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
An Analysis of Hedge Fund Strategies - Abstract

This PhD thesis analyses hedge fund strategies in detail by decomposing hedge

fund performance figures. Our aim is to present hedge funds, to understand what

managers expect to do and to understand how they make or destroy value over time. In

order to achieve this objective, we develop a multi-factor performance analysis model,

use it over several time periods and improve it over time. 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. Following this, I analyse

another specific aspect of hedge funds, their neutrality relative to equity markets in order

to validate hedge fund managers’ claims that they are market neutral. Finally, we

develop new efficient frontier measures, which not only include returns and volatility, but

also skewness and kurtosis in order to determine whether hedge funds are really

beneficial to investors.

i
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.

This paper was published in the Review of Financial Studies and was called Empirical

Characteristics of Dynamic Trading Strategies: the Case of Hedge Funds. At this time,

hedge funds were somewhat unknown investment vehicles providing very limited

information. Since then, the finance industry and the world of hedge funds in particular

have changed dramatically. Today, hedge funds are present almost everywhere. They

have been studied for years by dozens of academicians and professionals, who have tried

to understand the funds themselves, how hedge fund managers make money, the real

risks underlying alternative strategies, the use of hedge funds as diversification tools and

how they are really different from other investment products.

At the heart of my thesis are the performance figures of hedge fund managers. My

entire thesis is based on performance figures and their use as a way to understand hedge

fund strategies. 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
 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). Part two, on market neutral funds, looks at this prominent hedge

fund strategy. The third and final study is on the inclusion of hedge funds in a portfolio.

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, thereby allowing them to reach a higher efficient

frontier. Interestingly, 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. The main difference between this and my first

research on the subject is that both are based on the same underlying idea. The

difference is that in the final study I define and estimate an adapted efficient frontier that

takes into consideration not only volatility, as measured by the standard deviation, 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).

iii
 In coming to the end of this thesis, 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. I have really enjoyed exploring the world of hedge funds and this thesis is only a

start, since the world of hedge funds is continuously evolving in terms of the number of

funds, assets under management as well as the strategies employed.

iv
Introduction and Purpose

Hedge funds are private investment vehicles that can take long and short

positions in various markets, using various investment strategies and these funds are

accessible to large investors only. On the one hand, this definition is precise; on the

other, it is very broad. This is clear and focused. From another point of view, the funds

may use various kinds of securities on various markets. 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, when around 2,000 hedge funds were managing assets

totalling ca. $60 billion, the subsequent growth in the number and asset base of hedge

funds has never really been refuted. The industry only suffered from a relative slowdown

in 1998, but since then has enjoyed a renewed vitality with an estimated total of 10,000

funds managing more than a trillion US dollars by the end of 2006. The growing trend of

the sector remained remarkably sustained during the stock market collapse that started

in March 2000, when the NASDAQ Composite Index reached an all-time high of 5,132,

and finished three years later with a floor level of 1,253. In the meantime, the global net

asset value (NAV) of hedge funds continued to grow at a steady rate of 10.6% (Van

Hedge Funds Advisors International, 2002), contrasting with a decrease of 2.7% in the

worldwide mutual fund industry (Investment Company Institute, 2003). More recently, in

2001, Capocci & Hübner (2004) estimated that there were 6.000 HF managing around

$400b. In 2007, Capocci, Duquenne & Hübner (2007) estimate that there are 10.000 HF

managing around $1trillion. This is a growth of 11% in the number of funds and 26% in

assets over six years.

1
 The purpose of this doctoral thesis is clearly established: to understand hedge

fund strategies by looking at the performance numbers produced. Our first objective of

the studies is to understand clearly hedge fund managers and to explain how they create

alpha over time. This involves developing, testing and improving a performance analysis

model to understand hedge fund performance, while developing and adapting a

methodology to determine whether there is any persistence in hedge fund returns on the

other. I achieve this objective in three complementary studies grouped in Part 1

(Analysis of Hedge Fund Performance, Hedge Fund Performance and Persistence in Bull

and Bear Markets and Sustainability in Hedge Fund Performance: New Insights).

The second objective of the thesis is clearly linked to the first. Since the purpose

is to understand hedge fund strategies in detail, I perform a specific analysis on the most

represented and the most interesting, market neutral funds. By definition, market neutral

funds must a limited exposure to the market. I 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 (see Part two: An Analysis of Hedge Fund’s

Market Exposure).

2
 Finally, 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. 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. 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. Then, I

present the data issue before disserting on investing in hedge funds for the final

investors. Finally, we present the three parts of the Thesis in detail.

3
Table 1: Studies Specificities and Objectives

Specificity Objective 1 Objective 2

- Introduction of the extended multi-

An Analysis of Hedge - Determine if HF strategies sign. - Determine if HF strategies sign. and
factor model
Fund Performance outperform classical markets persis. outperform classical markets
- Analysis based on performance

Hedge Fund - Consider various market conditions &

- Determine if HF strategies sign. - Determine if HF strategies sign. and
Performance and adpated model (high yield & mortgage
outperform classical markets in bull persis. outperform classical markets in
Persistence in Bull factors)
and/or bear market conditions bull and/or bear market conditions
and Bear Markets - Analysis based on performance

The Sustainability of - Adapted model (option factors) - Determine if HF strategies sign. - Find a systematic way of buying HF in
Hedge Fund - Adapted meth. based on performance outperform classical markets using order to sign. and persis. outperform
Performance & other risk-adj. measures several risk-adjusted measures classical markets

- Focus on market neutral funds (28% of

The Neutrality of - Determine if high/low beta market
the industry) - Determine the market exposure of
Market Neutral neutral funds outperform their hig/low
- Focused on LT and ST periods and market neutral funds
Funds beta peers
various market conditions
- Analyse the impact of inserting HF in a
- Develop a methodology to determine if
classical portfolio taking abnormality - Determine if bond and/or equity
Diversifying using a portfolio can be diversified with
into account investors should include HF in their
Hedge Funds securities displaying abnormal return
- Distinguish between dir. undir HF and portfolio
dist. charact.
FoF
This Table reports the specificities and the objectives of the studies grouped in this PhD Thesis. Part 1: The Persistence in Hedge Fund Performance
contains three studies (An Analysis of Hedge Fund Performance, Hedge Fund Performance and Persistence in Bull and Bear Markets and the
Sustainability of Hedge Fund Performance: new insights). Part 2: An Analysis of Hedge Fund’s Market Exposure contains one study (The Neutrality
of Market Neutral Funds). Part 3: Hedge Funds as Diversification Tools contains one study (Diversifying Using Hedge Funds: A Utility-Based
Approach). For each Part we report the main specificity of the study and its first and second objective.
Global Literature Review

There have been many studies on hedge funds covering many different aspects of

the industry. In each of the specific studies reported in the heart of the thesis, we

perform literature reviews including studies whose results are directly linked to the

subject under analysis. In this introduction, we provide a global literature review on

hedge fund studies.

As illustrated in Figure 1, hedge fund academic studies can be classified into four

global categories: 1. Hedge fund performance, 2. Hedge fund investment style, 3.

Correlation analysis and diversification power and 4. Other studies. In the first global

category, we report studies that are focused on hedge fund performance. There are three

fields within this first category of hedge fund performance analysis. The first of these

fields includes studies that compare the performance of hedge funds with equity and

other indices (see for example Ackermann, McEnally and Ravenscraft, 1999; Brown,

Goetzmann and Ibbotson, 1999; Liang, 1999; Amin and Kat, 2001; Liang, 2001; Barès,

Gibson and Gyger, 2002; Liang, 2003; Agarwal and Naik, 2004). Results of such studies

are mitigated. Some authors (Brown et al, 1999; Liang, 1999; Capocci et al., 2005)

conclude that hedge funds have been able to outperform these indices, while others

(Ackermann et al., 1999; Agarwal and Naik, 2004) are more cautious in their conclusion.

Hübner and Papageorgiou (2006) find that there are three kinds of persistence in hedge

fund returns. Firstly, that there is statistical evidence of positive persistence based on

alphas for non-directional portfolios in the bullish period. Secondly, there is statistical

evidence of negative persistence for directional portfolios in both the bullish and the

bearish periods. Finally, the authors find statistical evidence of progressive positive

5
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. In this context, Ackermann, McEnally and Ravenscraft

(1999) and Liang (1999) find that hedge funds consistently achieve better performance

than mutual funds, although they are lower and more volatile than the reference market

indices considered.

The third field of hedge fund performance analysis includes the study of the

persistence of hedge fund returns. Persistence is particularly important in the case of

hedge funds because, as suggested by Brown, Goetzmann and Ibbotson (1999) and

Liang (2000, 2001), the hedge fund industry has a higher attrition rate than is the case

in mutual funds (see Brown, Goetzmann and Ibbotson, 1999). They prove that offshore

hedge funds have positive risk adjusted returns, 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. Agarwal and Naik (2000) analyse the presence of persistence in

hedge fund returns using a one-year moving average period. They find that there is proof

of persistence in hedge fund performance, particularly for poorly performing funds that

continue to underperform.

6
Figure 1: Four global categories of hedge fund academic studies

HEDGE FUND INVESTMENT CORRELATION ANALYSIS and

HEDGE FUND PERFORMANCE OTHER STUDIES
STYLE DIVERSIFICATION POWER

- Risks (Schneeweis and Spurgin,

- Comparison with classical markets
(Ackermann et al., 1999; Brown et
al., 1999; Liang, 1999; Amin and Kat,
2001; Liang, 2001; Barès et al.,
2002; Liang, 2003; Agarwal and
Naik, 2004).
- Sharpe style analysis (Fung and
Hsieh, 1997; Brown et al., 1998;
Brealy and Kaplanis, 2001, Brown
and Goetzmann, 2001, Liang 2001;
Ben Dor and Jagannathan, 2002;
Liang 2003).
- Correlation analysis (Fung and
Hsieh, 1997; Schneeweis and
Spurgin, 1997; Liang, 1999; Agarwal
and Naik, 1999).
1999; Jorion, 2000; Amenc et al.,
2002a; Amenc et al., 2002b, Berényi,
2002).
- Bias analysis (Liang, 2000; Fung
and Hsieh, 2000.

- Comparison with mutual funds - Hedge fund indices (Brooks and

- Rolling regression (McGuire, - Diversification power (Amin and Kat,
(Ackermann et al., 1999 and Liang, Kat, 2001; Amenc and Martellini,
Remolona and Tsatsaronis, 2005) 2001; Amenc and Martellini, 2002).
1999). 2002; Fung and Hsieh, 2002b).

- Persistence in performance - CTAs (see for example Edwards and

- Dynamic model (Swinkels and Van
(Agarwal and Naik, 2000; Brown et Park, 1996; Fung and Hsieh, 2001;
der Sluis, 2001; Posthuma and Van
al., 1999; Hübner and Papagergiou, Gregoriou and Rouah, 2003; Liang,
der Sluis)
2006; Liang 2001; Liang, 2000). 2003; Spurgin and Georgiev, 2001).

Figure 1 reports four global categories of hedge fund academic studies. We group studies on hedge fund performance, hedge fund investment
style, correlation analysis, diversification power and finally the other studies.
 The vast majority of performance studies on hedge funds have not focused solely

on the behaviour under different market conditions. The periods under review do not

favour this exercise, as periods of downward trending stock markets were rare and

discontinuous between 1994 and March 2000. For the period 1990-1998, Edwards and

Caglayan (2000) found that only three types of hedge fund strategies (Market Neutral,

Event Driven and Macro) provided protection to investors when stock markets decline.

More recently, Ennis and Sebastian (2003) contend that, in general, hedge funds did not

provide investor protection after the market downturn of March 2000; rather, their

superior performance was mostly due to the good market timing of their managers.

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, Fung and Hsieh, 1997; Brown, Goetzmann

and Park, 1998; Brealy and Kaplanis, 2001; Brown et al., 2001; Liang 2001; Ben Dor and

Jagannathan, 2002 and Liang 2003). In this context, Fung and Hsieh (1997) apply

Sharpe's style analysis (see Sharpe, 1992) to a large sample of hedge funds and

commodity trading advisors (CTAs). They assume that fund returns are linearly related to

the returns through a number of factors and measure those factors through eight

mimicking portfolios. 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. 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. A particular aspect that has been taken into account more

recenthly is the style drift in hedge fund returns. This effect comes from the fact that

hedge fund managers are opportunity driven and therefore change style over time.

Brown, Goetzmann and Park (1998) analyse hedge fund returns during the 1997-98

8
Asian crisis using rolling regression to take the style drift into consideration. The

methodology consists in realizing a set of linear regressions and moving the estimation

period of each of them by one observation. This simple technique enables one to observe

style variation of a manager over time. This methodology has one major drawback: the

choice of a number of observations used for the estimation. McGuire, Remolona and

Tsatsaronis (2005) apply the same methodology. To handle this issue, 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). This technique is adaptive in the sense

that changes in the style exposures are priced up automatically from the data. Unlike the

ad hoc rolling regression approach, the time variation in the exposures is explicitly

modelled. No restrictions are imposed on the betas. As stressed by Posthuma and Van

der Sluis, this model is a state-space model and can be estimated by using standard

Kalman filter techniques1. No window size and ad hoc chosen length need to be used.

The Kalman filter procedure chooses the optimal weighting scheme directly from the

data. The filter is an adaptive system based on the measurement and updating

equations.

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. 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, thanks to the weak correlation to the funds with other

financial securities. 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, but that a great majority of funds

1
See Pollock (1999) for a detailed presentation of the Kalman filtering and space-state models.

9
classified as inefficient on a stand-alone basis are able to produce an efficient payoff

profile when mixed with the S&P 500. They obtained the best results when 10-20% of

the portfolio value is invested in hedge funds. Kooli (2007) analyzes the power of hedge

funds as an efficient frontier enhancer. 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.

The author finds however that funds of hedge funds do bring diversification for mean-

variance investors. Taking all these results into consideration, hedge funds are seen as

good investment tool. 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.

In the fourth global category of hedge fund academic studies, “Other studies”,

other authors have analysed various other aspects of the hedge fund industry.

Schneeweis and Spurgin (2000), Jorion (2000), Amenc, Curtis and Martellini (2002),

Amenc, Martellini and Vaissié (2002) and Berényi (2002) study the risks involved in

hedge fund investments. Schneeweis and Spurgin (1999) as well as Amenc, Martellini

and Vaissié (2002) prove that hedge fund returns are not only exposed to the market

risk, but that other risks such as volatility risk, default risk or liquidity risk have to be

considered. Liang (2000) analyse the presence of survivorship bias in hedge fund data

and Fung and Hsieh (2000) include other biases in their analysis. Ackermann, McEnally

and Ravenscraft (1998) emphasise that stricter legal limitations for mutual funds rather

than for hedge funds hinder their performance. Some authors alos studied hedge fund

indices (see Brooks and Kat, 2001; and Amenc and Martellini, 2003). There are many

10
different hedge fund index providers such as EACM, HFR, CSFB/Tremont, Zurich Capital,

Van Hedge, the Hennessee Group, Hedgefund.net, LJH Global Investment, Mar, Altvest

and Magnum. Fung and Hsieh (2002b) looked at the natural biases present in hedge fund

indices.

Commodity trading advisors (CTAs) are a particular category in the hedge fund

world. Unlike hedge funds, which first appeared an academic journal in 1997, CTAs have

been studied for a longer period of time. Many studies were published in the late 80s and

in the early 90s (see for example, Elton et al., 1987, 1989, 1990, or Edwards and Ma,

1988). Since 1997, some authors have considered CTAs as part of the hedge fund world

(Fung and Hsieh, 1997; Schneeweis and Spurgin, 2000), whereas others have studied

them either by separating them from hedge funds (Liang, 2003) or on a stand-alone

basis (Fung and Hsieh, 2001; Gregoriou and Rouah, 2003 and Capocci, 2004b). Research

on CTAs is very sparse and it is difficult to present a complete literature review. However,

Billingsley and Chance (1996) and Edwards and Park (1996) demonstrate that CTAs can

add diversification to stocks and bonds in a mean-variance framework. Schneeweis,

Savanayana and McCarthy (1991) and Schneeweis (1996) stated that the benefits of

CTAs are similar to those of hedge funds, 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.

11
 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. According to Schneeweis,

Spurgin and Georgiev (2001), CTAs are known regularly to short stock markets. Fung

and Hsieh (2001) analyse CTAs and conclude that they are similar to a look-back call and

a look-back put. Gregoriou and Rouah (2003) examine whether the percentage changes

in the NAVs of CTAs follow random walks. They prove that all classifications (except the

diversified sub-index) behave in the same way as a random walk. The effectiveness of

CTAs in enhancing risk-return characteristics of portfolios could be compromised when

pure random walk behaviour is identified. Kat (2002) finds that allocating to managed

futures allows investors to achieve a very substantial degree of overall risk reduction at

limited cost. Managed futures appear to be more efficient diversifiers than hedge funds.

Regarding performance, the results are mitigated even though Edwards and

Caglayan (2001) conclude that, during bear markets, CTAs provide greater downside

protection than hedge funds, and have higher returns along with a negative correlation

with stock returns in bear markets. Schneeweis and Georgiev (2002) conclude that

careful inclusion of CTA managers in an investment portfolio can enhance its return

characteristics, especially during severe bear markets. Schneeweis, Spurgin and

McCarthy (1996) observe that performance persistence is virtually non-existent between

1987 and 1995. There is little information on the long-term diligence of these funds

(Edwards and Ma, 1998, Irwin, Zulauf and Ward, 1994, Kazemi, 1996). In his book

Managed Trading: Myths and Truths, Jack Schwager reviews the literature on whether

CTAs exhibit performance persistence and conducts his own analysis. He concludes that

there is little evidence that the top performing funds can be predicted. According to

Worthington (2001), between 1990 and 1998, the correlation of managed futures to the

S&P 500 during its best 30 months was 0.33 and that it was –0.25 during the worst 30

12
months. Georgiev (2001) underlines, however, that one of the drawbacks of CTAs is that,

during bull markets, their performance is generally inferior to that of hedge funds.

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. Finally, Capocci (2004b) proves that there is persistence in CTA

returns for badly performing funds, which tend to continue to significantly underperform

their peers.

13
The Data Issue

The main issue with hedge fund analysis is access to databases and the quality of

data2. 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, Hedge Fund Research (HFR) in Chicago, and Lipper TASS (TASS). As of

December 2004, TASS had 4,130 funds (2,431 live and 1,699 defunct), HFR had 5,158

funds (2,939 live and 2,219 defunct), and CISDM had 3,246 funds (1,315 live and 1,931

defunct). There are four other entrants to this field – The Barclay Group (Barclays),

Morgan Stanley Capital International (MSCI), Eureka Hedge, and Standard and Poors

(S&P). Because of their late entry to this field, their data were largely from reconstructed

history rather than real-time collection of hedge fund performance.

As stated by Liang (2000) and Fung and Hsieh (2006) many hedge funds report to

only a single database. Only few of them report to more than one database. 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, inception date, net assets

value, incentive fee, management fee, and investment styles across the two databases.

2
See for example Fung & Hsieh (2002b).

14
 In Figure 2, Fung and Hsieh (2006) compare the HFR, TASS and CISDM

databases. The Venn diagram divides the global hedge fund universe composed of five of

the main hedge fund databases. As shown in Figure 2, the overlap between the

databases is very low. This indicates that results obtained when performing an analysis

based on a specific database may be different if another database is used. Moreover,

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.

In this Thesis we use HFR, CISDM and Barclays (together and/or individually). 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. We have done for our extended multi-factor

performance decomposition model. We first apply it to a combination of the HFR and

CISDM databases. Then, we apply the adapted version to CISDM alone before using a

combination of the CISDM and the Barclay databases. Our results remain consistent

independently of the database used. The databases used and their characteristics are

reported in Table 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 period covered goes

from 1994/2000 for the shorter one to 1993/2003 for the longest (without considering

sub-period analysis).

15
 The second important aspect regarding the quality of hedge fund data is the

presence of biases in databases. First, 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. Voluntary participation means

that only a portion of the universe of hedge funds is observable. 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. This effect leads to a bias in

database that is called instant return history bias or the backfill bias.

The backfill bias on equity market data is commonly calculated by an indirect

approach. As stated by Posthuma and van der Sluis (2003), this indirect approach is

eliminating the first two years of reported data; see e.g. Fama & French (1993). Brown,

Goetzmann & Park (1998) use the method of Park to estimate an instant history of 15

months for the TASS database. Ackermann et al. (1999), Fung & Hsieh (2000), and

Edwards & Caglayan (2001) addressed the backfill bias for hedge funds in different

periods for different databases, and all used indirect approaches. Ackermann et al.

(1999) eliminate two years and find an average annual bias of 0.5% for MAR and HFR

database funds with different sample periods ending in 1995. Fung & Hsieh (2000)

calculate the backfill bias for the TASS database over the period 1994 to 1998. They

eliminate the first 12 months of returns, because they find a median 343 day incubation

period. The lasting mean performance was 1.4% lower over the period 1994–1998,

leading to a backfill bias of 1.4% for the TASS database 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 1.17% higher than the annual returns in subsequent years. We

estimate it by calculating the mean return of a portfolio investing in all funds and then we

16
make the same estimation by leaving the first 12, 24, 36 & 60 first returns of each fund.

We then compare the difference in performance.

Figure 2: Hedge fund database universe repartition

Source: Fung & Hsieh (2006)

17
Table 2: Database Comparison

Number of Percentage Funds of Percentage Analysis

Data provider Hedge funds
strategies dissolved funds dissolved period

An Analysis of Hedge
HFR/ CISDM 28 2449 28% 347 34% 1994/2000
Fund Performance

Hedge Fund Performance

and Persistence in Bull CISDM 16 2247 47% 647 33% 1994/2002
and Bear Markets

The Sustainability of CISDM/

16 3060 60% 907 72% 1994/2002
Hedge Fund Performance Barclays

The Neutrality of Market

CISDM 4 634 37% n/a n/a 1993/2002
Neutral Funds

Diversifying using Hedge

CISDM 16 4476 46% 2011 40% 1993/2003
Funds

This Table reports the comparison of the databases used over this PhD Thesis. HFR = Hedge Fund Research, Inc, CISDM = Center for International
Securities Derivatives Markets. Nb of strat = number of strategies used, Hedge funds = number of individual hedge funds, Percentage dissolved =
percentage of funds in the database that stop reporting to it before the end of the period under review. Funds of funds = funds of hedge funds and
analysis period = period under review for the corresponding study.
 Another very important bias is the survivorship bias. Funds disappearing from

database tend to have poorer performance than existing funds. Not taking these funds

into account leads to survivorship bias. Survivorship bias is calculated as the

performance difference between surviving funds and all funds in the dataset.

Survivorship is an issue in hedge fund analysis and this bias is estimated between 1.5%

(Fung & Hsieh, 1998) & 3% (Liang, 2001). We estimate these biases for our databases

and the results are reported in Table 3.

Our estimation of backfill bias lies between 1.2% and 1.3% when the 12 first

months of existence of each fund is removed to estimations between 2% and 2.2% when

more months are removed. 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. The difference can have several

reasons. First, they decide to eliminate the last month of existence of any fund by 50%,

this rule will clearly impact estimation. Second, the period they analyse starts in 1996

whereas ours (and many others estimations) is based on data starting in 1993 or 1994.

Third, since most hedge funds still report to one database only, there may be differences

in statistics of the different databases used. We use together or separately HFR, CISDM

and Barclays whereas Posthuma and van der Sluis (2003) has access to TASS. Finally,

Posthuma and van der Sluis (2003) use a direct method of examining the backfill bias.

Instead of eliminating the same average or median incubation period for all funds, the

direct method eliminates the individual incubation period per fund. The information that

they use is information from TASS and it is mainly based on qualitative information from

TASS employee.

19
Table 3: Backfill bias and survivorship bias estimation

Backfill bias Backfill bias Backfill bias

Survivorship bias
(12m) (24m) (36m)

Analysis of HF Perf. 1.22% 1.2% 1.6% 2.0%

HF Perf. and Pers. in

1.51% 1.3% 1.8% 2.2%
B&B Mkt

Sustainability of HF
1.08% n/a n/a n/a
Perf.

Neutrality of Mkt Ntl

1.68% n/a n/a n/a
Funds

This table reports the estimated survivorship bias and instant return history bias as estimated in four of our
studies. Survivorship bias comes from the fact that funds disappearing from database tend to have a worse
performance than existing funds. Not taking these funds into account lead to a survivorship bias. 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.

Without considering what methodology is the best, the differences between

Posthuma and van der Sluis (2003) methodology and ours explain the difference in

estimations. Througout this Thesis, we prefer to use the indirect approach with no

qualitative influence.

20
 A last element to stress is that a fund can be accessible to investors even if its

returns are not reported in any database. 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 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. The fact that many funds still report their performance to one database only

comforts us in this idea.

We are convinced that this bias may be quite low. Several authors (see Ackermann et

al, 1999; Fung and Hsieh, 2002b) argue that this bias can be counterbalanced by good

managers that stop reporting to databases when they close the funds to new investors.

Selection bias manifests itself in two basic ways. Hedge funds may enter a database on a

voluntary basis. On the one hand, presumably, only those funds that have good

performance and are looking to attract new investors want to be included in a database.

Therefore, hedge funds in a database tend to have better performance than those that

are excluded. On the other hand, hedge funds may not be participating in a database

because they are not looking to attract new investors. These self-excluded funds may

have better performance than the average hedge fund. Thus, the net effect of selection

bias on the returns of hedge funds in a database is ambiguous. Practically, there is no

way to mitigate this bias and we have to keep in mind that this bias may be present.

21
 Our estimation of survivorship bias lies between 1.22% and 1.68%. Such values

are in the lower end of recent estimation. Brown et al. (1999) report a bias of 3% for

offshore hedge funds per year. Fung & Hsieh (2000) use the TASS database and

calculate the annual survivorship bias to be 3% with a 15% drop out rate. Liang (2000)

examines this survivorship bias in hedge fund returns by comparing the TASS and the

HFR database. He finds that the survivorship bias exceeds 2% per year in the TASS

database, while the HFR database survivorship bias equals 0.6%, which is consistent with

the higher drop out rate in the TASS database. Ackermann, McEnally & Ravenscraft

(1999) suggest that two biases, the survivorship bias and the self-selection bias, offset

each other. The difference can be explained by several factors. First, most of our studies

analyses period ending by the end of 2002. Until recently, only the best managers

managed hedge funds. Since the demand for hedge funds exploded after the internet

bubble starting in 2000, more and more players entered the industry. The best managers

continue to launch funds, but less experienced individual are also attracted by the high

fee levels. As any other these funds get listed in databases but as they do not offer

attractive returns, most of them are dissolved after two or three years of data. This

element could explain an increase in the percentage of dissolved funds and a higher

survivorship bias after 2002 or 2003.

Finally, the three parts of this Thesis are based on individual hedge fund data;

several other researchers have used and/or looked at hedge fund indices (see Brooks and

Kat, 2001; Amenc and Martellini, 2003). The use of indices to analyse the hedge fund

world may also lead to measurement problems. There are many hedge fund indices

providers and most of them are reported in Table 4 with their main characteristics.

22
 The literature on the subject report five main potential problems with using hedge

fund indices3. First, since the quality of hedge fund data is poor, constructing indices

based on hedge fund data will result in biases in the index. As a result the returns of

hedge fund indices may not be meaningful. Second, some of the best hedge fund

managers do not disclose fund information to the public. If the assets held by these

managers make up a large portion of the assets in the hedge fund universe, then hedge

fund indices will under-represent the returns of the universe. Third, there is a debate on

how indices should be constructed, i.e. equally weighted or asset weighted. 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.

23
Table 4: Hedge Fund Indices Comparison

Nb of
Providers Launch Nb of Funds Website
Strategies

EACM 13 1996 100 eacmalternative.com

HFR 15 1994 1,1 hfr.com

CSFB/Tremont 9 1999 340 hedgeindex.com
Zürich Capital 5 2001 60 zcmgroup.com
Van Hedge 12 1995 750 vanhedge.com
Hennessee Group 22 1992 450 hedgefnd.com
Hedgefund.net 33 1979 1,8 hedgefund.net
LJH Global Investments 16 1992 800 ljh.com
MAR 15 1990 1,3 marhedge.com
Altvest 13 2000 1,4 altvest.com
Magnum 8 1994 NA magnum.com

This Table reports a comparison between the major hedge fund indices available. Nb of
strategies = number of strategies as defined by the index provider. Launch = launch
date of the indices for the corresponding index provider. Nb of funds = estimated num

In practice this Figure 2 is difficult to determine, since many hedge fund managers

have managed accounts and on/off-shore vehicles. Moreover, hedge funds may have

different levels of leverage and may vary their leverage employed through time.

Standardizing for leverage is problematic in index construction. Fourth, indices, suffer

from the problem that they overweight markets that have had strong historical

performance. Fifth, as shown by Amenc an Martellini (2003), there are significant

differences in return distribution for the same strategies4.

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
 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
Investing in Hedge Funds

Despite the growing interest in hedge funds, it is difficult for many individual and

institutional investors to participate in this area of the market. Several reasons can

explain this effect: high minimum wealth levels, sophisticated investor requirements or

complexity of the strategy applied. Funds of funds have several advantages in

comparison to individual hedge funds. They provide investors with diversification across

manager styles and professional oversight of fund operations that can provide the

necessary degree of due diligence. In addition, many funds of funds hold shares in hedge

funds closed to new investment allowing smaller investors the access to those managers.

Because of these, funds of funds constitute the only way of investing in hedge funds for

many investors.

There is however, one major drawback: the additional fees. The additional fees

has two main impacts. First, 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, other may be down

leading to a complex distribution of returns for the fund of funds. In addition, 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. 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. In case of bad

26
choice, the final outcome for investors will be less attractive than stated in academic

studies.

Several studies analyse the impact of fees in funds of funds. Brown et al. (2005)

concluded that funds of funds offered a relatively poor historical performance relative to

the hedge funds in which they invest. 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. (2005)

compare the performance of funds of funds with the one of portfolio constructed on the

basis of alpha, 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.

These elements stress out that there is a difference between conclusion of

academic paper and the reality for final investors. 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.

27
Abstract Part One: The Persistence in Hedge Fund Performance

As stated in the first part of the introduction, the first objective of this Thesis is to

understand hedge fund managers and to explain how they create alpha over time. This

involve developing, testing and improving a performance analysis model in order to

understand hedge fund performance on the one hand, while on the other developing and

adapting a methodology to determine whether there is any persistence in hedge fund

returns. We achieved this objective in three complementary studies (Analysis of Hedge

Fund Performance, Hedge Fund Performance and Persistence in Bull and Bear Markets

and Sustainability in Hedge Fund Performance: New Insights).

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,

Fama and French (1993)5 and Carhart (1997)6 models. Even if hedge funds are different

from mutual funds by the strategy they apply, the securities they use and the freedom

they have in their management, they remain investment funds. As such, 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.

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; 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.

29
 Since hedge funds do invest not only in US equities, we add several factors that

take into account the fact that hedge funds invest in non-US equities and in bonds

(government, corporate, high yield and default), as well as commodities. Our model

evolves over time for several reasons. First, some factors didn’t help explaining hedge

fund performance. Second, in some cases, the correlation between factors increased

leading to a risk of multicolinearity. Finally, some factors were added as they seem to

help decomposing the performance. The models we developed over the three studies are

described in Table 5.

As stated in Table 5, the first model has 11 factors, the second 10 and the one

used in the last part of the analysis 10 and 14. From model 1 to model 2, 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 high yield factor finally helps but the mortgage

factor does not.

As we will see in the papers, 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.

7
There has been a major move in the mortgage market in September 2002 and several hedge funds

have faced strong losses.

30
Table 5: Multi-factor Performance Decomposition Model

Model 1 Model 2 Model 3 Model 4

Capocci,
Capocci &
Corhay & Capocci Capocci
Hübner
Hübner (2006)-1 (2006)-2
(2004)
(2005)
Alpha X X X X
US Stock Market X X X X
Size X X X X
Style X X X X
International Style X n/a n/a n/a
Momentum X X X X

Non-US Stock Market X X X X

US Bond X n/a n/a n/a

Wd Gov Bond X X X X
EMBI X X X X
Lehman BAA X n/a n/a n/a
High Yield n/a X X X
Mortgage n/a X n/a n/a
GSCI X X X X
Currency n/a n/a X X
Option Factors n/a n/a n/a X
Number of factors 11 10 10 14

This Table reports a comparison of the multi-factor performance models used in this Thesis. Size = the
size market risk premium of Fama and French (1993), style = the style market risk premium of Fama
and French (1993), international style = the international style market risk premium of Fama and French
(1996), momentum = Carhart’s (1997) "momentum" factor, non-US stock market = MSCI World
excluding US, US bonds = the Lehman Aggregate US Bond Index, Wd Gov Bond = JPMorgan world
government bond index, EMBI = JP Morgan Emerging Market Bond Index, Lehman BAA = the Lehman
BAA Corporate Bond Index, High yield = Lehman High Yield Bond Index, Mortgage = Lehman Mortgage-
Backed Securities Index, Salomon World Government Bond Index, GSCI = Goldman Sach Commodity
Index, Currency = the Federal Reserve Bank Trade Weighted Dollar Index, Option factors = Agarwal and
Naik (2004) at-the-money (ATMC), out-of-the money (OTMC) European call option factors, and at-the-
money (ATMP) and out-of-the money (OTMP) European put option factors.

31
 Over our three studies on the persistence in hedge fund 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. Our

results indicate that most hedge fund strategies do offer significant alpha over a long

period of time. These results were not due to the lack of power of the model since, in all

cases, the adjusted r-squared were very high. The next logical step was to perform the

same analysis under various market conditions and we did that in our second study. More

precisely, we analyse hedge fund performance and persistence in performance over bull

market conditions, bear market conditions and over a full cycle. Our results indicate that

hedge funds tend to outperform during bull market conditions but that this out-

performance is no longer significant under bear market conditions. The only exception

was market neutral strategies, which needed further analysis that is reported later in this

Thesis.

Once we discussed that hedge fund strategies do significantly outperform classical

markets over time, we analysed the persistence of this performance, that is, we looked

at whether there was a repetitive way to isolate it over time. At this level, we reach the

second important basic concept also used in the second study: the decile classification of

Carhart (1997). 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. 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.”

32
 Carhart’s methodology is relatively straightforward to understand. Each year, all

funds are ranked in 10 equally weighted portfolios based on their previous year’s return.

The portfolios are held until the following January and then rebalanced again. The

combination of our multi-factor model with this methodology enables us to determine

whether there is persistence in hedge fund returns.

Our results indicate that there is some proofs of persistence for low volatility funds

that tend neither to be the best performers, nor the worst, but that offer relatively

consistent returns over time. This result was the first important conclusion of our thesis.

It needs deeper analysis over a shorter period of time, which was done in the second

study. Persistence analysis indicates that most of the predictability of superior

performance is found in bull market conditions (prior to March 2000). Our results confirm

several previous studies that found that persistence, if any, is mostly located among

medium performers. In bear market conditions, only negative persistence can be found

among the past losers, suggesting that bad performance has probably been the decisive

factor for hedge funds mortality.

In both studies, low volatility funds were the ones offering significant alpha. The

only issue is that these funds tend to be classified in the middle decile portfolios. 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. This is exactly what we did in our third study. We tested

several ways of classifying funds on the basis of their past performance: returns,

volatility, Sharpe ratio, alpha, beta, skewness and kurtosis. 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

checked the robustness of our results by performing the same analysis over sub-periods,

33
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, systematic way of

creating pure alpha using a simple classification methodology based on basic statistics:

risk-return trade-off measures (Sharpe score), pure volatility measures (standard

deviation) and, to a lesser extent, beta exposure, which appear to be better and more

stable ways of classifying hedge funds in order to detect persistency in the returns. Funds

offering stable returns, with limited volatility and/or with limited exposure to the equity

market consistently and significantly outperformed equity and bond markets.

