and commodity trading advisors (CTAs). Ben Dor and Jagannathan (2002) stress the importance of selecting the right style benchmarks and emphasise how the use of inappropriate style benchmarks may lead to the wrong conclusion. 1998. as periods of downward trending stock markets were rare and discontinuous between 1994 and March 2000. 2001. Edwards and Caglayan (2000) found that only three types of hedge fund strategies (Market Neutral. 2002 and Liang 2003). Ben Dor and Jagannathan.The vast majority of performance studies on hedge funds have not focused solely on the behaviour under different market conditions. For the period 1990-1998.

thanks to the weak correlation to the funds with other financial securities. The methodology consists in realizing a set of linear regressions and moving the estimation period of each of them by one observation. McGuire. Fung and Hsieh (1997) and Schneeweis and Spurgin (1998) prove that the insertion of hedge funds into a portfolio can significantly improve its risk-return profile. Remolona and Tsatsaronis (2005) apply the same methodology. The Kalman filter procedure chooses the optimal weighting scheme directly from the data. This technique is adaptive in the sense that changes in the style exposures are priced up automatically from the data. this model is a state-space model and can be estimated by using standard Kalman filter techniques1.Asian crisis using rolling regression to take the style drift into consideration. the time variation in the exposures is explicitly modelled. Unlike the ad hoc rolling regression approach. As stressed by Posthuma and Van der Sluis. This methodology has one major drawback: the choice of a number of observations used for the estimation. The third global category of hedge fund academic studies focuses on the correlation of hedge funds with other investment products and analyses the power of the diversification properties of hedge funds. No window size and ad hoc chosen length need to be used. No restrictions are imposed on the betas. but that a great majority of funds 1 See Pollock (1999) for a detailed presentation of the Kalman filtering and space-state models. This simple technique enables one to observe style variation of a manager over time. The filter is an adaptive system based on the measurement and updating equations. This low correlation is also emphasised by Liang (1999) as well as by Agarwal and Naik (2004). Amin and Kat (2001) find that stand-alone investment hedge funds do not offer a superior risk-return profile. To handle this issue. 9 . Posthuma and Van der Sluis (2005) propose to use a dynamic style model in which beta can vary over time developed by Swinkels et Van der Sluis (2001).

McEnally and Ravenscraft (1998) emphasise that stricter legal limitations for mutual funds rather than for hedge funds hinder their performance. and Amenc and Martellini. In the fourth global category of hedge fund academic studies. Schneeweis and Spurgin (1999) as well as Amenc. 2003). Liang (2000) analyse the presence of survivorship bias in hedge fund data and Fung and Hsieh (2000) include other biases in their analysis. There are many 10 . Martellini and Vaissié (2002) and Berényi (2002) study the risks involved in hedge fund investments.classified as inefficient on a stand-alone basis are able to produce an efficient payoff profile when mixed with the S&P 500. Some authors alos studied hedge fund indices (see Brooks and Kat. but that other risks such as volatility risk. other authors have analysed various other aspects of the hedge fund industry. Kooli (2007) analyzes the power of hedge funds as an efficient frontier enhancer. Martellini and Vaissié (2002) prove that hedge fund returns are not only exposed to the market risk. 2001. “Other studies”. They obtained the best results when 10-20% of the portfolio value is invested in hedge funds. hedge funds are seen as good investment tool. Amenc. Amenc. default risk or liquidity risk have to be considered. He finds that hedge funds as an asset class imprive the mean-variance frontier of sets of benchmarks portfolios but that investors who already hold a diversified portfolio do not improve their statistics using hedge funds. Jorion (2000). The author finds however that funds of hedge funds do bring diversification for meanvariance investors. Curtis and Martellini (2002). Amenc and Martellini (2002) prove on the basis of ex-post estimations that the inclusion of hedge funds in a portfolio can lead to a significant decrease in the volatility of the portfolio without leading to a significant change in the returns. This implies that a stronger risk control does not necessarily correspond to a decrease in return. Schneeweis and Spurgin (2000). Ackermann. Taking all these results into consideration.

Billingsley and Chance (1996) and Edwards and Park (1996) demonstrate that CTAs can add diversification to stocks and bonds in a mean-variance framework. Commodity trading advisors (CTAs) are a particular category in the hedge fund world. 1987. whereas others have studied them either by separating them from hedge funds (Liang. 1990. However. Savanayana and McCarthy (1991) and Schneeweis (1996) stated that the benefits of CTAs are similar to those of hedge funds. 2003 and Capocci. 1989. which first appeared an academic journal in 1997. Schneeweis.net. Research on CTAs is very sparse and it is difficult to present a complete literature review. HFR. 1997.. Fung and Hsieh (2002b) looked at the natural biases present in hedge fund indices. Schneeweis and Spurgin. 1988). Altvest and Magnum. Zurich Capital. Many studies were published in the late 80s and in the early 90s (see for example. some authors have considered CTAs as part of the hedge fund world (Fung and Hsieh. Van Hedge. 2001. LJH Global Investment. the Hennessee Group. in that they improve upon and can offer a superior risk-adjusted return trade-off to stock and bond indices while acting as diversifiers in investment portfolios.different hedge fund index providers such as EACM. 2000). 2004b). Since 1997. Hedgefund. 11 . or Edwards and Ma. Elton et al. Mar. CSFB/Tremont. 2003) or on a stand-alone basis (Fung and Hsieh. CTAs have been studied for a longer period of time. Unlike hedge funds. Gregoriou and Rouah.

Schneeweis. Fung and Hsieh (2001) analyse CTAs and conclude that they are similar to a look-back call and a look-back put. Kazemi. 1998. According to Schneeweis. Kat (2002) finds that allocating to managed futures allows investors to achieve a very substantial degree of overall risk reduction at limited cost. Schneeweis and Georgiev (2002) conclude that careful inclusion of CTA managers in an investment portfolio can enhance its return characteristics. the correlation of managed futures to the S&P 500 during its best 30 months was 0. They prove that all classifications (except the diversified sub-index) behave in the same way as a random walk. during bear markets. The effectiveness of CTAs in enhancing risk-return characteristics of portfolios could be compromised when pure random walk behaviour is identified. Jack Schwager reviews the literature on whether CTAs exhibit performance persistence and conducts his own analysis.25 during the worst 30 12 . between 1990 and 1998.Fung and Hsieh (1997) prove that a constructed CTA style factor has a persistent positive return when the S&P 500 has a negative return. He concludes that there is little evidence that the top performing funds can be predicted. In his book Managed Trading: Myths and Truths. According to Worthington (2001). Managed futures appear to be more efficient diversifiers than hedge funds. Gregoriou and Rouah (2003) examine whether the percentage changes in the NAVs of CTAs follow random walks. There is little information on the long-term diligence of these funds (Edwards and Ma. especially during severe bear markets. Spurgin and McCarthy (1996) observe that performance persistence is virtually non-existent between 1987 and 1995. Regarding performance. 1996). Spurgin and Georgiev (2001). Irwin. CTAs are known regularly to short stock markets.33 and that it was –0. 1994. the results are mitigated even though Edwards and Caglayan (2001) conclude that. Zulauf and Ward. CTAs provide greater downside protection than hedge funds. and have higher returns along with a negative correlation with stock returns in bear markets.

their performance is generally inferior to that of hedge funds.months. Capocci (2004b) proves that there is persistence in CTA returns for badly performing funds. during bull markets. 13 . that one of the drawbacks of CTAs is that. Finally. which tend to continue to significantly underperform their peers. Brorsen and Townsend (2002) have shown that a minimal amount of performance persistence is found in CTAs and that some advantages might exist in selecting CTAs based on past performance. when a long time series of data is available and accurate methods are used. however. Georgiev (2001) underlines.

Because of their late entry to this field. 14 .931 defunct). There are several hedge fund databases available but only three of them have more than ten years of actual data collection experience: the Centre for International Securities and Derivatives Markets (CISDM) at the University of Massachusetts in Amherst. net assets value.699 defunct).158 funds (2. As of December 2004. HFR had 5. Only few of them report to more than one database.219 defunct).The Data Issue The main issue with hedge fund analysis is access to databases and the quality of data2. Hedge Fund Research (HFR) in Chicago. and investment styles across the two databases. As stated by Liang (2000) and Fung and Hsieh (2006) many hedge funds report to only a single database.246 funds (1. There are four other entrants to this field – The Barclay Group (Barclays). inception date.939 live and 2.431 live and 1. and Lipper TASS (TASS). 2 See for example Fung & Hsieh (2002b). Morgan Stanley Capital International (MSCI). Eureka Hedge. management fee. incentive fee. Liang (2000) reports that not only most funds do not report to the two databases he compared but moreover that there are significant differences in returns. and CISDM had 3. and Standard and Poors (S&P). their data were largely from reconstructed history rather than real-time collection of hedge fund performance. TASS had 4.130 funds (2.315 live and 1.

The period covered goes from 1994/2000 for the shorter one to 1993/2003 for the longest (without considering sub-period analysis). We have done for our extended multi-factor performance decomposition model. the overlap between the databases is very low. The Venn diagram divides the global hedge fund universe composed of five of the main hedge fund databases.In Figure 2. Table 2 indicates that we use between 634 and 4476 individual funds over the five studies of this Thesis and between 347 and 2011 fund of funds. The databases used and their characteristics are reported in Table 2. As shown in Figure 2. CISDM and Barclays (together and/or individually). Our results remain consistent independently of the database used. We first apply it to a combination of the HFR and CISDM databases. we apply the adapted version to CISDM alone before using a combination of the CISDM and the Barclay databases. In this Thesis we use HFR. generalization based on a single database may not be true for the entire hedge fund industry since any database only represent part of the industry. Moreover. Fung and Hsieh (2006) compare the HFR. TASS and CISDM databases. 15 . Then. This indicates that results obtained when performing an analysis based on a specific database may be different if another database is used. Since most funds do not report to all the existing databases it in interesting to apply the same methodology to different databases in order to be sure that the results obtained do not depend on the particular database used.

Brown. Voluntary participation means that only a portion of the universe of hedge funds is observable.4% lower over the period 1994–1998. Edwards & Caglayan (2001) use the same indirect approach of eliminating 12 months of returns from the MAR database and find that the average annual returns of hedge funds in the first year are 1. The backfill bias on equity market data is commonly calculated by an indirect approach. Ackermann et al. and all used indirect approaches. This means that funds tend to report to databases only when their performance have been good and may stop reporting once the performance becomes less attractive.g. because they find a median 343 day incubation period. Fung & Hsieh (2000) calculate the backfill bias for the TASS database over the period 1994 to 1998. and Edwards & Caglayan (2001) addressed the backfill bias for hedge funds in different periods for different databases. (1999) eliminate two years and find an average annual bias of 0.17% higher than the annual returns in subsequent years.The second important aspect regarding the quality of hedge fund data is the presence of biases in databases.4% for the TASS database over the period 1994–1998. this indirect approach is eliminating the first two years of reported data. Fama & French (1993). Goetzmann & Park (1998) use the method of Park to estimate an instant history of 15 months for the TASS database.5% for MAR and HFR database funds with different sample periods ending in 1995. As stated by Posthuma and van der Sluis (2003). This effect leads to a bias in database that is called instant return history bias or the backfill bias. First. The lasting mean performance was 1. They eliminate the first 12 months of returns. leading to a backfill bias of 1. Fung & Hsieh (2000). We estimate it by calculating the mean return of a portfolio investing in all funds and then we 16 . Ackermann et al. hedge funds report their performance to hedge fund database providers on a voluntary basis and a result in statistical sampling theory is that voluntary participation can lead to sampling biases. see e. (1999).

Figure 2: Hedge fund database universe repartition Source: Fung & Hsieh (2006) 17 . We then compare the difference in performance.make the same estimation by leaving the first 12. 24. 36 & 60 first returns of each fund.

HFR = Hedge Fund Research. Inc. . CISDM = Center for International Securities Derivatives Markets. Funds of funds = funds of hedge funds and analysis period = period under review for the corresponding study. Nb of strat = number of strategies used. Percentage dissolved = percentage of funds in the database that stop reporting to it before the end of the period under review. Hedge funds = number of individual hedge funds.Table 2: Database Comparison Data provider Number of strategies Hedge funds Percentage dissolved Funds of funds Percentage dissolved Analysis period An Analysis of Hedge Fund Performance HFR/ CISDM 28 2449 28% 347 34% 1994/2000 Hedge Fund Performance and Persistence in Bull and Bear Markets CISDM 16 2247 47% 647 33% 1994/2002 The Sustainability of Hedge Fund Performance CISDM/ Barclays 16 3060 60% 907 72% 1994/2002 The Neutrality of Market Neutral Funds CISDM 4 634 37% n/a n/a 1993/2002 Diversifying using Hedge Funds CISDM 16 4476 46% 2011 40% 1993/2003 This Table reports the comparison of the databases used over this PhD Thesis.

5% (Fung & Hsieh. The information that they use is information from TASS and it is mainly based on qualitative information from TASS employee. Funds disappearing from database tend to have poorer performance than existing funds. they decide to eliminate the last month of existence of any fund by 50%. We estimate these biases for our databases and the results are reported in Table 3. 1998) & 3% (Liang. Survivorship bias is calculated as the performance difference between surviving funds and all funds in the dataset. Not taking these funds into account leads to survivorship bias. Our estimation of backfill bias lies between 1. the period they analyse starts in 1996 whereas ours (and many others estimations) is based on data starting in 1993 or 1994. Posthuma and van der Sluis (2003) use a direct method of examining the backfill bias. CISDM and Barclays whereas Posthuma and van der Sluis (2003) has access to TASS. Third.3% when the 12 first months of existence of each fund is removed to estimations between 2% and 2.Another very important bias is the survivorship bias. First. there may be differences in statistics of the different databases used. Survivorship is an issue in hedge fund analysis and this bias is estimated between 1. The results indicate that funds outperform over their first months of existence. Our estimations are in line with those of other studies even if Posthuma and van der Sluis (2003) recently estimate that the magnitude of the overall backfill bias is about 4% per annum on average. Instead of eliminating the same average or median incubation period for all funds. 2001).2% and 1.2% when more months are removed. since most hedge funds still report to one database only. Second. We use together or separately HFR. Finally. the direct method eliminates the individual incubation period per fund. 19 . The difference can have several reasons. this rule will clearly impact estimation.

6% 2.2% 1.68% n/a n/a n/a This table reports the estimated survivorship bias and instant return history bias as estimated in four of our studies.3% 1. 20 .08% n/a n/a n/a Neutrality of Mkt Ntl Funds 1. Not taking these funds into account lead to a survivorship bias. 1.51% 1.8% 2. Instant return history bias comes from the fact that hedge funds report their performance to hedge fund database providers on a voluntary basis and a result in statistical sampling theory is that voluntary participation can lead to sampling biases. Througout this Thesis.22% 1. and Pers. Without considering what methodology is the best. the differences between Posthuma and van der Sluis (2003) methodology and ours explain the difference in estimations. Survivorship bias comes from the fact that funds disappearing from database tend to have a worse performance than existing funds. we prefer to use the indirect approach with no qualitative influence.Table 3: Backfill bias and survivorship bias estimation Survivorship bias Backfill bias (12m) Backfill bias (24m) Backfill bias (36m) Analysis of HF Perf.0% HF Perf. 1. in B&B Mkt 1.2% Sustainability of HF Perf.

There are still good managers building a track record before actively marketing the fund but that will be open to new investors in case of demand. The fact that many funds still report their performance to one database only comforts us in this idea. These self-excluded funds may have better performance than the average hedge fund. hedge funds may not be participating in a database because they are not looking to attract new investors. Practically. 1999. Thus. We are convinced that this bias may be quite low. On the one hand. Hedge funds may enter a database on a voluntary basis. 21 . Therefore. Several authors (see Ackermann et al. 2002b) argue that this bias can be counterbalanced by good managers that stop reporting to databases when they close the funds to new investors. presumably. The use of data is to try to represent the hedge fund industry as closely to the reality as possible but the hedge fund industry is not limited to funds reporting to databases. hedge funds in a database tend to have better performance than those that are excluded. the net effect of selection bias on the returns of hedge funds in a database is ambiguous. Fung and Hsieh. there is no way to mitigate this bias and we have to keep in mind that this bias may be present. only those funds that have good performance and are looking to attract new investors want to be included in a database. On the other hand.A last element to stress is that a fund can be accessible to investors even if its returns are not reported in any database. Selection bias manifests itself in two basic ways.

He finds that the survivorship bias exceeds 2% per year in the TASS database. Such values are in the lower end of recent estimation. The difference can be explained by several factors. First. This element could explain an increase in the percentage of dissolved funds and a higher survivorship bias after 2002 or 2003. McEnally & Ravenscraft (1999) suggest that two biases.22% and 1. the survivorship bias and the self-selection bias. while the HFR database survivorship bias equals 0. Until recently. As any other these funds get listed in databases but as they do not offer attractive returns. the three parts of this Thesis are based on individual hedge fund data.Our estimation of survivorship bias lies between 1. (1999) report a bias of 3% for offshore hedge funds per year. only the best managers managed hedge funds. 2001. Brown et al.6%. 2003). The best managers continue to launch funds. more and more players entered the industry.68%. Ackermann. most of them are dissolved after two or three years of data. Amenc and Martellini. which is consistent with the higher drop out rate in the TASS database. Finally. but less experienced individual are also attracted by the high fee levels. The use of indices to analyse the hedge fund world may also lead to measurement problems. Fung & Hsieh (2000) use the TASS database and calculate the annual survivorship bias to be 3% with a 15% drop out rate. 22 . Liang (2000) examines this survivorship bias in hedge fund returns by comparing the TASS and the HFR database. offset each other. There are many hedge fund indices providers and most of them are reported in Table 4 with their main characteristics. several other researchers have used and/or looked at hedge fund indices (see Brooks and Kat. most of our studies analyses period ending by the end of 2002. Since the demand for hedge funds exploded after the internet bubble starting in 2000.

If the assets held by these managers make up a large portion of the assets in the hedge fund universe. equally weighted or asset weighted. 23 . constructing indices based on hedge fund data will result in biases in the index. i. since the quality of hedge fund data is poor.e. Some hedge fund indices use dollars under management as the weighting for the individual components. 3 See Liew (2003) and Amenc and Martellini (2003) for more information on the subject.The literature on the subject report five main potential problems with using hedge fund indices3. As a result the returns of hedge fund indices may not be meaningful. there is a debate on how indices should be constructed. then hedge fund indices will under-represent the returns of the universe. Second. Third. some of the best hedge fund managers do not disclose fund information to the public. First.

indices.3 1. since many hedge fund managers have managed accounts and on/off-shore vehicles.net ljh. Nb of strategies = number of strategies as defined by the index provider. as shown by Amenc an Martellini (2003).com vanhedge.com hedgeindex. hedge funds may have different levels of leverage and may vary their leverage employed through time.8 800 1. Fifth. Fourth.1 340 60 750 450 1. Moreover.4 NA Website eacmalternative.com hedgefund.com This Table reports a comparison between the major hedge fund indices available. Standardizing for leverage is problematic in index construction. suffer from the problem that they overweight markets that have had strong historical performance.Table 4: Hedge Fund Indices Comparison Nb of Strategies 13 15 9 5 12 22 33 16 15 13 8 Providers EACM HFR CSFB/Tremont Zürich Capital Van Hedge Hennessee Group Hedgefund.com magnum. Launch = launch date of the indices for the corresponding index provider.net LJH Global Investments MAR Altvest Magnum Launch 1996 1994 1999 2001 1995 1992 1979 1992 1990 2000 1994 Nb of Funds 100 1. Nb of funds = estimated num In practice this Figure 2 is difficult to determine. there are significant differences in return distribution for the same strategies4.com zcmgroup.com marhedge.com hedgefnd.com altvest. 4 Amenc and Martellini (2003) analysed the largest differences between the same indices of the hedge fund indices providers on a monthly basis and found differences up to 22% for a single month. 24 .com hfr.

All these aspects mentioned above mean that the real hedge fund world can be different from the one analysed in academic papers and that general conclusion cannot always be applied to particular hedge fund strategies or individual managers. 25 .

They provide investors with diversification across manager styles and professional oversight of fund operations that can provide the necessary degree of due diligence. Because of these. many funds of funds hold shares in hedge funds closed to new investment allowing smaller investors the access to those managers. The additional fees has two main impacts. In case of bad 26 . Funds of funds have several advantages in comparison to individual hedge funds. other may be down leading to a complex distribution of returns for the fund of funds. In addition. In addition. First. Several reasons can explain this effect: high minimum wealth levels. the fund of funds fee structure over the underlying fund one generally cut the profits by a performance fee once they reach a certain level (LIBOR for example) cutting the upside of the portfolio while on the downside there is no performance fees. sophisticated investor requirements or complexity of the strategy applied. the return distribution of fund of funds will be complex because strongly impacted by the level of fees and more importantly by the distribution of these fees. When some underlying funds will offer positive returns.Investing in Hedge Funds Despite the growing interest in hedge funds. one major drawback: the additional fees. The second and obvious element is that the additional fee will lower the performance of the fund of funds that can only be as attractive as the one of the hedge fund industry as a whole if the fund of funds managers make a good selection of underlying managers. There is however. it is difficult for many individual and institutional investors to participate in this area of the market. funds of funds constitute the only way of investing in hedge funds for many investors.

the final outcome for investors will be less attractive than stated in academic studies. Sharpe ratio and information ratio and find that a portion of the fund of funds available can be beaten through a simple selection strategy based on simple statistics. Care has to be given when trying to profit from academic conclusion based on individual hedge funds and these cannot always be applicable for funds of hedge funds. (2005) compare the performance of funds of funds with the one of portfolio constructed on the basis of alpha. Several studies analyse the impact of fees in funds of funds. (2005) concluded that funds of funds offered a relatively poor historical performance relative to the hedge funds in which they invest.choice. Brown et al. These elements stress out that there is a difference between conclusion of academic paper and the reality for final investors. They explain the poor performance of funds of funds by the performance fees charges by underlying funds when they offer a positive performance even if the fund of funds as a whole is negative. Gregoriou et al. 27 .

.

Abstract Part One: The Persistence in Hedge Fund Performance As stated in the first part of the introduction. testing and improving a performance analysis model in order to understand hedge fund performance on the one hand. 29 . Hedge Fund Performance and Persistence in Bull and Bear Markets and Sustainability in Hedge Fund Performance: New Insights). they remain investment funds. a style factor that takes into account the difference in performance between growth and value players 6 Carhart (1997) extension added a momentum factor that takes into account the fact that certain managers favour previously well performing stocks in their portfolio. We achieved this objective in three complementary studies (Analysis of Hedge Fund Performance. As such. Fama and French (1993)5 and Carhart (1997)6 models. we saw a model coming from the mutual fund literature as a good basis to build a new performance decomposition model specific to hedge funds. This involve developing. the first objective of this Thesis is to understand hedge fund managers and to explain how they create alpha over time. The basis of Part 1 is study 1 that aims at answering to one question: What factors might explain hedge fund returns? We base our multi-factor performance decomposition model on models that are used in the mutual fund literature for years. 5 Fama and French’s (1993) model includes the following: a size factor that takes into account the difference in performance between small and large companies. while on the other developing and adapting a methodology to determine whether there is any persistence in hedge fund returns. the securities they use and the freedom they have in their management. Even if hedge funds are different from mutual funds by the strategy they apply.

the model with the stronger power of explanation is model 3 that has a relatively limited number of factors but that covers almost all the aspects of hedge fund investing and that enable us to reach very high R². the correlation between factors increased leading to a risk of multicolinearity. From model 1 to model 2. Second. the second 10 and the one used in the last part of the analysis 10 and 14. First. we re-adjust the factor used in the model to integrate a high yield factor and a mortgage backed securities factor to take into consideration the significant increase in the number of funds exposed to the high yield market and to determine if the exposure of fixed income funds to the mortgage was high or not7. the first model has 11 factors. The high yield factor finally helps but the mortgage factor does not. some factors didn’t help explaining hedge fund performance. Our model evolves over time for several reasons. As we will see in the papers. 7 There has been a major move in the mortgage market in September 2002 and several hedge funds have faced strong losses. high yield and default). some factors were added as they seem to help decomposing the performance. we add several factors that take into account the fact that hedge funds invest in non-US equities and in bonds (government. Finally.Since hedge funds do invest not only in US equities. 30 . corporate. in some cases. The models we developed over the three studies are described in Table 5. as well as commodities. As stated in Table 5.

Salomon World Government Bond Index. 31 . non-US stock market = MSCI World excluding US. Corhay & Hübner (2005) X X X X n/a X X n/a X X n/a X X X n/a n/a 10 Model 3 Capocci (2006)-1 X X X X n/a X X n/a X X n/a X n/a X X n/a 10 Model 4 Capocci (2006)-2 X X X X n/a X X n/a X X n/a X n/a X X X 14 This Table reports a comparison of the multi-factor performance models used in this Thesis. GSCI = Goldman Sach Commodity Index. Lehman BAA = the Lehman BAA Corporate Bond Index. EMBI = JP Morgan Emerging Market Bond Index. and at-themoney (ATMP) and out-of-the money (OTMP) European put option factors. Option factors = Agarwal and Naik (2004) at-the-money (ATMC). Size = the size market risk premium of Fama and French (1993). Currency = the Federal Reserve Bank Trade Weighted Dollar Index. High yield = Lehman High Yield Bond Index. style = the style market risk premium of Fama and French (1993). Wd Gov Bond = JPMorgan world government bond index. US bonds = the Lehman Aggregate US Bond Index. momentum = Carhart’s (1997) "momentum" factor. Mortgage = Lehman MortgageBacked Securities Index. international style = the international style market risk premium of Fama and French (1996). out-of-the money (OTMC) European call option factors.Table 5: Multi-factor Performance Decomposition Model Model 1 Capocci & Hübner (2004) Alpha US Stock Market Size Style International Style Momentum Non-US Stock Market US Bond Wd Gov Bond EMBI Lehman BAA High Yield Mortgage GSCI Currency Option Factors Number of factors X X X X X X X X X X X n/a n/a X n/a n/a 11 Model 2 Capocci.

in all cases. we looked at whether there was a repetitive way to isolate it over time. Once we discussed that hedge fund strategies do significantly outperform classical markets over time. Our results indicate that hedge funds tend to outperform during bull market conditions but that this outperformance is no longer significant under bear market conditions. These results were not due to the lack of power of the model since. bear market conditions and over a full cycle. The only exception was market neutral strategies. Our results indicate that most hedge fund strategies do offer significant alpha over a long period of time. which needed further analysis that is reported later in this Thesis. we analysed the persistence of this performance. the objective is to explain long term hedge fund performance in order to determine whether some hedge fund strategies significantly outperform classical markets over time. At this level.” 32 . The next logical step was to perform the same analysis under various market conditions and we did that in our second study. the adjusted r-squared were very high. More precisely. we reach the second important basic concept also used in the second study: the decile classification of Carhart (1997). that is. we analyse hedge fund performance and persistence in performance over bull market conditions. The hypothesis that hedge funds with a superior average return in this period will also have a superior average return in the next period is called the hypothesis of persistence in performance. As we state in the study: “Active hedge fund selection strategies could increase the expected return on a portfolio if hedge fund performance is predictable.Over our three studies on the persistence in hedge fund performance.

but that offer relatively consistent returns over time. The only issue is that these funds tend to be classified in the middle decile portfolios. The portfolios are held until the following January and then rebalanced again. It needs deeper analysis over a shorter period of time. nor the worst. Our results clearly indicates that measures incorporating volatility display a very strong ability to assist investors in creating alpha and in consistently and significantly outperforming classical indices. We tested several ways of classifying funds on the basis of their past performance: returns. low volatility funds were the ones offering significant alpha. This result was the first important conclusion of our thesis. only negative persistence can be found among the past losers.Carhart’s methodology is relatively straightforward to understand. all funds are ranked in 10 equally weighted portfolios based on their previous year’s return. This is exactly what we did in our third study. Persistence analysis indicates that most of the predictability of superior performance is found in bull market conditions (prior to March 2000). In bear market conditions. In both studies. is mostly located among medium performers. Our results confirm several previous studies that found that persistence. beta. 33 . We checked the robustness of our results by performing the same analysis over sub-periods. The combination of our multi-factor model with this methodology enables us to determine whether there is persistence in hedge fund returns. Each year. skewness and kurtosis. if any. Sharpe ratio. alpha. which was done in the second study. volatility. This led us to the conclusion that we needed another way of classifying hedge funds in the persistence analysis in order to be able to clearly identify the funds that significantly and consistently outperformed. Our results indicate that there is some proofs of persistence for low volatility funds that tend neither to be the best performers. suggesting that bad performance has probably been the decisive factor for hedge funds mortality.

beta exposure. to a lesser extent. 34 . systematic way of creating pure alpha using a simple classification methodology based on basic statistics: risk-return trade-off measures (Sharpe score). We report the specificity. which appear to be better and more stable ways of classifying hedge funds in order to detect persistency in the returns.during bull and bear market conditions (defined as the up and down months of the S&P 500 and as consecutive bull and bear market periods) and by changing the month of classification (June instead of January). We found a consistent. pure volatility measures (standard deviation) and. Funds offering stable returns. with limited volatility and/or with limited exposure to the equity market consistently and significantly outperformed equity and bond markets. objectives and main conclusions of each study of Part 1 in Table 6.

Objectives and Conclusions Objective 1 Conclusion 1 Objective 2 Conclusion 2 An Analysis of Hedge Fund Performance .Determine if hedge fund strategies significantly and persistently outperform classical markets in bull and/or bear market conditions .Determine if hedge fund Hedge Fund Performance strategies significantly and Persistence in Bull outperform classical markets and Bear Markets in bull and/or bear market conditions . For each study we report the main specificities of the study aalong with its first and second objective and conclusions. standard deviation) This Table reports the specificities.Find a systematic way of .No significant outperformance in bear market conditions but for market neutral funds .Hedge funds tend to outperform during bull market conditions (not significantly in bear markets) . Hedge Fund Performance and Persistence in Bull and Bear Markets and the Sustainability of Hedge Fund Performance: new insights).Determine if hedge fund do offer significant alpha strategies significantly over the long term despite a outperform classical markets high R² .There is no proof of persistence for hedge funds but low volatile funds that tend to neither be the best performers nor the worst . objectives and conclusion of the studies grouped in Part 1: The Persistence in Hedge Fund Performance.Most hedge fund strategies buying hedge funds in order do offer significant alpha to significantly and over the long term despite a persistently outperform high R² classical markets .Most hedge fund strategies .Determine if hedge fund strategies significantly The Sustainability of outperform classical markets Hedge Fund Performance using several risk-adjusted measures .Determine if hedge fund strategies significantly and persistently outperform classical markets . . Part 1 countains three studies (An Analysis of Hedge Fund Performance.Table 6: Persistence Analysis Studies Specificities.Systematic outperformance of hedge fund portfolios invested in previous year' low volatile funds (measured by Sharpe score.

The study analyses a complete cycle as well as sub-periods (bull and bear market conditions). the funds that are significantly exposed can bias the results because a) the bulk of the funds are not significantly exposed to the market. Second. confirming that low volatility funds emerge over time. Market neutral funds represent a large part of the industry. This can be explained by two reasons.Abstract Part Two: The Neutrality of Market Neutral Funds Self-defined market neutral funds significantly and consistently outperformed the classical market over time. the aggregation of funds in indices leads to an increase in exposure to the equity market. Our results confirm that the betas obtained were low in absolute terms even though they were all significantly positive. At the individual fund level. The objective of this second part is clear: to analyse the exposure to the equity market of market neutral funds and to explore how to isolate funds that consistently outperform. We perform an analysis at the individual fund level in order to obtain this result because market neutral index analysis lead to controversial results. This result also stresses the importance of considering individual funds when performing an empirical analysis. only one third of the funds were. around 28% of our database used in the second Study of Part 1: Hedge Fund Performance and Persistence in Bull and Bear Markets. one third of the funds are significantly positively exposed to the market. and b) only five percent of the funds are significantly negatively exposed to the market. This result requires further analysis. 36 . First. while two thirds of the alphas are significantly positive. The decile analysis indicates that the more volatile funds (top and worst performing funds) have the highest market exposure.

most poor performing market neutral funds out-performed the equity market without being significantly exposed to the market.Real market neutral significantly exposed market exposure of neutral funds funds outperform dirty to the equity market market neutral funds outperform their market neutral funds (low in absolute hig/low beta peers terms) This Table reports the specificities. Our analysis leads to the conclusion that most market neutral funds are not significantly exposed to the equity market. but tend to be more exposed during bear markets than during bull markets without being negatively impacted. during the bullish period. Part 2 countains one study (The Neutrality of Market Neutral Funds). We report the main specificities of the study and its The sub-period analysis also reports two interesting results. Second.Determine if tend to be . 37 . all but the best performing deciles are significantly exposed to the market. We report the specificity. objectives and main conclusions of part 2 in Table 7.Market neutral funds . but they all (except the best performing funds) created significant alpha. First.Table 7: Neutrality of Market Neutral Fund Specificities. However. during the bear market.Determine the high/low beta market . objectives and conclusion of the study of Part 2: An Analysis of Hedge Fund’s Market Exposure. during the bearish period. but the best performing funds significantly out-perform the equity market and offer significantly positive returns. Objectives and Conclusions Objective 1 Conclusion 1 Objective 2 Conclusion 2 Neutrality of Mkt Ntl Funds . no index is significantly exposed to the market.

Hedge funds exhibit abnormal returns. This is the basic reason why traditional tools like the mean-variance efficient frontier analysis should not be used for their analysis. The only difference is that the risk factor is no more only defined by the standard deviation of the returns. the asymmetry of the return distribution (skewness) and the presence of fat tails (kurtosis). This new tool has the same underlying idea as the classical efficient frontier and can be illustrated the same way while taking into account more sophisticated statistics.Abstract Part Three: Hedge Funds as Diversification Tools Hedge fund performance decomposition and strategy analysis were the first two aims of this doctoral thesis. there is a significant improvement for diversified portfolios (20 to 80% allocated to the risky 38 . This decomposition enables us to define a new and extended risk measures that we use in a classical risk-return framework. In order to complete these analyses. Our results indicate that directional hedge funds should be considered separately from undirectional hedge funds and fund of hedge funds. we analysed in our third and final part the impact of inserting hedge funds into a classical portfolio of stock and bond mutual funds. the volatility. We decompose this function and take into account the mean return. Our methodology is based on the Taylor’s extension of the linex utility function developed by Bell (1988). When more than 20% is allocated to directional hedge funds. In this paper we develop the idea of an adapted capital market line in an extended risk-return framework that includes not only volatility as a measure of risk but also higher moments. Adding a small allocation to directional hedge funds does not significantly change the risk-return profile offered by the global portfolio.

undirectional strategies do not help diversifying and reaching higher return levels. adding undirectional hedge funds or fund of funds to a classical portfolio enables investors to reach higher levels of returns for low and medium risk levels for allocation as low as 10% to hedge funds. Our results confirm that undirectional strategies and funds of funds are diversificating low risk profile investments and should be used as such. For high allocation to the risky asset. 39 . it determines if hedge funds must be added to the existing portfolio. The new adapted efficient frontier opens new doors for asset allocators. Moreover it helps to determine what hedge fund strategy should be favoured. Over a allocation of 50% to directional hedge funds offers significantly more attractive returnsin every case. objectives and main conclusions of part 3 in Table 8. We report the specificity. Based on the clients’ objective and the market conditions.asset). However.

Adding small allocation of undirecitonal hedge funds and funds of hedge funds significantly improve the profile This Table reports the specificities. .Taylor's expansion of Bell's . Part 3 countains one study (Diversifying using Hedge Funds: a utility based approach).Develop a methodology to utility function enables us to determine if a portfolio can be take skewness and kurtosis Diversifying using diversified with securities into account hedge funds displaying abnormal return .High allocation to directional hedge funds do significantly improve the profile . We report the main specificities of the study and its first and second objective and conclusions.Table 8: Hedge Funds as Diversification Tools: Specificities. Objectives and Conclusions Objective 1 Conclusion 1 Objective 2 Conclusion 2 .Determine if bond and/or equity investors should include hedge funds in their portfolio .The adapted Capital Market distribution characteristics Line and new efficient frontier complete the development . objectives and conclusion of the study of Part 3: Hedge Funds as Diversification Tools.

.

.

Part One: The Persistence in Hedge Fund Performance .

Journal Empirical of Finance 11/1. Maastricht University. Analysis of Hedge Fund Performance. 55-89 . Limburg Institute of Financial Economics.J. and Luxembourg School of Finance. Capocci. CAPOCCI HEC-ULG Management School – University of Liège (Belgium) Georges HÜBNER HEC-University of Liège. Daniel and Georges Hübner. January 2004.Analysis of Hedge Fund Performance Daniel P. University of Luxembourg.

including the Asian Crisis period.Analysis of Hedge Fund Performance Abstract Using one of the largest hedge fund databases ever used (2796 individual funds including 801 dissolved). 45 . and for all funds in general. indicates limited evidence of persistence in performance but not for extreme performers. The performance of hedge funds for several individual strategies and different sub-periods. This addition is particularly suitable for more than half of the hedge fund categories. including an extension of Carhart’s (1997) specification combined with the Fama and French (1998) and Agarwal and Naik (2004) models and a new factor that takes into account the fact that some hedge funds invest in emerging bond markets. we investigate hedge fund performance using various asset pricing models.

These funds. about 90% of hedge fund managers are based in the US. leverage or arbitrage on different markets. hedge funds justify an increased attention in financial press as well as in the academic world. combined with the models of Fama and French (1998) and Agarwal and Naik (2004) and with a factor. While the number of funds has more than doubled since the mid-nineties. and around 50% are smaller than $12 million.000 funds managing around $400 billion in capital. including an extension of Carhart’s (1997) model. Scientific literature on performance-evaluation yields controversial results. Currently. never previously used in this context. 1999). accounting for the fact that some hedge funds invest in emerging bond markets. are not legally defined but share some common characteristics: they use a broad range of instruments like short selling. This lack of consensus on the « right » model puts researchers in a quandary (Metrick. that have been existing for more than fifty years. around 80% of hedge funds are smaller than $100 million. This reflects the high number of recent entries. Hedge funds require high minimum investments and their access is limited to individual investors or to institutions with large financial resources. This study investigate hedge funds performance levels and persistence using various asset-pricing models. 9% in Europe.Analysis of Hedge Fund Performance I Introduction With almost 6. This analysis is carried out for different sub-periods including the Asian Crisis period and for several individual hedge funds strategies. 46 . derivatives. and 1% in Asia and elsewhere.

Therefore. inducing that managers of mutual funds have access to sufficient information to recover their costs. and Grinblatt and Titman (1989. In general. depending on whether they conclude or deny that mutual funds have significantly higher realized returns that those obtained by following passive strategies8. 47 . 8 See a. Ippolito (1989). Section 6 reports the performance of hedge funds. few performance studies have been carried out on hedge funds comparing to other investment tools like mutual funds. Gruber (1996) and Carhart (1997) for studies reaching the opposite conclusion.1 Performance Studies Despite the increasing interest that hedge funds have originated due to their recent development. Section 8 concludes the study. Section 7 documents and explains the persistence in hedge fund returns. II Literature Review 2.The rest of the study is organised as follow. Section 3 sets out the performance models we will use. This can partly be explained by their private characteristics and the difficulties encountered to have access to individual funds data. it is interesting to succinctly consider the results obtained in the main performance studies of mutual funds before introducing results of studies on hedge funds. Section 2 reviews some of the major mutual and hedge funds performance studies with a focus on the evolution in the models used. performance studies can mainly be classified in two categories. Malkiel (1995). Section 5 brings some insights on hedge fund performance. 1992) for contenders of superior performance of hedge funds. and Jensen (1969).o. The next Section provides a thorough description of our database. Lehmann and Modest (1987).

Brown and Goetzmann (1995). (1993). (1993). studies on hedge funds persistence in performance are less frequent. Nevertheless. Agarwal and Naik (2000) sustain that persistence in hedge funds performance exists. although lower than the market indices considered. and Zheng (1999) report a predictability in the mutual funds returns over a longer period of time. Hendricks et al.Among studies finding superior mutual funds performance. Sirri and Tufano (1998). Brown and al. (1993). as emphasised by Brown and al. numerous studies investigate further its persistence. Considering the recent interest for this sector. This issue of persistence in performance is particularly important in the case of hedge funds because. 48 . (1996). They also indicate that the returns in hedge funds are more volatile than both the returns of mutual funds and those of market indices. hedge funds experience an attrition rate much higher than mutual funds. On the other hand. On the one hand. (1997) demonstrate that the momentum effect in the share’s returns explain the hot hands effect detected by Hendricks. Grinblatt and Titman (1989. Ackermann and 9 This effect means that the securities held by funds that had better performance one year realize superior returns than other funds the following year. et al. (1999) and Liang (1999) who compare the performance of hedge funds to mutual funds and several indices find that hedge funds constantly obtain better performance than mutual funds. Goetzmann and Ibbotson (1994). It is worth noting that Carhart (1997) and Daniel and al. and Wermers (1996) show persistence in mutual funds performance for a short period (1 to 3 years). Ackermann and al. Ippolito (1989). Elton et al. 1992). and attribute it to hot hands9 or to common investment strategies. Elton et al. (1999) prove that offshore hedge funds display positive returns adjusted for risk but they attribute this performance to style effect and conclude that there is hardly any evidence of the existence of differential manager skills. (1999. 2001) and Liang (1999).

Amin and Kat (2001) find that stand-alone investment hedge funds do not offer a superior risk-return profile. Fama and French. earnings/price (Basu. Breeden. Several studies10 report that the cross-section of average returns on U. 1981). and Litzenberger (1989). 1983). Fung and Hsieh (1997) and Schneeweis and Spurgin (1998) prove that the insertion of hedge funds in a portfolio can significantly improve its risk-return profile thanks to their weak correlation with other financial securities. 10 See e.Ravenscraft (1998) emphasise that the stronger legal limitations for mutual funds than for hedge funds hinder their performance. (2001).S. common stocks shows little relation to the betas of the Sharpe (1964)Lintner (1965) CAPM or the Breeden (1979) ICAPM. but that a great majority of funds classified as inefficient on a stand-alone basis are able to produce an efficient payoff profile when mixed with the S&P 500. Taking all these results into account. leverage (Bhetari. 1985. They obtain the best results when 10-20% of the portfolio value is invested in hedge funds. 49 .2 Evolution in Performance Measurement In the eighties. 1988). This low correlation is also emphasised by Liang (1999) and Agarwal and Naik (2000). hedge funds seem a good investment tool. hedge funds showing good performance in the first part of the year reduce the volatility of their portfolio in the second half of the year. 2. Instead. Reinganum (1981).g. Jegadeesh and Lakonishok (1996). performance measures based on the CAPM. Gibbons. like Jensen’s alpha (1968) and their extensions were commonly used in performance evaluation. According to Brown and al. book-to-market (Rosenberg et al. The recent interest in multi-factor models primarily comes from the literature on the cross-sectional variations in stock return. these authors identify other factors like the size of the company (Banz. Fama and French (1996) and Chan.

different models have also been used in performance evaluation. 1979. 1993. and the international model of Fama and French (1998). the 4-factor model from Carhart (1997). Fung and Hsieh (1997) extend Sharpe’s (1992) asset class factor model and find five dominant investment styles in hedge funds. For the non-parametric method. (1999) and Ackermann and al. the Grinblatt and Titman (1989) 8-factor model and the Fama and French (1993) 3-factor model. (1999) use a single factor model and focus only on total risk. dividend yield (Litzenberger and Ramaswamy. Brown and al. In addition. Schneeweis and Spurgin (1998) also use style analysis based on a multi-factor approach. the asset class factor model from Sharpe (1992). recent studies have cast doubt on the usefulness of these new models. Liang (1999) uses the 11 See Allen and Soucik (2000) for a description of the major models used in mutual funds performance studies. 50 .11 However. Carhart (1997) develops his own 4-factor model that proves to be superior to the classical CAPM. Carhart. 1995) that have reliable power to explain the cross-section of average returns. the 3-factor model from Fama and French (1993). In hedge funds literature. Kothari and Warner (2001) show that the Fama and French (1993) 3-factor model provides better results than the classical CAPM. Their parametric method regresses alphas (or appraisal ratios) on their lags. In an early study.1992). Subsequent multi-factor models include the 8-factor model developed by Grinblatt and Titman (1994). Agarwal and Naik (2000) use regression-based (parametric) and contingency-tablebased (non-parametric) methods. but document that it detects significant abnormal results (including timing) when none really exists. 1982) and more recently the momentum effect (Jegadeesh and Titman. they construct a contingency table of winners and losers depending on the alpha.

regressions based on fund characteristics. but there exists no unanimously accepted model. rather than simply use the CAPM. it is preferable to use several specifications in order to compare the results obtained.extension of Fung and Hsieh (1997) model. Agarwal (2001) uses a model consisting of trading strategy factors and location factors to explain the variation in hedge funds returns over time. Therefore. These results suggest that it is necessary to realize performance studies based on multi-factor models. 51 . and classical measure like the Sharpe ratio. Agarwal and Naik (2004) propose a general asset class factor model comprising of excess returns on passive option-based strategies and on buy-and-hold strategies to benchmark the performance of hedge funds.

2.1 The Capital Asset Pricing Model The first performance model we use is a single index model based on the classical CAPM developed by Sharpe (1964) and Lintner (1965). T (1) where RPt = return of fund P in month t.. our study of hedge funds performance starts with the CAPM. respectively. 3. 12 Following Bams and Otten (2002) we do not consider the Sharpe’s (1992) asset class factor model which is an asset allocation model and not an asset evaluation model. RFt = risk-free return on month t.. αP and βP are the intercept and the slope of the regression. εPt = error term. 1998) and the Carhart (1997) model because they are not dominated by any other model in the mutual funds performance literature. The basic multi-factor specifications are the Fama and French (1993) 3-factor model and its international version of 1998 (Fama and French...12 Finally. 52 . Its equation to estimate is the following: RPt − RFt = α P + β P (RMt − RFt ) + ε Pt t = 1. we construct a multifactor model that extends the Carhart (1997) model by combining it with factors proposed in Fama and French (1998) model and Agarwal and Naik (2000) and by adding an additional factor.III Performance Measurement Models For comparison purposes. RMt = return of the market portfolio on month t.

The formula is the following: 13 See Fama and French (1993) for a precise description of the construction of SMBt and HMLt. αp commonly called Jensen’s alpha (1968) is usually interpreted as a measure of out. It takes the size and the book-to-market ratio of the firms into account.2 The 3-factor Model of Fama and French (1993) and its international version of Fama and French (1998) The Fama and French (1993) 3-factor model is estimated from an expected form of the CAPM regression. In the international version of the model.2.T (2) where SMBt = the factor-mimicking portfolio for size (‘small minus big’) and HMLt = the factor-mimicking portfolio for book-to-market equity (‘high minus low’).The intercept of this equation.. It is estimated from the following extension of the CAPM regression: RPt − RFt =α P + β P1(RMt − RFt )+ β P2SMBt + β P3HMLt +ε Pt t =1... and Far East) countries and several emerging markets and propose an international factor mimicking for book-to-market equity (HML). 53 .13 These factors aim at isolating the firm-specific components of returns. Australia. Fama and French (1998) consider twelve major EAFE (Europe. 3.or under-performance relative to the market proxy used..

Grinblatt. According to Fama and French (1998).14 14 For a description of the construction of PR1YR see Carhart (1997).. but a one-state-variable international ICAPM that explains returns with the global market return and a risk factor for relative distress captures the value premium in country and global returns. This model is estimated with the following regression: RPt − RFt =α P + β P1(RMt − RFt )+ β P2 SMB t + β P3 HML t + β P4 PR 1 YRt +ε Pt t =1.3 The 4-Factor Model of Carhart (1997) Carhart’s (1997) 4-factor model is an extension of the Fama and French (1993) factor model.. Titman and Wermers (1995) define this effect as buying stocks that were past winners and selling past losers..2.. It takes into account size and book-to-market ratio.. 3.2.T (3) where IHMLt = an international version of HMLt. the international CAPM cannot explain the value premium in international returns... but also an additional factor for the momentum effect. 54 .RPt − RFt =α P + β P1(RMt − RFt )+ β P2 IHMLt +ε Pt t =1..T (4) Where PR1YRt = the factor-mimicking portfolio for the momentum effect.

the market proxy used is the Russel 3000 that represents over 95% of investable US equity market. and the model used by Agarwal and Naik (2004) and Agarwal (2001). the coefficients of four zero investment factor mimicking portfolios are appropriate measures of multidimensional systematic risk. the international model of Fama and French (1998).4 An Extended Multi-Factor Model In order to take into account the different characteristics of the hedge fund industry. 3. we implement a combination and an extension of Carhart’s (1997) 4-factor model. Fama and French (1998) international value and Carhart’s (1997) momentum factor. a default factor (Lehman BAA corporate bond index) as introduced by Agarwal and Naik (2004). It identifies a matching passive portfolio return for each fund return. Salomon world government bond index. three factors to take into account the fact that hedge funds invest in US and foreign bond indices (Lehman US aggregate bond index. This model contains the zero investment strategies representing Fama and French’s (1993) size and value.As stressed by Daniel et al. in the absence of stock selection or timing abilities. (1997). this model assumes that. 55 . and JP Morgan Emerging Market Bond Index) and finally a commodity factor (GSCI Commodity Index). Beyond the combination of existing models. a factor for non-US equities investing funds (MSCI World excluding US). In order account for the fact that hedge funds invest in a wide range of equities including small and large companies. the originality of this model is to feature a factor that takes into account the fact that hedge funds may invest in bonds on emerging markets.

SWGBIt = return on the Salomon World Government Bond Index. LEHBAAt = return of the Lehman BAA Corporate Bond Index. the Salomon Brothers Government and Corporate Bond Index. RPt − RFt =α P + β P1(RMt − RFt )+ β P2 SMB t + β P3HMLt + β P4 IHML t + β P4 PR 1YRt + β P5(MSWXUS t − RFt )+ β P7(LAUSBI t − RFt )+ β P8(SWGBI t − RFt ) + β P9(JPMEMBI t − RFt )+ β P10(LEHBAA t − RFt )+ β P11(GSCI t − RFt )+ε Pt (5) where RMt = return on the Russel 3000 index. Their high colinearity with other factors lead us not to test these indices further. MSWXUSt = return of the MSCI World Index excluding US. 56 . JPMEMBIt = return of the JP Morgan emerging market Bond Index. Following Agarwal (2001) we choose the Goldman Sachs Commodity index instead of a Gold index used by Fung and Hsieh (1997) as the former indicates better exposure of hedge funds in commodities especially considering the fact that hedge funds may not be investing solely in gold among commodities. and the Lehman High Yield Bond Index.Note that Agarwal and Naik (2004) and Agarwal (2001) take several additional factors such as the MSCI Emerging Markets. GSCIt = return of the Goldman Sachs Commodity Index. LAUSBIt = return on the Lehman Aggregate US Bond Index. Its components are weighted according to their impact on production in the world economy.

and ‘TASS Management’ (TASS. for ‘allowed’ persons depending on the country in which the fund wants to find investors. available to the qualifying public. 2001). and Agarwal (2001). MAR defines 9 strategies along with 15 sub-strategies. and Brown and Goetzmann (2001). start and ending date. Schneeweis and Spurgin (1998). in turn. 1400 funds). Agarwal and Naik (2000. Fortunately. The three databases have never been used together in a study. they record other useful information such as company name.’ (HFR. These databases are the most used in academic and commercial hedge fund studies. by Fung and Hsieh (1997). among others.IV Data 4. Data vendors do not only collect performance data. Liang (1999). Liang (2000) uses a combination of TASS and HFR. managers’ name etc. For a majority of funds. (1999) and Ackermann and Ravenscraft (1998) used a combination of HFR and MAR. The TASS database is used in Brown et al.1 Data Providers First. Some data collectors make them. 2200 funds). strategy followed. but Ackermann et al. management and incentive fees. The MAR database is used. (2001). ‘Hedge Fund Research. HFR defines sixteen 57 . Inc. and Amin and Kat (2001). These are ‘Managed Account Reports’ (MAR. There is no consensus on the definition of the strategy followed but there are similarities. many hedge funds release monthly information to inform existing investors or to attract new ones. it is important to stress all information on hedge funds is available exclusively on a voluntary basis. 2002). As stressed by Amin and Kat (2001). HFR is used by Schneeweis and Spurgin (1998). Fung and Hsieh (2000. 1500 funds). there are three main hedge fund database providers in the world. assets under management.

2 Hedge Funds We use hedge fund data from HFR and MAR. 15 This happened in three cases: when the same fund (same name. 12 non-directional and 5 directional strategies. plus the Funds of Funds and the Sector categories. This leaves a total of 2796 individual hedge funds. and returns) appeared twice in the database with two different company names. we use the database presenting more observations. Then. Finally. These funds include 1995 (71%) survived funds and 801 (29%) dissolved funds. and when the same fund (same company. we remove funds that appear twice15 and funds with quarterly returns. TASS defines 15 strategies. when the same fund (same name.different strategies in two categories. 58 . whose database ranges from 1988 to 1995. The HFR database provides 198 monthly returns on 1811 individual hedge funds plus 48 HFR indices (16 investment styles with 3 indices for each investment style: onshore. and the MAR database gives 2354 individual hedge funds plus 23 indices between January 1984 and June 2000. Both databases give monthly net-of-fee individual returns and other information on individual funds and group them in indices. (1999). This leaves us with 1639 individual hedge funds in the HFR database and 2014 hedge funds in the MAR database. in each database. For the 857 funds present in both databases. 4. company. and returns) appeared twice in the database. offshore and a combined index). This is one of the greatest database ever used in hedge funds performance studies. and returns) appeared twice in the database with two different fund names. similarly to Ackermann et al.

Schneeweis and Spurgin (1998) and Liang (1999) use style analysis based multifactor approach. The alternative is the valueweighted portfolio of all NYSE. Amex and NASDAQ stocks market proxy. The comparison of descriptive statistics of the two proxies suggests that both market proxies are very similar. 1996. 1993. 2000. We take the value-weighted portfolio of all NYSE. 16 The comparison is available upon request. Carhart. while Brown et al.16 The results of the study should not be influenced by the market proxy chosen.4. Amex and NASDAQ stocks usually used in mutual funds performance studies (see for example Fama and French. 59 . We have to choose a market performance index. 1997) or the Russel 3000 used in Agarwal and Naik (2004) and Agarwal (2001). (1999) address this issue by employing a Generalised Stylistic Classification (GSC) algorithm and grouping the managers on the basis of their realized returns. and the one-month T-bill from Ibbotson Associates as the risk-free rate.3 Risk-free Return and Market Performance As underlined by Agarwal (2001). a fundamental challenge in a risk-adjusted analysis of hedge funds is the identification of a meaningful benchmark. Fung and Hsieh (1997).

as pointed out in several other studies (see Brown et al. The second bias called instant history bias or backfilled bias (Fung and Hsieh 2000) occurs because a fund’s performance history is backfilled after inclusion.. since they are not allowed to advertise. (1999) and to Fung and Hsieh (2000). However.. 2001). This may cause an upward bias because funds with a poor track record are less likely to apply for inclusion than funds with good performance history.4. to avoid polemics. they consider inclusion in a database primarily as a marketing tool. Nevertheless. two upward biases exist in the specific case of hedge funds because. These providers do not eliminate defunct funds and should normally not suffer from survivorship bias for the years after the start of the databases. because they might be considered by the SEC as making illegal advertising. data vendors do backfill fund’s performance history when a new fund is added to the database. Moreover. Carhart et al. survivorship bias is very likely to be present and not negligible. This bias may severely hinder statistical inference (Hendricks et al. In response to this concern. The first one is called the self-selection bias is present because funds that realize good performance have less incentive to report their performance to data providers in order to attract new investors. 2000). 1997. we take all funds (both living and dissolved) into account. for the period before 1994. This allows them to provide data that go back before the starting date of the database itself (usually 1993).. 2000).. although hardly measurable.4 Biases in Hedge Funds Data Survivorship bias is an important issue in mutual funds performance studies (see Carhart and al. according to Ackermann et al. 60 .

The particularly low estimate for the bias in the 1984-1993 period is due to the extremely low attrition rate of hedge funds in the database.V Data analysis We will first discuss evidence of biases before and after 1994 in our database. We report the bias using both definitions for the whole period and for 2 sub-periods 1984-1993 and 1994-2000.. Ackermann et al. Ackerman et al. Liang.g. as evidence of a serious difference in these two sub-periods would cast doubt on the reliability of the global analysis for the 1984-2000 period.1 Survivorship bias Survivorship bias has received considerable attention in the academic literature. As our sample does not report data on dissolved funds prior to 1994 for MAR and prior to 1993 for HFR. This indicates the lack of reliability of the database for this period. Two definitions of this bias are commonly used in studies: the performance difference between surviving funds and dissolved funds (e. one is automatically left with funds that survived this period and thus survivorship is indeed qualitatively maximal.36% (or 4.9% per annum) using the second formula.45% per annum) for the whole period using the first formula and in Panel B a bias of 0. In Panel A of Table 9.07% per month (0.g. Their value (0. (1999) use the same combination of databases as we do and the first definition of survivorship bias. 5.9% and 61 .16% per year) is lower than the value we obtain for the whole period and for the second sub-period (respectively 0. we report a monthly survivorship bias of 0. A look at sub-period biases indicates that survivorship bias is much higher after 1994. 2000). 1999) and the performance difference between living funds and all funds (e.

i.30% monthly bias found by Fung and Hsieh (2000).e.5% from Fung and Hsieh (1998). The value reported using the second definition for period 1994-2000 (1. This difference can mostly be explained by the different time period analysed. 17 We find this consensus value quite high when compared to the 0.45% per year) and closer to the one we obtain for the first sub-period (0.17 Liang (2000) compares the survivorship bias in HFR and TASS database and finds significant differences between them. The issue of their suspiciously low survivorship bias has been extensively discussed by Fung and Hsieh (2000) and Liang (2000). the greatest part of their time window does not encompass funds that disappeared during the period. (1999) study the 1988-1995 period. but the TASS database is probably more complete as it covers a larger number of dissolved funds. the 3% bias found by Liang (2001) and the industry consensus bias of 3% stressed by Amin and Kat (2001). 62 . even though our sample covers a larger time period. He finds out that they are mostly non-overlapping.8-1. Ackermann et al.6% per year). Our relatively low survivorship is in line with Liang’s (2000) claim that the HFR and MAR databases taken together do not fully resolve the survivorship issue.5 bias reported by Malkiel (1995) and Brown and Goetzmann (1995) for US mutual funds. It is however lower than the 0.2%) is very close to the percentage of 1.4.

46 2.39 1.09 2.07 1.58 1. 3.8 1.33 5.50 0.75 1.23 Obs.61 2.79 -1 1.61 0. Dev.24 1.66 1.26 1.3 Obs.08 2.12 1.67 0.77 0.58 4.86 0.51 0.48 0.58 1.81 10.38 1.48 St.79 1.03 0.08 1.48 1.Table 9: Survivorship Bias in Hedge Funds All Funds Year 1984 1985 1986 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 Mean 84-93 Mean 94-00 Mean 84-00 Return 1.69 1.61 1.15 1.73 0.34 2.92 3. 1. Dev.57 1.96 0.58 0. Dev.74 2.55 1.15 1.72 1.55 4.66 0.74 0.65 0.4 1.6 0.52 0.67 1.36 1.47 0.87 1.2 2.16 1.26 1.79 1.35 -0.36 1.16 1.62 1.83 1.44 1.68 1.75 11.54 0.29 1.95 0.08 2.99 1.76 1.5 1.91 0.83 1.74 1.19 Dissolved Funds Std. 348 501 716 1069 1376 1931 2792 3880 5280 7245 9856 13528 17654 22454 25735 25826 12421 Return 0.14 0.55 Surviving Funds Std.16 1.46 1.69 0.64 1.75 0.77 2.64 0.12 2.32 1.6 1.73 0.07 0.44 1.23 1.13 0. 248 356 497 782 1020 1444 2063 2826 3860 5140 6931 9298 11942 15122 18440 21712 11880 Return 1. 76 121 195 263 332 463 705 1030 1396 2081 2901 4206 5688 7308 7271 4090 529 .62 1.22 1. 7.78 2.97 1.16 1.72 2.22 1.3 3.97 2.94 2.23 Obs.4 0.87 1.

45 per Month per Year per Month per Year per Month per Year Panel B: Living Funds .16 1.89 per Month per Year per Month per Year per Month per Year This Table reports the survivorship bias of calculated from our database.65 7.All Funds Bias 84-93 0.09 1.05 0. 64 . In Panel B survivorship bias is calculated as the performance difference between surviving funds and all funds. Numbers in the table are monthly percentage.Dead Funds Bias 84-93 0.Panel A: Living Funds . In Panel A survivorship bias is calculated as the performance difference between surviving funds and dissolved funds.36 4. All returns are net of fees and on a monthly basis.22 Bias 84-00 0. indicating inferior performance: it corresponds to a decrease of the mean return of almost 3.79 Bias 84-00 0. Our combined HFR/TASS database contains 2796 hedge funds. It shows a declining return pattern towards the date of exit. Figure 3 plots returns of the dissolved funds in our database over the 24month period before their exit dates. including 1995 survived funds and 801 dissolved funds as of June 2000.07 0.5% in two years. These biases suggest that poor performance could be the main reason for disappearance.6 Bias 94-00 0.96 Bias 94-00 0.

5 9 8 7 6 5 4 3 2 1 Months to dissolution Monthly returns are computed on the MAR/HFR database over the 24-months period before their exit dates. Data are obtained on 801 individual hedge funds between January 1984 and June 2000.5 1 Returns 0. 65 .5 0 24 23 22 21 20 19 18 17 16 15 14 13 12 11 10 -0.Figure 3: Monthly returns for dead funds towards the dissolution date 2 1.5 -1 -1.

This corresponds to a demand by fund managers who market themselves if they have good track records. Brown et al. The remaining difference can be explained by the difference in time period covered and in the database used. They find a 1. 1. we estimate the average monthly return using the portfolio which invests in all funds from our database each month (we called this portfolio the observable one).49%.20% (36 months). 1.9% per year. The bias is estimated for the whole period and splitting the time period in two in order to compare our results with those obtained by Fung and Hsieh (2000). On the other. historical returns are backfilled. This gives an estimate of approximately 0. the bias of 1. the bigger the bias. Results are reported in Table 10.5. we estimate the average monthly return from investing in all these funds after deleting the first 12.4% found by Fung and Hsieh (2000) for the instant history bias.42% (when deleting the 12 first months). while the adjusted observable one was 1. On the one hand.4% per year difference in returns for the 1994-1998 period. 24. Following Park (1995). Interestingly.26% (24 months). the observable monthly return averaged 1.15% (36 months). and 1. (1997) and Fung and Hsieh (2000). 36 and 60 months of returns (we called this portfolio the adjusted observable one).e. i. after compiling good performance. lower than the 1. For the 1/94-6/00 period. we estimate this bias for our hedge fund database in two steps. Fung and Hsieh (2000) estimate this bias using a 12 month incubation period. 66 .2% per year is closer to the one of Fung and Hsieh (2000).2 Instant Return History Bias When new funds are added into a database. our results indicate that the longer the estimation period. For the 1/84-6/00 period.

and only consider the 1984-1993 sub-period in isolation in order to compare the consistency of the results obtained and relate them to the presence of the biases of the database. Table 10 yields a consistently greater estimate of the instant history bias for the pre-1994 than the post1994 period when 24 months or more are deleted. 67 . Results for the whole 1984-2000 period are not reported because we may not soundly claim that the pre. Based on these pieces of evidence regarding the presence of biases.Contrarily to the results reported for the survivorship bias. the rest of this study will focus on performance level and persistence for the 1994-2000 sub-period.and post-1994 time windows exhibit the same biases with the same severity.

5.15%) and by the Sector (1.3 Basic Performance Before going in the heart of our work. 68 . Fama and French’s (1993) SML and HML. Short Sellers (0. Given that MAR and HFR classify differently the individual hedge funds.18 We contrast hedge funds data against the descriptive statistics of the market proxy.64%). we combine the data for strategies that exist across both databases (sometimes under different names) and we add the strategies or sub-strategies present in only one database. Strategies that offer the lowest mean return are Foreign Exchange (0. the MSCI World excluding US.71%) and Funds of Funds (0. Lehman BAA corporate bond index (default spread). JP Morgan Emerging Market Bond Index. These statistics are reported in panel B of Table 10. Carhart’s (1997) momentum factor.78%). panel A of Table 10 contains descriptive statistics of the funds in our database.58%) followed by the US Opportunistic Growth (2. Fama and French’s (1998) international IHML. The results are similar for mean excess returns. and Goldman Sachs Commodity Index. Panel A shows that the highest mean return was achieved by the US Opportunistic Small Caps (2.91%). Salomon World government bond index.29%. including living and dead funds. 18 The description of these strategies is available upon request. whereas the mean return of the whole database is 1. Lehman US aggregate bond index.

13% 1.20% 1.42% 1.04% 0.35% 0.19% 0. Nb of Fds per Month 1631 1308 1017 773 426 0.35% 1/84-12/93 Difference Av. including 1995 survived funds and 801 dissolved funds as of June 2000. All returns are net of fees and on a monthly basis. Nb of Fds per Month 768 644 527 425 269 This Table reports the Instant History Bias calculated from our database. Our combined database contains 2796 hedge funds.21% Mean Annual Return All Without 12M Without 24M Without 36M Without 60M 1.28% 0. 69 .29% 0.Table 10: Estimation of Instant Return History Bias 1/84-6/00 Mean Annual Return All Without 12M Without 24M Without 36M Without 60M 1.16% 0.08% 0.41% 1/94-6/00 Difference Av.23% 0.29% 1. Instant history bias is calculated as the performance difference between the average monthly return using the portfolio which invests in all funds each month (the observable portfolio) and the average monthly return from investing in these funds after deleting the first 12. Numbers in the table are monthly percentage.10% 0.27% 1.49% 1.16% 1.62% 1.13% 0. 36 and 60 months of returns (the adjusted observable portfolio). 24.10% 0.19% 1.35% 1.58% 1. Nb of Fds per Month 207 147 103 71 32 Difference Av.15% Mean Annual Return All Without 12M Without 24M Without 36M Without 60M 1.26% 1.

The mean excess premium of the MSCI World excluding US is an insignificant 0.When standard deviation is taken into account through the Sharpe measure (the ratio of excess return and standard deviation).0465).3633) and the Market Neutral Relative Value Arbitrage (0. 70 . They respectively produced an average premium of 0. The average SMB and HML returns are insignificant.26% per month (about 16% per year) and statistically different from zero. A look at the t-stats indicates that mean returns are significantly different from 0 at the 5% significance level for all funds but the Short Sellers and that the mean excess returns are significantly different from 0 at the 1% in a majority of cases. This large value indicates that the period under review is unambiguously bullish.15% per month. followed by the Market Neutral Convertible Arbitrage (0.. the latter being economically as well as statistically significant. unlike the results obtained by Fama and French (1993) and Carhart (1997). Their high variance suggests a very unstable behavior.2766). which are also in the worst performing funds when risk is not taken into account.43% per month. Funds offering the best trade-off between risk and return are the US Opportunistic Small Caps (0. results are somewhat different.28% and 1. Panel B of Table 11 shows that the mean excess return of the Market Proxy is 1.3814).19 The international HML and the momentum factor give more interesting values. The worst Sharpe ratio is obtained by the Short Sellers (0. 19 The differences in SMB and HML can be explained by the different periods covered by our studies.

98 3.11 2.11 3.49 1.56 71 .81 3.98 4.6% 55.75 3.69 4.78 1.58 1.91 0.4% 46.95 4.31 3.76 6.3% 100% 17.69 6.49 4.Table 11: Descriptive Statistics of Hedge Funds Strategies and Passive Investment Strategies Panel A: Hedge Funds strategies No of Fds Event Driven Distressed Securities Risk Arbitrage No sub-strategy Global International Emerging Regional Established Global Macro Market Neutral Long/short Convertible Arbitrage Fixed Income Stock Arbitrage Mortgage-Backed Sec.94 1.84 1.41 1.2 9.61 2.97 6.08 1.29 4.02 1.68 1.35 6.34 3.76 5.7 0.29 4.27 Std.87 2.49 3.9% 6.53 2.1% Living Funds 56 25 27 4 184 27 123 34 259 522 83 38 50 309 42 0 9 139 67 61 11 16 40 86 9 132 229 314 1995 Dead Funds 52 24 23 5 165 35 80 50 60 189 67 6 11 83 14 8 15 73 21 49 3 11 9 34 0 54 118 21 801 Mean Return 1.53 9.78 4.2 3. 5.14 6.78 5. Dev.1% 6.52 4.08 0.25 0.09 5.02 0.36 1.33 4.71 1.6% 100% 21.43 3.57 6.9% 1.06 1.5% 51.2 4.59 2.5 3.1% Substrategy 100% 45.31 0.73 1.29 t(mean) =0 3.1% 7.33 5.38 5.34 3.2 2.84 6.36 3.98 1.29 3.12 5.06 4.75 5.75 3.79 1.5 6.2% 24.3% 8.26 1.21 6.4 2.68 0.56 5.71 7.55 3.2 0.2% 8.1 1.42 1.15 1. Relative Value Arb.8% 58.64 1.94 1.89 0.33 4. Short Sellers US Opportunistic Growth Value Small Caps Long Only Leveraged Market Timing Equity non-Hedge Foreign Exchange Sector Funds of Funds Non classified All funds 108 49 50 9 349 62 203 84 319 711 150 44 61 392 56 8 24 212 88 110 14 27 49 120 9 186 347 335 2796 100% 41.