We report the specificity, objectives and main conclusions of each study of Part 1

in Table 6.

34
Table 6: Persistence Analysis Studies Specificities, Objectives and Conclusions

Objective 1 Conclusion 1 Objective 2 Conclusion 2

- There is no proof of
- Most hedge fund strategies - Determine if hedge fund
- Determine if hedge fund persistence for hedge funds
An Analysis of Hedge do offer significant alpha strategies significantly and
strategies significantly but low volatile funds that
Fund Performance over the long term despite a persistently outperform
outperform classical markets tend to neither be the best
high R² classical markets
performers nor the worst

- Determine if hedge fund

- Determine if hedge fund - Hedge funds tend to
strategies significantly and - No significant
Hedge Fund Performance strategies significantly outperform during bull
persistently outperform outperformance in bear
and Persistence in Bull outperform classical markets market conditions (not
classical markets in bull market conditions but for
and Bear Markets in bull and/or bear market significantly in bear
and/or bear market market neutral funds
conditions markets)
conditions

- Determine if hedge fund
strategies significantly
The Sustainability of
outperform classical markets
Hedge Fund Performance
using several risk-adjusted
measures
- Systematic
- Find a systematic way of
- Most hedge fund strategies outperformance of hedge
buying hedge funds in order
do offer significant alpha fund portfolios invested in
to significantly and
over the long term despite a previous year' low volatile
persistently outperform
high R² funds (measured by Sharpe
classical markets
score, standard deviation)

This Table reports the specificities, objectives and conclusion of the studies grouped in Part 1: The Persistence in Hedge Fund Performance. Part 1
countains three studies (An Analysis of Hedge Fund Performance, Hedge Fund Performance and Persistence in Bull and Bear Markets and the
Sustainability of Hedge Fund Performance: new insights). For each study we report the main specificities of the study aalong with its first and
second objective and conclusions.
Abstract Part Two: The Neutrality of Market Neutral Funds

Self-defined market neutral funds significantly and consistently outperformed the

classical market over time. This result requires further analysis. 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. The study

analyses a complete cycle as well as sub-periods (bull and bear market conditions).

Market neutral funds represent a large part of the industry, around 28% of our database

used in the second Study of Part 1: Hedge Fund Performance and Persistence in Bull and

Bear Markets.

Our results confirm that the betas obtained were low in absolute terms even

though they were all significantly positive. The decile analysis indicates that the more

volatile funds (top and worst performing funds) have the highest market exposure,

confirming that low volatility funds emerge over time. At the individual fund level, one

third of the funds are significantly positively exposed to the market, while two thirds of

the alphas are significantly positive. 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. This can be explained by two reasons. First, the

aggregation of funds in indices leads to an increase in exposure to the equity market.

Second, the funds that are significantly exposed can bias the results because a) the bulk

of the funds are not significantly exposed to the market; only one third of the funds

were, and b) only five percent of the funds are significantly negatively exposed to the

market.

36
Table 7: Neutrality of Market Neutral Fund Specificities,
Objectives and Conclusions

Objective 1 Conclusion 1 Objective 2 Conclusion 2

- Market neutral funds

- Determine if
tend to be
- Determine the high/low beta market - Real market neutral
Neutrality of significantly exposed
market exposure of neutral funds funds outperform dirty
Mkt Ntl Funds 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, objectives and conclusion of the study of Part 2: An Analysis of Hedge Fund’s Market
Exposure. 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. First, during the bear

market, most poor performing market neutral funds out-performed the equity market

without being significantly exposed to the market, but the best performing funds

significantly out-perform the equity market and offer significantly positive returns.

Second, during the bullish period, no index is significantly exposed to the market.

However, during the bearish period, all but the best performing deciles are significantly

exposed to the market, but they all (except the best performing funds) created

significant alpha.

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. We report the

specificity, objectives and main conclusions of part 2 in Table 7.

37
Abstract Part Three: Hedge Funds as Diversification Tools

Hedge fund performance decomposition and strategy analysis were the first two

aims of this doctoral thesis. In order to complete these analyses, we analysed in our third

and final part the impact of inserting hedge funds into a classical portfolio of stock and

bond mutual funds. 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. 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.

Our methodology is based on the Taylor’s extension of the linex utility function

developed by Bell (1988). We decompose this function and take into account the mean

return, the volatility, the asymmetry of the return distribution (skewness) and the

presence of fat tails (kurtosis). This decomposition enables us to define a new and

extended risk measures that we use in a classical risk-return framework. The only

difference is that the risk factor is no more only defined by the standard deviation of the

returns. 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.

Our results indicate that directional hedge funds should be considered separately

from undirectional hedge funds and fund of hedge funds. Adding a small allocation to

directional hedge funds does not significantly change the risk-return profile offered by

the global portfolio. When more than 20% is allocated to directional hedge funds, there is

a significant improvement for diversified portfolios (20 to 80% allocated to the risky

38
asset). Over a allocation of 50% to directional hedge funds offers significantly more

attractive returnsin every case.

However, 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. For high allocation to the risky asset,

undirectional strategies do not help diversifying and reaching higher return levels. Our

results confirm that undirectional strategies and funds of funds are diversificating low risk

profile investments and should be used as such.

The new adapted efficient frontier opens new doors for asset allocators. Based on

the clients’ objective and the market conditions, it determines if hedge funds must be

added to the existing portfolio. Moreover it helps to determine what hedge fund strategy

should be favoured.

We report the specificity, objectives and main conclusions of part 3 in Table 8.

39
Table 8: Hedge Funds as Diversification Tools: Specificities, Objectives and Conclusions

Objective 1 Conclusion 1 Objective 2 Conclusion 2

- High allocation to directional

- Taylor's expansion of Bell's
hedge funds do significantly
- Develop a methodology to utility function enables us to
- Determine if bond and/or improve the profile
determine if a portfolio can be take skewness and kurtosis
Diversifying using equity investors should - Adding small allocation of
diversified with securities into account
hedge funds include hedge funds in their undirecitonal hedge funds and
displaying abnormal return - The adapted Capital Market
portfolio funds of hedge funds
distribution characteristics Line and new efficient frontier
significantly improve the
complete the development
profile

This Table reports the specificities, objectives and conclusion of the study of Part 3: Hedge Funds as Diversification Tools. Part 3 countains one study (Diversifying
using Hedge Funds: a utility based approach). We report the main specificities of the study and its first and second objective and conclusions.
Part One: The
Persistence in Hedge
Fund Performance
Analysis of Hedge Fund Performance

Daniel P.J. CAPOCCI

HEC-ULG Management School – University of Liège (Belgium)

Georges HÜBNER

HEC-University of Liège, Limburg Institute of Financial Economics,

Maastricht University, and Luxembourg School of Finance, University of

Luxembourg.

Capocci, Daniel and Georges Hübner, January 2004, Analysis of Hedge Fund

Performance, Journal Empirical of Finance 11/1, 55-89

Analysis of Hedge Fund Performance

Abstract

Using one of the largest hedge fund databases ever used (2796 individual funds

including 801 dissolved), we investigate hedge fund performance using various asset

pricing models, 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. This

addition is particularly suitable for more than half of the hedge fund categories, and for

all funds in general. The performance of hedge funds for several individual strategies and

different sub-periods, including the Asian Crisis period, indicates limited evidence of

persistence in performance but not for extreme performers.

45
Analysis of Hedge Fund Performance

I Introduction

With almost 6.000 funds managing around $400 billion in capital, hedge funds

justify an increased attention in financial press as well as in the academic world. These

funds, that have been existing for more than fifty years, are not legally defined but share

some common characteristics: they use a broad range of instruments like short selling,

derivatives, leverage or arbitrage on different markets. Hedge funds require high

minimum investments and their access is limited to individual investors or to institutions

with large financial resources. Currently, about 90% of hedge fund managers are based

in the US, 9% in Europe, and 1% in Asia and elsewhere. While the number of funds has

more than doubled since the mid-nineties, around 80% of hedge funds are smaller than

$100 million, and around 50% are smaller than $12 million. This reflects the high

number of recent entries.

Scientific literature on performance-evaluation yields controversial results. This

lack of consensus on the « right » model puts researchers in a quandary (Metrick, 1999).

This study investigate hedge funds performance levels and persistence using various

asset-pricing models, including an extension of Carhart’s (1997) model, combined with

the models of Fama and French (1998) and Agarwal and Naik (2004) and with a factor,

never previously used in this context, accounting for the fact that some hedge funds

invest in emerging bond markets. This analysis is carried out for different sub-periods

including the Asian Crisis period and for several individual hedge funds strategies.

46
 The rest of the study is organised as follow. Section 2 reviews some of the major

mutual and hedge funds performance studies with a focus on the evolution in the models

used. Section 3 sets out the performance models we will use. The next Section provides

a thorough description of our database. Section 5 brings some insights on hedge fund

performance. Section 6 reports the performance of hedge funds. Section 7 documents

and explains the persistence in hedge fund returns. Section 8 concludes the study.

II Literature Review

2.1 Performance Studies

Despite the increasing interest that hedge funds have originated due to their

recent development, few performance studies have been carried out on hedge funds

comparing to other investment tools like mutual funds. This can partly be explained by

their private characteristics and the difficulties encountered to have access to individual

funds data. Therefore, 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.

In general, performance studies can mainly be classified in two categories,

depending on whether they conclude or deny that mutual funds have significantly higher

realized returns that those obtained by following passive strategies8, inducing that

managers of mutual funds have access to sufficient information to recover their costs.

8
See a.o. Lehmann and Modest (1987), Ippolito (1989), and Grinblatt and Titman (1989, 1992) for

contenders of superior performance of hedge funds, and Jensen (1969), Malkiel (1995), Gruber (1996)

and Carhart (1997) for studies reaching the opposite conclusion.

47
 Among studies finding superior mutual funds performance, numerous studies

investigate further its persistence. On the one hand, Hendricks et al. (1993), Goetzmann

and Ibbotson (1994), Brown and Goetzmann (1995), and Wermers (1996) show

persistence in mutual funds performance for a short period (1 to 3 years), and attribute

it to hot hands9 or to common investment strategies. It is worth noting that Carhart

(1997) and Daniel and al. (1997) demonstrate that the momentum effect in the share’s

returns explain the hot hands effect detected by Hendricks, et al. (1993). On the other

hand, Ippolito (1989), Grinblatt and Titman (1989, 1992), Elton et al. (1993), Elton et al.

(1996), Sirri and Tufano (1998), and Zheng (1999) report a predictability in the mutual

funds returns over a longer period of time.

Considering the recent interest for this sector, studies on hedge funds persistence

in performance are less frequent. Nevertheless, Agarwal and Naik (2000) sustain that

persistence in hedge funds performance exists. This issue of persistence in performance

is particularly important in the case of hedge funds because, as emphasised by Brown

and al. (1999, 2001) and Liang (1999), hedge funds experience an attrition rate much

higher than mutual funds. Brown and 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.

Ackermann and 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, although lower than the market indices

considered. They also indicate that the returns in hedge funds are more volatile than

both the returns of mutual funds and those of market indices. 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.

48
Ravenscraft (1998) emphasise that the stronger legal limitations for mutual funds than

for hedge funds hinder their performance. According to Brown and al. (2001), 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.

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. This low correlation is also

emphasised by Liang (1999) and Agarwal and Naik (2000). Amin and Kat (2001) find

that stand-alone investment hedge funds do not offer a superior risk-return profile, 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. They obtain the best

results when 10-20% of the portfolio value is invested in hedge funds. Taking all these

results into account, hedge funds seem a good investment tool.

2.2 Evolution in Performance Measurement

In the eighties, performance measures based on the CAPM, like Jensen’s alpha

(1968) and their extensions were commonly used in performance evaluation. The recent

interest in multi-factor models primarily comes from the literature on the cross-sectional

variations in stock return. Several studies10 report that the cross-section of average

returns on U.S. common stocks shows little relation to the betas of the Sharpe (1964)-

Lintner (1965) CAPM or the Breeden (1979) ICAPM. Instead, these authors identify other

factors like the size of the company (Banz, 1981), leverage (Bhetari, 1988),

earnings/price (Basu, 1983), book-to-market (Rosenberg et al. 1985; Fama and French,

10
See e.g. Reinganum (1981), Breeden, Gibbons, and Litzenberger (1989), Fama and French (1996)

and Chan, Jegadeesh and Lakonishok (1996).

49
1992), dividend yield (Litzenberger and Ramaswamy, 1979, 1982) and more recently the

momentum effect (Jegadeesh and Titman, 1993; Carhart, 1995) that have reliable power

to explain the cross-section of average returns.

Subsequent multi-factor models include the 8-factor model developed by Grinblatt

and Titman (1994), the asset class factor model from Sharpe (1992), the 3-factor model

from Fama and French (1993), the 4-factor model from Carhart (1997), and the

international model of Fama and French (1998).11

However, recent studies have cast doubt on the usefulness of these new models.

Kothari and Warner (2001) show that the Fama and French (1993) 3-factor model

provides better results than the classical CAPM, but document that it detects significant

abnormal results (including timing) when none really exists. In addition, Carhart (1997)

develops his own 4-factor model that proves to be superior to the classical CAPM, the

Grinblatt and Titman (1989) 8-factor model and the Fama and French (1993) 3-factor

model.

In hedge funds literature, different models have also been used in performance

evaluation. In an early study, Fung and Hsieh (1997) extend Sharpe’s (1992) asset class

factor model and find five dominant investment styles in hedge funds. Schneeweis and

Spurgin (1998) also use style analysis based on a multi-factor approach. Brown and al.

(1999) and Ackermann and al. (1999) use a single factor model and focus only on total

risk. Agarwal and Naik (2000) use regression-based (parametric) and contingency-table-

based (non-parametric) methods. Their parametric method regresses alphas (or

appraisal ratios) on their lags. For the non-parametric method, they construct a

contingency table of winners and losers depending on the alpha. Liang (1999) uses the

11
See Allen and Soucik (2000) for a description of the major models used in mutual funds

performance studies.

50
extension of Fung and Hsieh (1997) model, regressions based on fund characteristics,

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. Agarwal

(2001) uses a model consisting of trading strategy factors and location factors to explain

the variation in hedge funds returns over time.

These results suggest that it is necessary to realize performance studies based on

multi-factor models, rather than simply use the CAPM, but there exists no unanimously

accepted model. Therefore, it is preferable to use several specifications in order to

compare the results obtained.

51
III Performance Measurement Models

For comparison purposes, our study of hedge funds performance starts with the

CAPM. The basic multi-factor specifications are the Fama and French (1993) 3-factor

model and its international version of 1998 (Fama and French, 1998) and the Carhart

(1997) model because they are not dominated by any other model in the mutual funds

performance literature.12 Finally, 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.

3.1 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). Its equation to estimate is the

following:

RPt − RFt = α P + β P (RMt − RFt ) + ε Pt t = 1,2,..., T (1)

where RPt = return of fund P in month t; RFt = risk-free return on month t; RMt =

return of the market portfolio on month t; εPt = error term; αP and βP are the intercept

and the slope of the regression, respectively.

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.

52
 The intercept of this equation, αp commonly called Jensen’s alpha (1968) is

usually interpreted as a measure of out- or under-performance relative to the market

proxy used.

3.2 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. It takes the size and the book-to-market ratio of the firms into

account. 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 =1,2,...,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’).13 These

factors aim at isolating the firm-specific components of returns.

In the international version of the model, Fama and French (1998) consider twelve

major EAFE (Europe, Australia, and Far East) countries and several emerging markets

and propose an international factor mimicking for book-to-market equity (HML). The

formula is the following:

13
See Fama and French (1993) for a precise description of the construction of SMBt and HMLt.

53
 RPt − RFt =α P + β P1(RMt − RFt )+ β P2 IHMLt +ε Pt t =1,2,...,T (3)

where IHMLt = an international version of HMLt.

According to Fama and French (1998), the international CAPM cannot explain the

value premium in international returns, 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.

3.3 The 4-Factor Model of Carhart (1997)

Carhart’s (1997) 4-factor model is an extension of the Fama and French (1993)

factor model. It takes into account size and book-to-market ratio, but also an additional

factor for the momentum effect. Grinblatt, 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 =1,2,...,T (4)

Where PR1YRt = the factor-mimicking portfolio for the momentum effect.14

14
For a description of the construction of PR1YR see Carhart (1997).

54
 As stressed by Daniel et al. (1997), this model assumes that, in the absence of

stock selection or timing abilities, the coefficients of four zero investment factor

mimicking portfolios are appropriate measures of multidimensional systematic risk. It

identifies a matching passive portfolio return for each fund return.

3.4 An Extended Multi-Factor Model

In order to take into account the different characteristics of the hedge fund

industry, we implement a combination and an extension of Carhart’s (1997) 4-factor

model, the international model of Fama and French (1998), and the model used by

Agarwal and Naik (2004) and Agarwal (2001).

This model contains the zero investment strategies representing Fama and

French’s (1993) size and value, 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), a factor for non-US equities investing funds

(MSCI World excluding US), three factors to take into account the fact that hedge funds

invest in US and foreign bond indices (Lehman US aggregate bond index, Salomon world

government bond index, and JP Morgan Emerging Market Bond Index) and finally a

commodity factor (GSCI Commodity Index). Beyond the combination of existing models,

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.

In order account for the fact that hedge funds invest in a wide range of equities

including small and large companies, the market proxy used is the Russel 3000 that

represents over 95% of investable US equity market.

55
 Note that Agarwal and Naik (2004) and Agarwal (2001) take several additional

factors such as the MSCI Emerging Markets, the Salomon Brothers Government and

Corporate Bond Index, and the Lehman High Yield Bond Index. Their high colinearity with

other factors lead us not to test these indices further.

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. Its components are weighted according

to their impact on production in the world economy.

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; MSWXUSt = return of the MSCI World

Index excluding US; LAUSBIt = return on the Lehman Aggregate US Bond Index;

SWGBIt = return on the Salomon World Government Bond Index; JPMEMBIt = return of

the JP Morgan emerging market Bond Index; LEHBAAt = return of the Lehman BAA

Corporate Bond Index; GSCIt = return of the Goldman Sachs Commodity Index.

56
IV Data

4.1 Data Providers

First, it is important to stress all information on hedge funds is available

exclusively on a voluntary basis, for ‘allowed’ persons depending on the country in which

the fund wants to find investors. Fortunately, many hedge funds release monthly

information to inform existing investors or to attract new ones. Some data collectors

make them, in turn, available to the qualifying public. As stressed by Amin and Kat

(2001), there are three main hedge fund database providers in the world. These are

‘Managed Account Reports’ (MAR, 1500 funds), ‘Hedge Fund Research, Inc.’ (HFR, 1400

funds), and ‘TASS Management’ (TASS, 2200 funds). These databases are the most used

in academic and commercial hedge fund studies. The MAR database is used, among

others, by Fung and Hsieh (1997), Schneeweis and Spurgin (1998), and Amin and Kat

(2001). HFR is used by Schneeweis and Spurgin (1998), Liang (1999), Agarwal and Naik

(2000, 2002), and Agarwal (2001). The TASS database is used in Brown et al. (2001),

Fung and Hsieh (2000, 2001), and Brown and Goetzmann (2001). The three databases

have never been used together in a study, but Ackermann et al. (1999) and Ackermann

and Ravenscraft (1998) used a combination of HFR and MAR. Liang (2000) uses a

combination of TASS and HFR.

Data vendors do not only collect performance data. For a majority of funds, they

record other useful information such as company name, start and ending date, strategy

followed, assets under management, management and incentive fees, managers’ name

etc. There is no consensus on the definition of the strategy followed but there are

similarities. MAR defines 9 strategies along with 15 sub-strategies. HFR defines sixteen

57
different strategies in two categories, 12 non-directional and 5 directional strategies, plus

the Funds of Funds and the Sector categories. Finally, TASS defines 15 strategies.

4.2 Hedge Funds

We use hedge fund data from HFR and MAR, similarly to Ackermann et al. (1999),

whose database ranges from 1988 to 1995. Both databases give monthly net-of-fee

individual returns and other information on individual funds and group them in indices.

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,

offshore and a combined index), and the MAR database gives 2354 individual hedge

funds plus 23 indices between January 1984 and June 2000. Then, in each database, we

remove funds that appear twice15 and funds with quarterly returns. This leaves us with

1639 individual hedge funds in the HFR database and 2014 hedge funds in the MAR

database. For the 857 funds present in both databases, we use the database presenting

more observations.

This leaves a total of 2796 individual hedge funds. This is one of the greatest

database ever used in hedge funds performance studies. These funds include 1995

(71%) survived funds and 801 (29%) dissolved funds.

15
This happened in three cases: when the same fund (same name, company, and returns) appeared

twice in the database; when the same fund (same name, and returns) appeared twice in the database

with two different company names; and when the same fund (same company, and returns) appeared

twice in the database with two different fund names.

58
4.3 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), Schneeweis and Spurgin (1998) and Liang (1999) use style analysis based multi-

factor approach, while Brown et al. (1999) address this issue by employing a Generalised

Stylistic Classification (GSC) algorithm and grouping the managers on the basis of their

realized returns.

We have to choose a market performance index. The alternative is the value-

weighted portfolio of all NYSE, Amex and NASDAQ stocks usually used in mutual funds

performance studies (see for example Fama and French, 1993, 1996, 2000; Carhart,

1997) or the Russel 3000 used in Agarwal and Naik (2004) and Agarwal (2001). The

comparison of descriptive statistics of the two proxies suggests that both market proxies

are very similar.16 The results of the study should not be influenced by the market proxy

chosen.

We take the value-weighted portfolio of all NYSE, Amex and NASDAQ stocks

market proxy, and the one-month T-bill from Ibbotson Associates as the risk-free rate.

16
The comparison is available upon request.

59
4.4 Biases in Hedge Funds Data

Survivorship bias is an important issue in mutual funds performance studies (see

Carhart and al., 2000). In response to this concern, data vendors do backfill fund’s

performance history when a new fund is added to the database. This allows them to

provide data that go back before the starting date of the database itself (usually 1993).

These providers do not eliminate defunct funds and should normally not suffer

from survivorship bias for the years after the start of the databases. However, as pointed

out in several other studies (see Brown et al., 2001), survivorship bias is very likely to be

present and not negligible, although hardly measurable, for the period before 1994. This

bias may severely hinder statistical inference (Hendricks et al., 1997; Carhart et al.,

2000).

Moreover, according to Ackermann et al. (1999) and to Fung and Hsieh (2000),

two upward biases exist in the specific case of hedge funds because, since they are not

allowed to advertise, they consider inclusion in a database primarily as a marketing tool.

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, because they might be considered by the SEC as making illegal

advertising. 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. 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. Nevertheless, to avoid polemics, we

take all funds (both living and dissolved) into account.

60
V Data analysis

We will first discuss evidence of biases before and after 1994 in our database, 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.

5.1 Survivorship bias

Survivorship bias has received considerable attention in the academic literature.

Two definitions of this bias are commonly used in studies: the performance difference

between surviving funds and dissolved funds (e.g. Ackermann et al., 1999) and the

performance difference between living funds and all funds (e.g. Liang, 2000). We report

the bias using both definitions for the whole period and for 2 sub-periods 1984-1993 and

1994-2000.

In Panel A of Table 9, we report a monthly survivorship bias of 0.36% (or 4.45%

per annum) for the whole period using the first formula and in Panel B a bias of 0.07%

per month (0.9% per annum) using the second formula. A look at sub-period biases

indicates that survivorship bias is much higher after 1994.

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. As our sample does not

report data on dissolved funds prior to 1994 for MAR and prior to 1993 for HFR, one is

automatically left with funds that survived this period and thus survivorship is indeed

qualitatively maximal. This indicates the lack of reliability of the database for this period.

Ackerman et al. (1999) use the same combination of databases as we do and the

first definition of survivorship bias. Their value (0.16% per year) is lower than the value

we obtain for the whole period and for the second sub-period (respectively 0.9% and

61
4.45% per year) and closer to the one we obtain for the first sub-period (0.6% per year).

This difference can mostly be explained by the different time period analysed. Ackermann

et al. (1999) study the 1988-1995 period, i.e. the greatest part of their time window

does not encompass funds that disappeared during the period. The issue of their

suspiciously low survivorship bias has been extensively discussed by Fung and Hsieh

(2000) and Liang (2000).

The value reported using the second definition for period 1994-2000 (1.2%) is

very close to the percentage of 1.5% from Fung and Hsieh (1998). It is however lower

than the 0.30% monthly bias found by Fung and Hsieh (2000), the 3% bias found by

Liang (2001) and the industry consensus bias of 3% stressed by Amin and Kat (2001).17

Liang (2000) compares the survivorship bias in HFR and TASS database and finds

significant differences between them. He finds out that they are mostly non-overlapping,

but the TASS database is probably more complete as it covers a larger number of

dissolved funds. 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,

even though our sample covers a larger time period.

17
We find this consensus value quite high when compared to the 0.8-1.5 bias reported by Malkiel

(1995) and Brown and Goetzmann (1995) for US mutual funds.

62
Table 9: Survivorship Bias in Hedge Funds

All Funds Surviving Funds Dissolved Funds

Year Return St. Dev. Obs. Return Std. Dev. Obs. Return Std. Dev. Obs.

1984 1,08 3,26 348 0,97 1,57 248 1,36 7,6 76

1985 2,55 1,44 501 2,83 1,61 356 1,87 1,34 121
1986 1,8 1,58 716 1,99 1,55 497 1,58 2,33 195
1987 1,5 4,16 1069 1,4 4,26 782 1,64 5,39 263
1988 1,72 1,54 1376 1,83 1,73 1020 1,44 1,03 332
1989 1,52 0,79 1931 1,65 0,87 1444 1,13 0,76 463
1990 0,74 1,16 2792 0,77 1,22 2063 0,61 1,32 705
1991 2,08 1,67 3880 2,07 1,96 2826 2,12 1,4 1030
1992 1,23 0,61 5280 1,24 0,74 3860 1,2 0,64 1396
1993 1,95 0,47 7245 1,91 0,48 5140 2,07 0,66 2081
1994 0,16 0,69 9856 0,16 0,79 6931 0,15 0,73 2901
1995 1,62 0,86 13528 1,68 1,14 9298 1,48 0,58 4206
1996 1,69 0,75 17654 1,66 0,77 11942 1,75 0,97 5688
1997 1,51 0,81 22454 1,6 0,75 15122 1,35 1,3 7308
1998 0,46 10,08 25735 0,78 11,94 18440 -0,36 3,72 7271
1999 2,46 2,12 25826 2,58 2,09 21712 1,79 2,92 4090
2000 1,15 2,23 12421 1,22 2,23 11880 -1 3,3 529

Mean 84-93 1.62 1.67 1.50

Mean 94-00 1.29 1.38 0.74

Mean 84-00 1.48 1.55 1.19

 Panel A: Living Funds - Dead Funds Panel B: Living Funds - All Funds

Bias 84-93 0,16 per Month Bias 84-93 0,05 per Month
1,96 per Year 0,6 per Year
Bias 94-00 0,65 per Month Bias 94-00 0,09 per Month
7,79 per Year 1,22 per Year
Bias 84-00 0,36 per Month Bias 84-00 0,07 per Month
4,45 per Year 0,89 per Year

This Table reports the survivorship bias of calculated from our database. Our combined HFR/TASS
database contains 2796 hedge funds, including 1995 survived funds and 801 dissolved funds as of
June 2000. In Panel A survivorship bias is calculated as the performance difference between surviving
funds and dissolved funds. In Panel B survivorship bias is calculated as the performance difference
between surviving funds and all funds. All returns are net of fees and on a monthly basis. Numbers in
the table are monthly percentage.

These biases suggest that poor performance could be the main reason for

disappearance. Figure 3 plots returns of the dissolved funds in our database over the 24-

month period before their exit dates. It shows a declining return pattern towards the date

of exit, indicating inferior performance: it corresponds to a decrease of the mean return of

almost 3.5% in two years.

64
Figure 3: Monthly returns for dead funds towards the
dissolution date

1,5

1
Returns

0,5

0
24 23 22 21 20 19 18 17 16 15 14 13 12 11 10 9 8 7 6 5 4 3 2 1
-0,5

-1

-1,5
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.

65
5.2 Instant Return History Bias

When new funds are added into a database, historical returns are backfilled. This

corresponds to a demand by fund managers who market themselves if they have good

track records, i.e. after compiling good performance. Fung and Hsieh (2000) estimate this

bias using a 12 month incubation period. They find a 1.4% per year difference in returns

for the 1994-1998 period.

Following Park (1995), Brown et al. (1997) and Fung and Hsieh (2000), we

estimate this bias for our hedge fund database in two steps. On the one hand, 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). On the other, we

estimate the average monthly return from investing in all these funds after deleting the

first 12, 24, 36 and 60 months of returns (we called this portfolio the adjusted observable

one). 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). Results are

reported in Table 10.

For the 1/84-6/00 period, the observable monthly return averaged 1.49%, while

the adjusted observable one was 1.42% (when deleting the 12 first months), 1.26% (24

months), 1.20% (36 months), and 1.15% (36 months). This gives an estimate of

approximately 0.9% per year, lower than the 1.4% found by Fung and Hsieh (2000) for

the instant history bias. For the 1/94-6/00 period, the bias of 1.2% per year is closer to

the one of Fung and Hsieh (2000). The remaining difference can be explained by the

difference in time period covered and in the database used. Interestingly, our results

indicate that the longer the estimation period, the bigger the bias.

66
 Contrarily to the results reported for the survivorship bias, Table 10 yields a

consistently greater estimate of the instant history bias for the pre-1994 than the post-

1994 period when 24 months or more are deleted.

Based on these pieces of evidence regarding the presence of biases, the rest of this

study will focus on performance level and persistence for the 1994-2000 sub-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.

Results for the whole 1984-2000 period are not reported because we may not soundly

claim that the pre- and post-1994 time windows exhibit the same biases with the same

severity.

67
5.3 Basic Performance

Before going in the heart of our work, panel A of Table 10 contains descriptive

statistics of the funds in our database, including living and dead funds. Given that MAR

and HFR classify differently the individual hedge funds, 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.18 We contrast hedge funds data

against the descriptive statistics of the market proxy, the MSCI World excluding US, Fama

and French’s (1993) SML and HML, Fama and French’s (1998) international IHML,

Carhart’s (1997) momentum factor, Lehman US aggregate bond index, Salomon World

government bond index, JP Morgan Emerging Market Bond Index, Lehman BAA corporate

bond index (default spread), and Goldman Sachs Commodity Index. These statistics are

reported in panel B of Table 10.

Panel A shows that the highest mean return was achieved by the US Opportunistic

Small Caps (2.58%) followed by the US Opportunistic Growth (2.15%) and by the Sector

(1.91%). Strategies that offer the lowest mean return are Foreign Exchange (0.64%),

Short Sellers (0.71%) and Funds of Funds (0.78%), whereas the mean return of the

whole database is 1.29%. The results are similar for mean excess returns.

18
The description of these strategies is available upon request.

68
Table 10: Estimation of Instant Return History Bias

1/84-6/00
Mean Annual Av. Nb of Fds per
Difference
Return Month
All 1,49% 768

Without 12M 1,42% 0,08% 644

Without 24M 1,26% 0,23% 527

Without 36M 1,20% 0,29% 425

Without 60M 1,15% 0,35% 269

1/84-12/93
Mean Annual Av. Nb of Fds per
Difference
Return Month
All 1,62% 207

Without 12M 1,58% 0,04% 147

Without 24M 1,35% 0,28% 103

Without 36M 1,27% 0,35% 71

Without 60M 1,21% 0,41% 32

1/94-6/00
Mean Annual Av. Nb of Fds per
Difference
Return Month
All 1,29% 1631

Without 12M 1,19% 0,10% 1308

Without 24M 1,16% 0,13% 1017

Without 36M 1,13% 0,16% 773

Without 60M 1,10% 0,19% 426

This Table reports the Instant History Bias calculated from our database. Our combined
database contains 2796 hedge funds, including 1995 survived funds and 801 dissolved
funds as of June 2000. 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, 24, 36 and 60 months of returns (the adjusted
observable portfolio). All returns are net of fees and on a monthly basis. Numbers in the
table are monthly percentage.

69
 When standard deviation is taken into account through the Sharpe measure (the

ratio of excess return and standard deviation), results are somewhat different. Funds

offering the best trade-off between risk and return are the US Opportunistic Small Caps

(0.3814), followed by the Market Neutral Convertible Arbitrage (0.3633) and the Market

Neutral Relative Value Arbitrage (0.2766). The worst Sharpe ratio is obtained by the Short

Sellers (0.0465), which are also in the worst performing funds when risk is not taken into

account.

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..

Panel B of Table 11 shows that the mean excess return of the Market Proxy is

1.26% per month (about 16% per year) and statistically different from zero. This large

value indicates that the period under review is unambiguously bullish. The mean excess

premium of the MSCI World excluding US is an insignificant 0.43% per month. The

average SMB and HML returns are insignificant, unlike the results obtained by Fama and

French (1993) and Carhart (1997).19 The international HML and the momentum factor give

more interesting values. They respectively produced an average premium of 0.28% and

1.15% per month, the latter being economically as well as statistically significant.

19
The differences in SMB and HML can be explained by the different periods covered by our studies.

Their high variance suggests a very unstable behavior.

70
Table 11: Descriptive Statistics of Hedge Funds Strategies and
Passive Investment Strategies

Panel A: Hedge Funds strategies

Sub- Living Dead Mean t(mean)

No of Fds Std. Dev.
strategy Funds Funds Return =0
Event Driven 108 100% 56 52 1,25 3,49 5,06
Distressed Securities 49 45,4% 25 24 0,98 3,34 4,12
Risk Arbitrage 50 46,3% 27 23 1,49 3,76 5,59
No sub-strategy 9 8,3% 4 5 1,08 5,53 2,76
Global 349 100% 184 165 1,02 2,2 6,5
International 62 17,8% 27 35 1,02 2,26 6,35
Emerging 203 58,2% 123 80 0,94 1,98 6,69
Regional Established 84 24,1% 34 50 1,36 4,2 4,56
Global Macro 319 259 60 1,68 4,11 5,75
Market Neutral 711 100% 522 189 1,06 3,78 3,95
Long/short 150 21,1% 83 67 1,31 4,2 4,41
Convertible Arbitrage 44 6,2% 38 6 0,89 9,4 1,33
Fixed Income 61 8,6% 50 11 0,79 2,5 4,43
Stock Arbitrage 392 55,1% 309 83 1,08 3,98 3,81
Mortgage-Backed Sec. 56 7,9% 42 14 0,84 3,78 3,11
Relative Value Arb. 8 1,1% 0 8 1,2 5,94 2,84
Short Sellers 24 9 15 0,71 1,52 6,57
US Opportunistic 212 100% 139 73 1,87 4,29 6,14
Growth 88 41,5% 67 21 2,15 4,49 6,75
Value 110 51,9% 61 49 1,61 4,21 5,38
Small Caps 14 6,6% 11 3 2,58 6,36 5,71
Long Only Leveraged 27 16 11 1,73 3,31 7,33
Market Timing 49 40 9 1,1 3,29 4,69
Equity non-Hedge 120 86 34 1,68 3,53 6,7
Foreign Exchange 9 9 0 0,64 9,29 0,97
Sector 186 132 54 1,91 4,34 6,2
Funds of Funds 347 229 118 0,78 3,55 3,09
Non classified 335 314 21 1,42 3,75 5,33
All funds 2796 1995 801 1,29 3,27 5,56

71
Panel A (continued): Hedge Funds strategies

Mean exc. t(mean Sharpe

Median Min Max
return exc.) =0 ratio

Event Driven 0,94 -8,62 6,71 0,85 2,36 0,17

Distressed Securities 0,86 -9,75 6,37 0,57 1,96 0,14
Risk Arbitrage 1,02 -7,97 10,47 1,05 2,65 0,19
No sub-strategy 1,32 -8,79 6,47 0,68 3,47 0,25
Global 0,86 -17,46 10,94 0,61 1,32 0,09
International 0,88 -12,66 9,83 0,62 1,37 0,10
Emerging 0,75 -20,01 11,56 0,54 1,13 0,08
Regional Established 1,12 -8,97 9,54 0,96 2,96 0,21
Global Macro 1,36 -6,27 9,93 1,27 3,12 0,22
Market Neutral 0,92 -3,36 5,29 0,66 2,34 0,17
Long/short 0,98 -1,7 6,54 0,91 2,91 0,21
Convertible Arbitrage 0,84 -1,79 2,31 0,48 5,11 0,36
Fixed Income 0,65 -5,62 7,99 0,38 1,22 0,09
Stock Arbitrage 0,97 -4,59 6,1 0,67 2,49 0,18
Mortgage-Backed Sec. 0,75 -4,31 3,09 0,43 1,96 0,14
Relative Value Arb. 1,24 -5,82 6,62 0,79 3,89 0,28
Short Sellers 0,81 -10,4 10,96 0,31 0,65 0,05
US Opportunistic 1,6 -11,75 13,71 1,47 3,36 0,24
Growth 1,85 -12,41 17,52 1,75 3,65 0,26
Value 1,27 -11,74 10,2 1,2 3,15 0,22
Small Caps 2,41 -7,92 17,53 2,18 5,37 0,38
Long Only Leveraged 1,24 -14,47 15,64 1,32 2,54 0,18
Market Timing 0,77 -2,47 10,38 0,69 2,08 0,15
Equity non-Hedge 1,41 -14,535 11,2 1,28 2,68 0,19
Foreign Exchange 0,61 -5,46 3,05 0,23 3,42 0,24
Sector 1,47 -12,54 12,32 1,51 3,42 0,24
Funds of Funds 0,8 -8,7 6,83 0,38 1,71 0,12
Non classified 1,04 -2,91 6,71 1,02 2,69 0,19
All funds 1,01 -7,88 7,87 0,89 2,25 0,16

72
Panel B: Passive Strategies
Mean
Mean t(excess
t(mean) = 0 Std. Dev. Median Min Max excess Sharpe ratio
Return mean) = 0
return
Equity

Market Proxy 1,64 5,75 4 2,34 -15,31 7,84 1,26 4,44 0,315

MSCI World Excluding US 0,97 3,18 4,28 1,32 -13,61 12,9 0,43 1,41 0,100

F&F SMB Factor -0,16 -0,48 4,74 -0,45 -16,69 21,49 -0,61 -1,82 -0,129

F&F HML Factor 0 -0,02 3,61 0,1 -9,05 12,03 -0,41 -1,61 -0,115

F&F HML International Factor 0,28 1,28 3,04 0,12 -12,75 7,92 -0,19 -0,87 -0,062

Momentum Factor 1,15 3,7 4,39 1,03 -10,93 13,88 0,81 2,58 0,184

Bond

Lehman US Aggregate -0,12 -1,41 1,19 -0,02 -2,73 3,87 -0,54 -6,32 -0,449

Salomon World Government 0,45 3,64 1,72 0,27 -3,43 5,94 0,05 0,38 0,027

JPM Emerging Market Bond Index 0,97 2,64 5,15 1,68 -21,98 10,55 0,52 1,41 0,100

Lehman BAA Corp. 0,52 4,86 1,51 0,55 -3,12 4,8 0,13 1,18 0,084

Commodity

Goldman Sachs Commodity 0,75 2,06 5,14 0,95 -12,28 15,79 0,26 0,71 0,050
This table shows the inception date, mean returns, t-stat for mean = 0, standard deviation, medians, minimum, maximum, mean excess returns, t-stat for mean excess
return = 0, and Sharpe ratios for the individual hedge funds in our combined MAR/HFR database following 12 strategies and 15 sub-strategies, and for 11 passive
investment strategies for the 1/1994 and 6/2000 period. Sharpe ratio is the ratio of excess return and standard deviation. In panel A, No of Fds represent the number of
funds following a particular strategy (or sub-strategy), Living Funds and Dead Funds represents the number of surviving and dead funds (in June 2000, without
considering the new funds established in 6/2000). We calculate the Mean Excess Return and the Sharpe ratio considering Ibbotson Associates one-month T-bills.
Numbers in the table are monthly percentage.
 The highest mean return was obtained by the Market Proxy for equity, and by the

JP Morgan Emerging Market Bond Index for the bond. The Sharpe ratios bring the same

results, with the only difference that the Lehman BAA Corporate (0.0842) has a Sharpe

ratio very close to the one obtained by the JP Morgan Emerging Market Bond Index (0.1).

The Sharpe ratio obtained by our whole hedge fund database (0.1596) is lower

than the one for the Market Proxy (0.3154), and higher than for the MSCI World Excluding

US (0.0999).

5.4 Correlation

Fung and Hsieh (1997), Schneeweis and Spurgin (1998), Liang (1999) and Amin

and Kat (2001) report a weak correlation between hedge funds and other securities.

Hence, the addition of hedge funds to a traditional portfolio should improve its risk-return

trade-off.

Table 12 reports correlation coefficients among and between hedge funds and

passive investment strategies.

Panel A reports correlations among hedge funds strategies. There is a high

variability between different strategies, ranging from 0.97 (between Equity non-Hedge

and US Opportunistic) to –0.80 (between Short Selling and US Opportunistic). 39

correlation coefficients (40%) are greater than 0.80 and 14 (14%) are negative. In

particular, Short Sellers are negatively correlated with all the other hedge fund strategies.

74
 Panel B reports correlation coefficients between hedge funds and equity, bond and

commodity indices. The range is narrower than in the previous case (from –0.65 to 0.86).

Correlation coefficients between hedge funds strategies and the Market Proxy are, in

almost all cases, greater than 0.5 whereas they are always smaller than 0.3 with the MSCI

World excluding US and than 0.5 with bond indices (except for the default factor). These

results confirm that hedge funds strategies are weakly correlated with traditional

investment tools.20

Panel C displays correlations among Passive Investment strategies. All coefficients

are below 0.46 (and higher than –0.53) and almost 90% of them are below 0.3, too low to

raise serious multi-colinearity concerns.

20
Except with the market proxy, but this result can easily be understood since the market proxy

contains almost all the American market.

75
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. 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, the discussion of performance

models will be performed on the 1994-2000 period, when the database can be considered

as fully reliable.

6.1 Performance Measurement using the CAPM

The first performance model used is the CAPM based single index model. Panel A of

Table 13 reports the results for the strategies, sub-strategies and for the All Funds

category. We use equally weighted portfolio excess returns for each investment style and

for the All Funds category, and we estimate the model for each fund individually.21 The

last columns give the distribution of individually estimated alphas per strategy, with the

percentage of significantly positive, insignificant and negative alphas at the 5% level. This

approach enables us to analyse hedge funds performance in more details.

21
To make individual estimation, we require all funds to have consecutive monthly return history for at

least 24 months, so that relatively accurate risk measures can be estimated.

76
Table 12: Correlation between Hedge Funds and Passive Investment Strategies

Panel A: Correlation between Hedge Funds strategies

EVT GLB MAC MKN SHS OPP LOL MKT ENH FEX SEC FOF NCL ALL

EVT 1

GLB 0,74 1

MAC 0,85 0,74 1

MKN 0,83 0,82 0,89 1

SHS -0,7 -0,46 -0,73 -0,60 1

OPP 0,85 0,74 0,95 0,84 -0,8 1

LOL 0,85 0,78 0,91 0,87 -0,71 0,91 1

MKT 0,27 0,17 0,44 0,31 -0,22 0,39 0,31 1

ENH 0,86 0,76 0,94 0,83 -0,78 0,97 0,88 0,43 1

FEX 0,77 0,69 0,61 0,71 -0,51 0,62 0,65 0,15 0,65 1

SEC 0,87 0,77 0,94 0,85 -0,78 0,95 0,90 0,42 0,95 0,66 1

FOF 0,83 0,92 0,88 0,95 -0,58 0,84 0,89 0,26 0,83 0,72 0,86 1

NCL 0,1 0,13 0,38 0,35 -0,30 0,35 0,29 0,27 0,26 -0,15 0,30 0,25 1

ALL 0,89 0,87 0,96 0,94 -0,71 0,95 0,93 0,38 0,94 0,71 0,95 0,95 0,32 1
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 0,69 0,64 0,72 0,58 -0,54 0,77 0,67 0,39 0,82 0,58 0,71 0,65 0,10 0,73

WXU 0,13 0,29 0,10 0,25 0,04 0,06 0,18 -0,19 0,05 0,26 0,06 0,27 -0,07 0,16

SMB 0,54 0,36 0,57 0,55 -0,65 0,57 0,58 0,20 0,52 0,31 0,63 0,47 0,27 0,56

HML -0,32 -0,31 -0,45 -0,31 0,49 -0,48 -0,41 -0,30 -0,53 -0,32 -0,48 -0,32 0,00 -0,43

IHML -0,14 -0,05 -0,31 -0,26 0,34 -0,31 -0,30 -0,17 -0,21 -0,02 -0,30 -0,17 -0,29 -0,24

MOM 0,14 0,05 0,29 0,24 -0,35 0,25 0,29 0,28 0,18 0,03 0,26 0,18 0,20 0,23

LUS 0,09 -0,05 0,06 -0,03 -0,06 0,06 0,09 0,05 0,06 0,02 0,04 -0,01 0,02 0,02

SWG 0,02 -0,23 0,00 -0,12 -0,12 0,05 0,00 0,20 0,08 -0,10 0,04 -0,19 0,01 -0,06

MEM 0,20 0,31 0,16 0,24 -0,01 0,11 0,25 -0,18 0,09 0,35 0,08 0,30 -0,16 0,20

BAA 0,66 0,55 0,49 0,58 -0,40 0,51 0,56 0,19 0,54 0,86 0,53 0,57 -0,13 0,58

GSC 0,35 0,18 0,31 0,32 -0,27 0,28 0,34 0,15 0,31 0,39 0,31 0,25 0,03 0,30
Panel C: Correlation between Passive Investment strategies

MKT WXU SMB HML IHML MOM LUS SWG MEM BAA

MKT 1

WXU -0,1 1

SMB 0,02 0,09 1

HML -0,43 0,02 -0,28 1

IHML -0,05 0,23 -0,38 0,08 1

MOM 0,08 -0,23 0,31 -0,14 -0,52 1

LUS 0,13 0,04 -0,04 0,06 -0,06 0,04 1

SWG 0,13 -0,30 -0,10 -0,07 0,03 0,14 0,38 1

MEM -0,01 0,46 0,06 0,06 0,15 -0,12 0,24 -0,15 1

BAA 0,45 0,19 0,20 -0,27 0,00 -0,01 0,15 0,04 0,23 1

GSC 0,44 -0,03 -0,03 -0,06 -0,1 0,22 0,19 0,38 -0,01 0,43

This Table reports the correlation coefficient between hedge funds strategies (Panel A), between hedge funds strategies and passive investment strategies (Panel B) and
between passive investment strategies for the January 1994-June 2000 period. EVT = Event Driven, GLB = Global, MAC = Global Macro, MKN = Market Neutral, SHS =
Short Selling, OPP = US Opportunistic, LOL = Long only Leveraged, MKT = Market Timing, ENH = Equity non-Hedge, FEX = Foreign Exchange, SEC = Sector, FOF = Funds
of Funds, NCL = Non Classified, ALL = All hedge funds, MKT = Market Proxy, WXU = World excluding US, IHML = International HML, MOM = Momentum, LUS = Lehman
US Aggregate Bond Index, SWG = Salomon World Government Bond Index, MEM = JP Morgan Emerging Market Bond Index, BAA = Lehman BAA Corporate Bond Index
and GSC = Goldman Sachs Commodity Index.
 The betas estimated in Panel A are rather low, except for US Opportunistic Growth

and the Long Only Leveraged, suggesting the need to use a more detailed model. Overall,

two thirds of the strategies produce significantly positive alphas. The All Funds category

also significantly out-performs the market at the 1% level. In almost all out-performing

strategies, more than 30% of the alphas are significant. Surprisingly, for some strategies

(e.g. Equity non-Hedge or the Non Classified funds), more than 80% of the individual

funds do not significantly out-perform the market, inducing that the best funds must have

obtained extremely high returns.22

6.2 Performance Measurement using Multi-Factor Models

It is presumably better to use a multi-factor model to account for all possible

investment strategies. In Panel B of Table 13, 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

Panels B and C reveal that the premium on the SMB factor is, in almost all cases,

significantly positive. However, in the Short Sellers strategy, 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. The momentum factor does not prove to be a strong

indicator of hedge funds behaviour. Only 6 out of 28 investment styles exhibit significant

momentum loadings (at the 10% level). Moreover, the sign of the coefficient is in 3 cases

negative, indicating momentum contrarian strategies.

22
These results are remarkably similar to the 1984-1994 period (data available upon request).
23
The Fama and French (1993) model produces results very similar to the Carhart model.

80
 Panel C also indicates that the World excluding US, US Bond, Default, World

Government Bond and Commodity factors add explanatory power in 20 to 33% of the

cases. The default factor has, as we can expect, a negative impact in half of the cases.

The Emerging Bond factor adds explanatory power in 16 of the strategies and sub-

strategies. It is highly significant (at the 1% level) in more than half of the strategies, as

well as for the All Funds category.

These results provide some insight into the preferences of hedge funds managers

depending on the strategy followed:

ƒ Almost all managers seem to prefer smaller stocks;

ƒ Most Event Driven, Market Neutral and US Opportunistic managers prefer stocks with

high book-to-market ratios;

ƒ Some Event Driven and Market Neutral managers follow a momentum strategy and

others are momentum contrarian;

ƒ More than half of the managers invest in emerging bond markets.

81
Table 13: Performance Measurement using the CAPM,

Carhart’s 4-factor model and the combined model.

Alpha Distribution
Alpha Mkt R² adj No Fds
+/0/-

Event Driven 0,35% ** 0,42 *** 0,57 84 25% 73% 2%

Distressed Sec. 0,10% 0,39 *** 0,43 37 11% 89% 0%

Risk Arbitrage 0,55% *** 0,45 *** 0,53 38 42% 53% 5%

No sub-strategy 0,18% 0,42 *** 0,4 9 11% 89% 0%

Global -0,13% 0,62 *** 0,45 258 12% 86% 3%

International -0,03% 0,54 *** 0,54 50 10% 90% 0%

Emerging -0,26% 0,67 *** 0,38 151 9% 87% 4%

Regional Est. 0,43% ** 0,44 *** 0,53 57 19% 79% 2%

Global Macro 0,62% *** 0,55 *** 0,62 252 23% 75% 2%

Market Neutral 0,40% *** 0,21 *** 0,42 553 33% 66% 1%

Long/short 0,72% *** 0,16 *** 0,26 102 34% 64% 2%

Convertible Arb. 0,41% *** 0,06 *** 0,08 39 74% 26% 0%

Fixed Income -0,08% 0,39 *** 0,43 48 10% 88% 2%

Stock Arbitrage 0,38% *** 0,24 *** 0,38 310 31% 68% 1%

Mortgage-Back. 0,35% *** 0,07 *** 0,06 46 28% 70% 2%

82
Panel A (continued): Single index model

Alpha Distribution
Alpha Mkt R² adj No Fds
+/0/-

Rel. Value Arb. 0,51% ** 0,2 *** 0,19 8 50% 50% 0%

Short Sellers 0,86% *** -0,48 *** 0,39 18 28% 72% 0%

US Opport. 0,50% *** 0,8 *** 0,7 180 19% 81% 1%

Growth 0,60% *** 0,96 *** 0,65 68 24% 76% 0%

Value 0,36% ** 0,7 *** 0,74 100 14% 85% 1%

Small Caps 1,39% *** 0,66 *** 0,35 12 33% 67% 0%

Long Only Lev. 0,17% 0,96 *** 0,54 18 11% 83% 6%

Market Timing 0,47% *** 0,18 *** 0,14 34 26% 74% 0%

Equity non-Hed. 0,24% 0,86 *** 0,77 112 9% 89% 2%

Foreign Exch. 0,01% 0,19 *** 0,37 9 0% 67% 33%

Sector 0,68% *** 0,69 *** 0,62 128 30% 67% 2%

Funds of Funds -0,07% 0,37 *** 0,49 278 13% 83% 4%

Non classified 4,83% *** 0,08 0,01 230 13% 86% 0%

All funds 0,36% ** 0,44 *** 0,64 2154 24% 74% 2%

Mean R²adj 0,44 Mean 23% 74% 3%

83
 Alpha Mkt SMB HML

Event Driven 0,43% *** 0,44 *** 0,24 *** 0,11 ***

Distressed Sec. 0,24% 0,42 *** 0,29 *** 0,14 ***

Risk Arbitrage 0,62% *** 0,46 *** 0,22 *** 0,09 ***

No sub-strategy 0,08% 0,41 *** 0,14 *** 0,08

Global 0,03% 0,63 *** 0,26 *** 0,10

International 0,10% 0,54 *** 0,21 *** 0,06

Emerging -0,07% 0,68 *** 0,29 *** 0,09

Regional Est. 0,53% *** 0,44 *** 0,21 *** 0,07

Global Macro 0,67% *** 0,51 *** 0,27 *** 0,03

Market Neutral 0,42% *** 0,21 *** 0,14 *** 0,05 *

Long/short 0,71% *** 0,17 *** 0,15 *** 0,10 ***

Convertible Arb. 0,45% *** 0,08 *** 0,06 *** 0,06 ***

Fixed Income 0,02% 0,33 *** 0,16 *** -0,08

Stock Arbitrage 0,41% *** 0,24 *** 0,17 *** 0,06 *

Mortgage-Back. 0,29% *** 0,08 *** 0,01 0,04

84
Panel B (continued): Carhart's 4-factor model

Alpha Distribution
PR1YR R² adj No Fds
+/0/-

Event Driven -0,06 *** 0,78 84 27% 69% 4%

Distressed Sec. -0,11 *** 0,70 37 14% 81% 5%

Risk Arbitrage -0,04 * 0,68 38 42% 55% 3%

No sub-strategy 0,11 *** 0,50 9 22% 78% 0%

Global -0,12 ** 0,53 258 14% 84% 3%

International -0,09 *** 0,63 50 10% 88% 2%

Emerging -0,14 * 0,45 151 11% 87% 3%

Regional Est. -0,06 * 0,66 57 25% 72% 4%

Global Macro 0,03 0,83 252 31% 68% 2%

Market Neutral 0,00 0,64 553 37% 62% 1%

Long/short 0,01 0,59 102 36% 63% 1%

Convertible Arb. -0,04 *** 0,20 39 77% 23% 0%

Fixed Income -0,01 0,53 48 8% 92% 0%

Stock Arbitrage 0,00 0,59 310 38% 61% 1%

Mortgage-Back. 0,04 ** 0,07 46 30% 67% 2%

85
Panel B (continued 2x): Carhart's 4-factor model

Alpha Mkt SMB HML

Mortgage-Back. 0,29% *** 0,08 *** 0,01 0,04

Rel. Value Arb. 0,59% *** 0,24 *** 0,12 * 0,15 **

Short Sellers 0,77% *** -0,38 *** -0,35 *** 0,07

US Opport. 0,63% *** 0,75 *** 0,38 *** 0,02

Growth 0,78% *** 0,83 *** 0,43 *** -0,11

Value 0,46% *** 0,7 *** 0,29 *** 0,11 ***

Small Caps 1,57% *** 0,64 *** 0,55 *** 0,14 *

Long Only Lev. 0,25% 0,9 *** 0,54 *** 0,11

Market Timing 0,43% *** 0,14 *** 0,03 -0,06

Equity non-Hed. 0,44% *** 0,8 *** 0,35 *** -0,03

Foreign Exch. 0,07% 0,19 *** 0,08 *** 0,01

Sector 0,82% *** 0,63 *** 0,4 *** 0,01

Funds of Funds -0,02% 0,37 *** 0,19 *** 0,08 *

Non classified 4,76% *** 0,09 0,12 0,10

All funds 0,43% *** 0,42 *** 0,23 *** 0,04

86
Panel B (continued 3x): Carhart's 4-factor model

Alpha Distribution
PR1YR R² adj No Fds
+/0/-

Rel. Value Arb. -0,05 0,21 8 50% 50% 0%

Short Sellers -0,07 * 0,72 18 17% 83% 0%

US Opport. 0,00 0,90 180 31% 68% 2%

Growth 0,04 0,86 68 38% 60% 1%

Value -0,05 * 0,88 100 24% 74% 2%

Small Caps -0,06 0,64 12 42% 58% 0%

Long Only Lev. 0,06 0,77 18 17% 78% 6%

Market Timing 0,09 * 0,18 34 18% 82% 0%

Equity non-Hed. -0,06 *** 0,93 112 19% 80% 1%

Foreign Exch. -0,04 *** 0,42 9 11% 89% 0%

Sector 0,00 0,89 128 41% 59% 0%

Funds of Funds -0,02 0,63 278 20% 77% 3%

Non classified 0,06 0,06 230 43% 55% 2%

All funds -0,01 0,84 2154 30% 69% 2%

Mean R²adj 0,60 Mean 0,28 70% 2%

87
Panel C: The combined model

Alpha Mkt Wd x US SMB HML I HML

Event Driven 0,34% * 0,42 *** 0,01 0,25 *** 0,10 *** 0,05

Distressed Sec. 0,09% 0,40 *** 0,06 0,27 *** 0,12 *** 0,04

Risk Arbitrage 0,60% *** 0,44 *** -0,03 0,24 *** 0,08 *** 0,07

No sub-strategy -0,05% 0,43 *** 0,00 0,16 *** 0,10 0,05

Global -0,50% 0,68 *** 0,12 * 0,16 *** 0,07 0,04

International -0,21% 0,56 *** 0,13 ** 0,13 *** 0,03 -0,05

Emerging -0,71% 0,73 *** 0,13 * 0,17 *** 0,06 0,06

Regional Est. 0,27% 0,50 *** 0,10 ** 0,18 *** 0,09 * -0,02

Global Macro 0,56% *** 0,53 *** 0,06 * 0,25 *** 0,04 -0,09 *

Market Neutral 0,28% *** 0,21 *** 0,07 *** 0,11 *** 0,04 ** -0,04

Long/short 0,67% *** 0,16 *** 0,01 0,13 *** 0,08 *** -0,07 ***

Convertible Arb. 0,42% *** 0,05 *** 0,05 ** 0,05 *** 0,04 ** 0,02

Fixed Income -0,26% 0,36 *** 0,09 * 0,13 *** -0,07 * 0,04

Stock Arbitrage 0,22% * 0,25 *** 0,09 *** 0,13 *** 0,06 ** -0,05

Mortgage-Back. 0,23% 0,02 0,02 0,02 0,01 0,01

88
Panel C (continued): The combined model

PR1YR US Bd World GV Emerg Bd Default Comm. R² adj

Event Driven -0,03 -0,04 -0,01 0,06 *** 0,11 * 0,00 0,79

Distressed Sec. -0,08 ** -0,08 -0,03 0,07 ** 0,16 ** 0,00 0,74

Risk Arbitrage -0,02 0,01 -0,09 0,05 * 0,13 -0,01 0,66

No sub-strategy 0,12 *** -0,10 0,39 *** 0,06 -0,16 0,00 0,51

Global 0,01 -0,33 * -0,41 *** 0,13 *** -0,12 0,09 ** 0,67

International -0,02 -0,17 -0,21 * 0,06 * -0,05 0,08 *** 0,72

Emerging 0,02 -0,38 * -0,48 *** 0,16 *** -0,17 0,12 ** 0,6

Regional Est. -0,03 -0,22 ** -0,09 0,03 -0,16 * -0,03 0,71

Global Macro 0,03 -0,11 0,01 0,06 *** -0,06 -0,02 0,85

Market Neutral 0,02 -0,15 *** -0,06 0,03 0,07 0,01 0,72

Long/short 0,00 -0,12 * -0,05 0,02 0,10 * 0,02 * 0,61

Convertible Arb. -0,02 * 0,06 -0,11 *** 0,03 * 0,16 *** 0,00 0,49

Fixed Income 0,05 * -0,17 -0,02 0,04 -0,09 0,02 0,55

Stock Arbitrage 0,02 -0,19 *** -0,08 0,02 0,04 0,01 0,68

Mortgage-Back. 0,04 * -0,13 * -0,01 0,03 0,33 *** 0,00 0,22

89
Panel C (continued 2x): The combined model

Alpha Mkt Wd x US SMB HML I HML

Rel. Value Arb. 0,43% 0,21 *** 0,03 0,02 0,02 0,08

Short Sellers 0,75% *** -0,34 *** 0,01 -0,31 *** 0,10 0,07

US Opport. 0,46% *** 0,80 *** 0,09 ** 0,32 *** 0,04 -0,16 **

Growth 0,74% *** 0,85 *** 0,04 0,36 *** -0,12 -0,35 ***

Value 0,20% 0,77 *** 0,12 *** 0,26 *** 0,13 *** 0,02

Small Caps 1,34% *** 0,73 *** 0,08 0,46 *** 0,16 ** -0,25 ***

Long Only Lev. 0,00% 0,88 *** 0,14 * 0,47 *** 0,07 -0,14

Market Timing 0,48% ** 0,19 *** 0,02 0,09 *** 0,00 0,04

Equity non-Hed. 0,23% * 0,84 *** 0,08 ** 0,31 *** -0,02 -0,07

Foreign Exch. -0,02% 0,15 *** 0,00 0,07 *** -0,01 0,03

Sector 0,69% *** 0,63 *** 0,06 * 0,37 *** 0,00 -0,08

Funds of Funds -0,28% * 0,39 *** 0,07 * 0,14 *** 0,06 * -0,01

Non classified 4,86% *** 0,14 0,10 0,11 *** 0,14 -0,13

All funds 0,25% ** 0,44 *** 0,06 ** 0,19 *** 0,04 * -0,04

90
Panel C (continued 3x): The combined model

PR1YR US Bd World GV Emerg Bd Default Comm. R² adj

Rel. Value Arb. 0,02 0,03 -0,23 * 0,10 ** 0,05 0,05 * 0,36

Short Sellers -0,05 0,05 -0,14 -0,01 -0,07 -0,08 ** 0,72

US Opport. 0,00 -0,18 * 0,18 ** 0,04 * -0,29 ** 0,02 0,92

Growth -0,03 -0,21 0,15 0,10 *** -0,18 0,00 0,89

Value 0,01 -0,12 0,20 *** 0,00 -0,34 *** 0,04 ** 0,91

Small Caps -0,06 -0,27 0,14 0,05 -0,51 ** 0,06 0,65

Long Only Lev. 0,08 -0,17 0,00 0,19 *** 0,17 -0,01 0,82

Market Timing 0,06 0,02 0,21 * -0,05 -0,16 -0,12 ** 0,25

Equity non-Hed. -0,05 * -0,20 ** 0,21 *** 0,03 -0,21 *** 0,04 * 0,94

Foreign Exch. -0,03 -0,10 * -0,11 0,08 *** 0,23 *** -0,01 0,56

Sector 0,00 -0,18 0,06 0,02 0,01 0,03 0,89

Funds of Funds 0,04 -0,18 * -0,20 ** 0,08 *** -0,01 0,04 * 0,75

Non classified 0,03 0,06 0,08 -0,12 *** -0,20 -0,04 0,05

All funds 0,01 -0,16 ** -0,05 0,05 *** -0,03 0,01 0,88

Mean R²adj 0,66

91
Panel C (continued 4x): The combined model

No Fds Alpha Distr. + / 0 / -

Event Driven 84 27% 69% 4%

Distressed Sec. 37 19% 78% 3%

Risk Arbitrage 38 39% 55% 5%

No sub-strategy 9 11% 89% 0%

Global 258 10% 77% 12%

International 50 10% 82% 8%

Emerging 151 8% 75% 17%

Regional Est. 57 18% 77% 5%

Global Macro 252 42% 56% 1%

Market Neutral 553 30% 65% 5%

Long/short 102 27% 71% 2%

Convertible Arb. 39 59% 41% 0%

Fixed Income 48 8% 79% 13%

Stock Arbitrage 310 29% 65% 6%

Mortgage-Back. 46 33% 63% 4%

92
Panel C (continued 5x): The combined model

No Fds Alpha Distr. + / 0 / -

Rel. Value Arb. 8 50% 38% 13%

Short Sellers 18 44% 56% 0%

US Opport. 180 21% 76% 3%

Growth 68 28% 72% 0%

Value 100 16% 79% 5%

Small Caps 12 25% 75% 0%

Long Only Lev. 18 17% 78% 6%

Market Timing 34 12% 88% 0%

Equity non-Hed. 112 28% 71% 1%

Foreign Exch. 9 0% 78% 22%

Sector 128 30% 69% 2%

Funds of Funds 278 14% 72% 14%

Non classified 230 35% 62% 3%

All funds 2154 26% 68% 6%

Mean 25% 70% 5%

This Table presents the results of the estimation of the Single Index Model (Panel A), of Carhart’s
(1997) Model (Panel B) and of our Combined Model (Panel C) for the 1/1984-6/2000 period. We report
the OLS estimators for equally weighted portfolio’s per investment strategy, sub-strategy and for all
funds. 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. We report the percentage of significantly positive
alpha’s (+), significantly negative alpha’s (-) and alpha’s that are insignificantly different from zero (0) at
the 5% level. The next to last column reports the number of individual funds used for the individual
estimation of the last column. T-stat are heteroskedasticity consistent. *** Significant at the 1% level, **
Significant at the 5% level and * Significant at the 10% level.

93
 These results are close to those found by Mitchell and Pulvino (2001) and Agarwal

(2001) for the funds following Event Driven strategies. We find that 27% of the hedge

funds show significant excess return, matching the 27% they found. This independently

confirms that our approach is able to capture important risk exposure of hedge funds.

Carhart (1997) and Gruber (1996) examine US mutual fund strategies and report

that managers prefer smaller stocks as well as growth stocks. The first evidence is

consistent with our finding, while the second one is opposite. However, this difference

does not exist for all hedge funds managers, given that the HML factor is only significant

for some strategies.

Comparing the alpha distribution of Panels B and C shows that, taking more factors

into account induces that fewer individual funds significantly out-performed the market,

and more funds have insignificant or negative excess returns.

Evidence on alphas obtained in Panel C is contrasted. The excess returns for Event

Driven, Global, Long Only Leveraged, Equity non-Hedge, and Foreign Exchange strategies

are insignificant (at the 5% level), whereas the alpha for the Fund of Funds strategy is

negative and significant. All other categories display positive and significant alphas. Our

results are in most cases confirmed by the last column. In the All Funds category, for

example, more than 25% of the individual alphas are significantly positive at the 5% level.

Considering the All Funds category, we can observe that hedge funds as a whole:

ƒ Deliver significant excess returns (one fourth of the individual funds gave significant

positive excess return) ;

ƒ Seem to prefer smaller stocks ;

ƒ Invest in Emerging Market Bonds and suffer from the US Bond market.

94
 Overall it seems that the combined model does a very good job in describing hedge

funds behaviour. The average R²adj increases from 0.44 for the single factor model, to

0.60 for the 4-factor model and to 0.66 for our combined model. The combined model

gives adjusted R² higher than 0.65 for all strategies but the Market Timing (0.25), the

Non-classified funds (0.05) and Foreign Exchange funds (0.56). It seems particularly

adapted to Equity non-hedge (0.94), US Opportunistic (0.92), Sector (0.89), Global Macro

(0.85), Long only Leveraged (0.82) and Event Driven funds (0.79). The R²adj for the All

Funds category increases from 0.66 for the single factor model, to 0.84 for the 4-factor

model and to 0.88 for our combined model. The mean R²adj for the individual hedge funds

estimation is up too. Carhart’s (1997) model raises the R²adj by an average 10% over the

single index model, but our combined model increases it again by another 7%. For the All

Funds category, the increase from the CAPM to Carhart’s model is 10% and from Carhart’s

model to our combined model is another 7%.

Our R²adj are also higher than those obtained by Brown et al. (2001) and Fung and

Hsieh (1997). They report R² lower than 0.20 in all cases for groups of funds. Schneeweis

and Spurgin (1998) report R²adj between -0.09 and 0.67 with a mean of 0.31 for several

hedge funds strategies. Comparing their results with ours for strategies that exist across

the two databases, we get greater R²adj in all cases. For several HFR strategies, Liang

(1999) finds unadjusted coefficients ranging from 0.23 to 0.77, with an average of 0.49:

taking the same strategies, our R²adj range between 0.27 and 0.92 with an average of

0.63.

95
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 out-

perform 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 sub-

period; 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 Sub-

Periods

1 sub-period 2 sub-periods Asian crisis Pre-1994

1/94 - 6/00 1/94 - 12/96 1/97 - 6/00 1/97-12/98 1/97-6/98 1/84 - 12/93

Event Driven 0,34% * 0,41% * 0,08% 0,02% -0,05% -0,21%

Distressed Sec. 0,09% 0,17% -0,11% 0,19% -0,07% 0,07%

Risk Arbitrage 0,60% *** 0,60% ** 0,31% * -0,13% 0,02% -0,29%

No sub-strategy -0,05% 0,40% -0,45% -0,17% -0,25% -0,53% *

Global -0,50% 0,23% -1,21% *** -1,02% -1,30% ** 1,01% *

International -0,21% -0,33% -0,40% -0,99% *** -1,10% *** 1,29% ***

Emerging -0,71% 0,47% -1,70% *** -1,18% -1,55% ** 0,92%

Regional Est. 0,27% 0,38% *** -0,07% -0,46% -0,54% 1,21% ***

Global Macro 0,56% *** 0,45% * 0,56% *** 0,46% -0,07% 0,85% ***

Market Neutral 0,28% *** 0,20% 0,19% ** 0,15% -0,04% 0,63% ***

Long/short 0,67% *** 0,40% * 0,72% *** 0,66% *** 0,42% ** 0,05%

Convertible Arb. 0,42% *** 0,46% *** 0,36% *** 0,40% *** 0,40% *** 0,55% ***

Fixed Income -0,26% -0,47% -0,41% -0,72% -1,23% ** 1,01% ***

Stock Arbitrage 0,22% * 0,16% 0,13% 0,08% -0,11% 0,81% ***

Mortgage-Back. 0,23% 0,58% *** -0,17% -0,13% -0,12% 0,00%

97
 1 sub-period 2 sub-periods Asian crisis Pre-1994

1/94 - 6/00 1/94 - 12/96 1/97 - 6/00 1/97-12/98 1/97-6/98 1/84 - 12/93

Rel, Value Arb. 0,43% 1,20% *** 0,26% 0,54% 1,20% *** NA

Short Sellers 0,75% *** 0,70% 0,51% 0,66% ** 0,48% 0,13%

US Opport. 0,46% *** 0,57% *** 0,41% ** 0,40% -0,05% 0,73% ***

Growth 0,74% *** 0,94% *** 0,64% ** 0,65% ** 0,19% 1,13% ***

Value 0,20% 0,26% * 0,19% 0,21% -0,18% 0,47% **

Small Caps 1,34% *** 2,56% *** 0,60% * 0,44% -0,23% 1,70%

Long Only Lev. 0,00% 0,28% -0,21% -0,20% -0,42% NA

Market Timing 0,48% ** 0,10% 0,49% 0,22% 0,23% 0,22%

Equity non-Hed. 0,23% * 0,52% ** 0,07% 0,04% -0,33% 0,06%

Foreign Exch. -0,02% 0,20% -0,25% 0,13% 0,39% *** 0,83% ***

Sector 0,69% *** 0,77% *** 0,55% * 0,42% * 0,16% 1,10% **

Funds of Funds -0,28% * -0,28% -0,43% ** -0,33% -0,48% ** 0,61% ***

Non classified 4,86% *** 4,14% *** 5,57% *** 4,38% *** 3,79% *** 3,84% ***

All funds 0,25% ** 0,29% * 0,10% 0,04% -0,24% 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, sub-
strategies, 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 sub-

periods. 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 sub-

period.