57 1.71 6.17 0.91 5.08 2.19 0.2 17.05 0.Panel A (continued): Hedge Funds strategies Median Event Driven Distressed Securities Risk Arbitrage No sub-strategy Global International Emerging Regional Established Global Macro Market Neutral Long/short Convertible Arbitrage Fixed Income Stock Arbitrage Mortgage-Backed Sec.41 -11.54 -8.75 -7.12 1.98 0.46 -12.36 -1.36 3.92 0.23 1.75 1.01 -8.47 6. return 0.36 0.2 2.79 -5.24 0.1 3.61 1.47 0.34 2.04 1.22 2.82 -10.09 6.36 0.22 0.74 -7.05 12.24 0.71 2.54 2.38 0.4 -11.65 3.67 0.36 1.91 -7.96 2.38 0.15 0.37 1.97 0.8 1.09 0.25 Sharpe ratio 0.02 0.31 -5.12 0.66 0.21 0.14 0.87 Mean exc.96 3.47 -2.83 6.7 -2.41 0.535 -5.7 -1.13 2.59 -4.18 0.94 9.64 10.27 -3.71 17.01 Min -8.68 3.71 7.48 0.91 0.83 11.54 0.27 2.38 1.79 -17.09 0.89 t(mean exc.65 3.96 13.53 15.37 10.68 0.26 0.27 0.69 2.94 0.32 0.54 2.11 1.81 1.86 1.43 0. Short Sellers US Opportunistic Growth Value Small Caps Long Only Leveraged Market Timing Equity non-Hedge Foreign Exchange Sector Funds of Funds Non classified All funds 0.42 3.19 0.32 6.24 0.31 7.79 0.69 1.96 3.61 0.32 1.84 0.18 1.56 9. Relative Value Arb.96 1.47 1.65 3.47 -14.25 0.85 0.42 1.28 0.14 0.46 -12.21 0.24 0.05 0.18 0.29 6.97 -6.49 1.93 5.16 72 .2 3.62 -9.) =0 2.47 1.77 1.6 1.15 5.52 10.10 0.32 0.28 0.75 1.97 -8.65 0.17 0.19 0.51 0.62 -4.31 1.08 0.62 0.89 0.47 10.38 11.02 1.86 0.88 Max 6.92 -14.75 -12.12 2.62 10.41 1.54 9.99 6.85 1.75 1.24 0.22 0.66 -20.37 2.88 0.

03 7.98 -3. standard deviation.26 0. Sharpe ratio is the ratio of excess return and standard deviation.32 -0.92 13.03 -15.27 1.58 0.184 t(mean) = 0 Std.Panel B: Passive Strategies Mean Return Equity Market Proxy MSCI World Excluding US F&F SMB Factor F&F HML Factor F&F HML International Factor Momentum Factor Bond Lehman US Aggregate Salomon World Government JPM Emerging Market Bond Index Lehman BAA Corp. medians.84 12.05 0.43 -21.75 -10. We calculate the Mean Excess Return and the Sharpe ratio considering Ibbotson Associates one-month T-bills. mean returns.72 5.100 0.19 0.13 -6.12 0.43 -0.51 -0.97 0.38 1.48 -0.81 4.06 5.129 -0.14 0.87 5. t-stat for mean = 0.94 10.97 -0. No of Fds represent the number of funds following a particular strategy (or sub-strategy).44 1. Median Min Max Mean excess return t(excess mean) = 0 Sharpe ratio This table shows the inception date.79 0.34 1.02 0.18 -0.64 2. Dev.115 -0.61 -0. Living Funds and Dead Funds represents the number of surviving and dead funds (in June 2000.69 -9.61 -0.1 0.71 0.41 1.084 1.32 0. t-stat for mean excess return = 0.050 -0.41 -1.28 15.64 4.61 -16.16 0 0.61 3.7 4 4. without considering the new funds established in 6/2000).93 7.75 2.39 2. and for 11 passive investment strategies for the 1/1994 and 6/2000 period.100 -0.52 -1.45 0.49 12. maximum.15 5. Numbers in the table are monthly percentage.75 3. mean excess returns.12 1.86 1.55 4.04 4.449 0.315 0.28 3.55 -2.82 -1.95 -12. .45 0.02 1.05 -12.027 0.26 0.15 1.87 2.41 3.12 3.68 0. and Sharpe ratios for the individual hedge funds in our combined MAR/HFR database following 12 strategies and 15 sub-strategies.31 -13.28 4.41 -0.28 1.64 0.52 0.88 1. In panel A. minimum.74 3. Commodity Goldman Sachs Commodity 0.8 -0.9 21.062 0.19 1.54 0.18 -0.73 -3.

Short Sellers are negatively correlated with all the other hedge fund strategies. with the only difference that the Lehman BAA Corporate (0. and by the JP Morgan Emerging Market Bond Index for the bond. The Sharpe ratio obtained by our whole hedge fund database (0. In particular. Panel A reports correlations among hedge funds strategies. Table 12 reports correlation coefficients among and between hedge funds and passive investment strategies.80 and 14 (14%) are negative.The highest mean return was obtained by the Market Proxy for equity. There is a high variability between different strategies. 74 .97 (between Equity non-Hedge and US Opportunistic) to –0.1). ranging from 0.3154). Schneeweis and Spurgin (1998). The Sharpe ratios bring the same results.1596) is lower than the one for the Market Proxy (0.4 Correlation Fung and Hsieh (1997). Hence. the addition of hedge funds to a traditional portfolio should improve its risk-return trade-off.80 (between Short Selling and US Opportunistic). 5.0842) has a Sharpe ratio very close to the one obtained by the JP Morgan Emerging Market Bond Index (0. Liang (1999) and Amin and Kat (2001) report a weak correlation between hedge funds and other securities.0999). and higher than for the MSCI World Excluding US (0. 39 correlation coefficients (40%) are greater than 0.

86). but this result can easily be understood since the market proxy contains almost all the American market. Correlation coefficients between hedge funds strategies and the Market Proxy are. The range is narrower than in the previous case (from –0.5 whereas they are always smaller than 0.46 (and higher than –0. 75 .53) and almost 90% of them are below 0. These results confirm that hedge funds strategies are weakly correlated with traditional investment tools. too low to raise serious multi-colinearity concerns.65 to 0.3. All coefficients are below 0.3 with the MSCI World excluding US and than 0. 20 Except with the market proxy.Panel B reports correlation coefficients between hedge funds and equity.5 with bond indices (except for the default factor). bond and commodity indices. greater than 0.20 Panel C displays correlations among Passive Investment strategies. in almost all cases.

the discussion of performance models will be performed on the 1994-2000 period. 21 To make individual estimation. 6.1 Performance Measurement using the CAPM The first performance model used is the CAPM based single index model. 76 . with the percentage of significantly positive. when the database can be considered as fully reliable. so that relatively accurate risk measures can be estimated. Panel A of Table 13 reports the results for the strategies. We use equally weighted portfolio excess returns for each investment style and for the All Funds category. This approach enables us to analyse hedge funds performance in more details.VI Hedge Funds Performance The aim of this section is to determine whether or not hedge funds as a whole and depending on the strategy followed have out-performed the market. and we estimate the model for each fund individually. insignificant and negative alphas at the 5% level. we require all funds to have consecutive monthly return history for at least 24 months. We compute all estimations by using Newey-West (1987) standard errors to adjust for any autocorrelation in the returns. Following the discussion of the previous section. sub-strategies and for the All Funds category.21 The last columns give the distribution of individually estimated alphas per strategy.

77 0.91 0.43 0.58 -0.95 0.97 0.31 0.92 0.87 0.86 0.60 0.27 0.85 0.85 0.94 0.90 0.86 0.1 0.71 0.35 0.96 1 -0.71 -0.65 0.78 -0.74 0.82 -0.71 1 0.78 -0.Table 12: Correlation between Hedge Funds and Passive Investment Strategies Panel A: Correlation between Hedge Funds strategies EVT EVT GLB MAC MKN SHS OPP LOL MKT ENH FEX SEC FOF NCL ALL 1 0.89 1 0.83 0.95 0.72 -0.7 0.76 0.44 0.32 1 GLB MAC MKN SHS OPP LOL MKT ENH FEX SEC FOF NCL ALL .87 0.29 0.93 1 0.8 -0.13 0.84 0.95 0.91 0.89 -0.46 0.84 0.22 -0.85 0.35 0.25 0.15 0.15 0.95 1 0.51 -0.77 0.94 1 -0.74 0.95 0.38 1 0.83 0.94 0.27 0.85 0.74 0.39 0.78 0.89 0.31 0.94 1 0.42 0.88 0.95 1 0.83 0.88 0.62 0.73 0.26 0.38 0.65 0.61 0.30 -0.17 0.87 1 0.69 0.26 0.83 -0.66 0.30 0.95 1 0.71 1 0.

02 -0.72 0.02 -0.02 0.08 -0.15 0.54 0.01 0.28 0.77 0.30 .56 0.09 -0.29 0.23 0.24 MOM 0.29 0.25 0.26 0.67 0.25 -0.58 -0.48 -0.55 -0.00 -0.31 0.25 0.39 0.00 0.54 0.53 0.31 -0.32 0.66 0.05 0.71 0.25 0.54 0.04 -0.73 WXU 0.18 0.65 0.05 0.16 SMB 0.27 0.64 0.02 SWG 0.14 -0.30 -0.39 0.49 0.20 0.65 0.19 0.31 -0.30 -0.12 0.02 0.30 -0.20 0.14 0.47 0.51 0.05 0.19 0.31 0.27 -0.06 0.20 0.13 0.26 0.55 0.40 0.08 0.06 -0.01 0.04 -0.31 0.05 0.24 -0.21 -0.13 0.06 0.34 -0.23 LUS 0.09 0.18 0.58 -0.41 -0.03 -0.30 -0.18 0.20 BAA 0.03 0.Panel B: Correlation between Hedge Funds strategies and Passive Investments strategies EVT GLB MAC MKN SHS OPP LOL MKT ENH FEX SEC FOF NCL ALL MKT 0.43 IHML -0.82 0.04 0.06 0.10 0.16 0.63 0.45 -0.06 0.48 -0.18 -0.34 0.35 0.05 -0.57 -0.00 -0.58 GSC 0.58 0.11 0.86 0.27 0.35 0.29 -0.57 0.28 0.57 0.07 0.09 0.24 -0.05 0.10 0.49 -0.12 -0.17 -0.03 0.29 0.18 0.32 -0.32 -0.16 0.31 0.06 0.32 -0.17 -0.20 0.10 0.31 -0.31 0.52 0.26 0.01 -0.53 -0.69 0.06 MEM 0.56 HML -0.31 0.19 0.35 0.36 0.58 0.

19 -0. ALL = All hedge funds. SWG = Salomon World Government Bond Index.15 0. IHML = International HML. between hedge funds strategies and passive investment strategies (Panel B) and between passive investment strategies for the January 1994-June 2000 period. MKT = Market Proxy.04 -0. FEX = Foreign Exchange.04 0.43 -0. MKN = Market Neutral.45 0.02 -0.00 -0. MKT = Market Timing.07 0.23 -0.08 -0.13 -0.04 -0.09 0.19 1 -0.46 0.22 1 0. WXU = World excluding US. MEM = JP Morgan Emerging Market Bond Index.44 1 0.01 0.23 -0. GLB = Global. EVT = Event Driven. SHS = Short Selling. MOM = Momentum.03 1 0.13 0.38 1 0.03 0. LUS = Lehman US Aggregate Bond Index.27 -0. NCL = Non Classified. LOL = Long only Leveraged.38 0.23 0. .Panel C: Correlation between Passive Investment strategies MKT WXU SMB HML IHML MOM LUS SWG MEM BAA MKT WXU SMB HML IHML MOM LUS SWG MEM BAA GSC 1 -0.30 0. SEC = Sector. FOF = Funds of Funds.06 0.04 0.31 -0.01 1 0. OPP = US Opportunistic.06 0. BAA = Lehman BAA Corporate Bond Index and GSC = Goldman Sachs Commodity Index. ENH = Equity non-Hedge.28 -0.08 0.52 -0.06 -0. MAC = Global Macro.15 0.1 0.15 0.1 1 0.14 0.05 0.10 0.12 -0.24 0.38 0.01 0.02 0.14 -0.43 This Table reports the correlation coefficient between hedge funds strategies (Panel A).06 1 -0.20 -0.03 1 -0.06 -0.

Equity non-Hedge or the Non Classified funds). more than 80% of the individual funds do not significantly out-perform the market. Only 6 out of 28 investment styles exhibit significant momentum loadings (at the 10% level).The betas estimated in Panel A are rather low. in the Short Sellers strategy. The All Funds category also significantly out-performs the market at the 1% level. more than 30% of the alphas are significant. The Fama and French (1993) model produces results very similar to the Carhart model. inducing that the best funds must have obtained extremely high returns.22 6. In Panel B of Table 13. the premium is significantly negative. Panel C shows that the HML (respectively IHML) factor seems to add less explanatory power as only one third (respectively one seventh) of the factors is significantly positive at the 5% level.2 Performance Measurement using Multi-Factor Models It is presumably better to use a multi-factor model to account for all possible investment strategies. 22 These results are remarkably similar to the 1984-1994 period (data available upon request). suggesting the need to use a more detailed model. in almost all cases. Surprisingly. Overall. However. we report the results for Carhart’s 4-factor model and in Panel C the results for our combined model applied to hedge funds. 23 80 . In almost all out-performing strategies. significantly positive.g.23 Panels B and C reveal that the premium on the SMB factor is. The momentum factor does not prove to be a strong indicator of hedge funds behaviour. two thirds of the strategies produce significantly positive alphas. for some strategies (e. Moreover. the sign of the coefficient is in 3 cases negative. indicating momentum contrarian strategies. except for US Opportunistic Growth and the Long Only Leveraged.

Panel C also indicates that the World excluding US. Some Event Driven and Market Neutral managers follow a momentum strategy and others are momentum contrarian. as we can expect. Default. 81 . The Emerging Bond factor adds explanatory power in 16 of the strategies and substrategies. It is highly significant (at the 1% level) in more than half of the strategies. Market Neutral and US Opportunistic managers prefer stocks with high book-to-market ratios. The default factor has. More than half of the managers invest in emerging bond markets. as well as for the All Funds category. Most Event Driven. a negative impact in half of the cases. US Bond. These results provide some insight into the preferences of hedge funds managers depending on the strategy followed: Almost all managers seem to prefer smaller stocks. World Government Bond and Commodity factors add explanatory power in 20 to 33% of the cases.

54 0.Table 13: Performance Measurement using the CAPM.35% ** 0.07 *** R² adj No Fds 0.43% ** 0.08% 0.44 *** 0.18% -0. Global Macro Market Neutral Long/short Convertible Arb. 0.21 *** 0.55 *** 0.06 84 37 38 9 258 50 151 57 252 553 102 39 48 310 46 Alpha Distribution +/0/25% 11% 42% 11% 12% 10% 9% 19% 23% 33% 34% 74% 10% 31% 28% 73% 89% 53% 89% 86% 90% 87% 79% 75% 66% 64% 26% 88% 68% 70% 2% 0% 5% 0% 3% 0% 4% 2% 2% 1% 2% 0% 2% 1% 2% 82 .35% *** Mkt 0.42 0.62% *** 0.62 0.43 0. Risk Arbitrage No sub-strategy Global International Emerging Regional Est.24 *** 0.4 0.53 0.67 *** 0.39 *** 0. Alpha Event Driven Distressed Sec.54 *** 0.40% *** 0.06 *** 0.08 0.38% *** 0.10% 0.57 0.39 *** 0.42 *** 0.53 0.16 *** 0.03% -0.45 0.26% 0.38 0.43 0.55% *** 0.45 *** 0.62 *** 0. Fixed Income Stock Arbitrage Mortgage-Back.42 *** 0. Carhart’s 4-factor model and the combined model.26 0.41% *** -0.13% -0.38 0.72% *** 0.

83% *** 0.60% *** 0.36% ** 0.54 0.50% *** 0.36% ** 1. Sector Funds of Funds Non classified All funds 0.01 0. Market Timing Equity non-Hed.24% 0.66 *** 0.47% *** 0.Panel A (continued): Single index model Alpha Mkt R² adj No Fds Alpha Distribution +/0/- Rel.86 *** 0.77 0.19 *** 0.96 *** 0.68% *** -0.69 *** 0.86% *** 0.49 0.17% 0.14 0.37 0.48 *** 0.01% 0.2 *** -0.7 *** 0.37 *** 0.07% 4.51% ** 0.64 0.96 *** 0. Growth Value Small Caps Long Only Lev.08 0.35 0.65 0.19 0. Value Arb.74 0. Foreign Exch.44 *** Mean R²adj 0.8 *** 0.44 8 18 180 68 100 12 18 34 112 9 128 278 230 2154 Mean 50% 28% 19% 24% 14% 33% 11% 26% 9% 0% 30% 13% 13% 24% 23% 50% 72% 81% 76% 85% 67% 83% 74% 89% 67% 67% 83% 86% 74% 74% 0% 0% 1% 0% 1% 0% 6% 0% 2% 33% 2% 4% 0% 2% 3% 83 .39 0.62 0.18 *** 0.39% *** 0.7 0. Short Sellers US Opport.

Alpha

Mkt

SMB

HML

Event Driven Distressed Sec. Risk Arbitrage No sub-strategy Global International Emerging Regional Est. Global Macro Market Neutral Long/short Convertible Arb. Fixed Income Stock Arbitrage Mortgage-Back.

0,43% *** 0,24% 0,62% *** 0,08% 0,03% 0,10% -0,07% 0,53% *** 0,67% *** 0,42% *** 0,71% *** 0,45% *** 0,02% 0,41% *** 0,29% ***

0,44 *** 0,42 *** 0,46 *** 0,41 *** 0,63 *** 0,54 *** 0,68 *** 0,44 *** 0,51 *** 0,21 *** 0,17 *** 0,08 *** 0,33 *** 0,24 *** 0,08 ***

0,24 *** 0,29 *** 0,22 *** 0,14 *** 0,26 *** 0,21 *** 0,29 *** 0,21 *** 0,27 *** 0,14 *** 0,15 *** 0,06 *** 0,16 *** 0,17 *** 0,01

0,11 *** 0,14 *** 0,09 *** 0,08 0,10 0,06 0,09 0,07 0,03 0,05 * 0,10 *** 0,06 *** -0,08 0,06 * 0,04

84

Panel B (continued): Carhart's 4-factor model

PR1YR

R² adj

No Fds

Alpha Distribution +/0/-

Event Driven Distressed Sec. Risk Arbitrage No sub-strategy Global International Emerging Regional Est. Global Macro Market Neutral Long/short Convertible Arb. Fixed Income Stock Arbitrage Mortgage-Back.

-0,06 *** -0,11 *** -0,04 * 0,11 *** -0,12 ** -0,09 *** -0,14 * -0,06 * 0,03 0,00 0,01 -0,04 *** -0,01 0,00 0,04 **

0,78 0,70 0,68 0,50 0,53 0,63 0,45 0,66 0,83 0,64 0,59 0,20 0,53 0,59 0,07

84 37 38 9 258 50 151 57 252 553 102 39 48 310 46

27% 14% 42% 22% 14% 10% 11% 25% 31% 37% 36% 77% 8% 38% 30%

69% 81% 55% 78% 84% 88% 87% 72% 68% 62% 63% 23% 92% 61% 67%

4% 5% 3% 0% 3% 2% 3% 4% 2% 1% 1% 0% 0% 1% 2%

85

Panel B (continued 2x): Carhart's 4-factor model

Alpha

Mkt

SMB

HML

Mortgage-Back. Rel. Value Arb. Short Sellers US Opport. Growth Value Small Caps Long Only Lev. Market Timing Equity non-Hed. Foreign Exch. Sector Funds of Funds Non classified All funds

0,29% *** 0,59% *** 0,77% *** 0,63% *** 0,78% *** 0,46% *** 1,57% *** 0,25% 0,43% *** 0,44% *** 0,07% 0,82% *** -0,02% 4,76% *** 0,43% ***

0,08 *** 0,24 *** -0,38 *** 0,75 *** 0,83 *** 0,7 *** 0,64 *** 0,9 *** 0,14 *** 0,8 *** 0,19 *** 0,63 *** 0,37 *** 0,09 0,42 ***

0,01 0,12 * -0,35 *** 0,38 *** 0,43 *** 0,29 *** 0,55 *** 0,54 *** 0,03 0,35 *** 0,08 *** 0,4 *** 0,19 *** 0,12 0,23 ***

0,04 0,15 ** 0,07 0,02 -0,11 0,11 *** 0,14 * 0,11 -0,06 -0,03 0,01 0,01 0,08 * 0,10 0,04

86

Panel B (continued 3x): Carhart's 4-factor model Alpha Distribution +/0/50% 17% 31% 38% 24% 42% 17% 18% 19% 11% 41% 20% 43% 30% 0,28 50% 83% 68% 60% 74% 58% 78% 82% 80% 89% 59% 77% 55% 69% 70% 0% 0% 2% 1% 2% 0% 6% 0% 1% 0% 0% 3% 2% 2% 2%

PR1YR

R² adj

No Fds

Rel. Value Arb. Short Sellers US Opport. Growth Value Small Caps Long Only Lev. Market Timing Equity non-Hed. Foreign Exch. Sector Funds of Funds Non classified All funds

-0,05 -0,07 * 0,00 0,04 -0,05 * -0,06 0,06 0,09 * -0,06 *** -0,04 *** 0,00 -0,02 0,06 -0,01 Mean R²adj

0,21 0,72 0,90 0,86 0,88 0,64 0,77 0,18 0,93 0,42 0,89 0,63 0,06 0,84 0,60

8 18 180 68 100 12 18 34 112 9 128 278 230 2154 Mean

87

Panel C: The combined model Alpha Event Driven Distressed Sec. Risk Arbitrage No sub-strategy Global International Emerging Regional Est. Global Macro Market Neutral Long/short Convertible Arb. Fixed Income Stock Arbitrage Mortgage-Back. 0,34% * 0,09% 0,60% *** -0,05% -0,50% -0,21% -0,71% 0,27% 0,56% *** 0,28% *** 0,67% *** 0,42% *** -0,26% 0,22% * 0,23% Mkt 0,42 *** 0,40 *** 0,44 *** 0,43 *** 0,68 *** 0,56 *** 0,73 *** 0,50 *** 0,53 *** 0,21 *** 0,16 *** 0,05 *** 0,36 *** 0,25 *** 0,02 Wd x US 0,01 0,06 -0,03 0,00 0,12 * 0,13 ** 0,13 * 0,10 ** 0,06 * 0,07 *** 0,01 0,05 ** 0,09 * 0,09 *** 0,02 SMB 0,25 *** 0,27 *** 0,24 *** 0,16 *** 0,16 *** 0,13 *** 0,17 *** 0,18 *** 0,25 *** 0,11 *** 0,13 *** 0,05 *** 0,13 *** 0,13 *** 0,02 HML 0,10 *** 0,12 *** 0,08 *** 0,10 0,07 0,03 0,06 0,09 * 0,04 0,04 ** 0,08 *** 0,04 ** -0,07 * 0,06 ** 0,01 I HML 0,05 0,04 0,07 0,05 0,04 -0,05 0,06 -0,02 -0,09 * -0,04 -0,07 *** 0,02 0,04 -0,05 0,01

88

Panel C (continued): The combined model PR1YR Event Driven Distressed Sec. Risk Arbitrage No sub-strategy Global International Emerging Regional Est. Global Macro Market Neutral Long/short Convertible Arb. Fixed Income Stock Arbitrage Mortgage-Back. -0,03 -0,08 ** -0,02 0,12 *** 0,01 -0,02 0,02 -0,03 0,03 0,02 0,00 -0,02 * 0,05 * 0,02 0,04 * US Bd -0,04 -0,08 0,01 -0,10 -0,33 * -0,17 -0,38 * -0,22 ** -0,11 -0,15 *** -0,12 * 0,06 -0,17 -0,19 *** -0,13 * World GV Emerg Bd -0,01 -0,03 -0,09 0,39 *** -0,41 *** -0,21 * -0,48 *** -0,09 0,01 -0,06 -0,05 -0,11 *** -0,02 -0,08 -0,01 0,06 *** 0,07 ** 0,05 * 0,06 0,13 *** 0,06 * 0,16 *** 0,03 0,06 *** 0,03 0,02 0,03 * 0,04 0,02 0,03 Default 0,11 * 0,16 ** 0,13 -0,16 -0,12 -0,05 -0,17 -0,16 * -0,06 0,07 0,10 * 0,16 *** -0,09 0,04 0,33 *** Comm. 0,00 0,00 -0,01 0,00 0,09 ** 0,08 *** 0,12 ** -0,03 -0,02 0,01 0,02 * 0,00 0,02 0,01 0,00 R² adj 0,79 0,74 0,66 0,51 0,67 0,72 0,6 0,71 0,85 0,72 0,61 0,49 0,55 0,68 0,22

89

Panel C (continued 2x): The combined model Alpha Rel. Value Arb. Short Sellers US Opport. Growth Value Small Caps Long Only Lev. Market Timing Equity non-Hed. Foreign Exch. Sector Funds of Funds Non classified All funds 0,43% 0,75% *** 0,46% *** 0,74% *** 0,20% 1,34% *** 0,00% 0,48% ** 0,23% * -0,02% 0,69% *** -0,28% * 4,86% *** 0,25% ** Mkt 0,21 *** -0,34 *** 0,80 *** 0,85 *** 0,77 *** 0,73 *** 0,88 *** 0,19 *** 0,84 *** 0,15 *** 0,63 *** 0,39 *** 0,14 0,44 *** Wd x US 0,03 0,01 0,09 ** 0,04 0,12 *** 0,08 0,14 * 0,02 0,08 ** 0,00 0,06 * 0,07 * 0,10 0,06 ** SMB 0,02 -0,31 *** 0,32 *** 0,36 *** 0,26 *** 0,46 *** 0,47 *** 0,09 *** 0,31 *** 0,07 *** 0,37 *** 0,14 *** 0,11 *** 0,19 *** HML 0,02 0,10 0,04 -0,12 0,13 *** 0,16 ** 0,07 0,00 -0,02 -0,01 0,00 0,06 * 0,14 0,04 * I HML 0,08 0,07 -0,16 ** -0,35 *** 0,02 -0,25 *** -0,14 0,04 -0,07 0,03 -0,08 -0,01 -0,13 -0,04

90

Panel C (continued 3x): The combined model PR1YR Rel. Value Arb. Short Sellers US Opport. Growth Value Small Caps Long Only Lev. Market Timing Equity non-Hed. Foreign Exch. Sector Funds of Funds Non classified All funds 0,02 -0,05 0,00 -0,03 0,01 -0,06 0,08 0,06 -0,05 * -0,03 0,00 0,04 0,03 0,01 US Bd 0,03 0,05 -0,18 * -0,21 -0,12 -0,27 -0,17 0,02 -0,20 ** -0,10 * -0,18 -0,18 * 0,06 -0,16 ** World GV Emerg Bd -0,23 * -0,14 0,18 ** 0,15 0,20 *** 0,14 0,00 0,21 * 0,21 *** -0,11 0,06 -0,20 ** 0,08 -0,05 0,10 ** -0,01 0,04 * 0,10 *** 0,00 0,05 0,19 *** -0,05 0,03 0,08 *** 0,02 0,08 *** -0,12 *** 0,05 *** Default 0,05 -0,07 -0,29 ** -0,18 -0,34 *** -0,51 ** 0,17 -0,16 -0,21 *** 0,23 *** 0,01 -0,01 -0,20 -0,03 Comm. 0,05 * -0,08 ** 0,02 0,00 0,04 ** 0,06 -0,01 -0,12 ** 0,04 * -0,01 0,03 0,04 * -0,04 0,01 Mean R²adj R² adj 0,36 0,72 0,92 0,89 0,91 0,65 0,82 0,25 0,94 0,56 0,89 0,75 0,05 0,88 0,66

91

Panel C (continued 4x): The combined model

No Fds Event Driven Distressed Sec. Risk Arbitrage No sub-strategy Global International Emerging Regional Est. Global Macro Market Neutral Long/short Convertible Arb. Fixed Income Stock Arbitrage Mortgage-Back. 84 37 38 9 258 50 151 57 252 553 102 39 48 310 46 27% 19% 39% 11% 10% 10% 8% 18% 42% 30% 27% 59% 8% 29% 33%

Alpha Distr. + / 0 / 69% 78% 55% 89% 77% 82% 75% 77% 56% 65% 71% 41% 79% 65% 63% 4% 3% 5% 0% 12% 8% 17% 5% 1% 5% 2% 0% 13% 6% 4%

92

T-stat are heteroskedasticity consistent. + / 0 / 38% 56% 76% 72% 79% 75% 78% 88% 71% 78% 69% 72% 62% 68% 70% 13% 0% 3% 0% 5% 0% 6% 0% 1% 22% 2% 14% 3% 6% 5% This Table presents the results of the estimation of the Single Index Model (Panel A). The last column gives the distribution of individually estimated monthly alphas for all funds with 24 monthly data or more in a specific investment style. *** Significant at the 1% level. We report the percentage of significantly positive alpha’s (+). Growth Value Small Caps Long Only Lev. Market Timing Equity non-Hed. Sector Funds of Funds Non classified All funds 8 18 180 68 100 12 18 34 112 9 128 278 230 2154 Mean 50% 44% 21% 28% 16% 25% 17% 12% 28% 0% 30% 14% 35% 26% 25% Alpha Distr. 93 . Short Sellers US Opport.Panel C (continued 5x): The combined model No Fds Rel. We report the OLS estimators for equally weighted portfolio’s per investment strategy. of Carhart’s (1997) Model (Panel B) and of our Combined Model (Panel C) for the 1/1984-6/2000 period. ** Significant at the 5% level and * Significant at the 10% level. The next to last column reports the number of individual funds used for the individual estimation of the last column. sub-strategy and for all funds. Value Arb. significantly negative alpha’s (-) and alpha’s that are insignificantly different from zero (0) at the 5% level. Foreign Exch.

more than 25% of the individual alphas are significantly positive at the 5% level. Our results are in most cases confirmed by the last column. taking more factors into account induces that fewer individual funds significantly out-performed the market. Seem to prefer smaller stocks . This independently confirms that our approach is able to capture important risk exposure of hedge funds. In the All Funds category. while the second one is opposite. 94 . matching the 27% they found. The excess returns for Event Driven. we can observe that hedge funds as a whole: Deliver significant excess returns (one fourth of the individual funds gave significant positive excess return) . All other categories display positive and significant alphas. Equity non-Hedge. The first evidence is consistent with our finding. and Foreign Exchange strategies are insignificant (at the 5% level). whereas the alpha for the Fund of Funds strategy is negative and significant. Long Only Leveraged. this difference does not exist for all hedge funds managers. Comparing the alpha distribution of Panels B and C shows that. for example. We find that 27% of the hedge funds show significant excess return. and more funds have insignificant or negative excess returns.These results are close to those found by Mitchell and Pulvino (2001) and Agarwal (2001) for the funds following Event Driven strategies. given that the HML factor is only significant for some strategies. Global. Invest in Emerging Market Bonds and suffer from the US Bond market. Evidence on alphas obtained in Panel C is contrasted. Carhart (1997) and Gruber (1996) examine US mutual fund strategies and report that managers prefer smaller stocks as well as growth stocks. Considering the All Funds category. However.

88 for our combined model.44 for the single factor model. (2001) and Fung and Hsieh (1997).92 with an average of 0.66 for our combined model. to 0. Our R²adj are also higher than those obtained by Brown et al.66 for the single factor model.77. Comparing their results with ours for strategies that exist across the two databases. The mean R²adj for the individual hedge funds estimation is up too. The average R²adj increases from 0. Long only Leveraged (0. with an average of 0.49: taking the same strategies.31 for several hedge funds strategies. Sector (0. the increase from the CAPM to Carhart’s model is 10% and from Carhart’s model to our combined model is another 7%. Schneeweis and Spurgin (1998) report R²adj between -0. US Opportunistic (0. 95 . Liang (1999) finds unadjusted coefficients ranging from 0. to 0. we get greater R²adj in all cases. For several HFR strategies. the Non-classified funds (0.27 and 0. Carhart’s (1997) model raises the R²adj by an average 10% over the single index model. They report R² lower than 0.25).82) and Event Driven funds (0.60 for the 4-factor model and to 0. Global Macro (0.84 for the 4-factor model and to 0.85).94).20 in all cases for groups of funds. The R²adj for the All Funds category increases from 0. but our combined model increases it again by another 7%. For the All Funds category. The combined model gives adjusted R² higher than 0.67 with a mean of 0.09 and 0.92). It seems particularly adapted to Equity non-hedge (0. our R²adj range between 0.Overall it seems that the combined model does a very good job in describing hedge funds behaviour.65 for all strategies but the Market Timing (0.89).79).56).05) and Foreign Exchange funds (0.23 to 0.63.

6.3 Performance over Shorter Periods In the previous sub-sections, we discuss the performance of hedge funds using several specifications for the 1/94-6/00 period. In order to better interpret these results, Table 14 presents a summary of the same analysis over different sub-periods. We subdivide the 1994-2000 period in two sub-periods of equal lengths, and then report results of the same analysis for the Asian crisis period. The analysis of the Asian crisis period will enable us to determine if some strategies took advantage of the crisis and which one suffered (or not) from it. The 1984-1993 period is reported in the last column for comparison purposes. The last column of Table 14 confirms that the same strategies significantly outperform the market when the period before 1994 is considered, with only 4 exceptions. Three strategies (Short Sellers, Market Timing, and Equity non-Hedge) do not beat the market in the 1984-1993-period but do after 1994, while one (Foreign Exchange) does the reverse. When the time period is divided in two, it is interesting to note that the significance of the abnormal performance for the 1994-2000 period is mainly due to the first subperiod; some hedge funds strategies and sub-strategies even significantly under-perform the market in the second sub-period (Global, Emerging, and Funds of Funds). The latter is the only one whose significant underperformance extends to the whole 1994-2000 time window.

96

Table 14: Performance of Hedge Funds in different SubPeriods

1 sub-period 1/94 - 6/00 Event Driven Distressed Sec. Risk Arbitrage No sub-strategy Global International Emerging Regional Est. Global Macro Market Neutral Long/short Convertible Arb. Fixed Income Stock Arbitrage Mortgage-Back. 0,34% * 0,09% 0,60% *** -0,05% -0,50% -0,21% -0,71% 0,27% 0,56% *** 0,28% *** 0,67% *** 0,42% *** -0,26% 0,22% * 0,23%

2 sub-periods 1/94 - 12/96 0,41% * 0,17% 0,60% ** 0,40% 0,23% -0,33% 0,47% 0,38% *** 0,45% * 0,20% 0,40% * 0,46% *** -0,47% 0,16% 0,58% *** 1/97 - 6/00 0,08% -0,11% 0,31% * -0,45% -1,21% *** -0,40% -1,70% *** -0,07% 0,56% *** 0,19% ** 0,72% *** 0,36% *** -0,41% 0,13% -0,17%

Asian crisis 1/97-12/98 0,02% 0,19% -0,13% -0,17% -1,02% -0,99% *** -1,18% -0,46% 0,46% 0,15% 0,66% *** 0,40% *** -0,72% 0,08% -0,13% 1/97-6/98 -0,05% -0,07% 0,02% -0,25% -1,30% ** -1,10% *** -1,55% ** -0,54% -0,07% -0,04% 0,42% ** 0,40% *** -1,23% ** -0,11% -0,12%

Pre-1994 1/84 - 12/93 -0,21% 0,07% -0,29% -0,53% * 1,01% * 1,29% *** 0,92% 1,21% *** 0,85% *** 0,63% *** 0,05% 0,55% *** 1,01% *** 0,81% *** 0,00%

97

1 sub-period 1/94 - 6/00 Rel, Value Arb. Short Sellers US Opport. Growth Value Small Caps Long Only Lev. Market Timing Equity non-Hed. Foreign Exch. Sector Funds of Funds Non classified All funds 0,43% 0,75% *** 0,46% *** 0,74% *** 0,20% 1,34% *** 0,00% 0,48% ** 0,23% * -0,02% 0,69% *** -0,28% * 4,86% *** 0,25% **

2 sub-periods 1/94 - 12/96 1,20% *** 0,70% 0,57% *** 0,94% *** 0,26% * 2,56% *** 0,28% 0,10% 0,52% ** 0,20% 0,77% *** -0,28% 4,14% *** 0,29% * 1/97 - 6/00 0,26% 0,51% 0,41% ** 0,64% ** 0,19% 0,60% * -0,21% 0,49% 0,07% -0,25% 0,55% * -0,43% ** 5,57% *** 0,10%

Asian crisis 1/97-12/98 0,54% 0,66% ** 0,40% 0,65% ** 0,21% 0,44% -0,20% 0,22% 0,04% 0,13% 0,42% * -0,33% 4,38% *** 0,04% 1/97-6/98 1,20% *** 0,48% -0,05% 0,19% -0,18% -0,23% -0,42% 0,23% -0,33% 0,39% *** 0,16% -0,48% ** 3,79% *** -0,24%

Pre-1994 1/84 - 12/93 NA 0,13% 0,73% *** 1,13% *** 0,47% ** 1,70% NA 0,22% 0,06% 0,83% *** 1,10% ** 0,61% *** 3,84% *** 0,64% ***

This table reports a summary of the results for the estimation of the combined model for different time periods. We report monthly alphas for equally weighted portfolio's per investment strategies, substrategies, and for all funds. NA means that we could not make the estimation for the sub-period considered because the strategy has existed for less than 24 months. T-stat (not reported) are heteroskedasticity consistent. *** Significant at the 1% level, ** Significant at the 5% level, * Significant at the 10% level.

98

A closer look at the 1997-1998 suggests that most hedge funds strategies suffered during the Asian crisis. Nineteen strategies and sub-strategies out of the 29 face negative returns between 1997 and the middle of 1998, with five being significant (Global, Global International, Global Emerging, Market Neutral Fixed Income and Funds of Funds). Five strategies had significant positive returns during this period (Market Neutral Long/Short, Convertible Arbitrage, Non Classified funds, Relative Value Arbitrage and Foreign Exchange), but only the first three had also significant excess returns during other subperiods. The only sub-strategies that could have significantly benefited from the Asian crisis are the Relative Value Arbitrage and the Foreign Exchange. But one must be cautious because of the short estimation period. Moreover, the division in sub-periods indicates that over-performance is rarely sustainable over every shorter period of time: three strategies (Long/short Convertible Arbitrage and the Non Classified funds) out-performed the market in all sub-periods we considered; only the latter two extend their superior performance to the pre-1994 subperiod.

6.4 Comparison with other Studies Schneeweis and Spurgin (1998) and Liang (1999) find different results from us, but they are mainly due to differences in the period studied and to a smaller number of funds in their database.24 Agarwal and Naik (2004) find the same results as ours except that Fixed Income, Risk Arbitrage and Long only Leveraged strategies significantly underperformed the market on average, while we find only small percentages of underperforming funds. Finally, Agarwal (2001) finds results close to ours.

24

Liang (1999) finds insignificant positive excess return for the Convertible Arbitrage, Foreign

Exchange, Funds of Funds, Market Timing, Sector and Short Selling strategies but documents that Growth and Market Neutral strategies significantly under-perform the market.

99

VII Persistence in Performance
Our results show significant evidence of superior performance over long period of time for most individual strategies and sub-strategies. Nevertheless, results are not stable over shorter period of time, neither for hedge funds as a whole, nor for individual hedge funds. Active hedge funds selection strategies could increase the expected return on a portfolio if hedge fund performance is predictable. The hypothesis that hedge funds with a superior average return in this period will also have a superior average return in the next period is called the hypothesis of persistence in performance. Sirri and Tufano (1998) and Zheng (1999) stress the importance of persistence analysis in mutual funds. The formers document large inflows of money into last years best performers, and withdrawals from last year’s losers. The latter finds that newly invested money in these best performing mutual funds is a predictor of future performance.

7.1 Persistence in One-year Return-Sorted Hedge Funds Portfolios We follow the methodology of Carhart (1997) using our combined model. All funds are ranked based on their previous year return. Every January, we put all funds into 10 equally weighted portfolios, ordered from highest to lowest past returns. Portfolios 1 (High) and 10 (Low) are then further subdivided on the same measure. The portfolios are held till the following January and then rebalanced again. This yields a time series of monthly returns on each decile portfolio from 1/94 to 6/00.25 Funds that disappear during the course of the year are included in the equally-weighted average until their death, then portfolio weights are readjusted appropriately.

25

The corresponding table for the 1985-1993 period is available upon request.

100

Table 15: Hedge Funds Persistence based on 12 Month lagged Returns

Monthly Exc. Std. Dev. Return D 1a D 1b D 1c D1 D2 D3 D4 D5 D6 D7 D8 D9 D 10 D 10a D 10b D 10c 1-10 spread 1a-10c spread 1-2 spread 9-10 spread 1,68% 1,26% 1,11% 1,35% 0,89% 0,83% 0,82% 0,71% 0,57% 0,62% 0,57% 0,80% 0,89% 0,82% 0,84% 1,03% 0,45% 0,65% 0,46% ** -0,10% 0,07 0,05 0,04 0,05 0,03 0,03 0,02 0,02 0,02 0,02 0,02 0,02 0,03 0,03 0,03 0,04 0,05 0,07 0,02 0,02

Alpha

Mkt

Wd x US

SMB

HML

0,00 0,00 0,00 0,00 0,00 0,00 0,00 0,00 0,00 0,00 * 0,00 0,00 ** 0,00 0,00 0,00 0,00 0,00 0,00 0,00 0,00

0,74 *** 0,62 *** 0,55 *** 0,64 *** 0,55 *** 0,50 *** 0,47 *** 0,40 *** 0,32 *** 0,33 *** 0,36 *** 0,38 *** 0,50 *** 0,48 *** 0,54 *** 0,51 *** 0,13 0,24 0,08 -0,12 *

0,05 0,12 0,08 0,09 0,10 0,05 0,06 ** 0,05 * 0,06 ** 0,03 0,06 * 0,06 0,07 0,03 0,05 0,14 0,02 -0,09 -0,01 -0,01

0,66 *** 0,40 *** 0,39 *** 0,48 *** 0,33 *** 0,24 *** 0,17 *** 0,13 *** 0,13 *** 0,12 *** 0,13 *** 0,09 *** 0,04 0,09 ** 0,05 -0,03 0,45 *** 0,70 *** 0,15 *** 0,05

-0,03 -0,14 -0,02 -0,07 0,05 0,10 ** 0,09 *** 0,09 *** 0,09 *** 0,07 ** 0,04 0,00 -0,05 -0,05 0,01 -0,12 -0,01 0,09 -0,11 * 0,06

101

I HML D 1a D 1b D 1c D1 D2 D3 D4 D5 D6 D7 D8 D9 D 10 D 10a D 10b D 10c 1-10 spread 1a-10c spread 1-2 spread 9-10 spread -0,32 -0,29 *** -0,18 -0,26 * -0,12 * -0,05 0,01 -0,01 0,01 0,00 0,06 0,03 0,08 0,15 0,07 0,02 -0,35 ** -0,34 -0,14 -0,05

PR1YR 0,19 0,14 * 0,10 0,14 0,09 ** 0,03 -0,01 0,00 -0,03 -0,05 * -0,03 -0,05 -0,02 -0,02 -0,01 -0,04 0,16 0,23 0,05 -0,03

US Bd -0,93 ** -0,38 -0,30 -0,54 * -0,38 ** -0,14 -0,09 -0,22 *** -0,11 -0,05 -0,06 -0,07 -0,05 -0,12 -0,02 -0,01 -0,49 -0,92 * -0,15 -0,01

World GV Emerg Bd Bd -0,19 0,02 -0,18 -0,12 -0,09 0,03 0,00 -0,07 -0,04 -0,01 -0,08 -0,07 -0,09 -0,11 -0,06 -0,11 -0,03 -0,07 -0,03 0,02 0,12 0,05 0,03 0,07 0,06 0,03 0,03 0,07 *** 0,05 ** 0,05 ** 0,04 0,03 0,06 0,09 * 0,09 * 0,01 0,00 0,11 0,01 -0,03

Default 0,44 0,09 0,02 0,18 0,00 -0,02 -0,06 -0,04 0,07 0,04 -0,02 -0,12 -0,37 * -0,16 -0,34 -0,63 * 0,55 * 1,06 ** 0,18 0,24 *

Com. 0,13 0,11 ** 0,03 0,09 0,06 ** 0,01 0,01 0,01 0,00 -0,01 -0,01 -0,01 0,00 0,00 0,01 -0,01 0,09 0,14 0,03 -0,01

R²adj 0,74 0,81 0,79 0,79 0,85 0,88 0,87 0,84 0,82 0,80 0,75 0,65 0,44 0,61 0,44 0,22 0,52 0,48 0,43 0,04

This Table reports the result of the estimation of our combined model for the 1/94-6/00 sub-period. Each year, all funds are ranked based on their previous year's return. Portfolios are equally weighted and weights are readjusted whenever a fund disappears. Funds with the highest previous year's return go into portfolio D1 and funds with the lowest go into portfolio D10. Monthly Exc Return is the Monthly Excess Return of the portfolio, Std. Dev. is the Standard Deviation of the Monthly Excess Return. Mkt is the excess return on the Market Proxy. SMB, HML, IHML, PR1YR, W x US, US Bd, W Gvt BD, Emerg Mkt BD, Default and Comm. are factors added to adjust for size, book-to-market, one-year return momentum and for the fact that some funds invest in other than US equity market, in bonds not picked up by the risk-free rate or in commodities. All numbers in the Table are monthly percentage. *** Significant at the 1% level ** Significant at the 5% level * Significant at the 10% level.

102

The monthly average return to the strategy of investing in portfolio 1 would have been 1.35% for the 1/94-6/00 period. Conversely, the monthly (respectively maximum and minimum) return to the strategy that invested in the lowest decile would have been 0.89% over the same period. For the 1985-1993 period, decile portfolios 1 and 10 have earned 1.74% and 1.07% respectively. The monthly excess returns on the decile portfolios decrease monotonically between portfolio D1 and D6, but then increases again from portfolio D6 to portfolio D10. Monthly excess return of portfolio D2 is the same as the one of the last portfolio. The annualized spread is approximately 5.5% between portfolio D1 and D10. Portfolio D1a out-performs portfolio 10c by 0.65% per month. These spreads are not significant.26 Cross-sectional variation in returns is considerably larger among previous year’s best performing funds than previous year’s worst funds. The sub-portfolios of the bottom decile show a modest spread of 21 basis point (0.82 to 1.03), whereas the spread in the top decile is a substantial 57 basis point (1.68 to 1.11). The 1-2 spread is significant at the 5% level, indicating big differences between top performing funds and other portfolio funds, but the 1-2 spread alpha is not significant, suggesting no persistence. After controlling for the risk factors, a large part of the spread between high and low portfolios disappears. The 1-10 spread goes from a 0.45% to – 0.09%, the 1a and 10c spread decreases from 0.65 to –0.36%, still not significant, and the 1-2 spread reduces from a significant 0.46% to a 0.25% non significant spread.

26

Over the whole database, the D1-D10 and D1a-D10c spreads amount to 0.58% and 0.95%

respectively, both significant.

103

Column 7 suggests that all deciles portfolios (except the last) prefer small stocks. More important, however, is the pronounced pattern in the funds’ HML, PR1YR and Emerging Market Bond coefficients (Emerg Bd). First, portfolio D3 to D7 prefer stocks with high book-to-market ratios, whereas portfolios D1 and D10 prefer (but not significantly) those with low book-to-market. Second, returns of the top decile funds are strongly, positively correlated with the one-year momentum factor, while returns in the other deciles are not. Third, the returns of the D5 to D7 deciles are strongly, positively related with the Emerging Market Bond factor. This could explain why the monthly excess return diminishes after decile D3. Thus, the different financial crises covered by our time period may explain why funds investing in Emerging market Bond are not in the 3 best performing decile portfolio. Column 4 suggests that funds in top and bottom portfolios do not significantly outor under-perform the market, without persistence. This means that these funds are there more by chance or misfortune, rather than by their abilities. Moreover, the standard deviation of top and bottom deciles are the greatest of all, indicating more volatility in returns. Things are different for the funds in the middle deciles. Managers in these portfolios are more likely to stay in these deciles over time, and some of these managers significantly beat the market (deciles 7 and 9).27

27

These results are reinforced when one considers the global database and when we consider

strategies individually (see section 7.4). The qualitative results for the book-to-market and momentum strategies are unchanged, and evidence on an emerging bond market strategy for middle-decile portfolios is stronger. Alphas are significantly positive for funds in deciles 3 to 7 and 9. However, these values are mainly driven by the extremely good performance funds belonging to those deciles in the 1985-1993 period, which raises a legitimate suspicion over their economic significance.

104

this can be explained by dissolution or by the fact that they are more or less forced to take less risk. (2001). Table 15 leads to the following conclusions: Best performing funds follow momentum strategies whereas worst performing ones may follow momentum contrarian strategies.28 Best performing funds do not invest significantly in Emerging Market Bond. where some funds significantly beat 28 This result may not be stable over time. (2001) explain this behaviour of funds managers by the negative penalty linked to dissolution having a relatively stronger effect than the reward of the call-like feature of their compensation contract. After the strong market reversal that took place after March 2000 would be very indicative of the economic significance of this result. To summarize. even if some hedge funds managers take a lot of risk. whereas best and worst performing ones may prefer low book-to-market ones. 105 .This suggests that. These issues are being investigated as a part of our ongoing research. i. No persistence in performance exists for best and worst performing funds. which lead them to have very high or low returns for short period of time. Average return funds prefer high book-to-market stocks. Brown et al. with a positive alpha – tend to pursue a low variance strategy. Extreme performers tend not to remain in their situation. but our database unfortunately ends in June 2000.e. The stronger persistence for middle deciles with a substantially lower variance of returns than extreme deciles is in line with the conjecture of Brown et al. most managers follow less risky strategies that allow them to out-perform the market for long period of time. returning to the middle of the pack. but there is weak evidence of persistence for middle deciles. since the period under review covers a bull market when momentum strategies are likely to work. Funds that performed well – relative to the market.

but decile 1a is the only (sub)decile with positive (but not significant) alpha. In the 1/97-6/98 analysis. This indicates that some funds were not affected by the crisis. 106 . in 1997. but it is likely to be driven by the absence of dissolved funds in this period. and that there have more volatile returns than funds in lowest decile.2 Persistence over the Asian crisis The same analysis for the Asian crisis period shows that top performing funds of 1996 had significantly lower returns in 1997. may also explain these differences. probably because their investment strategies were relatively immune to it. etc. Evidence is more pronounced for the 1985-1993 period. 29 But one must be cautious with this result.the market with persistence. the Asian crisis in 1997-1998. 5 of them being significant.30 The alphas of decile D1 and the 1a-10c spread are significantly negative indicating that the best performing funds (of 1996) significantly under-performed the worst performing funds (of 1996). 30 Numerical results are available upon request. given that this is not the only reason possible. These results confirm our previous conclusion that funds in the first decile have no persistence in returns.29 7. all deciles’ alphas are negative. The Bond crisis in 1994.

This suggests that bad performance may be a major factor for dissolution. with the top and bottom decile divided in 3 for the Market Neutral strategy. This result is consistent the finding of no persistence in the best performing funds.4 One-Year Persistence for Hedge Fund Strategies This subsection focuses on the persistence in returns for some hedge funds strategies. Top 7 deciles have a more or less constant average rate of dissolution of 7%. we determine whether persistence in returns exists for these strategies. If a hedge fund ceases reporting returns at any time before the end of the year.3 Dissolution Frequencies Figure 4 shows a histogram of hedge funds dissolution frequencies in one year as a function of the previous year mean return decile. 7. Table 16 reports a summary of our results for these strategies. At the beginning of each year.7. We classify funds in 10 decile portfolios. but that persistence exists in the middle deciles. Results for the pre-1994 period are also indicated for comparison. In this sub-section. We test it for the 1994-2000 period and for the Asian crisis period (1996-97 and 19976/98). but that good performance is not a protection against it. it is 12. then this is counted as dissolved. 107 . We consider two strategies with more than 300 funds: Global Macro and Market Neutral that significantly out-performed the market for the 1/94-6/00 period. all hedge funds are put into decile rankings by their mean returns in the previous year.5% for the bottom 3 deciles.

00 1 2 3 4 5 6 7 8 9 10 Decile in Year t-1 Hedge funds dissolution frequencies by year t as a function of year t-1 decile.Figure 4: Hedge funds dissolution frequencies 0. They show that there is no significant difference between good and bad performing funds. or that they benefit from it. but not for the extreme deciles. At the beginning of year t. If a hedge fund ceases to report returns at any time before the end of year t. Panel A reports the results of the analysis for Macro funds. all funds are placed into decile rankings on the basis of their returns in year t-1. The 1996-97 Asian crisis sub-periods show that some funds (decile D3) have persistence in their return despite the Asian crisis. Half of the alphas obtained are significantly positive for the whole period. This confirms the previous findings for these funds. This could either mean that these funds returns were not affected by the crisis. it is counted as dissolved.08 0.12 0. 108 .04 0.16 Dissolution Frequency in Y 0. The second hypothesis is more plausible given that we are analyzing Macro funds that per definition anticipate market movements. These results are stronger than those obtained for the whole database.

Default and Commodity (positive) and World Government Bond (negative). This indicates that significantly positive difference existed between best and worst performing funds during the Asian crisis period. They are very close to those obtained for the whole hedge funds database in Table 16 but the results are stronger because we found significant alphas in some cases (recall that the Market Neutral strategy represents 712 funds that is 25% of our whole hedge funds database). but also that there has been a sharp reversal in strategies for this category of funds after 1994. In the previous period. but not the alphas. Those of the top decile are only significant for the 1985-93 sub-period. though they were insignificant in the all funds analysis: HML. There are some interesting differences between these numbers and those obtained for our whole hedge funds database (see Table 16). For the 1994-2000 period. This indicates that the results obtained for our whole hedge funds database may not be valid for one specific strategy of funds. 109 . their effect was also significant but with the opposite sign. The 1-10 and 1a-10c excess returns spreads are both significant in most of the period considered. four factors have a significant effect on spreads for the Market Neutral funds. except from the 97-6/98 period. Panel C reports Market Neutral’s spread decomposition for each of the sub-period considered. The only significantly positive alphas are those from the middle deciles.Panel B shows the results for the Market Neutral funds. As an illustration. with HML being replaced by IHML.

49% 2.82% ** 0.90% 1.03% 0.58% 3.83% 1.42% 1.07% 2.10% 1. Alpha 6.38% 3.55% 1.33% 2. Dev.03% * 3.55% 1.50% 5. Alpha Std.67% 2. Dev.98% 3.24% 6.61% 0.69% 2.13% 0.10% 3.95% -0.34% 0.37% 0.15% 3.44% Pre-1994 Period 1987-1993 Std.19% 2. Alpha Std.89% 2.43% 3.91% 3.93% 0.44% 1.66% 0.14% 1.29% 0.54% .24% 2.98% 3.54% 1997-6/1998 Mean excess Return Mean excess Return 1.06% 2.31% -0.62% 0.14% 3.84% 1.92% 2.86% *** 0.55% 4.88% 2.43% 1.99% 1.49% 1.46% 0.13% -0.12% 3.14% 6.Table 16: Hedge Funds Strategy Persistence based on 12 Month lagged Returns Panel A: Global Macro Whole Period 1994-2000 Mean excess Return D1 D2 D3 D4 D5 D6 D7 D8 D9 D 10 1-10 spread 1.43% 2.92% 2.79% 0.64% 1.12% 0.09% 0.77% * 0.39% 1.78% -0.64% Asian crisis 1996-1997 Mean excess Return 1.01% 1.36% 0.23% 5.48% 0.04% 1.36% *** 0.98% 1.38% 1.58% *** 0.28% -0.40% ** 0.06% 0.33% 0.57% 0.09% -0.17% 0.88% 1.39% 0.49% -0.49% -0.43% 1.01% 0.34% 0.16% 2.24% -0.51% 0.98% 6.84% 0.59% 1.13% 2. Dev.62% 1.03% 1.93% * 0.36% 2.14% 3.03% 3.21% -0.49% 1.97% 3.71% 0.20% 0.59% 0.63% * 0.33% 0.85% 0.73% 3.75% * -0.08% 2.32% 0.02% 2.37% 1.84% 0.66% 2.89% 1.35% 4. Alpha Std.57% 0. Dev.78% 3.81% 3.91% ** 1.22% 0.76% 0.46% -0.38% 1.05% 0.50% ** 0.07% 4.

37% 3. 4.29% 5.13% 0.00% 0.68% 3.56% 2. Dev.27% * 0.22% 0.04% ** 0.59% 0.16% 1.23% 0.91% 4.17% 1.05% 1.65% 0.04% 1.06% 1.59% * 0.59% 0.73% 0.98% 0.11% 1.63% 0.29% 2.86% 7.30% 1.42% 0.94% 1.27% 2.16% 0.98% 2.91% 1.12% 0.24% 0.10% 2.04% 1.45% 5.26% *** 1.20% 0.45% 1.22% 2.70% 3.33% Alpha 0. 4.95% 1.59% 0.15% 0.06% 3.78% 2.56% 0.07% 0.52% 0.Panel B: Market Neutral Whole Period 1994-2000 Mean excess Return D 1a D 1b D 1c D1 D2 D3 D4 D5 D6 D7 D8 D9 D 10 D 10a D 10b D 10c 1-10 spread 1a-10c spread 1.07% 0.97% 1.94% 1.19% 0.66% 1.86% 1.25% ** 0.65% *** 0.43% ** 1.60% 0.34% 0.42% 0.83% Pre-1994 Period 1987-1993 Std.22% 2.19% 0.21% 0.99% 0.15% 0.25% 1.43% 3.18% Std.55% 1.55% 0.04% -0.45% 0.42% 0.59% 0.10% 3.70% 0.37% 0.49% Alpha 0.56% 0.15% 0.08% 7.87% 3.30% 2. Dev.04% 0.63% 0.79% 1.31% *** 0.10% 3.83% 0.36% 1.21% 0.43% Mean excess Return 1.47% 111 .03% 2.54% 0.18% 0.60% 0.31% -0.75% 6.24% 0.90% 0.04% 0.95% -0.

56% 3.71% 0.31% 0. Dev.45% 1.14% *** 1.07% 0.95% 1.99% 0.67% *** 0.80% 0.15% -0.10% -0. 3.29% * Std.05% 1.06% -0.47% * 0.01% -0.38% -1.72% 0.90% 1.64% 1.60% 1.95% 1.82% 0.26% 0.23% 1.82% 1.52% 0.81% 0.82% Alpha -0.06% 1. Dev.48% -0.16% 0.70% 1.65% 0.61% 1.67% 0.09% -0.25% 0.05% *** 0.34% 0.60% * Mean excess Return 2.92% 0.54% -1.93% *** 0.26% 0.64% 0.77% 0.70% 3.65% 0.15% 1.28% -0.16% 112 .88% 0.60% 1.81% 1.52% 0.79% 3.21% -0.20% 0.16% 1.87% 2.81% 0.77% 0.02% -0.38% 0.19% 0.70% 0.20% 0.38% 1.48% 0.43% 0.26% ** 1.21% 1.13% Alpha 0.98% 0.61% 0.22% 0.Panel B (continued): Market Neutral Asian crisis 1997-6/1998 Mean excess Return D 1a D 1b D 1c D1 D2 D3 D4 D5 D6 D7 D8 D9 D 10 D 10a D 10b D 10c 1-10 spread 1a-10c spread 1.94% 0.48% -0.56% 1.34% 0.06% -0. 3.09% 0.18% 0.02% -0.86% 1.31% 1.15% 1.28% 2.38% 0.06% 0.23% 0.15% 0.45% 1.15% -0.69% 1.37% ** 0.67% 0.30% 1.90% 1.44% 1.47% * 0.72% 1996-1997 Std.98% 1.98% 1.39% *** 1.25% 0.

3 *** 0.05 0. Return Alpha Mkt W x US SMB HML I HML PR1YR 1994-2000 1-10 spr.41 1a-10c spr.17 * 0.39 *** 0.21% *** 1.03 1985-1993 1-10 spr.03 -0.06 0.Panel C: Market Neutral spread comparison Monthly Std.28 *** 1a-10c spr.13% 1.39% *** 1. 1.04 1a-10c spr.93 *** 0.29% * 3.18% ** 4.14% *** -0.60% * -0.13 -1. 1.91% 0.33% 0.16% 0.11 1997-6/1998 1-10 spr.82% -1.52 * 0.07 0.02 -0.04 0.27 ** 0.21% 0.5 .17 ** 0.66 0.31 0.47% 0. 0.04 1.6 -0.95% -0. 1.33 ** 0.1 -0.03 0.31 *** -0. 1. Exc.79% 0. Dev.1 -0.03 1a-10c spr.83% 6.41 *** -0.43 *** 0.43% 0.17 0.72% ** 3. 0.16 -0.84 ** 0.75% 0.08 0.71 *** 1996-1997 1-10 spr.63% ** 7.56% 1. 1.49% -0.46 -0.09 0.38% 0.26 0.16 *** 0.09 1.85 ** -0.3 * 0.15 1.55% * 2.36 0.38 ** -0. 1.03 ** 0.

75% 6.35 0.Panel C (continued): Market Neutral spread comparison Monthly Exc. 1a-10c spr.05 0.68 0.26 0. Portfolios are equally weighted and weights are readjusted whenever a fund disappears. W x US.99 * 1. US Bd. Std. Panel B contains the results for the Market Neutral funds.7 ** This Table reports the result of the estimation of our combined model.83% 2.16% 0.14 -0. 0. 1a-10c spr.34 0. Dev. 1996-1997 1-10 spr.36 ** -3. and Panel C contains a comparison of Market Neutral spread. SMB.72% ** 1.62 0.08 -0.07 1.04 0.67 *** -0.48 ** 0.72 ** 1.60% * 0.95% -0.79% 3.15 -0.33% 1.63 ** -0. one-year return momentum and for the fact that some funds invest in other than US equity market. are factors added to adjust for size.18% ** 1.43 0.05 -0.55% * 1.09 -0.85 ** -0.21% *** 1. Each year.39 ** 0.13% 1. 114 . HML.38 * 0. 1997-6/1998 1-10 spr.38% -1.19 -0. PR1YR.4 0.39% *** 1.37 -0. 1a-10c spr.45 *** 1.44 * 2. W Gvt BD. All numbers in the Table are monthly percentage.63% ** 0.43% 1. Emerg Mkt BD.56% 3. R²adj 0.12 -0.67 * 0.94 -2.74 *** -1.42 -1.47% -0.85 -0. Panel A contains the results for the Global Macro funds for different time period.49% Alpha US Bd W GV Bd Emer Bd Default Comm.47 0. Monthly Exc Return is the Monthly Excess Return of the portfolio. Return 1994-2000 1-10 spr. in bonds not picked up by the risk-free rate or in commodities. 1a-10c spr.21% 0. Mkt is the excess return on the Market Proxy.23 *** -0. Std.54 ** -0. book-to-market.91% 4. Default and Comm.01 -0. Dev. all funds are ranked based on their previous year's mean return.14% *** 1. 1985-1993 1-10 spr.29% * 1.54 ** 0. is the Standard Deviation of the Monthly Excess Return.16 ** -0.21 *** -0.12 0.39 -0.24 -0. IHML.21 * -0.82% 7.34 * -0.41 -0.

Global Macro. the analysis of the survivorship and instant history biases for the 1984-1993 period indicates that serious concerns can be raised about the statistical reliability of the observations reported in this time window. combined with Fama and French (1998). our combined model does a better job describing hedge funds behaviour. We demonstrate that our combined model is able to explain a significant proportion of the variation in hedge fund returns over time. Therefore. we investigated hedge funds performance using various assetpricing models including an extension form of Carhart’s (1997) model. Although our sample displays funds returns going back to 1984. This corresponds to the lack of data about funds that were dissolved prior to 1994 in the databases. Compared to models used in other hedge fund performance studies. 115 . It appears particularly good for the Equity non-hedge. Sector. with a possible reinforcement with results for the 1984-1993 period whenever similar. our conclusions can only be derived for the 1994-2000 period.VIII Conclusion Using one of the greatest hedge fund database ever used (2796 individual funds including 801 dissolved). Long only Leveraged and Event Driven funds. US Opportunistic. Agarwal and Naik (2004) models and an additional factor that takes into account the fact that hedge funds may invest in Emerging Market Bonds.

constant over time. average return funds prefer high book-to-market stocks. and that many hedge funds invest in Emerging Market Bonds. in most cases. on the one 116 . whereas lower decile funds do. Secondly. best performing funds follow momentum strategies whereas worst performing ones tend to be momentum contrarian. they significantly under-performed the worst performing funds (of 1996) during the crisis period. according to our results. The analysis of persistence in performance for our whole hedge funds database indicates 3 main findings that complete our results of the performance analysis. Our main conclusion from this evidence is that there is no persistence in performance for best and worst performing funds. Moreover. which lead them to have very high or low returns for short period of time. Firstly. This suggests that. Analyzing each strategy individually lead to the additional conclusions that 10 out of 13 strategies offer significantly positive excess return. whereas best and worst performing funds may prefer low book-to-market ones. The Asian crisis analysis suggests that top performing funds in the year before the crisis (1996) had significantly lower returns in the first year of the crisis. most funds suffered from the Asian crisis. even if some hedge funds managers take a lot of risk. A sub-period analysis showed that over-performance ability is. But.The performance analysis shows that one fourth of individual hedge funds delivers significant positive excess returns. that most of them seem to prefer smaller stocks. best performing funds do not significantly invest in Emerging Market Bond. but limited evidence of persistence exists for middle decile funds. The much stronger results obtained for data before 1994 raise additional suspicion about the presence of severe bias in this part of the database. Thirdly. many hedge fund managers follow less risky strategies that allow them to out-perform the market for longer period of time. The analysis of dissolution frequencies indicates.

as found by Fung and Hsieh (1997. 117 . and on the other. the study of two popular strategies that significantly out-perform the market confirms that there is a persistence in positive excess returns for middle decile funds. that bad performance may be a major factor for dissolution. The next step in the development of model that describes hedge funds behaviour. and that specific models could be developed for each investment style in the future. This stresses that funds in different strategy are not influenced by the same factors. including Trading Strategy (option writing/buying) factors.hand. A further analysis of the results for the Market Neutral strategy indicates that the differences in spread are not explained by the same factors as for the whole hedge fund database. that very good performance is not a protection against dissolution. All these results help in understanding the performance of hedge funds and the persistence in this performance over different time period. Finally. including a study of the Asian crisis. in order to account for the fact that some hedge funds exhibit non-linear option-like exposures to standard assetclasses. 2001).

.

Maastricht University. 2005. and Luxembourg School of Finance. Limburg Institute of Financial Economics. Georges Hübner and Albert Corhay. Hedge Funds Performance and Persistence in Bull and Bear Markets. and Luxembourg School of Finance. 361-392 . CAPOCCI HEC-ULG Management School – University of Liège (Belgium) Albert CORHAY HEC-University of Liège. University of Luxembourg.Study 2: Hedge Fund Performance and Persistence in Bull and Bear Markets Daniel P. Daniel. European Journal of Finance 11/5. Capocci. Maastricht University.J. University of Luxembourg. Georges HÜBNER HEC-University of Liège. Limburg Institute of Financial Economics.

despite the high attrition rate of funds observed in the down market. In contrast. The analysis of performance indicates that most hedge funds significantly out-performed the market during the whole test period. The analysis of persistence yields very similar results.Hedge Fund Performance and Persistence in Bull and Bear Markets Abstract This study tests the performance of 2894 hedge funds in a time period that encompasses unambiguously bullish and bearish trends whose pivot is commonly set at March 2000. with most of the predictability being found among middle performers during the bullish period. We apply an original ten-factor composite performance model that achieves very high significance levels. as abnormal performance is sustained throughout and significant persistence can be found between the 20% and 69% best performers in this category. the Market Neutral strategy represents a remarkable exception. probably due to an extreme adaptability and a very active investment behaviour. However. no significant under-performance of individual hedge funds strategies is observed when markets headed south. mostly thanks to the bullish sub-period. Our database proves to be fairly reliable with respect to the most important biases in hedge funds studies. 120 .

Kat and Amin. but has enjoyed since then a renewed vitality with an estimated size of 8400 funds managing $900 billion assets in 2004 (Van Hedge Funds Advisors International. and finished three years later with a floor level at 1253. corresponding to growth rates of respectively 10. This relatively positive attitude of investors is typically motivated by the perceptions that hedge funds are largely market neutral. 121 . the subsequent growth of number and asset base of hedge funds has never really been refuted. 2004.8% and 21. when around 2000 hedge funds were managing together assets of ca. 2002). 1997).Hedge Fund Performance and Persistence in Bull and Bear Markets I Introduction Since 1990. 2003).6% (Van Hedge Funds Advisors International.3% respectively. 2003b). The growing trend of the sector could be remarkably sustained during the stock market collapse that started in March 2000. making them a good diversification vehicle (Schneeweis and Spurgin. that their managers enjoy greater flexibility in their asset allocation that enables them to achieve a better market timing (Fung and Hsieh. $ 60 billion. the global net asset value (NAV) of hedge funds continued to grow at a steady 10. In the meantime.7% in worldwide mutual fund industry (Investment Company Institute. Capocci and Hübner. The industry only suffered from a relative slowdown in 1998. or that hedge funds have a relatively low covariance with other classes of financial assets. contrasting with a decrease of 2. when the NASDAQ Composite Index reached an all-time high of 5132. 2002). 1998.

More recently. as periods of downward trends on the stock market were rare and discontinuous. On the other hand. Event Driven and Macro) provide protection to investors when stock markets head south. as most empirical evidence reveals that data collected prior to 1994 by several data vendors displays a significant survivorship bias. their superior performance is mostly due to the good market timing of their managers during the US stock market bubble that preceded it. Ackermann and al. Agarwal and Naik (2000) find evidence of persistence in hedge funds performance. (1999) conclude that there is hardly any evidence of the existence of differential manager skills but persistence is rather due to style effects. Liang. rather. Liang (2000) and Capocci and Hübner (2004). as shown by Fung and Hsieh (2000). Few authors have attempted to estimate the behaviour of hedge funds in bear markets. In this context. although lower and more volatile than the reference market indices considered. hedge funds did not provide investor protection after the market downturn of March 2000.. Edwards and Caglayan (2001) find that only three hedge fund strategies (Market Neutral. the issue of persistence in performance is particularly important in the case of hedge funds because these funds experience a greater attrition rate than mutual funds (Brown and al. 2001. (1999) and Liang (1999) find that hedge funds constantly obtain better performance than mutual funds. This is generally due to the particularly bullish period corresponding to the time window under review.The vast majority of performance studies on hedge funds has not focused on their behaviour under different market conditions. 122 . For the period 1990-1998.1999. while Capocci and Hübner (2004) sustain that it can be mostly found among average performers and Brown and al. The periods under review do not favour this exercise. Ennis and Sebastian (2003) contend that in general. 1999).

Edwards and Caglayan (2001) and Liang (2003). Kao et al. Rao (2001). Our study will thus identify and separately analyse two sub-periods corresponding to upward and downward market trends. that preclude any analysis of persistence. Carhart (1997). This yields a model with 10 risk premia. nor from arbitrarily chosen definitions of a bear market. Firstly. which may look at first sight overspecified but one has to bear in mind that hedge funds families are very heterogeneous and. successively proposed by Fama and French (1993). Thus. In this study. with a pivot set at the end of March 2000. we introduce a modified asset pricing model encompassing the risk premia that proved to be relevant for assessing funds performance in previous studies. (1998). our analysis neither suffers from discontinuities between down periods. This study introduces several key modifications with respect to the previous studies on hedge funds performance and persistence. Secondly. we study the performance of hedge funds and its persistence during a time window that encompasses relatively long bullish and bearish periods.This study benefits from the fact that stock markets have experienced a long period of depression. involve investments in many types of assets and markets. since stock indices have been almost continuously going down for a period of three years. although Capocci and Hübner (2004) and Liang (2001) consider a relatively short bearish sub-period with the Asian crisis. as Fabozzi and Francis (1979). 123 . using a similar methodology as the one developed by Capocci and Hübner (2004). Agarwal and Naik (2004) and Capocci and Hübner (2004). their sample does not enable them to distinguish between unambiguously bullish and bearish sub-periods. unlike mutual funds.

Thirdly. Section 6 documents and explains the persistence in hedge fund returns over the same time windows. we provide a thorough analysis of the database. The fourth Section studies potential biases in the database. The study is organised as follows. It has also been studied in an unambiguously bullish setup by Capocci and Hübner (2004). following the results of Edwards and Caglayan (2001). with a special focus on the Market Neutral strategy. In Section 3. we perform an in-depth analysis of the level and persistence of its performance before and after the stock markets downturn. who find that this family of funds tends to out-perform the market during the 1994-2000 period. namely Market Neutral that. The next Section reports the performance of hedge funds for the whole period and the sub-periods considered. we specifically identify one hedge fund strategy. Section 2 sets out the performance models we will use. supposedly hedge investors against bearish markets. In this study. 124 . Section 7 concludes the study.

II Performance Measurement Models The starting point of our study of hedge funds performance is the original Sharpe (1964) – Lintner (1965) CAPM.. T (6) where RPt = return of fund P in month t. RFt = risk-free return on month t. εPt = error term. we use the Carhart (1997) model as it is that is widely used in practice and it is not dominated by any other model in the mutual funds performance literature.. RMt = return of the market portfolio on month t.1 The Capital Asset Pricing Model The first performance model we use is a single index model based on the classical CAPM developed by Sharpe (1964) and Lintner (1965).. 2. As the basic multi-factor specification. 125 . Finally.2. we construct a multifactor model that extends the Carhart (1997) specification by combining it with factors proposed in Agarwal and Naik (2004) and Capocci and Hübner (2004) and by adding an additional factor. respectively. αP and βP are the intercept and the slope of the regression.. Its equation to estimate is the following: RPt − RFt = α P + β P (RMt − RFt ) + ε Pt t = 1.

.2 The 4-Factor Model of Carhart (1997) The four-factor model of Carhart (1997) is an extension of the Fama and French (1993) 3-factor model. For a description of the construction of PR1YR see Carhart (1997).. HMLt = the factor-mimicking portfolio for book-to-market equity (‘high minus low’)31 and PR1YRt = the factor-mimicking portfolio for the momentum effect.2. but also an additional factor for the momentum effect. αp commonly called Jensen’s alpha (1968) is usually interpreted as a measure of out.32 These factors aim at isolating the firm-specific components of returns.or under-performance relative to the market proxy used. Titman and Wermers (1995) define this effect as buying stocks that were past winners and selling past losers. This model is estimated with the following regression: RPt − RFt =α P + β P1(RMt − RFt )+ β P2 SMB t + β P3 HML t + β P4 PR 1 YRt +ε Pt t =1. 31 See Fama and French (1993) for a precise description of the construction of SMBt and HMLt..T (7) where SMBt = the factor-mimicking portfolio for size (‘small minus big’)..The intercept of this equation. 32 126 . Grinblatt. It takes into account size and book-to-market ratio. 2.

Fama and French (1993) "size" and "value" factors. namely the Lehman Mortgage-Backed Securities Index to take into account the fact that various hedge funds strategies (fixed income arbitrage.2. 127 . two factors to account for the fact that hedge funds invest in US and foreign bond indices33 (Lehman High Yield Bond Index and Salomon World Government Bond Index) and one factor that Capocci and Hübner (2004) proved to be highly significant. mortgage-backed securities) are exposed to this market and the Lehman High-Yield Credit Bond Index. the JP Morgan Emerging Market Bond Index. we implement a combination and an extension of Carhart’s (1997) 4-factor model. Amex and NASDAQ stocks market proxy that is usually used in mutual funds performance studies. five factors introduced by Agarwal and Naik (2004): a factor for non-US equities investing funds (MSCI World excluding US). and finally a commodity factor (GSCI Commodity Index). Several additional factors.3 The Composite Model In order to take into account the complex characteristics of the hedge fund industry. was found to have an extremely high correlation with the Lehman BBA Corporate Bond Index and thus was removed from our study. the model used by Agarwal and Naik (2004) and the one used by Capocci and Hübner (2004). Carhart’s (1997) "momentum" factor. the Lehman BAA Corporate Bond Index and the Salomon Brothers Government and Corporate Bond Index proposed by Agarwal and Naik (2004) and Capocci and Hübner (2004) and the Gold 33 The Lehman US Aggregate Bond Index. The market proxy used is the value-weighted portfolio of all NYSE. such as the MSCI Emerging Markets Index. This model contains the market risk premium. that was used in several previous hedge funds studies. we add an additional bond index factor that is not used in previous studies. Furthermore.

34 Agarwal and Naik (2003) suggest that the Goldman Sachs Commodity index is a better approximation of the commodity market as the Gold index regarding hedge funds. JPMEMBIt = return of the JP Morgan Emerging Market Bond Index. and GSCIt = return of the Goldman Sachs Commodity Index. HYt = return of the Lehman High Yield Credit Bond Index et MORTt = return of the Lehman Mortgage-Backed Securities Index. SWGBIt = return of the Salomon World Government Bond Index. 128 .index used by Fung and Hsieh (1997) were not included in our extended model given their high colinearity with our set of indices.34 R Pt − R Ft = α P + β P1 (R Mt − R Ft ) + β P 2 SMB t + β P 3 HML t + β P 4 PR1YRt + β P 5 (MSWXUS t − R Ft ) + β P 6 (SWGBI t − R Ft ) + β P 7 (JPMEMBI t − R Ft ) + β P 8 (HYt − R Ft ) + β P 9 (MORT t − R Ft ) + β P10 (GSCI t − R Ft ) + ε Pt (8) where RMt = return on the Russel 3000 Index. MSWXUSt = return of the MSCI World Index excluding US.

management and incentive fees. HFR defines sixteen different strategies in two categories. ‘Hedge Fund Research. manager's name etc. start and ending date. they record other useful information such as company name.35 Data vendors do not only collect performance data. TASS defines 15 strategies.III Data 3. Finally. (1999) and Capocci and Hübner (2004) used a combination of HFR and MAR while Liang (2000) uses a combination of TASS and HFR. We use 108 monthly returns on 2894 individual hedge funds plus 48 indices (16 investment styles with 3 indices for each investment style: onshore. MAR defines 9 strategies with a total of 16 sub-strategies. 2003a). and ‘TASS Management’ (TASS) (Amin and Kat. The database gives monthly net-of-fee individual returns and other information on individual funds and groups them in indices. These funds include 1622 funds alive at the end of the period (56%) and 1272 dissolved funds (44%). as in Fung and Hsieh (1997).’ (HFR). but Ackermann and Ravenscraft (1998) and Ackermann et al. 129 . strategy followed. There is no consensus on the definition of the strategy followed but there are similarities. assets under management. offshore and a combined index).1 Database Three main hedge fund databases are available for empirical studies: ‘Managed Account Reports’ (MAR). Schneeweis and Spurgin (1998). plus the Funds of Funds and the Sector categories. For a majority of funds. and Amin and Kat (2003a). These databases are the most used in academic and commercial hedge fund studies. Inc. We use hedge fund data from MAR. 35 The three databases have never been used together in a study. 12 non-directional and 5 directional strategies.

Hedge funds are classified in two categories.g. Global Intern.2 Basic Performance Panel A of Table 17 contains descriptive statistics of the funds. Short Sales and No Category. Macro. Long Only Leveraged. 1997). Carhart. Global Emerging.36. See Capocci and Hübner (2004) for more a complete analysis of this correlation. 1996. The Funds of Funds category features 3 strategies: Niche. These hedge funds data are contrasted against the descriptive statistics of the factors introduced in equation (3) of Section 2.37 Finally.. 3. Amex and NASDAQ stocks market proxy that is usually used in mutual funds performance studies (see e. The Individual Funds category features 13 strategies: Event driven – Risk Arbitrage. whether living or dead. the latter one corresponding to funds with no stated strategy and funds whose strategy does not fill in any of the above. 37 130 . 36 This strategy has been suppressed in 1999. Diversified and Others.. Its almost perfectly correlation with the Russell 3000 index used in Agarwal and Naik (2004) suggests that both market proxies are very similar. 1993. US Opp. the one-month T-bill rate from Ibbotson Associates is taken as the riskfree rate. We take the value-weighted portfolio of all NYSE. Mkt Neutral. Fama and French. Sector. Global. Event-Driven – Distressed Securities. These statistics are reported in panel B of Table 16. in our database. Global Est.

Short Sellers the and Funds of Funds. The mean excess premium of the MSCI World excluding US is an insignificant 0.Panel A shows that the highest mean return was achieved by the Sector (1. The average SMB and HML returns are insignificant.70%. Panel B of Table 16 shows that the mean excess return of the Market Proxy is 0. The highest mean return for bond indices was obtained by the JP Morgan Emerging Market Bond Index. US Opportunistic Growth and Sector while the worst average performers are Foreign Exchange. This reasonable value indicates that the bullish sub-period has been almost totally offset by the market correction.29%) and Global Emerging (1. with the highest mean. Market Neutral funds appear to be the best performers.66%). without sub-strategy. only statistically different from zero at the 10% level.22% per month. only the Momentum factor. provides a significantly positive value. These descriptive statistics differ from the results obtained by Capocci and Hübner (2004) for the 1994-2000 period. The Sharpe measure (the ratio of excess return and standard deviation) offers a much different picture: accounting for risk.5% per year). This difference can be explained by the difference in the database used (MAR combined with HFR for Capocci and Hübner.78% per month (about 9.17%) follow. This pattern is similar for the mean excess returns. 131 . then the Global Est. only followed by the Global (0. Average returns of funds of funds are all around 0. who find that the best performers are US Opportunistic Small Caps. 2004) and the different time period studied. while the funds that achieve the highest absolute returns are only among the average risk-adjusted performers.45%) strategy that achieves the lowest mean return. (1. A look at the t-stats indicates that mean returns are significantly different from 0 at the 5% significance level for all funds and that the mean excess returns are significantly positive for all cases but the Global and Diversified funds of funds categories.

132 . During the up period. presented in Table 17. the monthly index return was positive in 70% of the months (52 out of 74) with an average yearly return of 19. respectively.48 during the session of March 24.The Sharpe ratio obtained by our whole hedge fund database (0.19).4%. During the down period.29) is higher than the ones for the Market Proxy (0. This month corresponds to the maximum observed value of the Russell 3000 Index that reached a value of 858. bearish and neutral months since these rules would obviously not match the ones used by funds managers for their market timing decisions.15). bullish and bearish without having to use a complex rule to separate bullish. the monthly index return was positive in 39% of the months (12 out of 34) and the average yearly return was 16. Those trends are sufficiently strong to allow us the consider the whole sub-periods as. reveals some interesting differences. The analysis of basic performance for the two sub-periods under review. 2000. and higher than for the MSCI World Excluding US (0.9%. 3.3 Analysis per Sub-periods The cutting point chosen for the identification of the up and down periods has been set at March 2000.

11% 0.97% 5.Table 17: Descriptive Statistics of Hedge Funds Strategies and Passive Investment Strategies Panel A: Hedge Funds strategies Jan.15% 0.Other Funds of Funds Total Total 136 106 175 499 72 157 39 144 635 33 190 37 24 2247 114 501 32 647 2894 % of the category 6.82% 1.50% 2.15 4.10% 4.14% Med.1994-December 2002 (108 months) Nb of Fds Event driven .11% 133 .99% 0.80% 3.49 1.23% 0.20% 3.99% t(mean) Std.80% 0.17% 0.98 4. Dev.17% 0.75% 0.77% 0.20% 7.60% 1.** Macro Market Neutral Long Only Lev.17 4.93% 1.40% 2. Global Emerging US Opp.87% 1.71% 0.08% 0.96 5.61% 1.Niche FoF .95 4.47 7. 0.80% 22.87 6. Sector Short Sales No category Individual Funds Total FoF .70% 6.48% 3.06 9.00% 1.3 4.92% 1.67% 0.66% 0.87 5.48 2.58% 1.93% 0.30% 1.74% 0.69% 1.48 1.70% 7.45% 1.77% 2.27 3.71% 1.04% 0.73 5.21% 0.42% 5.67% 0.1 3.02% 2.64% 1.15 5.29% 0.92% 1.25% 2.72% 0.07% 0.52 5.40% 28.88% 0.50% 1.33% 1.Diversified FoF .93% 1. 5.05% 1.25% 3.28% 1.43% 4.50% 8.Risk Arb Event driven Distressed Sec Global* Global Est.45% 2.09% 0.88% 1.80% 0.12 5. Global Intern.10% 100% 18% 77% 5% 100% 100% Living Funds 85 70 1 300 46 97 0 52 385 16 111 24 6 1186 86 349 1 436 1622 Dead Funds 51 42 174 199 26 60 39 92 250 17 79 13 18 1061 28 152 31 211 1272 Mean Return 0.74% 2.83% 4.

And Global Emerging categories. the funds disappearance is mostly due to category transfers.64% 1.2368 2.08 0.30% 8.38 0.28% 0.27 0.90% 13.74 3.92% 1.73 3.61% t(mean exc.55% 0.09 0.01 3.80% -8.04 2.11 -0.31 0.80% -9.00% 5.37 5.62 2.36% 0.7 0.18% 0.71% 1.02 0.) 3.75% 0.45 -0.85 3. Event driven .14 0.05% 1.58% 1.26 0.50% 2.59 1.74 0.67% 0.29 0.07% 0.32 0.74 0.34 3.60% -0.30% 19.25 21.18 0.77 1.22 -0.57 1.41 2.86 3.1994-December 2002 (108 months) Med.39% 0. ** US Opportunistics ended in 1999.00% 13.30% -6.16 2.96 7.11% -6.Panel A (continued): Hedge Funds strategies Jan.60% 12.16 0.60% -4.42 3.07% 0.11 0.05 0.44% 0.79% -0.09 1.00% 13.30% 7.34% 0.16 -2.38 Sharpe ratio 0.17% 0.80% -25.16 -0.10% 7.40% 7.61% 1. 134 . Sector Short Sales No category Individual Funds Total FoF .44 0.26% 0.87 1.66% 0.80% 0.11 0.44 3.55% 0.70% 8.00% 4.93% 1..27 0.67 0.54% 1.60% -7.31 0.40% -10.30% -7.00% 5.61 1.10% -7.19 0.2 2.06 2.58 -1.2 0.16 1.90% -6.28 0.70% -5.77 4.70% 0.54 4.09% 0.21 0.90% 14.05 3.Diversified FoF .60% 7.67% 0.20% 5.29 0.60% -4.50% 0. Therefore. Global Emerging US Opp.99 1.Niche FoF .20% 12.80% 0.23 0.50% 0.37 -1 2.21% 0. Kurt Excess return 0.76 3.** Macro Market Neutral Long Only Lev. Global Intern.09 0.80% -21.04 The Global category has been gradually suppressed and replaced by the Global Est.Risk Arb Event driven Distressed Sec Global* Global Est.Other Funds of Funds Total Total Min Max Skew.22 -0.10% -2.24 -0. Global Intern.12 -0.66 0.50% -17.80% 6.91% 0.89 -0.10% -13.17 -0.24 6.50% 7.61% 0.40% -13.11% 0.80% -8.

35 0.74 1.5 2.37 34. Median Market Proxy 0.Panel B: Passive Strategies Mean Return Equity Factors t(mean) Std.59 6.4 1.36 F&F HML Factor 0.14 5.66 Commodities Goldman Sachs Commodity 0.11 0.36 0.24 JPM EMBI Global 0.27 Bond Factors 1 month T-bill 0.4 Salomon WBGI 0.11 0.14 2.7 4.99 2.58 MSCI World Excluding US 0.78 1.67 1.89 0.59 1.04 4.5 4.82 1.22 4.83 1.68 Momentum Factor 1.02 0.16 0.06 5.81 4. Dev.77 1.6 1.47 F&F SMB Factor 0.83 0.61 135 .09 0.16 Lehman Mortgage 0.94 0.45 -0.69 0.66 Lehman High Yield Credit 0.

99 2.18 0.03 NA 0. maximum.9 10.28 15.15 -0.6 -7.9 3.07 0. US Opp. Living Funds and Dead Funds represents the number of surviving and dead funds (in December 2002).9 -16. mean excess returns.) Sharpe ratio This Table shows the mean returns.41 -0. No of Fds represent the number of funds following a particular strategy (or sub-strategy). medians.22 0.14 0.1 -3.Panel B (continued): Passive Strategies Min Equity Factors Market Proxy MSCI World Excluding US F&F SMB Factor F&F HML Factor Momentum Factor Bond Factors 1 month T-bill Salomon WBGI JPM EMBI Global Lehman Mortgage Lehman High Yield Credit Commodities Goldman Sachs Commodity -12.4 -24.6 5.2 0.82 0.67 -0.35 0.9 -0.77 0.9 -25.23 0. minimum.22 1.39 0.1 8. Sharpe ratio is the ratio of excess return and standard deviation with a risk-free rate set at 5%.44 0.2 0.21 2.72 0. t-stat for mean = 0.7 18. We calculate the Mean Excess Return considering Ibbotson Associates one-month T-bills. standard deviation. Numbers in the table are monthly percentage.3 21.08 -15.3 -8.17 NA 0.2 -2.24 Max Mean exc. Return t(mean exc.99 0.13 0.79 0. 136 .39 0.66 0.4 13.58 1.3 10.19 -0. In panel A.2 7.37 0. Funds ended in 01:1999.7 -12.49 NA 0. t-stat for mean excess return = 0.28 -0. and Sharpe ratios for the individual hedge funds in our MAR database for the whole period 01:199412:2002.

Finally. Preliminary evidence does not seem to indicate that the behaviour of the two Event Driven and of the Macro strategies out-perform the other ones in the bearish period. Secondly. with the best performers before March 2000 also displaying the worst returns after the market reversal too place. while the returns of the Market Neutral strategy are then significant but not when excess returns are considered. this does not support the findings of Edwards and Caglayan (2001) with a different definition of bearish market conditions. Although further evidence is obviously needed. 137 .Table 19 displays summary results for the bullish and bearish sub-periods. the US Opportunistics strategy did poorly in spite of favourable market conditions. Firstly. which explains the disappearance of this category. excess returns obtained for the majority of hedge funds strategies are mostly due to the bullish sub-period. There are three noticeable exceptions. the Short Sales strategy is the only one that records significant excess returns during bad times but at the expense of insignificant returns in good times. As expected from the nature of the time windows. the Global strategy seems to achieve returns that are much less dependent on the conjuncture than the other strategies.

Table 18: Descriptive statistics of hedge funds strategies for the bullish and bearish sub-periods Panel A: Sub-period 01:1994-03:2000 Nr of Fds Event driven . Macro Market Neutral Long Only Lev. Global Emerging US Opp.Diversified FoF . Global Intern.Dist. Sector Short Sales No category Individual Funds Total FoF .Risk Arb Event driven .Niche FoF . Sec Global Global Est.Other Funds of Funds Total Hedge funds Total 113 83 175 400 64 132 39 121 487 31 147 34 19 1845 81 408 31 520 2365 Living Funds 89 65 6 342 58 108 0 63 350 22 126 25 9 1263 57 333 18 408 1671 Dead Funds 24 18 169 58 6 24 39 58 137 9 21 9 10 582 24 75 13 112 694 138 .

06%*** 0.36 0.10%*** 1. Global Intern.35 0.3 0.47%*** 0.Diversified FoF .57%*** 0.Other Funds of Funds Total Hedge funds Total 1.27%*** 0.68%*** 0.85%*** 0.95%*** 1.57%** 1.16% 1.08 0.26% 0.96%*** 0.42 139 .42%*** 2.16%*** -0.89%*** 1. Sector Short Sales No category Individual Funds Total FoF .54%*** 0.23%*** 1.18% 0.51%** 0.26 0.18%*** 1.12 0.35 0.48 0.92%*** 0. Sec Global Global Est.54%** 0.28 0.83%*** 2.Panel A (continued): Sub-period 01:1994-03:2000 Mean Return Event driven .30%*** 1. Macro Market Neutral Long Only Lev.35%*** Excess return 0.92%*** 0.46 0.45 0.26 0.Dist.04 0.93%*** Sharpe ratio 0.50%** 0.5 0 0.56%*** 0.77%*** 1.21 -0. Global Emerging US Opp.2 0.82%*** 0.14%*** -0.39%* 1.75 0.21%*** 1.Risk Arb Event driven .23% 1.Niche FoF .96%*** 1.80%*** 1.02% 0.

Sector Short Sales No category Individual Funds Total FoF . Global Emerging US Opp. Global Intern.Risk Arb Event driven .Diversified FoF . Macro Market Neutral Long Only Lev.Panel B: Sub-period 04:2000-12:2002 Nr of Fds Event driven .Niche FoF .Dist. Sec Global Global Est.Other Funds of Funds Total Hedge funds Total 112 94 6 441 66 133 NA 86 498 24 169 28 14 1665 90 416 19 535 2214 Living Funds 85 70 0 300 46 97 NA 52 385 16 111 24 6 1186 86 349 1 436 1622 Dead Funds 27 24 6 141 20 36 NA 34 113 8 58 4 8 479 4 77 18 99 578 140 .

04 0. Sec Global Global Est.07 This Table shows the number of funds following a particular strategy (or sub-strategy). Sharpe ratio is the ratio of excess return and standard deviation with a risk-free rate set at 5%. the corresponding number of living and dead funds.0012 0.Panel B (continued): Sub-period 04:2000-12:2002 Mean Return Event driven .16 0.27% NA 0.0003 -0.0038 1. mean excess returns.71%*** -1.0012 0.17%*** -0.0006 0.14 -0.0008 0.07 0.0033 0.13 0.Dist. Global Emerging US Opp.05 -0.33 -0.01 NA -0.0019 0.0018 0. mean returns.99%*** 0.0016 -0.0019 Excess return -0.52 -0.48%*** -0.0025 0. ** statistically significant at the 5% level.52%** -0.23 -0.22 -0. Global Intern.0012 Sharpe ratio -0.0015 0. and Sharpe ratios for the individual hedge funds in our MAR database for the sub-periods 01:1994-03:2000 (bullish) and 04:2000-12:2002 (bearish).0019 0.44%* 0.0013 -0.004 -1.04% NA -0. Sector Short Sales No category Individual Funds Total FoF .0021 -0.68%*** -0.Niche FoF .0028 0. 141 .11 -0.0016 0.06 0.*** statistically significant at the 1% level.0003 -0. Macro Market Neutral Long Only Lev.0037 0.Risk Arb Event driven .69%**° 1. * statistically significant at the 10% level.Diversified FoF .03 -0.0039 -0.0014 -0.06% -0.Other Funds of Funds Total Hedge funds Total 0.0014 -0.01 -0.19 -0.03 -0.

4 Correlations The traditional hedge funds literature contends that. we chose to report ranges in correlations. This sub-section studies the ranges of correlation coefficients among and between hedge funds and passive investment strategies. Results are reported in Table 19. 1997. while the Book-to-Market and Lehman Mortgage factors obtained significant positive abnormal returns during the second one. The correlations have been computed for the whole 1994-2002 period and for two-sub-periods.and post-March 2000 sub-periods. Schneeweis and Spurgin.38 3. In each cell. Liang. In order to obtain periods with comparable lengths. 142 . while the lower part is split between the pre. that started in September 1998 (19 months) rather than the whole 1994-03:2000 period. 1999. 38 Detailed data are available upon request. hedge funds are likely to improve to the riskreturn trade-off when added to a traditional portfolio (see Fung and Hsieh. Because of the extremely large number of results to be reported. correlations increase as the colour is darker.For our factors. Poorest performers were the Lehman Aggregate US Bond Index during the up market trend period and the Market and World Excluding US factors during the down period. Amin and Kat. we took the bearish sub-period starting in April 2000 (33 months) and matched it against the most bullish time window. thanks to their weak correlation between hedge funds and other securities. 2003b). 1998. The upper part of the cell accounts for the correlation during the whole period. the same analysis shows that the Market and Momentum factors gave the highest excess returns during the first sub-period.

As typically reported (see Liang. However. These results confirm that hedge funds strategies are weakly correlated with most traditional investment tools The first line indicates that almost all hedge funds strategies tend to follow the market (US and international) more closely in the bearish sub-period. Capocci and Hübner. SMB and Lehman High Yield) have a high correlation with most hedge funds strategies. while increasing their exposure to the Lehman High Yield and the SMB factors. for its part. noticeably switches from a "Glamour" strategy (low loading with the HML factor) to a "Value" one in the last sub-period. to a lesser extent. the Global International strategies tend to decrease their correlation with other funds in bearish times. This strong tendency is not invalidated for the supposedly investor-protecting strategies. a closer look at their evolution over time indicates that the Global and. on the other hand.Panel A reports correlations among hedge funds strategies. 2003. the Global Emerging strategy follows the other strategies more closely during periods of down markets. The Short Sellers strategy. 2004) in the literature. 143 . This is the only strategy that consistently invests in Value stocks in bearish markets. with the exception of the Short Sellers strategy that systematically goes conversely – as expected. as they become momentum-contrarian during bad times. They also reduce their sensitivity towards the Emerging Market Bond factor. Only four indices (Market. hedge funds strategies sharply decrease and even reverse their loading with the momentum factor. In Panel B. In general. the behaviour of our explanatory variables is also of considerable interest. Panel C indicates that the correlation coefficients of our regressors are low enough to raise serious multi-colinearity concerns. The no Category and Niche strategies seem to be more correlated with the rest of hedge funds in the sub-periods than in the full period. World Excluding US. these strategies are in general highly correlated when indices are considered.

between hedge funds and passive investment strategies. GIN.Table 19: Correlation among hedge funds. GEM MAC MKN LOL SEC SHS NOC NIC DIV 144 . and among passive investment strategies Panel A: Correlation among Hedge Funds strategies ERA EDS GLO GES GIN GEM MAC MKN LOL SEC SHS NOC NIC EDS GLO GES.

Panel B: Correlation between Hedge Funds strategies and Passive Investments strategies ERA EDS GLO GES GIN GEM MAC MKN LOL SEC SHS NOC NIC DIV MKT WXU SMB HML MOM SWG EMB MOR HIY GSC .

GIN = Global International. between 50 and 75% in dark grey. the bottom left square represents the range of correlation coefficient for the most bullish sub-period (09:1998-03:2000). SHS = Short Selling. between 0 and 25% in medium grey. SEC = Sectors. WXU = World excluding US. NIC = Niche. DIV = Diversified. and the bottom right square represents the range of correlation coefficient for the most bearish sub-period (04:200012:2002). LOL = Long only Leveraged. GEM = Global Emerging. ERA = Event Driven – Risk Arbitrage. MKT = Market Proxy. MOM = Momentum.Panel C: Correlation among Passive Investment strategies MKT MKT WXU SMB HML MOM SWG EMB MOR HIY GSC WXU SMB HML MOM SWG EMB MOR HIY This Table reports the ranges of correlation coefficient among hedge funds strategies (Panel A). NOC = no Category. For each pair of strategies (in line and in column). between -25 and 0% in light grey and <-25% in white. Colour codes for correlations are: >75% in black. EDS = Event Driven – Distressed Securities. GLB = Global. S . the upper rectangle of the cell represents the range of correlation coefficient for the whole period (01:1994-12:2002). MAC = Macro. OPP. GES = Global Established. between hedge funds strategies and passive investment strategies (Panel B) and among passive investment strategies (Panel C). between 25 and 50% in medium-dark grey. MKN = Market Neutral.

o. we report the yearly returns of all funds..g. 2004).. Brown et al.. 2001 and Fung and Hsieh. 1997). before this starting date. surviving funds and dissolved funds. However. Carhart. Hedge funds experience extremely high returns in 1999. data vendors backfill each fund’s performance history prior to their addition to the database. Thus.o.1 Survivorship bias In order to reduce the severity of survivorship bias. We report the bias using both definitions for the whole period and for 2 sub-periods 1994-03:2000 and 04:2000-2002. Liang. when the stock market experienced a sharp positive return. leading to an increase in differences in hedge funds returns between the best and the worst performing managers. an important concern for mutual funds (see a.IV Analysis of biases 4.g. Data starting in 1994 appears to be more reliable according to this criterion (Capocci and Hübner. 2000). (2001). Ackermann et al. they provide data that go back before the starting date of the database itself. 1997) as well as hedge funds studies (see a. one is left with only surviving funds data. 147 . 1999) and the performance difference between living and all funds (e. and Capocci and Hübner (2004) for the combined MAR and HFR databases have shown that data for the pre-1994 period is indeed subject to non-negligible survivorship bias that is very likely to hinder statistical inference (see Hendricks et al. Two definitions of this bias are commonly used in mutual and hedge fund studies: the performance difference between surviving and dissolved funds (e. returns gradually reduced but mostly due to the negative returns of dissolved funds. for the TASS database. Ackermann et al. without great difference between surviving and dissolved funds. 2000). usually 1993. This is a clear effect of the bearish turn of the market after March 2000. In Panel A of Table 20. In the subsequent years.

8-1. It is however lower than the 3% bias found by Liang (2001). which is also the industry consensus as stressed by Amin and Kat (2003b). 148 . lower than the 0.In Panel B.1% in 2002). It is similar to the percentage of 1. This latter value is much higher than the very low value obtained by Ackermann et al. but this effect is somehow mitigated by the decrease in the proportion of returns from dissolved funds in the database (this proportion steadily decreases from 35. Thanks to this bias-reduction effect of recent data.51% per annum) with the second formula.41% (or 4. our results yield a monthly survivorship bias of 0. 39 This consensus value quite high when compared to the 0. the global behaviour of the database in relationship to survivorship bias is kept within reasonable bounds.39 A look at sub-period biases indicates that the level of this bias is mostly due to the bearish period.2% found by Capocci and Hübner (2004) for the 1994-2000 period.5 bias reported by Malkiel (1995) and Brown and Goetzmann (1995) for US mutual funds.92% per annum) for the whole period using the first formula while in Panel C the bias of 0. where its level sets at 2.30% monthly bias found by Fung and Hsieh (2000) and slightly higher than the percentage of 1.9% in 1999 to 9. for the period 1988-1995.5% from Fung and Hsieh (1998).13% per month (1.61%. The bias drifts up through the very high level of returns differential between surviving and dissolved funds.

1.5% 9. 8601 10641 13049 15860 17872 18798 20221 20591 20771 14137 20528 16267 Return 3.2% 1.3% -3.6% 21.2% 18.4% 1.4% Obs.9% 3.6% 33.0% 3. 1.2% 3.6% 1. 3419 4630 6200 8136 9954 12052 14395 16706 18899 7399 16667 10488 Return 1.3% -2.0% 1.6% 2.0% 9.4% 8.0% 2.7% 0.8% 14.4% 3.9% 4.9% 1. 1. 5182 6011 6849 7724 7918 6746 5826 3885 1872 6738 3861 5779 .9% 15.3% 2.0% 2.2% -4.3% 2.0% 19.0% -3.3% 2.0% 1.0% 13.4% 33.6% 22.9% 23.0% 17.5% 2.0% 5.6% 33.0% 1.0% Surviving Funds S.8% 18.1% 3.6% 1.3% 20.8% 1.3% 18.9% 1.D.4% 17.3% 1.0% Dissolved Funds S.3% 1.0% 15.9% Obs.4% 1.2% 2. D.8% 1.4% 2.7% 1.6% 5.0% 8.8% Obs.6% 2.2% 2.0% 2.4% 2.0% 1. surviving funds and dissolved funds) All Funds Year 1994 1995 1996 1997 1998 1999 2000 2001 2002 Mean 01:94-03:00 Mean 04:00-12:02 Mean 94-02 Return 1.0% 1.Table 20: Survivorship bias in hedge funds Panel A: Annual performance (all funds.0% S. D.7% 19.4% 4.

01 0.13 1. In Panel C survivorship bias is calculated as the performance difference between surviving funds and all funds.92 Return 0.02 0.02 0.01 0.16 1.Panel B: Living . Our MAR database contains 2894 hedge funds.98 11.41 4.04 0.51 per Month per Year per Month per Year per Month per Year Bias 04:00-12:02 0.61 0.75 Bias 94-02 0. Numbers in the table are yearly percentage unless otherwise indicated.04 0.02 0.01 0.02 0.All Funds Year 1994 1995 1996 1997 1998 1999 2000 2001 2002 Bias 01:94-03:00 Return 0.Dead Funds Panel C: Living .11 0.09 1. All returns are net of fees.01 0. In Panel B survivorship bias is calculated as the performance difference between surviving funds and dissolved funds. including 1622 survived funds and 1272 dissolved funds as of December 2002.03 0.00 0.02 0.22 2. 150 .05 0.17 0.01 0.06 0.93 This Table reports the survivorship bias of calculated from our database.02 0.

99%. 2000) that occurs because a fund’s performance history is backfilled after inclusion. the observable monthly return averaged 0. their managers consider inclusion in a database primarily as a marketing tool. This creates a positive instant history bias or backfilled bias (Fung and Hsieh. On the other hand. 0. On the one hand. Brown et al. (1997). and 0. The bias is estimated for the whole period and for the bullish and bearish sub-periods in order to compare our results with those obtained by Fung and Hsieh (2000) and Capocci and Hübner (2004). 151 .2% found by Capocci and Hübner (2004). 36.81% (36 and 48 months).80% (60 months). we estimate the average monthly return of the “observable portfolio” which invests in all funds from our database each month. Fung and Hsieh (2000) to estimate this bias for our hedge fund database. This gives an estimate of 1.88% (when deleting the 12 first months). We use the same two-step methodology as Park (1995). if possible.2 Instant Return History Bias As hedge funds are not allowed to advertise. we estimate the average monthly return of the “adjusted observable portfolio” obtained from investing in all these funds after deleting the first 12. The upward bias results from the likelihood that funds with a poor track record are less likely to apply for inclusion than funds with good performance history. 48 and 60 months of returns. Results are reported in Table 21. very much in line with the values of 1. For the whole period. while the adjusted observable one was 0.84% (24 months).32% per year. 24 and.4% found by Fung and Hsieh (2000) and 1. 0.4.

Our MAR database contains 2894 hedge funds.18% 0.99% 0. 152 .19% 0.32% 1.04% 2.23% NA 1.21% 0. including 1622 survived funds and 1272 dissolved funds as of December 2002.81% 0.17% 0.18% 0.80% Monthly Difference Annual Difference Av.13% 0.11% NA 1.22% 1.14% 1. Nb of Fds NA 0.10% 0.09% 0.35% 1.18% 1.76% 1197 945 752 620 539 498 NA 0.15% 0. 24. Instant history bias is calculated as the performance difference between the average monthly return using the portfolio which invests in all funds each month (the observable portfolio) and the average monthly return from investing in these funds after deleting the first 12.80% 2.56% 2.32% 1715 1489 1256 This Table reports the instant history bias calculated from our database.08% 1.52% 2.84% 0.14% 1.13% 0. All returns are net of fees and on a monthly basis unless otherwise indicated.88% 0.52% 2.16% 2.11% 0.16% 2.81% 0.21% 0.Table 21: Estimation of instant return history bias Mean Monthly Return Period 94-02 All Without 12M Without 24M Without 36M Without 48M Without 60M Sub-period 01:94-03:00 All Without 12M Without 24M Without 36M Without 48M Without 60M Sub-period 04:00-12:02 All Without 12M Without 24M 0. 36 and 60 months of returns (the adjusted observable portfolio).19% NA 1.08% 1.28% 1356 1174 999 837 694 570 NA 0.

the first sub-period is mostly responsible for the level of the bias. One may relate the level of this bias to the one of survivorship bias in this context: during unfavourable market conditions. with the highest returns being obtained around the end of the period. As only the most successful funds tend to remain during this time period. For the “bearish” sub-period. but the bias is kept at very reasonable levels. Because the period was increasingly bullish.Contrarily to the analysis of the survivorship bias. the bias starts at a fairly high level and increases as more returns are removed from the estimation. consistently with the phenomenon found by Capocci and Hübner (2004). there is a sharp difference in the returns between surviving and dissolved funds that can be explained by attrition of the latter funds because of their bad performance. while the observable portfolio returns include returns from subsequently dissolved funds. 153 . the corresponding instant history bias is likely to be mitigated by this self-selection of the most successful managers towards the end of the period. only partial results are available due to the small length of this period.

and especially thanks to the increase in the number of funds that has been observed over the same time window. the phenomenon of elimination of the poorest performers under unfavourable market conditions is also responsible for the remarkably low level of the instant history bias. Interestingly. while there is evidence of a more important instant history bias during the upward trending period. this examination of biases indicates that both survivorship and instant history biases are kept to very reasonable levels for the whole period as well as for the bullish and bearish sub-periods.4. Yet. but for very different reasons. survivorship bias is higher for the period of down market. but this is avoided by the particular behaviour of the database. 154 .3 Conclusion Overall. One could have suspected that the high failure rate of hedge funds after March 2000 would have lead survivorship bias to suspiciously high levels.

Overall. so that relatively accurate risk measures can be estimated. suggesting the need to use a more detailed model. except for the Long Only Leveraged. Table 22 reports the results for All Funds. have significantly out-performed the market. The estimated betas are rather low. and all R-squared are below 60%. The model is also estimated for each fund individually. We compute all estimations by using Newey-West (1987) standard errors to adjust for any autocorrelation in the returns. with the percentage of significantly positive. 32% of the alphas are significantly positive. we require all funds to have consecutive monthly return history for at least 24 months. Individual Funds and Funds of Funds. two thirds of the individual funds strategies produce significantly positive alphas. Overall. the last columns give the distribution of individually estimated alphas per strategy. hedge funds as a whole also significantly out-perform the market at the 1% level. and all funds strategies. 155 .V Hedge Funds Performance This Section aims at studying whether hedge funds.40 To analyse hedge funds performance in more details. insignificant and negative alphas at the 5% level..1 Performance Measurement using the CAPM Panel A of Table 22 reports performance estimates using the CAPM. 40 To make individual estimation. 5. as a whole or strategy by strategy. Taken individually. except for Long Only Leveraged and Global Est. while the two Funds of Funds strategies out-perform the market at the 10% level. with equally weighted portfolio excess returns for each investment style.

24 0.05 *** *** 0.73 0.12% 0.651 0.73 -0.43% 0.81% 0. Global Emerging US Opp.FoF Other .41 0. Global Intern.99% 0.Table 22: Performance measurement using the CAPM.26 NA * *** 0.FoF Diversified .Risk Arb Event driven .15 0.31% 0.39 0.48% Niche .614 Alpha Distrib.25% 0.174 0.31 *** *** * 0. Macro Market Neutral Long Only Lev.47 *** ** 0.37% 0.607 0.3 0.58 0.378 0. Sector Short Sales No category Individual Funds Total Mkt *** *** 0.24% NA 0.12 1.619 0.449 0. Sec Global Global Est.50% -0.419 0.719 0.427 NA 0.337 NA 0.3 0.331 0.62 NA ** *** 0.55% NA 0.FoF Funds of Funds Total Hedge funds Total .Dist.35 *** *** *** *** *** *** *** *** *** *** *** *** *** *** *** *** *** R² adj 0.69% 0.74 0. / 0 / 48% 34% 9% 30% 22% 20% NA 18% 44% 9% 31% 11% 21% 31% 43% 33% NA 34% 32% 51% 64% 78% 67% 74% 75% NA 72% 52% 80% 67% 88% 78% 65% 53% 62% NA 60% 64% 0% 1% 11% 2% 2% 4% NA 8% 2% 9% 1% 0% 0% 3% 2% 3% NA 4% 3% + 0. Carhart’s 4-factor model and the combined model Panel A: Single index model Alpha Event driven .62% 0.46% 0.285 0.55% 0.11% 0.445 0.26 0.515 0.32% 0.34 0.

Global Emerging US Opp.01 0.2 0.07 NA 0.07 0.11 0.4 0.07 0.13% 0.37 *** *** *** *** *** *** *** *** *** *** *** *** *** *** *** *** *** SMB 0.18 0.63 0.66 0.10% NA 0.58 0. 0 / 37% 26% 11% 27% 21% 18% NA 17% 41% 12% 26% 20% 21% 28% 33% 25% NA 26% 28% 60% 71% 77% 71% 71% 78% NA 74% 54% 80% 72% 79% 73% 67% 62% 68% NA 66% 67% 1% 2% 11% 1% 7% 3% NA 8% 3% 6% 1% 0% 4% 3% 3% 5% NA 6% 4% + / .41% 0.01 0.01 0.64 NA 0.31 NA 0.45 0.Dist.82 0.81% 0.26 0. Global Intern.87 0.19% 0.83% 0.02 0.83 0.39 0.07 0. Macro Market Neutral Long Only Lev.12 NA 0.41% -0.28 0. Sector Short Sales No category Individual Funds Total Niche .71 -0.87 0.01 0.26 0.29 NA 0. Sec Global Global Est.19 0.48 0.08 0.Panel B: Carhart's 4-factor model .09 0.individual funds Alpha Event driven .26% 0.33 0.01 0.33 0.06 ** ** ** ** ** *** ** *** *** PR1YR -0.02 0.05 0.02 0.80 Alpha Distrib.23 0.15 0.58% 0.37% *** *** ** *** *** *** *** *** *** Mkt 0.05 0.19 *** *** *** *** *** *** *** *** *** *** *** *** *** *** *** *** *** *** *** HML 0.11 0.FoF Funds of Funds Total Hedge funds Total 0.04 0.13 0.07 0.43% NA 0.31 0.67 0.64 NA 0.61 0.38 0.16% 0.54 0.08 0.15 0.16 NA 0.27 0.46% 0.09 0.24% 0.08 0.04 0.FoF Other .36 0.21% 0.16 0.02 NA 0.63 0.55% 0.FoF Diversified .36 0.55 0.07 -0.11 0.40 NA 0.15 -0.04 0.78 0.06 0.07 NA 0.15 0.09 0.Risk Arb Event driven .14 0.36 0.98 0.43 -0.05 *** *** *** ** *** *** ** *** *** *** * R² adj 0.11% 0.

-0. + / 0 / 37% 59% 2% 27% 71% 1% 9% 81% 9% 26% 71% 2% 28% 67% 4% 15% 84% 0% NA NA NA 16% 78% 4% 39% 57% 3% 6% 83% 9% 19% 78% 1% 23% 76% 0% 21% 78% 0% 26% 69% 3% -0.06 -0.03 PR1YR 0. 0.73% *** 0.07 0.61% NA 0.25 * -0.12 ** 0.11 0.84 *** SMB 0.01 -0.30 ** 0.41% *** 0.42 0.01 0.11 -0.Risk 0.66 0.02 0.33% 0.74 *** -0.17 *** 0.05 0.01 0.10 * Em.01 0. Sector Short Sales No category Individual Funds Total -0.31 ** NA 0.69 0.11 0.05 W Gv Bd 0.71 0.44 0.15 0.18 -0.64 0.25 *** 0.20 0.78 0.05 0.66 *** NA -0.01 -0.46% *** Mkt 0.56% *** 0.Dist.06 ** 0.08 0.15 *** -0.06 0.10 *** 0.08 -0.03 0.03 -0.05 *** W x US -0.43 *** HML 0.19 *** 0.34 *** 0.07 NA 0.10 *** 0.04 -0.04 -0.12 ** 0. Global Emerging US Opp.49 *** 0.18 0.22 *** NA 0.86 Alpha Distrib.14 * -0.54% *** -0.03 -0. Macro Market Neutral Long Only Lev.39% *** Arb Event driven .05 0.58 *** 0.07 *** 0.39 *** 0.88 0.21 *** 0.07 *** High Y.02 0.03 0.00 0.12% 0.02 0.16 0.12 *** 1.17 *** 0.18 *** 0.11 0.12 0.10% 0.06 ** 0.66 0.11% 0.05 * 0.24 *** 0.92% *** 0.17 *** .00 0.52 NA 0.01 0.34 ** NA 0.05 0.06 ** 0. Bd 0.11 NA 0.30 *** 0. 0.06 0.08 0.07 * 0.18 -0.25 *** 0.71 ** -0.85 0.25 *** 0.11 *** 0.04 ** R² adj 0.03 0.07 0.01 0.08 *** 0.14 ** 0.04 0.13 ** 0.05 Mortg.03 0.01 0.04 *** 0.05 Comm.07 * 0.16 0.07 *** 0.26 ** 0.individual funds Alpha Event driven .12 * NA -0.Panel C: The combined model .13 0.12 * 0.11 -0.05 0.02 0.04 0.88 0.05 0.03 0.13 *** 0.13 *** 0.09 -0.04 * 0.01 0.18 NA 0.02 0.05 *** -0.04 0. Global Intern.21 *** 0.38 *** 0. 0.01 0.26 *** -0.02 0.40% *** Sec Global Global Est.01 *** 0.20 ** NA 0.00 0.01 NA 0.06 0.

84 26% 70% 3% This Table presents the results of the estimation of the single index model (Panel A).08 *** 0. significantly negative alphas (-) and alphas insignificantly different from zero (0) at the 5% level. *** Significant at the 1% level.70 25% 70% 3% Hedge funds Total 0. of Carhart’s (1997) model (Panel B) and of our combined model (Panel C) for the 01:1994-12:2002 period. Mortg. Bd High Y. t-stats are heteroskedasticity consistent.2 *** 0.06 0.06 -0.24% ** 0.07 0. per type of funds and for all funds.12 *** 0. We report the OLS estimators for equally weighted portfolios per investment strategy.13 *** 0.51 32% 64% 2% Diversified .03 0. R² adj Alpha Distrib.05 *** 0. .31 *** 0.08 *** 0.04 -0.03 0.07 *** 0.05 * 0.16 0.03 0.1 0. ** Significant at the 5% level and * Significant at the 10% level.71 24% 71% 3% Other .21 *** 0.17 *** 0.FoF Alpha Mkt SMB HML PR1YR W x US W Gv Bd Em.FoF 0.06 *** 0.05 *** 0.16 ** 0.07 0.07 *** 0.08 *** 0.Panel C (continued): The combined model .07 0.05 -0. + / 0 / - Niche . respectively.FoF 0.12% 0. 0 and 1 living fund in the second sub-period.38% *** 0.06 -0. We report the percentage of significantly positive alphas (+).06 -0.08 *** 0. Results for the US Opportunistics and Other categories are not reported as they have.13 *** 0.04 ** 0. Comm.02 0.03 0. The last column gives the distribution of individually estimated monthly alphas for all funds with 24 monthly data or more in a specific investment style.10% 0.16 *** 0.04 0.07 0.FoF NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA Funds of Funds Total 0.11 ** 0.14 0.

respectively. The individual alpha distribution shows that taking more factors into account drives down that the proportion of individual funds and funds of funds that significantly out-performed the market. In Panel B of Table 22.1% per months. including in the Short Sellers strategy where it is negative.5. accounting for more risk premia reduces the average reported out-performance by 0. and to all funds of funds strategies. with only the Short Sales strategy being momentum contrarian over the whole period. while Global International and Emerging strategies share similar risk exposure characteristics except that the former is more momentum-driven and the latter is naturally heavily exposed to the Emerging Market Bond factor. In particular.2 Performance Measurement using Multi-Factor Models It is presumably better to use a multi-factor model to account for all possible investment strategies. significant. we report the results for Carhart’s 4-factor model and in Panel C the results for our combined model applied to hedge funds. Overall. Evidence on alphas obtained in Panel C is not favourable to Funds of Funds and to the Macro strategy. the Emerging Bond factor adds explanatory power in more than 50% of the strategies with high significance levels. The coefficients of the HML and Momentum factors are significant for four and six individual funds strategies. The Momentum factor remains a stronger indicator of hedge funds behaviour. Panels B reveals that the premium on the SMB factor is. Panel C shows that the explanatory power of the HML factor seems marginal as only three betas are significantly positive at the 5% level. in almost all cases. 160 . Event Driven strategies are more prone to bear a high exposure to high yield bond factors. as already outlined by Capocci and Hübner (2004). and the distribution of performance among strategies is also more concentrated. The results with our combined model also indicate that all additional factors add explanatory power to the regression.

Overall it seems that the combined model does a very good job in describing hedge funds behaviour.3 Performance over bullish and bearish sub-periods In order to analyse the performance components in the bullish and bearish market configurations.g. The average R²adj increases from 0. . 161 . 41 See e.84 for our combined model. we only report results for our combined model.61 for the single factor model. This coefficient is the best one reported in the literature so far.80 for the 4-factor model and to 0. Table 23 shows the value of the coefficients for the sub-periods 01:1994-03:2000 (Panel A) and 04:2000-12:2002 (Panel B).41 5. to 0.Liang (1999) and Amin and Kat (2003a).

07 0.03 0.38 0.08 0.08 -0.06 0.01 -0.07 -0.04 0.06 -0.17 0.00 -0.01 -0.31 0.25 0.05 -0.66 0.09 -0.07 0. Global Emerging US Opp.09% 0.03 0.19 -0.52 -0.26 0.06 0.53% *** 0.25% 0.04 * ** *** ** ** * W x US -0.51 -0.19 0.01 0.08 0.02 -0.01 0.10 0.44% *** 0.13 0.28 -0.07 0.07 -0.44 -0.05 0.00 0. Macro Market Neutral Long Only Lev.98 -0.07 0.11 0.20 0.30 0.50% *** 0.27% 0.26 -0.12 0.65 0.03 -0.07 0.17 -0.00 0.01 0.11 0.50 0. Sec Global Global Est.35% *** Mkt 0.09 0.03 0.22% 0.01 0. Global Intern.14 1.13 0.12 -0.18 -0.06 0.95% *** 0.02 -0.07% 0.13 0.13 0.03 0.02% 0.34 0.14 0.14 0.09 -0.FoF Other .50% *** -0.06 0. Sector Short Sales No category Ind.05% 0.Table 23: Performance of hedge funds during the bullish and bearish sub-periods Panel A: 01:1994-03:2000 Alpha Risk Arb Dist.27 0. of Funds Total Total 0.04 ** *** ** PR1YR -0.62% *** 0.19 *** *** *** *** *** *** *** ** ** 162 .36 0.25 -0.03% 0.06 0.05 -0.24 0.32 0.24% 0.03 0.92% *** 0.08 -0.10 0.65% ** 0.04 0.04 0.06 *** * *** ** ** W Gv Bd -0.36 0.47 0.82 -0.07 0.00 0.22 0.00 0.20 0.00 0.06 0.00 -0.17 -0.02 -0.39 0.03 0.45 0.04 0.32 -0. Funds Total Niche .FoF F.41 0.47% ** 0.05 0.01 0.29 0.FoF Diversified .06 -0.31 0.38 *** *** *** *** *** * *** *** *** *** *** *** *** *** *** *** ** *** SMB 0.02 0.13 0.18 ** *** *** *** *** *** *** *** ** *** ** *** *** *** ** *** *** HML 0.

33 *** *** ** *** -0.03 0.66 0.FoF Diversified .05 -0. + / 0 / 46% 33% 9% 25% 27% 10% 6% 15% 39% 3% 25% 17% 31% 26% 35% 21% 18% 23% 46% 51% 66% 81% 72% 70% 87% 82% 80% 59% 88% 74% 82% 68% 70% 60% 73% 77% 71% 51% 2% 0% 9% 1% 1% 1% 10% 3% 1% 7% 0% 0% 0% 2% 3% 4% 4% 4% 2% Risk Arb Dist.06 0.04 0.05 *** *** 0. Macro Market Neutral Long Only Lev.86 *** *** -0.74 0.48 0.00 0.08 0.11 -1. Funds Total Niche .07 -0.51 0.01 0.67 0. of Funds Total Total 0. Comm.08 0.07 ** *** 0.05 0.06 0.04 0. Bd High Y.10 -0.03 0.17 -0. Sector Short Sales No category Ind.34 -0.17 0.06 0. Global Intern.16 0.71 0.FoF F.15 -0.Panel A (continued): 01:1994-03:2000 Em.00 ** 0.53 * -0.23 -0.17 0.09 -0.06 163 .50 0.01 0.22 0.FoF Other .19 0.10 ** -0.51 0.03 0.04 ** ** ** * * *** 0.12 0. Global Emerging US Opp.75 0.04 0. R² adj Alpha Distrib.76 0.04 0.09 0.01 0.49 0.02 0.06 -0.01 0.26 0.30 -0.43 0.70 0.11 0.07 0.05 0.03 0.04 0.71 0.48 0.15 0.51 0.00 0.06 0.05 0.04 0.80 0.18 0.56 0.06 -0.91 0.87 *** -0.89 0.03 -0.10 -0. Sec Global Global Est.90 0.66 0.81 0.85 0.06 * -0. Mortg.03 0.03 0.07 0.04 0.06 * ** 0.02 -0.22 -0.32 *** *** 0.

19 0.08 0.07 -0. of Funds Total Total -0.19 0.29 *** *** * *** *** *** 0.02% 0.00 0.10 0.50% -0. Sector Short Sales No category Ind.12 ** 0.14 0.00 0. Macro Market Neutral Long Only Lev.18 0.03 0.10 NA 0.02 *** *** *** 0.71 0.29 ** * 0.27 0.05% -0.14 0.22 0.12% -0.02 -0.12 0.26 -0.00 -0. Global Intern.01 -0.16 -0.08 0.08 0.11 0.Panel B: 04:2000-12:2002 Alpha Mkt SMB HML PR1YR W x US W Gv Bd Risk Arb Dist.07 0.00 * *** *** *** 0.17 0.07 * * ** * * *** 0.FoF Diversified .05 -0.31 NA ** 0.49 NA 0.04 NA 0.03 -0.14 NA -0.11 -0.07 0.16 0.05% 0.03% -0.08 0.44 -0.08 0.08 ** *** *** -0.02 NA * 0.02 0.13 0.13 *** *** 0.40% -0. Global Emerging US Opp.18 0. Funds Total Niche .03 0.00 NA 0.12 0.01 0.02 0.38 -0.04 NA 0.01% 0.45 0.00 0.23 -0.19 -0.19 NA 0.12% -0.34 -0.25% 0.06 NA ** -0.09 * *** -0.19% 0.12 NA *** *** *** 0.37% *** -0.12 -0.27 0.02 0.05 -0.08 0.06 0.19 0.25% NA -0.97 -1.02 164 .01% -0.05 ** ** * *** ** 0.21 0.08 -0.FoF Other .06 NA *** ** -0.02 0.19% -0.39 0.36 0.14 0.03 0. Sec Global Global Est.13% NA -0.FoF F.15 0.03 0.25 -0.44 0.06 0.06 1.13 0.04 0.02 0.05 0.05 0.18 0.03 *** *** 0.12% -0.09 * * ** ** *** ** ** 0.32 0.10 0.08 0.04 0.

01 0.12 *** *** *** ** *** *** ** *** *** Mortg.92 0.05 0.01 0. *** Significant at the 1% level.04 NA 0.22 -0.69 0. Sec Global Global Est.01 0.90 0.05 -0.2 * * Comm.01 0. significantly negative alphas (-) and alphas insignificantly different from zero (0) at the 5% level.03 -0.65 0.25 0.15 0.04 NA -0.04 0.05 0.FoF Diversified .09 0.01 -0.02 -0.08 0. of Funds Total Total -0.21 NA 0. 0.86 0. per type of funds and for all funds.01 NA 0 0.08 0.95 0. 0.1 NA 0.88 0.09 0. Sector Short Sales No category Ind.11 0. 165 .76 0.08 0.Panel B (continued): 04:2000-12:2002 Em.01 0.24 -0. In Panel B.24 0.02 0.71 NA 0.94 7% 13% 0% 5% 9% 15% NA 6% 27% 0% 6% 0% 0% 13% 14% 10% NA 11% 13% Alpha Distrib.66 0. Macro Market Neutral Long Only Lev.FoF Other . The last column gives the distribution of individually estimated monthly alphas for all funds with 24 monthly data or more in a specific investment style. 0.03 -0. t-stats are heteroskedasticity consistent. We report the percentage of significantly positive alphas (+).05 0.92 0.89 NA 0. Global Emerging US Opp.27 0.07 0 0.04 0.06 -0.06 -0.12 0.01 0.27 0.75 -0.56 0.06 0.01 0 NA 0 0 ** * R² adj 0.06 0. Bd Risk Arb Dist.19 0.13 0. ** Significant at the 5% level and * Significant at the 10% level.44 0.78 0.23 0. We report the OLS estimators for equally weighted portfolios per investment strategy.04 -0.12 0. Funds Total Niche .1 0. + / 0 / 91% 86% 100% 89% 87% 80% NA 85% 68% 95% 89% 90% 100% 82% 80% 81% NA 81% 82% 0% 0% 0% 5% 3% 4% NA 7% 3% 4% 3% 9% 0% 4% 4% 7% NA 7% 4% This Table presents the results of the estimation of our combined model for the 01:1994-03:2000 (Panel A) and the 04:2000-12:2002 (Panel B) sub-periods.14 0.74 0. respectively.02 -0.43 0.08 NA 0.11 -0. results for the US Opportunistics and Other categories are not reported as they have.03 0.FoF F.38 0. 0 and 1 living fund in the second sub-period.01 -0.05 -0.09 NA 0. Global Intern.95 0.02 ** High Y.

A quick look at the alphas for the considered sub-period clearly indicates that the major part of the performance over the total 1994-2002 period is recorded prior March 2000, with the noticeable exception of the Market Neutral strategy that sustains positive out-performance for both sub-periods. This finding is consistent with the result obtained by Liang (2003) who investigates the behaviour of hedge funds strategies using a piecewise linear regression setup: Market Neutral funds obtain by far the largest standardized value for alpha, with at the same time a very low explanatory power of the regression. In contrast, Panel A shows that the Global strategy achieves significant negative performance during the bullish period. It is worth reporting that, although nine individual strategies and both funds of funds strategies recorded a negative alpha in Panel B, none of these values are shown to be significant. A look at individual alphas reinforces this finding, as the proportion of significantly positive alphas does not significantly differ from the total period to the first sub-period, but Panel B shows that 27% of the Market Neutral funds sustained positive out-performance while on average more than 80% of individual funds managers were in line with the market. The strategies followed by funds managers sharply differed from one sub-period to another. All but the No-category strategies individual funds significantly followed the market until March 2000; only the Long Only Leveraged and the Sectors funds increased their exposure thereafter. The Global, Global International, Market Neutral and No-category strategies were not even significantly loaded to the market risk premium. In contrast, the Funds of Funds strategies all increased their US stock market exposure after March 2000. Some general swings of exposures to several risk factors are observed from one period to another Exposure to the World excluding US usually becomes negative, although with low significance levels, in the bearish sub-period, except for Global International and Short Sales. On the other, the broadly negative exposure to the World Government Bond

166

Index in the first sub-period fades away after March 2000 except for Global and Sectors, although the Short Sales strategy follows a converse tendency. The loadings for the Emerging Market Bond Index and High Yield Bond Index are generally positive in the first and second sub-period, respectively, which possibly indicates a broad sliding of bond strategies of hedge funds managers. At the individual strategy level, some changes are also of particular interest after March 2000. Event driven strategies cease to be momentum-contrarian. All Global strategies increase their investments in small firms and reduce their exposures to bond factors; this latter statement also holds for Macro funds. Market Neutral funds managers relied more extensively on domestic bond indices. Strikingly, the Short Sales strategy left a pure market-contrarian profile for a much broader mix of exposures (positive for HML, World excluding US and Commodity Index; negative for SMB, Momentum and World Government Bond Index). For funds of funds strategies, the noticeable difference is the noticeable

reinforcement of the loadings to the Momentum factor and the High Yield Bond Index after March 2000, while the exposure to the World Government Bond Index goes from very negative to slightly positive.

167

VI Persistence in Performance
Our results show significant evidence of superior performance over the total period of time for most individual strategies. Nevertheless, results are mostly due to the first, bullish sub-period and the positive out-performance tends to fade away after March 2000. Nevertheless, active hedge funds selection strategies are likely to increase the expected return if performance is persistent, i.e. if a superior average return in a period is likely to be followed by a superior average return in the next period for a given fund. Sirri and Tufano (1998) document large inflows of money into last years best performers, and withdrawals from last years' losers. Zheng (1999) finds that newly invested money in these best performing mutual funds is a predictor of future performance. This indicates that persistence in performance is critical for mutual funds. This is all the more important given that we have found in the previous subsection a substantial break in performance at the peak of the stock markets. Is persistence sustainable over the total period, or is it likely to be observed only in a particular subperiod?

6.1 Persistence over the total period We follow the methodology of Carhart (1997) using our combined model. All funds are ranked based on their previous year total return. Every January, we put all funds into 10 equally weighted portfolios, ordered from highest to lowest past returns. Portfolios 1 (High) and 10 (Low) are then further subdivided on the same measure. The portfolios are held till the following January and then rebalanced again. Funds that disappear during the course of the year are included in the equally-weighted average until their death, then portfolio

168

weights are readjusted appropriately. This yields a time series of monthly returns on each decile portfolio from 01:1995 to 12:2002. The monthly average return to the strategy of investing in portfolios 1 and 10 would have been, respectively, 1.07% and 0.44% for the total period. The monthly excess returns on the decile portfolios decrease monotonically between portfolio D1 and D8, but the subdecile D10c obtains an excess return higher than 1%, slightly significant. The spreads between decile excess returns are not significant. Cross-sectional variation in returns is considerably larger for the extreme deciles than for the middle deciles, in line with the results of Brown et al. (2001) and Capocci and Hübner (2004). After controlling for the risk factors, the picture is dramatically altered. The D10c portfolio, i.e. the extreme losers, enjoy a remarkable monthly out-performance of 1.77%.42 The 1a-10c spread goes from an insignificant 0.05% to a significant – 1.6%. Aside from this extreme value, significant alphas are mostly to be found in the middle deciles, with the most significant values (at 1% level) being observed in portfolios D4 to D8. The pattern of loading to risk premia suggests that past winners more closely follow the market, invest more in small firms and in emerging bond markets but less in the world stock index than past losers. Quite naturally given the definition of the portfolios, past winners follow momentum strategies while past losers are momentum-contrarian.

42

The poor value of the adjusted R² for this decile portfolio suggests however a very unstable behaviour

of individual funds returns inside this decile portfolio.

169

In the middle deciles, where performance seems to be persistent when accounting for risk, we notice that these strategies are usually characterized by positive exposure to the HML factor (value strategy), negative exposure to the world bond index but positive exposure to the emerging bond markets, indicating arbitraging strategies on geographical bond markets. This is a possible source for their sustained performance.

6.2 Persistence over the sub-periods The same analysis as before is performed in Table 24 for the bullish (Panel A) and for the bearish (Table B) sub-periods.43 Panel A of Table 24 displays, not surprisingly, very comparable results with the ones of Table 23, but there are some important differences. Firstly, the alpha for portfolio D10c is not significant anymore; only middle-decile portfolios have a significant alpha. This result is consistent with Capocci and Hübner (2004) that analyse the 01:1994- 06:2000 period. There is no significant spread between decile portfolio returns. Loading to individual factors are to a large extent similar for stock indices, but not at all for bond indices. During the bullish period, past winners have not invested in any bond index, and even heavily divested from the world bond index. In contrast, they have also significant loadings with respect to the commodity index. At the same time, past average performers were mostly invested in the high yield bond market.

43

For the second sub-period, we used returns of the 04:1999-03:2000 period to form the first decile

portfolios.

170

Unfortunately, Panel B indicates that there is no evidence of persistence in good performance during the bearish period. The only sustained performance is the negative one, as past losers are found to persistently aggravate their losses in portfolios D10, D10a and D10b. This finding is in line with our analysis of survivorship biases, reinforcing the conjecture that there was a particularly high mortality rate after March 2000 due to poor performance of the disappeared funds. Top decile portfolios during that period had positive loadings in high yield bonds and in the Momentum factor, but negative loadings in the commodity market and the HML factor. The losing strategies, i.e. the ones persistently followed by bottom decile portfolios, had loading of opposite signs on the same factors and, additionally, very high loadings on the mortgage market. We also notice that, contrarily to the "conventional wisdom" concerning the correlation between the momentum factor and hedge funds performance during bearish periods, past winners consistently invested in momentum strategies and past winners consistently followed contrarian strategies after March 2000. The paradox is only illusory, as past winners typically hold winner stocks in their portfolio and are thus naturally positively exposed to the momentum factor with these securities. Our results simply suggest that these funds managers did not actively manage this particular component of their portfolio.

171

Table 24: Hedge funds persistence based on 12 month lagged returns
Panel A: Decile estimation Portfolio D1a D1b D1c D1 D2 D3 D4 D5 D6 D7 D8 D9 D10 D10a D10b D10C 1-10 spread 1a-10c spread 1-2 spread 9-10 spread Exc. return 1,12% * 1,10% ** 0,97% ** 1,07% ** 0,86% *** 0,78% *** 0,71% *** 0,61% *** 0,50% *** 0,45% *** 0,46% *** 0,31% 0,44% 0,54% 0,03% 1,07% * 0,63% 0,05% 0,21% -0,13% St. dev 6,63% 4,75% 3,97% 5,00% 3,27% 2,42% 2,10% 1,69% 1,33% 1,37% 1,67% 2,34% 3,95% 3,36% 4,03% 5,65% 5,03% 7,71% 2,12% 2,17% Alpha 0,17% 0,38% 0,47% ** 0,34% 0,32% * 0,31% ** 0,36% *** 0,35% *** 0,28% *** 0,30% *** 0,40% *** 0,34% ** 0,72% ** 0,53% ** 0,29% 1,77% *** -0,38% -1,60% ** 0,02% -0,38% Mkt 0,64 0,57 0,43 0,55 0,48 0,37 0,31 0,26 0,2 0,17 0,11 0,18 0,22 0,28 0,27 0,07 0,33 0,58 0,07 -0,04 ** ** *** *** *** *** *** *** *** *** *** *** *** *** * *** ** SMB 0,63 0,4 0,35 0,46 0,28 0,19 0,13 0,09 0,09 0,08 0,09 0,08 0,05 0,09 0,04 0,03 0,41 0,6 0,18 0,03 *** *** *** * *** *** *** *** *** *** *** *** *** *** *** *** HML -0,04 0,03 -0,02 -0,01 0,04 0,07 0,07 0,05 0,07 0,03 -0,01 0,02 -0,04 0 -0,01 -0,13 0,03 0,09 -0,04 0,06 ** ** * ***

172

Panel A (continued): Decile estimation

Portfolio

PR1YR

W x US

W Gv Bd

Em. Bd

High Y.

Mortg.

Comm.

R² adj

D1a D1b D1c D1 D2 D3 D4 D5 D6 D7 D8 D9 D10 D10a D10b D10C 1-10 spread 1a-10c spread 1-2 spread 9-10 spread

0,36 0,29 0,2 0,28 0,20 0,11 0,08 0,05 0,03 0,01 -0,02 -0,10 -0,19 -0,13 -0,14 -0,32 0,47 0,67 0,08 0,09

*** *** *** *** *** *** *** *** **

0,14 0,05 0,07 0,08 -0,04 0,01 0,04 0,01 0,02 0,04 0,11 *** *** ** ** ** *

-0,56 -0,19 -0,15 -0,29 -0,02 -0,07 -0,13 -0,05 -0,09 -0,10 -0,18 -0,23 -0,29 -0,02 -0,32 -0,53 0 -0,03 * -0,27 0,06

**

0,19 0,11 0,08

** * * **

0,07 0,16 0,15 0,12 0,08

0,51 -0,06 0,03 0,15 -0,2 0,12 0,05 -0,05 * 0,07 0,05 0,15 0,12 -0,15 0,14 -0,41 -0,26 0,3 0,76 0,35 0,27

0,13 0,07 0,04 0,07 0,04 0,03 0,04 0,02 0,02 0,03 0,05 0,04 0,08 0,03 0,07 0,1 0 0,03 0,03 -0,04

**

0,77 0,77 0,78

*

0,12 0,06 0,07

*

0,80 0,81 0,79

** ** ** *** *** **

0,05 0,05 0,09 0,08 0,05 0,08 0,05 0,12

*

0,06 0,05

*

0,75 0,70 0,77

** ** *** ***

0,05 0,04 0,06 0,05 0,04 0,09

* **

0,74 0,70 0,73 0,56 0,60 0,56 0,36 0,52 0,48 0,54 0,18

*** *** *** *** *** *** *** *** **

0,19 0,28 0,22 0,29 0,35 -0,2 -0,21 0,12 -0,09

*

0,04 0,09 0,25 0 -0,19

* *

0,03 -0,02 0,08 0,21

***

0,07 0,01

*

0,04 -0,07

This Table reports the result of the estimation of our combined model for the 01:1994-12:2002 period. Each year, all funds are ranked based on their previous year's return. Portfolios are equally weighted and weights are readjusted whenever a fund disappears. Funds with the highest previous year's return go into portfolio D1 and funds with the lowest go into portfolio D10. Monthly Exc Return is the Monthly Excess Return of the portfolio, Std. Dev. is the Standard Deviation of the Monthly Excess Return.. All numbers in the Table are monthly percentage. *** Significant at the 1% level ** Significant at the 5% level * Significant at the 10% level.

173

6.3 Analysis of the Market Neutral strategy This sub-section focuses on the persistence in returns for the only hedge funds strategy that has been found to provide significant abnormal returns for both sub-periods in the previous section, namely the Market Neutral strategy. We determine whether persistence in returns exists for this strategy for the whole period as well as for the subperiods. As for the total sample, we classify funds in 10 decile portfolios, with the top and bottom decile divided in 3. Table 26 reports our results for these strategies. Panel A reports the results of the analysis for the global period. They show that there is no significant difference between good and bad performing funds. All alphas but the one of portfolio D10c are significantly positive for the whole period, although the significance level is lower for the extreme sub-deciles. Compared to our results for the whole database, excess returns are higher for the extreme deciles and smaller for middle deciles; however, returns of Market Neutral funds exhibit a much lower variance, and higher alphas for the top deciles (from D1 to D5). In a nutshell, all but the poorest past performers exhibit a significant persistence in performance. The review of risk coefficients shows that, aside from the fact that past best performers had a significant loading on the stock market index and on the SMB factor while past losers had a greater focus on world stock markets, no other clear pattern emerges.