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 under-

performed the market on average, while we find only small percentages of under-

performing 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. Alpha Mkt Wd x US SMB HML
Return

D 1a 1,68% 0,07 0,00 0,74 *** 0,05 0,66 *** -0,03

D 1b 1,26% 0,05 0,00 0,62 *** 0,12 0,40 *** -0,14

D 1c 1,11% 0,04 0,00 0,55 *** 0,08 0,39 *** -0,02

D1 1,35% 0,05 0,00 0,64 *** 0,09 0,48 *** -0,07

D2 0,89% 0,03 0,00 0,55 *** 0,10 0,33 *** 0,05

D3 0,83% 0,03 0,00 0,50 *** 0,05 0,24 *** 0,10 **

D4 0,82% 0,02 0,00 0,47 *** 0,06 ** 0,17 *** 0,09 ***

D5 0,71% 0,02 0,00 0,40 *** 0,05 * 0,13 *** 0,09 ***

D6 0,57% 0,02 0,00 0,32 *** 0,06 ** 0,13 *** 0,09 ***

D7 0,62% 0,02 0,00 * 0,33 *** 0,03 0,12 *** 0,07 **

D8 0,57% 0,02 0,00 0,36 *** 0,06 * 0,13 *** 0,04

D9 0,80% 0,02 0,00 ** 0,38 *** 0,06 0,09 *** 0,00

D 10 0,89% 0,03 0,00 0,50 *** 0,07 0,04 -0,05

D 10a 0,82% 0,03 0,00 0,48 *** 0,03 0,09 ** -0,05

D 10b 0,84% 0,03 0,00 0,54 *** 0,05 0,05 0,01

D 10c 1,03% 0,04 0,00 0,51 *** 0,14 -0,03 -0,12

1-10 spread 0,45% 0,05 0,00 0,13 0,02 0,45 *** -0,01

1a-10c spread 0,65% 0,07 0,00 0,24 -0,09 0,70 *** 0,09

1-2 spread 0,46% ** 0,02 0,00 0,08 -0,01 0,15 *** -0,11 *

9-10 spread -0,10% 0,02 0,00 -0,12 * -0,01 0,05 0,06

101
 World GV
I HML PR1YR US Bd Emerg Bd Default Com. R²adj
Bd

D 1a -0,32 0,19 -0,93 ** -0,19 0,12 0,44 0,13 0,74

D 1b -0,29 *** 0,14 * -0,38 0,02 0,05 0,09 0,11 ** 0,81

D 1c -0,18 0,10 -0,30 -0,18 0,03 0,02 0,03 0,79

D1 -0,26 * 0,14 -0,54 * -0,12 0,07 0,18 0,09 0,79

D2 -0,12 * 0,09 ** -0,38 ** -0,09 0,06 0,00 0,06 ** 0,85

D3 -0,05 0,03 -0,14 0,03 0,03 -0,02 0,01 0,88

D4 0,01 -0,01 -0,09 0,00 0,03 -0,06 0,01 0,87

D5 -0,01 0,00 -0,22 *** -0,07 0,07 *** -0,04 0,01 0,84

D6 0,01 -0,03 -0,11 -0,04 0,05 ** 0,07 0,00 0,82

D7 0,00 -0,05 * -0,05 -0,01 0,05 ** 0,04 -0,01 0,80

D8 0,06 -0,03 -0,06 -0,08 0,04 -0,02 -0,01 0,75

D9 0,03 -0,05 -0,07 -0,07 0,03 -0,12 -0,01 0,65

D 10 0,08 -0,02 -0,05 -0,09 0,06 -0,37 * 0,00 0,44

D 10a 0,15 -0,02 -0,12 -0,11 0,09 * -0,16 0,00 0,61

D 10b 0,07 -0,01 -0,02 -0,06 0,09 * -0,34 0,01 0,44

D 10c 0,02 -0,04 -0,01 -0,11 0,01 -0,63 * -0,01 0,22

1-10 spread -0,35 ** 0,16 -0,49 -0,03 0,00 0,55 * 0,09 0,52

1a-10c spread -0,34 0,23 -0,92 * -0,07 0,11 1,06 ** 0,14 0,48

1-2 spread -0,14 0,05 -0,15 -0,03 0,01 0,18 0,03 0,43

9-10 spread -0,05 -0,03 -0,01 0,02 -0,03 0,24 * -0,01 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 out-

or 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 suggests that, even if some hedge funds managers take a lot of risk, which

lead them to have very high or low returns for short period of time, most managers follow

less risky strategies that allow them to out-perform the market for long period of time.

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. (2001). Funds

that performed well – relative to the market, i.e. with a positive alpha – tend to pursue a

low variance strategy. Extreme performers tend not to remain in their situation; this can

be explained by dissolution or by the fact that they are more or less forced to take less

risk, returning to the middle of the pack. Brown et al. (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.

To summarize, Table 15 leads to the following conclusions:

ƒ Best performing funds follow momentum strategies whereas worst performing ones

may follow momentum contrarian strategies;28

ƒ Best performing funds do not invest significantly in Emerging Market Bond;

ƒ Average return funds prefer high book-to-market stocks, whereas best and worst

performing ones may prefer low book-to-market ones;

ƒ No persistence in performance exists for best and worst performing funds, but there is

weak evidence of persistence for middle deciles, where some funds significantly beat

28
This result may not be stable over time, since the period under review covers a bull market when

momentum strategies are likely to work. After the strong market reversal that took place after March

2000 would be very indicative of the economic significance of this result, but our database unfortunately

ends in June 2000. These issues are being investigated as a part of our ongoing research.

105
 the market with persistence. Evidence is more pronounced for the 1985-1993 period,

but it is likely to be driven by the absence of dissolved funds in this period.29

7.2 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 1997.30 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), in 1997. These

results confirm our previous conclusion that funds in the first decile have no persistence in

returns, and that there have more volatile returns than funds in lowest decile. In the

1/97-6/98 analysis, all deciles’ alphas are negative, 5 of them being significant, but decile

1a is the only (sub)decile with positive (but not significant) alpha. This indicates that some

funds were not affected by the crisis, probably because their investment strategies were

relatively immune to it.

29
But one must be cautious with this result, given that this is not the only reason possible. The Bond

crisis in 1994, the Asian crisis in 1997-1998, etc. may also explain these differences.
30
Numerical results are available upon request.

106
7.3 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. At the beginning of each year, all hedge

funds are put into decile rankings by their mean returns in the previous year. If a hedge

fund ceases reporting returns at any time before the end of the year, then this is counted

as dissolved.

Top 7 deciles have a more or less constant average rate of dissolution of 7%, it is

12.5% for the bottom 3 deciles. This suggests that bad performance may be a major

factor for dissolution, but that good performance is not a protection against it. This result

is consistent the finding of no persistence in the best performing funds, but that

persistence exists in the middle deciles.

7.4 One-Year Persistence for Hedge Fund Strategies

This subsection focuses on the persistence in returns for some hedge funds

strategies. 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. In

this sub-section, we determine whether persistence in returns exists for these strategies.

We test it for the 1994-2000 period and for the Asian crisis period (1996-97 and 1997-

6/98). Results for the pre-1994 period are also indicated for comparison. We classify funds

in 10 decile portfolios, with the top and bottom decile divided in 3 for the Market Neutral

strategy. Table 16 reports a summary of our results for these strategies.

107
Figure 4: Hedge funds dissolution frequencies

0,16
Dissolution Frequency in Y

0,12

0,08

0,04

0,00
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. At the

beginning of year t, all funds are placed into decile rankings on the basis of their returns in

year t-1. If a hedge fund ceases to report returns at any time before the end of year t, it

is counted as dissolved.

Panel A reports the results of the analysis for Macro funds. They show that there is

no significant difference between good and bad performing funds. Half of the alphas

obtained are significantly positive for the whole period, but not for the extreme deciles.

These results are stronger than those obtained for the whole database. This confirms the

previous findings for these funds. The 1996-97 Asian crisis sub-periods show that some

funds (decile D3) have persistence in their return despite the Asian crisis. This could either

mean that these funds returns were not affected by the crisis, or that they benefit from it.

The second hypothesis is more plausible given that we are analyzing Macro funds that per

definition anticipate market movements.

108
 Panel B shows the results for the Market Neutral funds. 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). The only

significantly positive alphas are those from the middle deciles. Those of the top decile are

only significant for the 1985-93 sub-period. The 1-10 and 1a-10c excess returns spreads

are both significant in most of the period considered, but not the alphas, except from the

97-6/98 period. This indicates that significantly positive difference existed between best

and worst performing funds during the Asian crisis period.

As an illustration, Panel C reports Market Neutral’s spread decomposition for each

of the sub-period considered. 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, four factors have a significant effect on spreads for the Market Neutral

funds, though they were insignificant in the all funds analysis: HML, Default and

Commodity (positive) and World Government Bond (negative). In the previous period,

their effect was also significant but with the opposite sign, with HML being replaced by

IHML. This indicates that the results obtained for our whole hedge funds database may not

be valid for one specific strategy of funds, but also that there has been a sharp reversal in

strategies for this category of funds after 1994.

109
Table 16: Hedge Funds Strategy Persistence based on 12 Month lagged Returns

Panel A: Global Macro

Whole Period Asian crisis Pre-1994 Period

1994-2000 1997-6/1998 1996-1997 1987-1993

Mean Mean Mean

Mean excess
excess Std. Dev. Alpha Std. Dev. Alpha excess Std. Dev. Alpha excess Std. Dev. Alpha
Return
Return Return Return

D1 1,55% 6,07% 0,71% 1,92% 3,15% 0,17% 1,98% 3,92% -0,84% 1,64% 4,50% 0,91% **

D2 1,55% 4,58% 0,63% * 2,01% 3,13% 0,12% 1,59% 2,98% 0,62% 1,49% 5,03% 1,03%

D3 1,42% 3,97% 0,34% 1,37% 2,73% 0,46% 1,43% 3,12% 1,36% *** 1,14% 3,91% 0,57%

D4 1,33% 3,38% 0,86% *** 1,04% 3,14% -0,31% 1,39% 3,10% 0,51% 1,44% 3,23% 0,82% **

D5 0,99% 3,16% 0,06% 1,38% 3,07% -0,24% 1,38% 3,19% 0,05% 1,32% 5,14% 0,79%

D6 1,36% 2,67% 0,58% *** 1,59% 2,88% -0,13% 1,66% 2,33% 0,75% * 0,39% 6,14% 0,13%

D7 0,88% 2,49% 0,40% ** 0,76% 2,89% -0,78% 0,84% 2,69% -0,09% 0,89% 3,81% 0,93% *

D8 1,09% 2,43% 0,50% ** 0,85% 2,66% -0,62% 0,83% 2,36% -0,49% 1,03% 3,98% 0,77% *

D9 0,84% 2,43% 0,33% 0,93% 2,24% 0,21% 1,10% 2,02% -0,28% 1,43% 3,55% 0,57%

D 10 1,01% 3,24% 0,22% 0,90% 2,08% -0,61% 1,34% 2,06% -0,49% 1,20% 4,98% 0,37%

1-10 spread 0,54% 6,46% 0,49% 1,03% * 2,29% 0,78% 0,64% 2,95% -0,35% 0,44% 6,48% 0,54%
Panel B: Market Neutral

Whole Period Pre-1994 Period

1994-2000 1987-1993

Mean excess Mean excess

Std. Dev. Alpha Std. Dev. Alpha
Return Return

D 1a 1,60% 4,10% 0,59% * 1,94% 4,37% 0,16%

D 1b 0,79% 3,43% 0,24% 1,56% 3,06% 1,26% ***

D 1c 1,17% 3,10% 0,59% 0,56% 3,86% 1,19%

D1 1,15% 3,22% 0,45% 2,30% 7,27% 0,91%

D2 0,65% 2,36% 0,04% 1,07% 2,22% 1,43% **

D3 0,73% 1,95% 0,20% 0,83% 2,05% 1,04% **

D4 0,59% 1,45% 0,19% 0,70% 1,98% 0,15%

D5 0,52% 1,04% 0,25% ** 0,86% 2,29% 0,27% *

D6 0,54% 0,99% 0,31% *** 1,21% 2,87% 0,65% ***

D7 0,37% 0,94% 0,18% 0,42% 3,03% 0,23%

D8 0,34% 1,04% 0,13% 0,59% 2,06% 0,07%

D9 0,56% 1,25% 0,22% 0,59% 1,68% 0,31%

D 10 0,60% 1,97% 0,24% 0,66% 3,45% -0,04%

D 10a 0,55% 1,78% 0,15% 1,12% 5,30% -0,98%

D 10b 0,90% 2,70% 0,42% 0,04% 2,29% 0,00%

D 10c 0,42% 3,10% 0,16% 1,11% 5,08% 0,63%

1-10 spread 0,55% 2,91% 0,21% 1,63% 7,75% 0,95%

1a-10c spread 1,18% 4,33% 0,43% 0,83% 6,49% -0,47%

111
Panel B (continued): Market Neutral

Asian crisis

1997-6/1998 1996-1997
Mean
Mean excess
Std. Dev. Alpha excess Std. Dev. Alpha
Return
Return
D 1a 1,60% 3,28% 0,34% 2,16% 3,38% -0,77%

D 1b 1,45% 2,05% 0,45% 1,06% 1,90% 0,93% ***

D 1c 1,95% 1,81% 1,05% *** 1,82% 1,90% 0,20%

D1 1,61% 1,98% 0,47% * 1,64% 1,95% 0,10%

D2 1,15% 1,69% 0,06% 1,23% 1,60% -0,09%

D3 0,98% 1,56% 0,02% 1,07% 1,23% 0,25%

D4 0,65% 1,16% -0,37% ** 0,94% 0,80% 0,20%

D5 0,67% 0,92% 0,01% 0,77% 0,71% 0,47% *

D6 0,61% 0,99% -0,09% 0,82% 0,81% 0,18%

D7 0,48% 0,88% -0,06% 0,65% 0,81% 0,21%

D8 0,19% 0,64% -0,15% 0,38% 0,52% -0,06%

D9 0,26% 0,86% -0,48% 0,52% 0,70% -0,15%

D 10 0,22% 1,25% -0,67% *** 0,43% 0,98% -0,28%

D 10a 0,26% 0,70% 0,02% 0,31% 1,15% -0,48%

D 10b 0,34% 1,70% -0,54% 0,72% 1,87% -0,67%

D 10c 0,31% 3,15% -1,26% ** 0,44% 2,30% 0,38%

1-10 spread 1,39% *** 1,56% 1,14% *** 1,21% 1,79% 0,38%

1a-10c spread 1,29% * 3,13% 1,60% * 1,72% 3,82% -1,16%

112
Panel C: Market Neutral spread comparison

Monthly
Std. Dev. Alpha Mkt W x US SMB HML I HML PR1YR
Exc. Return

1994-2000 1-10 spr. 0,55% * 2,91% 0,21% 0,17 * 0,09 0,16 *** 0,03 -0,3 *** 0,03

1a-10c spr. 1,18% ** 4,33% 0,43% 0,3 * 0,17 0,17 ** 0,1 -0,27 ** 0,11

1997-6/1998 1-10 spr. 1,39% *** 1,56% 1,14% *** -0,05 0,08 0,43 *** 0,02 -0,38 ** -0,28 ***

1a-10c spr. 1,29% * 3,13% 1,60% * -0,04 0,07 0,46 -0,13 -1,41 *** -0,71 ***

1996-1997 1-10 spr. 1,21% *** 1,79% 0,38% 0,33 ** 0,03 0,39 *** 0,26 0,16 -0,04

1a-10c spr. 1,72% ** 3,82% -1,16% 0,93 *** 0,06 0,52 * 0,84 ** 0,66 0,03

1985-1993 1-10 spr. 1,63% ** 7,75% 0,95% -0,1 -0,15 1,03 ** 0,85 ** -0,31 0,41

1a-10c spr. 0,83% 6,49% -0,47% 0,6 -0,04 1,09 1,31 *** -0,36 0,5
Panel C (continued): Market Neutral spread comparison

Monthly Exc.
Std. Dev. Alpha US Bd W GV Bd Emer Bd Default Comm. R²adj
Return

1994-2000 1-10 spr. 0,55% * 2,91% 0,21% -0,15 -0,24 0,01 0,54 ** 0,05 0,43

1a-10c spr. 1,18% ** 4,33% 0,43% -0,37 -0,54 ** -0,09 0,72 ** 0,16 ** 0,35

1997-6/1998 1-10 spr. 1,39% *** 1,56% 1,14% *** -0,07 -0,74 *** -0,19 1,45 *** -0,23 *** 0,62

1a-10c spr. 1,29% * 3,13% 1,60% * 1,41 -1,21 *** -0,05 1,38 * -0,67 *** 0,47

1996-1997 1-10 spr. 1,21% *** 1,79% 0,38% -0,42 -0,63 ** -0,08 0,67 * -0,04 0,68

1a-10c spr. 1,72% ** 3,82% -1,16% -1,34 * -0,99 * -0,14 0,94 0,21 * 0,4

1985-1993 1-10 spr. 1,63% ** 7,75% 0,95% -0,85 ** 1,44 * -0,39 -2,36 ** -0,12 0,34

1a-10c spr. 0,83% 6,49% -0,47% -0,85 2,48 ** -0,12 -3,39 ** -0,7 ** 0,26

This Table reports the result of the estimation of our combined model. Panel A contains the results for the Global Macro funds for different time period. Panel B
contains the results for the Market Neutral funds, and Panel C contains a comparison of Market Neutral spread. Each year, all funds are ranked based on their
previous year's mean return. Portfolios are equally weighted and weights are readjusted whenever a fund disappears. 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.

114
VIII Conclusion

Using one of the greatest hedge fund database ever used (2796 individual funds

including 801 dissolved), we investigated hedge funds performance using various asset-

pricing models including an extension form of Carhart’s (1997) model, combined with

Fama and French (1998), Agarwal and Naik (2004) models and an additional factor that

takes into account the fact that hedge funds may invest in Emerging Market Bonds.

Although our sample displays funds returns going back to 1984, 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. This corresponds to the lack of data about funds that were

dissolved prior to 1994 in the databases. Therefore, our conclusions can only be derived

for the 1994-2000 period, with a possible reinforcement with results for the 1984-1993

period whenever similar.

We demonstrate that our combined model is able to explain a significant

proportion of the variation in hedge fund returns over time. Compared to models used in

other hedge fund performance studies, our combined model does a better job describing

hedge funds behaviour. It appears particularly good for the Equity non-hedge, US

Opportunistic, Sector, Global Macro, Long only Leveraged and Event Driven funds.

115
 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, and

that many hedge funds invest in Emerging Market Bonds. Analyzing each strategy

individually lead to the additional conclusions that 10 out of 13 strategies offer

significantly positive excess return. A sub-period analysis showed that over-performance

ability is, in most cases, constant over time. But, according to our results, most funds

suffered from the Asian crisis.

The analysis of persistence in performance for our whole hedge funds database

indicates 3 main findings that complete our results of the performance analysis. Firstly,

best performing funds follow momentum strategies whereas worst performing ones tend

to be momentum contrarian. Secondly, best performing funds do not significantly invest

in Emerging Market Bond, whereas lower decile funds do. Thirdly, average return funds

prefer high book-to-market stocks, whereas best and worst performing funds may prefer

low book-to-market ones. Our main conclusion from this evidence is that there is no

persistence in performance for best and worst performing funds, but limited evidence of

persistence exists for middle decile funds. This suggests that, even if some hedge funds

managers take a lot of risk, which lead them to have very high or low returns for short

period of time, many hedge fund managers follow less risky strategies that allow them to

out-perform the market for longer period of time. The much stronger results obtained for

data before 1994 raise additional suspicion about the presence of severe bias in this part

of the database. 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.

Moreover, they significantly under-performed the worst performing funds (of 1996)

during the crisis period. The analysis of dissolution frequencies indicates, on the one

116
hand, that bad performance may be a major factor for dissolution, and on the other, that

very good performance is not a protection against dissolution.

Finally, 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. 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. This stresses that funds in different strategy are not influenced by the same

factors, and that specific models could be developed for each investment style in the

future.

All these results help in understanding the performance of hedge funds and the

persistence in this performance over different time period, including a study of the Asian

crisis. The next step in the development of model that describes hedge funds behaviour,

including Trading Strategy (option writing/buying) factors, in order to account for the fact

that some hedge funds exhibit non-linear option-like exposures to standard asset-

classes, as found by Fung and Hsieh (1997, 2001).

117
Study 2: Hedge Fund Performance and

Persistence in Bull and Bear Markets

Daniel P.J. CAPOCCI

HEC-ULG Management School – University of Liège (Belgium)

Albert CORHAY

HEC-University of Liège, Limburg Institute of Financial Economics,

Maastricht University, and Luxembourg School of Finance, University of

Luxembourg.

Georges HÜBNER

HEC-University of Liège, Limburg Institute of Financial Economics,

Maastricht University, and Luxembourg School of Finance, University of

Luxembourg.

Capocci, Daniel, Georges Hübner and Albert Corhay, 2005, Hedge Funds Performance

and Persistence in Bull and Bear Markets, European Journal of Finance 11/5, 361-392
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. Our database proves to be fairly reliable with respect to the most important

biases in hedge funds studies, despite the high attrition rate of funds observed in the

down market. We apply an original ten-factor composite performance model that

achieves very high significance levels. The analysis of performance indicates that most

hedge funds significantly out-performed the market during the whole test period, mostly

thanks to the bullish sub-period. In contrast, no significant under-performance of

individual hedge funds strategies is observed when markets headed south. The analysis

of persistence yields very similar results, with most of the predictability being found

among middle performers during the bullish period. However, the Market Neutral

strategy represents a remarkable exception, as abnormal performance is sustained

throughout and significant persistence can be found between the 20% and 69% best

performers in this category, probably due to an extreme adaptability and a very active

investment behaviour.

120
Hedge Fund Performance and Persistence in Bull and

Bear Markets

I Introduction

Since 1990, when around 2000 hedge funds were managing together assets of ca.

$ 60 billion, the subsequent growth of number and asset base of hedge funds has never

really been refuted. The industry only suffered from a relative slowdown in 1998, 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, 2002), corresponding to

growth rates of respectively 10.8% and 21.3% respectively.

The growing trend of the sector could be remarkably sustained during the stock

market collapse that started in March 2000, when the NASDAQ Composite Index reached

an all-time high of 5132, and finished three years later with a floor level at 1253. In the

meantime, the global net asset value (NAV) of hedge funds continued to grow at a steady

10.6% (Van Hedge Funds Advisors International, 2002), contrasting with a decrease of

2.7% in worldwide mutual fund industry (Investment Company Institute, 2003).

This relatively positive attitude of investors is typically motivated by the

perceptions that hedge funds are largely market neutral, that their managers enjoy

greater flexibility in their asset allocation that enables them to achieve a better market

timing (Fung and Hsieh, 1997), or that hedge funds have a relatively low covariance with

other classes of financial assets, making them a good diversification vehicle (Schneeweis

and Spurgin, 1998; Capocci and Hübner, 2004; Kat and Amin, 2003b).

121
 The vast majority of performance studies on hedge funds has not focused on their

behaviour under different market conditions. This is generally due to the particularly

bullish period corresponding to the time window under review, as most empirical evidence

reveals that data collected prior to 1994 by several data vendors displays a significant

survivorship bias, as shown by Fung and Hsieh (2000), Liang (2000) and Capocci and

Hübner (2004).

In this context, Ackermann and al. (1999) and Liang (1999) find that hedge funds

constantly obtain better performance than mutual funds, although lower and more volatile

than the reference market indices considered.

On the other hand, 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.,1999, 2001; Liang, 1999). Agarwal and Naik (2000) find

evidence of persistence in hedge funds performance, while Capocci and Hübner (2004)

sustain that it can be mostly found among average performers and Brown and al. (1999)

conclude that there is hardly any evidence of the existence of differential manager skills

but persistence is rather due to style effects.

Few authors have attempted to estimate the behaviour of hedge funds in bear

markets. The periods under review do not favour this exercise, as periods of downward

trends on the stock market were rare and discontinuous. For the period 1990-1998,

Edwards and Caglayan (2001) find that only three hedge fund strategies (Market Neutral,

Event Driven and Macro) provide protection to investors when stock markets head south.

More recently, Ennis and Sebastian (2003) contend that in general, hedge funds did not

provide investor protection after the market downturn of March 2000; rather, their

superior performance is mostly due to the good market timing of their managers during

the US stock market bubble that preceded it.

122
 This study benefits from the fact that stock markets have experienced a long period

of depression, since stock indices have been almost continuously going down for a period

of three years. Thus, our analysis neither suffers from discontinuities between down

periods, that preclude any analysis of persistence, nor from arbitrarily chosen definitions

of a bear market, as Fabozzi and Francis (1979), Kao et al. (1998), Rao (2001), Edwards

and Caglayan (2001) and Liang (2003).

In this study, using a similar methodology as the one developed by Capocci and

Hübner (2004), we study the performance of hedge funds and its persistence during a

time window that encompasses relatively long bullish and bearish periods.

This study introduces several key modifications with respect to the previous studies

on hedge funds performance and persistence. Firstly, although Capocci and Hübner (2004)

and Liang (2001) consider a relatively short bearish sub-period with the Asian crisis, their

sample does not enable them to distinguish between unambiguously bullish and bearish

sub-periods. Our study will thus identify and separately analyse two sub-periods

corresponding to upward and downward market trends, with a pivot set at the end of

March 2000.

Secondly, we introduce a modified asset pricing model encompassing the risk

premia that proved to be relevant for assessing funds performance in previous studies,

successively proposed by Fama and French (1993), Carhart (1997), Agarwal and Naik

(2004) and Capocci and Hübner (2004). This yields a model with 10 risk premia, which

may look at first sight overspecified but one has to bear in mind that hedge funds families

are very heterogeneous and, unlike mutual funds, involve investments in many types of

assets and markets.

123
 Thirdly, we specifically identify one hedge fund strategy, namely Market Neutral

that, following the results of Edwards and Caglayan (2001), supposedly hedge investors

against bearish markets. It has also been studied in an unambiguously bullish setup by

Capocci and Hübner (2004), who find that this family of funds tends to out-perform the

market during the 1994-2000 period. In this study, we perform an in-depth analysis of the

level and persistence of its performance before and after the stock markets downturn.

The study is organised as follows. Section 2 sets out the performance models we

will use. In Section 3, we provide a thorough analysis of the database. The fourth Section

studies potential biases in the database. The next Section reports the performance of

hedge funds for the whole period and the sub-periods considered. Section 6 documents

and explains the persistence in hedge fund returns over the same time windows, with a

special focus on the Market Neutral strategy. Section 7 concludes the study.

124
II Performance Measurement Models

The starting point of our study of hedge funds performance is the original Sharpe

(1964) – Lintner (1965) CAPM. As the basic multi-factor specification, 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. Finally, 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.

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). Its equation to estimate is the

following:

RPt − RFt = α P + β P (RMt − RFt ) + ε Pt t = 1,2,..., T (6)

where RPt = return of fund P in month t; RFt = risk-free return on month t; RMt =

return of the market portfolio on month t; εPt = error term; αP and βP are the intercept and

the slope of the regression, respectively.

125
The intercept of this equation, αp commonly called Jensen’s alpha (1968) is usually

interpreted as a measure of out- or under-performance relative to the market proxy used.

2.2 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. It takes into account size and book-to-market ratio, but also an

additional factor for the momentum effect. Grinblatt, 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 =1,2,...,T (7)

where SMBt = the factor-mimicking portfolio for size (‘small minus big’), HMLt = the

factor-mimicking portfolio for book-to-market equity (‘high minus low’)31 and PR1YRt = the

factor-mimicking portfolio for the momentum effect.32 These factors aim at isolating the

firm-specific components of returns.

31
See Fama and French (1993) for a precise description of the construction of SMBt and HMLt.
32
For a description of the construction of PR1YR see Carhart (1997).

126
2.3 The Composite Model

In order to take into account the complex characteristics of the hedge fund

industry, we implement a combination and an extension of Carhart’s (1997) 4-factor

model, the model used by Agarwal and Naik (2004) and the one used by Capocci and

Hübner (2004).

This model contains the market risk premium, Fama and French (1993) "size" and

"value" factors, Carhart’s (1997) "momentum" factor, five factors introduced by Agarwal

and Naik (2004): a factor for non-US equities investing funds (MSCI World excluding US),

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, the JP Morgan

Emerging Market Bond Index, and finally a commodity factor (GSCI Commodity Index).

Furthermore, we add an additional bond index factor that is not used in previous studies,

namely the Lehman Mortgage-Backed Securities Index to take into account the fact that

various hedge funds strategies (fixed income arbitrage, mortgage-backed securities) are

exposed to this market and the Lehman High-Yield Credit Bond Index.

The market proxy used is the value-weighted portfolio of all NYSE, Amex and

NASDAQ stocks market proxy that is usually used in mutual funds performance studies.

Several additional factors, such as the MSCI Emerging Markets Index, 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, that was used in several previous hedge funds studies, was

found to have an extremely high correlation with the Lehman BBA Corporate Bond Index and thus was

removed from our study.

127
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 )
(8)
+ 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

where RMt = return on the Russel 3000 Index; MSWXUSt = return of the MSCI World Index

excluding US; SWGBIt = return of the Salomon World Government Bond Index; JPMEMBIt

= return of the JP Morgan Emerging Market Bond Index; HYt = return of the Lehman High

Yield Credit Bond Index et MORTt = return of the Lehman Mortgage-Backed Securities

Index, and GSCIt = return of the Goldman Sachs Commodity Index.

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.

128
III Data

3.1 Database

Three main hedge fund databases are available for empirical studies: ‘Managed

Account Reports’ (MAR), ‘Hedge Fund Research, Inc.’ (HFR), and ‘TASS Management’

(TASS) (Amin and Kat, 2003a). These databases are the most used in academic and

commercial hedge fund studies.35

Data vendors do not only collect performance data. For a majority of funds, they

record other useful information such as company name, start and ending date, strategy

followed, assets under management, management and incentive fees, manager's name

etc. There is no consensus on the definition of the strategy followed but there are

similarities. MAR defines 9 strategies with a total of 16 sub-strategies. HFR defines sixteen

different strategies in two categories, 12 non-directional and 5 directional strategies, plus

the Funds of Funds and the Sector categories. Finally, TASS defines 15 strategies.

We use hedge fund data from MAR, as in Fung and Hsieh (1997), Schneeweis and

Spurgin (1998), and Amin and Kat (2003a). The database gives monthly net-of-fee

individual returns and other information on individual funds and groups them in indices.

We use 108 monthly returns on 2894 individual hedge funds plus 48 indices (16

investment styles with 3 indices for each investment style: onshore, offshore and a

combined index). These funds include 1622 funds alive at the end of the period (56%) and

1272 dissolved funds (44%).

35
The three databases have never been used together in a study, but Ackermann and Ravenscraft

(1998) and Ackermann et al. (1999) and Capocci and Hübner (2004) used a combination of HFR and

MAR while Liang (2000) uses a combination of TASS and HFR.

129
 Hedge funds are classified in two categories. The Individual Funds category

features 13 strategies: Event driven – Risk Arbitrage, Event-Driven – Distressed

Securities, Global, Global Est., Global Intern., Global Emerging, US Opp.36, Macro, Mkt

Neutral, Long Only Leveraged, Sector, Short Sales and No Category, the latter one

corresponding to funds with no stated strategy and funds whose strategy does not fill in

any of the above. The Funds of Funds category features 3 strategies: Niche, Diversified

and Others.

We take the value-weighted portfolio of all NYSE, Amex and NASDAQ stocks market

proxy that is usually used in mutual funds performance studies (see e.g. Fama and

French, 1993, 1996; Carhart, 1997). Its almost perfectly correlation with the Russell 3000

index used in Agarwal and Naik (2004) suggests that both market proxies are very

similar.37 Finally, the one-month T-bill rate from Ibbotson Associates is taken as the risk-

free rate.

3.2 Basic Performance

Panel A of Table 17 contains descriptive statistics of the funds, whether living or

dead, in our database. These hedge funds data are contrasted against the descriptive

statistics of the factors introduced in equation (3) of Section 2. These statistics are

reported in panel B of Table 16.

36
This strategy has been suppressed in 1999.
37
See Capocci and Hübner (2004) for more a complete analysis of this correlation.

130
 Panel A shows that the highest mean return was achieved by the Sector (1.66%),

then the Global Est. (1.29%) and Global Emerging (1.17%) follow. Average returns of

funds of funds are all around 0.70%, only followed by the Global (0.45%) strategy that

achieves the lowest mean return. This pattern is similar for the mean excess returns.

These descriptive statistics differ from the results obtained by Capocci and Hübner

(2004) for the 1994-2000 period, who find that the best performers are US Opportunistic

Small Caps, US Opportunistic Growth and Sector while the worst average performers are

Foreign Exchange, Short Sellers the and Funds of Funds, without sub-strategy. This

difference can be explained by the difference in the database used (MAR combined with

HFR for Capocci and Hübner, 2004) and the different time period studied.

The Sharpe measure (the ratio of excess return and standard deviation) offers a

much different picture: accounting for risk, Market Neutral funds appear to be the best

performers, while the funds that achieve the highest absolute returns are only among the

average risk-adjusted performers.

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.

Panel B of Table 16 shows that the mean excess return of the Market Proxy is

0.78% per month (about 9.5% per year), only statistically different from zero at the 10%

level. This reasonable value indicates that the bullish sub-period has been almost totally

offset by the market correction. The mean excess premium of the MSCI World excluding

US is an insignificant 0.22% per month. The average SMB and HML returns are

insignificant; only the Momentum factor, with the highest mean, provides a significantly

positive value. The highest mean return for bond indices was obtained by the JP Morgan

Emerging Market Bond Index.

131
 The Sharpe ratio obtained by our whole hedge fund database (0.29) is higher than

the ones for the Market Proxy (0.19), and higher than for the MSCI World Excluding US (-

0.15).

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. This month corresponds to the maximum observed value of the Russell

3000 Index that reached a value of 858.48 during the session of March 24, 2000. During

the up period, the monthly index return was positive in 70% of the months (52 out of 74)

with an average yearly return of 19.4%. During the down period, the monthly index return

was positive in 39% of the months (12 out of 34) and the average yearly return was -

16.9%. Those trends are sufficiently strong to allow us the consider the whole sub-periods

as, respectively, bullish and bearish without having to use a complex rule to separate

bullish, bearish and neutral months since these rules would obviously not match the ones

used by funds managers for their market timing decisions.

The analysis of basic performance for the two sub-periods under review, presented

in Table 17, reveals some interesting differences.