174

28 0.39 0.45% -0.03 -0.71% 0.15 0.13 0.46% 3.13% ** ** ** ** *** *** *** ** Mkt 0.01% -0.00 0.30% St.59% -0.87% *** 1.75% *** 0.69 0.08 -0.31% 0.75% -0.30% 0.19 0.97% ** 1.10 0.15 -0.3 0.19% 2.11 -0.19 0.09 0.67% ** 0.42% 0.07 0.12 -0.41 0.62 0.24 0.09 0.44% *** 1.85% 4.08 *** *** *** *** *** *** *** *** *** *** *** *** *** *** * HML -0.46 0.01% -0.01% ** 0.32 0.09% ** 1.27% 0.68% *** 1.46% 0.03 -0.27% 2.89% *** 1.39 0.32 0.07 0.18% 3.19 -0.57 0.70 0.21 0.41 0.05 -0.35% 1.60% 3.35 0.56% 2.07 ** *** ** *** *** ** 175 .43% 0.17% 0.09 0.41% 0.89% 1.35 0.40% 0.08 0.77% * 1.45% 1.45% 1.87 0.11 0.Table 25: Hedge funds persistence during the bullish and bearish sub-periods Panel A: 01:1994-03:2000 Portfolio D1a D1b D1c D1 D2 D3 D4 D5 D6 D7 D8 D9 D10 D10a D10b D10C 1-10 spread 1a-10c spread 1-2 spread 9-10 spread Exc.54% 5.01 0.28% 0.97% 4.37% ** 0.36% 0.21 0.13 0.41% 0.97% 3.22% Alpha 0. return 2.11% 2.14 0.55% 0.48 0.25 0.51 0.04 0.24 0.11 0.21 0.04 -0.90% *** 0.27% 0.09 0.32% 7.51 0.35% 4.6 0.11 *** *** *** *** *** *** *** *** *** *** ** *** ** *** SMB 0.67% 2.28% 0.76% *** 0.08 0.08% 3.08% *** 0. dev 6.92% * 0.17 0.76% *** 0.17 -0.15% *** 1.

01 ** -0.3 -0.10 * -0.05 0.07 -0.06 0.36 0.23 0.41 *** 0.97 0.17 0.18 -0.94 *** -0.56 -0.28 -0.35 -0.26 0.10 ** 0.29 0.06 ** *** *** ** *** * R² adj 0.09 0.10 0.03 0.03 176 .51 -0.84 0.15 0.22 *** -0.40 0.Panel A: 01:1994-03:2000 Portfolio D1a D1b D1c D1 D2 D3 D4 D5 D6 D7 D8 D9 D10 D10a D10b D10C 1-10 spread 1a-10c spread 1-2 spread 9-10 spread PR1YR 0.11 * -0.02 -0.35 0.00 0.05 0.13 -0.33 0.03 ** *** *** *** ** *** *** * W x US 0.16 0.24 0.05 -0.18 0.84 ** 0.16 ** -0.30 -0.01 0.06 -0.01 0.07 ** -0.06 0.89 0.47 -0.18 -0.07 ** 0.19 0.27 0.78 -0.27 0.32 -0.01 0.84 0.07 -0.02 -0.55 0. 0.83 1.34 0.14 0.09 -0.14 ** -0.78 1.85 0.11 0.29 Comm.45 0.26 0.83 0.25 -0.54 0.20 0.09 ** -0.37 *** -0.32 0.89 0.31 *** -0.06 0.00 0.05 0.33 -0.06 0.03 0.17 0.10 ** 0.04 0.05 0.15 0. Bd 0.05 0.06 0.06 0.05 0.75 -0.06 0.30 * -0.06 -0.07 0.00 * ** *** * * * W Gv Bd -0. -0.07 0.00 * ** ** ** *** *** *** * High Y.02 0.46 Mortg.47 -0.25 0.21 ** -0.04 0.05 0.02 0.05 0.58 *** -0.05 0.88 0.08 0.05 0.02 0.12 0.05 0.14 * -0.41 0.57 0.21 *** -0.16 0. 1.05 0.23 -0.50 ** 0.63 -1.02 0.46 0.4 ** Em.02 0.12 0.03 0.04 -0.67 0.16 0.03 0.28 0.29 2.11 0.02 0.23 *** -0.06 0.70 0.79 ** -1.13 0.23 0.17 0.03 * 0.43 *** -0.78 0.17 0.17 0.81 0.17 0.

26% 0.04 * ** ** ** *** *** *** *** ** ** *** ** *** ** HML -0.37 0.29% 0.38% -0.46 -0.05 0.91% -0.68% 4.07 -0.36% -0.97% 1.04 0.06% -0.37% -0.06 0.5 -0.19 -0.07% Alpha 0.33 *** *** *** *** *** * * *** *** *** *** *** ** * SMB 0.02 0 0.33 -0.13% 1.69 0.23% 6.38 0.12% -0.18 0.11 0.29 -0.46 0.50% 0.61% 0.12 0.13% -0.42% 3.16% 0.42 0.22 0.93% 2.01 -0.19 0.05% 0.03% 1.20% 8.25% 0.01% 3.03 0.56% 4.95% 1.03% -0.14 0.14 0.89 0.02% 0.33 0.60% * * * ** Mkt 0.37% -0.18% * St.04% 0.09 0.19 *** *** *** *** ** *** *** * *** ** * ** * ** ** 177 .06% 0.03% 0.25% 0.08% 0.88% -1.Panel B: 04:2000-12:2002 Portfolio D1a D1b D1c D1 D2 D3 D4 D5 D6 D7 D8 D9 D10 D10a D10b D10C 1-10 spread 1a-10c spread 1-2 spread 9-10 spread Exc.06 0.08 0.12 0.75 -0.14 0.3 0.05% 0.27 0.36 -0.24% 0.13 -0.7 -0.34% -1.67% 4.15 0.01% 0.55% 2.04 -0.71% 1.14 0.83% 5. dev 5.30% 0.15 -0.28 -0.22% -0.40% 0.73% -0.42 0.04 0.33 0.14 0.23 0.36 0.03 0.23 0.34% 0.07% -1. return -0.36 -0.72 0.56% -0.12% -0.56% 6.74% 3.30% 1.28 0.4 0.28 0.31% -0.02% 0.93% 0.49% -0.2 0.

05 * 0.26 0.02 * * ** High Y.84 0.02 -0.66 0.19 ** -0.4 Comm.87 0.07 0.01 -0.4 0.09 -0.02 0.09 0.14 0. Monthly Exc Return is the Monthly Excess Return of the portfolio.17 *** -0.24 1.07 ** -0.15 0.02 -0.12 0.02 0.3 -0.01 -0. 0.07 -0.25 0.84 *** ** ** * 0.13 -0.19 -0.16 * -0.06 0.86 0.33 0.01 -0.Panel B: 04:2000-12:2002 Portfolio D1a D1b D1c D1 D2 D3 D4 D5 D6 D7 D8 D9 D10 D10a D10b D10C 1-10 spread 1a-10c spread 1-2 spread 9-10 spread PR1YR 0.2 * Em.05 -0.16 ** 0.1 0 -0.08 ** ** Mortg.14 0.75 0.03 -0. all funds are ranked based on their previous year's return.71 0.02 0.17 0.27 0.12 0.07 -0.27 * -0.2 -0.22 0.04 0.74 ** 0. 178 .84 0. -0.76 0.43 0. is the Standard Deviation of the Monthly Excess Return.03 0.85 0.02 0.27 -0.26 -0.2 -0.85 0. Bd 0. Dev.03 0.01 -0.14 ** 0.28 *** -0.88 0.61 ** -0.2 0.54 1.01 -0.77 0.81 0.05 0.16 0.26 *** *** *** *** *** *** *** *** *** *** *** * W x US 0. Std.79 ** -0.07 -0.19 0.12 ** -0.65 0.05 ** -0.04 -0.15 -0.33 0.95 ** -1.22 ** 0.03 0.04 -0.21 0.04 * -0.16 ** 0.54 0.42 -0.43 *** -0.01 0 0. Each year.1 ** R² adj 0.33 * -0.01 0.1 0.09 0.03 0. Funds with the highest previous year's return go into portfolio D1 and funds with the lowest go into portfolio D10.13 ** -0.11 *** This Table reports the result of the estimation of our combined model for the 01:1994-03:2000 (Panel A) and the 04:2000-12:2002 (Panel B) sub-periods.11 ** 0.2 -0.12 ** -0.73 0.32 ** 1.04 -0.02 -0.11 0.54 0.86 0.85 0.02 0.13 W Gv Bd -0..21 0. All numbers in the Table are monthly percentage.11 0.18 0.05 -0.28 *** -0.11 0.00 0.85 0.34 0.11 0. 0.82 0.11 -0.38 0.23 0.02 0.04 0.78 0.02 -0.28 ** -0.12 -0.05 ** -0.76 1.21 0.17 -0.03 ** 0.10 ** -0.03 * -0.11 0.08 -0. *** Significant at the 1% level ** Significant at the 5% level * Significant at the 10% level.18 0.09 -0.36 *** -0.03 0.33 0.08 0.71 *** 0.3 0.77 0. Portfolios are equally weighted and weights are readjusted whenever a fund disappears.09 0.

persistence in performance is observable for the medium-to-top past performers only. During the 1994-March 2000 period. it is worth mentioning that the poor adjustment of the model for the best performing decile portfolios during the down market period also signals that these funds managers tended to pursue very active and moving investment strategies. Since these decile portfolios had significant alphas for the first sub-period too. In contrast. 179 .Panels B and C display very different pictures. is not very compelling given the proven very high accuracy of our combined model. The alternative explanation of missing risk factors. Finally. except for the median decile (D4) whose loadings are significant for the High Yield factor (positive) and for the World Government Bond and Momentum factors (negative). Middle decile funds had a clearer focus on high yield bond markets. as our results suggest that this performance is not only sustained during positive or negative market conditions. Panel C shows that the persistence in performance during the market collapse was clearly sustained for portfolios D2 to D6. although theoretically possible. this indicates that the superior performance of these funds was predictable irrespective of the prevailing market conditions. These funds had no particularly remarkable investment pattern. with high significance levels. but during both. showing that only a very targeted investment behaviour in Market Neutral Funds would provide a sustained positive abnormal return. leaving a large importance to market timing and tactical allocation. with a relatively high adjusted R² of 57. the lowest decile funds did not out-perform the market.8%. which is also reported in Liang (2003). however. this particular aspect opens the way to additional research on hedge funds performance and persistence during unfavourable market conditions. We view this as a major result considering that Market Neutral funds have traditionally been assigned the role of protecting investors against negative market twists: this reinforces this claims on a double dimension. only the alphas of top decile funds were systematically higher than for the whole period. However.

04 0.03 0.03 -0.84% 0.48% 0.97% Alpha 0.07 0.16% 0.14 0.31% 1.06 0.08% -0.33% 0.30% 0.43% 0.69% 0.01 0.54% 0.14 0.14 0.37% 1.97% 1.31% 0.10 0.09 0.56% 0.51% 3.01 0. return 1.43% 1.18% 3.07 0.16 0.41% 0.61% 0.92% 0.12% 1.80% 0.04 0.44% 2.09 0.75% 3.17 0.34% 0.10 ** ** ** *** *** *** *** *** *** ** *** *** HML 0.04 0.81% 0.03 0.00% 4.45% 1.30% 2.28 0.55% 2.16% * ** *** *** *** *** *** *** *** *** *** *** *** *** *** Mkt 0.50% 0.04 0.Table 26: Hedge funds persistence for the Market Neutral strategy Panel A: 01:1994-12:2002 Portfolio D1a D1b D1c D1 D2 D3 D4 D5 D6 D7 D8 D9 D10 D10a D10b D10C 1-10 spread 1a-10c spread 1-2 spread 9-10 spread Exc.01 0 0.03 0.44% 1.06 ** * * *** * *** *** *** ** *** SMB 0.58% 0.42% 0.12 0.25 -0.45% 0.00 -0.06 0.00 0.14 -0.02 ** *** ** 180 .11 0.08% 0.10 0.02 -0.07 0.66% 0.4 0.07 -0.05 -0.69% 0.10 0.17 0.59% 1.13 0.06 0.04 0.30% 1.99% 0.30% 0.06 0.05 0. dev 3.74% 0.02 0.54% 0.15 0.25% 0.3 0.64% 0.04 -0.17 0.25 0.24 0.06 0.93% 1.37 0.21% -0.74% 1.08 0.09 0.19% *** ** *** *** *** *** *** *** *** *** *** *** *** *** *** ** St.01% 0.86% 1.40% 2.18% 5.44% 0.09 0.15 0.04% -0.50% 2.14% 0.

27 0.07 0.00 0.00 0.52 0.08 0.05 ** 0.00 0.14 181 .1 0.33 0.10 0.28 0.01 ** -0.03 -0.07 0.01 -0.11 *** -0.02 0.01 -0.44 0.40 0.17 ** -0.17 0.17 0.04 *** ** ** *** * W Gv Bd -0.00 0.04 0.53 -0.13 -0.05 -0.33 0.05 -0.06 0.16 0.27 0.24 0.11 -0.18 -0.89 -0.13 *** -0.03 0.17 -0.17 0.28 0.08 0.08 -0.16 0.21 -0.00 0. Bd -0.09 *** *** *** ** * *** *** *** W x US 0.09 -0.42 0.00 0.02 0.63 0.37 0.01 Em.42 * Comm.05 0.12 -0.05 0.01 *** *** *** ** ** *** High Y.12 -0.16 0.36 -0.08 0.03 -0.48 0.17 -0.00 0.03 -0.21 0.00 -0.03 0.02 0.02 0.02 0.10 -0.21 * 0.06 0.03 0.08 0.23 0.04 -0.64 0.06 0.04 -0.17 -0.10 -0.86 * -0.08 0.44 0.11 -0.33 0.05 0.10 0.43 0.02 0.02 0.06 0.06 0.19 0.16 *** -0.05 0.4 0.02 0.01 0.12 0.03 * R² adj 0.13 -0.02 -0.04 0.16 0.18 0.03 0.04 * * * ** Mortg.22 0.21 0.Panel A (continued): 01:1994-12:2002 Portfolio D1a D1b D1c D1 D2 D3 D4 D5 D6 D7 D8 D9 D10 D10a D10b D10C 1-10 spread 1a-10c spread 1-2 spread 9-10 spread PR1YR 0.01 0.03 0.92 * -0.36 0.02 0.15 0.03 0.11 -0.12 *** -0.02 -0.21 -0.07 -0.07 0.22 -0.26 0. 0.14 0.03 0.03 0.13 0.38 0.08 0.07 ** -0.02 -0.20 *** 0.10 0.12 0. 0.16 0.65 0. 0.12 0.02 -0.16 0.07 -0.11 0.19 0.07 0.00 0.20 0.05 -0.13 0.01 -0.04 0.

46% 0.11 0.30% 0.27% 0.40% 0.04 0.41% 1.25 ** *** *** ** * * 0.15 0.36% -0.04 -0.07 0.19 0. return St.72% 0.23 0.04% 5.63% 0.04 0.01% 1.16% *** *** *** *** *** *** *** *** *** *** *** *** * * 0.22% *** *** *** *** *** *** *** *** *** *** *** *** *** ** ** ** *** 3.88% 1.06 0.31 -0.65% 4.62% 2.06 0.99% 0.14% 0.22 0.14 -0.86% 0.04 -0.3 0.03 -0.01 0.34% 1.08 0.22 0.01 ** 0.01 0 0.23% 2.04 0.02 0.06 0.79% 1.10% 0.54% 0.43% 1.24% 2.08 0.50% 0.Panel B: 01:1994-03:2000 Portfolio Exc.70% 0.28 0.35% 0.97% 0.2 -0.19 0.67% -0.19 0.07 -0.08 -0.45% 2.24% 1.04 ** * 0.01 0. dev Alpha Mkt SMB HML D1a D1b D1c D1 D2 D3 D4 D5 D6 D7 D8 D9 D10 D10a D10b D10C 1-10 spread 1a-10c spread 1-2 spread 9-10 spread 2.11 0.38% 0.55% 0.01 * ** 2.03 -0.07 0.31% 1.31 0.11% 0.94% 182 .02 -0.03 -0.02 -0.04 0.34% 0.58% 3.50% 1.89% 0.07% 1.59% 0.01% 1.68% 1.07 0.30% 3.28 0.02 0.25 0.86% 0.58% 0.31 0.79% 0.07 *** ** *** *** *** *** 0.97% 1.62% 0.1 -0.37 0.52% 2.06 0.64% 0.71% 0.12 0.18 -0.31% 0.67% 1.04 0.47% 0.07 0.01 -0.41% 2.27% 1.17% 2.05 *** * *** *** *** ** * * *** ** 0.32 0.

61 0.1 0.05 0.13 0.27 0.13 1.08 0.05 -0.17 -0.07 0.51 0.01 0.21 0.07 -0.14 0.01 0.01 -0.03 0.01 -0.35 183 .52 ** ** 0. 1.03 0.12 0 -0.05 0.04 ** -0.38 0.39 0.11 0.06 0.47 0.04 0.22 -0.04 0.16 0.05 -0.47 0.01 0.07 0.06 0.29 -0.07 0.05 0.34 0.04 0.07 0.03 -0.07 -0.22 0.04 0.01 ** -0.39 0. -0.44 0.08 0.18 0.07 0.01 0.81 ** -0.24 0.01 -0.02 -0.02 -0.22 0.07 -0.09 0.11 0.58 0.05 -0.23 0.00 0.14 -0.71 -0.08 -0.01 -0.05 0.11 * -0.14 0.04 0.03 0.1 0.01 ** * R² adj 0.04 -0.03 0.19 0.59 0.04 -0.23 ** -0.23 0.35 -0.39 0.01 0.36 Comm.01 0.09 -0. 0.07 0.01 -0.02 -0.Panel B (continued): 01:1994-03:2000 Portfolio D1a D1b D1c D1 D2 D3 D4 D5 D6 D7 D8 D9 D10 D10a D10b D10C 1-10 spread 1a-10c spread 1-2 spread 9-10 spread PR1YR 0.27 -0.48 0.10 0.13 0.51 0.09 -0.06 0.14 0.44 0.08 0.41 ** -0.08 -0.13 -0.47 0.07 0.03 *** ** *** *** *** ** *** High Y.46 -0.14 0 * ** * *** * ** W x US 0.09 ** -0.06 -0.10 -0.09 -0.14 -0.15 0.03 -0.04 0.38 -0.01 0.18 -0.43 0.87 -0.13 -0.05 0.17 0.13 0.04 *** W Gv Bd -0.29 0.03 0.33 -0. Bd -0.18 0.06 * ** ** Mortg.23 0.11 0.11 ** -0.07 -0.09 0.16 Em.42 0.50 0.01 -0.04 0.25 -0.08 0.02 -0.31 -0.41 -0.11 0.31 -0.11 0.00 0.01 0.2 0.

Panel C: 03:2000-12:2002 Portfolio D1a D1b D1c D1 D2 D3 D4 D5 D6 D7 D8 D9 D10 D10a D10b D10C 1-10 spread 1a-10c spread 1-2 spread 9-10 spread Exc.12% * *** *** *** *** *** St.64% 0.36% 0.03 0.26 0.09 0.07% 0.1 * *** ** *** 184 .11 0.03% Mkt -0.19% -0.06 -0.22% -0.48 -1.21 -0.47% 0.21 -0.51% 0.01 0.06 0.51% 0.03 0.04 -0.02 -0.09 0.01 0.18 *** *** *** *** HML 0 -0.38 0.23 0.71% 3.67% -0.68% 2.37% 0.78% 0.73% 0. return 0.91% 0.28 0.26 -0.38 *** ** *** * ** *** SMB 0.05% 0.32% 0.02% 0.24% 0.71% 3.24% 2.05 0.57% 0.78% 2.28 -0.78% -0.49% *** 0.05% -0.01 0.23% 2.49% 0.00 0.36 0.58% 3.44 1.35% ** 0.25 -0.06% Alpha 0.28 0.83% -0.25 0.01 -0.73% 1.03 0.04 -0.55% 0.54% -0.25% 0.07 0 -0.05 0.25% -0.62% *** 0.1 0.01 0 0.12% -0.29% 0.98% 3.03 -0.02 0.02 -0.27% -0.01 -0.16 0.04 0.77% 0.05% 0.25% 0.17% 0.04 -0.17 0.72% *** 0.15 0.01 0.07 -0.25 -0.01 0 0. dev 3.23% 0.12 0.16 -0.51% *** 0.50% 0.33% 4.42% 0.03 0.49 0.15% 2.31% * 0.79% -0.38% 0.07 0.02% 0.95% 2.

08 0.12 0.18 0.29 0.13 ** ** W Gv Bd -0.06 -0.11 0.17 0. *** Significant at the 1% level ** Significant at the 5% level * Significant at the 10% 185 .2 -0.04 0.71 0.04 This Table reports the result of the estimation of our combined model for the Market Neutral strategy 01:199403:2002 (Panel A) period.29 -0.23 0.10 0. is the Standard Deviation of the Monthly Excess Return..1 0.01 0. All numbers in the Table are monthly percentage.04 0.7 1.47 0.13 0.16 0.97 -0.16 0.06 -0.46 0.36 0.4 0.19 0.01 -0. Dev.78 ** 0.84 0. 0.26 -0.33 0.01 0.03 -0.06 0.03 ** -0.08 -0.52 -0.24 -0. Bd 0.01 -0.42 0.18 -0.03 -0.11 -0.37 1.04 0.01 0 0.4 0.00 -0.55 0.26 0.03 -0.12 0.02 0.04 0.17 0.51 0.03 -0.11 -0.02 ** * ** ** ** * * High Y.04 -1 -0.74 -1.05 -0.05 -0.02 0 R² adj 0.05 -0.19 0.09 ** * ** ** ** Mortg.01 0.19 0.17 -0.09 * Em.Panel C (continued): 03:2000-12:2002 Portfolio D1a D1b D1c D1 D2 D3 D4 D5 D6 D7 D8 D9 D10 D10a D10b D10C 1-10 spread 1a-10c spread 1-2 spread 9-10 spread PR1YR 0.02 -0.06 0.03 -0.04 0. all funds are ranked based on their previous year's return.04 0.08 -0.01 -0.67 * Comm. Monthly Exc Return is the Monthly Excess Return of the portfolio.18 -0.01 -0.08 0.07 -0.16 -0. Std.43 0.36 0.02 -0.13 0.07 0.02 -0.38 0.04 -0.00 0.49 0.03 -0.05 -0.30 0.17 0.13 -0. 1.35 0.07 ** 0.11 * ** -0.15 0.79 0.00 0.04 -0. Portfolios are equally weighted and weights are readjusted whenever a fund disappears.07 0.03 -0.13 0.16 0.13 * -0. Funds with the highest previous year's return go into portfolio D1 and funds with the lowest go into portfolio D10.27 -0.12 -0.35 0.13 0.12 0. and the 01:1994-03:2000 (Panel B) and 04:2000-12:2002 (Panel C) sub-periods.01 0.01 -0.5 0.02 0 -0.11 * ** *** ** *** *** *** *** *** *** * ** ** *** W x US 0.12 0.53 0.56 0. * -0.15 0.01 -0.48 -0.39 * 0.10 -0.14 0.1 * 0.58 0.09 * -0.24 0.26 0.39 0.60 0.13 -0.51 0.12 -0.07 -0.01 -0.22 0.83 2.17 -0.26 1.36 0.12 0. Each year.01 0.51 0.44 0.02 0.59 0.

is mostly located among medium performers. however. Persistence analysis also indicates that most of the predictability of superior performance is to be found prior to March 2000. Firstly. if any.VII Conclusion The evolution of financial markets during the 1994-2002 period has been very rich in significant up and down market movements whose length and severity have been largely unprecedented. In contrast. The Market Neutral strategy provides a noticeable exception. as is sustains abnormal performance over both the bullish and the bearish sub-periods. 186 . our database constituted with 2894 funds obtained from MAR prove to be fairly trustable with respect to the most important biases in hedge funds studies. In the second sub-period. In this study. The pattern is somehow attenuated for funds of funds strategies. only negative persistence can be found among the past losers. The analysis of performance indicates that most hedge funds significantly outperformed the market during the whole test period. Our results confirm several previous studies that found that persistence. namely the survivorship and instant return history biases despite the high attrition rate of funds observed after March 2000. we have seized this opportunity to test whether hedge funds displayed significantly different patterns of performance levels and persistence during this time window as well as in undoubtedly bullish and bearish market situations. no significant under-performance of individual hedge funds of funds of funds strategies is observed when markets headed south. but this is mostly due to the bullish sub-period. Our original ten-factor composite performance model also raises little suspicion concerning its ability to explain returns as we achieve very high significance levels with very little correlation among regressors.

Obviously. but this particular field of investigation is left for future theoretical as well as empirical research. Our analysis of the performance of the Market Neutral strategy is remarkably encouraging and is confirmed and refined with the persistence analysis: for portfolios that were between the 20% and 69% best performers in this category. Market timing issues do matter for their risk exposure. probably thanks to an extreme adaptability and a very active investment behaviour. 187 . We believe that this study potentially opens the way to a deeper examination of the properties of these particular hedge funds during negative market conditions.suggesting that bad performance has probably been the decisive factor for hedge funds mortality. these very appealing results call for a much more detailed analysis of the Market Neutral strategy among individual hedge funds. abnormal performance and persistence are pervasive throughout the sub-periods. and traditional asset pricing models may not fully account for their highly unstable investment strategies.

.

Working Paper. The Sustainability in Hedge Fund Performance: new insights.J.The Sustainability in Hedge Fund Performance: new insights Daniel P. Daniel. HEC-ULG Management School . 2006. CAPOCCI HEC-ULG Management School – University of Liège (Belgium) Capocci.

Measures incorporating the volatility display very strong ability to assist investors in creating alpha as well as consistently and significantly outperforming classical indices. skewness and kurtosis. alpha. 190 . The methodology used is adapted from Capocci and Hübner (2004). volatility. The measures used include the returns. Sharpe score. beta.The Sustainability of Hedge Fund Performance: New Insights Abstract This study analyses and decomposes hedge fund returns in order to determine a systematic hedge fund selection criterion that enables investors to consistently and significantly outperform equity and bond indices over a full market cycle and over bull and bear market conditions.

000 hedge funds were managing assets totalling roughly $60 billion. 191 . when the NASDAQ Composite Index attained an all-time high of 5. It is estimated that today. In the meantime.253. (2001) found that this result is at least partly spurious. there are roughly 10.6% (see Capocci et al. 2003). contrasting with a decrease of 2.7% for the mutual fund industry worldwide (Investment Company Institute.The Sustainability of Hedge Fund Performance: New Insights Introduction The hedge fund industry has experienced explosive growth since the early 1990s when approximately 2. making them good diversification vehicles (Amin and Kat. The growing trend of the sector has been remarkably sustained during the stock market collapse that started in March 2000. 2003b). 2005). or that hedge funds have a relatively low covariance with other classes of financial assets.132 and finished three years later at a bottom of 1. 2006). the global net asset value (NAV) of hedge funds continued to grow at a steady 10. that their managers enjoy greater flexibility in their asset allocation enabling them to achieve superior market-timing skills (see Chen and Liang. This relatively positive attitude of investors is typically motivated by the perceptions that most hedge funds tend to be largely market neutral.. Assness et al.000 hedge funds managing close to a trillion USD in assets.

. The second field of hedge fund performance analysis compares the performance of hedge funds with that of mutual funds. the literature on the subject has expanded quickly. as suggested by Brown et al. Some authors (Brown et al. the hedge fund industry has a higher attrition rate than mutual funds. 2001. Since then. particularly for poor performing funds that continue to underperform. 2004). (1999) and Liang (2000. Agarwal and Naik.. Ackermann et al. (1999). They do not all converge in their findings. (1999) find that offshore hedge funds have positive risk-adjusted returns and attribute this result to style effect and conclude that there is no evidence of particular skill of some fund managers. the third group of hedge fund performance analysis examines the persistence of hedge fund returns. Agarwal and Naik (2000) analyse the presence of persistence in hedge fund returns using a one-year moving average period. Amin and Kat. Liang. whereas others (Ackermann et al. 1999 as well as Agarwal and Naik. 1999 and Liang. Capocci and Hübner (2004) conclude that some low-risk 192 . Brown et al. Ackermann et al. Three fields exist that examine hedge fund performance. Finally. Brown et al. The authors find evidence of persistence in hedge fund performance. 2003. Liang. 1999) conclude that hedge funds are able to outperform these indices. The first includes studies that compare the performance of hedge funds with equity and other indices (see for example. 1999.Hedge funds have been studied since 1997 and the early seminal studies were Fung and Hsieh (1997) and Ackermann et al. 2004) are more cautious in their conclusions. Our analysis focuses on hedge fund performance and persistence in hedge fund performance. Liang. 2003b. In this context.. although lower and more volatile returns than the reference market indices considered. 1999. Persistence is particularly important in the case of hedge funds because. 1999. (1999) and Liang (1999) find that hedge funds constantly obtain superior performance to mutual funds. 2001).

Fund of hedge funds are particular in the hedge fund world and need particular care. we find a systematic way to extract hedge funds that consistently beat traditional markets. event driven and macro) provide protection to investors when stock markets drop. as shown by Fung and Hsieh (2000). Capocci and Hübner (2004) indicated that some hedge funds consistently and significantly outperform the equity and bond markets over time. Edwards and Caglayan (2001) find that only three hedge fund strategies (market neutral. describe a systematic way of selecting hedge funds that outperform. as periods of downward trending stock markets have been rare and discontinuous between 1994 and March 2000. Our second contribution is our multi-factor model that is adapted from Capocci and Hübner’s (2004) performance analysis model. As stated by Fung and Hsieh (2006) and Liang (2003). The vast majority of performance studies on hedge funds have not focused on the behaviour of hedge funds under different market conditions. Our study makes two main contributions.44 For the period 19901998.managers have been able to persistently create alpha between January 1994 and December 2002. We test various ways of classifying funds and indicate how to select funds that consistently outperform the classical markets. 193 . 44 Most empirical evidence reveals that data collected prior to 1994 by several data vendors displays a significant survivorship bias. there is a double counting issue with funds of hedge funds that should be distinguished from individual hedge funds. however. First. The periods under examination do not favour this exercise. Liang (2000) and Capocci and Hübner (2004). They do not.

Section VIII concludes the study. Section I describes the methodology. The study is organised as following. Our results do not indicate that this is justified. we report the database and analyse the descriptive statistics. The attrition rate is calculated in Section III. In Section II. we separate individual funds from funds of funds in order to determine if there are significant differences in exposures and in alpha creation. In Section IV. we report the bias analysis. We report the global results in Sections V and VI and perform further analysis in Section VII.Following Liang (2003). 194 .

S. As reported in Table 27. at-themoney put option (ATMP) and out-of-the-money put option (OTMP). high correlations between these factors and the US market factor (ranging from 0.99 between ATMP and OTMP) indicate that the use of these factors may have two issues: 45 See Capocci and Hübner (2004) for a full description of their model as well as those that contributed to its creation.73 between ATMC and ATMP to 0.I Methodology The starting point of our study on hedge fund performance is the original Sharpe (1964) – Lintner (1965) CAPM. 195 .87 between the market and ATMP to 0. we add the four option factors of Agarwal and Naik (2004) that are at-the money call option (ATMC). and foreign bond markets (Lehman High Yield Bond Index. Salomon World Government Bond Index and the JP Morgan Emerging Market Bond Index) A commodity factor (GSCI Index) A currency factor (the Federal Reserve Bank Trade Weighted Dollar Index) In the second model. We use the Capocci and Hübner (2004) multifactor model that extends the Carhart (1997) specification by combining it with factors specific to hedge funds. respectively: Two equity market risk premium (US market and MSCI world excluding US) Fama and French (1993) size and value factors Carhart’s (1997) momentum factor Three factors to account for the fact that hedge funds invest in U.45 Our multi-factor models contain 10 and 14 factors. out-of-the-money call option (OTMC).78 between the market and ATMC) and between these four factors (from -0.

2) They may be redundant with each other. Several additional factors. because this factor has low explanatory power. 196 . Model 1 R Pt − R Ft = α P + β P1 (R Mt − R Ft ) + β P 2 SMB t + β P 3 HML t + β P 4 PR1YR t + β P 5 (MSWXUS t − R Ft ) + β P 6 (SWGBI t − R Ft ) + β P 7 ( JPMEMBI t − R Ft ) + β P 8 (HY t − R Ft ) + β P 9 (GSCI t − R Ft ) + β P10 (CUR t − R Ft ) (9) 46 Agarwal and Naik (2003) suggest that the Goldman Sachs Commodity Index is a better approximation of the commodity market than the Gold Index regarding hedge funds. the Lehman Baa Corporate Bond Index and the Salomon Brothers Government and Corporate Bond Index proposed by Agarwal and Naik (2004) and the Gold Index used by Fung and Hsieh (1997)46 were not included in our extended model given their high colinearity with our set of indices.1) They may be redundant with the market factor and not marginally contribute to the return attribution decomposition. We do not use the Lehman Mortgage-Backed Securities Index as suggested by Capocci et al. (2005). such as the MSCI Emerging Markets Index.

RFt corresponds to the risk-free return on month t. MSWXUSt is equal to the return of the MSCI World Index excluding US. These factors aim at isolating the firm-specific components of returns. 197 . SMBt is equal to the factor-mimicking portfolio for size (small minus big). SWGBIt corresponds to the return of the Salomon World Government Bond Index. HMLt corresponds to the factor-mimicking portfolio for book-to-market equity (high minus low) and PR1YRt equals the factor-mimicking portfolio for the momentum effect. JPMEMBIt equals the return of the JP Morgan Emerging Market 47 The market proxy used is the value-weighted portfolio of all NYSE. RMt equals the return of the market portfolio on month t47.Model 2 R Pt − R Ft = α P + β P1 (R Mt − R Ft ) + β P 2 SMB t + β P 3 HML t + β P 4 PR1YR t + β P 7 ( JPMEMBI t − R Ft ) + β P 8 (HY t − R Ft ) + β P 9 (GSCI t − R Ft ) + β P10 (CUR t − R Ft ) + β P 5 (MSWXUS t − R Ft ) + β P 6 (SWGBI t − R Ft ) (10) + β P11 ATMC t + β P12 OTMC t + β P13 ATMP t + β P14 OTMP t + ε Pt where RPt is equal to the return of fund P in month t. Amex and NASDAQ stocks – a market proxy that is usually used in mutual fund and hedge fund performance studies.

Every January.Table 27: Correlation between the Option Factors and the Market Factor Market ATM Call OTM Call ATM Put OTM Put 0. Every year we rank all funds based on their total return of the prior year. ordered from highest to lowest past returns. Portfolios 1 (low) and 10 (high) are then further subdivided in three sub-portfolios on the same measure. εPt equals the error term. and at-the-money and out-of-the money European put option factors.85 ATM Call 1 0. OTMCt. we place all funds into 10 equally weighted portfolios. we follow the methodology of Carhart (1997). Funds that disappear during the course of the year are included in the average until their 198 .73 -0.96 -0. GSCIt corresponds to the return of the Goldman Sachs Commodity Index. while αP and βP are the intercept and the slope of the regression.73 -0.78 0. HYt is equal to the return of the Lehman High Yield Credit Bond Index. In order to determine if some funds consistently and significantly create alpha over time. The portfolios are held until the following January and then rebalanced again. Bond Index.69 -0.7 OTM Call ATM Put OTM Put 1 -0.99 1 This table reports the correlation between the option factors and the market factor over the January 1994-December 2002 period. out-of-the money European call option factors.87 -0.65 1 0. respectively. CURt equals the return of the Federal Reserve Bank Trade Weighted Dollar Index and ATMCt. ATMPt and OTMPt correspond respectively to Agarwal and Naik’s (2004) at-the-money.

as measured by the standard deviation. standard deviation. and then portfolio weights are readjusted appropriately. 199 . the next to bottom (D9) and the bottom portfolio (D10) and finally between extremes alphas (D1a and D10c). we also test the significance of the difference between the alpha of the top (D1) and the bottom (D10) decile portfolios.death. alpha. the top (D1) and the next (D2) decile portfolio. skewness and kurtosis. beta. This yields a time series of monthly returns on each decile portfolio from January 1995 to December 2002. In each case. 48 48 The Sharpe ratio is the ratio of excess performance over the risk-free rate to risk. Since past performance is only one element in analysing past returns. we decide to perform the decile classification methodology incorporating the Sharpe ratio .

The Barclay’s database includes 1./Centre for International Securities Derivatives Markets (MAR/CISDM).246 individual funds including 1.271 funds that were still alive at the end of 2002. Inc. we removed funds that appear twice in the same database49 and funds with quarterly 49 This happened in three cases: when the same fund (same name. 696 individual dissolved funds as well as 469 funds of funds including 384 funds of funds that were still in existence at the end of 2002 and 85 dissolved funds of funds. the Managed Account Reports. In each database.967 individual funds including 1. Nevertheless.185 funds that were still living at the end of 2002 and 1. and returns) appeared twice in the database.061 dissolved individual funds as well as 647 funds of funds including 436 funds of funds that were still in existence at the end of 2002 and 211 funds of funds that have been dissolved. For a majority of funds. manager's address. investment strategy. We combine hedge fund data from MAR/CISDM with Barclay’s data. Hedge Fund Research. management and incentive fees. 200 . assets under management. Each provider proposes its classification. etc.II Database Four main hedge fund databases are available for empirical studies. manager's name. and when the same fund (same company and returns) appeared twice in the database with two different fund names. The data providers collect information supplied by hedge fund managers. The MAR database started with 2. company. but on many occasions there maybe some similarities. there is no consensus on the definition of the strategy followed. they record other useful information such as company name. The first three are the most used in academic studies. when the same fund (same name and returns) appeared twice in the database with two different company names. starting and ending date. Inc.. TASS Management and the Barclay’s database.

1. We obtain a total of 3.910 individual funds and 588 funds of funds from the MAR/CISDM database and 1. We separate them for three main reasons.190 (38. This means that funds of funds returns should be lower than individual fund returns by 1% to 3% annually only because of the presence of the survivorship bias in the underlying funds. we want to avoid double counting individual funds that are part of funds of funds and that are in the database.15%) surviving individual funds.153 funds that were present in the two databases (834 existing funds and 319 dissolved funds). which leads to the presence of a second survivorship bias. 2000). These funds include a total of 1. All the returns used are net of fees. More precisely. We further found 1.85%) dissolved individual funds. First. On the other hand. There is a double survivorship bias in their returns. 201 . Secondly. 653 (72%) surviving funds of hedge funds and 254 (28%) dissolved funds of funds. Most previous studies on hedge funds group individual funds and funds of funds. we separate individual funds from funds of funds because funds of funds returns can be biased.873 (61. individual funds in funds of hedge funds portfolios can also be dissolved.returns.060 individual funds and 907 funds of funds including 1. This database is one of the largest and most unique databases ever used in hedge fund performance studies. there are funds of funds that are dissolved and which lead to the presence of survivorship bias in the data when we do not consider dissolved funds of funds (this is the same notion of survivorship bias as suggested in many other academic studies like Fung and Hsieh. We follow Liang (2003) and analyse them together and separately. This would lead to what we call the fund of funds bias.150 individual funds and 319 funds of funds from the Barclay’s database.

. 202 . This should at least partly be explained by the double fee structure (see Brown et al. 2002). 1999).Finally. Depending on the performance of the individual funds compared to their hurdle rate and/or high watermark. the funds may charge a performance fee and the mix of these fees will impact the return distribution. the third reasons why we separate funds of funds from individual funds is that performance and persistence models are usually less precise for funds of hedge funds than they are for individual hedge fund strategies (see Liang.

we present the descriptive statistics. 3. whereas short sales and funds of hedge funds have offered the lowest returns (0.8%). 50 Ratio of dissolved funds over the period covered to the number of funds in the database for the corresponding strategy. 203 . The first column indicates that the bulk of the funds reported in the database are defined as market neutral funds or global funds that each represents around 24% of the global individual hedge funds database. Interestingly. the correlation analysis and the bias analysis.56% monthly).87 for short sales and topping out at 4.III Preliminary Analysis Before analysing the presence of bias in hedge fund data. This confirms the presence of fat tails in hedge funds return distribution. this ratio is also high for funds of funds (at 72%). The skewness column indicates that most return distributions are negatively skewed indicating that average negative returns are higher than the average positive returns. whereas the least volatile funds are market neutral funds (1. The excess kurtosis reported is all positive beginning with 0.22%).98 for distressed securities. Sector funds have offered the highest mean monthly return (1. The mortality ratios50 reported are almost all between 50% and 70% indicating high mortality levels for individual hedge funds. Funds not classified are the most volatile.1 Descriptive Statistics Panel A of Table 28 reports the descriptive statistics of the hedge fund database.

80 Std dev.08 2.60 2.81 1.Distressed sec .04 3.88 1.Event driven Event Driven Total Global emerging Global Macro Global Global Total Mkt ntl Equity Hedge Sector Short sales Long only leveraged Mkt timing Currency fund Option strategy No strategy Individual funds total Fd of Fds 122 168 59 349 226 190 702 1118 731 332 277 48 142 43 4 3 13 3060 907 % 4% 5% 2% 11% 7% 6% 23% 37% 24% 11% 9% 2% 5% 1% 0% 0% 0% 100% NA Dead funds 61% 77% 17% 61% 58% 69% 46% 49% 64% 67% 70% 71% 73% 93% 25% 0% 8% 60% 72% Mean 1.96 2.56 0.62 4. 2.35 3.89 1.26 1.70 1.14 1.89 1.68 1.73 1.81 204 .23 1.07 3.99 1.25 1.80 1.02 0.15 1.94 3.02 0.85 1.03 0.91 1.90 Median 1.52 2.09 1.04 1.12 2.15 0.01 0.36 0.00 1.85 1.05 1.53 2.16 0.01 1.Risk arb .13 1.Table 28: Descriptive Statistics Panel A: Hedge Funds Descriptive Statistics N .20 1.13 0.73 0.09 0.17 0.13 1.22 2.19 2.65 1.09 1.12 1.02 1.

89 15.16 -11.14 -6.08 9.32 0.29 3.54 0.40 3.94 1.39 0.83 17.27 -6.32 7.59 3.45 -11.19 -0.99 4.87 1.19 0.17 0.33 7.17 -0.51 1.18 8.32 0.25 0.27 0.24 -8.21 0.91 9.Panel A (continued): Hedge Funds Descriptive Statistics Skewness Kurtosis Min Max Sharpe score .46 -4.19 0.88 9.02 -10.66 -4.27 0.19 -0.08 -12.23 -7.13 1.61 8.72 3.33 11.21 0.65 0.20 -10.37 6.95 -0.00 -11.97 7.33 0.48 -0.27 -0.17 -9.24 -0.06 -0.32 2.89 -4.19 10.25 2.44 10.61 6.10 -0.73 0.27 -0.Risk arb .37 0.43 2.91 1.36 7.19 3.25 0.23 0.85 3.16 4.23 -0.57 0.98 2.62 4.Event driven -0.44 -5.75 -13.20 6.33 0.93 0.44 -10.04 -8.16 -8.25 0.22 Event Driven Total Global emerging Global Macro Global Global Total Mkt ntl Equity Hedge Sector Short sales Long only leveraged Mkt timing Currency fund Option strategy No strategy Individual funds total Fd of Fds 205 .82 2.63 -0.66 6.34 0.11 0.Distressed sec .32 0.11 2.

97 89.35 0.16 0.01 1. High Yield 0.11 0.02 -56.16 -1.11 0.59 5.66 -0.45 SMB 0.55 -59.6 4.39 1.72 MSCI w.34 0.51 -20.11 0.76 1. -0.58 Skewness -0.99 ATM Call 0.81 Broad Dollar Index 0.27 Risk-free rate 0.88 95.48 PR1YR 1.09 4.79 ATM Put -11.24 0.14 5.73 95. 4.77 Median 1.82 84.36 0.74 1.83 0.29 0.68 0. ex US 0.4 -0.Panel B: Naive Strategies Descriptive Statistics Mean Market 0.34 1.13 Sal.62 OTM Put -15.37 0.45 -0.13 0.58 OTM Call -2.7 GSCI 0.87 HML 0.36 0.45 JPM EMBI 0.5 1.35 0.73 Lehman Agg.33 206 .02 4.27 -0. WGBI 0.19 -0.78 Std dev.4 2.47 -0.82 4.57 -30.61 0.67 1.64 Leh.

01 0.85 386. Sal.26 -8.14 0.26 5.21 -2.15 -0.160 dissolved funds) and 907 funds of funds (including 254 dissolved funds of funds) over the January 1994-December 2002 period.64 1.66 3.91 -25.44 -24.21 3.060 individual funds (including 1.16 0.19 0.18 0.13 -2.29 -96.6 0.00 0.29 Min -15.41 2.47 7.38 13.56 1.28 0.02 400.87 5.58 0.67 18.1 -95. Our original MAR/CISDM/Barclay’s database consists of 3.33 10.37 0. High Yield GSCI Risk-free rate ATM Call OTM Call ATM Put OTM Put Broad Dollar Index 0.2 -7.79 Max 8.11 NA NA NA NA NA NA This table reports the descriptive statistics for hedge fund strategies (Panel A) and for naïve strategies (Panel B).27 21.04 -99. is for standard deviation.33 3.79 0. N stands for number of funds and Std dev.94 10.49 15.73 -3.69 -12.37 -12.02 0.14 4.87 7. WGBI JPM EMBI Leh.93 Sharpe score 0.27 0.56 216.18 3.78 -0. ex US SMB HML PR1YR Lehman Agg.61 0.26 0.11 -99.89 -16. 207 .91 246.Panel B: Naive Strategies Descriptive Statistics Kurtosis Market MSCI w.20 -0.

2 Correlation Analysis Traditional hedge fund literature contends that. hedge fund strategies are highly correlated. Liang. between the hedge fund strategies and the passive investment strategies and among the passive investment strategies. asymmetry to the left and presence of fat tails. hedge funds are likely to improve the riskreturn trade-off when added to a traditional portfolio (see Fung and Hsieh. 51 We do not report the results for the sake of brevity. The equity indices returns are more volatile than most hedge fund strategies.75. 208 . They are all reasonably low. with the exception of the correlation between the option factors of Agarwal and Naik (2004) and between these factors and the market. with the exception of short sellers that systematically has the opposite result. 1997. The skewness and kurtosis measures indicate the same pattern as for hedge funds. 2003). we also verified the correlation between the explanatory factors. but complete results are available directly from the author. To avoid any risk of multi-colinearity. 3. thanks to the weak correlation between hedge funds and other securities.5 (except short selling) and most of them are higher than 0.Panel B reports the same statistics for passive investment strategies. The coefficients reported are all higher than 0.51 As typically reported in the literature (see Liang. The bond and equity indices all have lower mean returns than the majority of the hedge fund strategies. This justifies the use of two different models. but the standard deviations of the bond indices are almost all lower than those of the hedge fund strategies. 1999 as well as Amin and Kat. 2003b). We analyse the correlation among the hedge funds strategies.

Survivorship bias is usually defined as the performance difference between living funds and all funds (Fung and Hsieh. Complete results are available upon request.54 On a yearly basis. there is a survivorship bias in the figures because dissolved funds tend to have inferior performance compared to surviving funds. 2000 and Liang.3 Survivorship Bias The most important bias in hedge fund data is the survivorship bias (see Fung and Hsieh. 2001). 52 By living funds we mean funds still in operation at the moment of the analysis. Fung and Hsieh. When only individual funds are considered.96%.5 bias reported by Malkiel 53 54 (1995) for US mutual funds. 2000). When only living funds52 are considered.44%.8-1.3.30% monthly bias found by Fung and Hsieh (2000). We find this consensus value quite high when compared to the 0. figures reported are a little higher at 1. 2000 and 2001 have been years of important differences in results between living and dissolved funds.53 This result is lower than the 0.08% per year. 209 . the 3% bias found by Liang (2001) and the industry consensus bias of 3% stressed by Amin and Kat (2003a). we obtain 0. We obtain a bias of 1. For funds of funds. 1997.

4. almost all strategies are significantly exposed to the US equity market (inversely for short sellers) with a beta ranging from 0. Every strategy but macro funds. we analyse the persistence in performance and look for a systematic way of classifying funds that enables investors to consistently and significantly outperform equity and bond indices over time.IV Global Results: Market Analysis Our analysis has two parts. This is perfectly in line with the strategy since option strategies tend to focus the S&P 500. First.25 for short sellers to 0. On the exposure side. Few strategies are significantly exposed to non-US equity markets. no sub-strategy funds and funds of funds create significant alpha at the 5% significance level.1 Performance Analysis Panel A of Table 29 reports the results obtained while using our multi-factor model (1). Option arbitrage strategies have been biased towards large companies. The size factor is significantly positive for every strategy. 210 . The results reported are particularly strong. The style factor indicates a value bias for risk arbitrage and market neutral funds and a growth bias for currency funds. the momentum factor is significantly positive for most strategies indicating that many hedge funds tend to be momentum players.66 for sector funds. Interestingly. we focus on the performance of hedge funds by grouping them per investment strategy. option strategy funds. except for option arbitrage funds and short sellers meaning than hedge funds profit from the small companies’ out-performance over the period under analysis by being small cap biased. Then.

On the bond side. The exception is short sellers that are slightly negatively exposed to this factor and currency funds. the exposure to the US equity market disappears. Second.26 and 0. On the other side. We performed the same analysis adding the option factors of Agarwal and Naik (2004) and report the results in Panel B of Table 29. The r-squared are particularly high between 0. the alpha is no longer significant. for global emerging markets funds. but no significant change in the alpha or in the other exposures. This result indicates that results may become unstable with the addition of auto-correlated factors. but that they offer at-the-money call option return features.72. First. whose exposure is not significant. The same result is found for the market exposure of market timers. but the exposure to the ATMC option factor is significant. but the exposure to the factor remains identical. for most strategies the alpha and the exposures to the factors (in terms of significance) remain the same. This result interestingly indicates that the returns offered by emerging market funds are not so highly correlated with the US equity market. 211 . there is a significant exposure to the ATM call factor. Another interesting element is that the alpha of global funds is no longer significant. A comparison of the two panels indicates two interesting patterns. The equity emerging market factor is strongly significantly positive for almost every strategy.89 with an average at 0. For currency funds. Such managers offer emerging market-like returns when the market is going up (like a call option on developed markets since the upside potential of emerging markets is much higher than that of developed markets in bull markets conditions). the world bond index factor is significantly negative for some strategies but no single strategy is significantly exposed to this factor indicating an inverse. but insignificant relation between hedge fund returns and the bond market. such managers tend to cut the exposure in bear market conditions.

28 0.09 0.06 0.31 0.12 0.03 0.24 0.09 0.47 0.42 0.08 0.51 0.02 0.03 *** *** *** PR1YR 0.13 0.32 0.01 0.04 0.09 0.25 0.07 0.7 0.09 0.05 0.03 0.06 -0.47 0.01 0.17 0.34 0.16 *** *** *** *** *** *** *** *** *** *** *** *** *** *** *** * *** *** *** *** HML 0.12 0.07 0.16 0.2 0 0.06 0.05 0.22 0.Table 29: Hedge Fund Strategies Performance Analysis (1/199412/2002) Panel A: Model (1) Alpha .07 ** *** *** *** *** *** ** *** * ** *** * *** *** *** ** * 212 .07 0.58 0.04 0.18 0.58 0.03 0 0.46 0.22 0.17 0.14 0.47 0.19 0.33 -0.01 -0.17 0.3 0.42 0.28 0.Distressed sec .04 0.23 0.06 0.08 0.08 0.02 0.08 0.16 0.01 0.44 0.47 0.08 0.28 0.08 0.12 0.43 0.29 0.14 0.19 0.06 0 0.05 0.01 0.12 -0.2 0.03 0.46 0.12 0.12 0.06 0.25 0.29 0.07 0.03 0.31 0.4 *** * *** *** *** *** *** ** * *** ** *** *** *** *** *** *** ** S&P 0.18 0.49 0.56 0.11 0.03 0.12 0 0.32 ** *** *** *** *** *** *** *** *** ** *** *** *** ** *** *** *** *** MSCI w ex US 0.66 -0.12 0.04 -0.06 0.Event driven Event Driven Total Global emerging Global Macro Global Global Total Mkt ntl Equity Hedge Sector Short sales Long only leveraged Mkt timing Currency fund Option strategy No strategy Individual funds total Fd of Fds Global Index 0.04 0.41 0.23 0.04 0.04 -0.3 0.45 0.04 * * * * *** *** * SMB 0.22 0.36 0.Risk arb .06 -0.05 0.14 0.

07 0.04 0.07 -0.11 * * * ** ** * ** 0.15 -0.17 -0.08 -0.01 -0.03 0.06 0.01 0 ** 0.04 -0.05 0.01 0 0.82 0.09 0.05 -0.05 0.07 -0.24 -0.08 -0.25 0.37 -0.28 -0.77 0.05 0.28 -0.08 -0.78 0.14 -0.4 * -0.02 -0.73 0.04 0.07 0.08 -0.04 0.09 0 0.04 0.06 0.07 -0.26 0.01 -0.17 -0.1 -0.04 0.08 -0.73 0.04 * -0.88 0.11 0.05 0.02 0.07 -0.23 * -0.02 0.76 0.06 0.11 0.08 0.03 0.02 -0.05 -0.52 0.87 0.07 0.07 0.65 0.07 0.82 0.47 0.Event driven Event Driven Total Global emerging Global Macro Global Global Total Mkt ntl Equity Hedge Sector Short sales Long only leveraged Mkt timing Currency fund Option strategy No strategy Individual funds total Fd of Fds Global Index -0.04 0.07 0 -0.15 -0.06 -0.16 0.Panel A (continued): Model (1) WGBI EMBI HY GSCI DOLLAR R² .24 0.1 * -0.86 0.12 0.Risk arb .07 0.01 0.03 -0.89 0.06 -0.09 0.05 *** 0.06 *** *** *** *** *** ** *** *** *** *** *** * *** *** * ** ** *** *** *** 0.48 0.22 0.39 -0.78 0.06 -0.1 0.02 0.02 0.86 ** * * 0.13 ** -0.88 0.03 213 .12 -0.06 -0.04 0.Distressed sec .63 0.1 -0.05 0.07 0.13 -0.11 *** *** *** ** *** * *** ** 0.04 0.01 0.04 -0.12 -0.06 -0.06 0.01 * -0.

01 0.49 *** 0.30 0.28 *** 0.05 0.32 *** -0.05 PR1YR 0.19 0.17 *** 0.04 0.03 0.07 *** 0.12 *** -0.15 *** 0.23 *** 0.26 *** 0.18 *** -0.37 *** 0.38 0.28 *** 0.08 0.09 *** -0.10 *** 0.14 *** 0.34 *** -0.14 *** HML 0.02 0.03 0.05 0.04 0.14 *** 0.45 *** 0.Distressed sec .06 *** 0.06 EMBI 0.09 *** 0.75 *** S&P 0.19 *** 0.06 0.07 *** -0.06 -0.00 0.02 0.30 *** 0.09 0.22 *** 0.44 *** 0.31 0.49 *** -0.09 0.19 *** 0.21 0.50 *** 0.14 *** 0.08 0.03 ** -0.04 -0.06 *** 0.04 *** 0.03 0.28 *** 0.12 *** 0.06 *** -0.08 *** -0.27 * 0.00 0.00 0.11 0.48 *** 0.08 *** -0.00 -0.04 0.42 *** 0.10 *** 0.08 0.07 *** -0.06 0.05 * * *** *** * * SMB 0.05 *** 0.18 *** 0.04 0.16 *** 0.07 *** 0.06 0.22 *** 0.12 * -0.29 *** 0.03 -0.15 ** 0.52 *** 0.39 *** 0.12 *** -0.PANEL B: Model (2) Alpha .08 *** 0.11 *** 0.39 ** 0.06 *** 0.Risk arb .03 0.08 0.05 0.08 ** -0.07 *** 0.48 *** 0.41 *** 0.23 *** 0.60 *** 0.13 *** 0.58 *** MSCI w ex US 0.01 0.18 *** 0.03 0.08 *** -0.12 * 0.12 0.24 *** 0.00 0.31 0.05 0.17 * 0.09 0.07 0.09 * 0.13 0.01 -0.42 *** 0.07 ** 0.07 *** 0.12 *** 214 .23 *** 0.04 0.04 * WGBI -0.04 ** 0.04 0.07 ** 0.06 *** -0.01 0.07 ** 0.07 *** -0.06 *** 0.Event driven Event Driven Total Global emerging Global Macro Global Global Total Mkt ntl Equity Hedge Sector Short sales Long only leveraged Mkt timing Currency fund Option strategy No strategy Individual funds total Fd of Fds Global Index 0.05 ** 0.07 -0.02 0.24 ** 0.03 0.29 0.01 0.25 *** 0.04 0.09 0.26 0.12 ** 0.19 *** 0.06 *** 0.10 * 0.29 ** 0.16 *** 0.20 *** 0.13 * 0.17 *** 0.06 -0.14 ** 0.67 *** 0.

07 *** 0. DOLLAR = return of the Federal Reserve Bank Trade Weighted Dollar Index and ATMCt.00 0.00 -0.01 -0.01 -0.10 *** 0.14 ** -0.07 *** 0.06 0.01 0.01 0.06 *** 0. out-of-the money (OTMC) European call option factors.74 0. HMLt = the factor-mimicking portfolio for book-to-market equity (‘high minus low’).15 * -0.02 0.10 * 0. GSCIt = return of the Goldman Sachs Commodity Index.00 0.01 * -0.24 ** 0.05 This Table presents the results of the performance estimation of the combined Model for the January 1994December 2002 period.10 * -0.00 0.00 0.03 0.00 -0.36 -0.06 *** 0.00 0.79 0.04 * 0.00 0.01 OTM CALL 0.48 0.06 -0. SMBt = the factor-mimicking portfolio for size (‘small minus big’).00 0.05 0.02 0.07 ** 0.12 0.02 ** 0.09 -0.00 0.03 -0. ATMPt and OTMPt = respectively Agarwal and Naik (2004) at-the-money (ATMC).01 0.03 * 0.00 0.50 0.00 0.87 0.00 0.15 ** -0.00 -0.14 * 0.00 0.05 ** -0.02 * 0.05 -0.28 ** -0.07 *** -0.01 0.Event driven Event Driven Total Global emerging Global Macro Global Global Total Mkt ntl Equity Hedge Sector Short sales Long only leveraged Mkt timing Currency fund Option strategy No strategy Individual funds total Fd of Fds Global Index HY GSCI DOLLAR 0.33 * -0.00 0.30 -0.87 0.00 0.02 * -0.05 *** 0.04 * 0.04 -0.00 0.87 0.00 0.10 -0.06 *** 0.04 * -0. ** Significant at the 5% level and * Significant at the 10% level.08 *** 0.00 0.00 OTM PUT -0.00 ATM PUT 0.05 * 0.04 0.00 0.00 0.04 0.01 *** 0.02 0.00 0.01 0.00 0.04 0.04 0.02 0. and at-the-money (ATMP) and out-of-the money (OTMP) European put option factors.77 0.06 *** 0.04 *** 0.00 0.00 0.07 *** -0. EMBIt = return of the JP Morgan Emerging Market Bond Index.07 -0.00 0.07 0.22 -0.04 0.09 *** 0.00 0.00 0.10 * 0.04 ** 0.89 0.83 0.10 0.83 0. *** Significant at the 1% level.00 0.02 -0.00 0.01 0.20 ** -0.78 0. HYt = return of the Lehman High Yield Credit Bond Index. 215 . MSWXUSt = return of the MSCI World Index excluding US.Risk arb .Distressed sec .00 0.02 0.01 0.04 -0.04 0.03 0.88 0.23 -0.79 0.66 0.00 0.00 0.54 0.00 -0.09 0.00 0.01 0.00 0. S&P stands for the market factor.14 * ATM CALL 0.06 * 0.04 -0.00 0.12 -0. OTMCt.67 0.89 0. PR1YRt = the factor-mimicking portfolio for the momentum effect.01 0.01 0.00 0.03 -0.06 -0.01 R² 0.01 0.00 0.00 0.13 * 0.PANEL B (continued): Model (2) EMBI .02 ** 0. T-statistics are heteroskedasticity consistent.00 0.00 0.05 0.01 0.16 -0.00 0.03 * 0.19 ** -0.01 0.29 0.02 0. SWGBIt = return of the Salomon World Government Bond Index.74 0.07 ** 0.00 0.20 *** 0.00 0.01 0.01 0.00 0.01 0.01 0.00 -0.00 -0.00 0.00 0.01 0.

we report only the results of model (1). the returns increase almost monotically between deciles D1 and D10. but not for the previous year’s best and worst performing funds. but that there is a serious risk of multi-colinearity. to high yield bonds and marginally to commodities. The factor exposures enable us to distinguish between the deciles. The volatility decreases between deciles D1 and D4 and then increases to attain its maximum at decile D10 indicating that the previous year’s best funds are also the most volatile. The previous year’s best performing funds are strongly exposed to small companies. 4. small cap biased and value oriented. 55 A closer look at the adjusted-R² indicates that it increased from 0 for most strategies to a maximum of 0.04 for market timers confirming the fact that option factors may be redundant with the market factor. 216 . Moreover. These alpha creators can be distinguished from the previous year’s poorly performing funds that do not create pure alpha and are exposed to the world excluding US equity markets while being momentum contrarian. As one would expect. pure momentum players with a short exposure to the USD.Panel B indicates that the option factors may be helpful in analysing hedge fund returns for some specific strategies55. in the rest of the study. Alpha creators (decile D2 to D8) have been long equities. they have been exposed to emerging market bonds.2 Persistence in Performance . Therefore.Returns 1. The alphas reported are significantly positive for deciles D2 to D8. Global Database Table 30 reports the results of the estimation for the global database (including individual funds and funds of hedge funds) for the entire January 1995 to December 2002 period.

All in one. the mean returns and the volatility reported are marginally higher and the alphas reported are significantly positive for all but the previous year’s best performing funds. Individual Hedge Funds Versus Funds of Hedge Funds We perform the same analysis by separating individual hedge funds from funds of hedge funds. 217 . investors should invest in low volatility funds but if the invest in higher volatility funds.56 For individual funds. 56 The corresponding tables are available upon request. The spreads of the alphas reported indicate that there is no significant difference in alpha between decile D1 and D10 or even between the extremes alphas (D1a and D10c) but that there is a significant difference between decile D9 and D10. The values obtained are very close to the results obtained for the whole database. the bottom one offering a significantly higher alpha (-0. This result indicates that there is no consistence in returns for the most volatile funds in the database. The volatility of these funds is so high that the returns obtained are no longer significantly different from zero.48% per month) but at a cost of higher volatility. Ideally. 2. our model clearly explains most of the hedge fund’s performance and enables us to distinguish between funds. A corollary is that investors looking at directional hedge funds should favour the one with the highest past performance independently of the volatility. the one offering the best past returns (decile D10) should be favoured to the one that offer less returns even if the volatility is lower.

93 1.02 -0.14 0.02 -0.85 0.08 0.27 0.91 1.31 4.09 0.30 1.03 -0.07 0.00 0.73 0.26 0.06 0.72 2.50 0.16 1.23 0.05 0.07 0.16 0.01 -0.45 *** *** ** * *** * ** *** *** *** *** *** ** * S&P 0.62 4.04 0.31 0.10 0.46 1.52 ** *** *** *** *** * *** *** *** *** *** *** *** *** *** MSCI w ex US 0.14 3.43 1.55 6.06 0.05 0.50 0.40 0.72 0.04 0.16 1.41 1.07 0.17 2.03 0.09 0.40 1.72 4.02 Alpha 0.07 0.45 0.34 Std dev.33 0.32 0.02 -0.43 0.35 0.05 0.00 -0.12 0.21 0.01 0.29 0.15 0.52 0.36 0.09 0.33 0.03 * * *** 218 .07 0.60 -0.05 -0.39 0.06 -0.14 -0.06 -0.05 0.81 3.37 0.12 -0.15 0.84 2.Table 30: Hedge Fund Persistence in Performance (1/199512/2002) Panel A: Multi-factor estimation .30 0.83 0.01 -0.58 0.17 0.21 0.Global Funds (1/1995-12/2002) Mean D1 D2 D3 D4 D5 D6 D7 D8 D9 D10 D1a D1b D1c D10a D10b D10c Spread 1-10 Spread 1-2 Spread 9-10 Spread 1a-10c 0.03 0.44 0.31 0.65 1.34 0.26 0.24 0.44 0.79 2.09 0.12 -0.50 0.01 0.19 0.36 -0.31 ** ** ** *** *** *** *** * SMB 0.38 0.30 0.48 0.07 0.12 -0.01 1.04 -0.67 0.02 0.32 1.01 -0.22 3.46 *** *** * *** *** *** *** ** *** *** *** *** *** *** *** *** HML -0.28 1.65 -0.36 0.34 0.82 5.32 0.21 0.14 -0.43 0.52 3. 3.09 0.06 0.90 7.56 -0.02 0.20 0.33 0.81 0.06 -0.19 0.48 1.02 1.

51 0.26 * * ** * *** *** *** *** *** *** *** HY 0. We report the OLS estimators for equally weighted portfolios per investment strategy.07 0.13 -0.58 0.32 -0.02 0.17 -0. SWGBIt = return of the Salomon World Government Bond Index.01 0.07 0.18 -0.07 0.06 -0.75 0.06 0.09 0. S&P stands for the market factor.13 -0.03 0.06 * *** * * DOLLAR -0.14 -0. HYt = return of the Lehman High Yield Credit Bond Index.02 0.04 0.05 0.05 0.06 0.16 -0.55 0.03 -0.25 -0.83 0.25 0. PR1YRt = the factor-mimicking portfolio for the momentum effect.15 0.13 0.06 0.00 -0.78 0.30 -0.11 -0.04 0.11 -0.14 -0.15 -0.07 * * * R² 0. HMLt = the factor-mimicking portfolio for book-to-market equity (‘high minus low’).80 0.Panel A (continued): Multi-factor estimation .24 -0. every fund is classified into a decile on the basis of its performance over the previous year.03 -0.07 -0.08 -0.05 0.82 0.04 0.30 -0.45 -0.05 -0.08 -0.07 0.54 0.01 0.19 0.05 0.04 -0.04 0.32 0.03 0.08 0. GSCIt = return of the Goldman Sachs Commodity Index.02 -0.05 0.82 0.51 This Table presents the results of the persistence in performance estimation of the combined Model for the January 1994-December 2002 period.07 0.49 -0.03 0. Each year.05 -0.10 -0. OTMCt.04 0.79 0.12 0.32 -0.00 *** *** *** *** * *** ** *** *** ** ** * * EMBI 0. and at-the-money (ATMP) and out-of-the money (OTMP) European p 219 .70 0.06 0.76 0. EMBIt = return of the JP Morgan Emerging Market Bond Index.07 -0.09 0.05 -0.07 0.55 0.09 -0. MSWXUSt = return of the MSCI World Index excluding US.21 -0.33 0.19 -0.Global Funds (1/1995-12/2002) PR1YR D1 D2 D3 D4 D5 D6 D7 D8 D9 D10 D1a D1b D1c D10a D10b D10c Spread 1-10 Spread 1-2 Spread 9-10 Spread 1a-10c -0.02 0.04 -0.07 0.83 0.09 0.20 0.01 0. DOLLAR = return of the Federal Reserve Bank Trade Weighted Dollar Index and ATMCt.09 -0.08 0.29 -0. ATMPt and OTMPt = respectively Agarwal and Naik (2004) at-the-money (ATMC).04 0.06 0.07 0.10 -0.82 0.05 0.03 0.05 0.12 0.02 0.16 -0.10 -0.35 -0.12 -0.82 0. SMBt = the factor-mimicking portfolio for size (‘small minus big’).69 ** *** *** *** *** *** *** *** *** *** *** *** *** *** *** *** *** *** WGBI -0.20 0.40 -0.03 0.01 0.01 0.32 0.02 0.15 0.02 0. sub-strategy and for all funds.85 0.27 -0.00 -0.07 * * ** GSCI 0.12 0.06 0.03 0.12 -0.12 -0.09 0.29 0.06 0. out-of-the money (OTMC) European call option factors.63 0.08 0.

Alphas are however significantly positive for deciles D2 to D5. no single fund of fund has been momentum contrarian. Finally. the emerging market bond factor is significant for every decile when funds of funds are considered. 220 . The exposure to ex-US equity markets is lower and no more significant for top deciles. alphas and standard deviations are lower than the corresponding statistic for individual funds. Secondly. 4. The alphas reported in Table 31 are consistent with our one-year classification period analysis. There are four main differences in the factor exposures.Funds of funds mean returns. There are however some differences in terms of exposures. Thirdly. the negative exposure to the government bond index is significantly lower. All in one. To take this constraint into account. The exposure to the emerging market bond index changes quite importantly. but the R² remains high. No significance for top and bottom deciles. The bottom deciles (D9 & D10) that were not significantly exposed to this factor are when funds are classified on three years of data. Funds cannot be distinguished by their exposure to the world excluding US equity markets or by their exposure to the emerging market bond index. individual funds outperform fund of funds and they can be distinguished by their respective factor exposure. Firstly.3 Persistence in Performance – Three Years of Data Many investors require funds to exist for at least three years before investing in order to be able to analyse their returns quantitatively. the currency factor is no longer significant at all. we performed the same analysis by considering three years of data to classify funds into deciles for the next 12 months. This classification is recalculated every year on the basis of the returns of the previous three years. Table 31 reports the results for the global database.

12 0.59 0.18 *** *** *** *** *** *** *** ** *** *** *** *** *** *** ** *** ** * HML 0.03 0.39 0.17 0.09 1.08 0.57 0.35 0.02 0.04 0.38 3.02 4.48 0.97 1.20 -0.09 0.95 1.38 0.00 0.03 *** *** *** *** *** *** *** *** *** *** *** *** *** *** * MSCI w ex US 0.06 0.10 -0.18 0.41 0.14 0.43 2.08 0.03 -0.11 0.02 -0.20 2.14 0.05 4.03 0.90 0.91 6.26 0.06 -0.86 1. 3.34 2.81 0.05 * ** * ** SMB 0.03 0.Table 31: Persistence in Performance Analysis Based on Threeyear Data (1/1997-12/2002) Panel A: Multi-factor estimation .40 0.18 0.11 0.77 ** *** * * * *** *** *** *** *** * S&P 0.03 -0.08 0.46 0.27 0.68 0.91 3.00 -0.36 0.02 0.69 Std dev.99 1.49 -0.90 0.50 -0.01 0.05 0.28 -0.06 0.83 0.19 0.19 0.06 0.04 1.79 0.01 -0.17 0.43 1.26 0.29 0.10 0.04 -0.61 4.30 0.32 0.01 -0.00 0.43 0.20 -0.02 -0.24 * * * * 221 .05 -0.07 0.06 0.06 -0.03 0.09 0.04 0.66 0.06 0.39 1.08 0.37 0.02 0.97 2.06 0.28 Alpha 0.81 2.67 5.07 0.68 0.50 -0.08 -0.23 0.13 0.43 0.01 -0.05 0.13 0.09 -0.56 0.15 0.33 0.19 0.33 0.01 3.02 0.41 0.57 0.48 0.Global Funds classified on their performance (1/1997-12/2002) Mean D1 D2 D3 D4 D5 D6 D7 D8 D9 D10 D1a D1b D1c D10a D10b D10c Spread 1-10 Spread 1-2 Spread 9-10 Spread 1a-10c 0.13 -0.02 1.38 1.41 0.25 0.07 0.90 4.91 0.02 0.59 -0.41 -0.59 0.

06 0. MSWXUSt = return of the MSCI World Index excluding US.31 -0.06 0. We report the OLS estimators for equally weighted portfolios per investment strategy.56 0.35 -0.31 0.01 0.34 -0.01 0.25 -0.07 -0.05 0.36 -0.22 -0.03 0.10 * *** ** * ** * GSCI 0.15 -0.06 0.26 -0.66 0.10 0.10 0.11 0.02 0.08 -0.27 -0.07 -0.35 *** *** *** *** *** *** * *** *** *** *** *** ** *** *** WGBI -0.31 -0.12 0.11 -0. GSCIt = return of the Goldman Sachs Commodity Index.06 0.07 0.08 0.21 -0.01 0.06 0.05 -0.10 -0.10 0.11 0.03 -0.19 0.01 0.10 0.05 0.05 0. DOLLAR = return of the Federal Reserve Bank Trade Weighted Dollar Index.04 0.04 -0.72 0.88 0. HMLt = the factor-mimicking portfolio for book-to-market equity (‘high minus low’).02 0.Panel A: Multi-factor estimation .02 0. EMBIt = return of the JP Morgan Emerging Market Bond Index. SMBt = the factor-mimicking portfolio for size (‘small minus big’).04 -0. *** Significant at the 1% level.08 -0. HYt = return of the Lehman High Yield Credit Bond Index.18 -0.27 -0.71 0.02 0.61 0.66 0.43 This Table presents the results of the persistence in performance estimation of the combined Model for the January 1994-December 2002 period.34 ** * ** R² 0.24 -0.21 0.03 ** ** ** *** ** ** *** * ** * EMBI 0.08 0. T-statistics are heteroskedasticity consistent.89 0.19 -0. 222 .09 0.09 0. every fund is classified into a decile on the basis of its performance over the previous three years.06 0.07 0.17 -0.01 0.16 0.14 -0.08 -0.89 0.03 0.20 -0.05 0.10 -0.08 -0.27 -0.15 0.04 0.00 * *** *** *** *** * *** *** *** *** *** *** *** *** HY 0.11 0. Each year.03 0.05 0.88 0. sub-strategy and for all funds. S&P stands for the market factor.01 -0.08 0.04 -0.05 0.49 0.18 -0.56 0.15 -0.49 -0.00 -0.02 0. ** Significant at the 5% level and * Significant at the 10% level.01 0.08 0.15 -0.05 0.86 0.85 0.14 0.07 0.14 0.87 0.12 0.04 -0.02 0.14 -0.10 0.08 0.16 0. PR1YRt = the factormimicking portfolio for the momentum effect.Global Funds classified on their performance (1/1997-12/2002) PR1YR D1 D2 D3 D4 D5 D6 D7 D8 D9 D10 D1a D1b D1c D10a D10b D10c Spread 1-10 Spread 1-2 Spread 9-10 Spread 1a-10c -0.06 * ** DOLLAR -0.80 0.14 -0.06 0.22 -0.10 0.02 0.23 -0.03 0. SWGBIt = return of the Salomon World Government Bond Index.17 -0.05 0.06 -0.80 0.83 0.00 0.19 -0.05 0.53 -0.13 -0.20 -0.

The spreads of the alphas reported are in line with the one obtained in the previous table. Individual funds outperform and create significant alpha in almost all cases. There is no significant difference in alpha between decile D1 and D10 or between the extremes alphas (D1a and D10c) but there is a significant difference between decile D9 and D10. 223 . This result confirms that if investors want to invest in high volatility/directional hedge funds they should favour the one with the highest past performance independently of the volatility. the bottom one offering a significantly higher alpha (-0.49% per month) but at a cost of higher volatility.The results obtained for individual funds and funds of hedge funds are in line with those obtained on a global level.

In order to go further in the analysis. Sharpe ratio57 2. 224 . we perform the same analysis with other ways of classifying funds into deciles. Our analysis indicates that some relatively low volatility hedge funds consistently and significantly outperform classical equity and bond indices over time. We would have obtained exactly the same decile classification if we would have used the Sharpe ratio since the same rate would have been deducted from the performance of all the funds considered. However.4 Persistence in Performance – Other Measures Absolute performance is the most important element of the hedge fund industry. we use the Sharpe score (ratio of mean returns over standard deviation). 58 We performed the same analysis using the minimum monthly return (for investors that want to limit very bad losses in any month) and the maximum monthly return (for investors that want to achieve punctual high returns) and the results obtained were in close to those obtained using the standard deviation as a rule for classifying the funds. some strategies are very risky whereas others attempt to offer stable returns. Alpha 4. An adapted Sharpe ratio 57 Practically. Beta 5. The corresponding Tables are available upon request. Skewness/Kurtosis 6. We use the following measures as classification parameters: 1. Standard deviation58 3.4.

The differences are reported in Figure 6. Disaggregation can happen if the deciles are not homogeneous that is when funds of the same strategies are grouped in the same decile. is reported in Figure 6. Figure 6 indicates that the two largest differences happen in decile D2 (-11% for global funds and +10% for market neutral funds). Figure 6 reports the difference between the percentage of funds of each strategy and in each decile. we have to check for the risk of disaggregation. 0. Figure 5 shows that there are some variations in the repartition of decile per strategy over time but these variations are not large. When we look at each strategy one by one. we see more funds of funds in decile D4 to decile D8 and less in the top deciles but marginal differences like this can also be explained by the fact that many funds of funds have been created recently and that the global database does not have the same unbiased repartition every year since new funds have been added or removed every year.9% of distressed funds.9%. 225 . The difference. We check this element and report the results for alpha and beta in Figure 5 to Figure 8.Before using these other ways of classifying hedge funds based on past performance. Note that the percentage of global funds decreased from around 21% to 16% and the percentage of market neutral funds increased from 16% to 21% over time. Consider the following example. There are 3% of distressed funds in the global database. There are for example 3% of distressed funds in the global database. The first two figures report the repartition of decile per strategy. Decile 1 has on average 3. We are only talking about some percentages of differences.

HE E Q DG MK NTL T G BAL LO MACRO G E L MER ED RA DS 10% 5% 0% D5 D6 D7 D8 D10 D9 DB D1 D2 D3 D4 226 .Figure 5: Decile Repartition per Strategy (Ranking base on Alpha) 35% 30% 25% 20% 15% FO F NOS TRAT O PTIO N CU RRENCY MK TIMING T LGLE V S RT HO S CTO E R E .

Figure 6: Strategy Repartition in the Alpha Ranking 12% 10% 8% 6% 4% 2% 0% -2% NOSTRAT CU RRENCY LGLEV SECTO R MKT NTL MACRO -4% -6% -8% -10% -12% D5 DS ED D1 D2 D3 D4 D6 D7 D8 D10 D9 227 .

E G M N KT TL G LOBAL M R AC O G E E L MR E D R A D S 10% 5% 0% D10 D8 D9 D6 D7 D5 D3 D4 DB D1 D2 228 .Figure 7: Decile Repartition per Strategy based on Beta 35% 30% 25% 20% 15% FO F N STR O AT OP TION C R EC URN Y M TIM G KT IN LG LE V SH R O T SE TO C R E HD E Q.

Figure CCC report the repatition in deciles based on alpha. Figure CCC+1 reports the difference with the global database. D1 D2 D3 D4 229 .Figure 8: Strategy Repartition in the Beta Ranking 12% 10% 8% 6% 4% 2% 0% -2% FO F NOS TRAT O PTIO N CU RRENCY MK TIMING T LGLE V SHO RT SE CTO R EQ HEDG . Figure CCC and CCC report the same information when funds are classified in deciles on the basis of their past beta. E MKT NTL G BAL LO MACRO G E L MER ED RA DS -4% -6% -8% -10% -12% D5 D6 D7 D8 D10 D9 These figures report the decile repartition between hedge fund strategies.

Figure 7 and Figure 8 show the same information when the ranking used to construct the deciles is based on beta. First. The Omega is like a return distribution that includes higher moments. We have also considered the omega from Keating and Shadwick (2002) as well as the Sortino ratio but we choose not to use them for the same reason. these 14% are ranked in middle decile. We could have decided to take a longer period under review to estimate the quality of the manager (at least five years or ten years) but then we would have limited the period of test of the ranking methodology since the global period under review covers 10 years of data in total. Results are a little more mitigated even if we remain at relatively low variations. 59 The value of the Omega function at level r is the ratio of probability weighted gains relative to r. we cannot use the omega. as a measure based on a distribution of returns. the omega59 needs a relatively long dataset to be estimated. In the present case. It gives a risk-reward measure for which returns are weighted by their probability and shows the interest of combining different investments in a single portfolio. Figure 5 and Figure 6 lead us to conclude that there is no risk of aggregation when alpha is used as a ranking measure. These results comfort us in the use of beta as a ranking tool for hedge funds.For funds of funds for example the sum of the differences reported is 14% meaning that there are 14% funds of funds more in the deciles reported based on alpha than in the global database. Funds of funds tend not to be in the top decile and global funds tend not to be in the bottom deciles. Since our persistence analysis methodology is based on one year and three years of data to rank the funds before analysing the persistence in their performance. 230 . to probability weighted losses relative to r.

By definition. it uses Downside Deviation. Another time. On aggregate 2576 hedge funds (65% of the 3967 funds of our database) have less than five years of data meaning that taking five years of data or more to rank funds will lead us not to consider these funds in our analysis. taking longer period of ranking to classify funds will lead to a survivorship bias in the database because dissolved funds will not be taken into account. 1075 funds out of theses 2576 funds have been dissolved before the end of the period. Three years of data seems to be the best tradeoff between stable statistics and the length of the period under review. most fund will or at least should suffer only a few months down. 231 . our estimation of the negative standard deviation will be unstable if we consider 1 or 3 years of data. Moreover. except that instead of using standard deviation as the denominator. hedge funds aim at offering positive returns irrespective of the market. Because of this. In addition. it is important to stress out that the issue is even more important is our case because individual funds are considered mainly because most funds exist for a few years only. 60 The Sortino ratio is similar to the Sharpe Ratio. the data available do not allow us to cover 5 or 10 years of data to classify funds before analysis the persistence.The same issue emerges with the Sortino ratio60. On a three year period.

93 for decile D10. Funds included in decile D1 offer an average monthly performance of 0. Various studies used the Sharpe ratio while analysing hedge fund returns (see for example. Liang. Amenc et al. Ackermann et al. while decile D10 includes the funds offering the most attractive risk-return trade-off over the three years used to classify funds. The sub-deciles reported in the second part of the Table 32 indicate that the increase in the Sharpe score is particularly for sub-decile 10a to 10c indicating that a limited number of funds strongly distinguish themselves. It is reasonable to use this measure while analysing the persistence in hedge fund returns. Investors want to see if hedge funds are able to produce consistent returns over time. 2003). The Sharpe score decreases slightly between decile D2 and decile D5 and increases monotically thereafter to reach 0. al. 1999.Sharpe score Many investors investigating hedge funds focus on their risk/return characteristics.79% against a mean monthly return of 0. 232 .93% for the best performing funds.7% with a monthly standard deviation of 2. 2003a) and most of them stressed the shortcut of this measure for alternative strategies like those applied by hedge fund managers (see for example.. 2003 and Goetzmann et. The idea behind this analysis is the following: will investors outperform if they focus at the previous year’s risk/return trade-offs of the hedge fund industry instead of focusing on pure performance only.. Decile D1 contains the funds having the lowest Sharpe score calculated using three years of data. 1999 as well as Amin and Kat. The Sharpe score – defined as the ratio of the average return to the standard deviation – aims at taking risk and returns into account. the use of this measure remains very predominant in the industry despite its weaknesses. the risk and the average Sharpe score of each decile.. The first three columns of Table 32 report the return. Nevertheless. Such investors pay a particular attention to the trade-off between the risk and the returns offered.

The results indicate that even if there are around 30% of the funds that offer significant alpha based on the Sharpe ratio.The alpha column indicates that there is persistence for the previous year’s top funds. The spreads of the alphas reported are not significant but for the spread between the most attractive funds decile D9 and decile D10. They also have been momentum contrarian. 233 . underperforming funds have not been exposed to the US equity market but have been exposed to the world excluding US. you would not only offer significant positive alpha but you would also significantly outperform significantly the next 10%. Table 32 interestingly indicates that if you invest in the top 10% of the funds that offered the highest Sharpe ratio over the last three years. A comparison with the results obtained when we classified hedge funds on the basis of their performance confirm the need to take risk and return into account in order to identify what kind of funds offer persistence in returns. Results obtained by separating individual funds from fund of funds are in line. This means that is it important to make a good selection and also that you can make a good selection (decile D8 and D9) or a top selection significantly better (decile D10). Funds classified in deciles D8 to D10 create alpha. the top one are apart from the others. On the other side. short world government bonds and short the USD. they have been value-oriented and that they have been exposed to the high yield market. The factor exposure indicates that top risk-return performers have had a limited exposure to the US equity market.

Standard Deviation Some investors focus on volatility using hedge funds as pure risk diversifiers. while being much more volatile – indicating that higher volatility funds do not necessarily lack of consistency in performance. 3) High volatility funds that do not offer persistent over-performance. Every year we classify funds on the basis of the standard deviation of their previous three years of returns.1. The mean and standard deviation columns of Panel A indicate that returns increase with volatility for low volatility funds. do volatile hedge funds consistently and significantly outor under-perform less volatile hedge funds? The answer is in Table 33. Stated differently. but that the trend is unclear for more volatile funds (decile D5 to D10). the alpha created by decile D5 is strongly significant (as for example. investors may use hedge funds instead of bonds in their portfolio and the most important aspect to them is to limit the volatility. decile D1). These results enable us to divide the hedge fund industry into three segments: 1) Low volatility funds that offer persistent over-performance. In this case. This analysis is interesting because it aims at determining if there is any proof of good or bad performance in hedge fund return on the basis of the volatility of their returns only. 2) Low/medium volatility funds that offer persistent over-performance. Interestingly. 234 . The alpha column confirms that low volatility funds create significant alpha and that high volatility funds do not. We perform the same analysis by focusing purely on volatility.

you will not only offer significant alpha but also significantly outperform the next alpha creators. The spreads of the alphas reported show the same pattern as with the Sharpe ratio. Unlike other deciles they also have not been momentum players. alpha creators again exhibit slight exposures to the US equity markets and they have been long high yield. The results indicate that even if there are between 30% and 50% of the individual funds that outperform. Our results indicate that less volatile funds tend to outperform the classical markets and to become more volatile when they cannot offer the expected performance. In terms of exposures. Higher volatility funds that offer consistent returns will have a relatively stable volatility (since returns are relatively stable) and consequently remain in segment two. the results of individual funds and funds of hedge funds are in line with those obtained for the global database with higher alpha reported for individual funds. The r-squared are very high. while volatile and inconsistent managers fall in segment three. the top one are apart from the others. 235 . Low volatile funds that cannot create consistent performance tend to become more volatile and switch from segment one to two or three. As before. This result confirms that it is not sufficient to be a momentum player to make money.This result leads to the conclusion that low volatility funds tend to consistently and significantly outperform over time. Selecting low volatility funds enable investors to create significant and consistent alpha over time but if you select the lowest funds that are available.

04 -0.01 0.23 * 0.16 1.03 0.68 0.47 *** 0.12 0.10 0.16 0.58 1.14 MSCI w ex US 0.99 1.01 -0.08 ** 0.02 -0.09 -0.87 0.06 -0.07 -0.03 0.08 *** -0.25 0.37 *** 0.03 0.01 0.39 0.59 0.Table 32: Persistence in Performance Analysis Based on the Sharpe score (1/1997-12/2002) Panel A: Multi-factor estimation .49 3.21 *** HML -0.10 1.03 0.86 -0.72 0.46 *** 0.89 0.12 *** 0.02 -0.93 0.11 * -0.34 0.26 Alpha 0.03 0.19 *** 0.04 0.23 0.03 0.81 0.11 ** 0.44 *** 0.06 *** 0.27 0.15 0.08 0.25 0.01 0.19 *** 0.20 -0.21 *** -0.23 0.94 0.12 *** 0.25 0.Global Funds classified on their Sharpe Score (1/1997-12/2002) Sharpe Score 0.05 *** 0.05 0.12 2.07 -0.04 0.08 *** 0.06 *** 0.74 0.01 -0.51 *** 0.03 0.42 1.66 ** 0.06 *** 0.41 0.27 * 0.41 *** 0.05 -0.48 *** 0.30 *** -0.07 *** 0.21 ** 0.56 0.45 *** 0.02 0.89 0.93 0. 2.01 0.07 236 .94 0.71 0.20 0.98 3.03 0.06 ** 0.23 -0.46 Std dev.00 0.03 ** -0.01 0.41 2.24 3.79 3.10 *** 0.03 0.58 2.01 0.09 0.02 0.02 -0.60 * 0.77 0.20 0.25 0.06 0.28 *** -0.28 0.01 0.23 ** 0.12 -0.04 -0.16 *** 0.47 0.01 -0.16 *** 0.49 *** 0.25 0.25 ** S&P 0.13 ** 0.02 -0.01 0.40 0.05 0.39 ** 0.71 3.71 2.00 3.25 0.02 0.30 *** SMB 0.70 0.52 *** 0.24 Mean D1 D2 D3 D4 D5 D6 D7 D8 D9 D10 D1a D1b D1c D10a D10b D10c Spread 1-10 Spread 1-2 Spread 9-10 Spread 1a-10c 0.33 *** 0.31 0.16 *** 0.82 0.01 -0.91 2.04 0.62 3.02 0.82 0.03 -0.93 0.06 *** 0.71 0.

75 0.02 0.03 0.09 0.10 -0.02 -0.87 0.06 -0.05 -0.09 0. We report the OLS estimators for equally weighted portfolios per investment strategy.01 0.06 0.65 -0.04 0.14 *** *** *** ** *** *** *** *** * ** *** *** *** *** *** *** ** ** WGBI -0.01 0.10 0.07 0.22 -0.06 0.06 0.38 -0.48 0.05 0.01 0. EMBIt = return of the JP Morgan Emerging Market Bond Index. HYt = return of the Lehman High Yield Credit Bond Index.15 0.07 -0.11 -0.09 -0.33 -0.05 0.05 0.03 0.01 0. SWGBIt = return of the Salomon World Government Bond Index.02 0.51 0. S&P stands for the market factor.06 0.74 0.06 0.12 -0.03 -0. GSCIt = return of the Goldman Sachs Commodity Index.13 ** * * * ** * * GSCI 0.89 0.Global Funds classified on their Sharpe Score (1/199712/2002) PR1YR D1 D2 D3 D4 D5 D6 D7 D8 D9 D10 D1a D1b D1c D10a D10b D10c Spread 1-10 Spread 1-2 Spread 9-10 Spread 1a-10c -0.55 -0.14 0.09 0.11 -0.86 0. HMLt = the factor-mimicking portfolio for book-to-market equity (‘high minus low’).30 -0. DOLLAR = return of the Federal Reserve Bank Trade Weighted Dollar Index.14 0.10 0.05 -0. *** Significant at the 1% level.02 ** ** * DOLLAR -0.09 -0.26 -0.10 0.13 *** ** ** *** *** * *** EMBI 0.08 -0. PR1YRt = the factor-mimicking portfolio for the momentum effect.00 0.02 0.07 0.60 0.36 0.36 -0.43 0.01 -0.71 0.51 0.07 0. SMBt = the factor-mimicking portfolio for size (‘small minus big’).07 0.09 0.13 -0.52 -0.03 0.61 ** * ** ** ** ** ** * * ** R² 0. every fund is classified into a decile on the basis of its Sharpe score over the past three years.28 -0.01 -0. ** Significant at the 5% level and * Significant at the 10% level.12 0.01 *** *** *** *** *** *** *** *** *** *** *** *** HY 0.15 0.06 0.06 -0.01 0.06 0.06 -0.07 0.59 -0.01 0.10 0.11 0.81 0. Each year.03 0.08 0.04 0.07 -0.02 -0.04 -0. 237 .06 0.10 0.56 -0.01 0.01 0.19 -0.04 0.04 -0. MSWXUSt = return of the MSCI World Index excluding US.13 -0.19 -0.02 0.90 0.07 0.09 0.04 -0.03 -0. sub-strategy and for all funds.72 0.06 0.04 0.10 -0.90 0.09 0.02 0.12 0.02 0.42 0.Panel A (continued): Multi-factor estimation .26 -0.02 0.15 -0.04 0.11 0.10 0. T-statistics are heteroskedasticity consistent. with the Sharpe score defined as the ratio of performance to volatility.27 This Table presents the results of the persistence in performance estimation of the combined Model for the January 1994-December 2002 period.89 0.69 0.03 0.03 -0.09 -0.11 -0.03 0.03 0.68 0.06 0.

22 0.01 0.28 0.00 0.17 -0.03 0.30 0.33 0.36 -0.76 0.46 0.12 0.07 0.47 0.01 0.07 0.11 0.04 0.22 0.06 0.11 0.02 0.94 6.02 0.23 0.30 0.04 0.69 0.01 -0.07 -0.20 0.04 *** *** *** *** * * *** *** *** *** *** S&P 0.03 -0.12 0.02 Std dev.19 0.06 0.81 7.45 -0.03 -0.16 -0.31 0.70 0.01 -0.02 0.14 0.08 0.04 -0.21 6.67 0.45 0.91 0.12 -0.05 0.05 0.77 0.18 -1.41 0.45 -0.13 0.17 1.62 2.67 -0.09 0.47 0.97 0.63 0.74 0.41 0.76 -0.06 0.06 -0.13 0.02 0.03 0.01 *** SMB 0.44 2.56 0.14 *** ** * * ** * *** * 238 .71 0.13 0.27 0.68 0.58 2.02 0.33 0.02 0.91 0.09 0.01 0. 0.02 0.11 -0.85 0.04 0.87 0.52 0.02 0.59 0.01 0.Table 33: Persistence in Performance Analysis Based on the Standard Deviation (1/1997-12/2002) Panel A: Multi-factor estimation .01 -0.65 0.17 0.69 0.29 0.01 -0.44 ** *** *** *** *** *** *** *** *** *** *** *** *** *** *** *** *** HML 0.72 6.05 *** *** *** *** *** *** *** *** *** *** *** *** *** *** *** MSCI w ex US 0.01 0.96 1.31 0.16 -0.00 0.34 0.77 0.02 0.70 Alpha 0.02 0.01 -0.27 7.06 -0.Global Funds classified on their standard deviation (1/1997-12/2002) Mean D1 D2 D3 D4 D5 D6 D7 D8 D9 D10 D1a D1b D1c D10a D10b D10c Spread 1-10 Spread 1-2 Spread 9-10 Spread 1a-10c 0.32 0.36 0.07 0.71 0.03 0.04 0.02 0.21 6.46 0.15 -0.16 0.82 0.01 0.34 0.58 1.28 2.99 3.09 0.45 4.

18 * * *** *** *** *** *** *** *** *** *** *** *** *** * HY 0.53 0.13 0. SWGBIt = return of the Salomon World Government Bond Index.07 0. We report the OLS estimators for equally weighted portfolios per investment strategy.90 0.10 -0.02 0. GSCIt = return of the Goldman Sachs Commodity Index.02 0. 239 .35 0.12 0.89 0.08 0.08 -0. S&P stands for the market factor.02 0.61 0.00 0.01 0.06 0.06 -0.28 -0.51 0.25 R² 0. *** Significant at the 1% level.07 0. ** Significant at the 5% level and * Significant at the 10% level.75 This Table presents the results of the persistence in performance estimation of the combined Model for the January 1994-December 2002 period.07 0.12 0.02 -0.00 0.89 0.01 -0.63 0.04 -0.02 -0.85 0.08 0.04 0. DOLLAR = return of the Federal Reserve Bank Trade Weighted Dollar Index.01 0.09 -0.03 0.00 -0.24 -0.16 -0.12 0. T-statistics are heteroskedasticity consistent.07 0.82 0.07 0.09 0.07 -0.07 -0.09 0.10 0.12 0.10 0.07 0.Panel A (continued): Multi-factor estimation .05 0.53 0.15 -0.11 0.02 0.82 0.10 0.01 -0.01 -0.79 0. Each year. HYt = return of the Lehman High Yield Credit Bond Index.51 0. PR1YRt = the factor-mimicking portfolio for the momentum effect.02 0. sub-strategy and for all funds.01 0.26 -0.80 0.14 -0.01 0. SMBt = the factor-mimicking portfolio for size (‘small minus big’).10 0.07 0.13 0.11 -0.01 -0.59 0.07 0.04 0. every fund is classified into a decile on the basis of its standard deviation over the previous three years.09 0.32 0.01 0.45 EMBI 0.06 0.05 0.17 0.07 -0.01 0.76 0.05 0.01 0.10 -0.08 0.12 -0. EMBIt = return of the JP Morgan Emerging Market Bond Index.02 0.08 0.67 0.21 -0.06 -0.04 -0.07 *** *** *** *** *** *** *** *** * WGBI -0.28 *** ** *** *** *** *** * * GSCI 0.01 0.06 0.02 0.85 0.17 0.14 -0.15 -0.20 -0. HMLt = the factor-mimicking portfolio for book-to-market equity (‘high minus low’).03 0.11 DOLLAR -0.09 -0.Global Funds classified on their standard deviation (1/199712/2002) PR1YR D1 D2 D3 D4 D5 D6 D7 D8 D9 D10 D1a D1b D1c D10a D10b D10c Spread 1-10 Spread 1-2 Spread 9-10 Spread 1a-10c 0.12 0.02 0.28 -0.77 0.02 -0.01 0.06 -0.81 0.07 -0.03 0.07 -0.11 -0. MSWXUSt = return of the MSCI World Index excluding US.12 -0.12 0.11 -0.23 0.03 0.07 0.04 0.04 0.11 0.08 0.01 -0.04 0.30 -0.12 0.16 0.

Alpha Hedge funds aim at offering absolute returns in any market environment. The exposures reported are in line with the ones previously obtained. the standard deviation decreases between decile D1 and D6 before increasing again until decile D10. The only major difference is that alpha creators are momentum players according to those results. pure alpha creators will or will not be considered momentum players.2. The trend in alpha is not clear. The current analysis is based on the following statement: do managers that create higher alpha over a period persistently outperform the equity. but significant for the middle deciles D4 to D8. 240 . bond and commodity markets over the next period? The mean return decreases between decile D1 and D10. Table 34 reports the decomposition of hedge fund returns classified on the basis of the alpha created over a three-year period. The volatility column confirms once again that low volatility funds can create significant and consistent value over time. This result confirms the instability of this factor in distinguishing pure alpha creators since depending on the way of classifying funds. No spread is significant meaning that the funds that offered the highest alpha or lowest alpha in the past do not distinguish from each other in the future. Investors who use them as a return enhancer compare the part of their portfolio allocated to hedge funds to the proportion of their portfolio invested in equities to determine if they are a good return enhancer.

The r-squared obtained are particularly high. The factor exposures indicate that alpha creators have not all been long US equities (decile D10 is even net short the U. The spread between the high beta funds (decile D1) and the low beta funds (decile D10) is significantly positive at 0. pure hedge funds should have a limited exposure to the equity market. there is a difference between the good and the top ones.94. As one would expect. Beta Table 35 reports the results obtained by classifying funds on the basis of their beta (high beta funds in decile D1). Almost all deciles from deciles D5 to D10 significantly and consistently create value. 241 . equity market but not significantly) but that these funds have been value oriented and exposed to the high yield market.S.. Spread 9-10 is also significant but negative. The funds that offer the highest alpha are much better than the one outperform the underperforming funds. Our results suggest that low beta funds that also a limited volatility can significantly and consistently create value over time. The spreads report very interesting results.3. with a maximum of 0. High beta funds are the most volatile and they do not outperform classical markets over time.92 meaning that high beta funds significantly underperform low beta funds by more than ninety basis points a month when the factor exposure in taken into account. This result indicates another time that in the alpha producers.

This means that the distribution goes from negatively skewed to positively skewed. The first interest of Table 36 is that the volatilities reported in the second column are relatively stable over the 10 deciles. The kurtosis reported in the next column goes from high values for negatively skewed returns. In Panel A. the alpha column is not easy to interpret. the skewness increases from decile D1 to decile D10.4. Second. Panel A of Table 36 reports the results obtained by classifying funds on the basis of their skewness (the lower the skewness. to low values for unskewed return distributions. skewed distributions have fat tails. In short. In both cases. 242 . Since there is no formal difference in the volatility of the deciles considered. Skewness and Kurtosis Investors like positive skewness but do not like fat tails. in each case we report the skewness and the kurtosis because our results indicate that there is a direct link between these two measures. Volatility is on the downside. this result again confirms the need to take volatility into account when analysing hedge funds and when looking for funds offering persistent performance over time. Classifying funds on the basis of their skewness or their kurtosis is not the best way to try to understand and predict hedge fund returns. It confirms these findings since funds with fat tails also have the highest skewness. This result is very different from what we obtained previously. Finally. Panel B reports the same analysis and focuses on the kurtosis. the lower the decile). performing deciles are spread out. This relation is even stronger when considering negatively skewed distributions. Funds tend to have fat tails on the left-hand side of their mean.

43 0.01 0.03 -0.18 -0.12 *** -0.73 2.28 0.05 -0.78 1.25 0.33 *** 0.03 -0.12 0.78 0.93 0.36 0.12 *** 0.29 *** 0.03 0.09 *** 0.11 0.05 0.33 *** 0.46 *** 0.81 0.00 0.24 0.00 -0.53 0.58 1.03 0.15 *** 0.27 HML -0.51 0.26 *** 0.30 *** 0.81 0.56 4.21 *** 0.10 *** 0.59 *** 0.Global Funds classified on their alpha (1/1997-12/2002) Mean D1 D2 D3 D4 D5 D6 D7 D8 D9 D10 D1a D1b D1c D10a D10b D10c Spread 1-10 Spread 1-2 Spread 9-10 Spread 1a-10c 0.69 0.84 0.44 0.01 -0.36 *** 0.39 *** -0.98 2.21 0.50 0.07 SMB 0.10 0.67 2.22 *** 0.10 *** 0.17 -0.22 3.10 -0.26 0.82 0.77 0.05 0.36 5.90 *** 0.14 *** 0.12 -0.92 0.53 3.44 5.31 *** 0.24 *** 0.08 0.18 *** 0.47 3.04 * 0.87 0.15 -0.03 0.11 0.10 6.07 0.88 2.08 -0.17 Std dev.01 0.11 -0.15 ** ** *** *** *** S&P 0.89 1.01 0.10 *** 0.20 * -0.66 *** 0.10 *** 0.23 *** 0.01 3.22 *** 0.19 0.76 1.02 0.05 0.03 0.33 *** 0.85 Alpha 0.06 0.25 *** 0.54 *** 0.13 -0.10 -0.00 0.17 -0.51 * MSCI w ex US -0.08 * -0.02 0.17 1.11 0.01 -0.18 *** 0.02 0.01 -0.19 -0.03 -0.40 4.06 0.75 0.03 0.02 -0.79 0.23 1.04 243 .04 -0.34 *** 0.12 *** 0.94 2.08 0.30 *** 0.02 0.79 0.65 0.37 ** 0. 5.Table 34: Persistence in Performance Analysis Based on the Alpha (1/1997-12/2002) Panel A: Multi-factor estimation .04 -0.

08 0.06 0. 244 .13 -0.04 -0.11 -0.81 0.07 0.04 -0.06 -0.07 0.05 -0. HMLt = the factor-mimicking portfolio for book-to-market equity (‘high minus low’).03 0.01 0.04 0.07 0.15 0.04 -0.03 -0.11 0.06 0.27 EMBI 0.02 0.02 0.03 0.04 -0.04 0.09 -0.02 0.10 0. MSWXUSt = return of the MSCI World Index excluding US.10 -0. HYt = return of the Lehman High Yield Credit Bond Index.28 This Table presents the results of the persistence in performance estimation of the combined Model for the January 1994-December 2002 period.06 0.14 0.62 -0.10 0.06 0. PR1YRt = the factor-mimicking portfolio for the momentum effect.Panel A (continued): Multi-factor estimation .21 0.00 -1.13 0.02 0.13 * * GSCI 0.04 0.13 0. SMBt = the factor-mimicking portfolio for size (‘small minus big’).15 * DOLLAR -0.13 0.25 0.06 0.77 0.33 -0.04 0.02 -0.10 -0.12 0.00 0.07 0.12 0.01 0.03 0.08 -0.12 -0.40 0.07 0.89 0.00 0.81 0.75 0.07 -0.09 -0.02 0.12 -0.37 -0.10 0.13 -0.67 -0.05 0.85 0.08 0.02 0.79 0.06 -0.08 0.08 -0.05 0.23 -0.02 0.62 -0.07 0.76 0. EMBIt = return of the JP Morgan Emerging Market Bond Index.01 0. every fund is classified into a decile on the basis of its alpha (relative to the S&P 500 equity index) over the past three years.13 -0.01 ** *** * * ** *** *** *** *** *** *** ** WGBI -0.06 -0.02 0.07 0.09 0.78 0.10 0.Global Funds classified on their alpha (1/1997-12/2002) PR1YR D1 D2 D3 D4 D5 D6 D7 D8 D9 D10 D1a D1b D1c D10a D10b D10c Spread 1-10 Spread 1-2 Spread 9-10 Spread 1a-10c -0.07 0.11 -0.87 0.09 ** ** *** *** ** *** *** ** *** *** *** *** HY 0.19 -0.07 0.11 0.12 -0.07 0.76 0. *** Significant at the 1% level.41 0.13 -0. Each year.18 -0.02 0. T-statistics are heteroskedasticity consistent.57 -0.64 0.04 0.76 0. GSCIt = return of the Goldman Sachs Commodity Index.16 0.15 -0.01 -0. S&P stands for the market factor.05 -1.07 0.08 0.04 -0.51 ** ** ** R² 0.05 0.79 0.74 0.06 0.02 -0.11 0.07 0. DOLLAR = return of the Federal Reserve Bank Trade Weighted Dollar Index. SWGBIt = return of the Salomon World Government Bond Index.81 0.24 -0.16 0. ** Significant at the 5% level and * Significant at the 10% level.02 0.06 0.02 0.08 0.13 -0.31 0.07 0.

06 *** 0.34 1.51 *** 0.07 ** -0.24 3.08 0.10 *** 0.80 3.05 0.94 *** MSCI w ex US -0.13 0.36 0.07 0.05 1.36 *** 0.19 *** 1.41 *** 0.00 -0.05 0.02 0.61 0.07 *** 0.61 2.18 0.48 *** 0.92 *** -0.36 *** 0.01 0.06 0.24 0.10 0.02 0.55 6.74 0.19 *** 0.21 S&P 1.36 *** 0.19 0.07 5.84 0.11 *** 0.03 -0.04 -0.06 0.24 ** 0.23 *** 0.00 -0.11 *** 0.50 *** 0.30 0.42 *** 0.30 *** 0.19 -0.79 1.02 0.64 1.08 0.03 -0.56 0.21 *** 0. 6.69 0.44 *** 0.39 ** 0.29 ** -0.18 -0.13 0.35 *** 0.13 -0.81 0.79 0.Table 35: Persistence in Performance Analysis Based on Beta (1/1997-12/2002) Panel A: Multi-factor estimation .59 *** HML -0.21 *** 0.43 3.03 0.06 0.02 0.22 -0.00 8.Global Funds classified on their beta (1/1997-12/2002) Mean D1 D2 D3 D4 D5 D6 D7 D8 D9 D10 D1a D1b D1c D10a D10b D10c Spread 1-10 Spread 1-2 Spread 9-10 Spread 1a-10c 0.75 1.01 0.31 * 0.85 1.60 1.02 -0.32 *** 0.72 *** -0.03 *** -0.69 0.71 ** 0.31 1.06 -0.03 0.28 ** 0.46 *** 0.05 *** -0.39 *** 0.08 *** 0.03 0.88 0.34 2.05 0.04 -0.52 *** -1.21 *** -0.02 0.31 -0.91 ** 0.18 1.14 0.27 0.45 *** 0.76 0.32 ** * *** * *** * * 245 .09 0.30 0.90 7.01 -0.29 0.08 0.17 -0.02 -0.10 0.10 0.82 Alpha 0.10 * SMB 0.12 Std dev.49 *** 0.12 0.94 *** 0.09 -0.06 *** 0.05 0.04 0.13 2.04 0.06 *** 0.04 0.07 0.99 0.35 6.

03 0.13 -0.09 0.35 *** ** *** *** *** ** * ** ** *** *** ** *** *** *** ** WGBI -0.00 0.08 0.02 -0.80 0. ** Significant at the 5% level and * Significant at the 10% level. Tstatistics are heteroskedasticity consistent.25 -0.07 0.78 0.04 -0. SWGBIt = return of the Salomon World Government Bond Index.03 0. PR1YRt = the factor-mimicking portfolio for the momentum effect. every fund is classified into a decile on the basis of its beta (relative to the S&P 500 equity index) over the past three years.02 0.06 -0.05 0. GSCIt = return of the Goldman Sachs Commodity Index.28 R² 0.87 0.19 -0.11 0.02 0.36 -0.05 0.13 -0.04 0.18 -0.13 -0.32 EMBI 0. HYt = return of the Lehman High Yield Credit Bond Index.09 0.01 0.01 0.08 0.09 0.07 -0.07 0.07 0.21 0.12 -0.03 0.06 0.28 -0.92 0.91 0.08 -0.11 0.72 0. DOLLAR = return of the Federal Reserve Bank Trade Weighted Dollar Index.78 0.04 0.11 0.04 * DOLLAR -0.13 0.01 -0.30 0.09 0.04 0.12 -0.Global Funds classified on their beta (1/1997-12/2002) PR1YR D1 D2 D3 D4 D5 D6 D7 D8 D9 D10 D1a D1b D1c D10a D10b D10c Spread 1-10 Spread 1-2 Spread 9-10 Spread 1a-10c 0.02 0.03 -0.02 0.92 0.02 0.94 0.05 0.18 -0.01 0. MSWXUSt = return of the MSCI World Index excluding US. HMLt = the factormimicking portfolio for book-to-market equity (‘high minus low’).03 0.04 0. EMBIt = return of the JP Morgan Emerging Market Bond Index.12 -0.09 0.10 0.74 0. SMBt = the factor-mimicking portfolio for size (‘small minus big’).05 0.84 0.04 0.26 -0.01 -0.02 -0.08 -0.09 0.05 -0.11 0.12 -0.08 0.02 0.01 0.02 0.04 0.07 0.25 0.03 0.14 -0.01 -0.07 0.05 -0.05 -0.11 0.08 0.05 -0.03 0.11 * ** *** * ** *** *** ** *** *** *** ** *** *** ** HY 0.04 0.18 -0.87 This Table presents the results of the persistence in performance estimation of the combined Model for the January 1994-December 2002 period.11 0.00 0.08 -0.28 -0.29 -0.08 0.05 0.04 -0. *** Significant at the 1% level.05 0.02 0. S&P stands for the market factor.06 0.02 -0.11 0.06 0.08 * * *** *** *** *** * GSCI 0.04 0.36 0.68 0.59 0.07 -0.82 0.00 0.07 0.01 0.10 0.75 0.24 0.03 -0.01 0.02 0.02 -0.92 0.14 0. Each year.14 -0.12 -0. 246 .95 0.Panel A (continued): Multi-factor estimation .11 0.

82 2.66 -0.96 2.76 2.13 *** D5 0.06 * 0.41 *** -0.31 0.24 1.36 0.26 * 0.02 0.39 ** 0.26 -1.03 0.27 ** 0.65 -0.65 0.28 *** 0.32 0.15 *** D6 0.04 * D3 0.Table 36: Persistence in Performance Analysis Based on the Skewness and Kurtosis (1/1997-12/2002) Panel A: Multi-factor estimation .1 0.71 0.03 0.42 3.03 0.02 0.88 0.83 2.34 0.14 *** D4 0.75 2.19 0.32 *** 0.92 2.51 0.05 ** D2 0.26 1.21 *** 247 .24 0.01 0.76 2.31 *** 0.02 0.25 0.85 -0.29 2.09 0.79 1.22 *** D8 0.19 *** D10 0.39 * 0.38 ** 0.25 *** 0.59 0.32 *** 0.04 0.35 1.83 2.32 5.38 *** 0.03 0.94 2. Skewness Kurtosis Alpha S&P MSCI w ex US SMB D1 0.89 3.98 -2.47 0.28 0.54 2.03 2.Global Funds classified on their skewness (1/1997-12/2002) Mean Std dev.5 *** 0.32 12.16 *** 0.33 ** 0.18 *** D7 0.37 *** -0.18 *** D9 0.28 0.9 0.37 *** 0.

06 ** -0.17 ** 0.06 * -0.1 0.03 0.17 0.09 *** -0.85 D7 0.11 0.11 * 0.08 *** 0.08 *** 0 0.16 ** 0.01 -0.07 *** 0.05 0.02 0.04 0.08 *** 0.06 *** 0.1 *** 0.86 D6 0 0.09 0.03 -0.01 0.22 ** 0.09 *** -0.03 0.03 -0.23 *** 0.04 0.11 * 0.85 D4 0.16 *** 0.09 0.03 0.07 *** -0.09 *** -0.86 D5 0 0.06 0 -0.08 *** 0.Panel A (continued): Multi-factor estimation .12 *** -0.87 248 .05 ** -0.02 -0.07 0.05 *** 0.04 0.68 D2 0.04 -0.02 0.05 0.09 0.2 * 0.12 0.05 0.07 *** 0.83 D9 0.01 -0.01 0.85 D10 0.02 0.01 -0.05 0.08 *** 0.16 ** 0.15 *** 0.83 D3 0.36 ** 0.06 *** 0.09 0.2 0.01 -0.Global Funds classified on their skewness (1/1997-12/2002) HML PR1YR WGBI EMBI HY GSCI DOLLAR R² D1 0.08 0.01 -0.87 D8 -0.

04 0.24 0.18 0.73 0.15 *** D3 0.11 *** D9 0.88 2.08 0.85 2.06 0.05 0.Global Funds classified on their kurtosis (1/1997-12/2002) Mean Std dev.68 0.43 ** 0.36 *** 0.39 *** -0.18 *** D5 0.75 2.27 *** 0.14 -0.25 1.96 2.47 0.04 0.84 0.21 0.82 4.19 *** 0.37 ** 0.77 2.39 2.9 0.8 -0.33 *** 0. Skewness Kurtosis Alpha S&P MSCI w ex SMB D1 0.89 4.37 ** 0.04 0.21 0.75 1.44 *** 0.15 *** D8 0.22 ** D6 0.72 3.03 0.37 *** 0.3 * 0.82 2.36 *** 0.16 *** D7 0.3 2.62 -0.27 * 0.55 0.01 0.2 *** D2 0.86 0.1 *** 249 .01 0.37 *** 0.86 3.42 0.2 1.Panel B: Multi-factor estimation .93 -0.06 0.78 0.16 1.26 *** 0.26 -0.41 0.14 *** D4 0.87 2.11 *** D10 0.36 0.1 0.05 0.03 0.92 2.32 *** 0.77 2.34 *** -0.76 0.

88 D6 0.08 0.09 *** 0.03 -0.18 0.14 0.09 *** 0.08 0.09 *** 0.07 ** -0.06 0.03 -0. every fund is classified into a decile on the basis of its skewness (Panel A) and kurtosis (Panel B) over the previous three years.09 *** 0.06 *** 0. Each year.09 0.09 *** -0.04 0.1 0.02 -0.01 -0.Panel B (continued): Multi-factor estimation . SWGBIt = return of the Salomon World Government Bond Index. T-statistics are heteroskedasticity consistent.13 * 0.04 0. HYt = return of the Lehman High Yield Credit Bond Index. We report the OLS estimators for equally weighted portfolios per investment strategy.88 D5 0.86 D9 0.07 *** 0.1 * 0.04 0.07 0. EMBIt = return of the JP Morgan Emerging Market Bond Index.07 0.02 0.86 D2 0 0.07 *** -0. DOLLAR = return of the Federal Reserve Bank Trade Weighted Dollar Index.84 D4 0.24 * 0. PR1YRt = the factor-mimicking portfolio for the momentum effect.2 *** 0.01 0. S&P stands for the market factor.06 *** 0.13 0.02 -0.06 *** -0.13 *** 0.19 0. ** Significant at the 5% level and * Significant at the 10% level.07 0.8 This Table presents the results of the persistence in performance estimation of the combined Model for the January 1994-December 2002 period.11 0. sub-strategy and for all funds.07 *** -0.05 ** 0.18 *** 0. *** Significant at the 1% level.06 *** -0.86 D8 0.07 *** -0.02 0.11 ** 0.02 -0.86 D7 0 0. SMBt = the factor-mimicking portfolio for size (‘small minus big’).84 D3 0.02 -0.03 0.06 *** 0.07 *** 0.07 ** -0. 250 .07 *** -0.04 * -0.02 0.09 0.Global Funds classified on their kurtosis (1/1997-12/2002) HML PR1YR WGBI EMBI HY GSCI DOLLAR R² D1 0 0.12 * 0.06 0.01 -0.02 -0.83 D10 0.03 * 0.06 *** 0.18 0. GSCIt = return of the Goldman Sachs Commodity Index.07 0.22 * 0. MSWXUSt = return of the MSCI World Index excluding US.08 *** 0. HMLt = the factor-mimicking portfolio for book-to-market equity (‘high minus low’).03 -0.

and the (wealth-related) proportion of risky investment to initial wealth The risk measure reported in (3) is a measure that integrates not only the variance as a measure of risk but also skewness and kurtosis. please see Part 3 of the Thesis. Capocci et al. 61 For more details. I / W0 .5. c. (2007)61. 251 . we use the new risk measure developed by Capocci et al. this risk measure accounts for volatility. (2006) estimate C as 10 for a dynamic investor that does not put much weight on 62 the extreme risks. An Adapted Sharpe Ratio As a final and combined risk measure. We focus on average investors but results are in line for dynamic and protective investors. The level of C depends on the appetite for risk of the investor62. As stated previously investors do not like volatility. This measure R x is defined as: 1 1 1 R x = V x − CS x + C 2 K x 2 6 24 (11) C≡c Where I W0 is the product of the (intrinsic) risk perception coefficient. asymmetry and the fat tails. negative skewness and positive kurtosis. 35 for an average investor and 60 for a very protective investor who mostly fears the tail event of a distribution. In other words. Capocci et al. (2007) consider three representative levels.

Such an adapted Sharpe score would be defined as mean return divided by the new risk measure: R x as defined in (11)63. depending on the decile considered negative (decile D1) or positive skewness (decile D2 and D3) and positive kurtosis. There are still around one third of the funds in our database that consistently outperform over time but it is not easy to extract them. standard deviation. Results are reported in Table 37 (the adapted Sharpe Score are multiplied by 100 to be comparable). The factor exposure of these deciles indicate that the funds included in deciles D1 to D3 distinguish themselves from the other funds (at least for decile D2 and D3) by having been relatively lowly exposed to the equity market. skewness. The first columns indicates that these deciles have had a limited volatility. The spreads help us to understand why decile D1 is so different from decile D2 and D3. the R² reported for them are lower than for the other deciles. The five first columns report the mean return. These results indicate that when higher moments are considered results are mitigated. the alpha of the deciles that include the funds with the lower adapted Sharpe over the ranking period tend to significantly and consistently outperform over time. 252 . that they have not been exposed to emerging market bond and that more globally. There is no clear trend in the adapted Sharpe reported. with no momentum or contrarian bets. Fund with the highest Sharpe score have been classified in the bottom deciles while those with the lowest one in the top deciles.This risk measure can also be used as an adapted Sharpe ratio that takes into account the other aspects of risk that are not taken into account by the standard deviation. More interestingly. 63 An adapted version of the Sharpe ratio would be defined as the excess mean return divided by the adapted risk measure. kurtosis and adapted Sharpe for each decile and each spread.

On the other side. Volatility and returns should be considered together to easily identify funds that should outperform classical markets in the future.Spread 1-2 and spread 9-10 are significantly negative. Other Measures – Conclusion Table 38 summarises the persistence analysis. This result is directly confirmed by the Sharpe score. A negative spread 1-2 indicates that there is a significant difference between the least attractive funds in terms of adapted Sharpe ratio. However. These results provide a simple way to create pure alpha using hedge funds. the pure volatility analysis. Moreover. Panel A interestingly indicates that classifying hedge funds on the basis of their past returns and investing in the previous year’s best performing funds does not allow the investor to significantly and consistently outperform classical bond and equity indices. 253 . 6. One can significantly and consistently outperform stock and bond markets by buying hedge funds that offer the most appealing risk-return characteristics over a three-year period and by rebalancing the portfolio every year. Panel A reports the alpha column and Panel B the U. Such a result is astonishing because the extremes (decile D1 and D10) outperform the neighbour at the two end of Table 37. a deeper analysis indicates that lower volatility funds (classified in the middle deciles) do significantly and consistently outperform the indices considered. the alpha and the beta classifications. equity exposure. the skewness and kurtosis analysis showed that when funds cannot be differentiated by their volatility. pure alpha creators cannot be identified. The worst ones (decile D1) are much less attractive than the next to worst one (decile D2).S. the next to top one (decile D9) also significantly underperform the top one (decile D10).

00 1.59 0.20 0.50 0.50 0.90 -0.10 -0.00 254 .40 2.40 0.02 0.32 0.80 1.10 3.50 1.Table 37: Persistence in Performance Analysis Based an adapted Sharpe score (1/1997-12/2002) Mean D1 D2 D3 D4 D5 D6 D7 D8 D9 D10 D1a D1b D1c D10a D10b D10c Spread 1-10 Spread 1-2 Spread 9-10 Spread 1a-10c 1.40 0.80 0.20 0.00 -3.60 2.01 0.30 2.10 -0.80 0.80 2.60 0.40 1.10 0.30 0.70 2. 2.73 0.90 1.00 3.10 2.70 Adapted Sharpe 0.00 0.70 0.50 0.50 1.50 2.40 2.10 2.08 1.85 0.10 -0.10 1.30 2.10 Kurtosis 3.80 0.40 -0.10 3.04 0.05 1.60 1.90 -0.20 2.81 0.03 Std dev.50 -1.00 0.70 1.20 3.05 -0.60 -0.20 -0.00 4.40 0.14 1.85 0.65 0.05 0.10 1.10 2.97 1.30 3. -0.70 2.50 1.70 0.40 1.20 1.01 0.10 2.50 0.00 2.50 3.00 1.71 1.10 Skew.70 -0.60 9.80 3.04 1.80 1.40 1.90 1.50 0.12 0.10 1.10 1.10 0.

63 0.07 0.29 This Table presents the results of the persistence in performance estimation of the combined Model for the January 1994-December 2002 period. SWGBIt = return of the Salomon World Government Bond Index.02 0.23 -0. *** Significant at the 1% level. S&P stands for the market factor.08 -0.84 0.12 -0.04 -0.09 -0.02 0.08 0.08 0.02 -0.04 0. SMBt = the factor-mimicking portfolio for size (‘small minus big’). We report the OLS estimators for equally weighted portfolios per investment strategy.10 0.00 -0.08 0.06 0.03 0.05 * ** R² 0.13 0.01 0.91 0.06 0.08 -0.76 0.08 0.01 0.80 0. 255 .87 0. ** Significant at the 5% level and * Significant at the 10% level.04 0.09 0.01 0.01 -0.06 0. sub-strategy and for all funds.02 0.09 0.10 0.26 -0.18 -0.07 0.13 -0.03 -0. GSCIt = return of the Goldman Sachs Commodity Index.08 0.15 -0.76 0.04 0.14 -0.34 0.17 -0.04 -0.09 -0.02 0.22 -0.01 -0.08 0.00 0.82 0. HYt = return of the Lehman High Yield Credit Bond Index.09 0.13 -0.06 0.09 -0.09 *** *** *** * GSCI 0.10 0. HMLt = the factor-mimicking portfolio for book-to-market equity (‘high minus low’).10 0.85 0.09 0.04 0.05 0.17 -0.52 0.61 0.03 0.06 -0.02 0. DOLLAR = return of the Federal Reserve Bank Trade Weighted Dollar Index.78 0.01 -0.02 0.04 0.14 0.13 -0.16 -0.02 0.01 -0.12 -0.88 0.86 0. every fund is classified into a decile on the basis of its adapted Sharpe score (Panel A) over the previous three years.01 *** *** *** *** *** *** *** ** *** ** * *** ** *** HY 0.10 -0.06 -0.35 -0.07 -0.Panel B (continued): Multi-factor estimation .08 0. MSWXUSt = return of the MSCI World Index excluding US.65 0.76 0.03 DOLLAR -0.02 0.Global Funds classified on their beta (1/199712/2002) WGBI D1 D2 D3 D4 D5 D6 D7 D8 D9 D10 D1a D1b D1c D10a D10b D10c Spread 1-10 Spread 1-2 Spread 9-10 Spread 1a-10c -0.03 0.02 0.08 -0.01 0.07 -0.82 0.03 0. Each year.03 0.07 -0.06 0.06 -0.03 0.11 -0.18 -0.09 -0. PR1YRt = the factor-mimicking portfolio for the momentum effect. EMBIt = return of the JP Morgan Emerging Market Bond Index.08 0.01 0.01 0.09 0. Tstatistics are heteroskedasticity consistent.07 0.02 *** *** EMBI 0.09 0.

25 0.37 *** 0.25 ** 0.36 *** 0.05 Kendall Tau -0.51 *** D7 0.27 0.32 *** 0.41 *** 0.37 *** 0.36 *** D8 0.28 D4 0.6 * 0.33 * 0.47 0.Table 38: Summary of Alpha Creation and Equity Exposure Panel A: Alpha creation Returns Sharpe Score Std dev.34 *** 0.5 ** D5 0.43 *** 0.33 *** 0.43 -0.56 -0.22 0.44 ** D6 0.25 0.17 D10 0.57 * 0.19 -0.35 *** 0.16 0.41 * 0.19 D2 0.41 *** 0.49 *** 0.29 * 0.07 -0. Alpha D1 0.26 *** D9 0.23 * 0.35 256 .15 0.27 0.43 D3 0.4 0.29 * 0.38 *** 0.

46 *** 0.39 ** 0.37 ** 0.09 0.31 -0.19 0.39 * 0.3 0.37 *** -0.7 ** D3 0.31 * 0.27 ** 0.13 D9 0.36 *** 0.43 ** 0.16 D5 0.24 0.28 0.Panel A (continued): Alpha creation Beta Skewness Kurtosis Adapted Sharpe D1 0.39 ** 0.45 *** D2 0.01 D6 0.44 *** 0.37 ** 0.03 D10 0.38 ** 0.21 0.47 257 .27 * 0.18 0.33 ** 0.25 Kendall Tau 0.18 0.24 0.08 D7 0.36 *** 0.31 -0.53 *** D4 0.26 * 0.3 * 0.25 0.5 *** 0.39 *** 0.54 -0.09 D8 0.

56 *** 0.48 *** 0.48 *** 0.3 *** 0. Alpha D1 0.01 0.41 *** 0.36 *** D4 0.45 *** 0.29 *** D5 0.46 *** D3 0.39 0.26 *** 0.59 *** 0.25 *** D8 0.13 *** 0.77 *** 0.51 *** 0.18 *** 0.08 0.33 *** 0.44 *** 0.33 *** 0.46 *** 0.Panel B: US equity exposure Returns Sharpe Score Std dev.28 *** 0.3 *** D10 0.34 *** Kendall Tau 0.66 *** D2 0.39 ** 0.17 * 0.39 258 .23 *** 0.12 *** 0.78 -0.36 *** 0.23 *** D6 0.57 *** 0.24 *** D9 0.82 -0.03 0.21 *** D7 0.2 *** 0.07 *** 0.08 *** 0.16 *** 0.

43 *** Kendall Tau -1 -0.25 *** 0.06 *** 0.Panel B (continued): US equity exposure Beta Skewness Kurtosis Adapted Sharpe D1 1.39 *** D8 0.21 *** 0.08 *** 0. 259 .02 0.27 *** 0. Each year.36 *** 0.11 *** D4 0.18 0.32 *** 0.37 *** 0.33 *** D2 0. every fund is classified into a decile on the basis of its data over the past three years.26 *** 0.46 *** D10 -0.31 *** 0.16 *** 0.28 *** 0.34 *** 0. sub-strategy and for all funds.36 *** 0.37 *** 0.52 0.49 *** 0.11 *** 0.27 *** D6 0.41 *** 0. ** Significant at the 5% level and * Significant at the 10% level.32 *** 0. *** Significant at the 1% level. T-statistics are heteroskedasticity consistent.4 *** D9 0.32 *** 0. Panel A reports the alpha.37 *** 0.19 *** 0. We report the OLS estimators for equally weighted portfolios per investment strategy. Panel B reports the equity exposure.31 *** D7 0.32 *** 0.39 *** 0.38 *** 0.17 *** D5 0.11 -0. These results are based on the January 1994-December 2002 period.71 ** 0.19 *** D3 0.66 This Table summarises the persistence in performance analysis.33 *** 0.42 *** 0.

Panel A indicates that the only significant correlations exists between the alphas produced by the standard deviation and alpha series (the coefficient is at 0. skewness and kurtosis have not been correlated confirming that these ways of ranking are completely independent from any other. Panel A contains the Kendall coefficients for the alphas Panel B contains the same factors obtained for the equity market exposure. More globally. it measures the strength of association of the cross tabulations.In order to quantitatively determine if there is a correlation between the alphas produced and/or the market exposures. Kendall Tau is used to measure the degree of correspondence between two rankings and assessing the significance of this correspondence. 260 . These results indicate that the corresponding series evolves linearly even if a closer look at the significance level indicates that the corresponding significant alphas do not correspond to the same deciles. We use this tool to determine if there is redundancy in the alpha and in the equity exposure produced by different ways of ranking funds in the persistence analysis. It seemed to us at first sight that the ranking based on standard deviation and the beta were simply opposite but the Kendall Tau coefficients do not confirm our believes. In other words. the beta.52 significantly positive) and between returns and the adapted Sharpe (0. we report the Kendall Tau correlation coefficients in Table 39.51 significantly positive). This coefficient depends upon the number of inversions of pairs of objects which would be needed to transform one rank order into the other64. 1948) or Abdi (2007). that is to determine if some measures are redundant. 64 For more information please see Kendall (1938. Table 39 reports the results. it is interesting to note that the Sharpe score. Kendall’s coefficient evaluates the degree of similarity between two sets of ranks given to a same set of objects.

36 *** 0.49 * *** 0.64 *** -0. *** Significant at the 1% level.25 -0.02 0.12 0.00 *** -0.39 0.29 *** -0.00 0.43 *** -0.20 0. alpa.09 -0.04 1 *** -0.07 -0.22 -0.18 1 [8] 0.49 * -0.51 ** 0.36 1 *** 1 *** -0.23 -0.09 0. Alpha Beta Skewness Kurtosis [2] [3] [4] [5] [6] [7] -0. ** Significa 261 .78 *** -0.Table 39: Kendall Tau Estimation Panel A: Kendall Tau between the Alphas [1] [1] [2] [3] [4] [5] [6] [7] Returns Sharpe Score Std dev.14 0. standard devition.49 * [6] 0.39 1 [8] Adapted Sharpe Panel B: Kendall Tau between the US equity market betas [1] [1] [2] [3] [4] [5] [6] [7] Returns Sharpe Score Std dev.33 0. skewness. Sharpe score.87 *** -0.64 *** -0.07 -0.60 ** 0.14 1 *** [7] 0. kurtosis and adapted Sharpe).45 * 0.75 *** -1.33 1 -0.02 *** -0.25 0.33 *** -0.07 *** 0. Alpha Beta Skewness Kurtosis [2] [3] -0.87 *** -0.14 1 *** 1 *** 0.14 -0.38 1 0.33 1 -0.22 0.52 ** 1 [5] -0.16 1 *** [8] 0.02 -0. beta.07 1 [8] Adapted Sharpe This Table reports that Kendall Tau between the alphas obtained usingthe various ways o ranking funds into deciles (returns.49 * 0.13 *** -0.27 0.52 0.09 0.33 1 *** -0.09 1 [4] -0.

Sharpe score and adapted Sharpe. up-down-up or down-up-down) in several cases (Sharpe score. This result can be explained by the fact that the beta exposures evolve linearly (up-down. panel B shows a different pattern. Independently of the basis of the ranking most beta exposures are significantly correlated between each other (returns and standard deviation. return and adapted Sharpe. 262 . In this case. In two cases they even evolve perfectly linearly (standard deviation and beta with a Kendall Tau of 1). standard deviation and adapted Sharpe and beta and adapted Sharpe). down-up.As expected. the Sharpe ratio is a good approximation for one of its extended version. alpha and skewness). returns and beta. standard deviation and beta. This result suggest even if the Sharpe ratio does not count for skewness and kurtosis. most hedge fund managers (at least when aggregated in deciles) offer the same skewness and kurtosis characteristics. being the only to differentiate completely from all the others in terms of evolution of the market exposure. Sharpe score and skewness. The alpha and the kurtosis are the only series that are correlated with no other. It is interesting to note that the adapted Sharpe score is highly correlated with its adapted version.

We perform two additional checks. 5.1 Sub-Period Analysis To determine if the results obtained are due to the analysis period. Firstly. we rank the funds based on past statistics each July instead of each January. to the time period covered or to the market conditions. we divide this period in two ways. we will define bull market as positive months for the S&P 500 and bear market as negative months for the S&P 500. Firstly. 66 We do not report the results for the sake of brevity. we perform several further analyses that should confirm the stability of our findings.V Further Analysis Our results may be due to specific aspects of our methodology. but all the tables are available upon request. Secondly. To handle these issues. we perform the same analysis but and divide the analysis period into subperiods in order to determine if the period covered or the market conditions can explain at least partly our results.66 65 We perform the same analysis using September 2000 (the top of the S&P 500) instead of April 1st as the turning point and the results obtained are very similar. we will take March 2000 as the turning point between bull and bear market conditions. 263 .65 Secondly.

After 2000. The results are clear: before 2000 almost half of the hedge funds significantly outperform and create pure alpha. (2005) who found that hedge funds considered as an industry have not been able to outperform traditional markets during the market turmoil between 2000 and 2002 even if 20% to 30% of the funds do. only 20% to 30% of the funds – with the same characteristics – do outperform. There are 39 months before the end of March 2000 (starting from January 1997) and 31 months after this date (starting from April 2000). These results confirm the results obtained by Capocci et al. 264 . These funds have had a limited exposure to the equity markets and a low volatility. March 2000 as a Turning Point March 2000 coincides with the top of the NASDAQ composite index. on the other side. We perform the same analysis as previously twice using on one side data before this date.1. data after this date.

These results confirm that hedge funds face difficulties beating classical indices during bull markets. 265 . An investor investing in the previous year’s best or worst performing funds. the one that looks for positive skewness or investors who avoid fat tails would not have been able to create significant alpha in bull market conditions. These results confirm that integrating volatility in the way of classifying funds helps investors in creating consistent and significant value over time. in funds offering the best alpha. Interestingly. in funds offering the best risk-return trade-off. Up and Down Months of the S&P 500 Another way of separating bull and bear markets is to take the positive months of the S&P 500 as bull market conditions and negative months bear market conditions.2. 39 months have been positive for the S&P500 while 33 months have headed south. During bear market conditions there is consistent alpha creation for around one third of the deciles. that there are still possibilities to create pure alpha. Top performers even underperform (but not significantly) during market rally. the standard deviation and the beta results suggest however. These are another time the low volatile decile confirming our previous results. Over the period covered. Our results indicate that no single decile significantly and consistently outperforms the classical markets during bull market conditions. Low volatility funds unexposed to the equity markets have been able to create significant and consistent alpha over bull market periods.

6 to 0. there is a long exposure to emerging market bonds and to the high yield market. Alpha creators are still mildly volatile. The exposures to the majority of factors do remain the same using these three best performing measures. Inversely. Moreover. but only marginally so using the January classification. 1994). 1992 and Kramer. Bhardwaj and Brooks. These results should be due to the slight difference in the analysis period since January to June 1994 is omitted when classifying funds in June. with a limited exposure to the market. 266 . this section confirms that classifying funds in July rather than in January does not change our results significantly. The r-squared are high. ranging from 0.67 The first important point to stress is that the same low volatility middle deciles create alpha. decile D10 alpha is significantly positive using the June classification. a small cap bias and value oriented.9. The best measure remains the Sharpe score. Altogether. There are some marginal differences in exposures (from slightly exposed to significantly exposed and inversely) but nothing major.5. 67 The corresponding tables are available on request. The differences with previous results are that deciles D8 and D9 offer significantly positive alpha using the January classification date. we perform the same analysis by classifying the funds in July instead of January. see for example. whereas these alphas are only marginally significant using the June date. Factor exposure analysis indicates that classifying funds in January or in June leads to very close results. the standard deviation and the beta.2 Date Impact To remove any effect that could arise from the January Effect (stock prices tend to go up between December and January.

In this best case. they get an estimation of performance at the end of the by the middle of January. In the best case. In the worst case. This is a very attractive conclusion to investors but. It ranges from 10% to 30% meaning that redemptions will be blocked in case of large sell-off. Hedge funds are not liquid investments even if they tend to become more liquid (at least for equity listed related strategies). most hedge funds have lockup period in place meaning that you cannot sell the position before the lockup period ends. We call this issue “frictions the real world”. if investors decide to close a position based on ranking early in the year. Finally. it will take them at least one month before being able to sell and maybe 2 or 3 weeks to get the cash. In this particular case. and there is naturally a but. Why can investors not easily profit from our results? The first element to this conclusion is liquidity. Another concern regarding the liquidity of hedge funds is that. This period lengths usually minimum one year and goes up to three years. A gate is the maximum percentage of assets that can be sold by investors at any single liquidity date. hedge funds offer monthly liquidity with a 30-day notice period meaning that it takes minimum one month to get out of a fund. many funds have gate in place. sell the position before the month end for the end of February and get the case in March to be able to invest for April 1st. 267 . meaning that they cannot be invested before the end of March. it is not easy for investors to profit from this opportunity even in applying the simple strategy of buying the funds that offered the greatest risk-return trade-off or the funds with low betas. funds offer yearly liquidity with six months of notice.VI Frictions in the Real World Hedge funds can offer persistence over time.

aso. even if most funds report real returns net of all fees. In fact. Finally. there is always a data problem with hedge fund databases. All these issues confirm the need to take all our results with care. These are always based on underlying hypothesis regarding liquidity. it stops reporting to databases but its past performance remains in the database with false figures. Hedge fund managers report their performance on a voluntary basis meaning that fraud is even facilitated compared to mutual fund databases. This means that at least some funds that are considered in studies like ours cannot be considered for investments. This may lead to a wrong choice. most funds that are closed to new investors continue to report to databases in order to promote the name of the company (particularly in case of launch of a new fund).Funds that report to the database have not to be open to investment or to have minimum assets. some of them use pro-forma estimation or manage account performances that cannot be replicated easily. the funds that are closed to investments reach their target size because of their good past performance. Moreover. Usually. as it has been mentioned in many papers in the past. Moreover. When a fund frauds. the quality of data. 268 .

Previous studies have all been focused on past performance as the unique tool to analyse alpha creation. the beta exposure appear to be the best and most stable way of classifying hedge funds in order to detect persistency in the returns. Another idea would be to develop and use a measure combining the returns. The important element used to detect these funds is the methodology by which they are classified. These results hold not only for a full market cycle. This analysis is of particular interest because it clearly proves that some funds consistently and significantly outperform classical markets. beta. funds with a limited volatility and/or funds with a limited exposure to the equity market consistently and significantly outperform equity and bond markets. We go one step further in decomposing hedge funds returns and analysing the persistence in hedge fund returns. We find a consistent. The next steps would be to focus on specific strategies and determine what way of classification is most suitable for each hedge fund strategy. skewness and kurtosis to classify funds. Such a measure for classifying hedge funds may be the final step in the decomposition of hedge fund returns. skewness and kurtosis. systematic way of creating pure alpha using a simple classification methodology based on basic statistics: risk-return trade-off measure (the Sharpe score).VII Conclusion This study brings new insights in the research analyzing the persistence in hedge fund returns. Our study uses other measures such as the Sharpe score. Funds offering the highest Sharpe score. volatility. volatility and to a lesser extent. alpha. 269 . but also when separating bull and bear market conditions. standard deviation.

.

Part two: Analysis of Hedge Fund’s Market Exposure .

forthcoming Global Finance Journal 272 .The Neutrality of Market Neutral Funds Daniel P. 2006.J. The Neutrality of Market Neutral Funds. Daniel. CAPOCCI HEC-ULG Management School – University of Liège (Belgium) This paper has been awarded Best Doctoral Paper at the Global Finance Conference in Dublin (2005) Capocci.

this study formally analyses the market neutrality of market neutral funds which are particular in the hedge fund universe since the only objective of these funds is to provide positive returns completely independent of the market conditions. 273 . We start by analysing this neutrality using various market neutral indices before focusing on individual fund returns. We perform this analysis over the global January 1993. an analysis based on ex-post beta helps us explaining and confirming our previous results. Finally.The Neutrality of Market Neutral Funds Abstract Using an original database of 634 market neutral hedge funds.December 2002 period as well as on bull and bear market periods.