132
Table 17: Descriptive Statistics of Hedge Funds Strategies and
Passive Investment Strategies

Panel A: Hedge Funds strategies Jan.1994-December 2002 (108 months)

Nb of % of the Living Dead Mean
t(mean) Std. Dev. Med.
Fds category Funds Funds Return
Event driven - Risk
136 6,10% 85 51 0,93% 5,15 1,74% 0,92%
Arb
Event driven -
106 4,70% 70 42 0,99% 5,48 2,25% 1,17%
Distressed Sec

Global* 175 7,80% 1 174 0,45% 2,47 3,80% 0,64%

Global Est. 499 22,20% 300 199 1,29% 7,15 3,25% 1,09%

Global Intern. 72 3,20% 46 26 0,88% 4,87 2,42% 0,93%

Global Emerging 157 7,00% 97 60 1,17% 6,48 5,02% 1,80%

US Opp.** 39 1,70% 0 39 0,23% 1,3 2,40% 0,21%

Macro 144 6,40% 52 92 0,82% 4,52 2,15% 0,61%

Market Neutral 635 28,30% 385 250 1,04% 5,73 0,97% 1,05%

Long Only Lev. 33 1,50% 16 17 0,92% 5,06 5,83% 1,50%

Sector 190 8,50% 111 79 1,66% 9,17 4,43% 2,07%

Short Sales 37 1,60% 24 13 0,88% 4,87 4,48% 0,67%

No category 24 1,10% 6 18 0,93% 5,12 3,45% 0,58%

Individual Funds
2247 100% 1186 1061 1,08% 5,98 2,28% 1,11%
Total

FoF - Niche 114 18% 86 28 0,74% 4,1 1,33% 0,67%

FoF - Diversified 501 77% 349 152 0,71% 3,95 1,87% 0,75%

FoF - Other 32 5% 1 31 0,77% 4,27 1,69% 0,80%

Funds of Funds
647 100% 436 211 0,72% 3,96 1,77% 0,71%
Total
Total 2894 100% 1622 1272 0,99% 5,49 2,14% 1,11%

133
Panel A (continued): Hedge Funds strategies Jan.1994-December 2002 (108 months)

Excess t(mean Sharpe

Med. Min Max Skew. Kurt
return exc.) ratio

Event driven - Risk

0,92% -6,40% 5,50% -0,58 2,24 0,55% 3,05 0,32
Arb
Event driven -
1,17% -10,80% 7,00% -1,16 6,25 0,61% 3,38 0,27
Distressed Sec

Global* 0,64% -25,80% 13,60% -2,74 21,61 0,07% 0,37 0,02

Global Est. 1,09% -9,90% 12,30% 0,11 1,59 0,91% 5,04 0,28

Global Intern. 0,93% -6,80% 8,90% 0,24 1,44 0,50% 2,77 0,21

Global Emerging 1,80% -21,70% 14,30% -0,66 3,18 0,79% 4,37 0,16

US Opp.** 0,21% -5,60% 7,40% 0,09 0,7 -0,18% -1 -0,08

Macro 0,61% -4,10% 7,00% 0,45 0,57 0,44% 2,42 0,2

Market Neutral 1,05% -2,50% 4,00% -0,22 1,05 0,66% 3,62 0,67

Long Only Lev. 1,50% -17,40% 13,30% -0,44 0,16 0,54% 2,96 0,09

Sector 2,07% -13,10% 19,90% 0,31 2,74 1,28% 7,06 0,29

Short Sales 0,67% -13,60% 13,20% 0,09 0,74 0,50% 2,76 0,11

No category 0,58% -7,80% 12,70% 1,17 3,2 0,55% 3,01 0,16

Individual Funds
1,11% -8,60% 8,00% -0,26 2,77 0,70% 3,87 0,31
Total

FoF - Niche 0,67% -4,10% 5,10% 0,22 1,73 0,36% 1,99 0,27

FoF - Diversified 0,75% -7,30% 7,00% -0,12 3,54 0,26% 1,41 0,14

FoF - Other 0,80% -7,30% 5,80% -0,89 4,86 0,39% 2,16 0,23

Funds of Funds
0,71% -6,80% 6,60% -0,11 3,34 0,34% 1,85 0,19
Total

Total 1,11% -8,20% 7,60% -0,2368 3,04 0,61% 3,38 0,29

The Global category has been gradually suppressed and replaced by the Global Est., Global Intern. And
Global Emerging categories. Therefore, the funds disappearance is mostly due to category transfers. ** US
Opportunistics ended in 1999.

134
Panel B: Passive Strategies

Mean Return t(mean) Std. Dev. Median

Equity Factors

Market Proxy 0,78 1,7 4,77 1,58

MSCI World Excluding US 0,09 0,22 4,36 0,47

F&F SMB Factor 0,02 0,04 4,45 -0,36

F&F HML Factor 0,6 1,5 4,16 0,68

Momentum Factor 1,14 2,06 5,74 1,27

Bond Factors

1 month T-bill 0,37 34,69 0,11 0,4

Salomon WBGI 0,5 2,83 1,83 0,24

JPM EMBI Global 0,82 1,81 4,67 1,16

Lehman Mortgage 0,59 6,94 0,89 0,66

Lehman High Yield Credit 0,4 1,99 2,11 0,66

Commodities

Goldman Sachs Commodity 0,59 1,14 5,35 0,61

135
Panel B (continued): Passive Strategies

Mean exc.
Min Max t(mean exc.) Sharpe ratio
Return

Equity Factors

Market Proxy -15,7 8,3 0,41 0,9 0,19

MSCI World Excluding US -12,9 10,3 -0,28 -0,67 -0,15

F&F SMB Factor -16,3 21,4 -0,35 -0,82 -0,18

F&F HML Factor -8,9 13,7 0,23 0,58 0,14

Momentum Factor -25,1 18,2 0,77 1,39 0,24

Bond Factors

1 month T-bill 0,1 0,6 NA NA NA

Salomon WBGI -3,4 5,9 0,13 0,72 0,39

JPM EMBI Global -24,2 10,9 0,44 0,99 0,21

Lehman Mortgage -2,6 3,2 0,22 2,66 2,99

Lehman High Yield Credit -7,37 7,49 0,03 0,17 0,08

Commodities

Goldman Sachs Commodity -12,28 15,79 0,22 1,07 0,2

This Table shows the mean returns, t-stat for mean = 0, standard deviation, medians,
minimum, maximum, mean excess returns, t-stat for mean excess return = 0, and Sharpe
ratios for the individual hedge funds in our MAR database for the whole period 01:1994-
12:2002. US Opp. Funds ended in 01:1999. Sharpe ratio is the ratio of excess return and
standard deviation with a risk-free rate set at 5%. In panel A, No of Fds represent the number
of funds following a particular strategy (or sub-strategy), Living Funds and Dead Funds
represents the number of surviving and dead funds (in December 2002). We calculate the Mean
Excess Return considering Ibbotson Associates one-month T-bills. Numbers in the table are
monthly percentage.

136
 Table 19 displays summary results for the bullish and bearish sub-periods. As

expected from the nature of the time windows, excess returns obtained for the majority of

hedge funds strategies are mostly due to the bullish sub-period, with the best performers

before March 2000 also displaying the worst returns after the market reversal too place.

There are three noticeable exceptions. Firstly, the US Opportunistics strategy did poorly in

spite of favourable market conditions, which explains the disappearance of this category.

Secondly, the Global strategy seems to achieve returns that are much less dependent on

the conjuncture than the other strategies. Finally, 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.

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, while

the returns of the Market Neutral strategy are then significant but not when excess returns

are considered. Although further evidence is obviously needed, this does not support the

findings of Edwards and Caglayan (2001) with a different definition of bearish market

conditions.

137
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 Living Funds Dead Funds

Event driven - Risk Arb 113 89 24

Event driven - Dist. Sec 83 65 18

Global 175 6 169

Global Est. 400 342 58

Global Intern. 64 58 6

Global Emerging 132 108 24

US Opp. 39 0 39

Macro 121 63 58

Market Neutral 487 350 137

Long Only Lev. 31 22 9

Sector 147 126 21

Short Sales 34 25 9

No category 19 9 10

Individual Funds Total 1845 1263 582

FoF - Niche 81 57 24

FoF - Diversified 408 333 75

FoF - Other 31 18 13

Funds of Funds Total 520 408 112

Hedge funds Total 2365 1671 694

138
Panel A (continued): Sub-period 01:1994-03:2000

Mean Return Excess return Sharpe ratio

Event driven - Risk Arb 1,23%*** 0,82%*** 0,46

Event driven - Dist. Sec 1,27%*** 0,85%*** 0,36

Global 0,57%** 0,16% 0,04

Global Est. 1,96%*** 1,54%*** 0,48

Global Intern. 1,30%*** 0,89%*** 0,35

Global Emerging 1,57%*** 1,16%*** 0,21

US Opp. 0,23% -0,18% -0,08

Macro 1,10%*** 0,68%*** 0,3

Market Neutral 1,18%*** 0,77%*** 0,75

Long Only Lev. 1,83%*** 1,42%*** 0,26

Sector 2,56%*** 2,14%*** 0,5

Short Sales 0,39%* -0,02% 0

No category 1,21%*** 0,80%*** 0,2

Individual Funds Total 1,47%*** 1,06%*** 0,45

FoF - Niche 0,92%*** 0,51%** 0,35

FoF - Diversified 0,96%*** 0,26% 0,12

FoF - Other 0,92%*** 0,50%** 0,26

Funds of Funds Total 0,95%*** 0,54%** 0,28

Hedge funds Total 1,35%*** 0,93%*** 0,42

139
Panel B: Sub-period 04:2000-12:2002

Nr of Fds Living Funds Dead Funds

Event driven - Risk Arb 112 85 27

Event driven - Dist. Sec 94 70 24

Global 6 0 6

Global Est. 441 300 141

Global Intern. 66 46 20

Global Emerging 133 97 36

US Opp. NA NA NA

Macro 86 52 34

Market Neutral 498 385 113

Long Only Lev. 24 16 8

Sector 169 111 58

Short Sales 28 24 4

No category 14 6 8

Individual Funds Total 1665 1186 479

FoF - Niche 90 86 4

FoF - Diversified 416 349 77

FoF - Other 19 1 18

Funds of Funds Total 535 436 99

Hedge funds Total 2214 1622 578

140
Panel B (continued): Sub-period 04:2000-12:2002

Mean Return Excess return Sharpe ratio

Event driven - Risk Arb 0,0025 -0,0006 -0,04

Event driven - Dist. Sec 0,0037 0,06% 0,03

Global 0,0016 -0,0014 -0,05

Global Est. -0,0021 -0,52%** -0,19

Global Intern. -0,0008 -0,0039 -0,22

Global Emerging 0,27% -0,04% -0,01

US Opp. NA NA NA

Macro 0,0019 -0,0012 -0,07

Market Neutral 0,71%*** 0,004 0,52

Long Only Lev. -1,17%*** -1,48%*** -0,23

Sector -0,0038 -0,69%**° -0,16

Short Sales 1,99%*** 1,68%*** 0,33

No category 0,0028 -0,0003 -0,01

Individual Funds Total 0,0019 -0,0012 -0,06

FoF - Niche 0,0033 0,0003 0,03

FoF - Diversified 0,0015 -0,0016 -0,13

FoF - Other 0,44%* 0,0014 0,14

Funds of Funds Total 0,0018 -0,0013 -0,11

Hedge funds Total 0,0019 -0,0012 -0,07

This Table shows the number of funds following a particular strategy (or sub-strategy), the corresponding
number of living and dead funds, mean returns, mean excess returns, 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). Sharpe ratio is the ratio of excess return and standard deviation with a risk-free rate set at
5%.*** statistically significant at the 1% level, ** statistically significant at the 5% level, * statistically
significant at the 10% level.

141
 For our factors, the same analysis shows that the Market and Momentum factors

gave the highest excess returns during the first sub-period, while the Book-to-Market and

Lehman Mortgage factors obtained significant positive abnormal returns during the second

one. 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.38

3.4 Correlations

The traditional hedge funds literature contends that, thanks to their weak correlation

between hedge funds and other securities, hedge funds are likely to improve to the risk-

return trade-off when added to a traditional portfolio (see Fung and Hsieh, 1997;

Schneeweis and Spurgin, 1998; Liang, 1999; Amin and Kat, 2003b).

This sub-section studies the ranges of correlation coefficients among and between

hedge funds and passive investment strategies. The correlations have been computed for

the whole 1994-2002 period and for two-sub-periods. In order to obtain periods with

comparable lengths, we took the bearish sub-period starting in April 2000 (33 months) and

matched it against the most bullish time window, that started in September 1998 (19

months) rather than the whole 1994-03:2000 period. Because of the extremely large

number of results to be reported, we chose to report ranges in correlations. Results are

reported in Table 19.

In each cell, correlations increase as the colour is darker. The upper part of the cell

accounts for the correlation during the whole period, while the lower part is split between

the pre- and post-March 2000 sub-periods.

38
Detailed data are available upon request.

142
 Panel A reports correlations among hedge funds strategies. As typically reported (see

Liang, 2003; Capocci and Hübner, 2004) in the literature, these strategies are in general

highly correlated when indices are considered, with the exception of the Short Sellers

strategy that systematically goes conversely – as expected. However, a closer look at their

evolution over time indicates that the Global and, to a lesser extent, the Global

International strategies tend to decrease their correlation with other funds in bearish times;

on the other hand, the Global Emerging strategy follows the other strategies more closely

during periods of down markets. 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.

In Panel B, the behaviour of our explanatory variables is also of considerable

interest. Only four indices (Market, World Excluding US, SMB and Lehman High Yield) have

a high correlation with most hedge funds strategies. 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. This strong tendency

is not invalidated for the supposedly investor-protecting strategies. In general, hedge funds

strategies sharply decrease and even reverse their loading with the momentum factor, as

they become momentum-contrarian during bad times. They also reduce their sensitivity

towards the Emerging Market Bond factor, while increasing their exposure to the Lehman

High Yield and the SMB factors. The Short Sellers strategy, for its part, noticeably switches

from a "Glamour" strategy (low loading with the HML factor) to a "Value" one in the last

sub-period. This is the only strategy that consistently invests in Value stocks in bearish

markets.

Panel C indicates that the correlation coefficients of our regressors are low enough to

raise serious multi-colinearity concerns.

143
Table 19: Correlation among hedge funds, between hedge
funds and passive investment strategies, 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.

GIN.

GEM

MAC

MKN

LOL

SEC

SHS

NOC

NIC

DIV

144
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
Panel C: Correlation among Passive Investment strategies

MKT WXU SMB HML MOM SWG EMB MOR HIY

MKT

WXU

SMB

HML

MOM

SWG

EMB

MOR

HIY

GSC

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

4.1 Survivorship bias

In order to reduce the severity of survivorship bias, an important concern for mutual

funds (see a.o. Carhart, 1997) as well as hedge funds studies (see a.o. Ackermann et al.,

2001 and Fung and Hsieh, 2000), data vendors backfill each fund’s performance history

prior to their addition to the database. Thus, they provide data that go back before the

starting date of the database itself, usually 1993. However, before this starting date, one is

left with only surviving funds data. Brown et al. (2001), for the TASS database, 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., 1997). Data starting in 1994 appears to be

more reliable according to this criterion (Capocci and Hübner, 2004).

Two definitions of this bias are commonly used in mutual and hedge fund studies:

the performance difference between surviving and dissolved funds (e.g. Ackermann et al.,

1999) and the performance difference between living and all funds (e.g. Liang, 2000).

We report the bias using both definitions for the whole period and for 2 sub-periods

1994-03:2000 and 04:2000-2002.

In Panel A of Table 20, we report the yearly returns of all funds, surviving funds and

dissolved funds. Hedge funds experience extremely high returns in 1999, when the stock

market experienced a sharp positive return, without great difference between surviving and

dissolved funds. In the subsequent years, returns gradually reduced but mostly due to the

negative returns of dissolved funds. This is a clear effect of the bearish turn of the market

after March 2000, leading to an increase in differences in hedge funds returns between the

best and the worst performing managers.

147
 In Panel B, our results yield a monthly survivorship bias of 0.41% (or 4.92% per

annum) for the whole period using the first formula while in Panel C the bias of 0.13% per

month (1.51% per annum) with the second formula. This latter value is much higher than

the very low value obtained by Ackermann et al. for the period 1988-1995. It is similar to

the percentage of 1.5% from Fung and Hsieh (1998), lower than the 0.30% monthly bias

found by Fung and Hsieh (2000) and slightly higher than the percentage of 1.2% found by

Capocci and Hübner (2004) for the 1994-2000 period. 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).39

A look at sub-period biases indicates that the level of this bias is mostly due to the

bearish period, where its level sets at 2.61%. The bias drifts up through the very high level

of returns differential between surviving and dissolved funds, 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 35.9% in 1999 to 9.1% in 2002). Thanks to this

bias-reduction effect of recent data, the global behaviour of the database in relationship to

survivorship bias is kept within reasonable bounds.

39
This consensus value quite high when compared to the 0.8-1.5 bias reported by Malkiel (1995) and

Brown and Goetzmann (1995) for US mutual funds.

148
Table 20: Survivorship bias in hedge funds

Panel A: Annual performance (all funds, surviving funds and dissolved funds)

All Funds Surviving Funds Dissolved Funds

Year Return S.D. Obs. Return S. D. Obs. Return S. D. Obs.

1994 1,8% 1,6% 8601 3,0% 1,4% 3419 1,0% 1,8% 5182

1995 18,6% 1,0% 10641 19,9% 1,0% 4630 17,7% 1,0% 6011

1996 21,3% 1,4% 13049 23,6% 1,3% 6200 19,3% 1,6% 6849

1997 20,4% 2,0% 15860 22,4% 1,9% 8136 18,3% 2,2% 7724

1998 3,6% 3,2% 17872 4,4% 3,0% 9954 2,6% 3,4% 7918

1999 33,5% 2,5% 18798 33,8% 2,2% 12052 33,3% 2,9% 6746

2000 9,6% 2,9% 20221 14,4% 2,3% 14395 -2,3% 4,3% 5826

2001 5,7% 1,6% 20591 8,0% 1,4% 16706 -3,2% 2,6% 3885

2002 0,4% 1,3% 20771 1,2% 1,2% 18899 -4,9% 2,0% 1872

Mean 01:94-03:00 17,0% 2,0% 14137 18,0% 1,8% 7399 15,0% 2,1% 6738

Mean 04:00-12:02 5,0% 1,9% 20528 8,0% 1,7% 16667 -3,0% 3,0% 3861

Mean 94-02 13,0% 1,9% 16267 15,0% 1,8% 10488 9,0% 2,4% 5779
 Panel B: Living - Dead
Panel C: Living - All Funds
Funds

Year Return Return

1994 0,02 0,01

1995 0,02 0,01

1996 0,04 0,02

1997 0,04 0,02

1998 0,02 0,01

1999 0,01 0,00

2000 0,17 0,05

2001 0,11 0,02

2002 0,06 0,01

Bias 01:94-03:00 0,16 0,09 per Month

1,92 1,03 per Year

Bias 04:00-12:02 0,98 0,22 per Month

11,75 2,61 per Year

Bias 94-02 0,41 0,13 per Month

4,93 1,51 per Year

This Table reports the survivorship bias of calculated from our database. Our MAR database contains
2894 hedge funds, including 1622 survived funds and 1272 dissolved funds as of December 2002. In
Panel B survivorship bias is calculated as the performance difference between surviving funds and
dissolved funds. In Panel C survivorship bias is calculated as the performance difference between
surviving funds and all funds. All returns are net of fees. Numbers in the table are yearly percentage
unless otherwise indicated.

150
4.2 Instant Return History Bias

As hedge funds are not allowed to advertise, 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, 2000) that occurs because a fund’s performance history is

backfilled after inclusion. 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.

We use the same two-step methodology as Park (1995), Brown et al. (1997), Fung

and Hsieh (2000) to estimate this bias for our hedge fund database. On the one hand, we

estimate the average monthly return of the “observable portfolio” which invests in all funds

from our database each month. On the other hand, we estimate the average monthly return

of the “adjusted observable portfolio” obtained from investing in all these funds after

deleting the first 12, 24 and, if possible, 36, 48 and 60 months of returns. 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). Results are reported in Table 21.

For the whole period, the observable monthly return averaged 0.99%, while the

adjusted observable one was 0.88% (when deleting the 12 first months), 0.84% (24

months), 0.81% (36 and 48 months), and 0.80% (60 months). This gives an estimate of

1.32% per year, very much in line with the values of 1.4% found by Fung and Hsieh (2000)

and 1.2% found by Capocci and Hübner (2004).

151
Table 21: Estimation of instant return history bias

Mean Monthly Monthly Annual

Av. Nb of Fds
Return Difference Difference
Period 94-02

All 0,99% NA NA 1356

Without 12M 0,88% 0,11% 1,32% 1174

Without 24M 0,84% 0,15% 1,80% 999

Without 36M 0,81% 0,18% 2,16% 837

Without 48M 0,81% 0,18% 2,16% 694

Without 60M 0,80% 0,19% 2,28% 570

Sub-period 01:94-03:00

All 1,35% NA NA 1197

Without 12M 1,22% 0,13% 1,56% 945

Without 24M 1,18% 0,17% 2,04% 752

Without 36M 1,14% 0,21% 2,52% 620

Without 48M 1,14% 0,21% 2,52% 539

Without 60M 1,13% 0,23% 2.76% 498

Sub-period 04:00-12:02

All 0,19% NA NA 1715

Without 12M 0,10% 0,09% 1,08% 1489

Without 24M 0,08% 0,11% 1,32% 1256

This Table reports the instant history bias calculated from our database. Our MAR database contains
2894 hedge funds, including 1622 survived funds and 1272 dissolved funds as of December 2002.
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, 24, 36 and 60 months of returns
(the adjusted observable portfolio). All returns are net of fees and on a monthly basis unless otherwise
indicated.

152
 Contrarily to the analysis of the survivorship bias, the first sub-period is mostly

responsible for the level of the bias. Because the period was increasingly bullish, with the

highest returns being obtained around the end of the period, 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). For the “bearish” sub-period, only

partial results are available due to the small length of this period, but the bias is kept at

very reasonable levels.

One may relate the level of this bias to the one of survivorship bias in this context:

during unfavourable market conditions, 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. As only the most successful funds tend to remain during this time

period, 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, while the observable

portfolio returns include returns from subsequently dissolved funds.

153
4.3 Conclusion

Overall, 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, but for very different reasons. Interestingly, survivorship

bias is higher for the period of down market, while there is evidence of a more important

instant history bias during the upward trending period.

One could have suspected that the high failure rate of hedge funds after March 2000

would have lead survivorship bias to suspiciously high levels, but this is avoided by the

particular behaviour of the database, and especially thanks to the increase in the number of

funds that has been observed over the same time window. Yet, 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.

154
V Hedge Funds Performance

This Section aims at studying whether hedge funds, as a whole or strategy by

strategy, have significantly out-performed the market. We compute all estimations by using

Newey-West (1987) standard errors to adjust for any autocorrelation in the returns.

Table 22 reports the results for All Funds, Individual Funds and Funds of Funds, and

all funds strategies, with equally weighted portfolio excess returns for each investment

style. The model is also estimated for each fund individually.40 To analyse hedge funds

performance in more details, the last columns give the distribution of individually estimated

alphas per strategy, with the percentage of significantly positive, insignificant and negative

alphas at the 5% level.

5.1 Performance Measurement using the CAPM

Panel A of Table 22 reports performance estimates using the CAPM. The estimated

betas are rather low, except for the Long Only Leveraged, and all R-squared are below

60%, except for Long Only Leveraged and Global Est., suggesting the need to use a more

detailed model. Overall, two thirds of the individual funds strategies produce significantly

positive alphas, while the two Funds of Funds strategies out-perform the market at the 10%

level. Overall, hedge funds as a whole also significantly out-perform the market at the 1%

level. Taken individually, 32% of the alphas are significantly positive.

40
To make individual estimation, we require all funds to have consecutive monthly return history for at

least 24 months, so that relatively accurate risk measures can be estimated.

155
Table 22: Performance measurement using the CAPM, Carhart’s 4-factor model and the combined
model

Panel A: Single index model

Alpha Distrib. +
Alpha Mkt R² adj
/ 0 / -
Event driven - Risk Arb 0,46% *** 0,26 *** 0,515 48% 51% 0%
Event driven - Dist. Sec 0,50% *** 0,3 *** 0,41 34% 64% 1%
Global -0,11% 0,47 *** 0,331 9% 78% 11%
Global Est. 0,69% *** 0,58 *** 0,719 30% 67% 2%
Global Intern. 0,37% ** 0,34 *** 0,449 22% 74% 2%
Global Emerging 0,55% 0,62 *** 0,337 20% 75% 4%
US Opp. NA NA NA NA NA NA
Macro 0,32% ** 0,3 *** 0,445 18% 72% 8%
Market Neutral 0,62% *** 0,12 *** 0,378 44% 52% 2%
Long Only Lev. 0,12% 1,05 *** 0,73 9% 80% 9%
Sector 0,99% *** 0,73 *** 0,619 31% 67% 1%
Short Sales 0,81% *** -0,74 *** 0,607 11% 88% 0%
No category 0,43% 0,31 *** 0,174 21% 78% 0%
Individual Funds Total 0,55% *** 0,39 *** 0,651 31% 65% 3%
Niche - FoF 0,31% *** 0,15 *** 0,285 43% 53% 2%
Diversified - FoF 0,24% * 0,26 *** 0,427 33% 62% 3%
Other - FoF NA NA NA NA NA NA
Funds of Funds Total 0,25% * 0,24 *** 0,419 34% 60% 4%
Hedge funds Total 0,48% *** 0,35 *** 0,614 32% 64% 3%
Panel B: Carhart's 4-factor model - individual funds

Alpha Distrib. + /
Alpha Mkt SMB HML PR1YR R² adj
0 / -

Event driven - Risk Arb 0,41% *** 0,28 *** 0,16 *** 0,08 *** -0,01 0,67 37% 60% 1%
Event driven - Dist. Sec 0,41% *** 0,33 *** 0,23 *** 0,11 *** 0,01 0,61 26% 71% 2%
Global -0,19% 0,48 *** 0,19 *** 0,08 0,02 0,36 11% 77% 11%
Global Est. 0,58% *** 0,58 *** 0,26 *** 0,04 0,07 *** 0,87 27% 71% 1%
Global Intern. 0,24% 0,38 *** 0,15 *** 0,09 ** 0,05 * 0,54 21% 71% 7%
Global Emerging 0,43% 0,64 *** 0,31 *** 0,12 0,02 0,40 18% 78% 3%
US Opp. NA NA NA *** NA NA NA NA NA NA
Macro 0,16% 0,33 *** 0,13 *** 0,04 0,11 *** 0,63 17% 74% 8%
Market Neutral 0,55% *** 0,15 *** 0,08 *** 0,07 *** 0,01 0,55 41% 54% 3%
Long Only Lev. 0,13% 0,98 *** 0,39 *** -0,02 0,02 0,82 12% 80% 6%
Sector 0,81% *** 0,71 *** 0,43 *** 0,01 0,15 *** 0,87 26% 72% 1%
Short Sales 0,83% *** -0,66 *** -0,36 *** 0,09 -0,09 ** 0,78 20% 79% 0%
No category 0,26% 0,31 *** 0,26 *** 0,01 0,14 *** 0,36 21% 73% 4%
Individual Funds Total 0,46% *** 0,4 *** 0,2 *** 0,06 ** 0,05 *** 0,83 28% 67% 3%
Niche - FoF 0,21% ** 0,18 *** 0,11 *** 0,07 ** 0,04 ** 0,45 33% 62% 3%
Diversified - FoF 0,10% 0,29 *** 0,16 *** 0,07 ** 0,07 *** 0,64 25% 68% 5%
Other - FoF NA NA NA *** NA NA NA NA NA NA
Funds of Funds Total 0,11% 0,27 *** 0,15 *** 0,07 ** 0,07 *** 0,63 26% 66% 6%
Hedge funds Total 0,37% *** 0,37 *** 0,19 *** 0,06 ** 0,05 *** 0,80 28% 67% 4%
Panel C: The combined model - individual funds
Alpha Distrib.
Alpha Mkt SMB HML PR1YR W x US W Gv Bd Em. Bd High Y. Mortg. Comm. R² adj
+ / 0 / -
Event driven - Risk
0,39% *** 0,25 *** 0,13 *** 0,06 ** 0,01 -0,02 0,04 0,04 * 0,17 *** -0,07 0,03 0,71 37% 59% 2%
Arb
Event driven - Dist.
0,40% *** 0,21 *** 0,19 *** 0,06 0,02 0,06 -0,11 0,10 *** 0,21 *** 0,18 0,03 0,69 27% 71% 1%
Sec

Global -0,10% 0,30 ** 0,12 0,01 0,04 0,18 -0,11 0,14 ** 0,16 -0,09 0,12 ** 0,42 9% 81% 9%

Global Est. 0,56% *** 0,58 *** 0,24 *** 0,03 0,08 *** 0,01 0,03 0,07 *** 0,00 -0,08 0,03 0,88 26% 71% 2%

Global Intern. 0,41% *** 0,12 ** 0,10 *** 0,05 0,06 ** 0,30 *** -0,26 *** 0,05 0,05 0,15 0,05 * 0,66 28% 67% 4%

Global Emerging 0,61% 0,31 ** 0,20 ** 0,01 0,07 0,34 ** -0,66 *** 0,22 *** 0,18 0,11 0,12 * 0,52 15% 84% 0%

US Opp. NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA

Macro 0,12% 0,25 *** 0,13 *** 0,01 0,11 *** 0,05 -0,14 * 0,04 0,05 0,49 *** -0,02 0,66 16% 78% 4%

Market Neutral 0,54% *** 0,12 *** 0,07 *** 0,05 *** 0,01 0,01 -0,06 0,04 *** 0,07 * 0,11 0,02 0,64 39% 57% 3%

Long Only Lev. -0,11% 1,01 *** 0,38 *** -0,08 0,00 -0,11 0,00 0,17 *** 0,06 0,71 ** 0,05 0,85 6% 83% 9%

Sector 0,73% *** 0,84 *** 0,43 *** 0,02 0,15 *** -0,12 * 0,03 0,04 -0,11 -0,18 0,06 ** 0,88 19% 78% 1%

Short Sales 0,92% *** -0,74 *** -0,39 *** 0,08 -0,07 * 0,04 -0,01 -0,07 0,25 * -0,13 0,01 0,78 23% 76% 0%

No category 0,33% 0,16 0,25 *** -0,01 0,13 ** 0,26 ** 0,01 0,18 *** -0,20 0,02 0,03 0,44 21% 78% 0%

Individual Funds
0,46% *** 0,34 *** 0,17 *** 0,03 0,05 *** 0,05 -0,10 * 0,07 *** 0,05 0,05 0,04 ** 0,86 26% 69% 3%
Total
Panel C (continued): The combined model - FoF

Alpha Distrib.
Alpha Mkt SMB HML PR1YR W x US W Gv Bd Em. Bd High Y. Mortg. Comm. R² adj
+ / 0 / -

Niche - FoF 0,24% ** 0,13 *** 0,08 *** 0,05 * 0,05 *** 0,04 -0,07 0,05 *** 0,06 -0,1 0,02 0,51 32% 64% 2%

Diversified - FoF 0,10% 0,21 *** 0,13 *** 0,03 0,08 *** 0,06 -0,17 *** 0,08 *** 0,07 0,16 0,03 0,71 24% 71% 3%

Other - FoF NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA

Funds of Funds
0,12% 0,2 *** 0,12 *** 0,04 0,08 *** 0,06 -0,16 ** 0,07 *** 0,07 0,14 0,03 0,70 25% 70% 3%
Total

Hedge funds Total 0,38% *** 0,31 *** 0,16 *** 0,03 0,06 *** 0,05 -0,11 ** 0,07 *** 0,06 0,07 0,04 ** 0,84 26% 70% 3%

This Table presents the results of the estimation of the single index model (Panel A), of Carhart’s (1997) model (Panel B) and of our combined model (Panel C) for the
01:1994-12:2002 period. We report the OLS estimators for equally weighted portfolios per investment strategy, per type of funds and for all funds. 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. Results for the US Opportunistics and Other
categories are not reported as they have, respectively, 0 and 1 living fund in the second sub-period. We report the percentage of significantly positive alphas (+),
significantly negative alphas (-) and alphas insignificantly different from zero (0) at the 5% level. t-stats are heteroskedasticity consistent. *** Significant at the 1% level, **
Significant at the 5% level and * Significant at the 10% level.
5.2 Performance Measurement using Multi-Factor Models

It is presumably better to use a multi-factor model to account for all possible

investment strategies. In Panel B of Table 22, we report the results for Carhart’s 4-factor

model and in Panel C the results for our combined model applied to hedge funds.

Panels B reveals that the premium on the SMB factor is, in almost all cases,

significant, including in the Short Sellers strategy where it is negative. The coefficients of

the HML and Momentum factors are significant for four and six individual funds strategies,

respectively, and to all funds of funds strategies.

Panel C shows that the explanatory power of the HML factor seems marginal as only

three betas are significantly positive at the 5% level. The Momentum factor remains a

stronger indicator of hedge funds behaviour, with only the Short Sales strategy being

momentum contrarian over the whole period. The results with our combined model also

indicate that all additional factors add explanatory power to the regression. In particular, as

already outlined by Capocci and Hübner (2004), the Emerging Bond factor adds explanatory

power in more than 50% of the strategies with high significance levels. Event Driven

strategies are more prone to bear a high exposure to high yield bond factors, 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.

Evidence on alphas obtained in Panel C is not favourable to Funds of Funds and to

the Macro strategy. Overall, accounting for more risk premia reduces the average reported

out-performance by 0.1% per months. 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, and the distribution of performance

among strategies is also more concentrated.

160
 Overall it seems that the combined model does a very good job in describing hedge

funds behaviour. The average R²adj increases from 0.61 for the single factor model, to 0.80

for the 4-factor model and to 0.84 for our combined model. This coefficient is the best one

reported in the literature so far.41

5.3 Performance over bullish and bearish sub-periods

In order to analyse the performance components in the bullish and bearish market

configurations, we only report results for our combined model. 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
See e.g. .Liang (1999) and Amin and Kat (2003a).