Goetzmann and Ibbotson (1999) are in this category. Liang. McEnally and Ravenscraft (1999) and Liang (1999) find that hedge funds constantly obtain better performance than mutual funds. 2001.The Neutrality of market neutral funds Introduction Hedge funds considered as a whole have been studied since 1997. In this context. 1999. They prove that offshore hedge funds have positive risk adjusted returns. Barès. 1999. Gibson and Gyger. the hedge fund industry has an attrition rate higher than mutual funds. Finally. 2002. Hedge fund studies can be classified in four global categories. Amin and Kat. Liang. but they input this result to style effect and conclude that there is no proof of particular capacity of some fund 274 . McEnally and Ravenscraft (1999) in their Journal of Finance article. Results of such studies are mitigated. although lower and more volatile than the reference market indices considered. Since then. Brown. performance analysis includes the study of the persistence of hedge fund returns. Liang. 2001). 1999. The second field of hedge fund performance analysis compares the performance of hedge funds with the one of mutual funds. This includes mainly studies that compare the performance of hedge funds with equity and other indices (see for example Ackermann. McEnally and Ravenscraft. Ackermann. Capocci and Hübner. we report studies that are focused on hedge fund performance. 2004. Goetzmann and Ibbotson (1999) and Liang (2000. 2001. Agarwal and Naik. as suggested by Brown. 2004). The precursors of the academic world were Fung and Hsieh (1997) in their Review of Financial Studies paper and Ackermann. Brown. Persistence is particularly important in the case of hedge funds because. the literature on the subject has expanded strongly and more and more researchers are focusing on these investment products. Goetzmann and Ibbotson. In the first one. 2003.

It is important to stress that few authors have attempted to estimate the behaviour of hedge funds in bear markets. On the other hand. Brown. Brown and Goetzmann. The periods under review do not favour this exercise. Corhay and Hübner (2004) use the decile analysis developed in Carhart (1997) in order to determine if persistence is present in hedge fund returns. Event Driven and Macro) provide protection to investors when stock markets head south. 1998. Fung and Hsieh (1997) applied Sharpe's style analysis (see Sharpe. their superior performance is mostly due to the good market timing of their managers. 1997. hedge funds did not provide investor protection after the market downturn of March 2000. 2001. Brealy and Kaplanis. 2001. Agarwal and Naik (2000) analyse the presence of persistence in hedge fund returns using a one-year moving average period. Capocci and Hübner (2004) and Capocci. In this context. rather. Edwards and Caglayan (2001) found that only three hedge fund strategies (Market Neutral. as periods of downward trends on the stock market were rare and discontinuous before March 2000.managers. Liang 2003). Liang 2001. The first concluded that some low-risk managers have been able to consistently create alpha over the period studied (1/946/00). Goetzmann and Park. the second global category includes authors that both try to analyse and describe hedge funds investment style and to explain these features with style models (see for example Fung and Hsieh. For the period 1990-1998. excess return creation was present in most of the cases and there was a clear proof of persistency in hedge fund returns. 1992) to a large sample of hedge funds and CTA. More recently. They prove that there is proof of persistence in hedge fund's performance. They assumed that fund 275 . The second focused on the period starting 1/94 but ending 12/02 and concluded that over this global period that included a bear market. Ennis and Sebastian (2003) contend that in general. particularly for bad performing funds that continue to underperform.

returns are linearly related to the returns on a number of factors and measured the factor by 8 mimicking portfolios. this model is a state-space model and can be estimated by using standard Kalman filter techniques68. To handle with this issuer. No window size and ad hoc chosen length need to be used. This methodology has one major drawback: the choice of a number of observations used for the estimation. No restrictions are imposed on the betas. A particular aspect that has been taken into account more recenthly is the style drift in hedge fund returns. found that the regressions have little explanatory power and suggested that the low R²adj is due to the fund's trading strategy. Brown. 276 . This effect comes from the fact that hedge fund managers are opportunity driven and therefore change style over time. As stressed by Posthuma and Van der Sluis. Unlike the ad hoc rolling regression approach. They use rolling regression to take the style drift into account. This simple technique enables to observe style variation of a manager over time. the time variation in the exposures is explicitly modelled. Goetzmann and Park (1998) analyse hedge fund returns during the 1997-98 Asian crisis. Remolona and Tsatsaronis (2005) apply the same methodology. The methodology consists in realiazing a set of linear regressions and moving the estimation perdio of each of them by one observation. This technique is adaptive in the sense that changes in th style exposures are piced up automatically from the data. McGuire. 68 See Pollock (1999) for a detailed presentation of the Kalman filtering and space-state models. Posthuma and Van der Sluis (2005) propose to use a dynamic style model in which beta can vary over time developed by Swinkels et Van der Sluis (2001). The filter is an adaptive system based on the measurement and updateing equations. The Kalman filter procedure chooses the optimal weigthing scheme directly from the data.

Amenc and Martellini (2002) proved on the basis of ex-post estimations that the inclusion of hedge funds in a portfolio can lead to a significant decrease in the volatility of a portfolio without leading to a significant change in the returns. This category includes the ”other studies”. Amenc. Finally other authors have analysed various other aspect of the hedge fund industry. There are many different hedge funds’ indices providers like EACM. Agarwal and Naik (1999) and Capocci and Hübner (2004). Amin and Kat (2001) found that stand-alone investment hedge funds do not offer a superior risk-return profile. Martellini and Vaissié (2002) and Berényi (2002) have studied the risks involved in hedge fund investing. They obtain the best results when 10-20% of the portfolio value is invested in hedge funds. 277 . Various authors study hedge fund indices. Schneeweis and Spurgin and Amenc. Curtis and Martellini (2002). Taking all these results into account. Fung and Hsieh (1997) and Schneeweis and Spurgin (1998) proved that the insertion of hedge funds in a portfolio could significantly improve its risk-return profile thanks to their weak correlation with other financial securities. Liang (2000) analysed the presence of survivorship bias in hedge fund data and Fung and Hsieh (2000) included other biases in their analysis. Martellini and Vaissié (2002) proved that hedge fund returns are not only exposed to the market risk. but that a great majority of funds classified as inefficient on a stand-alone basis are able to produce an efficient payoff profile when mixed with the S&P 500. Amenc.A third part of the literature finally focuses on the correlation of hedge funds with other investment products and analyses the power of diversification of hedge funds. This means that a stronger risk control does not correspond with a decrease in return. hedge funds seem to be a good investment tool. Ackermann and Ravenscraft (1998) emphasised that the stronger legal limitations for mutual funds than for hedge funds hinder their performance. Schneeweis and Spurgin (1999). This low correlation is also emphasised by Liang (1999). default risk or liquidity risk have to be considered. but that other risks like volatility risk.

Hennessee Group. 2004b).HFR. Gregoriou and Rouah. 1997. LJH Global Investment. Mar. CSFB/Tremont. 2003) or study them on a stand-alone basis (Fung and Hsieh. In this study. 69 See Amenc and Martellini (2002) for a complete description of these hedge fund indices providers.net. 2000a. Hedgefund. Zurich Capital. whereas others study them by separating them from hedge funds (Liang.69 Brooks and Kat (2001) Amenc and Martellini (2002) studied this aspect in detail. 2003. 278 . we only study market neutral funds and do not include CTAs. Altvest and Magnum. Some authors consider them as part of the hedge fund world (Fung and Hsieh. CTAs are a particular category in the hedge fund world. 2000). Van Hedge. Capocci. Schneeweis and Spurgin.

Market neutral funds should also be studied on a stand-alone basis to understand their particularities. it can seem quite reductive to focus on a particular strategy since the bulk of the literature considers hedge funds as a whole.3% of the global MAR/CISDM individual funds in the database are market neutral funds.I Interest of the study This study aims at analysing the exposure to the equity market for market neutral funds. more and more authors however consider individual strategies to better understand their particularities. Fung and Hsieh (2002a) on fixed income arbitrage funds. Mitchell and Pulvino (2002) only focused on risk arbitrage. 2004). A high number of market neutral funds allow to analyse them in more detail and to obtain global results. Corhay and Hübner. Three independent reasons justify this choice. Capocci (2004a) has analysed the inclusion of convertible arbitrage funds in a classical portfolio. Gatev. Navone’s (2001) look at the diversification benefits from 279 . market neutral funds are particular since their objective is to create alpha while completely hedging the exposure to the market. market neutral funds represent a high percentage of the hedge fund industry. Capocci. Thirdly. Goetzmann and Rouwenhorst (1999) study relative value funds on a stand-alone basis. Recently. Other hedge fund strategies are generally at least partially exposed to the market and results particular to market neutral funds may not be true for other hedge fund strategies. in their study based on a very large database of a total of 2894 individual funds. At first sight. Corhay and Hübner (2004) concluded that equity market neutral funds present interesting characteristics that need further investigation. Secondly. According to our database described hereafter. Other authors also report high numbers of market neutral funds in their database (see for example Capocci and Hübner. Firstly. 2004 and Capocci. as much as 28. as suggested in several studies. This is exactly the objective of the present study.

In theory. Many managers of such strategies balance their long and shorts in the same sector or industry. By being sector neutral. This category includes.e. Beta neutral strategies target a zero total portfolio beta (i. 280 . Managers take advantage of relative price discrepancies. beta neutral or both.adding market neutral funds to a portfolio of mutual funds could significantly help the risk return trade-off but he did not look at their neutrality. while hedging all intrinsic value and fixed income arbitrage funds that exploit pricing anomalies in the global fixed income. among other things. While dollar neutrality has the virtue of neutrality. Another sub-strategy is long/short market neutral funds also named equity market neutral funds. market risk is greatly reduced but it is very difficult to make a profit on a large diversified portfolio. so stock picking is critical. they avoid the risk of market movements affecting some sectors or industries differently than others. Dollar neutral strategies have zero net investment. These funds simultaneously take long and short positions of the same size within the same market. equal dollar amount in long and short positions. Usually they focus on AAA-rated mortgage bonds. beta neutrality better defines a strategy uncorrelated with the market return. Market neutral funds are defined as funds that take long and short positions in various securities while trying to avoid exposure to the equity market. convertible arbitrage funds that buy undervalued convertibles. beta of the long positions equals the beta on the short side. The final sub-strategy is mortgage-backed funds that invest in mortgage-backed securities but also in futures and options. Sub-strategies are arbitrage market neutral funds. which are managers that apply market neutral arbitrage strategies. i.e. Market neutrality implies dollar neutral. Typically the strategy is based on quantitative models for selecting specific stocks with equal dollar amounts comprising the long and short sides of the portfolio.

We perform a subperiod analysis in section VIII. In section V we present our methodology. 281 .The study is organised as followed. We report the global results in section VI and the results obtained for individual funds in section VII. Section IX concludes the study. In section IV we report the descriptive statistics and analyse the attrition rate present in our database. Section III describes that database.

strategy followed. they record other useful information such as company name. The database gives monthly net-of-fee individual returns and other information on individual funds and groups them in indices. and Capocci. 11 non-directional and 5 directional strategies. plus the Funds of Funds and the Sector categories.II Database Four main hedge fund databases are available for empirical studies. MAR/CISDM defines 9 strategies with a total of 16 substrategies. Hedge Fund Research. Inc. management and incentive fees. Inc/Centre for International Securities Derivatives Markets. The first three ones are the most used in academic studies. The whole database consists of 634 individuals market neutral funds including 398 surviving funds (62. HFR defines 16 different strategies in 2 categories. Amin and Kat (2001). We use hedge fund data from MAR/CISDM. the Managed Account Reports. For a majority of funds.2%) dissolved funds. manager's name. TASS defines 15 strategies. The data provider collects various information on the funds included. The Barclay database is currently under investigation by various authors but it has not yet been used in published studies. We use 120 monthly returns between January 1993 and December 2002. There is no consensus on the definition of the strategy followed but there are similarities. Barclays defines 20 individual strategies. etc. Schneeweis and Spurgin (1998). Corhay and Hübner (2004). assets under management. start and ending date. 282 . as in Fung and Hsieh (1997). Finally.8%) and 236 (37. manager's address. TASS Management and the Barclays database.

54% for market neutral funds except for mortgage backed funds (-8.94%. We report the statistics for the whole market neutral database and we divide this database in arbitrage funds. Panel B of Table 41 reports the same descriptive statistics when funds were classified in decile based on the average performance of the funds over the whole period. On the other side. long/short funds. particularly for 70 There are 3 funds with no sub-strategy. This result can be explained but the low skewness and extremely large kurtosis of mortgage-backed securities funds. we present the descriptive statistics. Median returns indicate the same pattern except that mortgagebacked funds have an impressively high median return compared with the other funds studied. All average returns are significantly positive.70 These hedge fund data are contrasted with the descriptive statistics of the S&P 500 that represent the equity market.16% with a monthly volatility of 1. the correlation analysis and the attrition rates in this section.1 Descriptive statistics In Panel A of Table 41 contains descriptive statistics of the funds. Mortgage fund returns are more volatile than the other market neutral funds.16%.08% over the period studied with a monthly standard deviation of O.III Descriptive statistics and attrition rates Before analyzing the presence of bias in hedge fund data. Panel A of Table 40 indicates that market neutral funds offered an average monthly return of 1.74%). Arbitrage funds offered almost the same return with the same volatility whereas long/short funds offered 1. whether alive or dead. mortgage backed securities funds and non-classified funds. the maximum is also relatively low for all strategies. The data regarding these funds ended in January 2001. 3. 283 . in our database. Minimum monthly returns are relatively high with a total of –2.

which.69%. This return is weakly and significantly positive with a relatively high volatility of 4. which is in line with the long-term return of the equity market of around 8%. Panel B reports the same descriptive statistics when we classify all these funds in decile on the basis of their average performance over the whole period studied.93%) and mortgage backed funds (2. Decile 1 contains the worst performing funds over the period and decile 10 the best performing ones. By comparison. in turn offer a better trade-off as mortgage-backed securities funds. 52% of the funds in the worst 3 deciles are dissolved against 31% on the average for the top 3 deciles.47%. The median return is however as high as the median return of market neutral funds considered as a whole and higher than all the sub-strategies except the mortgage-backed index. All classifications also have significantly positive kurtosis which leads to the presence of fat tails (a least to some extent). Moreover 62% of the funds in decile 1 are dissolved. This is in line with previous results indicating normality of returns for the stocks indices (see for example Fung and Hsieh.45). Mortgage backed securities have largely negatively skewed returns’ distribution (skewness of –4. The dead fund column indicates that bad performing funds have more chance to be dissolved. Interestingly. 1999). On the average.arbitrage funds (2.11) with fat tails (kurtosis of 26. the equity index has a low average monthly return of 0. All other sub-strategies except long/short funds and the strategy considered as a whole all have significantly negative skewness. nor has fat tails. which leads to negatively skewed return distribution. fewer funds from deciles 5 to 9 have been dissolved compared 284 . Long/short funds have positively skewed return distribution. The Sharpe score (it is defined as the ratio of average return over the standard deviation) indicates that arbitrage funds offer a better trade-off than long/short funds.84%). The minimum and maximum for the index are more extreme than those obtained for the hedge funds and the return distribution of the index is neither skewed.

The first step of the analysis of the neutrality of equity market neutral funds is to check the correlation of these funds with the equity index as approximate by the S&P 500 index. hedge funds are likely to improve the risk-return trade-off when added to a traditional portfolio (see Fung and Hsieh. Skewness and kurtosis results do no lead to particular remarks. Amin and Kat. 4. 1997. 3. These results are in line with the one obtained by Capocci and Hübner (2004) in their dissolution frequencies analysis but their deciles were constructed on the basis of previous year’s performance. Liang. 1997. thanks to the weak correlation between hedge funds and other securities. We can however note that some skewness are significantly positive (deciles 2. 5 and 6) and that all kurtosis that are significant are positive. There is only one insignificant negative kurtosis. The minimum and maximum columns indicate that very low minimum will lead to lower decile (decile 1 minimum is extremely low compared to other minimums). 285 . but that high maximum does not necessarily lead to higher decile (decile 1 maximum is the second highest maximum over the 10 deciles).2 Correlation analysis As suggested in the introduction. Schneeweis and Spurgin. the traditional hedge funds literature contends that. 7 and 9) whereas others are significantly negative (deciles 1. 1999.to decile 10 and middle decile fund returns are more stable than worst and top performing decile (see standard deviation). 2001). The correlation coefficients are reported in the Table 41.

47% 286 .No sub-strategy 3 0.06% *** 1.Table 40: Descriptive statistics and decile descriptive statistics Panel A: Strategy.30% 197 109 1. sub-strategies and index descriptive statistics No of Fds % of the strategy Living Funds Dead Funds Mean Return S.Long/short funds 306 48.11% Index NA NA NA NA 0.50% 0 3 0.73% *** 1. Market neutral funds 634 100% 398 236 1.69% * 4.16% .90% 26 18 1. D.92% .94% .32% .16% *** 1.Arbitrage funds 281 44.08% *** 0.30% 175 106 1.03% *** 0.Mortgage backed funds 44 6.

92% .11% Index NA NA NA NA 0.30% 175 106 1.73% *** 1.50% 0 3 0.Arbitrage funds 281 44.03% *** 0.Panel A (continued): Strategy.08% *** 0.90% 26 18 1.69% * 4.Long/short funds 306 48.16% *** 1. D.30% 197 109 1.32% .06% *** 1.No sub-strategy 3 0.94% . sub-strategies and index descriptive statistics No of Fds % of the strategy Living Funds Dead Funds Mean Return S.16% .47% 287 . Market neutral funds 634 100% 398 236 1.Mortgage backed funds 44 6.

84% Decile 6 64 72% 28% 1.95% Decile 5 64 80% 20% 0.55% *** 0.72% *** 0.89% *** 0.10% Decile 9 63 81% 19% 1.28% Decile 8 63 71% 29% 1.75% *** 1.34% 3.21% Decile 3 63 62% 38% 0.40% *** 1.10% *** 1.43% Decile 10 63 56% 44% 2.53% Decile 2 63 43% 57% 0. D. Decile 1 63 38% 62% -0.59% Decile 4 63 56% 44% 0.Panel B: Decile descriptive statistics (120 months) No of Fds Living Funds Dead Funds Mean Return S.21% *** 1.04% Decile 7 63 71% 29% 1.28% *** 1.41% 288 .70% *** 2.

02 *** 1.00% 2.50% 4.30% -2. sub-strategy and index descriptive statistics.80% -2.D.95 Decile 8 1.11 *** 1.20% 0.23 Decile 10 2. the mean return.40% 0.50% -4. percentage of the strategy (% of the strategy).30% 4.40% 0.11 0.36 *** 2.10% 0. the standard deviation (S.50% 8.05 *** 0. the median.01 -0.9 *** 10. the maximum (max).70% 5.60% -1.3 ** 0.94 *** 0.1 Decile 2 0.93 Decile 4 0.70% -1.28 *** 5. the minimum (min).31 *** 1.40% -1.02 *** -0.20% -2.23 Decile 3 0.40% 6.30% 0.31 1. ** Significant at the 5% level and * Significant at the 10% level.51 *** 0. the number of living funds. the kurtosis and the Sharpe score for the individual hedge funds in our MAR/CISDM database for the whole period 01:1993-12:2002. 289 .05 Decile 6 1.20% -0.72 *** 1.20% 2.Panel B (continued): Decile descriptive statistics (120 months) Median Min Max Skewness Kurtosis Sharpe score Decile 1 -0.).40% 2. the number of dead funds.00% -4.27 Decile 9 1. the skewness. *** Significant at the 1% level.70% 4. Panel A focus on strategy.01 1. Sharpe score is the ratio of return and standard deviation.76 Decile 5 1.10% 9.12 This Table shows the number of funds (No of Fds).27 *** 1.20% -2.20% -17.78 *** 0. Panel B on decile descriptive statistics.90% 0.50% -1. No sub-strategy ended in 01:12001.40% -2.02 0.07 Decile 7 1.70% -2.24 *** 1.

arbitrage and long/short funds is higher than 90%.No sub-strategy Index 1 0.92 0.7 (between arbitrage funds and long/short funds).3 1 This Table reports the correlation coefficients among market neutral sub-strategies and between market neutral funds strategies and the index. The correlation between market neutral funds (considered as a whole or per sub-strategies) ranges from 0. This result is logical since these 2 sub-strategies represent a total of more than 90% of the funds in the database.37 (between mortgage-backed funds and no sub-strategy funds) to 0.7 0. Table 41 indicates that the correlation between the market neutral index and its main components.54 (between market neutral funds 290 .58 0.62 0.46 1 0.54 1 0. These figures are reasonably high and suggest that the subclassification has a sense since no sub-strategies are completely correlated.49.54 Arbitrage funds Long/short funds Mortgage backed funds No substrategy Index 1 0. including 398 survived funds and 236 dissolved funds as of December 2002.Arbitrage funds .19 (between no sub-strategy funds and the index) to 0.53 0. The correlation between the market neutral sub-strategies ranges from 0.Mortgage backed funds .Table 41: Correlation between market neutral strategies and equity index Market neutral funds Market neutral funds .37 0.Long/short funds .49 0. The period covered is 01:1993-12/2002 period. The database consists of 634 individual market neutral funds.19 1 0. The correlation between the market neutral index and the mortgage funds is significantly lower at 0.91 0.45 0.3 0.

Liang. 71 We removed 3 funds with few data to perform the correlation analysis at the individual fund level. Figure 9 reports the average.3 for market neutral funds. 291 . The bottom line of the block represents the minimum individual correlation between market neutral funds and the equity index. the top of the box is the maximum individual correlation between a single market neutral fund and the equity index. This result indicates that even if individual market funds are not correlated with the equity index. 1998. median. 0. The upper part of the grey block ended at the average correlation between market neutral funds and the index. 2004).and the index). -0. This point is particularly important for the beta analysis. 1999. These figures are reasonable but relatively high for the index. Amin and Kat. We do not report the result of the no sub-strategy index because it includes only 3 individual funds. The median is reported below the white part of the block at 0.29 in the first case.71 We describe Figure 9 using market neutral funds as an example. it is important to consider individual funds because index aggregation can increase the correlation and the exposure to the equity market. Figure 9 indicates that the individual correlation between individual hedge funds and the equity index vary strongly for the strategy and the sub-strategies considered and that the individual correlation between hedge funds and the index are much lower than the correlation between hedge funds indices and the equity index. Finally.15 for market neutral funds. 2001 and Capocci and Hübner. minimum and maximum individual correlation between market neutral fund index and its sub-strategies and the equity index. These results are in line with previous results obtained by Schneeweis and Spurgin. The sub-strategy correlation analysis also reported in Figure 9 confirms this topic since the correlation between the sub-indices and the equity market is lower than the correlation of the global market neutral index that contains all these sub-strategies.

median and mean individual correlations between market neutral funds and the equity index between January 1993 and December 2002. Liang 2003). maximum.Figure 9: Minimum.3 Birth and attrition rates Attrition rates of hedge funds are largely publicized in academic studies (see for example Fung and Hsieh. Liang. This rate can be defined as the percentage of funds in the database that are dissolved each year. maximum. 1997. 292 . 2000. median and mean individual correlations between market neutral funds and the equity index 80% 60% 40% 20% 0% -20% -40% -60% Market Neutral Funds Arbitrage Funds Long/short Funds Mortgage backed Funds This figure reports the minimum. Table 42 reports the birth rate and the attrition rate of the fund in our database for each year under review. 3.

The attrition rate is defined as the ratio of dissolved funds to number of funds at year start. 293 .20% 32.10% 8.10% 1.10% 11.40% 17.60% 8.20% 9. including 398 survived funds and 236 dissolved funds as of December 2002.90% 56. The database consists of 634 individual market neutral funds. The period covered is 01:1993-12/2002.40% 35.60% 19.00% 8.10% 30.10% 9. Birth rate is defined as the ratio between the number of new funds and the number of funds at year start.10% 45.70% 11.50% 13.70% This Table reports the number of funds at year start.Table 42: Birth and attrition rates Year start New Dissolved Year end Birth rate Attrition rate 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 63 89 137 163 224 285 324 346 361 396 Total 27 50 44 74 80 78 58 47 69 44 571 1 2 18 13 19 39 36 32 34 42 236 89 137 163 224 285 324 346 361 396 398 Average 42.70% 27.90% 13.20% 13. the number dissolved funds. the number of new funds.40% 10.60% 2. the birth rate and the attrition rate for the market neutral funds in our database.

294 .Table 42 indicates that birth rates are much higher than attrition rate but that this rate diminished over time from more than 50% in 1994 to 11% in 2002. Table 42 also indicates that the total number of funds has increased linearly over the period. On the other side. The high level of dissolution in 1998 should probably be due to the Russian. Goetzmann and Ibbotson (1999). These results are in line with the results obtained by Liang (2000) for the TASS database but lower than the 14% obtained by Brown.7%.7% of the fund in the industry are dissolved. the Asian crisis and the bailout of Long Term Capital Management. on average around 8.7% indicating that each year. LTCM. The average birth rate in around 30% and the average attrition rate is 8. the attrition rate is very low for the first 2 years under investigation and is almost stable between 1995 and 2002 ranging from 8% to 13.

Ackermann et al. since we focus on market neutral funds. In this study. Fung and Hsieh. Asian crisis and the LTCM bailout) and 2002 (3rd year of bearish market).22%. Standard deviation is interesting because it indicates that in recent years the volatility of the dissolved funds has increased dramatically compared to surviving funds. 295 .77%. higher than the whole funds in the database average return of 1. 1998 (Russian.. Fung and Hsieh.g. we will analyse the presence of survivorship bias for the funds in our database. 1999.IV Survivorship bias analysis Survivorship bias is particularly important in the case of hedge funds (see Fung and Hsieh. Ackermann. McEnally and Ravenscraft. Survivorship bias can be defined in 2 ways: the performance difference between surviving and dissolved funds (e. 1997. 2000). Usually this bias is studied on a global basis for full databases including a variety of different strategies. Panel A of Table 43 indicates that the returns have been less interesting in recent years compared to the past. 2004). This indicates that volatile funds have been more exposed to dissolution in 2001 and 2002. 1999) and the performance difference between living and all funds (e. The average return for the funds that survived over the whole period is 1.g. We report the bias using both definitions. The most difficult years for market neutral funds have been 1994 (bond market crisis).08% and higher than the returns of funds that have been dissolved during the period studied with a performance of 0. 2000. Capocci and Hübner.

Panel B reports the difference of performance between living and dead funds (left side) and living and all funds (right side). 296 .72 72 This consensus value is quite high when compared to the 0.2% found by Capocci and Hübner (2004) for the 1994-2000 period. Table 43 indicates that the difference in performance between living funds and dead funds has increased dramatically compared to the difference between living funds and all funds in recent years to almost 1% in 2000 and 2002. lower than the 0. This latter value is much higher than the very low value obtained by Ackermann et al. It is however lower than the 3% bias found by Liang (2001). for the period 1988-1995.4% using the first formula and 1.30% monthly bias found by Fung and Hsieh (2000) and slightly higher than the percentage of 1. The total bias is 5. which is also the industry consensus as stressed by Amin and Kat (2001).5 bias reported by Malkiel (1995) and Brown and Goetzmann (1995) for US mutual funds. It is similar to the percentage of 1. The latest figure is the generally used one in hedge fund studies and can be compared to past results.5% from Fung and Hsieh (1998).8-1.68% using the general one.

Return S.36% 0.D.80% 1.02% 1998 0.58% 1.68% 1995 1. D.44% 0.55% 1.43% 1.74% 0.16% 0.20% 1.63% 1997 1.88% 0.08% 0.22% 0.37% 0.Table 43: Survivorship bias Panel A: Annual performance (all funds.70% 1.10% 0.94% 1.25% 0.37% 0.47% 1994 0.14% 1.33% 0.91% 2000 1.76% 0.58% 0.54% 0.13% 0.59% -0.54% 0.41% 0.58% 1.55% 2.44% 2001 0.39% 2002 0.47% 1.75% 1996 1. 1993 1.41% 1.78% 1.26% 0.42% 1.42% 0.19% 297 .46% 1.64% 1.41% 0.92% 1.77% 1.62% 0.96% Average 1.61% 1999 1.45% 0. surviving funds and dissolved funds) All Funds Surviving Funds Dissolved Funds Year Return S.77% 1. Return S.43% 0.85% 0.63% 0.69% 0.56% 1. D.65% 0.56% 0.64% 1.48% 1.65% 0.47% 0.80% 0.

Panel B: Survivorship bias Living .15% 0.16% 0.40% Return 0.73% 0.45% 5.17% 0. 298 .05% 0.06% 0.20% 0. All returns are net of fees.23% 0.09% 0.68% per Month per Year This Table reports the survivorship bias of our database. Numbers in the table are yearly percentage unless otherwise indicated.32% 0. In Panel B survivorship bias is calculated as the performance difference between surviving funds and dissolved funds (lefthand side) and as the performance difference between surviving funds and all funds (right-hand side).94% 0. including 398 survived funds and 236 dissolved funds as of December 2002.97% 0.07% 0.32% 0. Our MAR/CISDM database contains 634 individual market neutral funds.29% 0.06% 0.All Funds Year 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 Bias 93-02 Return 0.14% 1.28% 0.35% 0.38% 0.10% 0.Dead Funds Living .

We classify the funds in decile on the basis of their ex-post betas and analyse their descriptive statistics. In the second part of the analysis we perform the same analysis for individual funds and by extracting a bull and a bear market period. The methodology used in this study is two-fold. we classify them in decile on the basis on the global performance over the period studied in order to determine which funds were the most exposed to the market. we analyse market neutral funds grouped per sub-strategy. we also perform an ex-post beta analysis.V Methodology The objective of this study is to determine if market neutral funds are exposed to the equity market or if they are really market neutral. Secondly. First. we classify them using Carhart’s (1997) methodology in order to determine if there is a pattern to detect portfolio’s constructed on the basis of previous year’s performance. This part of the analysis has the objective to determine if it is the same funds that performed well in bear/bull market and in the whole period. we classify funds on the basis of their previous year performance and we create return series. Thirdly. The first part of this analysis has the objective to analyse the exposure of market neutral funds to the equity market. We perform this analysis in three steps. Does the beta exposure explain better or worst returns? 299 . This methodology uses the same results as previously but differently. Each year. we estimate the exposure to the market for all these deciles in order to determine if best or worst performing funds are more or less exposed to the market. Then. For individual funds.

εPt = error term. RMt = return of the equity market portfolio. It is a relative measure. αp commonly called Jensen’s alpha (1968) is usually interpreted as a measure of out. We compute all estimations by using NeweyWest (1987) standard errors to adjust for any autocorrelation in the returns 300 . T (12) Where RPt = return of fund P in month t.2..The model used is a single index model based on the classical CAPM developed by Sharpe (1964) and Lintner (1965).or under-performance relative to the market proxy used.. respectively. in our case the S&P 500 on month t. Its equation to estimate is the following: R Pt Ft = α P + β P * R Mt + ε Pt t = 1.. αP and βP are the intercept and the slope of the regression. The intercept of this equation.. The beta is interpreted as a measure of the dependence of a fund’s return to the index.

08 301 .Arbitrage funds 1.Long/short funds 1.29 .03 .No sub-strategy 0. Alpha Market R²adj Market neutral funds 1.62% *** 0.06 * 0.Mortgage backed funds 1.16% 1.08% *** 0.29 .02% *** 0.11% 0.11 *** 0.14 *** 0.Table 44: Market exposure analysis Panel A: Market exposure for market neutral funds Average return S.08 *** 0.06% *** 1.00% *** 0.21 .92% 0.05% *** 0.16% *** 1.03% *** 0.32% 1.94% 1. D.73% *** 1.97% *** 0.1 *** 0.

59% 0. The period covered is 01:1993-12:2002.. D.06 Decile 2 0.95% 0. Panel A presents the average return.1 *** 0.2 *** 0.13 The table reports the market exposure analysis.31 Decile 7 1.19 Decile 9 1.11 *** 0.21 *** 0.21% *** 1.04% 1.08 Decile 6 1.10% *** 1.08 Decile 8 1. the standard deviation (S. *** Significant at the 1% level.85% *** 0. including 398 survived funds and 236 dissolved funds as of December 2002.13 *** 0.41% 2.40% *** 1.66% *** 0.11 *** 0.53% -0.75% *** 1.28% 1.89% *** 0.10% 1.15% *** 0.Average return S.70% *** 2.19 Decile 10 2.29 Decile 5 0. Panel B presents the same information when funds are classified in deciles on the basis of their performance over the period studied.84% 0.12 Decile 4 0. Alpha Market R²adj Decile 1 -0. t-stats are heteroskedasticity consistent.32% *** 0.14 *** 0. the market exposure (the beta) and the R²adj. Our MAR/CISDM database contains 634 individual market neutral funds.34% 3.12 Decile 3 0.22% * 0.09 *** 0. the alpha.72% *** 0.56% *** 0.05 *** 0.06 *** 0. 302 .51% *** 0.01% *** 0.D).28% 1.21% 0.55% *** 0. ** Significant at the 5% level and * Significant at the 10% level.43% 1.64% *** 0.48% 0.

The betas indicate however that all indices except mortgage-backed securities funds are significantly exposed to the equity market at the 1% significance level. As previously. These results are based on decile constructed on the basis of the performance over the period studied. 303 .14 for long/short funds.VI Strategy and decile analysis This section reports the results of the exposure analysis obtained using strategy and sub-strategy indices and on the basis of the decile classification. the R²adj are relatively low. the beta obtained is low between 0.06 for mortgage backed securities and 0. which indicates that only a small part of market neutral returns can be explained by the market. Panel A reports the average return and standard deviation of returns over the period studied as well as the results of the regression analysis. Panel B indicates the same pattern as Panel A. but all market betas are significantly positive even if they are all low in absolute term. On an absolute term however. most market neutral indices indicating that the market cannot explain a big proportion of the average return of the indices even if the 29% R²adj obtained for market neutral funds is relatively high for this kind of model. The R²adj are relatively low. the difference between the alpha and the average return is relatively low. top performing funds are reported in Decile 10 and worst performing funds are reported in Decile 1. A comparison between the average return and the alpha indicates that they are relatively close to each other. Panel B reports the results obtained using the decile classification. Panel A of Table 44 reports the results for the market neutral strategy and various market neutral sub-strategies. Top and worst performing funds have however the highest market exposure.

304 . Then. The objective of this analysis is to determine if last year best or bad performers are more exposed to the market. we classify the funds on the basis of their previous year performance and we create return series. In a second step. Each year. The left-hand axis reports the number of funds per year. The right-hand axis reports the number of funds per decile. The average is for the January 1994 to December 2002 period. we perform the same analysis on the basis of Carhart (1997) methodology. we estimate the exposure to the market for all these deciles in order to determine if best or worst performing funds are more or less exposed to the market.Figure 10: Lagged 12-month decile description 450 400 350 300 250 200 150 100 50 0 Av er ag e 19 97 19 98 19 99 19 93 19 94 19 95 19 96 20 00 20 01 50 45 40 35 30 25 20 15 10 5 0 Funds per year Funds per decile This figure reports the yearly lagged 12-month decile description.

These results suggest that the exposure of market neutral funds to the equity market is not clear. the market factor does not explain a large part of the alpha created by market neutral funds and the low R²ADJ obtained suggest that the exposure does not describe market neutral fund returns well. the market column indicates that 6 out of 10 deciles are significantly exposed to the market. we report the number of funds in each decile in Figure 10. the market factors are another time very low. It reports the number of funds in the analysis each year (left axis) and the number of funds per decile each year (right axis) as well as the average number of funds per year over the whole period and the average number of funds per decile. which confirms the results. Over the whole period.69% for the index. decile 1 (worst performing funds) would have offered 0. Even if the 6 out of 10deciles's exposure are significantly positive. Moreover. In absolute term. the R²adj obtained is very low.Before analysing the results of this analysis. Figure 10 indicates that the average number of funds per year is 264 with decile of 28 funds on average. A comparison between the average return and the alpha obtained in Table 45 indicates that the market exposure cannot explain the performance of market neutral funds since the alphas obtained before and after the regression are very close.72% per month against 1. in absolute term.26% for decile 10 (best performing funds) and 0. 305 . On the other hand however. The spreads reported on the bottom of Table 45 indicate that the spread between decile 1 and decile 10 is significantly positive but that the spreads between decile 1 and decile 2 and between decile 9 and decile 10 are not significantly different from zero. These deciles are spread but the 3 top performing deciles are all significantly exposed to the market.

69% 2.15% *** *** *** *** *** *** *** *** *** *** ** 2.65% 0.26% 0.04 -0.29% *** *** *** *** *** *** *** *** *** *** *** 0. 306 .Table 45: Lagged 12-month decile exposure Portfolio Average return S. including 398 survived funds and 236 dissolved funds as of December 2002. the standard deviation (S.06 0.75% 0.63% -0.66% 2.01 The table reports the market exposure using deciles constructed on the lagged-12 month performance (see Carhart 1997).D). Our MAR/CISDM database contains 634 individual market neutral funds.62% 0.32% 0.60% 0.63% 0.06 0.04 0.48% 0.95% 1.80% 0.08 0.11% 1.89% 1.03 0.09% 1. *** Significant at the 1% level.08 0.03 0.61% 0.07 0.11 0.03% 0. It reports the average return.01 -0.71% 1.03 0.07% 0. The period covered is 01:1993-12:2002.22% 1.14% 1. Alpha Market R² adj D1 D2 D3 D4 D5 D6 D7 D8 D9 D10 1-10 spread 1-2 spread 9-10 spread 0.10% 0.1 0. the market exposure (the beta) and the R²adj.11 0.82% 0.72% 0.03 0.67% 0.03 0.01 0. ** Significant at the 5% level and * Significant at the 10% level.33% 0.70% 0.67% 0.79% 0.54% -0. D. the alpha.06 0.04% 1.83% 0. t-stats are heteroskedasticity consistent.88% 1.11 0.11 0 -0.01 * *** ** ** 0.38% 2.04 -0.05 0.49% 0.01 *** * ** * *** 0.61% 0.

In the first sub-section. 214 long/short funds. we look at the ex-post beta exposure. Table 46 reports the average return. 7.73 Results are reported in Table 46. This lead us to remove 172 funds (74 arbitrage funds. significantly negative or not significant. the return distribution which is the percentage of average monthly returns that are significantly positive (+). Like average return. 73 We require all funds to have at least 24 months of data to perform this analysis. To check this. 5 mortgage backed funds and 1 no sub-strategy fund) from the database and leave us with a total of 462 funds (207 arbitrage funds. 92 long/short funds. we base our analysis on the individual fund return series in order to determine if the results previously obtained using strategy and sub-strategy indices are still valid. it is important to check if the results obtained on a strategy or sub-strategy basis are still valid on an individual fund basis. significantly negative (-) and not significant (0) on the one side. individual alpha and beta are classified as significantly positive at the 5% significance level. we perform the same analysis using the returns of individual funds. In the second one. we analyse the market exposure. and the results of the regression of the other. 39 mortgage backed securities funds and 2 no sub-strategy funds). 307 .VII Individual fund analysis In this section.1 Market exposure As we suggested in the correlation analysis.

The percentage of market neutral funds average return not significantly different increases from 21% to 38% when the market impact is taken into account. Results are similar for the main sub-strategies. First. Alphas are always lower than the corresponding average return but the differences are relatively small. 308 . On an absolute term. the aggregation of funds in indices leads to an increase in the exposure to the equity market. These two remarks are of particular interest. The average returns and alphas obtained for individual funds are higher compared to the ones obtained while studying funds agglomerated in indices. This result is not inline with the previous one and it indicates that almost two-third of the individual beta exposure for market neutral funds are not significantly different from zero and not exposed to the market. This result could be explained by 2 reasons. the market exposure remains obviously the same here as it was in previous analysis. much more alphas are concentrated around 0. being not significantly different from it. but they explain more precisely the ones previously obtained. The inclusion of the market index puts many significantly positive average return back to no significance even if the quantitative impact on alpha is small. since the bulk of the funds are not significantly exposed to the market but one third of the funds are and only five percent of the funds are negatively significantly exposed to the market. Very few average returns are significantly negative over the period studied. Secondly. Nevertheless. It is however particularly interesting to underline that the beta distribution for individual funds indicates that the majority of betas are not significantly different from zero.The left-hand side of Table 46 indicates that the vast majority of individual average returns are significantly positive and that around 20% of them are not significantly different from 0. the funds significantly exposed could bias the results. This first reason is in line with the results obtained in the correlation analysis and stresses the importance that it is of major importance to consider individual funds in empirical analysis because results based on indices could be biased.

.03% 71% 27% 2% 4.94% 71% 29% 0% 0. The period covered is 01:1993-12:2002.30% 1.No sub-strategy 0.05 .11 8% 92% 0% 0. the alpha.50% 0. Our MAR/CISDM database contains 634 individual market neutral funds. 0 not significantly different from 0 and – significantly negative).36% 50% 50% 0% 0.09 33% 63% 3% 0.08 R²adj Market neutral funds 1. (+ are significantly positive.00% 52% 47% 0% 0.18 50% 50% 0% 0.30% 0.13 32% 0 63% 5% 0.Mortgage backed funds 0. 0 not significantly different from 0 and – significantly negative). Alpha Alpha distribution (5%) + 0 38% 0% Beta distribution (5%) Market + 0.Table 46: Individual fund market exposure Average return Return distribution (5%) + 0 21% 1% S. alpha distribution at the 5% level (+ are significantly positive.02% 78% 3.03 . the beta distribution (+ are significantly positive. the market exposure (the beta). 0 not significantly different from 0 and – significantly negative) and the R²adj.Arbitrage funds 1.30% 0. It reports the average return.D.12 The table reports the individual fund market exposure. T-stats are heteroskedasticity consistent. the return distribution at the 5% level.D).16 36% 57% 7% 0.75% 67% 33% 0% 0.95% 62% . the standard deviation (S.45% 50% 50% 0% 1.1 .Long/short funds 1.89% 85% 13% 3% 3. including 398 survived funds and 236 dissolved funds as of December 2002.03% 84% 16% 0% 2.00% 0.

the beta distribution (+ are significantly positive.Mortgage backed funds 1. the standard deviation (S.21 The table reports the individual fund market exposure.13 5% 73% 22% -1. Our MAR/CISDM database contains 634 individual market neutral funds. including 398 survived funds and 236 dissolved funds as of December 2002.Arbitrage funds 1. T-stats are heteroskedasticity consistent.35 7% 82% 11% 0.10 .No sub-strategy 0. the alpha.09 4% 77% 19% -0.01 0% 100% 0% -4.30 15% 83% 1% 6.59 50% 50% 0% 148.08 5% 79% 15% -2.34 15% 85% 0% -4.27% 0% 50% 50% 0. the return distribution at the 5% level.Table 47 : Individual Fund Market Exposure – Multi factor model Alpha <0 0 >0 Mkt <0 0 >0 Mkt² <0 0 >0 Mkt³ <0 0 >0 R² adj Market neutral funds 1.Long/short funds 1. the market exposure (the beta).47% 0% 8% 92% 0.10% 0% 25% 75% 0. .12 .69 12% 83% 5% -1. (+ are significantly positive. alpha distribution at the 5% level (+ are significantly positive.47 10% 87% 3% 0.D).21% 0% 35% 65% 0. The period covered is 01:1993-12:2002.13 14% 82% 3% 2. 0 not significantly different from 0 and – significantly negative). It reports the average return.17 7% 69% 24% -1.09 .16 11% 82% 7% 0.98 0% 100% 0% 0. 0 not significantly different from 0 and – significantly negative).09 .51 9% 82% 9% 0.77% 0% 50% 50% -0. 0 not significantly different from 0 and – significantly negative) and the R²adj.

15% 1% 38% 38% 0.04 7% 81% 12% 0.11 .24% 1% 60% 60% 0.01 0% 100% 0% 0.Long/short funds 1.09 8% 76% 14% 0. including 398 survived funds and 236 dissolved funds as of December 2002. 0 not significantly different from 0 and – significantly negative). the standard deviation (S. 0 not significantly different from 0 and – significantly negative).Arbitrage funds 1.10 The table reports the individual fund market exposure. It reports the average return.22 .Mortgage backed funds 1.19 5% 64% 31% 0. The period covered is 01:1993-12:2002.14 6% 60% 33% 0.05 Alpha <0 0 >0 <0 0 >0 <0 0 >0 R² adj Market neutral funds 1.30 . the alpha.20 2% 70% 27% -0.31 . the market exposure in bull markets (bull beta estimated when the S&P500 is up over the month under review) and in bear markets((bear beta estimated when the S&P500 is down over the month under review ) as well as the beta distribution (+ are significantly positive.Table 48: Bull and Bear Beta Estimation Bear Beta 0.10 7% 86% 7% 0.58% 2% 23% 23% 0. alpha distribution at the 5% level (+ are significantly positive.18 0% 33% 67% 0. 0 not significantly different from 0 and – significantly negative) and the R²adj. . Our MAR/CISDM database contains 634 individual market neutral funds.22% 1% 47% 47% 5% 62% 32% 8% 79% 13% 0.No sub-strategy 0.17 Bull Beta 0. T-stats are heteroskedasticity consistent. the return distribution at the 5% level.90% 0% 33% 33% 0.D). (+ are significantly positive.

7. the impact is negative for around 15% of the funds considered. The objective of this calculation is to have an estimation of the co-skewness (market squared) and the co-kurtosis (market beta cubed) between the fund or the strategy considered and the market. Table 47 firstly confirms that the market exposure tend to be low for the bulk of the managers even if up to one fourth of the fund may be significantly exposed to the market.2 Non linearities in market exposure In order to complete our analysis we perform two additional analysis. all the strategies considered have a negative impact on skewness indicating that considered globally. The beta cubed reports the same information for the co-kurtosis. Interestingly. Results are reported in Table 47. market neutral strategies tend to lead to negatively skewed distribution. All in one. The results are almost the same when co-kurtosis is considered in the sense that the bulk of the individual funds do not have any impact on the kurtosis of the market return distribution but that around 10% will turn the kurtosis to positive or negative levels. the repartition of the beta squared shows that more than 80% of the individual estimations have no significant impact on skewness. these results show 312 . A negative beta to the market squared indicates that the fund or strategy considered tend to move the distribution of the return of the index to the left (negative asymmetry) while a positive co-skewness indicates a positive impact of the asymmetry of the return distribution. Secondly. Fang and Lai (1997). However. We first check the co-skewness and the co-kurtosis of market neutral funds with the equity market using a multi-factor CAPM based on the underlying idea developed by Kraus and Litzenberger (1976). Our four-moment CAPM use not only the beta of the market as a regression factor but also the beta of the market factor squared as well as the beta of the market cubed. Harvey and Siddique (2000) and Dittmar (2002). The inclusion of the fund or the strategy may lead to bigger tails (positive beta) or to lower tails (negative beta).

the bear betas are much higher than the bull betas confirming that hedge funds tends to have a higher exposure to markets in turbulent market conditions. decile 10 and decile 6. Table 49 indicates that there is no clear pattern in the average return column. The individual beta repartitions reported also indicate a higher beta exposure in bear market conditions where around one third of the exposures are positive while it is around 10% in bull market conditions.2 Ex-post beta analysis In the second step of the analysis we use the results obtained in the first part of this section to study the link between the beta and the performance of the funds. We separated the bull and the bear market beta to determine if the beta of the fund or the strategy considered remains constant over bull or bear market conditions. On the beta side. We use the individual results obtained in the previous section but we classify funds in Table 49 on the basis of the betas from the lowest to the highest one. These results indicate some presence of non-linearities even if there are not as high as one could expect. Results are reported in Table 48. decile 1. Table 48 reports the bull and bear alphas and betas. The first element to stress out is that the alphas reported are relatively high and that a little more than half of the funds in the dataset offers significantly positive alpha (mainly arbitrage and mortgage funds). 7. The return distribution however indicates that more insignificant returns are offered by low beta funds (see decile 1) and by high beta funds (decile 9 and 313 .some impact on the distribution of the returns for some funds but that the impact is marginal for most individual funds considered. For our second additional estimation we apply two types of piecewise regressions following the underlying idea of Faff (2001) and Galagedera and Faff (2004). The figure indicates that the highest average returns are spread between decile 8.

314 . This result is also true for other deciles but less precise. The distribution of the alpha indicates the same pattern as the return distribution with higher figures for alpha’s not significantly different from zero. in other words that the market can explain a major part of their returns. strongly exposed or negatively and significantly exposed deciles (decile 1. A closer look at the regression results indicates that the highest alphas are offered by the same deciles that offered high average returns but with lower absolute values. This could be explained by the fact that funds reported in decile 1 benefited mainly from bear market and funds in deciles 9 and 10 did so from bull market conditions. Negatively exposed and strongly positive exposed funds are more volatile than market neutral funds. This result is in line with the objective of market neutral funds which is to offer low volatility positive returns over time. These results also indicate an interesting pattern. decile 9 and decile 10) indicating that these funds were not able to offer significantly positive returns when the market impact was taken into account or.10). decile 9 and decile 10. This means that the best performing market neutral funds are those that are not exposed to the market. The standard deviation column indicates that the same decile 1. This remark is particularly true. Please note that the alpha for middle decile funds is weakly exposed to the market and very close to the values obtained for the average returns confirming the fact that these funds are unexposed to the market.

the beta distribution columns indicate a clear pattern. 315 . This logical result confirms our previous suggestion that low decile funds profit from the bear market.As we could expect. All over. No betas are significantly positive for the more weakly exposed deciles (deciles 1 to decile 3). this analysis suggests that most market neutral funds are not significantly exposed to the market. and by construction. As before the R²adj are low except for decile 10. This result can be explained by the fact that funds reported in decile 10 are more exposed to the market. the beta analysis indicates that the betas are increasing over the decile. More interestingly. the number of significantly positive returns increase monotonically and the percentage of significantly negatively exposed to the market funds go to zero after decile 4. indicating that there has been more positively market exposed funds than negatively exposed funds over the period studied. Then.

16 0.80% 0.14% 0. the beta distribution (+ are significantly positive. Our MAR/CISDM database contains 634 individual market neutral funds. alpha distribution at the 5% level (+ are significantly positive. (+ are significantly positive.05% 0.24% 1.01 0 0 0.01 0.00% 2. including 398 survived funds and 236 dissolved funds as of December 2002. the standard deviation (S.04 0.07% 0.21 0.89% 0.81% 1.80% 1.31 R²adj Decile 1 Decile 2 Decile 3 Decile 4 Decile 5 Decile 6 Decile 7 Decile 8 Decile 9 Decile 10 1.11% 59% 85% 98% 91% 89% 89% 78% 85% 61% 41% 4.60% 1.D.37 0. It reports the average return.Table 49: Ex-post beta analysis Average return Return distribution (5%) + 0 39% 15% 2% 9% 9% 9% 20% 13% 39% 57% 2% 0% 0% 0% 2% 2% 2% 2% 0% 2% S.80% 3.00% 1.07% 0. 0 not significantly different from 0 and – significantly negative). the alpha.12 0. the return distribution at the 5% level.07 0. 0 not significantly different from 0 and – significantly negative).90% 0. . 0 not significantly different from 0 and – significantly negative) and the R²adj.02 0.1 0. The period covered is 01:1993-12:2002. the market exposure (the beta).06 -0.07 0.00% 1.96% 0. t-stats are heteroskedasticity consistent.1 0.01 -0.96% 0.96% 1. Deciles are constructed on the basis of the market exposure of the individual funds.90% 1.00% 1.D).03 0.27 -0.90% 2.85% 39% 70% 83% 83% 81% 83% 59% 67% 26% 26% The table reports the ex-post beta.17% 0.70% 8.90% 2.60% 1.20% 1.90% 4.00% 1.8 0% 0% 0% 4% 13% 30% 48% 63% 83% 83% 0 54% 96% 100% 96% 87% 70% 52% 37% 17% 17% 46% 4% 0% 0% 0% 0% 0% 0% 0% 0% 0. Alpha Alpha distribution (5%) + 0 61% 30% 17% 17% 17% 17% 41% 33% 74% 74% 0% 0% 0% 0% 2% 0% 0% 0% 0% 0% Beta distribution (5%) Market + -0.

bearish and neutral months since these rules would obviously not match the ones used by fund managers for their market timing decisions. During the up period. Those trends are sufficiently strong to allow us to consider the whole sub-periods as.1 Market exposure Panel A of Table 50 reports the result of the beta analysis for funds considered in indices and for funds classified on the basis of their performance over the period studied in order to determine if best or worst performing funds have been more or less exposed to the equity market over the bullish and bearish period studied. This month corresponds to the maximum observed value of the Russel 3000 (that contains 95% of the capitalisation of the American equity market) 500 Index that attained a value of 858. the monthly index return was positive in 39% of the months (12 out of 34) and the average yearly return was -16.VIII Sub-period analysis Since our analysis period covers a bull and a bear period.9%. bullish and bearish without having to use a complex rule to separate bullish. 317 . 2000.4%. We chose the bull and a bear market as defined by Capocci. respectively. Corhay and Hübner (2004). During the down period. the monthly index return was positive in 70% of the months (52 out of 74) with an average yearly return of 19. The cutting point chosen for the identification of the up and down periods has been set at March 2000. 8.48 during the session of March 24. we will perform the same analysis as in section VI in dividing the analysis period in 2 sub-periods.

74 We use the term absolute to underlie the fact that these funds were not significantly exposed to the market. which means that these figures have to be considered on an absolute basis rather than in comparison with the index considered. The second part of Panel A of Table 50 reports the deciles’ regression analysis for the bull period. This means that these funds do not offer absolute74 significantly positive returns over this period. This result is quite astonishing because it seems to indicate that market neutral funds have been able to outperform the market index over the period studied. On the other side. Mortgage backed funds’ results even indicate that the funds were negatively (but not significantly) exposed to the market and that they create absolute alpha. the R²adj are very low indicating that the market factor cannot explain the returns. It indicates the same pattern as over the whole period with higher absolute numbers with the same level of significance indicating that all but the worst performing funds were able to create significantly positive alpha over the period studied with no exposure to the market as indicated by the market and the R²adj columns. the funds reported in the other deciles offered significantly positive absolute returns over the period studied. Moreover. The market and R²adj columns however indicate that this significant alpha was created totally independently of the market direction because all market factors are not significantly different from zero. this result confirms the non-exposure to the market for market neutral funds but the ability to create absolute performance over the bullish period studied.The results obtained for the bullish period (see Panel A of Table 50) indicates that the alphas reported are higher than those obtained over the whole period. More precisely. 318 .

This result is inverse from the one obtained for the bull market period and is logical since managers face more difficulties in finding investment opportunities when stocks’ prices (good and bad one) were decreasing. On the other side. The market exposure can explain a bigger part of the alphas generated since the R²adj are much higher than the one obtained over the bull market. This result is particularly important. the bullish period. Standard deviations are also in almost all cases higher. In the bull market period the market exposure could not explain the alpha created by the funds leading to significantly positive alphas with extremely low R²adj. in the second case. Alphas indicate the same pattern. the beta reported are significantly positive in all cases over this time period indicating that the funds were significantly exposed to the market over this period. the funds significantly outperformed the market with a significant exposure to it and comparatively high R²adj. On an absolute term however. 319 . The reasons explaining this result are however completely different from the one explaining the alpha in the bull market.Panel B of Table 50 shows that the average return for every strategy (the exception being the no sub-strategy category) was higher over the whole period than over the bear market period studied. the alphas generated by the market neutral funds are significantly positive. Interestingly.

39% 2.39% 1.01 -0.16% 0.02 0.41% 0.05 0.04 1.52% 1.01 0.12 -0.24% 1.60% 1.03 -0.23% 1.08% 1.09 -0.62% 2.03 -0.73% 2.34% 1.60% 0.74% *** -0.46% 1.41% 0.48% *** -0.04% *** *** *** *** *** *** *** *** -0.87% 0.06 -0.41% 1.12 0 -0.94% 1.04 -0.01% 1.27% *** *** *** *** *** *** *** *** 0.58% 2.No substrategy Decile 1 Decile 2 Decile 3 Decile 4 Decile 5 Decile 6 Decile 7 Decile 8 Decile 9 Decile 10 1.40% *** 1.03 -0.D.Long/short funds .64% *** 1.03 0.02 -0.04 -0.96% 1.07% 0.03 0.33% -0.03 -0.05 -0.03 320 .02 -0.11% 1.03 -0.03 -0.01% 1.03 -0.Mortgage backed funds .60% 4.53% 4.42% 0.49% *** 1.06 1.03 1.52% 1.36% 0.04% *** -0.09 0.54% 0.05 -0.85% 1.03 0.72% *** 1.52% *** -0. Alpha Market R²adj Market neutral funds .Table 50: Indices and deciles sub-period analysis Panel A: 9/98-03/00 Average return S.Arbitrage funds .65% *** 0.06 -0.18% 1.

70% 0.88% 1. the market exposure (the beta) and the R²adj.17 0.83% 0.15 0.78% 0.67 0.06 0.07 0.43 0. Alpha Market R²adj 0.11 0 0.05 0.21 0.17% 1.06% 1. ** Significant at the 5% level and * Significant at the 10% level.No substrategy Decile 1 Decile 2 Decile 3 Decile 4 Decile 5 Decile 6 Decile 7 Decile 8 Decile 9 Decile 10 0.15% 0.Mortgage backed funds . Panel A presents the bullish period (09:1998-03:2000).11 *** 0.Arbitrage funds .37% 0.29 0.92% 0. the alpha.84% *** 0.D).31% 3.06% *** * 0.Long/short funds .82% *** 0.26% 2.69% -0.03% 0.07 0.99% 0.D.75% *** 0.11 *** *** *** *** *** *** 0.Panel B: 4/0-12/02 Average return Market neutral funds . t-stats are heteroskedasticity consistent.65% *** 1.35 0. Each Panel presents the average return.09 *** 0. 321 .02 This table reports the market exposure analysis over sub-periods.22 0.97% 1.06 ** * 0.06% 0.13 0.06 0.1 0.71% *** S.67% 1.61% 0.97% *** *** 0.65% 0.43 0.85% *** 0.52% 0. *** Significant at the 1% level.42 -0.67% 0.03 *** ** *** *** *** *** *** *** 0.40% 0.14 0.83% 1.06% 1.48% 0.58% 0.87% 0.77% 0.95% 0.54% 0. the standard deviation (S.59% -0.08% 0.27 0.07 *** 0. including 398 survived funds and 236 dissolved funds as of December 2002.81% 0.08 *** 0.61% 0.42% *** *** *** *** *** *** *** *** 0. Panel B presents the bearish period (04:2000-12:2002).03% -0. Panel B presents the same information when funds are classified in deciles on the basis of their performance over the period studied. Our MAR/CISDM database contains 634 individual market neutral funds.08 0.80% 1.

Interestingly. An interesting point to stress here is that these results are not exact to performing funds that could create alpha with no exposure to the market. The R²adj are low. this strong bull market has a negative (but not significant) impact on the performance. These funds create positive absolute returns (while remaining market neutral funds) whereas worst performing market neutral funds create alpha compared to the equity market (which was strongly negative over this period). no deciles but decile 9 are exposed to the market. Table 51 reports the same results based on the decile classification from Carhart’s (1997). This result interestingly stresses that most bad performing market neutral funds out-perform the equity market without offering significantly positive performance but that the best performing funds significantly out-performed the equity market and offered significantly positive returns over the bearish market after March 2000. No alpha is significantly positive when the market impact is taken into account. The average return and standard deviation column indicates that the top and bottom decile funds.The second part of Panel B indicates exactly the same pattern. lower alpha for most decile in absolute term significant (in statistical term but low in absolute term) exposure to the equity market for the same deciles and relatively high R²adj. Moreover. This result is confirmed by the regression results. 322 . All these results suggest that low volatility market neutral funds create value on an absolute basis (without exposure to the equity market) but that almost none of them benefit from the market sharp increase. the most volatile one has not offered significantly positive return over the period covered. Most deciles are even negatively (but not significantly) exposed to the equity market.

the other deciles offer a significantly positive alpha. The spread between decile 1 and decile 2 is significantly negative confirming this remark. top funds however offered better (but not significantly positive) alphas than bad ones. the average return is lower than the alpha. 323 . Over the bearish period. the exposure of the fund to the market has unable the manager to offer significantly positive alpha compared to the index whereas its average return was not different from zero. confirming that volatile funds cannot create persistent value over time. In absolute term. Bottom and particularly top deciles had the highest volatility confirming our previous results. This result can be explained by the fact that funds in this decile were significantly exposed to the market. no decile but the one containing last year worst performing funds have been significantly exposed to the market. the alpha of this decile is significantly positive whereas its average return was not.Panel B of Table 51 indicates that even if the average return of the previous year top and worst performing funds are not significantly different from zero. Interestingly. These results are stronger than before. In fact. Since the market went down over the period studied.

02% -0.15% -0.01 9-10 spread 0.02% -0.1 324 .02% ** 1. Alpha Market R²adj Decile 1 0.17% 0.05 -0.02 -0.18% 0.74% 0.70% 0.D.01 -0.38% 2.26% -0.01 -0.15% -0.41% -0.14% 0.00% 0.06 1-2 spread 0.02 -0.88% 0.00% 2.16% 1.03 -0.04 0 Decile 6 0.56% * 1.11 Decile 10 1.46% ** 0.06 1-10 spread 1.50% 0.59% ** 1.03 -0.09% 0.03 Decile 4 0.86% 0.40% 1.05 Decile 8 0.44% * 1.Table 51: Lagged 12-month decile sub-period analysis Panel A: 9/98-03/00 Average return S.05 Decile 3 0.11% -0.35% ** 0.40% * 3.12% -0.05 -0.28% ** 0.03 -0.01 Decile 7 0.12% 0.11 ** 0.57% 0 -0.02 Decile 5 0.57% 0.07% -0.43% *** 0.93% 0.28% 0.06 Decile 9 1.14% 0.11 0.15% -0.05 Decile 2 0.

01 -0.79% 0.32% 0.67% *** 0.96% *** 0.13% 3. *** Significant at the 1% level.02 Decile 4 0.72% *** 0. the alpha.53% ** 1. the market exposure (the beta) and the R²adj.03 Decile 6 0.83% 1.53% 2. Panel B presents the bearish period (04:2000-12:2002).67% *** 0.03 1-2 spread -0.31% 0.93% *** 0. 325 .67% 2.44% 0.00 Decile 8 0.83% *** 0.03 Decile 3 0. It reports the average return.20% *** 0.23% 0.33% *** 0.73% *** 0. including 398 survived funds and 236 dissolved funds as of December 2002.D. Panel A presents the bullish period (09:1998-03:2000).03 Decile 5 0.60% 0.13 0.71% 0.03 The table reports the market exposure using deciles constructed on the lagged-12 month performance (see Carhart 1997).65% 2. t-stats are heteroskedasticity consistent.62% 0.04 0.81% 0.01 -0.50% 1.12% 2.53% *** 0.05% 0. ** Significant at the 5% level and * Significant at the 10% level. Our MAR/CISDM database contains 634 individual market neutral funds. the standard deviation (S.09% -0.11 Decile 2 0.94% * 0.D).01 -0.56% *** 0.03 -0.01 Decile 9 1.Panel B: 04/00-12/02 Average return S.12 9-10 spread -0.02 Decile 7 0.02 -0.01 -0.37% *** 0.45% *** 0.09 ** 0.64% 0.11 0.03 -0.76% *** 0.03 Decile 10 0.01 1-10 spread -0.82% *** 0.09 ** 0. Alpha Market R²adj Decile 1 0.78% 0.77% *** 0.92% *** 0 -0.

Whereas it was on the average positive over the whole period with one third of the beta being significantly positive. 13 mortgage backed funds and 1 no sub-strategy fund) from the database and leave us with a total of 262 funds (146 arbitrage funds. 326 . 162 long/short funds and 14 mortgage backed securities funds) for the 4/00-12/02 period analysis. 29 mortgage backed securities funds and 2 no sub-strategy funds) for the 9/98-3/00 period analysis. This lead us to remove 200 funds (90 arbitrage funds. This result indicates that individual market neutral funds are not exposed to the market in bull markets.2 Individual funds Table 52 reports the sub-period analysis obtained for individual funds. Panel B’s average returns are much lower but the average returns are spread in the same way as before for market neutral funds arbitrage funds and long/short funds. 96 long/short funds. This lead remove 125 funds (60 arbitrage funds. The market exposure over the bull market are more spread with up to 31% significantly positive alphas (arbitrage funds) and there are up to 18% of significantly negative alpha’s 75 We require all funds to have returns for the whole period between September 1998 and March 2000.8. the average return obtained for the bull market are higher than those obtained over the whole period (except for mortgage backed funds) and a higher proportion of theses returns are significantly positive. more average returns are significantly positive over the bear period. The alpha’s obtained are also lower and there are some differences in the proportion of significantly positive alphas and alphas not significantly different from zero for arbitrage funds (more alphas are not significantly different from zero over the bear period) and long/short funds (more alphas are significantly different from zero over the bear period).75 As one could expect. The alpha column and distribution reports the same pattern. 138 long/short funds. The market exposure however indicates differences. 52 long/short funds. 11 mortgage backed funds and 2 no sub-strategy fund) from the database and leave us with a total of 337 funds (147 arbitrage funds. The R²adj obtained confirms this result. almost all betas are not significantly different from zero over the bull market. For mortgage backed funds.

On the other side when the market was down they were able to create positive alpha but are more exposed to the market.for some strategy (mortgage backed funds for example) indicating that these funds did not short the market. market neutral managers considered as a whole seemed more to be exposed to the market. on the long on or the short side. 327 . Interestingly. in bear market. regarding market exposure. On the other side. These results indicate that Arbitrage funds and mortgaged backed funds seemed to perform better in bull market (with more significantly positive alpha) without being exposed to the market. market neutral funds were significantly and negatively exposed to the market only during bear market.

03 .55% 84% 16% 0% -0.52% 77% 23% 0% 2.76% 1.01 3% 94% 3% 0 .02 .10% 65% 35% 0% -0. Alpha Alpha distribution (5%) + 0 - Beta distribution (5%) Market + 0 R²adj Market neutral funds 1.No sub-strategy 0.38% 1.87% 21% 79% 0% -0.Arbitrage funds 1.02 6% 94% 0% 0.64% 100% 0% 0% 1.16% 0.01 .83% 38% 62% 0% 4.Mortgage backed funds 0.Long/short funds 1.81% 23% 77% 0% 4.57% 58% 42% 0% 3.D.Table 52: Sub-period individual funds results Panel A: 9/98-03/00 Average return Return distribution (5%) + 0 - S.55% 1.74% 100% 0% 0% -0.04 0% 100% 0% -0.11 0% 100% 0% 0.02 .12% 1.03 6% 92% 2% 0.65% 61% 38% 0% -0.

the beta distribution (+ are significantly positive.55% 84% 16% 0% -0.12% 1. the standard deviation (S.87% 21% 79% 0% -0.No sub-strategy 0. t-stats are heteroskedasticity consistent. including 398 survived funds and 236 dissolved funds as of December 2002.01 . Panel B presents the bearish period (04:2000-12:2002). Alpha Alpha distribution (5%) + 0 - Beta distribution (5%) Market + 0 R²adj Market neutral funds 1.D. the return distribution at the 5% level.03 The table reports the individual fund market exposure.04 0% 100% 0% -0.Mortgage backed funds 0.16% 0.03 6% 92% 2% 0.Panel B: 4/00-12/02 Average return Return distribution (5%) + 0 - S. 0 not significantly different from 0 and – significantly negative) and the R²adj.Long/short funds 1.11 0% 100% 0% 0. Our MAR/CISDM database contains 634 individual market neutral funds. .01 3% 94% 3% 0 .52% 77% 23% 0% 2.Arbitrage funds 1. the alpha. 0 not significantly different from 0 and – significantly negative).57% 58% 42% 0% 3. (+ are significantly positive. Panel A presents the bullish period (09:1998-03:2000).02 . the market exposure (the beta).55% 1.02 . alpha distribution at the 5% level (+ are significantly positive.74% 100% 0% 0% -0.83% 38% 62% 0% 4.10% 65% 35% 0% -0.65% 61% 38% 0% -0.81% 23% 77% 0% 4.38% 1.64% 100% 0% 0% 1.02 6% 94% 0% 0.76% 1. It reports the average return. 0 not significantly different from 0 and – significantly negative).D).

The standard deviation is close to those obtained over the whole period and higher than those obtained over the bull market period. All in one these results indicate that market neutral funds could offer absolute alpha during the bull market without being exposed to the market. Panel A reports higher average returns for the bull market period compared to the whole period but fewer of these returns are significantly positive.8. The R²adj obtained are also relatively high for top and bottom deciles indicating that fund’s market exposure help explaining returns for the highest negatively and positively exposed funds. The standard deviations of the returns are lower for the bull market compared to the global period. The beta distribution confirms this result. The regression results report higher alphas than previously for all deciles but fewer of them are significantly positive. market neutral fund’s beta are higher in absolute term. The R²adj are close to zero for all deciles except the high beta decile. As before and by construction the exposure to the market increases but the absolute level over the bull market are lower than before indicating that market neutral funds are less exposed to the market during bull market than during a whole investment cycle. Panel B of Table 53 indicates that over the bear market period covered. 330 . The market exposure increases logically by construction with extreme’s beta value higher (respectively lower) for high (respectively low) beta funds. market neutral funds offered higher average returns compared to the whole period and to the bull market period but fewer returns are significantly positive compared to the whole period and less strict compared to the bull market. Fewer individual betas are significantly positive over the bull market conditions compared to the whole period studied. This confirms our previous results that in the bear market period.3 Ex-post beta sub-period analysis Table 53 reports the beta ex-post analysis over the 2 sub-periods considered. Alphas are lower than over the whole period and fewer of them are significantly positive.

05 -0.02 0 -0.20% 1.68% 15% 27% 50% 62% 96% 85% 69% 69% 50% 50% 9.06 0.05 -0.60% 1.02 0.10% 5.17 0.01% 1.13 -0.34% 1.02 -0.42% 1.40% 0.23 -0.16 R²adj Decile 1 Decile 2 Decile 3 Decile 4 Decile 5 Decile 6 Decile 7 Decile 8 Decile 9 Decile 10 1.01 0.16% 1.50% 2.30% 3.44% 19% 50% 81% 69% 100% 81% 69% 69% 50% 23% .96% 1.47% 1.99% 1.04 0.24% 1.66% 1.69% 1.80% 3.D.01 0.Table 53: Ex-post beta sub-period analysis Panel A: 9/98-03/00 Average return Return distribution (5%) + 0 85% 73% 50% 38% 4% 15% 31% 27% 50% 50% 0% 0% 0% 0% 0% 0% 0% 4% 0% 0% S.36% 1.08 -0.54 -0.60% 2.37% 1.02 0.03% 2.66% 1.5 0% 0% 0% 0% 0% 0% 0% 4% 12% 46% 0 96% 100% 92% 96% 100% 100% 100% 96% 88% 54% 4% 0% 8% 4% 0% 0% 0% 0% 0% 0% 0.30% 2.40% 1.90% 5.89% 2. Alpha Alpha distribution (5%) + 0 81% 50% 19% 31% 0% 19% 31% 27% 50% 77% 0% 0% 0% 0% 0% 0% 0% 4% 0% 0% Beta distribution (5%) Market + -0.20% 3.01 0.04 -0.32% 1.

37 -0.03 -0.96% 0.33% -0.13 0. the market exposure (the beta).Panel B: 04/00-12/02 Average return Return distribution (5%) + 0 64% 39% 9% 9% 9% 24% 24% 47% 94% 94% 0% 0% 0% 0% 0% 0% 0% 0% 3% 6% S.36 R²adj Decile 1 Decile 2 Decile 3 Decile 4 Decile 5 Decile 6 Decile 7 Decile 8 Decile 9 Decile 10 1.00% 1.73% 0.70% 0. .86% 0.40% 1.95% 0.D).D.87% 0.75% 0.17 0.98% 0.10% 0. t-stats are heteroskedasticity consistent.55% 36% 61% 91% 91% 91% 76% 76% 53% 3% 0% 5.01 0.93% 0. 0 not significantly different from 0 and – significantly negative) and the R²adj.00% 4. the beta distribution (+ are significantly positive.20% 1.88 0% 0% 0% 0% 0% 9% 50% 71% 47% 88% 0 39% 76% 82% 100% 100% 91% 50% 29% 53% 12% 61% 24% 18% 0% 0% 0% 0% 0% 0% 0% 0.73% 0. the return distribution at the 5% level.13 0. the alpha. 0 not significantly different from 0 and – significantly negative). (+ are significantly positive.04% 0. Panel B presents the bearish period (04:2000-12:2002).06 0.28 0.11 -0.07 0. Panel A presents the bullish period (09:1998-03:2000).88% 0.76% 27% 33% 85% 85% 91% 76% 74% 56% 29% 24% The table reports the individual fund market exposure.10% 2.50% 1.02 0 0. It reports the average return.84% 0.78% 1.63% 0. 0 not significantly different from 0 and – significantly negative).80% 2. Alpha Alpha distribution (5%) + 0 73% 67% 15% 15% 9% 24% 26% 44% 68% 70% 0% 0% 0% 0% 0% 0% 0% 0% 3% 6% Beta distribution (5%) Market + -0.79% 0.03 0.40% 8.02 -0. the standard deviation (S. including 398 survived funds and 236 dissolved funds as of December 2002.06 0.10% 1.50% 0. alpha distribution at the 5% level (+ are significantly positive.14 0. Our MAR/CISDM database contains 634 individual market neutral funds.04 0 0.

Decile analysis indicates that top and worst performing funds (over the whole period) have the highest market exposure. The individual fund analysis results indicate that on the average. we will focus on the literature.IX Conclusion This study focuses on the neutrality of market neutral funds. In all cases. The core of the study is based on a methodology that use classical exposure measures like the beta in an original way. Then. on the descriptive statistics. the interest of the study. attrition and birth rates and on the presence of survivorship bias in the data. In the first descriptive part of the analysis. It has the objective to determine if so-called market neutral funds are really not exposed to the equity market. We will analyse this topic over a complete market cycle going from January 1993 to December 2002 and over bull and bear market conditions covering respectively the September 1998 to March 2002 and the April 2000 to December 2002 periods. one third of the individual funds were significantly positively exposed to the market while two third of the alphas are significantly positive (especially for worst and best performing funds that also offer the more volatile returns). The results obtained using sub-strategy indices indicate the beta obtained are low on an absolute term but significantly positive. the ex-post beta analysis indicates that negatively exposed and strongly positive exposed funds are more volatile than market neutral 333 . We will perform the analysis using strategy and sub-strategy indices and using individual funds data in order to determine if we obtained the same results. market factors are significantly positive but they do not explain a major part of the alphas that are relatively the same as the average return and the calculations usually give low R²adj. The analysis of the decile constructed in the previous year’s performance as a tool of classification suggests that the exposure of market neutral funds to the equity market is not clear.

the funds significantly exposed could bias the results. Over the bearish period. on a strategy and sub-strategy basis. This result could be explained by 2 reasons. Thirdly. Secondly. since the bulk of the funds are not significantly exposed to the market but one third of the funds are and only five percent of the funds are negatively significantly exposed to the market. but most market neutral funds are not significantly exposed to the market. most bad performing market neutral funds out-perform the equity market without being significantly exposed to the market but best performing funds significantly out-performed the equity market and offered significantly positive returns over the bearish market after March 2000. First. all but the best performing deciles have been significantly exposed to the market but they all. The decile classification based on previous year’s performance interestingly add to these results that they were very few funds significantly positively exposed to the market during the bullish period and that there was no clear pattern in the bear period results for these calculations. Secondly. except the best performing funds create significant alpha.funds. This first reason is in line with the results obtained in the correlation analysis and stresses the point that it is of major importance to consider individual funds in empirical analysis because results based on indices could be biased. We had to perform an analysis at the individual fund level to find out this result because market neutral index analysis lead to more controversial results. First. Some funds are strongly positively or negatively exposed to the market over the period studied. the aggregation of funds in indices leads to an increase in the exposure to the equity market. arbitrage funds and mortgaged-backed funds perform better in bull market 334 . The sub-period analysis also reports very interesting results. Middle decile funds returns are the only real market neutral funds. This study stresses the importance of considering individual funds when performing market neutral empirical analysis. over the bullish period no index or decile has been significantly exposed to the market.

regarding market exposure.without being exposed to the market. but that they tend to be more exposed during bear market than during bull market without being negatively impacted. some market neutral funds are significantly negatively exposed to the market only during bear market. On the other side when the market is down they are able to create positive alpha but are exposed to the market. 335 . Our analysis leads to the conclusion that most market neutral funds are not significantly exposed to the equity market. Interestingly. market neutral managers considered as a whole tend more to be exposed to the market. in bear market. on the long or the short side. On the other side.

.

Part 3: Hedge Funds as Diversification Tools .

Daniel. Working paper.J. Capocci. Frédéric Duquenne and Georges Hübner. Maastricht University. and Luxembourg School of Finance. Diversifying using Hedge Funds: A Utility-Based Approach.Diversifying using Hedge Funds: A UtilityBased Approach Daniel P. 2006. University of Luxembourg. CAPOCCI HEC-ULG Management School – University of Liège (Belgium) Frédéric Duquenne HEC-ULG Management School – University of Liège (Belgium) Georges Hübner HEC-University of Liège. Limburg Institute of Financial Economics. HEC-ULG Management School .

We apply this new methodology to determine if the inclusion of hedge funds in a classical portfolio of equity and bond mutual funds allows investors to reach a significantly higher level of satisfaction when not only volatility but also skewness and kurtosis are included in the portfolio. various levels of allocation to the risky asset and by separating different kind of hedge fund investments. This new efficient frontier is based on the Taylor’s decomposition of Bell’s (1988.Diversifying using Hedge Funds: A Utility-Based Approach Abstract This paper develops an adapted efficient frontier that integrates higher moments in the measurement of risk. 1995) utility function. 339 . Estimations are performed for various levels of risk aversion.

. Brown et al. Brealy and Kaplanis. The first one. 2004. 2001. 2001. 2005). 2001. encompasses papers that focus on hedge fund performance (see for example Ackermann et al. The second global category includes articles that analyse and describe hedge funds investment style and explain these features with style models (see for example Fung and Hsieh. Liang 2001.Diversifying using Hedge Funds: A Utility-Based Approach Introduction Hedge funds have been widely studied since 1997. Liang.76 Hedge fund studies can be classified in four global categories. 1997. From performance to risk analysis. Liang 2003). Capocci and Hübner. 2003. Almost all areas of the industry are under research. A third stream of the literature focuses on the correlation of hedge funds with other investment products and analyse the power of diversification of hedge funds (see for 76 “Angels” because hedge funds have been seen as the new perfect tool create returns when the traditional markets can no more offer attractive returns. 1999. 2001. 340 . Agarwal and Naik. Edwards and Caglayan. Malkiel and Saha. 2004. Amin and Kat. “Dark” because many investors have invested in these products without being aware of the underlying strategy or the real risk involved with the strategy applied. which is by large the richest in quantity.. 2000. Brown and Goetzmann. Agarwal and Naik. 2001. the interest in hedge fund analysis is growing and the literature more and more explain the specificities of these “dark angels”. from style analysis to their use as a diversification tool.

This category includes the “other studies” focusing on the risks involved in hedge fund investing (see Jorion. The underlying idea of our analysis is the same as the spanning tests developed by Huberman and Kandel (1987). or the difference between the indices available (see for example Amenc and Martellini. 2001).example Fung and Hsieh. 1999. The main differences introduced by our methodology are twofold. 2002a). Our final objective is to determine if the integration of hedge funds in a diversified portfolio of stocks and bonds helps the investor to increase significantly his utility for a reasonable specification. 2000. Schneeweis and Spurgin. 2000 or Lo. 2005). The paper is organised as following. 1997. the presence of bias in the data (see for example Liang. the original spanning is based on the mean and the variance and on the hypothesis of quadratic preferences whereas ours is based on Bell’s (1988. 2000 and Malkiel and Saha. 341 . Section III reports the results and Section IV concludes the paper. Liang. First. In section I we summarize the use of utility functions and introduce the characteristics of the Bell (1985) utility function. The methodology used as well as the database are described in Section II. 2002 and Fung and Hsieh. other studies analyse miscellaneous aspects of the hedge fund industry. that combines a linear and an exponential form. we not only look at the impact of inserting a particular security in a portfolio’s efficient frontier but we include a single risk specification including skewness and kurtosis in the analysis. It aims at determining if an investor that combines hedge funds with traditional investments really improves its satisfaction by combining the two families of investments. 2000). Finally. Fung and Hsieh. 1998. The current study belongs to the third stream. 1995) utility function. Agarwal and Naik. Second. but study a wide range of risk measures corresponding to different investor profiles.

In actual applications. These authors conclude that manipulation-proof performance measures can be completely characterized as the weighted average of a utility-like function. 1997). which is exactly the focus of our study.I Utility functions and Spanning 1.1 Utility functions and the Bell function Several strudies have stressed the limitations of traditional performance measures focusing on risk (as measured by the standard deviation and the return). Levy and Markowitz (1979) justify the practice of using mean-variance 342 . returns are typically not normally distributed and utility functions have a higher order than quadratic. portfolio selection theory examines how a rational investor behaves when choosing his investments in a framework of uncertainty about his futures outcomes. (2006) show that statistics can yield attractive statistics in terms of classical measures like the Jensen’s alpha using a simple rebalancing strategy. etc. It offers a lot of applications in various disciplines including finance. Many authors have devoted their time searching for alternative ways of modelling an investor’s preference and deriving the composition of his ideal portfolio. More recently Goetzmann et al. In finance. Jagannathan and Korajczyk (1986) demonstrate that it is possible for fund managers to construct portfolios that show artificial timing ability when no true timing ability exists while investing in options or levered securities. Based on its mean-variance approach. and particularly in the case of hedge funds. The utility function concept is widely used to model the process of decision making. microeconomics and operational research. the CAPM is still used as a benchmark “…but mean-variance analysis is only appropriate when returns are normally distributed or investors’ preferences are quadratic (Fund and Hsieh.

There are various classes of utility functions. a decision maker can be thought of as comparing alternatives based on his expected utility. The utility function concept used in the framework among uncertain investment outcomes is sustained by a strong theoretical justification.analysis by showing that mean-variance analysis can be regarded as a second-order Taylor series approximation of standard utility functions. Each has specific properties. power and quadratic specifications. The most common are the exponential. A usable utility function has must satisfy at least some basic desirable properties: (a) Non-satiety with respect to wealth: an increase in wealth always increases the utility. drawbacks and advantages (see Spanier and Oldman. He will always give more utility to a certain gain than to a random result with the same expected return. logarithmic. Von Neuman and Morgenstern (1947) proved that. The standard mean variance utility function of Markowitz (1959) and Sharpe (1970) can be viewed as an approximation to the more basic von Neuman and Morgenstern utility function and more particularly to the isoelastic family of utility functions. This property involves that the second-order derivative to be negative. 1987 for a detailed presentation of more than 400 utility functions). (c) Risk assets are not inferior goods. when we accept some hypothesis about the concept of rational choice. 343 . The first derivative of the utility function of wealth should positive (b) Risk aversion: the speculator is not a gambler. He invests despite the underlying risks and always requires a compensation.

thus an investor with a non-constant RRA cannot be characterized. The main issue with the exponential utility function is that the ARA77 is constant (and that its range of variation is the largest possible. ARA is the ratio of the second-order derivative to the first order one.e.Classical utility functions like the exponential. See Arrow (1971) for more details. and the risk R. RRA for Relative Risk Aversion. [0. of an alternative x. the quadratic utility function involves satiety and growing risk aversion. 344 . Friend and Blume (1975) and Morin and Suarez (1983) find empirical support for decreasing RRA. The logarithmic utility function and the power utility function have a constant RRA and are isoelastic. but in decreasing proportional amounts. power and quadratic do not offer all these desirable properties. i. Finally. R( ~ ). Considering these elements. Bell (1988. logarithmic. +∞[) implying that investors invest constant amounts in risky assets as their wealth increases. E ⎣U (W0 . W0 ] x x x ( 13 ) 77 ARA stands for Absolute Risk Aversion. Cohn et al. ~ ) ⎦ = f [r ( ~ ). This is not the case with Bell’s function. In other words. (1975). 1995) introduces a new kind of utility function that displays appealing characteristics and permits easy interpretation of alternatives in terms of risk and return. it seems that none of the popular utility functions listed above has all the desirable properties. Bell’s utility function is such that its expected value can be expressed as a function of the wealth increase r.

aka linex functions: U (W ) = W − be − cW (14) with b. ~ ) ⎦ = E W0 + ~ − e − aW e − ax = W(0 ) + r ( ~ ) − be − cW e − cr ( x ) − be − c [W0 + r ( x ) ] R( ~ ) x x x x [ ] (15) ~ where the measure of wealth increase r (x ) is the expected return of the alternative E (x ) .Bell’s function is a combination of a linear and an exponential utility function and more commonly known as “linear plus exponential”. c ≥ 0 ~ The expected utility in the case of an additive random wealth x is then equal to ~ ~ ~ E ⎣U (W0 . ~ 345 .

as a weight assigned to risk by the investor in his utility function. Bell (1988. parameter c constitutes relative riskiness and b aversiveness to that riskiness. 1995) shows that the linear plus exponential utility function has the feature that expected utility can be written as a function of mean final wealth and risk of the final wealth distribution. On the other side. parameter b does control the degree of aversion to risk. and particularly on the tail of the distribution: investors with a high c place more emphasis on the possibility of bad outcomes. which is specific to the investor.The measure of risk can be isolated from this expression: ~ 1 R ( ~ ) = ln E (e −c ( x − x ) ) x c [ ] (16) The measure of risk incorporates one parameter c. Thus. Parameter c depends on the distribution of return only. 346 .

The function provides a new form of second order approximation that has very satisfying properties. they will be judged equivalent at all levels of wealth. (2) (1) (2) (1) (2) It is a proper utility function as the successive derivatives alternative in sign . It is a non-polynomial function as it includes an exponential. The RRA (relative risk aversion) is decreasing then increasing with respect to W. 347 . (1) It is an increasing and concave function of wealth. With this function. if two alternatives are judged to be equivalent at two different levels of wealth. (1) Alternatives can be compared using a function of risk and return. It has a decreasing ARA (absolute risk aversion) and its range of variation is the largest possible. (2) The two parameters b and c leaves flexibility to match personal attitudes.

When investors condition their investment choices on the first two moments of the distribution of returns. This issue is addressed by the notion of spanning introduced by Huberman and Kandel (1987). One should estimate the linear relation that exists between the returns of the existing portfolio and of the additional securities considered. the underlying hypothesis is similar. There are two approaches to test for spanning: the regression approach and the stochastic discount factor approach.1. we want to determine whether or not the addition of a security (in our case hedge funds) enables investors to access a significantly superior efficient frontier. Under their approach. an investor with a quadratic utility function is indifferent to adding securities in his portfolio if the minimum-variance frontier of his initial portfolio coincides with the minimum variance portfolio that includes the new securities. “Can investors benefit from investing a new set of N assets conditional on K assets?” The spanning properties of different classes of financial assets have been studied by several authors. Mean-variance spanning is attained if the new frontier does not significantly differ from the old one. they want to know whether adding risky assets will improve the minimum-variance frontier. Ferson et al. (2001) test for spanning with futures contracts and non-traded assets and Kan and Zhou (2001) study tests of meanvariance spanning in detail. Huberman and Kandel (1987) propose a multivariate regression-based approach to test of this hypothesis. (1993) generalise the test and integrate the issue of heteroskedasticity in the testing while De Roon et al. The null hypothesis of absence of additional diversification with the additional portfolio (spanning) involves that 348 . The real question to be asked is the following: “Can investors maximise utility by holding just a smaller set of K assets conditional on the existence of N+K assets?” or said differently.2 Spanning Altough our methodology is not based on spanning.

and 349 . 1993 by considering non-normality and heteroskedasticity. from a meanvariance perspective. Hansen’s (1982) GMM is the common viable alternative that relies on the moment conditions of the model. (1993) and Kan and Zhou (2001). They consider non-normality and heteroskedasticity. When the residuals exhibit conditional heteroskedasticity. the three classical three tests will no longer be asymptotically Chi² distributed under the null hypothesis. a version of the two-step Wald test that uses a finite sample corrected estimate of the variance of the two-step GMM estimator. These results have been stressed by Ferson et al. Newey and West (1987) show that the GMM version of the likelihood ratio test and the Lagrange multiplier test have exactly the same form as the Wald test. Note all three tests are increasing transformation of each other.. the LM test. the minimum-variance frontier that includes the added securities does not statistically differ. two bootstrapped versions of these Wald tests.the intercept of the regression and the slope coefficient are not significantly different from 0 and 1 repsectively. The null hypothesis can be tested in three ways. If the conditions are both satisfied. Generalised methods of moments versions of the tests have been developed by Ferson et al. Huberman and Kandel (1987) provide a likelihood ratio test of spanning and derive its finite sample distribution under normality assumptions. In this case. The Wald test and the Lagrange multiplier test are two alternative asymptotic tests described in detail by Kan and Zhou (2001). from the original portfolio the investors including new securities cannot increase their utility by adding the new securities considered. These conditions imply that the regression line passes through the origin and that the sum of the coefficients of the independent variables equal 1 for each estimation. Bond and Windmeijer (2005) compare the finite sample performance of a range of tests of linear restrictions for linear panel data models estimated using the generalised method of moments (including standard asymptotic Wald tests based on one-step and two-step GMM estimators.

350 . while the bootstrapped one-step Wald test. The second approach. If the rejection is due to the first test. 2001). the stochastic discount factor approach has been developed by DeSantis (1993) and Bekaert and Urias (1996). but the null hypothesis remains equivalent. The corrected two-step Wald test performs similarly to the standard one-step Wald test. The null hypothesis of Huberman and Kandel (1987) can be proved (see Kan and Zhou. and a simple criterion-difference test can provide more reliable finite sample inference in some cases. the LM test.three criterion-based tests that have recently been proposed (see Kan and Zho. Kan and Zhou (2001) propose a step-down procedure to separate the two approaches to test for spanning (regression approach and stochastic discount factor approach). Bekaert and Urias (1996) exploit the duality of Hasen-Jagannathan (1991) bound and the mean-variance frontier. Following these results and Kan and Zho (2001) we will focus on the GMM Wald test. first one has to test whether the intercept is equal to zero before testing if the sum regression coefficients is equal to one conditional on the intercept being equal to zero. one can consider the two global minimum variance portfolios are very different. 2001) and is equivalent to the null of Bekaert and Urias (1996). If the rejection is due to the second test. then the two tangency portfolios are very different. They suggest that the stochastic discount factor based spanning tests using GMM have equivalent hypotheses as Huberman and Kandel (1987) and they project a stochastic discount factor and test whether the N-test assets can explain the variance of the stochastic discount factor. The advantages are 1) cause of the rejection is known and 2) the approach gives the flexibility to assign different significant levels to each test. Under this methodology. DeSantis (1993) also projects the stochastic discount factor on the gross return and the expected return does not appear as a parameter in the DeSantis (1993) specification.

351 .Kooli (2007) applies spanning to the diversification benefits of hedge funds and funds of hedge funds and finds that hedge funds as an asset class improve the meanvariance frontier of sets of benchmarks portfolios but that investors who already hold a diversified portfolio do not improve their statistics using hedge funds. The author further observes that funds of hedge funds offer diversification for mean-variance investors.

Only recently. (1997). Sun and Yan (2003). (2003). Harvey et al. however.. These approaches to multi-moment analysis are highlighted below before we present our utility-based methodology using a Taylor expansion.θ ) . Chan et al. Guidolin and Timmermann (2002) use a Taylor series expansion of the investors’ objective function to determine the optimal portfolios. (2005) incorporate investor preferences for higher moments of a return distribution into a polynomial goal programming optimization model.II Methodology and data Portfolio selection within a mean-variance-skewness framework has been considered by Lai (1991). Chunhachinda et al. Harvey et al. Davies et al. Jurczenko and Maillet (2006) build a four-moment efficient frontier for portfolio including hedge funds using a non-parametric methodology. (2004). They consider an m-th order Taylor series expansion of a generic utility function U (Wt +T . 2002. 352 . (2004) analyse the impact of inserting/deleting hedge funds in a hedge fund portfolio focusing on co-moments while Davies et al. have studies focused on a fourmoment framework (see. Guidolin and Timmermann. 2004). These studies all focus on portfolio selection but do not specifically study the spanning properties of a particular portfolio or a class of assets.

1 Classical portfolio selection approaches We consider the problem of an investor who selects a portfolio from risky assets in the mean-variance-skewness-kurtosis framework. we do not make the assumption of quadratic or power utility function but consider the Bell (1988. we do not only look at the slope of the capital market line but we also compare the level of returns achieved for a certain level of risk. Investors must be able to provide a detailed specification of their tastes. described by a joint subjective probability distribution. The problem with this approach is that it is difficult to implement in practice because the information requirements are strict. Considering the set of Von Neumann-Morgenstern’s axioms about rationale choice. our tests are based on the same underlying idea as the spanning test but the risk measure is extended to include the third and fourth central moment of the portfolio returns distribution. 1995) utility function. As the returns may be infinite.2. represented by a specific utility function as well as complete pattern of their beliefs. In our model. First. Moreover. we set that investors maximise their expected utility function at each period to derive their optimal investment choice. Second. it is impossible to reliably define such a probability distribution and to calculate the exact value of the expected utility. 353 . This is the most general and commonly accepted framework of portfolio selection theory referred as the E (U ) approach. Moreover. maximizing expected utility among all feasible portfolios is often a complex mathematical problem that requires a considerably greater number of calculations. we combine and adapt the tools presented in the previous section.

logarithmic.... m k ) (17) A sufficient condition for Vk to be a preference ordering consistent with E (U ) is that the utility function of the individual is a polynomial of degree n.. These parameters correspond in most cases to the first k moments of the distribution: m1 . m k provided that they exist. . Amongst usual utility functions. m 2. The other more appropriate functions (exponential... To isolate the impact of each moment. V k (m1 . The general approximating approach to expected utility consists in estimating the subjective distributions of the investment outcomes by k parameters so that the preference ordering Vk is defined over a set of k- values vectors. In this case there is no approximation to be made.This is why portfolio selection methods aim at being as close as possible to the model that maximises the expected utility of investors without encountering informational and computational problems. these non-polynomials are typically expanded using a Taylor series approximation of the utility function U . 354 . power and Bell’s) are non-polynomials so that defining a preference ordering among the first n moments of the distributions of uncertain outcomes will generally only result from an approximation to the actual expected utility of the investor. the quadratic is the only one satisfying this condition. . m 2... This technique consists in fitting a polynomial of degree k to U (W ) at one value (either 0 or W ).

yet no single conclusive approach has emerged. The first one is that the allocation problem solved cannot be precisely related to the expected utility function.. Several alternative approaches have been developed. 1999. Not only are hedge fund return distributions asymmetric and leptokurtic. The inclusion of higher moments in the efficient frontier analysis has been studied for years.A first attempt to approximate E (U ) is to estimate the probability distribution by its first two moments. Huberman and Kandel. since the choice of the parameter used to weight the moment deviations is not related to the parameters of the utility function. 2004. 2002 and Adcock.” Primal approaches include the polynomial goal programming (PGP) approach to determine the set of the mean-variance-skewness-kurtosis (see Davies et al.. this methodology cannot be used when the returns are not normally distributed as it is the case for hedge funds (see for example Ackermann et al. 2005 and Kat and Miffre. Agarwal and Naik. As stated in Jurczenko and Maillet (2006): “we can distinguish between primal and dual approaches for determining the mean-variance-skewnesskurtosis efficient frontier. leading to the mean-variance approximation (see for example Tobin. 2003. 2004). 1969. Malkiel and Saha. and Davies et al. 2001. 2005). 2004a). Minimizing deviations from the first four moments simultaneously only guarantees a solution that is close to the mean-variance-skewness-kurtosis efficient frontier. 2002. due to option-like features of alternative investments (see Weisman. Goetzmann et al. Agarwal and Naik. they also tend to display co-skewness and cokurtosis with the return of other assets classes. However.. These approaches are partial since 355 . 1987). There are two main shortcomings of this approach. 2004. Other authors tend to analytically solve the mean-variance-skewness-kurtosis portfolio optimization problem (see Athayde and Flôres. Fung and Hsieh. 2005). The second shortcoming is that the estimation does not comply with the Pareto-optimal definition of an efficient portfolio frontier.

we decide to select the mean-variance-skewness-kurtosis order of truncation as it seem to be the most comprehensive. The integration of skewness and kurtosis will enable us to take these statistics into account. Ederington (1986) finds that extending the expansion to a fourth term generally improves the approximation. 2. hedge fund returns may be skewed and have fat tails. 2005 and Jurczenko and Maillet. even when the series converges. Hlawitschka (1994) finds that adding terms to a Taylor-series expansion does not necessarily improve the approximation. however. as the descriptive statistics indicate. 2006). and never worsens it. Dual approaches use Taylor series expansion of the investors’ objective functions to determine the optimal portfolios (see Jondeau and Rockinger. The main drawback of this approach is that the Taylor series expansion may converge to the expected utility only under restrictive conditions (such as for the exponential).they are mainly focused on one objective of the optimization program at the cost of the others.2 Taylor approximation and risk measure Following Loistl (1979) and Hlawitschka (1994) we approximate the distribution on the basis of the first four moments. 356 . We base our analysis on a dual approach since our utility function includes an exponential. Bell’s function is approximated using a Taylor series expansion on the choice of a four-order approximation is justified as we analyse hedge fund return and. Moreover there is no rule in general for selecting the order of truncation. the first and main drawback of the approach does not apply and secondly.

the most common use of this formula is the expansion of f around its mean argument x . 2 n! (19) 357 .. Now consider a utility function of the random variable U (W ) . then the Taylor series expansion is a finite series with n terms.. if f is non-polynomial. The Taylor’s series expansion of this function around the mean end-of-period wealth W = W0 (1 + R ) – where R is the mean rate of return of the investment – equals: U (W ) = U (W ) + (W − W )U ' (W ) + (W − W ) 2 '' (W − W ) n n U (W ) + . + U (W ) + . then the Taylor series expansion is an infinite series (see Loistl.The Taylor approximation expresses any function f which is n+1 times differentiable over an open interval I that contains x and x + h .. If the function f is a polynomial of degree n.. On the other hand. 1976): f ( x + h) = ∑ f j =0 ∞ ( j) ( x) hj j! (18) When x is viewed as a random variable.

358 . it is often sufficient to only consider the first four moments of the wealth distribution since the associated derivatives of the utility function and moments of the wealth distribution are more intuitive to interpret. c >0 78 As stated by Guidolin and Timmermann (2002). we have78: 1 1 1 E [U (W )] ≅ U (W ) + σ 2 (W )U '' (W ) + S (W )U ''' (W ) + K (W )U '''' (W ) 2 6 24 (20) with S (W ) = E (W − W ) 3 and K (W ) = E (W − W ) 4 Applied to Bell’s function. the decomposed function including four moments is: U (W ) = W + be − cW c(W − W ) − be − cW c2 c3 c4 (W − W ) 2 + be − cW (W − W ) 3 − be − cW (W − W ) 4 2 6 24 (21) with b. the closer the approximation will be to the expected utility value.An intuitive idea is that the more terms are kept in the expansion. If we truncate the series to keep the first four terms.

1) we test whether the returns on the capital market line (CML) achieved for a certain risk level are significantly higher when new securities are included in the set of risky assets. Practically. the utility function becomes: 359 . If we define the risk premium as x = Θ − r . We perform two sets of tests. If invested in the risky asset. as we consider that everything that is not invested in the risky asset is invested in the riskfree asset. Define W0 as the initial value of the investors’ wealth and I as the global amount W0 − I . 2.3 Risk Measure Our objective is to determine if including hedge funds in a portfolio consisting of equity and bond mutual funds significantly improves the risk-return trade-off for the investor. The total amount invested in the risk-free asset is the risk-free rate gives a return equal to r and the risky asset a return equal to m . The difference between the capital market line determined in our case and the classical one is the nature of the risk measure used. 2) we test whether the slope of the capital market line for a defined level of risk changes when new securities are included in the set of risky assets. We can extract the risk measure from equation (9).Our objective is to determine if including hedge funds in a portfolio consisting of equity and bond mutual funds significantly improves the risk-return trade-off for the investor. we test whether the returns achieved on one side and the slope of the capital market line (CML) on another significantly change for a defined level of risk changes when new securities are included in the set of risky assets.

U (W ( x)) = W0 (1 + r ) + Ix − be − cW0 (1+ r ) E (e − cIx ) (22) and the expected utility is: E (U (W ( x))) = W0 (1 + r ) + Ix − be − c ( w0 (1+ r ) E (e − cIx ) ) (23) E (U (W ( x))) = W0 (1 + r ) + Ix − be − c ( w0 (1+ r )+ Ix ) E (e − cIx ( x− x ) ) The Taylor series expansion of this expression around m yields: (24) − cIx 1 1 1 E (U (W )) = W0 (1 + r ) + Ix ) − be −cW0 (1+ r ) e (1 + c 2 I 2Vx − c 3 I 3 S x + c 4 I 4 K x ) 2 6 24 (25) where V x = E (θ − θ ) 2 . S x = E (θ − θ ) " and K x = E (θ − θ ) 4 . 360 .

and draw the adapted CML with the risk-free rate. we identify the tangency portfolio. Based on experiments on a sample of 40 students. c. the (wealth-related) proportion of risky investment This risk measure accounts for volatility.e. C=35 for an average investor and C=60 for a very protective investor who mostly fears the tail event of a distribution. asymmetry and fat tails. we consider three representative levels. as in the CAPM. C=10 for a dynamic investor that does not put much weight on the extreme risks. for a standard unit of wealth (normalized 1 1 1 R x = V x − CS x + C 2 K x 2 6 24 (26) C≡c where I W0 is the product of the (intrinsic) risk perception coefficient. The level of C depends on the appetite for risk of the investor. and I /W0 . we get the risk measure to) R x . i. Then we fix a value of the risk measure and test the difference in the returns achievable before looking at the slope of the two capital market line (one 361 . Our spanning test is performed as follows. as the locus attainable combinations of the risk free rate and this risky portfolio) and the efficient frontier are tangent. the efficient portfolio where the adapted Capital Market Line or CML (defined. For each frontier.From (12).

Figure 12 illustrates the test performed. as risk is not measured with the variance. In our example.with hedge funds and one without) for that measure of risk. at X the test is fully conclusive since hedge funds and mutual funds do significantly better than mutual funds alone and mutual funds alone do significantly worse than hedge funds and mutual funds. Secondly. The underlying idea is illustrated in Figure 11. but mutual funds alone do not significantly worse than hedge funds plus mutual funds. 362 . we plot in bold a CML go through the 95% percentile and we add (in bold again) another CML that goes through the 5% percentile for the portfolio containing hedge funds and mutual funds. Our main challenge is to estimate the slope of the adapted CML because. The portfolio of hedge funds and mutual funds do significantly better than mutual funds alone. Both called “average CML” are in dotted lines in Figure 12. We first consider a risk level X and we let a CML containing only mutual funds through point A and another including hedge funds and mutual funds through point B. We plot point s C and D Plot points C and D at another risk value Y. while at Y the test is partly conclusive. The capital market line reported seems to be classic. The only particularity to this Figure is that the risk measure used is no more the standard deviation of the returns but the extended risk measure as defined in formula 26. this function is not a straight line.

In the simpliest form of bootstrapping.Figure 11: Illustration of the adapted Capital Market Line Return Capital Market Line Efficient Frontier Risk Measure 2. is an allusion to the expression “pulling oneself up by one’s bootstraps” – in this case.4 Estimation methodology The term “bootstrapping” due to Efron (1979). 363 . using the sample data as a population from which repeated samples are drawn. It is a general approach to statistical inference based on building a sampling distribution for a statistic by resampling from the data. one repeatedly analyses subsamples of the data. Each subsample is a random sample with replacement from the full sample.

net framework. The algorithm that makes the calculations was written in FORTRAN and C# and was compiled with the Microsoft. On the basis of the algorithm construct sets of portfolios having the required characteristics with regard to the proportion of funds of each type (for example 20 % hedge funds. constructing portfolios each with ten funds and we repeat the process as many times as required. equity mutual funds and bond mutual funds described below. 40 % equity mutual funds and 40 % bond mutual funds). All the funds from the three sets are randomly selected.Figure 12: Illustration of the test of the adapted Capital Market Line Return B C CML (HF+MF) CML95% (MF) CML5% (HF+MF) D CML (MF) A Y X Risk measure Three sets of funds are taken from a database: hedge funds. For each portfolio the algorithm determines two elements: 1) as a first step the CML generated from the funds included 364 .

The curve is then estimated with the locus of these risk/return pairs79. each slope corresponding to a particular value of Rt .78%. At each step the algorithm implements an optimization problem with the required return as objective function and the risk function as constraint. 365 . The calculation of the objective is worked out in accordance with the classical CAPM. The calculation of the CML is carried out point by point. The iterative process establishes a series of risk levels whose maximum required return must be calculated. x n −1 ) = R f + ( x0 β 0 + K x n −1 β n −1 )( Rm − R f ) (27) for a portfolio rp = ∑ wi ri i =1 n The calculation of the various centred moments function is carried out using the following general formula: m p .80 rb for the beta calculations is S&P 500 and the risk E ( x0 .in this portfolio and 2) a series of ten slopes for each CML.K . 80 This 2.l appearing in the risk 79 The specificity of this pair in comparison with the classical risk/return pair is that the risk measure has been improved by the including of skewness and kurtosis.78% risk free rate is the level of the 3-month T-bill rate from Ibbotson and associates such as the one used by Fama and French (1993). The benchmark index return free rate Rf is 2.

These elements imply that we have to solve a linear problem with non-linear constraints.. The selected resolution method is the SQP (Sequential Quadratic Programming..+ k n =l ) k + . 4 = λRb 2 6 24 w1 + ...t . 2002) and is based on an iterative reformulation of the non-linear problem into a problem of quadratic programming which is solved by the quadratic approximation of the Lagrangian of the objective function.m p . 2 − Cm p ... + wn = 1.3 + C 2 m p .wnkn (r1t − r1 ) k1 .. see Schittkowski.k n s. and the 366 .. wn max E (rp ) 1 1 1 m p .....l = ∑( k1 .. + k T t =1 1 n (28) And the optimization problem is: w1. R p ≡ (29) where λ is the target proportion of the (fixed) risk measure of the benchmark portfolio Rb ... 1 T k1 l! ∑ w1 ....t. wi ≥ 0∀i s.(rnt − rn ) kn k1 +.

20%. rank the slopes or the returns depending on the case considered by decreasing values for each value of risk. 50% and 100%) and separate directional.linearization of the constraints. undirectional and fund of funds that do not have the same objectives and distribution characteristics. summarized with values of parameter λ. the SQP method requires only the final solution to respect the constraints. Practically. Therfore. Our procedure for the calculation of the CML does not give us an analytical expression of this one which would enable us to use its derivative for the calculation of the slopes and the returns. we estimate the returns and slope of 2. In order to determine the reference p-value at the any confidence level.000 estimations for 5 different levels of lambda (5 different levels of allocation to the risk free rate . The main interest with this method lies in its “non-feasible way”: the progression towards an optimum solution is achieved on the basis of the “feasible” but also the “nonfeasible” intermediate points near the field of the constraints. is selected within the whole range of the risk values. Then. contrarily to many methods where the constraints are verified at each iteration. The slope is then given by ∆R p R p = λRb . Thus a set of risk values.000 CML (using bootstrapping). 80% and 100%). We perform 2. The resulting quadratic problem is then solved for every iteration. and select the level of confidence corresponding to the numerical p-values. we perform the same algorithm including various levels of hedge funds (10%. For each of these values the expected yield value is calculated for R p − ε and for R p + ε with ε = 1E − 8 ∆E ( R p ) in accordance with the above procedure. we estimate the 95% percentile 367 .20%. 60%. 40%.

Complete tables are available. 81 We made the estimation for the 90%. 368 .returns and slopes for a portfolio consisting of mutual funds only81 and we compare these levels with the mean82 return or slope of the results obtained when hedge funds are added. Tables are available upon request. The results are not reported because they are perfectly in line. 95% and 99% confidence levels but we only report the 95% level. 82 We make the same estimation using the median.

we use data from Micropal. As this bias favors the performance of the surviving funds. this introduces survivorship bias. For hedge funds. We report the descriptive statistics of our databases in Table 54. 83 Obviously. Note. equity hedge and CTAs (almost 60% on aggregate). our conclusions are very likely to embellish the risk-return trade-off of hedge funds. A large number of estimations are required and we therefore postpone this step. This particular period was especially difficult for hedge funds along with the Nasdaq bubble of Srping 2000 and the enuing bear market of 2000-2002. For mutual funds.III The database We attempt to determine if adding hedge funds in a classical portfolio can enhance the utility of an investor that is invested in stock and bond mutual funds. Table 54 indicates that the bulk of the hedge funds in the database are funds of funds. There is no simple way to mitigate this problem though. also that a similar bias is likely to hold for our sample of mutual funds as well. and should be considered with caution.8%. The best performers over the period under review have been directional funds led by emerging markets.5% respectively. equity nonhedge and equity hedge managers providing net annualized returns of 21. we use the Hedge Fund Research. This would be the next step in such an analysis. Following Amin and Kat (2003). A way to mitigate this problem would be to replace a fund from the dataset when it is dissolved (of before it exists) by another fund that exists at this moment.308 surviving bond mutual funds. 83 Note that our sample includes the period 1997-1998 which includes crises in Asia and Russia and the subsequent collapse of LTCM.2% and 15. 18. The lower return providers have been equity market neutral. Inc database.418 surviving equity mutual funds and 1. We use monthly performance data from 2. 369 . we concentrate on the 431 funds that exist during the entire January 1996-May 2006 period.

The same analysis on kurtosis indicates that kurtosis is always positive for hedge funds and that it is largely positive in most cases as indicated by Figure 14. 370 . a large majority of the directional strategies have been the most volatile.34% and 9. fixed income arbitrage. there is a diversification effect and the individual skewness reported has a large range.5% for relative value to 33. In terms of volatility.1% for convertible arbitrage and from 2.4% for short sellers respectively.47% respectively.merger arbitrage and convertible arbitrageurs that provided net returns of 7. as one could expect.85%. fund of funds and relative value) skewness tends to be negative. These results confirm that at least for some strategies (event driven. Next. The minimum and maximum monthly returns range range from -25. which is illustrated in Figure 13. The results interestingly show that even if skewness reported per strategy is close to zero. Pure arbitrage strategies have offered the most stable returns.4% for emerging markets to -2. 9. equity market neutral. we report skewness and kurtosis for each strategy as well as the minimum and maximum skewness and kurtosis obtained for individual funds in each strategy and globally.

6 2.10% 10.30% 11.07 0.90% 21.50% 11.70% 13.40% 7.30% 20.20% 9. DS EM ENH EH ED MAC SEC SS CTA MT UNDIR.59 1.2% 12. Sharpe Ratio DIRECT.94 1.88 1.10% 3.10% 3.0% 0.6% 4.09 .50% 11.70% 13.23 0.8% 5.10% 11.00% 7.60% 36.86 2.2% 3.40% 13.5% 0.6% 18.90% 18.60% 12. MA RV EMN FIA CB FOF GLB 245 11 19 20 81 25 18 13 1 54 3 70 8 12 14 17 19 116 431 1.8% 4.96 2.9% 100% 9.70% 3.20% 5.24 1.2% 3.80% 10.91 1.30% 14.60% 11.96 1.4% 26.30% 5. Dev.10% 5.7% 12.10% 6.70% 9.27 1. Return Std.20% 13.44 1.90% 10.10% 3.97 1.80% 7.8% 3.00% 1.20% 2.20% 15.9% 4.4% 4.25 1.9% 2.Table 54 : Descriptive statistics of the hedge fund (Panel A) and of the Equity and Bond Mutual Funds (Panel B) Panel A: Hedge Fund Descriptive Statistics (1/1996-5/2006) N % Mean Ann.12 0.

20% -5.Panel A (continued): Hedge Fund Descriptive Statistics (1/1996-5/2006) Min DIRECT.80% -4. DS EM ENH EH ED MAC SEC SS CTA MT UNDIR.10% -5.09 0.10% -2.40% -13.00% 9.02 0.60% 3.60% 0.02 0.20% 9.19 0.70% -9.90% -21.02 -0.01 0.80% 3.02 0.10% 4.01 1.01 Max Skewness Kurtosis 372 .20% -3.02 -0.01 0 0.50% -10.02 -0.20% 11.40% 6.01 0.01 0.80% 19.50% 33.60% 5.60% 7.20% -3.90% -6.01 -0.70% 7.02 0.01 0.01 0.01 0.04 -0.06 -0.01 0.01 0.02 0.80% 0.80% -6.06 0.03 0.16 0.90% 2.02 -4.04 0.70% -2.80% -25.40% -14.01 0. MA RV EMN FIA CB FOF GLB -5.02 0.00% 8.02 0.02 0.70% 3.60% 7.60% 7.

Std. Multi = mutli-strategy.5% 16.EqLong = equity long biaised and equity long only. MAC = macro.5% -0.30% 0.68% 0. 2418 equity mutual funds and 1308 bond funds over the January 1996 to May 2006 period.2% 0.=undirectional strategies. MA = merger arbitrage. CB = convertible arbitrage. DIR. ED = event driven. EMN = equity market neutral.39 Bd MF 1 308 4. Sharpe Ratio (2%) Eq MF 2 418 9. ELS= equity long/short. DS = distressed securities.37 Min Max Skewness Kurtosis Eq MF -16. Return Ann.62 0. Dev.51 This table reports the descriptive statistics of our database.Panel B: Mutual Fund Descriptive Statistics (1/1996-5/2006) N Mean Ann.85% 7.2% 7.= directional strategies. Others = other strategies. StatArb = statistical arbitrage.98 Bd MF -4. Our database covers 431 hedge funds. FI = fixed income strategies. EqShort = equity short. EM = emerging markets.22 0. FoF = fund of funds and HF = global database. 373 .60% 19. UNDIR. SEC = sector funds.

The skewness figures (more particularly the minimum and maximum skewness) indicate a positive skewness trend. These results indicate that hedge funds seems to offer attractive risk-return tradeoffs but that the asymmetry in their return distribution and the presence of large fat tails may impede their attractiveness when all these aspects are taken into consideration. Bond mutual funds offer much lower returns with a relatively limited volatility. Kurtosis column indicates the presence of fat tails.3%. the net mean annualized return is at 9.6% for equity funds with volatility at 19. 374 . positive skewness and positive kurtosis.On the mutual fund side. Such returns are relatively low when compared to hedge funds with a high volatility.

DS = distressed securities. CB = convertible arbitrage. FoF = fund of funds and HF = global database.Figure 13: Repartition of the skewness of individual funds 4 3 2 1 0 Eq Lo ng EM N ac s M A St at Ar b EM ul ti ED H F or t S C B D S SE C th er FI EL Eq Sh M M -1 -2 -3 -4 This figure reports the repartition of the skewness of individual funds. FI = fixed income strategies. ELS= equity long/short. StatArb = statistical arbitrage. Our database covers 431 hedge funds.318 bond funds over the January 1996 to May 2006 period.=undirectional strategies. Others = other strategies.= directional strategies. Multi = mutlistrategy.EqLong = equity long biaised and equity long only. MAC = macro. EM = emerging markets. DIR. MA = merger arbitrage. SEC = sector funds. EMN = equity market neutral. O Fo F 375 .418 equity mutual funds and 1. ED = event driven. EqShort = equity short. 2. UNDIR.

EqLong = equity long biaised and equity long only.318 bond funds over the January 1996 to May 2006 period. ELS= equity long/short. MA = merger arbitrage. 2. DIR. Our database covers 431 hedge funds. CB = convertible arbitrage. UNDIR. FI = fixed income strategies.= directional strategies. Others = other strategies. ED = event driven.418 equity mutual funds and 1. EqShort = equity short.Figure 14: Repartition of the kurtosis of individual funds 12 10 8 6 4 2 0 EM Eq Lo ng M ul ti M ac er s N M A St at Ar b H F Eq Sh or t EL S SE C -2 -4 -6 This figure reports the repartition of the kurtosis of individual funds. StatArb = statistical arbitrage. MAC = macro. Multi = mutlistrategy. FoF = fund of funds and HF = global database. SEC = sector funds.= undirectional strategies. DS = distressed securities. O th EM Fo F ED C B D S FI 376 . EM = emerging markets. EMN = equity market neutral.

Then. we report the results of the estimation increasing the allocation to hedge funds from 0 to 100% for three types of investor risk profiles and for the five different levels of allocation to the risk-free asset. the level over which the inclusion of hedge funds in a portfolio enables the investor to significantly improve his satisfaction. We first have to determine the significance levels. We test if the mean returns obtained for each level of risk are higher than the 95% level of confidence of a mutual funds only portfolio when hedge funds are added. We then compare the mean return obtained for portfolios including hedge funds with p-value in order to determine if hedge funds significantly add value to the initial portfolio. 4. The p-values 377 .1 Significance level We want to ascertain if the inclusion of hedge funds in a portfolio of stocks and bonds enables an investor to significantly improve the level of return he gets for a predetermined maximum level of risk. We start this section by explaining how we determine the levels of the p-values. that is. then we estimate the p-values by calculating the value corresponding to the 95% percentile of a portfolio consisting of mutual funds only (50% equity mutual funds and 50% bond mutual funds) for each value of lambda. This 95% confidence level is obtained by estimating 2.IV Results The objective of this study is to test if the inclusion of hedge funds in a diversified portfolio of stock and bond mutual funds enables investors to attain a significantly higher level of return for the same risk level. by ranking them to extract the level of confidence (95% corresponding to the 1900th estimation).000 capital market line for each level of lambda. A further analysis is based on the slope of the capital market line estimated and it test the inverse relation whereby the p-values are estimated the same way.

For inverse relations. For an allocation of 40% to the risky asset.3% annualized 95% of the time. 5% of the time. Finally Panel C the mean returns when funds of hedge funds are added to the portfolio.for the slope analysis are estimated for mutual fund portfolios only and are then ranked. A full exposure to the risky asset would have yield returns as high as 16%. Average investor The first line of Table 56 reports the p-value for the 95% confidence level for a portfolio of mutual funds only. 378 . a portfolio consisting of equity mutual funds and bond mutual funds would have offered a return lower than 5. Panel B the same results when undirectional hedge funds are added. we consider three categories of investors: average. a. we estimate the p-values for a portfolio consisting of only hedge funds and compare this level with the mean returns obtained when mutual funds are added to the portfolio. 4. Panel A reports the mean returns when only directional hedge funds are added. We report the results obtained for the three categories of investors separately. protective and progressive. The p-value line indicates that 95% of the time. the mean annualized return would have been lower than 8.4% annualized if 20% of the assets were allocated to the risky asset. For each estimation we report a Table that provides comparison and figures that illustrate the results.2 Including hedge funds in mutual funds portfolio As previously stated.

The first line of Panel A indicates that the addition of 10% of directional hedge funds would not enable the investor to reach a higher mean return. the levels of return achievable are lower than the corresponding p-values. This result holds for any level of allocation to the risky asset. directional hedge funds do not enable the investor to attain the risk level corresponding to the 20% lambda. The 100% directional hedge fund line confirms that on a stand alone basis directional hedge funds are more attractive than a diversified portfolio of mutual funds. They clearly indicate that large allocations to directional hedge funds enable investors to reach significantly higher levels of returns for the same level of risk. For allocation to the risky asset between 40% and 60% however. some investors can reach significantly higher level of return when 20% of their portfolio is allocated to directional hedge funds. The next two lines report 50% and 100% levels of allocation to hedge funds. The first column of Panel A indicates that as they tend to be risky. The second line of Panel A shows that results are mitigated for an allocation of 20% to directional hedge fund strategies. 379 . For low levels of allocation to the risky asset and for high allocation to the risky asset. This result should be explained by the fact that directional strategies are exposed to the market and adding only 10% of directional strategies to a directional portfolio does not enable the investor to attain significantly higher levels of returns. This indicates a first inflection point whereby depending on the risk appetite and the allocation to the risky asset. the inclusion of hedge funds helps for medium risks levels.

8% 11.1% 9.0% 14.5% PANEL B: UHF Mean Return 45% 45% 10% 5.3% 60% 11.2% 12.0% All 14.7% 10.0% 14.4% 13.9% 8.6% 12.5% 12.5% 13.2% 11.6% 14.6% 10.5% 80% 14.3% 18.6% 17.7% 7.8% Mean Return 40% 40% 20% 6.7% 11.9% 14.7% 20.9% 13.7% 12.1% 12.4% 12.8% 9.2% 15.5% Mean Return 25% 25% 50% 8.6% PANEL A: DHF Mean Return 45% 45% 10% 4.9% 10.4% 40% 8.5% 18.3% 13.3% Mean Return 0% 0% 100% NA 12.6% 13.1% 13.4% 11.Table 55: Mean return estimation for average investors Lambdas EMF BMF HF 20% P-value (95%) 50% 50% 0% 5.6% .7% 16.4% Mean Return 0% 0% 100% 9.5% Mean Return 40% 40% 20% 5.2% 11.4% 16.1% 16.0% Mean Return 25% 25% 50% 5.6% 15.4% 100% 16.3% 15.3% 10.

5% Mean Return 40% 40% 20% 5.6% 14.2% 13. Our database covers 431 hedge funds.1% 12.3% 12.3% 14. Panel B reports the result obtained when including undirectional hedge funds in the initial portfolio and Panel C reports the result obtained when including funds of hedge funds in the initial portfolio. 80% and 100%).418 equity mutual funds and 1. 2. P-values are obtained by estimating 2.3% 13.0% 17.3% 14. 60%.9% 10. S x = the skewness of the return distribution and x = the kurtosis of the return distribution.5% Mean Return 25% 25% 50% 6.2% 13. P-values estimated with a mutual fund portfolio divided equally between equity mutual funds (EMF) and bond mutual funds (BMF). An average investor is defined as an investor with a C value of 35 in our extended risk measure defined as 1 1 1 R x = V x − CS x + C 2 K x K 2 6 24 with V x =the variance of the return.7% 13.318 bond funds over the January 1996 to May 2006 period. Panel A reports the result obtained when including directional hedge funds in the initial portfolio.5% 11. Numbers in the table are annualized mean returns.8% 16. Lambda is the level of allocation to the risk free rate (20%.000 CML for a portfolio of mutual funds for any level of lambda and taking the 95% percentage level value. .Lambdas EMF BMF HF 20% PANEL C: FOF 40% 60% 80% 100% All Mean Return 45% 45% 10% 5.9% This table reports the mean return obtained for various allocation to hedge funds and for various level of allocation to the risky asset for an average investor. 40%.5% Mean Return 0% 0% 100% 8.2% 11.5% 12.6% 14.2% 12.6% 15.0% 13.4% 9.

undirectional strategies do not help diversifying. Panel C reports exactly the same pattern as Panel B. 382 . The difference is however that allocating 50% to funds of hedge funds can enable the investor to significantly improve his returns when 80% of the portfolio is allocated to risky strategies. Figure 15 confirms that a small allocation to directional hedge funds does not enable the investor to reach significantly higher returns while higher allocation do. Low risk allocation can significantly improve the returns offered while adding low levels (10-20%) of undirectional hedge funds or funds of funds while higher risk level should give high allocation to directional strategies. particularly for higher risk level.Panel B reports a different pattern. These results are in line with our expectations. Adding only 10% of a portfolio to funds of hedge funds enables an investor to reach significantly higher levels (for allocation to the risky assets up to 60%). For higher risk allocation. The first three of Panel B indicate that the addition of undirectional hedge funds to a portfolio of mutual funds enable investors to reach higher levels of returns for allocation to the risky assets up to 60%. Figures 15 to 17 report the same results visually. Undirectional strategies and funds of funds are low risk profile investments with return distribution completely different from the equity markets and to a certain extent to directional hedge funds that are exposed to the market. indicating that funds of hedge funds are marginally more attractive than unidirectional strategies. This first analysis interestingly and clearly indicates that hedge funds can help an investor to reach a higher level of returns for any level of risk depending on his appetite for risk and its allocation to the risky asset. The last line of Panel B however indicates that a 100% allocation to undirectional hedge funds is however more attractive than a 100% mutual fund portfolio.

Lambda 40% 383 . 40%. P-values are the return distribution and obtained by estimating 2. 10%.01 0. The 0% hedge fund portfolio is divided equally between equity mutual funds (EMF) and bond mutual funds (BMF).000 CML for a portfolio of mutual funds for any level of lambda and taking the 95% percentage level value.Figure 15: Directional Hedge Funds . 80% and 100%).01 -0.02 -0. P-values estimated with a mutual fund portfolio divided equally between equity mutual funds (EMF) and bond mutual funds (BMF). 2.Mean Percentile Return 0. P-values are reported in the 0% allocation to hedge funds part of the Figure.318 bond funds over the January 1996 to May 2006 period.00 -0. 50% and 100%) and for various levels of allocation to the risk-free asset (20%.04 All lambdas Lambda 100% Lambda 80% 100%D F H 50%D F H 20%D F H 10%D F H Lambda 20% Lambda 60% This figure reports the difference between the mean hedge fund return and the critical value obtained for various levels of allocation to directional hedge funds (0%. 40%. S x = the skewness of K x = the kurtosis of the return distribution. 20%.03 -0. Numbers in the table are annualized mean returns.05 0.02 0. 60%. Our database covers 431 hedge funds.04 0. Lambda is the level of allocation to the risk free rate (20%. An average investor is defined as an investor with a C value of 35 in our extended risk measure defined as 1 1 1 R x = V x − CS x + C 2 K x 2 6 24 with V x =the variance of the return.418 equity mutual funds and 1. 60%.03 0. 80% and 100%) for an average investor.

More than 20% of the portfolio has to be allocated to hedge funds. b. 384 . For higher allocation to the risky asset large exposure to undirectional hedge funds have to be considered to be as attractive as the classical portfolio of stock and bond mutual funds. Results are reported in Table 57 and Table 5884. For higher allocation to the risky asset. The lambda 20% series of values increases monotically with the inclusion of undirectional hedge funds confirming that allocating 10% of the portfolio to undirectional hedge funds strategies enables investors to obtain significantly higher returns. A progressive investor does not care as much about the risk but is focused on returns that can be attained. particularly for low values of lambda. integrating 10% or 20% of funds of funds does not help the investor to attain higher returns. As indicated by Figure 17. Protective & progressive investors By definition. funds of hedge funds offer the same characteristics as undirectional hedge funds. and enable investors to attain higher returns for the same level of risk for low allocation to the risky asset (lambda up to 60%) even when only 10% of the portfolio is allocated to hedge funds. a protective investor does not like riskand he would put more weight to an investment that enables him to diversify the risk of its portfolio away.Figure 16 shows that including undirectional hedge funds enable investors to earn higher returns. 84 We do not report the figures for the sake of brevity but they are available.

04 0. Numbers in the table are annualized mean returns. S x = the skewness of K x = the kurtosis of the return distribution. An average investor is defined as an investor with a C value of 35 in our extended risk measure defined as 1 1 1 R x = V x − CS x + C 2 K x 2 6 24 with V x =the variance of the return. P-values estimated with a mutual fund portfolio divided equally between equity mutual funds (EMF) and bond mutual funds (BMF).Figure 16: Undirectional Hedge Funds . 80% and 100%) for an average investor.000 CML for a portfolio of mutual funds for any level of lambda and taking the 95% percentage level value.02 -0. 50% and 100%) and for various levels of allocation to the risk-free asset (20%. 80% and 100%).03 0.04 All lambdas Lambda 100% 100%U F H 50%U F H Lambda 80% Lambda 60% 20%U F H Lambda 40% 10%U F H Lambda 20% This figure reports the difference between the mean hedge fund return and the critical value obtained for various levels of allocation to undirectional hedge funds (0%. Lambda is the level of allocation to the risk free rate (20%. 40%.318 bond funds over the January 1996 to May 2006 period. P-values are reported in the 0% allocation to hedge funds part of the Figure.01 -0.03 -0.418 equity mutual funds and 1. 2.Mean Percentile Return 0. Our database covers 431 hedge funds.05 0.01 0.02 0.00 -0. 385 . 40%. 20%. P-values are the return distribution and obtained by estimating 2. 60%. The 0% hedge fund portfolio is divided equally between equity mutual funds (EMF) and bond mutual funds (BMF). 60%. 10%.

The 0% hedge fund portfolio is divided equally between equity mutual funds (EMF) and bond mutual funds (BMF). S x = the skewness of K x = the kurtosis of the return distribution.318 bond funds over the January 1996 to May 2006 period.02 -0. 386 .Mean Percentile Return 0. Lambda is the level of allocation to the risk free rate (20%. 10%.03 0.06 0. 40%.000 CML for a portfolio of mutual funds for any level of lambda and taking the 95% percentage level value. P-values are the return distribution and obtained by estimating 2.02 0.04 0. 20%.04 S6 100%FO F All lambdas 50%FO F Lambda 80% Lambda 60% 20%FO F Lambda 40% 10%FO F Lambda 20% This figure reports the difference between the mean hedge fund return and the critical value obtained for various levels of allocation to funds of hedge funds (0%. 40%.03 -0. 60%.01 -0.05 0. 80% and 100%). 50% and 100%) and for various levels of allocation to the risk-free asset (20%. P-values are reported in the 0% allocation to hedge funds part of the figure. 80% and 100%) for an average investor.418 equity mutual funds and 1.00 -0. Our database covers 431 hedge funds. P-values estimated with a mutual fund portfolio divided equally between equity mutual funds (EMF) and bond mutual funds (BMF).01 0. An average investor is defined as an investor with a C value of 35 in our extended risk measure defined as 1 1 1 R x = V x − CS x + C 2 K x 2 6 24 with V x =the variance of the return. 60%. 2.Figure 17: Funds of Hedge Funds . Numbers in the table are annualized mean returns.

1% 10.4% 11.7% EMF P-value (95%) 50% BMF 50% HF 0% PANEL A: DHF Mean Return 45% 45% 10% 4.5% 10.8% 13.5% PANEL B: UHF Mean Return 45% 45% 10% 6.4% 18.2% 11.5% 12.4% 12.1% 11.7% 8.2% 16.4% 13.0% 11.5% 14.7% 16.0% 8.0% 14.4% .2% 13.8% 11.5% 15.9% Mean Return 25% 25% 50% 5.9% 9.0% 15.1% 13.3% 40% 8.5% 80% 14.5% 100% 16.9% 14.6% 12.0% 15.3% All 14.1% 16.7% Mean Return 40% 40% 20% 6.6% 13.Table 56: Mean return estimation for protective investors Lambdas 20% 5.7% 20.1% Mean Return 0% 0% 100% NA 12.9% 11.6% 10.1% 12.9% 14.7% Mean Return 40% 40% 20% 5.6% 10.2% 18.0% 9.5% 12.5% 17.3% 60% 11.3% Mean Return 0% 0% 100% 9.3% 12.4% Mean Return 25% 25% 50% 8.

60%.5% 12. Numbers in the table are annualized mean returns. Panel A reports the result obtained when including directional hedge funds in the initial portfolio.3% 9. A protective investor is defined as an investor with a C value of 60 in our extended risk 1 1 1 R x = V x − CS x + C 2 K x 2 6 24 with V x =the variance of the return.5% 13.8% 16.000 CML for a portfolio distribution and of mutual funds for any level of lambda and taking the 95% percentage level value.418 equity mutual funds and 1. Lambda is the level of allocation to the risk free rate (20%.1% 13.4% 11. 40%.1% 14.3% 14.5% 14. Our database covers 431 hedge funds. 2. Panel B reports the result obtained when including undirectional hedge funds in the initial portfolio and Panel C reports the result obtained when including funds of hedge funds in the initial portfolio. P-values estimated with a mutual fund portfolio divided equally between equity mutual funds (EMF) and bond mutual funds (BMF).0% 13.7% This table reports the mean return obtained for various allocation to hedge funds and for various level of allocation to the risky asset for a protective investor.6% 15.5% Mean Return 0% 0% 100% 8.6% Mean Return 40% 40% 20% 5. . P-values are obtained by estimating 2. 80% and 100%).4% Mean Return 25% 25% 50% 6.7% 10.4% 13.EMF BMF HF Lambdas 20% 40% 60% 80% 100% All PANEL C: FOF Mean Return 45% 45% 10% 5.318 bond funds over the January 1996 to May 2006 period.5% 15.8% 12.2% 12.4% 11.9% 14. S x = the skewness of the return measure defined as K x = the kurtosis of the return distribution.1% 12.5% 13.

2% 80% 14.4% 11.7% 15.7% 10.6% 14.8% 17.6% 13.4% 40% 8.9% 10.0% 14.9% 11.4% 11.4% 18.3% 11.3% 100% 16.5% 12.0% 9.7% 13.4% 12.8% 20.7% 12.4% 13.8% 11.2% 60% 11.3% 18.5% 13.5% 16.3% 12.1% 8.7% 16.2% 13.9% Mean Return 25% 25% 50% 5.5% Mean Return 0% 0% 100% 9.1% 16.8% Mean Return 40% 40% 20% 7.1% 13.1% 9.7% 11.5% Mean Return 25% 25% 50% 8.6% .1% 11.7% 14.6% PANEL B: UHF Mean Return 45% 45% 10% 6.7% 7.Table 57: Mean return estimation for progressive investors Lambdas 20% P-value (95%) 50% 50% 0% 5.3% 15.7% 10.0% 14.5% Mean Return 40% 40% 20% 5.6% EMF BMF HF PANEL A: DHF Mean Return 45% 45% 10% 4.2% Mean Return 0% 0% 100% NA 12.0% 15.6% 12.9% All 14.

8% 15.2% 14. Lambda is the level of allocation to the risk free rate (20%.5% 11.1% 12. Numbers in the table are annualized mean returns. K .6% 14. A progressive investor is defined as an investor with a C value of 10 in our extended risk measure defined as 1 1 1 R x = V x − CS x + C 2 K x 2 6 24 with V x =the variance of the return. 40%.Lambdas EMF BMF HF 20% PANEL C: FOF 40% 60% 80% 100% All Mean Return 45% 45% 10% 5. Our database covers 431 hedge funds.0% 12.6% Mean Return 0% 0% 100% 8. P-values are obtained by estimating 2.418 equity mutual funds and 1. 80% and 100%).2% 11.9% This table reports the mean return obtained for various allocation to hedge funds and for various level of allocation to the risky asset for a progressive investor. P-values estimated with a mutual fund portfolio divided equally between equity mutual funds (EMF) and bond mutual funds (BMF). 60%.4% 13.5% Mean Return 40% 40% 20% 5. Panel A reports the result obtained when including directional hedge funds in the initial portfolio.000 CML for a return distribution and portfolio of mutual funds for any level of lambda and taking the 95% percentage level value.3% Mean Return 25% 25% 50% 6.3% 14.7% 10. S x = the skewness of the x = the kurtosis of the return distribution.1% 17.8% 16.2% 13.4% 11. Panel B reports the result obtained when including undirectional hedge funds in the initial portfolio and Panel C reports the result obtained when including funds of hedge funds in the initial portfolio.2% 9.7% 13. 2.318 bond funds over the January 1996 to May 2006 period.8% 14.9% 13.5% 12.3% 13.

undirectional hedge funds or funds of funds. and to what extent they should be used to achieve higher levels of satisfaction. 391 .The results obtained for progressive and protective investors are very close to those obtained for an average investor. Corresponding Tables for protective and progressive investors are available upon request. The results indicate that it is the level of lambda (the global risk level) that has the most importance and the main impact on our results. Slope comparison Table 59 reports the same analysis based on the slope of the capital market line as a tool to measure investor satisfaction rather than simply using returns85. d. the investor should decide if he will use directional. The use of the slope as a tool to measure satisfaction can be justified by the fact that it can be viewed as an adapted Sharpe ratio with the extended risk measure as denominator as defined hereafter: Adapted Sharpe ratio: Re turn / Risk (30) 1 1 1 R x = V x − CS x + C 2 K x 2 6 24 With the risk measure defined as (31) 85 We report the results for average investors. Depending on the global risk level and the kind of investor considered.

010 Mean Return 0% 0% 100% NA 0.008 0.079 0.008 Mean Return 40% 40% 20% 0.002 All 0.033 0.001 0.001 0.008 0.005 0.038 60% 0.009 0.118 0.001 0.038 0.028 0.002 0.010 0.009 0.155 0.006 0.010 0.003 0.Table 58: Slope estimation for average investors Lambdas 20% P-value (95%) 50% 50% 0% 0.013 0.001 0.078 40% 0.020 80% 0.001 0.004 100% 0.009 Mean Return 25% 25% 50% 0.008 0.001 0.007 Mean Return 40% 40% 20% 0.161 0.020 Mean Return 0% 0% 100% 0.002 0.003 0.024 0.003 0.068 0.009 PANEL B: UHF Mean Return 45% 45% 10% 0.088 0.017 0.012 0.008 0.031 .001 0.003 0.001 0.022 EMF BMF HF PANEL A: DHF Mean Return 45% 45% 10% 0.003 0.008 Mean Return 25% 25% 50% 0.001 0.

019 0.286 0.003 0.003 0.001 0.001 0.002 0.008 Mean Return 25% 25% 50% 0. P-values estimated with a mutual fund portfolio divided equally between equity mutual funds (EMF) and bond mutual funds (BMF). Panel A reports the result obtained when including directional hedge funds in the initial portfolio.005 0.007 Mean Return 40% 40% 20% 0. Panel B reports the result obtained when including undirectional hedge funds in the initial portfolio and Panel C reports the result obtained when including funds of hedge funds in the initial portfolio.EMF BMF HF 20% 40% 60% Lambdas 80% 100% All PANEL C: FOF Mean Return 45% 45% 10% 0.001 0.099 0.418 equity mutual funds and 1. Lambda is the level of allocation to the risk free rate (20%. P-values are obtained by estimating 2. 2. Our database covers 431 hedge funds.000 CML for a portfolio of mutual funds for any level of lambda and taking the 95% percentage level value.160 0. 40%.009 Mean Return 0% 0% 100% 0. S x = the skewness of the return distribution and defined as K x = the kurtosis of the return distribution. 80% and 100%).001 0.001 0. 60%.001 0.014 0.004 0. An average investor is defined as an investor with a C value of 35 in our extended risk measure 1 1 1 R x = V x − CS x + C 2 K x 2 6 24 with V x =the variance of the return. .001 0.016 0.318 bond funds over the January 1996 to May 2006 period.020 0.008 This table reports the mean slope obtained for various allocation to hedge funds and for various level of allocation to the risky asset for an average investor.507 0. Numbers in the table are annualized mean returns.

The main difference is that the inclusion of high levels of undirectional hedge funds and funds of funds increases rapidly the slope in absolute and relative bases. the lambda 100% column indicates that for high allocation to the risky asset. Results are however completely different when we go further to the right and consider higher risk allocation.Table 59 reports mitigated results. Directional hedge funds enable average investors to improve their satisfaction (as measured by the slope) by giving high allocation to directional hedge funds for low levels of risky assets (lambda 20%). This is perfectly in line with the visual aspect of an efficient frontier. the slope decreases as we increase the risk level. Second. the slope is low on absolute terms and close to zero indicating that we are close to the flat tangent. but the results are unclear for higher levels of risk since figures reported are very low in absolute terms. Two remarks. Results requires careful analysis when returns are not the only measure considered as a measure of investor wealth for a specific level of risk level. confirming that hedge funds enable investors to reach significantly higher levels of satisfaction (as measured by the ratio of return to risk) when the risk level is low. first. the curve drawn will be concave. Panel B and C report the same pattern for undirectional hedge funds and for funds of hedge funds. 394 . Directional hedge funds do not enable investors to reach higher levels of satisfaction. We obtain exactly the same pattern for protective and progressive investors. An investor who accepts higher risks with the objective of higher returns will see risk/return ratio decrease as the risk increase more rapidly than the return. Panel A reports the slopes of the capital market line obtained when directional hedge funds are added to the portfolio. These results have to be mitigated because we are looking at very small numbers on absolute terms.

Adding mutual funds to hedge fund portfolio is attractive. a. This result is in line with the underlying characteristics of undirectional strategies.3 Including mutual funds in hedge fund portfolios We perform the inverse test as a complementary analysis. For high risk portfolio however. Results are more controversial with funds of funds. When higher allocation are made to mutual funds. Over this level. Average investors Table 59 interestingly indicates that in some cases the inclusion of mutual funds in a hedge fund portfolio may enable investors to reach higher returns. the returns offered for each level of risk are more attractive when only funds of funds are considered.4. In the case of directional hedge funds. For higher level of risks. by starting with a portfolio of hedge funds and test if the inclusion of mutual funds in this portfolio enables the investor to reach a significantly higher level of satisfaction. The inclusion of mutual funds in a fund of hedge funds portfolio is only attractive for relatively low levels of allocation to mutual funds (up to 50%). We perform this test for average. To offer riskier returns. Such strategies are low risk strategies by definition. Up to 80% of the portfolio can be allocated to mutual funds and the portfolio would still offer attractive returns for low level of risks (lambda up to 40%). 395 . the results are even more interesting. such strategies should be diversified with directional high risk strategies. For directional hedge funds. the returns offered remains attractive even for low allocation to hedge funds. protective and progressive investors as well as use mean returns as a measure of satisfaction and slope. the investor can reach higher level of returns for any level of risk when 50% of the portfolio is allocated to mutual funds (half equity and half bon mutual funds). returns are less attractive indicating that high allocation should be made to directional hedge funds when investors are considering directional hedge funds a diversification tool to mutual funds.