161
Table 23: Performance of hedge funds during the bullish and
bearish sub-periods

Panel A: 01:1994-03:2000

Alpha Mkt SMB HML PR1YR W x US W Gv Bd

Risk Arb 0,50% *** 0,31 *** 0,19 *** 0,03 -0,07 ** -0,01 -0,01

Dist. Sec 0,50% *** 0,25 *** 0,22 *** 0,09 ** -0,07 * 0,11 ** -0,18 **

Global -0,65% ** 0,50 *** 0,02 -0,03 0,01 0,14 -0,39 **

Global Est. 0,62% *** 0,66 *** 0,31 *** -0,01 0,06 0,03 0,00

Global Intern. 0,47% ** 0,20 ** 0,13 ** 0,00 0,01 0,30 *** -0,44 ***

Global Emerging 0,27% 0,51 *** 0,13 0,09 -0,02 0,36 ** -0,98 ***

US Opp. 0,07% -0,07 0,38 ** 0,00 -0,13 0,00 -0,25

Macro 0,09% 0,32 *** 0,14 *** 0,00 0,10 ** 0,06 -0,32 ***

Market Neutral 0,53% *** 0,14 *** 0,07 *** 0,07 *** 0,02 0,00 -0,05

Long Only Lev. 0,03% 1,03 *** 0,47 *** -0,01 -0,04 -0,02 -0,08

Sector 0,95% *** 0,82 *** 0,52 *** 0,03 0,17 *** -0,05 0,00

Short Sales 0,92% *** -0,65 *** -0,45 *** 0,08 -0,05 -0,07 0,09

No category 0,25% 0,26 * 0,36 *** -0,13 0,03 0,34 *** -0,06

Ind. Funds Total 0,44% *** 0,41 *** 0,20 *** 0,03 0,04 0,06 * -0,17 ***

Niche - FoF 0,22% 0,17 *** 0,08 ** 0,10 ** 0,04 0,05 -0,12

Diversified - FoF 0,02% 0,29 *** 0,13 *** 0,06 0,08 ** 0,06 -0,28 ***

Other - FoF 0,24% 0,24 *** 0,11 ** 0,04 -0,06 -0,01 -0,19

F. of Funds Total 0,05% 0,27 *** 0,12 *** 0,07 0,07 * 0,06 -0,26 ***

Total 0,35% *** 0,38 *** 0,18 *** 0,04 0,04 0,06 -0,19 ***

162
Panel A (continued): 01:1994-03:2000

Alpha Distrib.
Em. Bd High Y. Mortg. Comm. R² adj
+ / 0 / -

Risk Arb 0,04 ** 0,11 -0,02 0,00 0,74 46% 51% 2%

Dist. Sec 0,07 *** 0,51 *** -0,11 0,04 0,81 33% 66% 0%

Global 0,08 0,86 *** -1,10 ** 0,17 *** 0,67 9% 81% 9%

Global Est. 0,06 *** -0,18 -0,01 0,04 0,91 25% 72% 1%

Global Intern. 0,05 0,00 0,23 0,06 0,66 27% 70% 1%

Global Emerging 0,22 ** 0,48 -0,71 0,17 * 0,56 10% 87% 1%

US Opp. 0,00 0,34 0,50 -0,09 0,09 6% 82% 10%

Macro 0,04 -0,05 0,53 ** -0,01 0,70 15% 80% 3%

Market Neutral 0,04 *** 0,15 * -0,03 0,03 * 0,66 39% 59% 1%

Long Only Lev. 0,17 *** -0,10 0,85 * 0,01 0,80 3% 88% 7%

Sector 0,03 -0,26 0,10 0,07 ** 0,90 25% 74% 0%

Short Sales -0,06 0,33 -0,43 -0,04 0,71 17% 82% 0%

No category 0,19 *** -0,48 0,22 0,04 0,51 31% 68% 0%

Ind. Funds Total 0,06 *** 0,03 -0,05 0,05 ** 0,89 26% 70% 2%

Niche - FoF 0,05 ** 0,12 -0,30 0,01 0,51 35% 60% 3%

Diversified - FoF 0,08 *** 0,16 -0,06 0,03 0,76 21% 73% 4%

Other - FoF 0,02 0,32 * -0,06 0,03 0,49 18% 77% 4%

F. of Funds Total 0,07 *** 0,15 -0,06 0,03 0,75 23% 71% 4%

Total 0,07 *** 0,06 -0,06 0,04 ** 0,87 46% 51% 2%

163
Panel B: 04:2000-12:2002

Alpha Mkt SMB HML PR1YR W x US W Gv Bd

Risk Arb -0,12% 0,17 ** 0,12 *** 0,08 ** -0,02 -0,05 0,04

Dist. Sec -0,01% 0,22 * 0,18 *** 0,01 0,02 -0,16 0,10

Global -0,12% 0,36 0,29 * 0,16 -0,08 -0,08 0,71 *

Global Est. -0,19% 0,45 *** 0,15 *** 0,00 0,02 -0,08 0,08

Global Intern. -0,02% 0,06 0,12 *** 0,08 0,03 0,25 ** 0,00

Global Emerging 0,25% 0,49 ** 0,31 *** 0,06 0,04 -0,14 0,00

US Opp. NA NA NA NA NA NA NA

Macro -0,25% 0,27 ** 0,14 ** 0,03 0,05 -0,07 0,23

Market Neutral 0,37% *** 0,06 0,05 ** 0,03 0,00 0,01 -0,07

Long Only Lev. -0,50% 1,14 *** 0,18 * -0,19 -0,03 -0,44 ** 0,27

Sector -0,19% 0,97 *** 0,26 *** -0,02 0,08 ** -0,44 *** 0,34 **

Short Sales 0,40% -1,00 *** -0,21 ** 0,19 * -0,13 *** 0,38 ** -0,39 *

No category -0,05% 0,18 0,12 0,00 0,12 *** -0,05 0,13

Ind. Funds Total 0,03% 0,32 *** 0,14 *** 0,02 0,02 -0,09 * 0,06

Niche - FoF -0,05% 0,19 *** 0,08 *** 0,05 * 0,02 * -0,11 ** 0,08

Diversified - FoF -0,13% 0,19 *** 0,12 *** 0,02 0,04 *** -0,06 0,10 *

Other - FoF NA NA NA NA NA NA NA

F. of Funds Total -0,12% 0,19 *** 0,11 *** 0,03 0,04 *** -0,07 0,10 *

Total -0,01% 0,29 *** 0,13 *** 0,02 0,03 ** -0,09 * 0,07

164
Panel B (continued): 04:2000-12:2002
Alpha Distrib.
Em. Bd High Y. Mortg. Comm. R² adj
+ / 0 / -

Risk Arb -0,01 0,23 *** 0,19 0,03 0,86 7% 91% 0%

Dist. Sec 0,12 ** 0,22 *** 0,04 -0,05 * 0,74 13% 86% 0%

Global 0,27 -0,15 -0,56 0,02 0,27 0% 100% 0%

Global Est. 0,06 0,14 ** 0,38 * -0,01 0,95 5% 89% 5%

Global Intern. -0,02 0,08 0,02 -0,01 0,78 9% 87% 3%

Global Emerging -0,04 0,21 0,08 -0,04 0,71 15% 80% 4%

US Opp. NA NA NA NA NA NA NA NA

Macro 0,07 0,06 0,44 -0,05 0,65 6% 85% 7%

Market Neutral 0 0,1 *** 0,25 * -0,01 0,69 27% 68% 3%

Long Only Lev. 0,13 0,24 0,75 0,06 0,92 0% 95% 4%

Sector 0,06 -0,01 -0,01 0,05 0,92 6% 89% 3%

Short Sales -0,05 0,09 0,66 0,11 ** 0,90 0% 90% 9%

No category 0,03 0,04 0,11 -0,05 0,43 0% 100% 0%

Ind. Funds Total 0,03 0,12 *** 0,24 -0,01 0,95 13% 82% 4%

Niche - FoF -0,01 0,08 ** 0,04 0,01 0,76 14% 80% 4%

Diversified - FoF 0,01 0,1 *** 0,09 0 0,89 10% 81% 7%

Other - FoF NA NA NA NA NA NA NA NA

F. of Funds Total 0 0,09 *** 0,08 0 0,88 11% 81% 7%

Total 0,02 0,12 *** 0,2 0 0,94 13% 82% 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. We report the OLS estimators for equally weighted portfolios per
investment strategy, per type of funds and for all funds. 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. In Panel B,
results for the US Opportunistics and Other categories are not reported as they have, respectively, 0 and 1
living fund in the second sub-period. We report the percentage of significantly positive alphas (+), significantly
negative alphas (-) and alphas insignificantly different from zero (0) at the 5% level. t-stats are
heteroskedasticity consistent. *** Significant at the 1% level, ** Significant at the 5% level and * Significant at
the 10% level.

165
 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 sub-

period?

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 sub-

decile 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 Exc. return St. dev Alpha Mkt SMB HML

D1a 1,12% * 6,63% 0,17% 0,64 *** 0,63 *** -0,04

D1b 1,10% ** 4,75% 0,38% 0,57 *** 0,4 *** 0,03

D1c 0,97% ** 3,97% 0,47% ** 0,43 *** 0,35 *** -0,02

D1 1,07% ** 5,00% 0,34% 0,55 *** 0,46 *** -0,01

D2 0,86% *** 3,27% 0,32% * 0,48 *** 0,28 *** 0,04

D3 0,78% *** 2,42% 0,31% ** 0,37 *** 0,19 *** 0,07 **

D4 0,71% *** 2,10% 0,36% *** 0,31 *** 0,13 *** 0,07 **

D5 0,61% *** 1,69% 0,35% *** 0,26 *** 0,09 *** 0,05 *

D6 0,50% *** 1,33% 0,28% *** 0,2 *** 0,09 *** 0,07 ***

D7 0,45% *** 1,37% 0,30% *** 0,17 *** 0,08 *** 0,03

D8 0,46% *** 1,67% 0,40% *** 0,11 *** 0,09 *** -0,01

D9 0,31% 2,34% 0,34% ** 0,18 *** 0,08 *** 0,02

D10 0,44% 3,95% 0,72% ** 0,22 * 0,05 -0,04

D10a 0,54% 3,36% 0,53% ** 0,28 *** 0,09 * 0

D10b 0,03% 4,03% 0,29% 0,27 ** 0,04 -0,01

D10C 1,07% * 5,65% 1,77% *** 0,07 0,03 -0,13

1-10 spread 0,63% 5,03% -0,38% 0,33 ** 0,41 *** 0,03

1a-10c spread 0,05% 7,71% -1,60% ** 0,58 ** 0,6 *** 0,09

1-2 spread 0,21% 2,12% 0,02% 0,07 0,18 *** -0,04

9-10 spread -0,13% 2,17% -0,38% -0,04 0,03 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 0,36 *** 0,14 -0,56 ** 0,19 ** 0,07 0,51 0,13 ** 0,77

D1b 0,29 *** 0,05 -0,19 0,11 * 0,16 -0,06 0,07 0,77

D1c 0,2 *** 0,07 -0,15 0,08 * 0,15 0,03 0,04 0,78

D1 0,28 *** 0,08 -0,29 * 0,12 ** 0,12 0,15 0,07 * 0,80

D2 0,20 *** -0,04 -0,02 0,06 0,08 -0,2 0,04 0,81

D3 0,11 *** 0,01 -0,07 0,07 ** 0,05 0,12 0,03 0,79

D4 0,08 *** 0,04 -0,13 * 0,06 ** 0,05 0,05 0,04 * 0,75

D5 0,05 *** 0,01 -0,05 0,05 ** 0,09 -0,05 0,02 0,70

D6 0,03 ** 0,02 -0,09 ** 0,05 *** 0,08 * 0,07 0,02 0,77

D7 0,01 0,04 -0,10 ** 0,04 *** 0,05 0,05 0,03 * 0,74

D8 -0,02 0,11 *** -0,18 *** 0,06 ** 0,08 0,15 0,05 ** 0,70

D9 -0,10 *** 0,19 *** -0,23 *** 0,05 0,05 0,12 0,04 0,73

D10 -0,19 *** 0,28 ** -0,29 0,04 0,12 -0,15 0,08 0,56

D10a -0,13 *** 0,22 ** -0,02 0,09 * 0,04 0,14 0,03 0,60

D10b -0,14 *** 0,29 ** -0,32 * 0,03 0,09 -0,41 0,07 0,56

D10C -0,32 *** 0,35 * -0,53 * -0,02 0,25 -0,26 0,1 0,36

1-10 spread 0,47 *** -0,2 0 0,08 0 0,3 0 0,52

1a-10c spread 0,67 *** -0,21 -0,03 0,21 -0,19 0,76 0,03 0,48

1-2 spread 0,08 *** 0,12 * -0,27 *** 0,07 * 0,04 0,35 0,03 0,54

9-10 spread 0,09 ** -0,09 0,06 0,01 -0,07 0,27 -0,04 0,18

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 sub-

periods. 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
Table 25: Hedge funds persistence during the bullish and
bearish sub-periods

Panel A: 01:1994-03:2000

Portfolio Exc. return St. dev Alpha Mkt SMB HML

D1a 2,09% ** 6,85% 0,17% 0,69 *** 0,70 *** -0,07

D1b 1,87% *** 4,46% 0,46% 0,62 *** 0,41 *** 0,19 **

D1c 1,68% *** 3,97% 0,55% ** 0,48 *** 0,41 *** -0,01

D1 1,89% *** 4,97% 0,40% 0,6 *** 0,51 *** 0,04

D2 1,44% *** 3,27% 0,41% ** 0,57 *** 0,32 *** 0,13 **

D3 1,15% *** 2,56% 0,30% ** 0,46 *** 0,21 *** 0,15 ***

D4 1,08% *** 2,35% 0,31% ** 0,39 *** 0,14 *** 0,09 **

D5 0,90% *** 1,89% 0,28% ** 0,35 *** 0,09 *** 0,09 ***

D6 0,75% *** 1,45% 0,27% *** 0,24 *** 0,10 *** 0,07 ***

D7 0,76% *** 1,45% 0,36% *** 0,21 *** 0,11 *** 0,03

D8 0,76% *** 1,67% 0,42% *** 0,13 ** 0,11 *** -0,05

D9 0,67% ** 2,08% 0,28% 0,21 *** 0,08 * -0,08

D10 0,97% ** 3,60% 0,41% 0,32 ** 0,00 -0,15

D10a 1,01% ** 3,18% 0,43% 0,35 *** 0,09 -0,12

D10b 0,77% * 3,54% 0,27% 0,25 0,04 -0,11

D10C 1,37% ** 5,35% 0,75% 0,39 -0,17 -0,24

1-10 spread 0,92% * 4,32% -0,01% 0,28 0,51 *** 0,19

1a-10c spread 0,71% 7,11% -0,59% 0,3 0,87 *** 0,17

1-2 spread 0,45% 2,19% -0,01% 0,03 0,19 *** -0,08

9-10 spread -0,30% 2,22% -0,13% -0,11 0,08 0,07

175
Panel A: 01:1994-03:2000

Portfolio PR1YR W x US W Gv Bd Em. Bd High Y. Mortg. Comm. R² adj

D1a 0,34 *** 0,25 -0,94 *** 0,15 -0,56 1,18 0,27 *** 0,78

D1b 0,28 *** 0,18 * -0,4 ** 0,05 -0,28 -0,03 0,15 *** 0,84

D1c 0,17 ** 0,16 * -0,43 *** 0,05 -0,18 0,41 0,10 ** 0,81

D1 0,26 *** 0,19 * -0,58 *** 0,08 -0,35 0,51 0,17 *** 0,84

D2 0,13 *** 0,06 -0,21 ** 0,02 -0,05 -0,32 0,06 * 0,89

D3 0,06 * 0,06 -0,21 *** 0,05 ** 0,11 -0,16 0,04 0,89

D4 0,05 0,05 -0,22 *** 0,05 ** 0,29 ** -0,29 0,06 ** 0,88

D5 0,03 0,00 -0,07 0,05 ** 0,35 ** -0,50 ** 0,01 0,83

D6 0,02 0,02 -0,10 * 0,05 *** 0,20 ** -0,07 0,02 0,85

D7 -0,02 0,05 -0,14 ** 0,05 *** 0,16 ** 0,00 0,01 0,84

D8 -0,04 0,12 ** -0,23 *** 0,07 *** 0,06 0,27 0,06 ** 0,70

D9 -0,09 ** 0,17 *** -0,31 *** 0,06 * -0,01 0,25 0,04 0,67

D10 -0,06 0,17 -0,24 0,07 0,45 -0,78 0,07 0,46

D10a -0,14 * 0,17 0,05 0,11 * 0,23 -0,23 -0,02 0,57

D10b -0,02 0,23 -0,3 0,05 0,36 -0,63 0,06 0,40

D10C 0,02 0,09 -0,47 0,03 0,75 -1,78 0,17 0,26

1-10 spread 0,33 *** 0,03 -0,33 0,01 -0,79 ** 1,29 * 0,10 0,54

1a-10c spread 0,32 0,16 -0,47 0,12 -1,30 * 2,97 ** 0,10 0,41

1-2 spread 0,13 ** 0,14 * -0,37 *** 0,06 -0,30 0,83 ** 0,11 *** 0,55

9-10 spread -0,03 0,00 -0,07 0,00 -0,46 1,03 * -0,03 0,09

176
Panel B: 04:2000-12:2002

Portfolio Exc. return St. dev Alpha Mkt SMB HML

D1a -0,73% 5,83% 0,16% 0,46 * 0,33 ** -0,29 *

D1b -0,37% 5,01% 0,61% 0,28 0,14 -0,36 **

D1c -0,37% 3,67% 0,26% 0,33 * 0,18 ** -0,19 *

D1 -0,49% 4,74% 0,34% 0,36 * 0,22 ** -0,28 **

D2 -0,25% 3,03% 0,05% 0,37 *** 0,14 ** -0,15 **

D3 0,05% 1,95% 0,12% 0,3 *** 0,12 *** -0,04

D4 0,02% 1,30% -0,02% 0,27 *** 0,11 *** 0,05

D5 0,06% 1,03% 0,08% 0,2 *** 0,09 *** 0,03

D6 0,03% 0,93% 0,04% 0,15 *** 0,08 *** 0,06 **

D7 -0,12% 0,97% 0,01% 0,14 ** 0,06 ** 0,03

D8 -0,13% 1,55% 0,06% 0,14 0,12 ** 0,04

D9 -0,38% 2,68% -0,31% 0,38 *** 0,19 *** 0,23 ***

D10 -0,56% 4,42% -0,91% * 0,72 *** 0,23 ** 0,42 ***

D10a -0,34% 3,56% -0,88% * 0,69 *** 0,28 *** 0,4 ***

D10b -1,40% * 4,56% -1,36% ** 0,89 *** 0,14 0,42 ***

D10C 0,50% 6,23% -0,13% 0,5 0,33 * 0,46 **

1-10 spread 0,07% 6,20% 1,25% -0,36 -0,02 -0,7 ***

1a-10c spread -1,22% 8,71% 0,30% -0,04 0 -0,75 ***

1-2 spread -0,24% 1,93% 0,29% -0,01 0,07 -0,13 *

9-10 spread 0,18% 2,07% 0,60% * -0,33 *** -0,04 -0,19 ***

177
Panel B: 04:2000-12:2002

Portfolio PR1YR W x US W Gv Bd Em. Bd High Y. Mortg. Comm. R² adj

D1a 0,4 *** 0,02 -0,15 0,21 0,54 ** 0,4 -0,11 0,85

D1b 0,38 *** 0,04 -0,15 0,3 ** 0,65 ** 0,19 -0,13 ** 0,82

D1c 0,21 *** -0,09 0,2 0,09 0,43 ** -0,16 -0,1 ** 0,81

D1 0,33 *** -0,01 -0,03 0,2 * 0,54 ** 0,14 -0,12 ** 0,86

D2 0,22 *** -0,19 0,17 0,12 * 0,33 ** 0,03 -0,07 ** 0,88

D3 0,11 *** -0,16 * 0,14 0,03 0,19 ** 0,14 -0,04 * 0,85

D4 0,05 *** -0,12 ** 0,11 * 0,02 0,07 0,1 -0,03 * 0,86

D5 0,02 * -0,10 ** 0,07 -0,01 0,09 ** 0 -0,01 0,85

D6 0,01 -0,04 -0,02 -0,02 0,09 ** 0,05 0 0,84

D7 -0,02 -0,03 0,01 -0,02 0,06 -0,02 0,02 0,77

D8 -0,05 ** 0,00 -0,04 -0,04 0,08 0,01 0,03 0,71

D9 -0,17 *** 0,01 0,03 -0,05 -0,08 0,3 0,05 0,84

D10 -0,43 *** -0,13 0,23 -0,07 -0,32 ** 1,09 * 0,16 *** 0,84

D10a -0,28 *** -0,12 0,16 -0,03 -0,27 * 1,25 ** 0,11 ** 0,77

D10b -0,36 *** -0,2 0,26 -0,07 -0,33 * 0,54 0,18 ** 0,78

D10C -0,71 *** -0,02 0,27 -0,12 -0,33 1,61 0,17 * 0,76

1-10 spread 0,76 *** 0,11 -0,26 0,27 0,85 ** -0,95 -0,28 *** 0,73

1a-10c spread 1,1 *** 0,04 -0,42 0,34 0,87 ** -1,22 -0,28 ** 0,75

1-2 spread 0,11 *** 0,18 -0,2 0,08 0,21 ** 0,11 -0,05 0,66

9-10 spread 0,26 *** 0,13 -0,2 0,02 0,24 ** -0,79 ** -0,11 *** 0,74
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. 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.

178
 Panels B and C display very different pictures. During the 1994-March 2000 period,

only the alphas of top decile funds were systematically higher than for the whole period. In

contrast, the lowest decile funds did not out-perform the market. Middle decile funds had a

clearer focus on high yield bond markets.

Panel C shows that the persistence in performance during the market collapse was

clearly sustained for portfolios D2 to D6, with high significance levels. These funds had no

particularly remarkable investment pattern, 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), with a relatively high adjusted R² of 57.8%.

Since these decile portfolios had significant alphas for the first sub-period too, this

indicates that the superior performance of these funds was predictable irrespective of the

prevailing 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, as our results suggest that

this performance is not only sustained during positive or negative market conditions, but

during both; however, persistence in performance is observable for the medium-to-top past

performers only, showing that only a very targeted investment behaviour in Market Neutral

Funds would provide a sustained positive abnormal return.

Finally, 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, leaving a large

importance to market timing and tactical allocation. The alternative explanation of missing

risk factors, which is also reported in Liang (2003), although theoretically possible, is not

very compelling given the proven very high accuracy of our combined model. However, this

particular aspect opens the way to additional research on hedge funds performance and

persistence during unfavourable market conditions.

179
Table 26: Hedge funds persistence for the Market Neutral
strategy

Panel A: 01:1994-12:2002

Portfolio Exc. return St. dev Alpha Mkt SMB HML

D1a 1,44% *** 3,75% 0,93% ** 0,28 * 0,25 *** 0,14

D1b 1,43% *** 3,50% 1,08% *** 0,4 *** 0,11 0,12

D1c 1,30% *** 2,55% 0,92% *** 0,24 *** 0,17 *** 0,03

D1 1,37% *** 2,59% 0,97% *** 0,3 *** 0,17 *** 0,09

D2 1,16% *** 1,74% 1,01% *** 0,16 ** 0,13 *** 0,14 ***

D3 0,81% *** 1,12% 0,61% *** 0,17 *** 0,06 *** 0,06 **

D4 0,66% *** 1,30% 0,54% *** 0,05 0,09 *** 0,03

D5 0,54% *** 0,99% 0,44% *** 0,04 0,06 *** 0,03

D6 0,45% *** 0,80% 0,33% *** 0,08 *** 0,04 ** 0,04 **

D7 0,31% *** 0,74% 0,25% *** 0,03 0,04 *** 0,01

D8 0,41% *** 0,86% 0,42% *** -0,01 0,02 0,00

D9 0,50% *** 1,44% 0,48% *** 0 0,01 0,00

D10 0,69% *** 2,40% 0,64% *** 0,05 0,10 ** -0,02

D10a 0,56% ** 2,51% 0,58% *** -0,15 * 0,07 -0,09

D10b 0,45% 3,00% 0,43% 0,06 0,14 ** 0,04

D10C 1,14% *** 4,18% 0,84% * 0,37 ** 0,10 0,07

1-10 spread 0,69% ** 3,18% 0,34% 0,25 * 0,07 0,10

1a-10c spread 0,30% 5,30% 0,08% -0,09 0,15 0,07

1-2 spread 0,21% 2,31% -0,04% 0,14 0,04 -0,06

9-10 spread -0,19% 1,97% -0,16% -0,06 -0,10 ** 0,02

180
Panel A (continued): 01:1994-12:2002

Portfolio PR1YR W x US W Gv Bd Em. Bd High Y. Mortg. Comm. R² adj

D1a 0,17 *** 0,04 -0,16 -0,04 0,16 0,36 0,01 0,28

D1b 0,10 -0,21 0,21 0,12 0,18 -0,92 * 0,06 0,13

D1c 0,19 *** 0,02 -0,08 0,02 0,05 -0,02 0,03 0,52

D1 0,16 *** -0,05 -0,01 0,03 0,13 -0,22 0,03 0,44

D2 0,03 0,07 -0,11 -0,02 -0,04 -0,05 0,03 0,33

D3 0,03 * -0,05 -0,03 0,06 *** 0,06 -0,08 0,00 0,44

D4 0,00 -0,02 -0,17 ** 0,02 0,17 ** 0,17 0,00 0,27

D5 0,00 0,02 -0,09 0,07 *** 0,06 0,11 0,00 0,33

D6 0,00 0,02 -0,07 0,05 *** 0,01 0,08 0,00 0,43

D7 -0,02 0,02 -0,11 *** 0,03 ** 0,07 * 0,16 * 0,00 0,40

D8 -0,01 0,08 *** -0,13 *** 0,03 ** 0,08 * 0,07 0,02 0,38

D9 -0,03 0,12 ** -0,10 0,05 0,10 0,22 0,05 * 0,27

D10 -0,12 *** 0,16 ** -0,12 0,06 0,15 0,64 ** 0,01 0,42

D10a -0,07 ** 0,33 *** -0,21 0,19 *** 0,21 * 0,65 ** -0,01 0,48

D10b -0,16 *** 0,18 * -0,26 0,10 0,16 0,63 -0,04 0,37

D10C -0,20 *** -0,13 0,24 -0,11 0,04 0,89 0,08 0,20

1-10 spread 0,28 *** -0,21 * 0,11 -0,03 -0,02 -0,86 * 0,02 0,23

1a-10c spread 0,36 *** 0,17 -0,4 0,08 0,12 -0,53 -0,07 0,14

1-2 spread 0,13 *** -0,12 0,1 0,05 0,17 -0,17 0,00 0,10

9-10 spread 0,09 ** -0,04 0,01 -0,01 -0,04 -0,42 0,03 0,14

181
Panel B: 01:1994-03:2000

Portfolio Exc. return St. dev Alpha Mkt SMB HML

D1a 2,17% *** 3,58% 1,41% *** 0,22 0,32 *** 0,04

D1b 2,01% *** 3,23% 1,67% *** 0,37 ** 0,19 * 0,06

D1c 1,79% *** 2,62% 1,01% *** 0,31 *** 0,19 *** 0,02

D1 1,97% *** 2,45% 1,34% *** 0,3 *** 0,23 *** 0,04

D2 1,30% *** 2,07% 0,99% *** 0,28 *** 0,15 *** 0,22 ***

D3 0,97% *** 1,27% 0,54% *** 0,25 *** 0,07 ** 0,11 ***

D4 0,79% *** 1,50% 0,59% *** 0,07 0,08 * 0,12 **

D5 0,62% *** 1,14% 0,40% *** 0,04 0,06 * 0,04

D6 0,55% *** 0,89% 0,35% *** 0,11 *** 0,06 *** 0,06 **

D7 0,38% *** 0,72% 0,27% *** 0,01 0,04 ** 0,02

D8 0,58% *** 0,88% 0,46% *** 0 0,02 -0,03

D9 0,64% *** 1,52% 0,47% *** 0,03 -0,01 -0,04

D10 0,86% *** 2,24% 0,63% * -0,01 0,04 -0,03

D10a 0,70% ** 2,41% 0,50% * -0,14 -0,01 -0,1

D10b 0,68% ** 2,65% 0,34% -0,07 0,08 -0,07

D10C 1,31% ** 4,30% 1,10% * 0,28 0,01 0,25

1-10 spread 1,11% *** 3,04% 0,71% * 0,31 ** 0,19 ** 0,07

1a-10c spread 0,86% 5,43% 0,31% -0,07 0,31 * -0,2

1-2 spread 0,67% ** 2,24% 0,36% 0,02 0,08 -0,18 *

9-10 spread -0,22% 1,94% -0,16% 0,04 -0,05 -0,01

182
Panel B (continued): 01:1994-03:2000

Portfolio PR1YR W x US W Gv Bd Em. Bd High Y. Mortg. Comm. R² adj

D1a 0,27 ** 0,02 -0,18 -0,1 -0,46 1,08 0,14 * 0,39

D1b -0,11 -0,07 -0,07 0,04 -0,33 -0,06 0,11 0,11

D1c 0,23 *** 0,07 -0,25 -0,01 -0,31 0,19 0,08 0,61

D1 0,13 * 0,00 -0,17 -0,03 -0,38 0,35 0,10 ** 0,50

D2 -0,01 0,10 -0,21 -0,04 -0,07 -0,36 0,01 0,47

D3 0,04 -0,09 ** 0,02 0,07 *** 0,23 ** -0,41 ** -0,01 0,59

D4 0,04 -0,07 -0,13 0,03 0,51 ** -0,23 -0,01 0,34

D5 0,03 -0,01 -0,02 0,08 *** 0,29 ** -0,07 0,01 0,39

D6 -0,02 0,03 -0,08 0,05 *** 0,05 0,01 0,00 0,51

D7 -0,01 0,01 -0,06 0,04 *** 0,23 ** 0,12 0,01 0,52

D8 -0,01 0,06 -0,11 * 0,05 ** 0,2 ** 0 0,04 ** 0,48

D9 -0,05 0,09 -0,11 0,05 0,38 ** -0,13 0,07 ** 0,42

D10 -0,05 0,13 0,05 0,08 0,44 0,22 0,03 0,15

D10a -0,03 0,27 *** -0,11 0,22 *** 0,39 0,14 0,01 0,43

D10b 0,01 0,09 0,14 0,14 ** 0,58 * 0,47 -0,01 0,16

D10C -0,17 -0,07 0,29 -0,14 0,41 0,03 0,08 0,04

1-10 spread 0,18 * -0,13 -0,22 -0,1 -0,81 ** 0,13 0,07 0,24

1a-10c spread 0,44 ** 0,09 -0,47 0,04 -0,87 1,05 0,06 0,18

1-2 spread 0,14 * -0,09 0,04 0,01 -0,31 0,71 0,09 0,07

9-10 spread 0 -0,04 -0,16 -0,03 -0,06 -0,35 0,04 -0,11

183
Panel C: 03:2000-12:2002

Portfolio Exc. return St. dev Alpha Mkt SMB HML

D1a 0,05% 3,71% 0,05% -0,16 0,01 0

D1b 0,32% 3,78% 0,79% -0,06 0,09 -0,09

D1c 0,37% 2,15% -0,02% 0,23 0,12 0,01

D1 0,24% 2,51% 0,25% 0,01 0,07 -0,03

D2 0,91% *** 0,78% 0,72% *** 0,07 0,04 0,04

D3 0,51% *** 0,64% 0,62% *** 0 -0,01 -0,02

D4 0,42% *** 0,73% 0,49% *** -0,05 0,03 -0,03

D5 0,38% *** 0,57% 0,51% *** 0,01 0,01 0,01

D6 0,25% *** 0,55% 0,35% ** 0 0 -0,01

D7 0,17% 0,77% 0,31% * 0,02 0,03 0,00

D8 0,07% 0,73% 0,05% -0,02 0,05 0,04

D9 0,23% 1,24% -0,22% 0,11 0,1 0,15 **

D10 0,36% 2,68% -0,25% 0,49 *** 0,28 *** 0,25 ***

D10a 0,29% 2,71% -0,19% 0,17 0,28 *** 0,16

D10b 0,02% 3,58% -0,27% 0,44 ** 0,36 *** 0,38 ***

D10C 0,83% 3,98% -0,49% 1,04 *** 0,26 0,25

1-10 spread -0,12% 3,33% 0,50% -0,48 * -0,21 -0,28 *

1a-10c spread -0,78% 4,95% 0,54% -1,21 ** -0,26 -0,25

1-2 spread -0,67% * 2,23% -0,47% -0,06 0,03 -0,07

9-10 spread -0,12% 2,06% 0,03% -0,38 *** -0,18 *** -0,1

184
Panel C (continued): 03:2000-12:2002

Portfolio PR1YR W x US W Gv Bd Em. Bd High Y. Mortg. Comm. R² adj

D1a 0,16 * 0,55 -0,48 0,36 * 0,4 1,74 * -0,12 0,30

D1b 0,23 ** 0,03 -0,04 0,38 * 0,5 -1,08 -0,13 0,13

D1c 0,12 ** -0,14 0,18 0,11 0,19 0,56 -0,05 0,43

D1 0,17 *** 0,13 -0,10 0,27 ** 0,36 ** 0,4 -0,1 * 0,46

D2 0,00 -0,04 0,03 -0,01 0,1 0,29 0,01 0,19

D3 0,01 0,01 -0,07 0,01 0,12 ** -0,02 0 0,13

D4 -0,02 0,04 -0,10 -0,03 0,13 ** 0,17 0,02 0,59

D5 -0,01 0,04 -0,09 * -0,02 0,04 -0,01 0 0,49

D6 0,01 0,04 -0,08 0,01 0,07 0,05 -0,01 0,16

D7 -0,01 0,00 -0,12 -0,05 0,07 -0,08 -0,02 0,35

D8 -0,03 * 0,06 -0,12 -0,07 ** 0,06 0,18 -0,02 0,44

D9 -0,08 ** 0,00 0,03 -0,04 -0,03 0,7 * 0 0,16

D10 -0,18 *** -0,13 -0,06 -0,06 -0,12 1,37 ** 0,03 0,84

D10a -0,11 ** 0,15 -0,2 0,04 0,01 1,51 ** -0,01 0,60

D10b -0,26 *** 0,02 -0,39 * -0,03 -0,24 * 0,83 0,07 0,79

D10C -0,24 *** -0,71 ** 0,52 -0,15 -0,19 2,16 ** 0,05 0,51

1-10 spread 0,35 *** 0,26 -0,04 0,33 ** 0,47 ** -0,97 -0,13 * 0,53

1a-10c spread 0,39 *** 1,26 ** -1 * 0,51 * 0,58 -0,42 -0,17 0,22

1-2 spread 0,17 *** 0,17 -0,12 0,29 ** 0,26 0,11 -0,11 * 0,36

9-10 spread 0,11 *** 0,13 0,09 0,02 0,09 -0,67 ** -0,04 0,78
This Table reports the result of the estimation of our combined model for the Market Neutral strategy 01:1994-
03:2002 (Panel A) period, and the 01:1994-03:2000 (Panel B) and 04:2000-12:2002 (Panel C) sub-periods.
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%

185
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 this study, 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.

Firstly, 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, namely

the survivorship and instant return history biases despite the high attrition rate of funds

observed after March 2000. 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.

The analysis of performance indicates that most hedge funds significantly out-

performed the market during the whole test period, but this is mostly due to the bullish

sub-period. The pattern is somehow attenuated for funds of funds strategies. In contrast, no

significant under-performance of individual hedge funds of funds of funds strategies is

observed when markets headed south. The Market Neutral strategy provides a noticeable

exception, however, as is sustains abnormal performance over both the bullish and the

bearish sub-periods.

Persistence analysis also indicates that most of the predictability of superior

performance is to be found prior to March 2000. Our results confirm several previous

studies that found that persistence, if any, is mostly located among medium performers. In

the second sub-period, only negative persistence can be found among the past losers,

186
suggesting that bad performance has probably been the decisive factor for hedge funds

mortality.

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, abnormal performance

and persistence are pervasive throughout the sub-periods, probably thanks to an extreme

adaptability and a very active investment behaviour.

Obviously, these very appealing results call for a much more detailed analysis of the

Market Neutral strategy among individual hedge funds. Market timing issues do matter for

their risk exposure, and traditional asset pricing models may not fully account for their

highly unstable investment strategies. 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, but this particular field of investigation is left for future theoretical as

well as empirical research.

187
The Sustainability in Hedge Fund

Performance: new insights

Daniel P.J. CAPOCCI

HEC-ULG Management School – University of Liège (Belgium)

Capocci, Daniel, 2006, The Sustainability in Hedge Fund Performance: new insights,

Working Paper, HEC-ULG Management School

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. The methodology used is adapted from Capocci and Hübner

(2004). The measures used include the returns, volatility, Sharpe score, alpha, beta,

skewness and kurtosis. Measures incorporating the volatility display very strong ability to

assist investors in creating alpha as well as consistently and significantly outperforming

classical indices.