Protective and progressive investors86 The results obtained for protective and progressive investors are the same as those obtained for average investors. This result is logical since funds of hedge funds are low risk investments that should be diversified to be considered risky.this is not true. 86 Complete results are available upon request. 396 . Adding mutual funds in portfolio of hedge funds can be attractive mainly for high risk investors and the inclusion of mutual funds in a undiversified portfolio of hedge funds is attractive for relatively low levels of allocation to mutual funds. b.

2% 15.9% 14.8% Mean Return 50% 50% 0% 5.1% All 7.5% Mean Return 50% 50% 0% 4.8% 11.7% 10.9% Lambda 11.5% 18.8% 9.Table 59: Mean return estimation for average investors (including mutual funds in a hedge fund portfolio) Lambda Panel A: DHF EMF BMF HF 20% Percentile 5% 0% 0% 100% NA 40% 9.9% 13.0% 12.7% Percentile 5% 0% 0% 100% 6.8% Mean Return 25% 25% 50% 5.4% 13.6% 10.1% 12.7% 12.8% .4% Mean Return 40% 40% 20% 6.1% 16.6% 12.2% 11.5% 6.5% 12.9% 10.7% 100% 11.9% 8.7% 12.4% 11.2% 12.5% 13.2% 80% 14.0% 60% 10.9% 100% 15.6% 10.7% 11.8% 8.3% 10.1% 9.3% Mean Return 40% 40% 20% 5.2% 11.5% 80% 10.6% 14.9% 8.7% 7.3% Mean Return 25% 25% 50% 8.3% 10.3% 15.4% 11.0% Mean Return 45% 45% 10% 4.0% 14.8% All 11.6% 13.7% 10.4% 12.2% Panel B: UHF EMF BMF UHF 20% 40% 10.5% Mean Return 45% 45% 10% 5.6% 60% 13.

8% 8.5% 11.3% 13. Our database covers 431 hedge funds. .418 equity mutual funds and 1.5% M ean R eturn 45% 45% 10% 5.3% 12.9% 11. Panel B reports the result obtained when the initial portfolio consists of unidirectional hedge funds and Panel C reports the result obtained when the initial portfolio consists of hedge funds of funds. Panel A reports the result obtained when the initial portfolio consists of directional hedge funds. An average investor is defined as an investor with a C value of 35 in our extended risk measure 1 1 1 R x = V x − CS x + C 2 K x 2 6 24 with V x =the variance of the return.7% M ean R eturn 25% 25% 50% 6.0% 13. 40%.3% 14.5% 6.318 bond funds over the January 1996 to May 2006 period.3% 10.000 CML for a portfolio of hedge fund for any level of lambda and taking the 95% percentage level value.1% 12.5% 12.6% 14. 60%.2% This table reports the mean return obtained for various allocation to mutual funds and for various level of allocation to the risky asset for an average investor. P-values are obtained by estimating 2.2% 11. The allocation to mutual fund is equally weighted between equity mutual funds (EMF) and bond mutual funds (BMF).9% 10.4% 9.7% 11.0% 12. 2.6% 15.3% 12.2% 12. Lambda is the level of allocation to the risk free rate (20%.5% M ean R eturn 50% 50% 0% 4.5% M ean R eturn 40% 40% 20% 5. S x = the skewness of the return distribution and defined as K x = the kurtosis of the return distribution.2% 13.6% 13.Lambda Panel C: FoF E MF B MF FO F 20% 40% 60% 80% 100% All P ercentile 5% 0% 0% 100% 6.5% 11.7% 13. Numbers in the table are annualized mean returns.2% 13. 80% and 100%).

c. 399 . Mutual funds may enable hedge fund investors to significantly increase their level of satisfaction. Slope comparison As reported in Table 61. the results obtained using the slope as a measure of satisfaction are in line with the previous one based on mean return for any specific lambda.

1551 0.0074 Mean Return 50% 50% 0% 0.0079 Mean Return 50% 50% 0% 0.0046 0.0197 Mean Return 40% 40% 20% 0.0006 0.0079 0.0027 All 0.0008 0.0138 80% 0.0694 60% 0.0170 0.1178 0.1807 Percentile 5% 0% 0% 100% 0.0791 0.0104 0.0284 0.0684 0.0077 0.0028 0.0025 0.Table 60: Slope estimation for average investors (including mutual funds in a hedge fund portfolio) Lambda Panel A: DHF EMF BMF HF 20% Percentile 5% 0% 0% 100% NA 40% 0.0235 0.0017 0.0091 0.0335 Mean Return 25% 25% 50% 0.3578 Mean Return 25% 25% 50% 0.0273 60% 0.0022 0.0026 0.0078 0.0098 0.0011 0.0024 0.0099 Mean Return 40% 40% 20% 0.0375 0.0011 0.0088 0.0884 0.0091 Mean Return 45% 45% 10% 0.0023 100% 0.0013 All 0.0197 0.0520 0.0065 Panel B: UHF EMF BMF HF 20% 40% 0.0197 0.0067 80% 0.0022 0.0012 0.0095 Lambda 0.0065 .0055 0.0053 100% 0.0131 0.0011 0.0082 Mean Return 45% 45% 10% 0.0095 0.0011 0.0520 0.0011 0.

The allocation to mutual fund is equally weighted between equity mutual funds (EMF) and bond mutual funds (BMF). 80% and 100%).0190 0.0015 0.0065 This table reports the mean slope obtained for various allocation to mutual funds and for various level of allocation to the risky asset for an average investor.0315 60% 0.0033 0.0007 0.Lambda Panel C: FoF EMF BMF FOF 20% Percentile 5% 0% 0% 100% 1.0095 0.0068 Mean Return 50% 50% 0% 0. P-values are obtained by estimating 2. Lambda is the level of allocation to the risk free rate (20%. 40%.418 equity mutual funds and 1.0007 0.1598 0.0093 Mean Return 40% 40% 20% 0. Our database covers 431 hedge funds.318 bond funds over the January 1996 to May 2006 period. Panel A reports the result obtained when the initial portfolio consists of directional hedge funds.0059 80% 0.0238 Mean Return 25% 25% 50% 0.0158 0.000 CML for a portfolio of hedge fund for any level of lambda and taking the 95% percentage level value.6006 40% 0. Numbers in the table are annualized mean returns.0014 0.0985 0.0011 0. S x = the skewness of the return distribution and defined as K x = the kurtosis of the return distribution.0198 0.0022 0. An average investor is defined as an investor with a C value of 35 in our extended risk measure 1 1 1 R x = V x − CS x + C 2 K x 2 6 24 with V x =the variance of the return. Panel B reports the result obtained when the initial portfolio consists of unidirectional hedge funds and Panel C reports the result obtained when the initial portfolio consists of hedge funds of funds.0048 0. .2858 0.0025 100% 0.0038 0.0014 All 0. 60%.0077 Mean Return 45% 45% 10% 0.0520 0. 2.0014 0.0008 0.0197 0.

Our methodology is based on the Taylor’s extension of the linex utility function developed by Bell (1988. Our results indicate that directional hedge funds should be considered separately from undirectional hedge funds and fund of hedge funds. This new methodology opens a new field in portfolio analysis research. hedge funds exhibit an attractive risk/return profile. however this result needs careful care when higher moments are considered. In this paper we develop the idea of an adapted capital market line in an extended risk-return framework that includes not only volatility as a measure of risk but also higher moments. It enables investors to determine if they should include alternative investments in their portfolio depending on their level of risk aversion. Adding undirectional hedge funds or fund of funds to a classical portfolio enables investors to reach higher levels of returns for low and medium risk levels for allocation as low as 10% to hedge funds. This result in a significant improvement of satisfaction for diversified portfolios. This result is in contrast with the one obtained for directional hedge funds 402 . 1995). On one hand adding small allocations to directional hedge funds does not significantly change the risk-return profile of the global portfolio but this changes when we attain 20% an allocation of 20% to directional hedge funds. For high allocation to the risky asset undirectional strategies do not help diversifying and reaching higher return levels. undirectional hedge funds and fund of funds can be grouped together because they have the same pattern. Over a 50% allocation to directional hedge funds provides a significantly more attractive returns in all cases. When the risk measure is limited to volatility. This is the basic reason why traditional tools like the mean-variance efficient frontier analysis should not be used to analyse alternative assets such as hedge funds. On the other hand.V Conclusion Hedge funds exhibit abnormal returns.

Undirectional strategies and funds of funds are low risk profile investments and should be used as such. Second. it determines if hedge funds must be added to the existing portfolio and in the future we should be able to determine what hedge fund strategy should be favoured. 403 . The next step will be first to distinguish between hedge fund strategies while determining the impact of inserting hedge funds on the adapted efficient frontier.and confirms the need to differentiate directional from unidirectional hedge funds. A complementary analysis on the impact of the insertion of mutual funds in hedge fund portfolios confirms that mutual funds do not significantly diversify hedge fund portfolios when an extended risk measure is considered. Based on the clients’ objective and the market conditions. the identical tests should be performed on various analysis period to determine what kind of fund should be favoured in specific market conditions. The new adapted efficient frontier opens new doors for asset allocators. Another important future step would be to go into more precision regarding the estimation of the aversion to risk in order to enable investors to precisely determine if or not they should use hedge funds.

General Conclusion .

to risk analysis. and so on. Our first objective is to clearly understand hedge fund managers and to explain how they create alpha over time. test and improve an extended multi-factor performance analysis model in order to understand hedge fund performance on the one hand. This result indicates that hedge funds have been attractive investments in various market conditions. Three studies cover this aspect of the analysis (Analysis of Hedge Fund Performance. Hedge Fund Performance and Persistence in Bull and Bear Markets and Sustainability in Hedge Fund Performance: New Insights). understand how manager create performance over time and if the extracted is consistent over time. We analyse hedge fund returns. To extract pure alpha we develop. We divide this Thesis in three complementary parts. Depending on the period under review and on the strategy considered between 20% and 50% of the individual managers do outperform classical indices. With these results in hand. Almost every aspect of the hedge fund world has been studied from performance decomposition.General Conclusion Hedge funds have attracted the attention of many academicians over the last decade. we looked for a systematic way of investing in 405 . regulatory aspects. Globally. strategy explanation. Our contribution is on the quantitative analysis of hedge fund strategies. our results confirm that some hedge fund managers do create pure alpha over time and persistently do so independently of the market conditions considered and of the performance decomposition model considered. while on the other to develop and adapt a methodology to determine whether there is any persistence in hedge fund returns. The purpose of this doctoral thesis is clearly established: to understand hedge fund strategies by looking at the number produced.

hedge funds that would enable the investor to consistently and significantly outperform the classical markets over time. Our first persistence analysis based exclusively on performance showed that around half of the funds create significant and persistent alpha over time and that these funds tend to be low volatile funds with a limited exposure to the equity market. Then, we performed a further analysis that uses other tools than simply returns (like the Sharpe ratio, standard deviation, alpha, beta, skewness, kurtosis and an adapted Sharpe ratio ) as a way of determining if hedge funds offer consistence in alpha creation. We find a consistent, systematic way of creating pure alpha using a simple classification methodology based on basic statistics: risk-return trade-off measure (the Sharpe score), volatility and to a lesser extent, the beta exposure appear to be the best and most stable way of classifying hedge funds in order to detect persistency in the returns. Funds offering the highest Sharpe score, funds with a limited volatility and/or funds with a limited exposure to the equity market consistently and significantly outperform equity and bond markets. These results hold not only for a full market cycle, but also when separating bull and bear market conditions. This analysis is of particular interest because it clearly proves that some funds consistently and significantly outperform classical markets. The important element used to detect these funds is the methodology by which they are classified. The hedge fund industry may in attractive and outperform classical markets if the investors select the good managers. The second objective analysed is clearly linked to the first. We perform a specific analysis on the most represented strategy: market neutral funds. By definition, market neutral funds have a limited relative exposure to the equity market. We check for this neutrality and analyse what kind of funds consistently outperform over time: the pure market neutral funds, market timers or funds with a more directional bias. This aspect is analysed in part 2 (Analysis of Hedge Fund’s Market Exposure).

406

The core of the study is based on a methodology that uses classical exposure measures like the beta in an original way. Our analysis of market neutral funds indicates first that the betas obtained are low on an absolute term but significantly positive and that the more volatile funds tend to have the highest market exposure. Second, individual fund analysis results indicate that around one third of the funds have been significantly exposed to the market while two thirds of the alphas have been significantly positive. Finally, sub-period analysis confirm that market neutral funds tend to have a higher exposure to the markets in market turmoil and no exposure in strong bull markets even if they are some exceptions. This result confirms that the correlation of market neutral fund tend to increase when the market in going down even if there may still be alpha creation depending on the sub-strategy considered. On the other hand, they were few funds significantly positively exposed to the market during the bullish period. This relation is not what investors are looking for but as many market neutral funds continue to create significant alpha there is a trade-off between the increase in the beta (unwanted added risk) and value creation (significantly positive alpha). Another time, fund selection remains the central point. Finally, the third complementary objective of the thesis is to determine whether hedge fund strategies should be included in a classical portfolio of stocks and bonds. In Part three, Diversifying Using Hedge Funds: A Utility-Based Approach, we analyse the inclusion of hedge funds in a portfolio of stocks and bonds. The main originality of this study centres upon the development of a new efficient frontier, based not only on volatility but also on higher moments (skewness and kurtosis) and on a utility function that more closely corresponds to that of the investor without normality or other strong assumption. We develop a adapted capital market line that includes not only the volatility in the risk measure but that integrates the asymmetry of returns and the fat tails in order to determine if investors should include hedge funds in their portfolio once almost

407

the whole hedge fund return distribution characteristics are taken into account. Our results confirm that hedge funds are attractive investment tools for almost any kind of investors but that investors specific needs have to be considered in order to determine what kind of funds should be used and what allocation should be given to hedge funds. Hedge funds can be good investments on a stand alone basis and as a diversification tool but everything depends on the way of investing in hedge funds and on the quality of the selection of the underlying hedge fund managers.

408

References

Abdi, H., 2007, Kendall rank correlation, in N.J. Salkind (Ed.): Encyclopedia of Measurement and Statistics. Thousand Oaks (CA): Sage. Ackermann, C., McEnally, R., Ravenscraft, D., 1999, The Performance of Hedge Funds: Risk, Return, and Incentives, Journal of Finance 54, n°3, 833-874. Ackermann, C., Ravenscraft, D., 1998, The Impact of Regulatory Restrictions on Fund Performance: A Comparative Study of Hedge Funds and Mutual Funds, Working Paper, University of Notre-Dame. Adcock, C., 2006, Asset Pricing and Portfolio Selection based on the Multivariate SkewStudent Distribution, in Multi-moment Asset Pricing Models and Related Topics, ed. Bertrand Maillet and Emmanuel Jurczenko, Wiley. Agarwal, V., 2001, Intertemporal Variation in the Performance of Hedge Funds employing a Contingent-claim-based Benchmark, Working Paper, London Business School. Agarwal, V., Naik, N.Y., 2000, Multi-period Performance Persistence Analysis of Hedge Funds, Journal of Financial and Quantitative Analysis 35, n°3, 327-342. Agarwal, V., Naik, N.Y., 2004. Risks and Portfolio Decisions Involving Hedge Funds, Review of Financial Studies 17, n°1, 63-98. Allen, D. E., Soucik, V., 2000, In Search of true Performance: Testing Benchmark-model Validity in Managed Funds Context, Working Paper, Edith Cowan University. Amenc, N., Curtis S., and Martellini L., 2002, The Alpha and Omega of Hedge Fund Performance Measurement, Working Paper, Edhec Risk and Asset Management Research Centre. Amenc N., Martellini, L., 2002, Portfolio Optimization and Hedge Fund Style Allocation Decision, Journal of Alternative Investment 5, n°2, 7-20.

Amenc N., El Bied, S., Martellini, L., 2003, Predictability in Hedge Fund Returns, Financial Analysts Journal 59, n° 5, 32-46. Amenc N., Martellini L., 2003, The Brave New World of Hedge Funds Indices, EDHEC Risk & Asset Management Research Centre, Working Paper. Amenc N., Martellini, L., Vaissié, M., 2002, Benefits and Risks of Alternative Investment Strategies, Journal of Asset Management 4, n°2, 96-118. Amin, G.S., Kat, H.M., 2003a, Hedge Fund Performance 1990-2000: Do the ‘Money Machine’ really add Value?, Journal of Financial and Quantitative Analysis 38, n°2, 251274. Amin, G. S., Kat, H. M., 2003b, Stocks, Bonds, and Hedge Funds, Journal of Portfolio Management 29, n°4, 113-120. Arrow, K.J, 1971, The Theory of Risk Aversion, in Essays in the Theory of Risk-Bearing. Chicago: Markham publishing. Co, 91-120. Assness, C, Krail, R., Liew, J., 2001, Do Hedge Funds Hedge, Journal of Portfolio Management 28, n°1, 6-19. Athayde, G., Flôres, R., 2002, Portfolio Frontier with Higher Moments: the Undiscovered Country, Discussion Paper EPGE-FGV. Bams, D., Otten, R., 2002, European Mutual Fund Performance, European Financial Management 8, n°1, 75-101. Banz, R.W., 1981, The Relationship between Returns and Market Value of common Stocks, Journal of Financial Economics 9, n°1, 3-18. Barès, P-A, Gibson, R., Gyger, S., 2003, Performance in the Hedge Fund Industry: An Analysis of Short and Long Term Persistence, The Journal of Alternative Investments 6, n°3, 25-41.

Basu, S., 1983. The Relationship between Earnings Yield, Market Value, and Return for NYSE Common Stocks: Further Evidence, Journal of Financial Economics 12, n°1, 129156. Bekaert, G, Urias, M., 1996, Diversification, Integration and Emerging Market Closed-end Funds, Journal of Finance 51, n°3, 835-869. Bell, D.E, 1988, One-Switch Utility Functions and a Measure of Risk, Management Science 34, n°12, 1416-1424. Bell, D.E, 1995, Risk, Return and Utility, Management Science 41, n°1, 23-30. Bhandari, L.C., 1988, Debt/equity Ratio and expected Common Stocks Returns: Empirical Evidence, Journal of Finance 43, n°2, 507-528. Bhardwaj, R. K., Brooks, L. D., 1992, The January Anomaly: Effects of low Share Price, Transaction Costs, and Bid-ask Bias, Journal of Finance 47, n°2, 553-575. Brealy, R.A, Kaplanis, E., 2001, Hedge Funds and Financial Stability: An Analysis of their Factor Exposures, International Finance 4, pp. 161-187. Breeden, D.T., 1979, An Intertemporal Asset Pricing Model with Stochastic Consumption and Investment Opportunities, Journal of Financial Economics 7, n°3, 265-276. Breeden, D.T., Gibbons, M.R., Litzenberger, R.H., 1989, Empirical Tests of the Consumption-oriented CAPM, Journal of Finance 44, n°2, 231-296. Brown, S.J., Goetzmann, W.N., 1995, Performance Persistence, Journal of Finance 50, n°2, 679-698. Brown S.J., Goetzmann, W.N., Hiraki, T., Otsuki, T., Shiraishi, N., 2001, The Japanese open-end Fund Puzzle, Journal of Business 74, n°1, 59-78. Brown S.J., Goetzmann, W.N., Ibbotson, R.G., 1999, Offshore Hedge Funds: Survival and Performance 1989-1995, Journal of Business 72, n°1, 91-118.

Brown S.J., Goetzmann, W.N., Liang, B., 2002, Fees on Fees in Funds-on-Funds, Journal of Investment Management, 2, n°4, 39-56. Brown S.J., Goetzmann, W.N., Park, J., 2001, Careers and Survival: Competition and Risk in the Hedge Fund and CTA Industry, Journal of Finance 56, n°5, 1869-1886. Brooks C., Kat, H., 2002, The Statistical Properties of Hedge Fund Index Returns and their Implications for Investors, Journal of Alternative Investment 5, n°2, 26-44. Capocci D., 2004a, Inserting Convertible Arbitrage Funds in a classical Portfolio, in Hedge Funds: Strategies, Risk Assessment, and Returns, Beard Books, Washington. Capocci D., 2004b, CTA Performance, Survivorship Bias and Dissolution Frequencies, in Commodity Trading Advisors: Risk, Performance, Analysis and Selection, John Wiley and Sons, New York. Capocci, D., Hübner, G., 2001, L’Univers des Hedge Funds, une Perspective Empirique, Revue Bancaire et Financière, September, n°6, 361-369. Capocci D., Hübner, G., 2006, Funds of Hedge Funds: Bias and Persistence in Returns, in Fund of Hedge Funds: Performance, Assessment, Diversification and Statistical Properties, Elsevier Press. Capocci D., Nevolo, V., 2005, Funds of Hedge Funds versus Portfolios of Hedge Funds, in Hedge Funds: Insights in Performance Measurement, Risk Analysis, and Portfolio Allocation, John Wiley & Sons, New York. Carhart, M., 1995, Survivor Bias and Mutual Funds Performance, Ph.D. Dissertation, School of Business Administration, University of Southern California. Carhart, M.M., 1997, On Persistence in Mutual Fund Performance, Journal of Finance 52, n°1, 57-82. Carhart, M.M., Carpenter, J.N., Lynch, A.W., Musto, D.K., 2002, Mutual Fund

Survivorship, Review of Financial Studies 15, n°5, 149-1463.

Chan, L.K., Jegadeesh, N., Lakonishok, J., 1996, Momentum Strategies. Journal of Finance 51, n°5, 1681-1714. Chang C., T. Pactwa and A. Prakash, 2003, Selecting a Portfolio with Skewness: Recent Evidence from US, European, and Latin American Equity Markets, Journal of Banking and Finance 27, n°7, 1375-1390. Chen, Y, Liang, B., 2005, Do Market Timing Hedge Funds Time the Market?, Journal of Financial and Quantitative Analysis 40, n°1, 493–517. Chunhachinda, P., Dandapani, K., Hamid, S., Prakash, A.J., 1997, Portfolio Selection and Skewness: Evidence from International stock markets, Journal of Banking and Finance 21, n°2, 143-167. Cohn, R. A., Lewellen, W. G., Lease R. C., Schlarbaum, G. G., 1975, Individual Investor Risk Aversion and Investment Portfolio Composition, The Journal of Finance 5, n°2, 605-620. Daniel, K., Grinblatt, M., Titman, S., Wermers, R., 1997, Measuring Mutual Fund Performance with Characteristic-based Benchmark, Journal of Finance 52, n°3, 10351058. Davies, R., H. Kat, Lu, S., 2006, Single Strategy Fund of Hedge Funds: How Many Funds?, in Fund of Hedge Funds: Performance, Assessment, Diversification and Statistical Properties, ed. Greg N. Gregoriou, Elsevier, 203-210. DeSantis, G., 1993, Volatility Bounds for Stochastic Discount Factors: Tests and Implications from International Financial Markets, PhD Thesis. Dittmar, R. F., 2002, Nonlinear Pricing Kernels, Kurtosis Preference, and Evidence from the Cross Section of Equity Returns, Journal of Finance 57, n°1, 369-403. Edwards F. R., Caglayan, M. O., 2001, Hedge Fund and Commodity Fund Investments in Bull and Bear Markets, Journal of Portfolio Management, 27, n°4, 97-108.

1243-1250. Journal of Finance 51. Das. Fama. Journal of Finance 34.. Blake.. Ennis. Efficiency with costly Information: A Reinterpretation of Evidence from Managed Portfolios. E. M. 1993. Elton. Gruber. 2003. Sebastian.. Financial Review 36. n°2.. Gruber.F. 2001. Fama... 1996. Hlavka. Journal of Finance 47. Review of Financial Studies 6... Mutual Fund Systematic Risk for Bull and Bear Markets: An Empirical Examination. Multifactor Explanations of Asset Pricing Anomalies. 1992.. n°6. 1993. 1979. Fama.R. K. Fabozzi. Das. F. J. C. B. Fama.. French. n°2. Elton. C. E. . n°1. Journal of the American Statistical Association. The Cross-section of expected Returns. n°382.R.R. M. 316-331. Efron.. B.. An Introduction to the Bootstrap. 55-84. S. M. Common risk Factors in the Returns on Stocks and Bonds. 157-174. n°1. 1979. K.. A Multivariate Test of a Dual-Beta CAPM: Australian Evidence. The Annals of Statistics 7. M. 103-112. 3-56. Journal of Business 69... Journal of Finance 53. 1998.. R. Journal of Portfolio Management 29. Journal of Financial Economics 33. 427-465. E. R. The Persistence of Risk-adjusted Mutual Fund Performance. 78.. French... 1975-1999. E. 1996. K. Value versus Growth: The International Evidence. D. and Tibshirani. n°1.F. E. M. n°4. 1993. n°1.J. 1983. R. E.R. Bootstrap Methods: Another Look at the Jackknife. and Francis J. London: Chapman & Hall.. French.F. Efron. French. 1-21. n°5. 1-26.. S..Efron. A Critical Look at the Case for Hedge Funds. Faff.. 133-157. Estimating the Error Rate of a Prediction Rule: Improvement on Crossvalidation. B.F. K.

. I. Empirical Characteristics of Dynamic Trading Strategies: The Case of Hedge Funds. K.A. 900-922. Fung. Performance Characteristics of Hedge Funds and Commodity Funds: Natural vs.A. Fung. Fung.. The American Economic Review 65. 1998. Monash University. Hsieh.F. Duke University. Faff...A.. Hsieh. Characteristics. D. Hsieh. Fung. Financial Analyst Journal 58. n°2.. Hsieh. Fang. Durham. 1997. French. W. The Risks in Fixed Income Hedge Fund Styles. 389-406. Hsieh.. 1975. n°3.. n°1.R. Economic Review 91.A. E. n°1. 2000. Blume.A. 313-341. n°4 Galadera. 1997.. 2001. W. 293-307. Hsieh. Review of Financial Studies 14... Review of Financial Studies 10. 1999. The Risk of Hedge Funds Strategies: Theory and Evidence from Trend Followers. Fung.U. H. 2002b.. 275-302. n°2. Benchmarks of Hedge Fund Performance: Information Content and Measurement Biases. T. W... 2004. Journal of Finance 55. Fung. n°2. spurious Biases. D. D. D. . 291-307. D. D. Lai.. Fung. Hedge Funds: An Industry in its Adolescence. W.A. NC. R. Covariances. 53-58. 2002a.A. Modelling the Risk and Return Relation Conditional on Market Volatility and market Conditions. Friend..A. The Demand for Risky Assets. Is Mean-Variance Analysis applicable to Hedge Funds?. Journal of Quantitative and Financial Analysis 35. Co-Kurtosis and Capital Aseet Pricing. 1997. W. W. Hsieh.. and Average Returns: 1929-1997. D.. W. Fung. The Financial Review 32. Journal of Fixed Income 12.. D. W. 2000. D.A. 2006. n°2. 6-27. A Risk Neutral Approach to Valuing Trend Following Trading Strategies. and M.Fama. Working Paper.. 22-34. Working Paper 08/04. Hsieh. Economics Letters 62. n°5. E.

1029-1054. Titman. Journal of Finance 51. Journal of Portfolio Management 20. The Persistence of Mutual Fund Performance. R. W. Momentum Investment Strategies. Yale School of Management.... n° 2. 1994. 2006. Duke University. M. J. Hansen. Titman. W.N. mimeo. Pairs Trading: Performance of a Relative Value Rule. Grinblatt. 2002.I. Rouah.. and P. Rouwenhorst..N. Goetzmann.N. 51-56 Grinblatt. Econometrica 50... 1999. I. Gregoriou... Random Walk Behavior of CTAs Returns. G. S. W. M. n°3. S. 1994. 419-444. Goetzmann.. J.E. R. M. S. n°5.. Welch. Yale ICF Working Paper No... Journal of Alternative Investments 6. Goetzmann.. Large Sample Properties of the Generalised Method of Moments Estimators.. . n°5. 1088-1105. F. Liechty. M. Titman. 1977-1984.. Ibbotson. Ingersoll. M. L.Gatev E. 2003. 1995. Portfolio Performance and Herding: A Study of Mutual Fund Behaviour.. Journal of Financial and Quantitative Analysis 29. Liechty. Grinblatt.G. Grinblatt. 393-416. M. S. 1992. Do Winners Repeat? Patterns in Mutual Performance. Spiegel. 1989. working paper. n°4. 02-08. n°3.. Another Puzzle: the Growth in Actively Managed Mutual Funds. American Economic Review 85. Müller. J. Portfolio Performance Manipulation and Manipulation-Proof Performance Measures. Mutual Fund Performance: An Analysis of Quarterly Portfolio Holdings.. 9-18. 1982. n°2. C. Portfolio Selection with Higher Moments. Journal of Finance 47. Harvey. Journal of Business 62. Titman... Wermers. n°3. M. 1996. A Study of Monthly Mutual Fund Returns and Performance Evaluation Techniques. Gruber. 783-810.. K..

W.. Patel.. ICI Statistics and Research quarterly report. Rockinger. Journal of Finance 48. 2006. Hendricks..C. the Pricing of Capital Assets and Evaluation of Investment Portfolios.. G. Jensen. Risk. M. D. R. Journal of Business 42. R. Optimal Portfolio Allocation Under Higher Moments.C. Tests of Mean-Variance Spanning. 127-144. A.A. 1968. 1993. www. R.. Wiley.. Maillet. International Mutual Fund Selectivity and Market Timing during Up and Down Markets. n°2. 2006. W. Journal of Finance 48... Worldwide Mutual Funds Assets and Flows. . Kan. A. n°2. Zeckhauser. 29-55. n°1.. N. G. 2001. 93-130. 1998..Harvey. in Multi-moment Asset Pricing Models and Related Topics. 2000. Jondeau. 1989. M. B. The Four-Moment Capital Asset Pricing Model: between Asset Pricing and Asset Allocation. E. Second Quarter 2003. Siddique.C. n°1... Assessing the Market Training Performance of Managed Portfolios.. Washington University Working Paper. Financial Review 33. Journal of Finance 23. 2003. S. Chan K. Quarterly Journal of Economics 104. Jagannathan. 167-247.ici.. Titman.. M. J. n°2. Zhou. n°1.. n°1. Returns to Buying Winners and Selling Losers: Implications for Stock Market Efficiency. Jurczenko. Journal of Finance 55. E. 1993.. n°2. 1969.org/stats. L. European Financial Management 12. The Performance of Mutual Funds in the Period 1945-1964.T. Cheng. Bertrand Maillet and Emmanuel Jurczenko. Investment Company Institute. The Journal of Business 59. Conditional Skewness in Asset Pricing Tests. 217-235. Efficiency with Costly Information: A Study of Mutual Fund Performance. Ippolito. 389-416.. 1-24. Kao. Jegadeesh. n°1. Hot Hands in Mutual Funds: Short-run Persistence of Performance: 1974-88. 93-130... 1986. ed. C. 1263-1295. R. Jensen. R. Korajczyk.

n°4. C. Skewness Preference and the Valuation of Risk Assets. H. On the Performance of Hedge Funds. n°4. M. Charles Griffin & Company Limited. Cass Business School. Keating. n°4. 1987. 2005. Modest.P. S. n°5. 1985-2010.. n°4. Harry. American Economic Review 69. n°3. Evaluating Mutual Fund Performance. 589-611. 1938. 293-305. Stocks. Journal of Finance 42. 81-89. 2001. Hedge Fund Performance: The Role of Non-Normality Risks and Conditional Asset Allocation. M...N. n°2. 1994.. Journal of Finance 56. T. H. 113-120. Lintzenberger. Review of Quantitative Finance and Accounting 1.. Kothari. 1993. Shadwick. 59-84 Kendall. H. n°3. R. 1948.. Portfolio Selection with Skewness: A Multiple-objective Approach. Journal of Portfolio Management 29. Kramer. 72-85. 233-245. Kendall. 30. Bonds. Working Paper WP-FF-21-2005. Econometrica 61. The diversification benefits of hedge funds and funds of hedge funds. Journal of Performance Measurement 6. Kat. A Universal Performance Measure. 1991. 1999.B. Trading & Regulation 12.Y.Kat. Lehmann. n°1. Markowitz.. Approximating Expected Utility by a Function of Mean and Variance. W. 308-317. Lai. B. 2002. Journal of Finance 31. Warner. D... A. M. 1883-1891. n°5.. 1979.M.. n°3. M. Levy. Standard Risk Aversion. 290-300.. Kraus.. A New Measure of Rank Correlation.. 1085-1100. G. S. 2007.. Mutual Fund Performance Evaluation: A Comparison of Benchmarks and Benchmarks Comparisons. Biometrica. and Hedge Funds.. 1976. Kooli. C. Macroeconomic Seasonality and the January Effect.. J. Journal of Finance 49. Joëlle Miffre. Derivatives Use.. . M. Amin. M. 2003. Financial Analysts Journal 55. B. Kimball. Rank Correlation Methods. Liang. H.

Cleveland. 1965. Financial Analyst Journal 57.G. 1976. The Effects of Personal Taxes and Dividends on Capital Asset Prices: Theory and Empirical Evidence. Hedge Funds: The Living and the Dead. Litzenberger R. n°2. B.. 2000. A. n°2. Liew.. 309-325.. Journal of Portfolio Management 29. 2003. J. 1995. 16-33. Remolona E. Liang. n°1.. n°6... 2001. and Funds-of-Funds. 80-88. A. Case Western Reserve University. Working Paper. The American Economic Review 66. The Erroneous Approximation of Expected Utility by Means of a Taylor’s Series Expansion Analytic and Computational Results. 163-195. Journal of Finance 37. Financial Analyst Journal 61. Malkiel. Risk Management for Hedge Funds: Introduction and Overview. K. Journal of Financial and Quantitative Analysis 35. Lo. Tsatsaronis. Litzenberger R. 113-123. Financial Analysts Journal 57. Loistl. 549-572.. The Valuation of Risk Assets and the Selection of Risky Investments in Stock Portfolio and Capital Budgets. K. 2001. Hedge Funds. Review of Economics and Statistics 47.. On the Performance of Alternative Investments: CTAs.Liang. .. Hedge Funds: Risk and Return. Hedge Fund Index Investing Examined. The Effects of Dividends on Common Stocks Returns: Tax Effects or Information's Effects ?. Returns from Investing in Equity Mutual Funds 1971 to 1991. Liang. J. P. 1982. 904-910. Malkiel. 429-443. B. B. Time-Varying Exposures and Leverage in Hedge Funds . 59-72.. K. Hedge Funds Performance: 1990-1999.. n°6.. Ramaswamy. 13-37. n°3. B. Journal of Finance 50.. Journal of Financial Economics 7. BIS Quarterly Review. Lintner.. 2005. 2005. B. O.. Ramaswamy. 11-18. n°5. 1979. McGuire. n°2. 2003. n°2. Saha. n°1..

Signal Processing and Dynamics. Reinganum. U. working paper ABP Invstments. West. Morin.S. G. Journal of Financial and Quantitative Analysis 16. Navone. Journal of Finance 54. n°3. J. 2005. 1987. Posthuma. n°2. Gregoriou. 2135-2175. n°4. Ph. 1999. 2001. 800pages. A. . Werker. 1995. Risk Aversion Revisited. n°2. The Journal of Finance 56.R. M. Pollock. Managed Futures as an Investment Set..D. 1201-1216. A new Empirical Perspective on the CAPM. Mutual Fund Performance during Up and Down Market Conditions. 703-708. Characteristics of Risk and Return in Risk Arbitrage. P.. Academic Press... in Hedge Funds. M. W. N. 1981. N. Pulvino.. Hübner. n°5. Van der Sluis. Rao.. Performance Evaluation with Transactions Data: The Stock Selection of Investment Newsletters. Wiley.. B. 2001. Newey. G. 439-462. F. Mitchell. J.. The Journal of Finance 38. K. 1999. Nijman.. 721-742. A Handbook of Time-series Analysis.Metrick E. P. A Simple Positive-definite Heteroskedasticity and Autocorrelation Consistent Covariance Matrix.. 62-65.G. 2001. Econometrica 55. 2003. Rouah. R. n°6. F.. 1743-1775. T. Journal of Finance 56. Diversifying Market Risk through Market Neutral Strategies. Review of Business 22. Posthuma.. N.. M. A Reality Check on Hedge Fund returns. 1983. Van der Sluis. J. Columbia University. Analyzing Style Drift in Hedge Funds ». Roon. T.. A. Fernandez Suarez. Park. Dissertation. D. Ed. n°4. Testing for Mean-Variance Spanning with Short Sales Constraints and Transaction Costs: The Case of Emerging Markets.. 2001.. Papageorgiou... working paper.. IEIF Bocconi University.

Multifactor Analysis of hedge funds. Costly Search and Mutual Fund Flows. Quantitative Analysis of Hedge Funds and Managed Futures Return and Risks Characteristics. Van Hedge Funds Advisors International. University of Bayreuth.. Persuasive Evidence of Market Inefficiency. Tiburg University.hedgefund. Asset Allocation: Management Style and Performance Measurement. K. W. Return-based Style Analysis with Time- varying Exposures.. discussion paper 96. n°3.. T. Sharpe. Journal of Portfolio Management 11. Reid. Sharpe.. 1998. Hemisphere Publishing Corporation.. 2002. Journal of Finance 53. 7-19. Yan.. 1589-1622. Size of the Hedge Fund Universe. 1964.. Journal of Finance 19. R. www. L. Center for Economic Research.R. E. Sirri. Institutional Investor Publishing.. R.P. T.. Spurgin. Q. n°1. Van der Sluis. . Sun..A. Spurgin. 2003. R.J. Oldham. 1985. Journal of Banking and Finance 27. 1992. Report.. W. K. Tufano. 1-24. Department of Mathematics. 2002. Swinkels. in Evaluating and Implementing Hedge Fund Strategies. K.F.Rosenberg. n°5. n°3.F. Managed Futures and Mutual Fund Return and Risk Characteristics. An Atlas of Functions.. 425-442. 1111-1121. B.com/abouthfs/universe/universe. Schneeweis. 1987. Lanstein.htm. Schittkowski. Schneeweis. P. 1998. Skewness Persistence with Optimal Portfolio Selection. n°2.. Capital Asset Prices: A Theory of Market Equilibrium under Conditions of Risk. Spanier. Journal of Alternative Investments 1. 9-17. Y. n°6. J. Implementation of a Sequential Quadratic Programming Algorithm for Parallel Computing. 3rd Edition: The Experience of Managers and Investors. Journal of Portfolio Management 18. 2001. P. 2000.

A. Is Money Smart? A Study of Mutual Fund Investor’ fund Selection Ability. University of Colorado. Wermers. Journal of Portfolio Management 28. 1999. 901-933.. . L. Journal of Finance 54.. Momentum Investment Strategies of Mutual Funds.. Performance Persistence. Working paper. Zheng. and Survivorship Bias.Weisman. R. 1996. 2002. n°4. Graduate School of Business and Administration. 80-91. Informationless Investing and Hedge Fund Performance Measurement Bias. n°3.

....................... 108 FIGURE 5: DECILE REPARTITION PER STRATEGY (RANKING BASE ON ALPHA) ..................................................... 362 ................................... 229 FIGURE 9: MINIMUM.................... 226 FIGURE 6: STRATEGY REPARTITION IN THE ALPHA RANKING .......................................................................... 228 FIGURE 8: STRATEGY REPARTITION IN THE BETA RANKING ............................................................... 17 FIGURE 3: MONTHLY RETURNS FOR DEAD FUNDS TOWARDS THE DISSOLUTION DATE .......................................................... 65 FIGURE 4: HEDGE FUNDS DISSOLUTION FREQUENCIES .............. MAXIMUM......... MEDIAN AND MEAN INDIVIDUAL CORRELATIONS BETWEEN MARKET NEUTRAL FUNDS AND THE EQUITY INDEX ............................................. 292 FIGURE 10: LAGGED 12-MONTH DECILE DESCRIPTION ......An Analysis of Hedge Fund Strategies – List of Figures FIGURE 1: FOUR GLOBAL CATEGORIES OF HEDGE FUND ACADEMIC STUDIES ....................... 227 FIGURE 7: DECILE REPARTITION PER STRATEGY BASED ON BETA .......................................................... 304 FIGURE 11: ILLUSTRATION OF THE ADAPTED CAPITAL MARKET LINE . 7 FIGURE 2: HEDGE FUND DATABASE UNIVERSE REPARTITION ...

... 375 FIGURE 14: REPARTITION OF THE KURTOSIS OF INDIVIDUAL FUNDS ......................MEAN PERCENTILE RETURN . 363 FIGURE 13: REPARTITION OF THE SKEWNESS OF INDIVIDUAL FUNDS .................... 383 FIGURE 16: UNDIRECTIONAL HEDGE FUNDS ....... 376 FIGURE 15: DIRECTIONAL HEDGE FUNDS ............... 386 ..................MEAN PERCENTILE RETURN ...An Analysis of Hedge Fund Strategies – List of Figures FIGURE 12: ILLUSTRATION OF THE TEST OF THE ADAPTED CAPITAL MARKET LINE ............................. 385 FIGURE 17: FUNDS OF HEDGE FUNDS .......................................................................................MEAN PERCENTILE RETURN ..

.................................................................................................................... 77 TABLE 14: PERFORMANCE MEASUREMENT USING THE CAPM..................................................................................... OBJECTIVES AND CONCLUSIONS ............... OBJECTIVES AND CONCLUSIONS .. 71 TABLE 13: CORRELATION BETWEEN HEDGE FUNDS AND PASSIVE INVESTMENT STRATEGIES ...................................... 82 .. 37 TABLE 8: HEDGE FUND AS DIVERSIFICATION TOOLS SPECIFICITIES........................... 31 TABLE 6: PERSISTENCE ANALYSIS STUDIES SPECIFICITIES................................................................. 18 TABLE 3: BACKFILL BIAS AND SURVIVORSHIP BIAS ESTIMATION ... .................................................... 35 TABLE 7: NEUTRALITY OF MARKET NEUTRAL FUND SPECIFICITIES.............An Analysis of Hedge Fund Strategies – List of Tables TABLE 1: STUDIES SPECIFICITIES AND OBJECTIVES ............... 69 TABLE 12: DESCRIPTIVE STATISTICS OF HEDGE FUNDS STRATEGIES AND PASSIVE INVESTMENT STRATEGIES ............. 20 TABLE 4: HEDGE FUND INDICES COMPARISON ................ CARHART’S 4-FACTOR MODEL AND THE COMBINED MODEL............. 40 TABLE 9: SURVIVORSHIP BIAS IN HEDGE FUNDS ............................... 24 TABLE 5: MULTI-FACTOR PERFORMANCE DECOMPOSITION MODEL ..... OBJECTIVES AND CONCLUSIONS .................................................................................................... 4 TABLE 2: DATABASE COMPARISON .. 63 TABLE 11: ESTIMATION OF INSTANT RETURN HISTORY BIAS .................

.................... 152 TABLE 23: PERFORMANCE MEASUREMENT USING THE CAPM...... 144 TABLE 21: SURVIVORSHIP BIAS IN HEDGE FUNDS ............................. CARHART’S 4-FACTOR MODEL AND THE COMBINED MODEL ...................... 110 TABLE 18: DESCRIPTIVE STATISTICS OF HEDGE FUNDS STRATEGIES AND PASSIVE INVESTMENT STRATEGIES ................................................. 101 TABLE 17: HEDGE FUNDS STRATEGY PERSISTENCE BASED ON 12 MONTH LAGGED RETURNS ............ 162 TABLE 25: HEDGE FUNDS PERSISTENCE BASED ON 12 MONTH LAGGED RETURNS ............................................................................................... 172 ......................... 138 TABLE 20: CORRELATION AMONG HEDGE FUNDS............................. BETWEEN HEDGE FUNDS AND PASSIVE INVESTMENT STRATEGIES........................... 97 TABLE 16: HEDGE FUNDS PERSISTENCE BASED ON 12 MONTH LAGGED RETURNS....................................................................... 133 TABLE 19: DESCRIPTIVE STATISTICS OF HEDGE FUNDS STRATEGIES FOR THE BULLISH AND BEARISH SUBPERIODS ................ 149 TABLE 22: ESTIMATION OF INSTANT RETURN HISTORY BIAS ................................................................An Analysis of Hedge Fund Strategies – List of Tables TABLE 15: PERFORMANCE OF HEDGE FUNDS IN DIFFERENT SUB-PERIODS ..................................... AND AMONG PASSIVE INVESTMENT STRATEGIES ............. 156 TABLE 24: PERFORMANCE OF HEDGE FUNDS DURING THE BULLISH AND BEARISH SUB-PERIODS ......................

.................... 238 TABLE 35: PERSISTENCE IN PERFORMANCE ANALYSIS BASED ON THE ALPHA (1/1997-12/2002) ................................. 204 TABLE 30: HEDGE FUND STRATEGIES PERFORMANCE ANALYSIS (1/1994-12/2002) ..........................................................................................................................................................................................An Analysis of Hedge Fund Strategies – List of Tables TABLE 26: HEDGE FUNDS PERSISTENCE DURING THE BULLISH AND BEARISH SUB-PERIODS. 212 TABLE 31: HEDGE FUND PERSISTENCE IN PERFORMANCE (1/1995-12/2002)......... 221 TABLE 33: PERSISTENCE IN PERFORMANCE ANALYSIS BASED ON THE SHARPE SCORE (1/199712/2002)................................ 198 TABLE 29: DESCRIPTIVE STATISTICS ......................................................................................................... 245 TABLE 37: PERSISTENCE IN PERFORMANCE ANALYSIS BASED ON THE SKEWNESS AND KURTOSIS (1/1997-12/2002) ............................................... 236 TABLE 34: PERSISTENCE IN PERFORMANCE ANALYSIS BASED ON THE STANDARD DEVIATION (1/199712/2002)................ 175 TABLE 27: HEDGE FUNDS PERSISTENCE FOR THE MARKET NEUTRAL STRATEGY .................... 243 TABLE 36: PERSISTENCE IN PERFORMANCE ANALYSIS BASED ON BETA (1/1997-12/2002) ....................................... 247 ............. 218 TABLE 32: PERSISTENCE IN PERFORMANCE ANALYSIS BASED ON THREE-YEAR DATA (1/199712/2002)............... 180 TABLE 28: CORRELATION BETWEEN THE OPTION FACTORS AND THE MARKET FACTOR ............................

............................................................ 290 TABLE 43: BIRTH AND ATTRITION RATES .......... 256 TABLE 40: KENDALL TAU ESTIMATION ............................................................................................................ 286 TABLE 42: CORRELATION BETWEEN MARKET NEUTRAL STRATEGIES AND EQUITY INDEX ........................................ 310 TABLE 49: BULL AND BEAR BETA ESTIMATION ......................................................................................................... 297 TABLE 45: MARKET EXPOSURE ANALYSIS........................ 293 TABLE 44: SURVIVORSHIP BIAS ...................................................... 306 TABLE 47: INDIVIDUAL FUND MARKET EXPOSURE ......................... 301 TABLE 46: LAGGED 12-MONTH DECILE EXPOSURE ........................................................................... 254 TABLE 39: SUMMARY OF ALPHA CREATION AND EQUITY EXPOSURE ............................. 316 ........................ 311 TABLE 50: EX-POST BETA ANALYSIS ..........................................................................................................................An Analysis of Hedge Fund Strategies – List of Tables TABLE 38: PERSISTENCE IN PERFORMANCE ANALYSIS BASED AN ADAPTED SHARPE SCORE (1/199712/2002)............. 261 TABLE 41: DESCRIPTIVE STATISTICS AND DECILE DESCRIPTIVE STATISTICS ......................................................... 309 TABLE 48 : INDIVIDUAL FUND MARKET EXPOSURE – MULTI FACTOR MODEL .................

........................................ 331 TABLE 55 : DESCRIPTIVE STATISTICS OF THE HEDGE FUND (PANEL A) AND OF THE EQUITY AND BOND MUTUAL FUNDS (PANEL B) ................................................................................................................................................ 392 TABLE 60: MEAN RETURN ESTIMATION FOR AVERAGE INVESTORS (INCLUDING MUTUAL FUNDS IN A HEDGE FUND PORTFOLIO)........................................................................................ 387 TABLE 58: MEAN RETURN ESTIMATION FOR PROGRESSIVE INVESTORS .................................... 400 .... 320 TABLE 52: LAGGED 12-MONTH DECILE SUB-PERIOD ANALYSIS................................. 389 TABLE 59: SLOPE ESTIMATION FOR AVERAGE INVESTORS ...................................... 397 TABLE 61: SLOPE ESTIMATION FOR AVERAGE INVESTORS (INCLUDING MUTUAL FUNDS IN A HEDGE FUND PORTFOLIO).......................................................................... 380 TABLE 57: MEAN RETURN ESTIMATION FOR PROTECTIVE INVESTORS .................An Analysis of Hedge Fund Strategies – List of Tables TABLE 51: INDICES AND DECILES SUB-PERIOD ANALYSIS ............... 328 TABLE 54: EX-POST BETA SUB-PERIOD ANALYSIS .......... 324 TABLE 53: SUB-PERIOD INDIVIDUAL FUNDS RESULTS .......................................... 371 TABLE 56: MEAN RETURN ESTIMATION FOR AVERAGE INVESTORS ........................................................................................................

..........................................................................................................An Analysis of Hedge Fund Strategies – Detailed Table of Contents An Analysis of Hedge Fund Strategies ................................................................................ i Preface........................................................................................ i Acknowledgements ......................................................................Abstract......................................................................... i An Analysis of Hedge Fund Strategies – Table of Contents ..... 1 Global Literature Review....................................................................................................................................... 14 Investing in Hedge Funds ................................... 26 ....... i Introduction and Purpose .................................. 5 The Data Issue.........................................

. .. . . ... ... . . .. ........... ........ ........ .. ... .. 5 3 3 . . ........ 52 3 .. . 29 Abstract Part Two: The Neutrality of Market Neutral Funds .. . .. ...... .. 36 Abstract Part Three: Hedge Funds as Diversification Tools .. . .. . . . . . . .... .... . .... .. ... .. . ... ..... .. .. . . ..... ... ... . . ..... ... ..... .. .. .. . .... ...... ... .. ... .. ..... ... .... 46 II Literature Review ...... . ....... .. ... . . . ... .... ..... . .. .. .. .. .. ... . ......... .. ... .. . ....... ....... .. ..Abstract Part One: The Persistence in Hedge Fund Performance ... . . . .. . . . ... ...... 2 Ev o lu tion in P erf o rma nc e M e asu rement . .... 2 Th e 3... . .... .... . . .... . .... . .. . . . .. . 5 2 3 .. .. . . . . . . .. .. ... . . . ...... ... . ..... ... . 38 Analysis of Hedge Fund Performance .. . .. ... ..... ...... .. . .. . ... . .. . 4 A n Ext en d ed Mu lt i-F ac tor M od e l . . ... . .. . ... .. .F act o r Mod e l of C ar h ar t ( 1 99 7 ) . ......... .. .. 1 Th e Ca p it a l A sset Pr ic ing M o d e l . ..... .. ......... .. . 45 I Introduction. .. .. .... .. 1 P e rf o rm a nce Stu d ies . . . . . ... . .... .... ... . . . . .... ..... . ... . ....... . ... .. ..... 47 2 . . .. ... .... . . . . ..... . .. ..... .. . . ...... ... . .. ... .... .. . ..... .. 5 4 3 ....... . ..... ..... .. 3 Th e 4... ... . ..... . . ... 4 7 2 .. . ......... ... ....... . . .... . . . .. ... . .. ...... . . . ... .fact or Mod el of Fama and F ren ch ( 1 993) an d its int ern at ion a l v e r s ion of Fa m a and F r en ch ( 1 99 8) ... ... .. .. .. . . .. .. .. ....... 5 5 . . . ... . . .. ..... .. ... . 4 9 III Performance Measurement Models . .. . . ... ... .

.... .y ea r Ret u rn -S o rt e d H ed g e Fu n d s P or t f o lios .. . . .. .. . .. .. .. . . . . .. .. .. ... . ... ..IV Data ... 1 P e rs ist en c e in On e. . ... .. . . . . . . . . 6 1 5 . . . .. .. .. ...... . .. . . . . . . . . . .. . . .. ..... ..... . . .. . . . . . ... . . . .... . . . ..... . . . ...... . .... .. .. .. . .. . .. . ...... . .. .. . . . .. .. . ... . 76 6 ... . . . . . . . .. . . 9 9 VII Persistence in Performance. .. .. . . . ... . . . . . . .. ... .. .. . . . . .. .. . . .... . 3 P e rf o rm a nc e ove r Sh or t er P er io ds . .. . . .. . . .. 7 6 6 ... .. 5 8 4 .. . . . .. . ... . .. . . . . 4 C or r e la t ion . . . . . . . . ... . . . . . . . . . . ... . . . . . .... . . 1 0 0 . . . .. ... . . . . 2 P e rf o rm a nc e M ea su r e m ent us ing Mu lt i-F ac tor M od e ls . ... . 61 5 .. . .. . . . . 5 9 4 . . . . . . . . . . . 6 8 5 . .. . . . . ..... . . . . .. . .. . ... . . . .. .. . . ... ... . . . ..... .. . .. ... .. . .. .. .. ... ... . .. . . .. .. .. .. .. . . . . . . . ... . . . . ... . . .. .... .. . . . .. . . . ... . . . . .. . 6 0 V Data analysis . . . . . . .. . 8 0 6 .. .. . . . . . . . .. . . .. ... ... .... ..... . . . . . .. . . .. .. ... . .. . . . . . . . . .. . 1 D at a P rov id e rs . ... .. 9 6 6 . ... . . .. .. ..... . . 3 Bas ic P e rf o r manc e . .. . . . . . . . . 2 Ins tant R e tur n H is t o ry Bias . .. .. ... . . . . ... .. ... .. . ..... . .... . . .. . ... .. .. . .. . . . . .. . ... . 100 7 . .. .. . . . . . . .. .. . . . 7 4 VI Hedge Funds Performance . ... . . . .... .. .. . . . . .. 2 H ed g e Fund s . . ... . .. .. . . ... . .. . . . . . . .. . . .. .. .. . . . . ... . . . .... .. . .. .. .. . ... . .. . ... . .. .. . . .. 1 S ur v iv o rs h ip b ia s . .. .. .. .. ..... . ... . . 5 7 4 . . .. . . .. . .. 1 P e rf o rm a nc e M ea su r e ment us ing the C APM . ... . . ......... 4 C om pa r is on w ith o th er S tu d ies . .. . . . . .... . . . . . . .. . . . .. .... . .. .. . ... . . . . . . .... . .. ... . 3 R is k-fr e e R e tu rn an d M a rk e t P e rf o rm a nc e .. . .. .. 57 4 . . . . . ..... . . . . .. .. .. . .. . . . .. . ... . . . . . . . .. . . . .. . . . . 4 B ias es in H edge Fund s Da t a . . . .. .. . . .. . .. . . . . ... . . . . . . .. . . .. . ...... . . . ... ..... .. . .. . . .. . .... . . .. . . 6 4 5 .. . . . . . ..... . . . . . . . . .. .. . . . . . ... .

. ... .. . .. . . . . . ... ... ...... . .. . . .... . . .. . ..... ... .. .. .... ... .. .. .. 3 D iss o lu t ion Fr e qu en c ies .. .. . . . . . . .. 1 Th e C a p it a l A sset P r ic ing M o d e l .. ....... .. .. ..... . .. . . .. . . ....... .. 2 Ba s ic P e rf o r man c e . . . ........... ... ... . .. ... . 1 D at aba s e... ... ... ...... .. 1 0 7 VIII Conclusion ... . ..... .. .. . . . .. . .. . . .... . ... . . . ... . . ... .. . .... .. .. .. .. . . .... . .... ..... . .. . .. . .. . .. . .. . .. ... . .. . . .... . .. . ..... ... .... . . ..... . . .. .. . . . .... .. ... .. .. .... .. . . .. . 1 3 2 . .. .. . .. . .. .. . . . . . 2 P e rs ist en c e ov er the Asian c r is is . . . .... . . . . .. . ... ... .. ....... . ... . . .. .. . . 121 II Performance Measurement Models . ... .. .. ... . ... . .... . .. ... . . .. . . . . .. . 120 Abstract . . ....... ..... ....... .. .. .. ... . . .. . . . ..... . . . .. . .. . 115 Hedge Fund Performance and Persistence in Bull and Bear Markets . .. .. . ... . . . .. 1 3 0 3 ... .. .... . . 1 2 5 2 ....... .. . .. ......... 120 I Introduction. .. ... . .... . .. . . ... ... . . . .. ... . ..... . .. . .. ..... .. . . . . . .. ... ... . 1 0 7 7 ... .. .. .. .. . ... ... . . . . .. .... .. . .. .. . .. .. 1 0 6 7 .. 3 A n a lys is p e r Sub -p e r io ds .. . .. ..7 .. 2 Th e 4. . . . . .. . . . . 3 Th e C o m p osit e M od e l . . . . .... . 1 2 7 III Data.. .. . ... ...... . .... ..... . .. .. . ... . . . .. .. . ... . .. . ... ... 1 2 6 2 .. ... .... .. 125 2 . . .. .. . .. . . . .......F act o r M o d e l of C ar h ar t ( 1 99 7 ) . .. . ... . . . .. .. ... 1 2 9 3 . . .......... .. .. .. . . . . .. . .... ... . .. .. . . .. . .. .. ... . .. . .. .. ...... .... ....... .. .. . . ... ... .. .. .... . .... . . . .. ... . . . .. . . . . .. . . .. .. .. . . .... . .. .. 4 On e-Y e a r P e rs ist en c e f or H e d g e Fun d S t ra t eg ie s .. ... ... 129 3 . ... . ..... . .. .. .. . .... . .. .. . ... .. .. . . . . ..... . . .... . ... . . .... ..... .. ... .. ....

. . .. . .. .. .. .. .. .. .. . . ... ..... . .. ...... . .. . .. . ... . 2 Ins t ant R e t u r n History Bias ..... .. . .. . . . .. . . . . . . . .... . . . .. . . . ... . . .. . . . .. . .. .. .. ... 2 P e rf o rm a n c e M ea su r e m ent u s ing M u lt i-F ac t or M od els . . .. . .. . . .. . . .. ... .. . ... . . . . ... . . . ... 4 C or r e lat ion s . . . ..... . . . . .. ... .. 1 4 2 IV Analysis of biases . 168 6 . . . .. . .. . ... . ... . . . .. . . . .. ... .... 1 7 4 VII Conclusion. . ... . . .. ... . . . .. ..... .. . . . . . .. . . .... . .. .. .. .. ... .. . . . . .. .. .... 1 P e rf o rm a n c e M ea su r e ment u s ing t he C AP M ... ... . . . . . .. . ... ... .3 Analys is of the Mark et N eu t r a l strateg y . .... . . . .. ... . . 1 5 4 V Hedge Funds Performance .... . . . . ... ... ... . ... . . .. . ... . ..... . .. . . . ... . . . . . .. . . ..... .. .. ... . . .. . ... . . .. .. . 3 C on c lu s ion .. .... .. .. . . . . . ... .... ... .. . . 2 P e rs ist en c e ov er t h e su b-p er io ds .... .. . .... . .. ... . .. .. .. .... ....... 1 5 5 5 . . .. . .. . . . . . .. .... ... . ... . . .. .. . ... ... . ...... .. ..... . . ... ..... .. . . . . .. . . . . .. . . . . .. ... . .. .. 1 P e rs ist en c e ov er t h e t ot a l p er io d .. .. . . . .. 147 4 . . .. . . . . . . ... .. ... .. . . . .. . . . ... .. . 186 .. . . .. . 1 5 1 4 . ... . .. .. ... . . . . . .. 1 6 8 6 ... .. .. .. ........ .... . . .... . .. . ... .. .. . . . .. 1 S u r v iv o rs h ip b ia s . . . .... .. 1 6 0 5.. . . . 1 6 1 VI Persistence in Performance . . . . . . 1 4 7 4 . . . .. . .. .3 . .. .. . .... ... . ... . 155 5 .. . . .. ... . ... . . . . . . . .. ..... . . . .. . 1 7 0 6. . . . ... . . ... 3 P erf o rm a n c e over b u l l i s h a nd b e a ris h s ub-periods .. .. . . .. . . . ..... . .. . . . ... ... . ... .. ... . . . . . .. . . .. .. . . ... . . . .. . . . .. . . . .. ... ... . .. ..

. . ... . 2 1 0 4 .... .... ... . 4 P e rs ist en c e in P e r f o rm an ce – Oth e r M e asu r e s .... . ....... . . . 210 4 . .... .... .. .. .. ... .. . ... .. ........ . .. ... . 2 0 9 IV Global Results: Market Analysis.. .... . .... . ........ ..... . ..... . ... . . .... . . . .. ... 2 0 8 3 .. ... ...... 2 2 0 4 .. ... . . ...... . . ........ ... . ... .. ..... . ... .... .. ...... ... ... .. . .. . ... 2 0 3 3 . .. ...... .. . ... .. . .. 1 D es c r ip t ive S t a t is t ic s .. . .... .. .Period An a ly s is . 195 II Database . .... . . . .. .. .. .. . . . . . ... ... . .. . .... . . ... . .. . .. .. ... . . . . .... . ..... .... . ........ . ... ................... .. .. . .. 190 Introduction.. .... .... . . ..... . ... .........R et u rns ... . .. . ... . .... .. . . 203 3 . ........ ... . ..... .... ..... ... .. .... . .. . .. ... . . . . .. .. . .. .. ...... .... ..The Sustainability of Hedge Fund Performance: New Insights . . ... ... ... .. . . . . . . .. 2 C or r e la t ion An a ly s is ... . .... . 2 1 6 4 .. .. ...... .. . 2 6 3 .... . ... ... . ... . .... ..... .. .... .. ..... .... . . . ... .. ... ... .. ... .. . .. .. . . ... ...... . . .. . ....... 3 P e rs ist en c e in P e r f o rm an ce – Three Years of D at a ...... . . . .. . ... ....... 200 III Preliminary Analysis . .. .. .... ....1 Sub... .. ....... ..... ... ... . ............ ... . . ... 2 P e rs ist en c e in P e r f o rm an ce .... ...... .... ..... ..... . . .. .. . . 191 I Methodology ... .. . ... . . .. .. .... .... .. .. . . ... ..... .. ... ...... ... 190 Abstract .. .. .. . . ... . .. .... ... .... ......... . ............... .. .. . ... ..... ..... ..... .. . .. .. ..... . .. . .. ..... ..... ....... .. .. ...... . ... . . ...... ............ . ..... ....... .. .. ..... .... . ..... .. . ...... ....... .. .. ...... . . . ... ... . ... ....... .. . . . . ... 1 P e rf o rm a n c e Ana ly s is . .... . .... ... . . ... 2 2 3 V Further Analysis . . . ... . ... .... .. .. ... .. .. .. . . ... . .... .... .. .... ... .. ... . ... . ... .. . 2 S u r v iv o rs h ip B ia s . ..... ... .. 263 5.. ......

... ..... .... .. ..... .. ..... ...... ..... . .. .. ...... . ... .. ..... ...... ... .. ..... ... . ........ ..... ... 2 9 0 3 ........... 269 The Neutrality of Market Neutral Funds .... . .. .... . ... .. . 2 8 3 3 .... .... . ....... ......... ........ ....... .... .... 2 C or r e la t ion a n a ly s is ..... .. ... 2 9 2 IV Survivorship bias analysis.......... ............ .. ....... 282 III Descriptive statistics and attrition rates .... . ......... ..... .. .. .. .... . ......... ........... . ..... ..... .... .... ........ . .. .. ..... ...... . ... .................. .... ... .. ... ...... .... ... . 273 Introduction.. ..... ... . ....................... .. ..... .. . .. .. . .......... 1 D es c r ip t ive s t a t is t ics . . 283 3 . ........... 295 V Methodology ..................... ...... . . .. . .... . ........ ...... ... ...... .... . . ............ ... ....... . ... ....... ....... ........ ... .... ...... ......... ...... ... ...... .. .. .. ...... .. .. .... 279 II Database . .. ... .. . .... .......... ....... ... ..... ......... ..... ..... . ..... .............. . .. ......... .... . . ...... ........ ..... 267 VII Conclusion. ..... ....... ......... ............. ....... . .. . ..5 .... . . 2 D at e Imp a ct ... 3 B irth a n d a t t rit ion rat e s .... ... .... . ..... . .......... 273 Abstract ........ .. ..... .... . .. .. . . ....... . .. ............. ... ... ...... . .. .. .. .... . ... .... ........... ..... ....... ... .... .. .... . 2 6 6 VI Frictions in the Real World . .... .......... ...... .. 274 I Interest of the study ...... . ... ......... ....... ........ ..... . ... ........ .. . 299 . . . ......... ... ... .. . .. . .. ........ ............. . .. .. .. ............ ... ... . . .... ........................ ....... . . .... ........... .....

...... ...... .. .. ... ........ .. . .. ..... .. .. . . ... .................. ...... ... .. ... .. .. .... 333 Diversifying using Hedge Funds: A Utility-Based Approach . ... .. .. ... .. ... . ...... . ... ......... . . .... ...... .............. .. .... ... .. . . . . ..... .. ... ..... ... .. ..... .. .. ...... .. . .. . . . . . ... ..... ... ... .. .. ....... . . ... ..... ...... ... .. . . ... .... . .... . .. . .1 Utility functions an d the Bell function ..... .. 3 4 8 II Methodology and data . .. ...... . . . .... .. . ...... ....... ............. . ............ .. ....... 1 C las s ic a l p o rt f o lio select ion app r o ach e s .. . .... ......... .. ............ . ..... . ..... . ... ..... ........... ... 3 1 7 8 .... .. ... .... ........ ... .VI Strategy and decile analysis.... .... . . .. . . ..... .... . . .. .... .... . ..... ....... .... . .. ...... .. .... .. . . ...... . . ........ ........ ... 3 5 3 ........... . .. .. ........ . . ... . .. ........... ..... .... . ...... . . . . . .. ... .. ..... . ......... . ........ . . ... . . .. ..... .. .. 1 M ar k et exp o su re .. . 2 Ind iv id u a l funds .. .. 3 E x.. . .. .... ..... ... . . . 307 VIII Sub-period analysis . .. . . . ... ... 3 4 2 1 .. .. . .. . .... .......... ............. ... . ......... ..... . 3 2 6 8 .... .. ... ... ... . .... ...... .. ........... 342 1.. ... .... . . 317 8 .. ... ....... ... ....... . . .. . 3 3 0 IX Conclusion.. .... . ..... 303 VII Individual fund analysis... . .. . ...... ... ..... .. ... . 2 S p an n in g..... .... . .. .... ..... . ..... .... ....... .... ....... .. 339 Abstract ..... ... ...p os t b et a s u b -period an a ly s is ... . ... ..... . ... .. ... .. .. .......... ....... . . . ... ... ... .... . . .. ...... . . ... . ......... . ....... 352 2 . ..... . . .. 339 Introduction..... .. .. . .. . ... . . . ... . ... ... .... . ... 340 I Utility functions and Spanning....... ..... .

. . . . .. . ... ..... . ..... . ....... ... ...... 402 General Conclusion . 2 T ay lo r a pp r ox ima t ion a n d r is k m e asu r e . . 3 9 5 V Conclusion ... ..... ..... . .. . ..... ...... ....... ... ........... .. . . .... .... ....... 377 4 . ...... ... 3 6 3 III The database.. .... ..... ... . . . .. ... . ... . ..... . .. ... .. . ..... . .. . . .. ... . ... . .... .... ....... . .. . .... .. . ... ..... . . . .. ... .. ...... . ..... ... ..... .......... ... ..... . 3 7 7 4... . .... . ....... .. . . . .... . . ...... ..... .... ... .. .. ........ ......... .. .... ...... 426 An Analysis of Hedge Fund Strategies – Detailed Table of Contents... ...... .. . . ..... .... .. .. . ........ .. .. . .. . ... . ... 3 R is k M easu r e . .. .... ... .. . ....... ....... ...... ....... . .. 431 An Analysis of Hedge Fund Strategies – Detailed Table of Contents.... .. ...... . ... .. ... ... . ... ..... . ........ . .... ... . ..... ... .... . ... ............. ... ....... ...... ... 4 E s t ima t ion method olo gy ... ... .. ...... . .. . . . .. ..... ... . ... . .. 431 ... ..... .. .... . . ............ 1 S ign ific a nc e level .. ...... . ........... ........ .... ...... . ... ... ... 3 5 9 2 . . ... . ....... .. .. ...... ......... . . 369 IV Results. ... .... .. ..... .. . . .. . .. . . . ..... 3 Inc lu din g mu t u al f u n ds in hedg e fund p o rtf o lios ....... .. . . 410 An Analysis of Hedge Fund Strategies – List of Figures ........... . .. 2 Inclu din g h edg e f u n d s in mutual funds p or t f olio ......... ...2 .. . .. ................ ........ 3 7 8 4 ... ........... .. .. 3 5 6 2 . ... ... .. .... . . 424 An Analysis of Hedge Fund Strategies – List of Tables .... . ..... . ... . . .................. .. . ....... . ............ ...... ....... . . . 404 References .. ... . . . . .. . ........ . . . . .... . .. ...

jlbphoto.© Copyright by Daniel Capocci 2007 All rights reserved © Copyright by www.net for Daniel Capocci’s picture on the back 440 .

Agefi. Figures are added when more illustrations are needed. and he has contributed to several hedge fund readers and presented his research to several conferences..D. Global Finance Journal. This PhD Thesis focuses on analyzing hedge fund strategies and focuses more precisely on hedge fund returns. Part 2: An Analysis of Hedge Fund Market Exposure analyses the market exposure of so-called market neutral hedge funds in depth.Hedge Funds are relatively unknown investment vehicles. Ph. The Thesis is divided in three parts. Daniel Capocci.Chartered Alternative Investment Analyst. This Thesis is an aggregation of his main research in the hedge fund area. Daniel has published several papers in academic and business journals such as the Journal of Empirical Finance. the European Journal of Finance. We develop an original methodology and present it and its application in details. 441 . Part 1: The Persistence in Hedge Fund Performance focuses on the explanation of hedge fund returns. Finally. l’Echo. aso. This Thesis covers the whole analysis of hedge fund returns. Part 3: Hedge Funds as Diversification Tools aims at determining if investors should include hedge funds in a classical portfolio of stocks and bonds. holds the positions of Senior Portfolio Manager at Kredietbank Luxembourg. Daniel has been studying the world of hedge funds for almost ten years and has worked as a fund of hedge funds manager since the early 2000. Haute Finance.

You're Reading a Free Preview

Download
scribd
/*********** DO NOT ALTER ANYTHING BELOW THIS LINE ! ************/ var s_code=s.t();if(s_code)document.write(s_code)//-->