190
The Sustainability of Hedge Fund Performance: New

Insights

Introduction

The hedge fund industry has experienced explosive growth since the early 1990s

when approximately 2,000 hedge funds were managing assets totalling roughly $60

billion. It is estimated that today, there are roughly 10,000 hedge funds managing close

to a trillion USD in assets. The growing trend of the sector has been remarkably

sustained during the stock market collapse that started in March 2000, when the

NASDAQ Composite Index attained an all-time high of 5,132 and finished three years

later at a bottom of 1,253. In the meantime, the global net asset value (NAV) of hedge

funds continued to grow at a steady 10.6% (see Capocci et al., 2005), contrasting with a

decrease of 2.7% for the mutual fund industry worldwide (Investment Company

Institute, 2003).

This relatively positive attitude of investors is typically motivated by the

perceptions that most hedge funds tend to be largely market neutral, that their

managers enjoy greater flexibility in their asset allocation enabling them to achieve

superior market-timing skills (see Chen and Liang, 2006), or that hedge funds have a

relatively low covariance with other classes of financial assets, making them good

diversification vehicles (Amin and Kat, 2003b). Assness et al. (2001) found that this

result is at least partly spurious.

191
 Hedge funds have been studied since 1997 and the early seminal studies were

Fung and Hsieh (1997) and Ackermann et al. (1999). Since then, the literature on the

subject has expanded quickly. Our analysis focuses on hedge fund performance and

persistence in hedge fund performance. Three fields exist that examine hedge fund

performance. The first includes studies that compare the performance of hedge funds

with equity and other indices (see for example, Ackermann et al., 1999; Brown et al.

1999; Liang, 1999; Amin and Kat, 2003b; Liang, 2001; Liang, 2003; Agarwal and Naik,

2004). They do not all converge in their findings. Some authors (Brown et al., 1999 and

Liang, 1999) conclude that hedge funds are able to outperform these indices, whereas

others (Ackermann et al., 1999 as well as Agarwal and Naik, 2004) are more cautious in

their conclusions.

The second field of hedge fund performance analysis compares the performance of

hedge funds with that of mutual funds. In this context, Ackermann et al. (1999) and

Liang (1999) find that hedge funds constantly obtain superior performance to mutual

funds, although lower and more volatile returns than the reference market indices

considered.

Finally, the third group of hedge fund performance analysis examines the

persistence of hedge fund returns. Persistence is particularly important in the case of

hedge funds because, as suggested by Brown et al. (1999) and Liang (2000, 2001), the

hedge fund industry has a higher attrition rate than mutual funds. Brown et al. (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. Agarwal and Naik (2000) analyse the presence of persistence in hedge

fund returns using a one-year moving average period. The authors find evidence of

persistence in hedge fund performance, particularly for poor performing funds that

continue to underperform. Capocci and Hübner (2004) conclude that some low-risk

192
managers have been able to persistently create alpha between January 1994 and

December 2002.

The vast majority of performance studies on hedge funds have not focused on the

behaviour of hedge funds under different market conditions. The periods under

examination do not favour this exercise, as periods of downward trending stock markets

have been rare and discontinuous between 1994 and March 2000.44 For the period 1990-

1998, Edwards and Caglayan (2001) find that only three hedge fund strategies (market

neutral, event driven and macro) provide protection to investors when stock markets

drop.

Fund of hedge funds are particular in the hedge fund world and need particular

care. As stated by Fung and Hsieh (2006) and Liang (2003), there is a double counting

issue with funds of hedge funds that should be distinguished from individual hedge funds.

Our study makes two main contributions. First, we find a systematic way to

extract hedge funds that consistently beat traditional markets. Capocci and Hübner

(2004) indicated that some hedge funds consistently and significantly outperform the

equity and bond markets over time. They do not; however, describe a systematic way of

selecting hedge funds that outperform. We test various ways of classifying funds and

indicate how to select funds that consistently outperform the classical markets. Our

second contribution is our multi-factor model that is adapted from Capocci and Hübner’s

(2004) performance analysis model.

44
Most empirical evidence reveals that data collected prior to 1994 by several data vendors displays a

significant survivorship bias, as shown by Fung and Hsieh (2000), Liang (2000) and Capocci and

Hübner (2004).

193
 Following Liang (2003), we separate individual funds from funds of funds in order

to determine if there are significant differences in exposures and in alpha creation. Our

results do not indicate that this is justified.

The study is organised as following. Section I describes the methodology. In

Section II, we report the database and analyse the descriptive statistics. The attrition

rate is calculated in Section III. 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. Section

VIII concludes the study.

194
I Methodology

The starting point of our study on hedge fund performance is the original Sharpe

(1964) – Lintner (1965) CAPM. We use the Capocci and Hübner (2004) multifactor model

that extends the Carhart (1997) specification by combining it with factors specific to

hedge funds.45 Our multi-factor models contain 10 and 14 factors, 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.S. 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 add the four option factors of Agarwal and Naik (2004)

that are at-the money call option (ATMC), out-of-the-money call option (OTMC), at-the-

money put option (ATMP) and out-of-the-money put option (OTMP). As reported in Table

27, high correlations between these factors and the US market factor (ranging from -

0.87 between the market and ATMP to 0.78 between the market and ATMC) and between

these four factors (from -0.73 between ATMC and ATMP to 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.

195
 1) They may be redundant with the market factor and not marginally contribute to

the return attribution decomposition;

2) They may be redundant with each other.

We do not use the Lehman Mortgage-Backed Securities Index as suggested by

Capocci et al. (2005), because this factor has low explanatory power. Several additional

factors, such as the MSCI Emerging Markets Index, 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.

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 ) (9)
+ 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 )

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.

196
 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 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 ) (10)

+ β P 9 (GSCI t − R Ft ) + β P10 (CUR t − R Ft )

+ 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; RFt corresponds to the risk-free

return on month t; RMt equals the return of the market portfolio on month t47, SMBt is

equal to the factor-mimicking portfolio for size (small minus big), 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. These factors aim at

isolating the firm-specific components of returns. MSWXUSt is equal to the return of the

MSCI World Index excluding US; SWGBIt corresponds to the return of the Salomon World

Government Bond Index; JPMEMBIt equals the return of the JP Morgan Emerging Market

47
The market proxy used is the value-weighted portfolio of all NYSE, Amex and NASDAQ stocks – a

market proxy that is usually used in mutual fund and hedge fund performance studies.

197
Table 27: Correlation between the Option Factors and the

Market Factor

Market ATM Call OTM Call ATM Put OTM Put

ATM Call 0,78 1

OTM Call 0,73 0,96 1

ATM Put -0,87 -0,73 -0,69 1

OTM Put -0,85 -0,7 -0,65 0,99 1

This table reports the correlation between the option factors and the market factor over the January
1994-December 2002 period.

Bond Index; HYt is equal to the return of the Lehman High Yield Credit Bond Index, GSCIt

corresponds to the return of the Goldman Sachs Commodity Index, CURt equals the

return of the Federal Reserve Bank Trade Weighted Dollar Index and ATMCt, OTMCt,

ATMPt and OTMPt correspond respectively to Agarwal and Naik’s (2004) at-the-money,

out-of-the money European call option factors, and at-the-money and out-of-the money

European put option factors. εPt equals the error term, while αP and βP are the intercept

and the slope of the regression, respectively.

In order to determine if some funds consistently and significantly create alpha

over time, we follow the methodology of Carhart (1997). Every year we rank all funds

based on their total return of the prior year. Every January, we place all funds into 10

equally weighted portfolios, 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. The portfolios are held until the following January and then rebalanced again.

Funds that disappear during the course of the year are included in the average until their

198
death, and then portfolio weights are readjusted appropriately. This yields a time series

of monthly returns on each decile portfolio from January 1995 to December 2002.

In each case, we also test the significance of the difference between the alpha of

the top (D1) and the bottom (D10) decile portfolios, the top (D1) and the next (D2)

decile portfolio, the next to bottom (D9) and the bottom portfolio (D10) and finally

between extremes alphas (D1a and D10c).

Since past performance is only one element in analysing past returns, we decide to
48
perform the decile classification methodology incorporating the Sharpe ratio , standard

deviation, alpha, beta, skewness and kurtosis.

48
The Sharpe ratio is the ratio of excess performance over the risk-free rate to risk, as measured by

the standard deviation.

199
II Database

Four main hedge fund databases are available for empirical studies, the Managed

Account Reports, Inc./Centre for International Securities Derivatives Markets

(MAR/CISDM), Hedge Fund Research, Inc., TASS Management and the Barclay’s

database. The first three are the most used in academic studies. The data providers

collect information supplied by hedge fund managers. For a majority of funds, they

record other useful information such as company name, starting and ending date,

investment strategy, assets under management, management and incentive fees,

manager's name, manager's address, etc. Nevertheless, there is no consensus on the

definition of the strategy followed. Each provider proposes its classification, but on many

occasions there maybe some similarities.

We combine hedge fund data from MAR/CISDM with Barclay’s data. The MAR

database started with 2,246 individual funds including 1,185 funds that were still living at

the end of 2002 and 1,061 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. The Barclay’s database includes 1,967 individual funds

including 1,271 funds that were still alive at the end of 2002, 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. In each database, 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, company, and returns) appeared

twice in the database; when the same fund (same name and returns) appeared twice in the database

with two different company names; and when the same fund (same company and returns) appeared

twice in the database with two different fund names.

200
returns. We further found 1,153 funds that were present in the two databases (834

existing funds and 319 dissolved funds). We obtain a total of 3,060 individual funds and

907 funds of funds including 1,910 individual funds and 588 funds of funds from the

MAR/CISDM database and 1,150 individual funds and 319 funds of funds from the

Barclay’s database. This database is one of the largest and most unique databases ever

used in hedge fund performance studies. These funds include a total of 1,873 (61,15%)

surviving individual funds, 1,190 (38,85%) dissolved individual funds, 653 (72%)

surviving funds of hedge funds and 254 (28%) dissolved funds of funds. All the returns

used are net of fees.

Most previous studies on hedge funds group individual funds and funds of funds.

We follow Liang (2003) and analyse them together and separately. We separate them for

three main reasons. First, we want to avoid double counting individual funds that are

part of funds of funds and that are in the database. This would lead to what we call the

fund of funds bias. Secondly, we separate individual funds from funds of funds because

funds of funds returns can be biased. There is a double survivorship bias in their returns.

More precisely, 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, 2000). On the other hand, individual funds in funds of hedge

funds portfolios can also be dissolved, which leads to the presence of a second

survivorship bias. 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.

201
 Finally, 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, 1999). This should at

least partly be explained by the double fee structure (see Brown et al., 2002). 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.

202
III Preliminary Analysis

Before analysing the presence of bias in hedge fund data, we present the

descriptive statistics, the correlation analysis and the bias analysis.

3.1 Descriptive Statistics

Panel A of Table 28 reports the descriptive statistics of the hedge fund database.

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. The mortality ratios50 reported are almost all between

50% and 70% indicating high mortality levels for individual hedge funds. Interestingly,

this ratio is also high for funds of funds (at 72%).

Sector funds have offered the highest mean monthly return (1.56% monthly),

whereas short sales and funds of hedge funds have offered the lowest returns (0.8%).

Funds not classified are the most volatile, whereas the least volatile funds are market

neutral funds (1.22%). The skewness column indicates that most return distributions are

negatively skewed indicating that average negative returns are higher than the average

positive returns. The excess kurtosis reported is all positive beginning with 0.87 for short

sales and topping out at 4.98 for distressed securities. This confirms the presence of fat

tails in hedge funds return distribution.

50
Ratio of dissolved funds over the period covered to the number of funds in the database for the

corresponding strategy.

203
Table 28: Descriptive Statistics

Panel A: Hedge Funds Descriptive Statistics

N % Dead funds Mean Std dev. Median

- Distressed sec 122 4% 61% 1,03 2,65 1,09

- Risk arb 168 5% 77% 0,99 1,81 1,01

- Event driven 59 2% 17% 1,13 1,96 0,85

Event Driven Total 349 11% 61% 1,04 2,07 1,09

Global emerging 226 7% 58% 1,02 3,12 1,36

Global Macro 190 6% 69% 0,88 2,04 0,73

Global 702 23% 46% 1,13 3,08 1,12

Global Total 1118 37% 49% 0,89 2,70 1,05

Mkt ntl 731 24% 64% 1,02 1,22 1,00

Equity Hedge 332 11% 67% 1,13 2,35 1,23

Sector 277 9% 70% 1,56 3,60 1,73

Short sales 48 2% 71% 0,80 2,19 0,68

Long only leveraged 142 5% 73% 1,15 2,89 1,17

Mkt timing 43 1% 93% 1,01 1,94 0,91

Currency fund 4 0% 25% 1,20 3,53 1,14

Option strategy 3 0% 0% 1,02 2,62 1,16

No strategy 13 0% 8% 0,85 4,52 0,26

Individual funds total 3060 100% 60% 1,09 2,25 1,15

Fd of Fds 907 NA 72% 0,80 1,90 0,81

204
Panel A (continued): Hedge Funds Descriptive Statistics

Skewness Kurtosis Min Max Sharpe score

- Distressed sec -0,95 4,98 -12,14 7,37 0,25

- Risk arb -0,51 2,72 -6,89 6,08 0,34

- Event driven 1,27 3,40 -4,16 9,66 0,39

Event Driven Total -0,63 3,85 -8,75 6,44 0,32

Global emerging -0,73 3,94 -13,66 10,97 0,21

Global Macro 0,57 1,25 -4,44 7,33 0,25

Global 0,19 2,19 -10,16 11,88 0,25

Global Total -0,48 3,82 -11,46 9,62 0,19

Mkt ntl -0,23 2,11 -4,04 4,61 0,54

Equity Hedge -0,27 2,43 -8,45 8,89 0,32

Sector 0,17 2,37 -11,44 15,33 0,33

Short sales 0,06 0,87 -5,17 7,91 0,19

Long only leveraged -0,10 1,13 -9,27 9,61 0,27

Mkt timing -0,27 1,65 -6,02 6,19 0,33

Currency fund -0,24 0,99 -10,00 10,36 0,23

Option strategy -0,91 4,29 -11,20 7,83 0,21

No strategy 1,19 3,32 -10,24 17,18 0,11

Individual funds total -0,17 2,59 -8,23 8,20 0,32

Fd of Fds -0,16 3,08 -7,32 6,93 0,22

205
Panel B: Naive Strategies Descriptive Statistics

Mean Std dev. Median Skewness

Market 0,78 4,77 1,58 -0,72

MSCI w. ex US 0,09 4,36 0,47 -0,45

SMB 0,02 4,45 -0,36 0,87

HML 0,6 4,16 0,68 0,48

PR1YR 1,14 5,74 1,27 -0,73

Lehman Agg. -0,01 1,13 0,19 -0,13

Sal. WGBI 0,5 1,83 0,24 0,45

JPM EMBI 0,82 4,67 1,16 -1,64

Leh. High Yield 0,4 2,11 0,66 -0,7

GSCI 0,59 5,35 0,61 0,27

Risk-free rate 0,37 0,11 0,4 -0,99

ATM Call 0,82 84,51 -20,34 0,58

OTM Call -2,97 89,57 -30,35 0,79

ATM Put -11,88 95,02 -56,76 1,62

OTM Put -15,73 95,55 -59,39 1,81

Broad Dollar Index 0,34 1,11 0,29 0,33

206
Panel B: Naive Strategies Descriptive Statistics

Kurtosis Min Max Sharpe score

Market 0,37 -15,69 8,33 0.16

MSCI w. ex US 0,26 -12,89 10,27 0.02

SMB 5,56 -16,26 21,38 0.00

HML 1,14 -8,91 13,67 0.14

PR1YR 4,6 -25,13 18,21 0.20

Lehman Agg. 0,61 -2,73 3,87 -0.01

Sal. WGBI 0,47 -3,44 5,94 0.27

JPM EMBI 7,18 -24,2 10,87 0.18

Leh. High Yield 3,58 -7,37 7,49 0.19

GSCI 0,26 -12,28 15,79 0.11

Risk-free rate 0,15 0,11 0,56 NA

ATM Call -0,78 -99,04 216,91 NA

OTM Call -0,41 -99,1 246,85 NA

ATM Put 2,66 -95,29 386,02 NA

OTM Put 3,64 -96,21 400,33 NA

Broad Dollar Index 1,29 -2,79 3,93 NA

This table reports the descriptive statistics for hedge fund strategies (Panel A) and for naïve strategies (Panel B).
N stands for number of funds and Std dev. is for standard deviation. Our original MAR/CISDM/Barclay’s
database consists of 3,060 individual funds (including 1,160 dissolved funds) and 907 funds of funds (including
254 dissolved funds of funds) over the January 1994-December 2002 period.

207
Panel B reports the same statistics for passive investment strategies. The bond and

equity indices all have lower mean returns than the majority of the hedge fund

strategies. The equity indices returns are more volatile than most hedge fund strategies,

but the standard deviations of the bond indices are almost all lower than those of the

hedge fund strategies. The skewness and kurtosis measures indicate the same pattern as

for hedge funds, asymmetry to the left and presence of fat tails.

3.2 Correlation Analysis

Traditional hedge fund literature contends that, thanks to the weak correlation

between hedge funds and other securities, hedge funds are likely to improve the risk-

return trade-off when added to a traditional portfolio (see Fung and Hsieh, 1997; Liang,

1999 as well as Amin and Kat, 2003b). We analyse the correlation among the hedge

funds strategies, between the hedge fund strategies and the passive investment

strategies and among the passive investment strategies.51 As typically reported in the

literature (see Liang, 2003), hedge fund strategies are highly correlated, with the

exception of short sellers that systematically has the opposite result. The coefficients

reported are all higher than 0.5 (except short selling) and most of them are higher than

0.75.

To avoid any risk of multi-colinearity, we also verified the correlation between the

explanatory factors. 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. This justifies the use of two different models.

51
We do not report the results for the sake of brevity, but complete results are available directly from

the author.

208
3.3 Survivorship Bias

The most important bias in hedge fund data is the survivorship bias (see Fung and

Hsieh, 1997; Fung and Hsieh, 2000 and Liang, 2001). When only living funds52 are

considered, there is a survivorship bias in the figures because dissolved funds tend to

have inferior performance compared to surviving funds. Survivorship bias is usually

defined as the performance difference between living funds and all funds (Fung and

Hsieh, 2000). We obtain a bias of 1.08% per year.53 This result is lower than the 0.30%

monthly bias found by Fung and Hsieh (2000), the 3% bias found by Liang (2001) and

the industry consensus bias of 3% stressed by Amin and Kat (2003a).54 On a yearly

basis, 2000 and 2001 have been years of important differences in results between living

and dissolved funds. When only individual funds are considered, figures reported are a

little higher at 1.44%. For funds of funds, we obtain 0.96%.

52
By living funds we mean funds still in operation at the moment of the analysis.
53
Complete results are available upon request.
54
We find this consensus value quite high when compared to the 0.8-1.5 bias reported by Malkiel

(1995) for US mutual funds.

209
IV Global Results: Market Analysis

Our analysis has two parts. First, we focus on the performance of hedge funds by

grouping them per investment strategy. Then, 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.

4.1 Performance Analysis

Panel A of Table 29 reports the results obtained while using our multi-factor model

(1). The results reported are particularly strong. Every strategy but macro funds, option

strategy funds, no sub-strategy funds and funds of funds create significant alpha at the

5% significance level. On the exposure side, almost all strategies are significantly

exposed to the US equity market (inversely for short sellers) with a beta ranging from -

0.25 for short sellers to 0.66 for sector funds. Few strategies are significantly exposed to

non-US equity markets. The size factor is significantly positive for every strategy, 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. Option arbitrage strategies have been biased towards large companies. This is

perfectly in line with the strategy since option strategies tend to focus the S&P 500.

The style factor indicates a value bias for risk arbitrage and market neutral funds

and a growth bias for currency funds. Interestingly, the momentum factor is significantly

positive for most strategies indicating that many hedge funds tend to be momentum

players.

210
 On the bond side, 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. The

equity emerging market factor is strongly significantly positive for almost every strategy.

The exception is short sellers that are slightly negatively exposed to this factor and

currency funds, whose exposure is not significant. The r-squared are particularly high

between 0.26 and 0.89 with an average at 0.72.

We performed the same analysis adding the option factors of Agarwal and Naik

(2004) and report the results in Panel B of Table 29. A comparison of the two panels

indicates two interesting patterns.

First, for most strategies the alpha and the exposures to the factors (in terms of

significance) remain the same. Second, for global emerging markets funds, the alpha is

no longer significant, the exposure to the US equity market disappears, but the exposure

to the ATMC option factor is significant. This result interestingly indicates that the returns

offered by emerging market funds are not so highly correlated with the US equity

market, but that they offer at-the-money call option return features. 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). On the other side, such managers

tend to cut the exposure in bear market conditions.

Another interesting element is that the alpha of global funds is no longer

significant, but the exposure to the factor remains identical. This result indicates that

results may become unstable with the addition of auto-correlated factors. The same

result is found for the market exposure of market timers. For currency funds, there is a

significant exposure to the ATM call factor, but no significant change in the alpha or in

the other exposures.

211
Table 29: Hedge Fund Strategies Performance Analysis (1/1994-

12/2002)

Panel A: Model (1)

MSCI w ex
Alpha S&P SMB HML PR1YR
US

- Distressed sec 0,44 *** 0,31 *** 0,06 0,22 *** 0,01 0,05 *

- Risk arb 0,43 *** 0,28 *** 0 0,14 *** 0,07 *** 0,01

- Event driven 0,58 *** 0,23 *** 0,09 * 0,12 *** 0,03 0,07 **

Event Driven Total 0,46 *** 0,28 *** 0,04 0,16 *** 0,04 0,04

Global emerging 0,47 ** 0,29 *** 0,2 *** 0,14 *** 0,01 0,08 ***

Global Macro 0,23 * 0,29 *** 0 0,12 *** 0,02 0,12 ***

Global 0,45 *** 0,47 ** 0,08 * 0,22 *** 0,03 0,08 **

Global Total 0,31 ** 0,3 *** 0,12 *** 0,14 *** 0 0,09 ***

Mkt ntl 0,49 *** 0,19 *** 0 0,09 *** 0,06 *** 0,03 *

Equity Hedge 0,51 *** 0,36 *** 0,06 * 0,17 *** 0,05 0,06 **

Sector 0,7 *** 0,66 ** -0,09 * 0,33 *** 0,01 0,12 ***

Short sales 0,56 *** -0,25 *** 0,02 -0,19 *** 0,07 -0,03 *

Long only leveraged 0,46 *** 0,47 *** 0,04 0,22 *** 0,03 0,08 ***

Mkt timing 0,41 *** 0,25 *** -0,01 0,11 *** 0,04 0,06 ***

Currency fund 0,58 ** 0,42 *** -0,05 0,17 *** -0,18 *** 0,12 ***

Option strategy 0,42 0,16 0,08 0,12 * 0,06 0,06

No strategy 0,28 0,18 0,3 * 0,32 *** -0,07 0,12 **

Individual funds total 0,47 *** 0,34 ** 0,04 0,17 *** 0,03 0,06 ***

Fd of Fds 0,2 * 0,24 *** 0,05 0,13 *** 0,04 0,08 ***

Global Index 0,4 *** 0,32 *** 0,04 0,16 *** 0,03 0,07 ***

212
Panel A (continued): Model (1)

WGBI EMBI HY GSCI DOLLAR R²

- Distressed sec -0,06 0,09 *** 0,22 *** 0,04 0,07 0,82

- Risk arb 0,01 0,04 *** 0,08 0,01 -0,04 0,73

- Event driven -0,13 ** 0,07 *** -0,06 0 -0,14 ** 0,65

Event Driven Total -0,04 0,07 *** 0,11 ** 0,02 -0,01 0,82

Global emerging -0,39 *** 0,09 *** 0,16 * 0,06 * -0,28 * 0,76

Global Macro -0,08 0,05 ** 0,09 * -0,01 -0,04 0,73

Global -0,12 ** 0,07 *** 0 0,04 * -0,17 ** 0,86

Global Total -0,28 *** 0,08 *** 0,11 0,04 -0,23 ** 0,78

Mkt ntl -0,06 * 0,04 *** 0,06 0,02 * -0,08 0,77

Equity Hedge -0,07 0,05 *** 0,06 0,02 -0,13 * 0,87

Sector -0,02 0,07 *** -0,04 0,05 ** -0,02 0,88

Short sales -0,05 -0,05 * 0,05 -0,01 -0,1 0,52

Long only leveraged -0,07 0,07 *** 0,01 0 -0,17 * 0,89

Mkt timing -0,07 0,07 *** 0,15 ** 0,03 -0,03 0,63

Currency fund 0 0,12 * -0,24 -0,06 0,37 0,48

Option strategy -0,06 0,1 ** -0,01 -0,04 -0,08 0,26

No strategy -0,08 0,24 ** -0,25 0,04 -0,4 0,47

Individual funds total -0,1 *** 0,06 *** 0,05 0,03 * -0,12 * 0,88

Fd of Fds -0,15 *** 0,07 *** 0,07 0,02 -0,1 0,78

Global Index -0,11 *** 0,06 *** 0,05 0,03 * -0,11 0,86

213
PANEL B: Model (2)

MSCI w ex
Alpha S&P SMB HML PR1YR WGBI EMBI
US

- Distressed sec 0,39 ** 0,28 *** 0,06 * 0,22 *** 0,02 0,05 ** -0,06 0,09 ***

- Risk arb 0,42 *** 0,22 *** 0,00 0,15 *** 0,07 *** 0,01 -0,01 0,04 **

- Event driven 0,60 *** 0,28 *** 0,09 * 0,11 *** 0,02 0,07 ** -0,12 * 0,07 ***

Event Driven Total 0,44 *** 0,25 *** 0,04 0,17 *** 0,05 0,04 * -0,06 0,06 ***

Global emerging 0,38 0,17 * 0,21 *** 0,16 *** 0,01 0,08 *** -0,37 *** 0,10 ***

Global Macro 0,27 * 0,23 *** 0,01 0,14 *** 0,03 0,12 *** -0,12 0,04

Global 0,49 *** 0,42 *** 0,09 * 0,23 *** 0,03 0,08 ** -0,15 ** 0,06 ***

Global Total 0,26 0,19 *** 0,13 *** 0,16 *** 0,00 0,09 *** -0,29 *** 0,08 ***

Mkt ntl 0,52 *** 0,13 *** 0,00 0,10 *** 0,06 *** 0,03 ** -0,07 ** 0,04 ***

Equity Hedge 0,50 *** 0,30 *** 0,07 * 0,18 *** 0,05 0,06 *** -0,08 0,05 ***

Sector 0,75 *** 0,58 *** -0,08 0,34 *** 0,02 0,12 *** -0,04 0,07 ***

Short sales 0,49 *** -0,23 *** 0,01 -0,20 *** 0,06 -0,03 0,00 -0,04

Long only leveraged 0,48 *** 0,39 *** 0,04 0,24 *** 0,03 0,08 *** -0,09 * 0,07 ***

Mkt timing 0,45 *** 0,08 0,00 0,14 *** 0,05 0,06 *** -0,11 0,07 **

Currency fund 0,67 *** 0,29 ** -0,04 0,18 *** -0,19 *** 0,12 ** 0,03 0,13 *

Option strategy 0,31 0,03 0,09 0,14 ** 0,07 0,06 -0,06 0,10 *

No strategy 0,29 0,31 0,30 0,32 *** -0,08 0,12 * -0,09 0,24 **

Individual funds total 0,48 *** 0,28 *** 0,05 0,18 *** 0,03 0,07 *** -0,12 *** 0,06 ***

Fd of Fds 0,19 0,19 *** 0,05 0,14 *** 0,04 0,08 *** -0,14 *** 0,07 ***

Global Index 0,41 *** 0,26 *** 0,05 0,17 *** 0,04 0,07 *** -0,12 *** 0,06 ***

214
PANEL B (continued): Model (2)
ATM ATM OTM OTM
EMBI HY GSCI DOLLAR R²
CALL PUT CALL PUT
- Distressed sec 0,09 *** 0,20 *** 0,03 0,04 0,00 0,01 0,01 -0,02 * 0,83

- Risk arb 0,04 ** 0,07 0,01 -0,07 0,00 0,01 0,00 -0,01 0,74

- Event driven 0,07 *** -0,04 0,01 -0,09 0,01 0,00 -0,01 0,00 0,66

Event Driven Total 0,06 *** 0,10 * 0,02 -0,03 0,00 0,01 0,00 -0,01 0,83

Global emerging 0,10 *** 0,14 * 0,07 ** -0,33 * 0,01 *** 0,01 -0,01 * -0,02 0,77

Global Macro 0,04 0,09 0,00 -0,06 0,00 0,00 0,00 0,00 0,74

Global 0,06 *** 0,00 0,04 * -0,19 ** 0,00 0,00 0,00 0,00 0,87

Global Total 0,08 *** 0,10 0,05 * -0,28 ** 0,01 0,01 0,00 -0,01 0,79

Mkt ntl 0,04 *** 0,06 * 0,02 ** -0,10 * 0,00 -0,01 0,00 0,01 0,78

Equity Hedge 0,05 *** 0,05 0,02 -0,15 * 0,00 0,00 0,00 0,00 0,87

Sector 0,07 *** -0,05 0,05 ** -0,06 0,00 -0,01 0,00 0,01 0,89

Short sales -0,04 0,04 -0,02 -0,10 0,00 0,00 0,00 0,00 0,54

Long only leveraged 0,07 *** 0,01 0,00 -0,20 ** 0,00 0,00 0,00 0,00 0,89

Mkt timing 0,07 ** 0,14 ** 0,03 -0,12 0,00 -0,01 0,00 0,00 0,67

Currency fund 0,13 * -0,22 -0,04 0,36 0,02 ** -0,04 * -0,02 * 0,04 * 0,50

Option strategy 0,10 * -0,04 -0,04 -0,16 0,00 0,02 0,00 -0,02 0,29

No strategy 0,24 ** -0,23 0,05 -0,30 0,01 0,02 0,00 -0,01 0,48

Individual funds total 0,06 *** 0,04 0,03 * -0,15 ** 0,00 0,00 0,00 0,00 0,88

Fd of Fds 0,07 *** 0,06 0,03 -0,12 0,00 0,00 0,00 0,00 0,79

Global Index 0,06 *** 0,05 0,03 * -0,14 * 0,00 0,00 0,00 0,00 0,87

This Table presents the results of the performance estimation of the combined Model for the January 1994-
December 2002 period. S&P stands for the market factor, MSWXUSt = return of the MSCI World Index excluding
US, SMBt = the factor-mimicking portfolio for size (‘small minus big’), HMLt = the factor-mimicking portfolio for
book-to-market equity (‘high minus low’), PR1YRt = the factor-mimicking portfolio for the momentum effect, EMBIt
= return of the JP Morgan Emerging Market Bond Index; SWGBIt = return of the Salomon World Government
Bond Index; HYt = return of the Lehman High Yield Credit Bond Index, GSCIt = return of the Goldman Sachs
Commodity Index, DOLLAR = return of the Federal Reserve Bank Trade Weighted Dollar Index and ATMCt,
OTMCt, ATMPt and OTMPt = respectively Agarwal and Naik (2004) at-the-money (ATMC), out-of-the money
(OTMC) European call option factors, and at-the-money (ATMP) and out-of-the money (OTMP) European put
option factors. T-statistics are heteroskedasticity consistent. *** Significant at the 1% level, ** Significant at the 5%
level and * Significant at the 10% level.

215
 Panel B indicates that the option factors may be helpful in analysing hedge fund

returns for some specific strategies55, but that there is a serious risk of multi-colinearity.

Therefore, in the rest of the study, we report only the results of model (1).

4.2 Persistence in Performance - Returns

1. 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. As one would expect, the returns increase almost monotically between

deciles D1 and D10. 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 alphas reported are significantly positive for deciles D2 to D8, but not for the

previous year’s best and worst performing funds. The factor exposures enable us to

distinguish between the deciles. Alpha creators (decile D2 to D8) have been long

equities, small cap biased and value oriented. Moreover, they have been exposed to

emerging market bonds, to high yield bonds and marginally to commodities. 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. The previous year’s best performing funds are strongly

exposed to small companies, pure momentum players with a short exposure to the USD.

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
All in one, our model clearly explains most of the hedge fund’s performance and enables

us to distinguish between funds.

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 bottom one

offering a significantly higher alpha (-0.48% per month) but at a cost of higher volatility.

This result indicates that there is no consistence in returns for the most volatile funds in

the database. Ideally, investors should invest in low volatility funds but if the invest in

higher volatility funds, 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. A corollary is

that investors looking at directional hedge funds should favour the one with the highest

past performance independently of the volatility.

2. Individual Hedge Funds Versus Funds of Hedge Funds

We perform the same analysis by separating individual hedge funds from funds of

hedge funds.56 For individual funds, 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. The volatility of these funds is so high that the

returns obtained are no longer significantly different from zero. The values obtained are

very close to the results obtained for the whole database.

56
The corresponding tables are available upon request.

217
Table 30: Hedge Fund Persistence in Performance (1/1995-
12/2002)

Panel A: Multi-factor estimation - Global Funds (1/1995-12/2002)

Mean Std dev. Alpha S&P MSCI w ex US SMB HML

D1 0,85 3,79 0,72 * 0,21 * 0,30 *** 0,07 -0,03

D2 0,73 2,28 0,36 ** 0,20 *** 0,16 *** 0,07 ** 0,01

D3 0,81 1,65 0,39 *** 0,14 *** 0,09 *** 0,09 *** 0,00

D4 0,83 1,41 0,31 *** 0,19 *** 0,04 * 0,07 *** 0,03

D5 0,93 1,46 0,35 *** 0,21 *** 0,02 0,09 *** 0,05 ***

D6 1,02 1,84 0,34 *** 0,27 *** 0,02 0,10 *** 0,05

D7 1,16 2,17 0,43 *** 0,33 *** 0,02 0,15 *** 0,06 *

D8 1,16 2,52 0,31 ** 0,40 *** -0,01 0,19 *** 0,07 *

D9 1,32 3,31 0,34 * 0,50 *** -0,05 0,29 *** 0,04

D10 1,43 4,82 0,40 0,52 *** 0,09 0,43 *** -0,02

D1a 1,26 5,14 1,36 *** 0,15 0,38 ** 0,05 -0,07

D1b 0,50 3,81 0,32 0,23 ** 0,30 ** 0,07 0,01

D1c 0,91 3,22 0,67 *** 0,21 *** 0,26 *** 0,09 * -0,06

D10a 1,30 3,72 0,44 ** 0,44 *** 0,06 0,33 *** -0,01

D10b 1,48 4,55 0,45 * 0,50 *** 0,08 0,37 *** 0,00

D10c 1,60 6,62 0,36 0,65 *** 0,12 0,56 *** -0,05

Spread 1-10 -0,58 4,72 -0,12 -0,32 0,24 -0,33 *** -0,03

Spread 1-2 0,12 2,01 -0,06 0,04 0,14 0,02 -0,06

Spread 9-10 -0,12 1,90 -0,48 *** 0,01 -0,17 ** -0,14 *** 0,06

Spread 1a-10c -0,34 7,02 0,45 -0,52 0,31 -0,46 *** -0,03

218
Panel A (continued): Multi-factor estimation - Global Funds (1/1995-12/2002)

PR1YR WGBI EMBI HY GSCI DOLLAR R²

D1 -0,18 *** -0,30 ** 0,02 0,07 0,07 * -0,07 0,63

D2 -0,07 *** -0,19 *** 0,06 *** 0,05 0,03 -0,06 0,76

D3 -0,02 -0,12 *** 0,05 *** 0,09 * 0,03 *** 0,06 0,75

D4 0,01 -0,08 ** 0,04 *** 0,07 * 0,02 * 0,00 0,80

D5 0,04 ** -0,08 ** 0,04 *** 0,08 ** 0,01 -0,09 0,82

D6 0,07 *** -0,11 * 0,06 *** 0,07 0,01 -0,15 * 0,78

D7 0,09 *** -0,12 * 0,06 *** 0,03 0,03 -0,13 0,82

D8 0,13 *** -0,09 0,07 *** 0,05 0,01 -0,21 * 0,83

D9 0,20 *** -0,06 0,05 0,05 0,03 -0,24 * 0,85

D10 0,30 *** -0,32 *** 0,10 * 0,12 0,06 -0,25 0,83

D1a -0,27 *** -0,45 * -0,02 0,19 0,09 -0,15 0,51

D1b -0,16 *** -0,35 *** 0,02 0,04 0,07 * 0,00 0,58

D1c -0,12 *** -0,11 0,06 0,04 0,05 -0,04 0,70

D10a 0,20 *** -0,16 0,04 0,08 0,01 -0,14 0,82

D10b 0,29 *** -0,32 *** 0,09 * 0,15 0,05 -0,29 0,82

D10c 0,40 *** -0,54 *** 0,17 ** 0,14 0,12 -0,32 0,79

Spread 1-10 -0,49 *** 0,02 -0,13 -0,10 0,03 0,25 0,55

Spread 1-2 -0,12 *** -0,08 -0,05 -0,03 0,04 0,07 0,32

Spread 9-10 -0,10 *** 0,33 *** -0,05 -0,05 -0,03 0,02 0,55

Spread 1a-10c -0,69 *** 0,00 -0,26 * -0,07 -0,06 0,07 0,51

This Table presents the results of the persistence in performance estimation of the combined Model for the
January 1994-December 2002 period. We report the OLS estimators for equally weighted portfolios per
investment strategy, sub-strategy and for all funds. Each year, every fund is classified into a decile on the basis of
its performance over the previous year. S&P stands for the market factor, MSWXUSt = return of the MSCI World
Index excluding US, SMBt = the factor-mimicking portfolio for size (‘small minus big’), HMLt = the factor-mimicking
portfolio for book-to-market equity (‘high minus low’), PR1YRt = the factor-mimicking portfolio for the momentum
effect, EMBIt = return of the JP Morgan Emerging Market Bond Index; SWGBIt = return of the Salomon World
Government Bond Index; HYt = return of the Lehman High Yield Credit Bond Index, GSCIt = return of the
Goldman Sachs Commodity Index, DOLLAR = return of the Federal Reserve Bank Trade Weighted Dollar Index
and ATMCt, OTMCt, ATMPt and OTMPt = respectively Agarwal and Naik (2004) at-the-money (ATMC), out-of-the
money (OTMC) European call option factors, and at-the-money (ATMP) and out-of-the money (OTMP) European p

219
 Funds of funds mean returns, alphas and standard deviations are lower than the

corresponding statistic for individual funds. Alphas are however significantly positive for

deciles D2 to D5. There are four main differences in the factor exposures. Firstly, no

single fund of fund has been momentum contrarian. Secondly, the negative exposure to

the government bond index is significantly lower. Thirdly, the emerging market bond

factor is significant for every decile when funds of funds are considered. Finally, the

currency factor is no longer significant at all. All in one, individual funds outperform fund

of funds and they can be distinguished by their respective factor exposure.

4.3 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. To take this constraint into

account, 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.

The alphas reported in Table 31 are consistent with our one-year classification

period analysis. No significance for top and bottom deciles. There are however some

differences in terms of exposures. The exposure to ex-US equity markets is lower and no

more significant for top deciles. The exposure to the emerging market bond index

changes quite importantly. 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, but the R² remains high.

220
Table 31: Persistence in Performance Analysis Based on Three-
year Data (1/1997-12/2002)

Panel A: Multi-factor estimation - Global Funds classified on their performance (1/1997-12/2002)

MSCI w ex
Mean Std dev. Alpha S&P SMB HML
US

D1 0,68 3,20 0,57 * 0,17 * 0,18 0,10 * 0,00

D2 0,68 2,09 0,41 * 0,08 0,17 ** 0,06 -0,03

D3 0,81 1,38 0,43 *** 0,13 *** 0,02 0,06 *** 0,01

D4 0,79 1,43 0,35 *** 0,18 *** 0,03 0,06 *** 0,04 *

D5 0,83 1,97 0,37 *** 0,23 *** 0,05 0,09 *** 0,03

D6 0,90 1,97 0,38 *** 0,26 *** 0,02 0,09 *** 0,04

D7 0,91 2,34 0,36 *** 0,33 *** 0,01 0,14 *** 0,05 *

D8 0,99 2,91 0,29 * 0,48 *** -0,06 0,19 *** 0,07 *

D9 1,02 3,67 0,25 0,57 *** -0,07 0,26 ** 0,04

D10 1,11 5,02 0,40 0,56 *** 0,08 0,41 *** -0,03

D1a 0,59 4,38 0,39 0,41 *** 0,08 0,14 ** 0,06

D1b 0,90 3,43 1,00 *** -0,02 0,27 ** 0,02 -0,01

D1c 0,66 2,90 0,46 * 0,19 *** 0,15 * 0,12 *** -0,05

D10a 0,86 4,61 0,19 0,48 *** 0,08 0,32 *** -0,06

D10b 1,04 4,91 0,33 0,59 *** 0,06 0,41 *** 0,02

D10c 1,28 6,01 0,50 0,59 *** 0,13 0,50 *** -0,06

Spread 1-10 -0,43 3,95 -0,03 -0,39 *** 0,11 -0,30 *** 0,02

Spread 1-2 0,01 1,81 -0,20 0,09 0,03 0,08 *** 0,00

Spread 9-10 -0,10 2,05 -0,49 ** -0,02 -0,13 * -0,20 *** 0,07

Spread 1a-10c -0,69 4,28 -0,77 -0,03 -0,05 -0,18 ** 0,24 *

221
Panel A: Multi-factor estimation - Global Funds classified on their performance (1/1997-12/2002)

PR1YR WGBI EMBI HY GSCI DOLLAR R²

D1 -0,12 *** -0,31 ** 0,06 0,09 0,06 -0,35 ** 0,66

D2 0,01 -0,22 *** 0,05 * 0,11 0,01 -0,18 0,66

D3 0,00 -0,08 * 0,06 *** 0,11 *** 0,01 -0,05 0,72

D4 0,02 -0,08 ** 0,05 *** 0,08 ** 0,02 -0,03 0,80

D5 0,04 * -0,10 * 0,07 *** 0,10 * 0,01 -0,13 0,80

D6 0,06 *** -0,08 0,07 *** 0,10 ** 0,01 -0,14 0,83

D7 0,06 *** -0,10 0,08 *** 0,10 * 0,02 -0,23 * 0,87

D8 0,10 *** 0,00 0,10 *** 0,04 0,02 -0,15 0,88

D9 0,16 *** -0,06 0,08 *** 0,05 0,02 -0,17 0,89

D10 0,20 *** -0,24 ** 0,12 *** 0,03 0,06 -0,18 0,89

D1a -0,15 ** -0,20 0,05 0,09 0,05 -0,21 0,56

D1b -0,19 *** -0,49 *** 0,01 0,16 * 0,05 -0,53 ** 0,56

D1c -0,06 -0,14 0,11 *** 0,05 0,07 ** -0,19 0,71

D10a 0,21 *** -0,27 ** 0,14 *** 0,11 0,03 -0,25 0,86

D10b 0,14 *** -0,26 ** 0,19 *** -0,04 0,05 -0,36 0,88

D10c 0,27 *** -0,34 ** 0,14 *** -0,02 0,08 * -0,27 0,85

Spread 1-10 -0,31 *** -0,04 -0,03 0,07 -0,02 -0,22 0,61

Spread 1-2 -0,10 *** -0,05 0,07 * -0,08 0,01 -0,15 0,31

Spread 9-10 -0,04 0,15 -0,03 0,08 -0,04 -0,17 0,49

Spread 1a-10c -0,35 *** 0,03 0,00 0,10 -0,06 -0,34 0,43

This Table presents the results of the persistence in performance estimation of the combined Model for the
January 1994-December 2002 period. We report the OLS estimators for equally weighted portfolios per
investment strategy, sub-strategy and for all funds. Each year, every fund is classified into a decile on the
basis of its performance over the previous three years. S&P stands for the market factor, MSWXUSt = return
of the MSCI World Index excluding US, SMBt = the factor-mimicking portfolio for size (‘small minus big’),
HMLt = the factor-mimicking portfolio for book-to-market equity (‘high minus low’), PR1YRt = the factor-
mimicking portfolio for the momentum effect, EMBIt = return of the JP Morgan Emerging Market Bond Index;
SWGBIt = return of the Salomon World Government Bond Index; HYt = return of the Lehman High Yield
Credit Bond Index, GSCIt = return of the Goldman Sachs Commodity Index, DOLLAR = return of the
Federal Reserve Bank Trade Weighted Dollar Index. T-statistics are heteroskedasticity consistent. ***
Significant at the 1% level, ** Significant at the 5% level and * Significant at the 10% level.

222
 The results obtained for individual funds and funds of hedge funds are in line with

those obtained on a global level. Individual funds outperform and create significant alpha

in almost all cases. The spreads of the alphas reported are in line with the one obtained

in the previous table. 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, the bottom one offering a significantly higher alpha (-0.49%

per month) but at a cost of higher volatility. 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.

223
4.4 Persistence in Performance – Other Measures

Absolute performance is the most important element of the hedge fund industry.

However, some strategies are very risky whereas others attempt to offer stable returns.

Our analysis indicates that some relatively low volatility hedge funds consistently and

significantly outperform classical equity and bond indices over time. In order to go

further in the analysis, we perform the same analysis with other ways of classifying funds

into deciles. We use the following measures as classification parameters:

1. Sharpe ratio57

2. Standard deviation58

3. Alpha

4. Beta

5. Skewness/Kurtosis

6. An adapted Sharpe ratio

57
Practically, we use the Sharpe score (ratio of mean returns over standard deviation). 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.
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. The corresponding Tables are available upon request.

224
 Before using these other ways of classifying hedge funds based on past

performance, we have to check for the risk of disaggregation. Disaggregation can happen

if the deciles are not homogeneous that is when funds of the same strategies are

grouped in the same decile. We check this element and report the results for alpha and

beta in Figure 5 to Figure 8. The first two figures report the repartition of decile per

strategy. There are for example 3% of distressed funds in the global database. Figure 5

shows that there are some variations in the repartition of decile per strategy over time

but these variations are not large. We are only talking about some percentages of

differences.

The differences are reported in Figure 6. Figure 6 reports the difference between

the percentage of funds of each strategy and in each decile. Consider the following

example. There are 3% of distressed funds in the global database. Decile 1 has on

average 3.9% of distressed funds. The difference, 0.9%, 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). 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. 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.

225
Figure 5: Decile Repartition per Strategy (Ranking base on

Alpha)

35%

30%

25%

20%

15%

FOF
NOSTRAT
OPTION 10%
CURRENCY
MKT TIMING
LGLEV
SHORT
SECTOR 5%
EQ. HEDGE
MKT NTL
GLOBAL
MACRO 0%
GL EMER
ED
D10
D9
D8

RA
D7
D6
D5

DS
D4
D3
D2
D1
DB

226
Figure 6: Strategy Repartition in the Alpha Ranking

12%

10%

8%

6%

4%

2%

0%

-2%

-4%
NOSTRAT
CURRENCY -6%
LGLEV
SECTOR -8%

MKT NTL -10%

MACRO
-12%
ED
D10
D9
D8
D7
D6

DS
D5
D4
D3
D2
D1

227
Figure 7: Decile Repartition per Strategy based on Beta

35%

30%

25%

20%

15%

FOF
NO STRAT
OPTION 10%
CURRENCY
MKT TIMING
LG LEV
SHORT
SECTOR 5%
EQ. HEDGE
MKT NTL
GLOBAL
MACRO
GL EMER
0%
ED
D9 D10
RA
D6 D7 D8
DS D5
D2 D3 D4
DB D1

228
 Figure 8: Strategy Repartition in the Beta Ranking

12%

10%

8%

6%

4%

2%

0%

-2%

FOF -4%
NOSTRAT
OPTION
CURRENCY -6%
MKT TIMING
LGLEV
SHORT -8%
SECTOR
EQ. HEDGE
MKT NTL -10%
GLOBAL
MACRO
GL EMER -12%
ED
D10
D9

RA
D8
D7
D6

DS
D5
D4
D3
D2
D1

These figures report the decile repartition between hedge fund strategies. Figure CCC report

the repatition in deciles based on alpha, Figure CCC+1 reports the difference with the global

database. Figure CCC and CCC report the same information when funds are classified in

deciles on the basis of their past beta.

229
 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. In the present case, these 14% are ranked in middle decile.

Figure 5 and Figure 6 lead us to conclude that there is no risk of aggregation when alpha

is used as a ranking measure.

Figure 7 and Figure 8 show the same information when the ranking used to

construct the deciles is based on beta. Results are a little more mitigated even if we

remain at relatively low variations. Funds of funds tend not to be in the top decile and

global funds tend not to be in the bottom deciles. These results comfort us in the use of

beta as a ranking tool for hedge funds.

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. First, as a measure

based on a distribution of returns, the omega59 needs a relatively long dataset to be

estimated. 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, we

cannot use the omega. 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.

59
The value of the Omega function at level r is the ratio of probability weighted gains relative to r, to

probability weighted losses relative to r. The Omega is like a return distribution that includes higher

moments. 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.

230
 The same issue emerges with the Sortino ratio60. By definition, hedge funds aim

at offering positive returns irrespective of the market. On a three year period, most fund

will or at least should suffer only a few months down. Because of this, our estimation of

the negative standard deviation will be unstable if we consider 1 or 3 years of data.

Another time, the data available do not allow us to cover 5 or 10 years of data to classify

funds before analysis the persistence.

Moreover, 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. 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. In addition, 1075 funds out

of theses 2576 funds have been dissolved before the end of the period, 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. Three years of data seems to be the best

tradeoff between stable statistics and the length of the period under review.

60
The Sortino ratio is similar to the Sharpe Ratio, except that instead of using standard deviation as

the denominator, it uses Downside Deviation.

231
 Sharpe score

Many investors investigating hedge funds focus on their risk/return

characteristics. Such investors pay a particular attention to the trade-off between the risk

and the returns offered. The Sharpe score – defined as the ratio of the average return to

the standard deviation – aims at taking risk and returns into account. It is reasonable to

use this measure while analysing the persistence in hedge fund returns. 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. Investors want to see if hedge funds are able to produce consistent returns over

time. Various studies used the Sharpe ratio while analysing hedge fund returns (see for

example, Ackermann et al., 1999; Liang, 1999 as well as Amin and Kat, 2003a) and

most of them stressed the shortcut of this measure for alternative strategies like those

applied by hedge fund managers (see for example, Amenc et al., 2003 and Goetzmann

et. al., 2003). Nevertheless, the use of this measure remains very predominant in the

industry despite its weaknesses.

The first three columns of Table 32 report the return, the risk and the average

Sharpe score of each decile. Decile D1 contains the funds having the lowest Sharpe score

calculated using three years of data, while decile D10 includes the funds offering the

most attractive risk-return trade-off over the three years used to classify funds. Funds

included in decile D1 offer an average monthly performance of 0.7% with a monthly

standard deviation of 2.79% against a mean monthly return of 0.93% for the best

performing funds. The Sharpe score decreases slightly between decile D2 and decile D5

and increases monotically thereafter to reach 0.93 for decile D10. 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.

232
 The alpha column indicates that there is persistence for the previous year’s top

funds. Funds classified in deciles D8 to D10 create alpha. The factor exposure indicates

that top risk-return performers have had a limited exposure to the US equity market,

they have been value-oriented and that they have been exposed to the high yield

market. On the other side, underperforming funds have not been exposed to the US

equity market but have been exposed to the world excluding US. They also have been

momentum contrarian, short world government bonds and short the USD.

The spreads of the alphas reported are not significant but for the spread between

the most attractive funds decile D9 and decile D10. 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, you would not only offer significant positive alpha but you would also

significantly outperform significantly the next 10%. 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). The results indicate that even if there

are around 30% of the funds that offer significant alpha based on the Sharpe ratio, the

top one are apart from the others.

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.

233
 1. Standard Deviation

Some investors focus on volatility using hedge funds as pure risk diversifiers. In

this case, investors may use hedge funds instead of bonds in their portfolio and the most

important aspect to them is to limit the volatility. We perform the same analysis by

focusing purely on volatility. Every year we classify funds on the basis of the standard

deviation of their previous three years of returns.

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. Stated differently, do volatile hedge funds consistently and significantly out-

or under-perform less volatile hedge funds? The answer is in Table 33.

The mean and standard deviation columns of Panel A indicate that returns

increase with volatility for low volatility funds, but that the trend is unclear for more

volatile funds (decile D5 to D10). The alpha column confirms that low volatility funds

create significant alpha and that high volatility funds do not. Interestingly, the alpha

created by decile D5 is strongly significant (as for example, decile D1), while being much

more volatile – indicating that higher volatility funds do not necessarily lack of

consistency in performance. These results enable us to divide the hedge fund industry

into three segments:

1) Low volatility funds that offer persistent over-performance;

2) Low/medium volatility funds that offer persistent over-performance;

3) High volatility funds that do not offer persistent over-performance.

234
 This result leads to the conclusion that low volatility funds tend to consistently and

significantly outperform over time. Low volatile funds that cannot create consistent

performance tend to become more volatile and switch from segment one to two or three.

Higher volatility funds that offer consistent returns will have a relatively stable volatility

(since returns are relatively stable) and consequently remain in segment two, while

volatile and inconsistent managers fall in segment three.

In terms of exposures, alpha creators again exhibit slight exposures to the US

equity markets and they have been long high yield. Unlike other deciles they also have

not been momentum players. This result confirms that it is not sufficient to be a

momentum player to make money. The r-squared are very high.

The spreads of the alphas reported show the same pattern as with the Sharpe ratio.

Selecting low volatility funds enable investors to create significant and consistent alpha

over time but if you select the lowest funds that are available, you will not only offer

significant alpha but also significantly outperform the next alpha creators. The results

indicate that even if there are between 30% and 50% of the individual funds that

outperform, the top one are apart from the others.

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.

As before, 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.

235
Table 32: Persistence in Performance Analysis Based on the
Sharpe score (1/1997-12/2002)

Panel A: Multi-factor estimation - Global Funds classified on their Sharpe Score (1/1997-12/2002)

Sharpe MSCI w ex
Mean Std dev. Alpha S&P SMB HML
Score US

D1 0,70 2,79 0,25 0,60 * 0,03 0,23 ** 0,06 -0,01

D2 0,82 3,24 0,25 0,47 0,33 *** 0,10 0,11 ** -0,03

D3 0,72 3,62 0,20 0,15 0,51 *** 0,04 0,16 *** 0,03

D4 0,81 3,49 0,23 0,27 0,48 *** 0,04 0,21 *** 0,03

D5 0,77 3,12 0,25 0,25 0,41 *** 0,03 0,21 *** -0,01

D6 0,82 2,91 0,28 0,16 0,44 *** -0,04 0,19 *** 0,03

D7 0,87 2,58 0,34 0,23 * 0,39 ** -0,02 0,19 *** 0,06 **

D8 0,89 2,16 0,41 0,25 ** 0,30 *** -0,02 0,16 *** 0,07 ***

D9 0,89 1,58 0,56 0,37 *** 0,16 *** 0,01 0,12 *** 0,04

D10 0,93 1,00 0,93 0,49 *** 0,08 *** 0,01 0,06 *** 0,02

D1a 0,40 3,41 0,12 0,25 0,08 0,27 * 0,11 * 0,07

D1b 0,74 2,98 0,25 0,59 0,01 0,20 -0,01 -0,02

D1c 0,71 3,10 0,23 0,66 ** 0,02 0,28 *** 0,08 ** -0,05

D10a 0,94 1,31 0,71 0,45 *** 0,12 *** -0,01 0,10 *** 0,03

D10b 0,93 0,94 0,99 0,52 *** 0,06 *** 0,03 0,06 *** 0,01

D10c 0,86 0,71 1,20 0,47 *** 0,01 0,03 0,03 ** 0,02

Spread 1-10 -0,23 2,42 -0,09 -0,02 -0,05 0,21 ** -0,01 -0,03

Spread 1-2 -0,12 1,68 -0,07 -0,01 -0,30 *** 0,13 ** -0,05 *** 0,02

Spread 9-10 -0,04 0,71 -0,06 -0,46 *** 0,08 *** 0,00 0,06 *** 0,01

Spread 1a-10c -0,46 3,26 -0,14 -0,39 0,07 0,24 0,09 0,05

236
Panel A (continued): Multi-factor estimation - Global Funds classified on their Sharpe Score (1/1997-
12/2002)

PR1YR WGBI EMBI HY GSCI DOLLAR R²

D1 -0,09 ** -0,38 *** 0,06 0,15 * 0,03 -0,56 ** 0,60

D2 0,01 -0,19 0,07 0,09 0,02 -0,22 0,72

D3 0,06 * -0,15 0,14 *** 0,10 -0,01 -0,28 * 0,87

D4 0,06 ** -0,12 0,09 *** 0,07 0,03 -0,26 * 0,89

D5 0,07 *** -0,11 0,07 *** 0,04 0,02 -0,08 0,89

D6 0,12 *** -0,06 0,09 *** 0,09 0,02 -0,10 0,90

D7 0,11 *** -0,04 0,06 *** 0,04 0,03 -0,13 0,90

D8 0,11 *** -0,05 0,06 *** 0,09 * 0,02 -0,09 0,86

D9 0,07 *** -0,10 * 0,06 *** 0,10 ** 0,03 -0,01 0,81

D10 0,04 *** -0,07 0,05 *** 0,09 * 0,02 * 0,06 0,71

D1a -0,11 ** -0,26 0,02 -0,04 0,03 -0,65 ** 0,43

D1b -0,05 -0,52 *** 0,02 0,14 0,00 -0,55 ** 0,36

D1c -0,10 *** -0,36 *** 0,12 *** 0,15 * 0,06 ** -0,48 ** 0,69

D10a 0,06 *** -0,07 0,05 *** 0,10 0,04 ** 0,07 0,74

D10b 0,03 *** -0,06 0,04 *** 0,08 ** 0,02 0,05 0,75

D10c 0,03 ** -0,09 *** 0,03 *** 0,10 * 0,01 0,04 0,51

Spread 1-10 -0,13 *** -0,30 ** 0,01 0,06 0,01 -0,59 ** 0,42

Spread 1-2 -0,10 *** -0,19 ** -0,01 0,06 0,01 -0,33 * 0,51

Spread 9-10 0,03 *** -0,03 0,02 0,01 0,01 -0,11 ** 0,68

Spread 1a-10c -0,14 *** -0,13 -0,01 -0,13 0,02 -0,61 ** 0,27

This Table presents the results of the persistence in performance estimation of the combined Model for the
January 1994-December 2002 period. We report the OLS estimators for equally weighted portfolios per investment
strategy, sub-strategy and for all funds. Each year, every fund is classified into a decile on the basis of its Sharpe
score over the past three years, with the Sharpe score defined as the ratio of performance to volatility. S&P stands
for the market factor, MSWXUSt = return of the MSCI World Index excluding US, SMBt = the factor-mimicking
portfolio for size (‘small minus big’), HMLt = the factor-mimicking portfolio for book-to-market equity (‘high minus
low’), PR1YRt = the factor-mimicking portfolio for the momentum effect, EMBIt = return of the JP Morgan Emerging
Market Bond Index; SWGBIt = return of the Salomon World Government Bond Index; HYt = return of the Lehman
High Yield Credit Bond Index, GSCIt = return of the Goldman Sachs Commodity Index, DOLLAR = return of the
Federal Reserve Bank Trade Weighted Dollar Index. T-statistics are heteroskedasticity consistent. *** Significant
at the 1% level, ** Significant at the 5% level and * Significant at the 10% level.

237
Table 33: Persistence in Performance Analysis Based on the

Standard Deviation (1/1997-12/2002)

Panel A: Multi-factor estimation - Global Funds classified on their standard deviation (1/1997-12/2002)
MSCI w ex
Mean Std dev. Alpha S&P SMB HML
US

D1 0,69 0,63 0,32 *** 0,01 0,02 0,02 0,01

D2 0,76 0,96 0,34 *** 0,07 *** 0,02 0,04 *** 0,04 ***

D3 0,77 1,17 0,33 *** 0,12 *** 0,02 0,06 *** 0,03 *

D4 0,91 1,62 0,41 *** 0,20 *** 0,02 0,09 *** 0,05

D5 0,97 2,28 0,41 *** 0,28 *** 0,06 0,12 *** 0,08 *

D6 0,85 2,58 0,27 0,36 *** 0,05 0,11 *** 0,09 **

D7 0,87 2,99 0,29 * 0,45 *** 0,03 0,16 *** 0,07 *

D8 0,91 3,45 0,33 * 0,46 *** 0,02 0,22 *** 0,01

D9 0,82 4,21 0,22 0,59 *** 0,00 0,30 *** -0,01

D10 0,70 6,56 0,19 0,77 *** 0,06 0,47 *** -0,14

D1a 0,65 0,52 0,30 *** 0,00 0,02 0,01 0,01 **

D1b 0,69 0,74 0,31 *** 0,02 0,02 0,01 0,02 *

D1c 0,71 0,72 0,34 *** 0,01 0,04 *** 0,02 ** 0,01

D10a 0,68 6,21 0,17 0,67 *** 0,09 0,47 *** -0,15

D10b 0,71 6,27 0,31 0,58 *** 0,13 0,46 *** -0,16

D10c 0,67 7,94 0,11 1,04 *** 0,03 0,45 *** -0,13

Spread 1-10 -0,01 6,23 0,06 -0,76 *** -0,04 -0,45 *** 0,16

Spread 1-2 -0,07 0,44 -0,36 *** -0,07 *** 0,01 -0,03 *** -0,03 ***

Spread 9-10 0,12 2,81 -0,11 -0,18 -0,06 -0,17 *** 0,13

Spread 1a-10c -0,02 7,70 -0,04 -1,05 *** -0,01 -0,44 *** 0,14

238
Panel A (continued): Multi-factor estimation - Global Funds classified on their standard deviation (1/1997-
12/2002)

PR1YR WGBI EMBI HY GSCI DOLLAR R²

D1 0,02 -0,09 0,03 *** 0,10 *** 0,01 -0,02 0,61

D2 0,02 -0,11 0,04 *** 0,12 *** 0,02 -0,08 0,67

D3 0,04 *** -0,06 0,05 *** 0,12 *** 0,01 -0,01 0,76

D4 0,05 *** -0,07 0,07 *** 0,07 * 0,01 -0,12 0,77

D5 0,09 *** -0,11 0,07 *** 0,10 * 0,01 -0,16 0,81

D6 0,08 *** -0,09 0,07 *** 0,08 0,01 -0,24 0,82

D7 0,07 *** -0,07 0,07 *** 0,07 0,01 -0,15 0,89

D8 0,08 *** -0,14 0,12 *** 0,11 0,02 -0,30 0,89

D9 0,08 *** -0,15 0,09 *** 0,03 0,03 -0,32 0,90

D10 0,10 * -0,26 0,13 * 0,07 0,08 0,05 0,85

D1a 0,00 -0,06 0,02 0,11 *** 0,01 0,00 0,59

D1b 0,02 -0,10 0,04 *** 0,12 ** 0,02 0,01 0,53

D1c 0,02 -0,12 0,03 *** 0,10 *** 0,01 -0,06 0,63

D10a 0,12 *** -0,14 0,16 *** 0,12 0,07 0,23 0,85

D10b 0,12 -0,28 0,04 0,21 0,06 0,07 0,79

D10c 0,07 -0,51 0,20 * -0,17 0,11 -0,28 0,80

Spread 1-10 -0,09 0,17 -0,10 0,04 -0,07 -0,06 0,82

Spread 1-2 0,00 0,01 -0,01 -0,01 -0,01 0,02 0,53

Spread 9-10 -0,02 0,13 -0,04 -0,04 -0,06 -0,35 0,51

Spread 1a-10c -0,07 0,45 -0,18 * 0,28 -0,11 0,25 0,75

This Table presents the results of the persistence in performance estimation of the combined Model for the
January 1994-December 2002 period. We report the OLS estimators for equally weighted portfolios per
investment strategy, sub-strategy and for all funds. Each year, every fund is classified into a decile on the basis of
its standard deviation over the previous three years. S&P stands for the market factor, MSWXUSt = return of the
MSCI World Index excluding US, SMBt = the factor-mimicking portfolio for size (‘small minus big’), HMLt = the
factor-mimicking portfolio for book-to-market equity (‘high minus low’), PR1YRt = the factor-mimicking portfolio for
the momentum effect, EMBIt = return of the JP Morgan Emerging Market Bond Index; SWGBIt = return of the
Salomon World Government Bond Index; HYt = return of the Lehman High Yield Credit Bond Index, GSCIt =
return of the Goldman Sachs Commodity Index, DOLLAR = return of the Federal Reserve Bank Trade Weighted
Dollar Index. T-statistics are heteroskedasticity consistent. *** Significant at the 1% level, ** Significant at the 5%
level and * Significant at the 10% level.

239
 2. Alpha

Hedge funds aim at offering absolute returns in any market environment.

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. Table 34 reports the decomposition of

hedge fund returns classified on the basis of the alpha created over a three-year period.

The current analysis is based on the following statement: do managers that create higher

alpha over a period persistently outperform the equity, bond and commodity markets

over the next period?

The mean return decreases between decile D1 and D10, the standard deviation

decreases between decile D1 and D6 before increasing again until decile D10. The trend

in alpha is not clear, but significant for the middle deciles D4 to D8. The volatility column

confirms once again that low volatility funds can create significant and consistent value

over time. 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.

The exposures reported are in line with the ones previously obtained. The only

major difference is that alpha creators are momentum players according to those results.

This result confirms the instability of this factor in distinguishing pure alpha creators

since depending on the way of classifying funds, pure alpha creators will or will not be

considered momentum players.

240
 3. Beta

Table 35 reports the results obtained by classifying funds on the basis of their

beta (high beta funds in decile D1). High beta funds are the most volatile and they do not

outperform classical markets over time. Almost all deciles from deciles D5 to D10

significantly and consistently create value. The factor exposures indicate that alpha

creators have not all been long US equities (decile D10 is even net short the U.S. equity

market but not significantly) but that these funds have been value oriented and exposed

to the high yield market.. The r-squared obtained are particularly high, with a maximum

of 0.94.

The spreads report very interesting results. The spread between the high beta

funds (decile D1) and the low beta funds (decile D10) is significantly positive at 0.92

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. Spread 9-10

is also significant but negative. This result indicates another time that in the alpha

producers, there is a difference between the good and the top ones. The funds that offer

the highest alpha are much better than the one outperform the underperforming funds.

As one would expect, pure hedge funds should have a limited exposure to the

equity market. Our results suggest that low beta funds that also a limited volatility can

significantly and consistently create value over time.

241
 4. Skewness and Kurtosis

Investors like positive skewness but do not like fat tails. Panel A of Table 36

reports the results obtained by classifying funds on the basis of their skewness (the lower

the skewness, the lower the decile). The first interest of Table 36 is that the volatilities

reported in the second column are relatively stable over the 10 deciles. This result is very

different from what we obtained previously. Second, in each case we report the skewness

and the kurtosis because our results indicate that there is a direct link between these two

measures. In Panel A, the skewness increases from decile D1 to decile D10. This means

that the distribution goes from negatively skewed to positively skewed. The kurtosis

reported in the next column goes from high values for negatively skewed returns, to low

values for unskewed return distributions. In short, skewed distributions have fat tails.

This relation is even stronger when considering negatively skewed distributions.

Panel B reports the same analysis and focuses on the kurtosis. It confirms these findings

since funds with fat tails also have the highest skewness. Funds tend to have fat tails on

the left-hand side of their mean. Volatility is on the downside. Finally, the alpha column

is not easy to interpret. In both cases, performing deciles are spread out. Since there is

no formal difference in the volatility of the deciles considered, 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.

242
Table 34: Persistence in Performance Analysis Based on the
Alpha (1/1997-12/2002)

Panel A: Multi-factor estimation - Global Funds classified on their alpha (1/1997-12/2002)

MSCI w ex
Mean Std dev. Alpha S&P SMB HML
US

D1 0,93 5,47 0,19 0,66 *** -0,03 0,22 *** -0,03

D2 0,92 3,73 0,43 0,46 *** 0,06 0,12 *** -0,03

D3 0,82 2,94 0,28 0,36 *** 0,05 0,12 *** -0,01

D4 0,87 2,17 0,50 ** 0,29 *** 0,03 0,10 *** 0,00

D5 0,79 1,89 0,44 ** 0,23 *** 0,05 0,10 *** -0,01

D6 0,84 1,76 0,51 *** 0,21 *** 0,06 0,10 *** 0,00

D7 0,78 1,78 0,36 *** 0,25 *** 0,03 0,14 *** 0,03

D8 0,77 1,88 0,26 *** 0,24 *** 0,02 0,15 *** 0,04 *

D9 0,79 2,22 0,17 0,30 *** 0,00 0,18 *** 0,09 ***

D10 0,69 3,10 -0,05 0,34 *** 0,01 0,30 *** 0,11

D1a 0,58 6,44 -0,53 0,90 *** -0,12 0,11 0,13

D1b 1,08 5,56 0,25 0,59 *** 0,02 0,26 *** -0,04

D1c 0,81 4,67 0,21 0,54 *** 0,02 0,18 *** -0,07

D10a 0,81 2,53 0,18 0,33 *** -0,05 0,22 *** 0,10 ***

D10b 0,65 3,40 -0,10 0,33 *** 0,03 0,31 *** 0,11

D10c 0,75 4,01 -0,10 0,37 ** 0,04 0,39 *** 0,12

Spread 1-10 0,24 3,98 0,11 0,33 *** -0,01 -0,08 -0,20 *

Spread 1-2 0,02 2,23 -0,17 0,15 0,01 0,08 * -0,08

Spread 9-10 0,10 1,36 -0,19 -0,02 -0,03 -0,12 *** -0,01

Spread 1a-10c -0,17 5,85 -0,15 0,51 * -0,07 -0,27 -0,04

243
Panel A (continued): Multi-factor estimation - Global Funds classified on their alpha (1/1997-12/2002)

PR1YR WGBI EMBI HY GSCI DOLLAR R²

D1 -0,01 -0,06 0,14 *** 0,16 0,01 -0,67 0,81

D2 0,08 * -0,11 0,08 ** 0,11 0,03 -0,24 0,81

D3 0,06 * -0,11 0,10 *** 0,07 0,01 -0,23 0,81

D4 0,07 ** -0,09 0,08 *** 0,08 0,03 -0,18 0,75

D5 0,09 *** -0,15 0,07 ** 0,10 * 0,02 -0,13 0,74

D6 0,06 *** -0,13 0,07 *** 0,10 * 0,02 -0,06 0,79

D7 0,07 *** -0,10 0,05 *** 0,07 0,02 -0,03 0,85

D8 0,07 *** -0,13 0,04 *** 0,05 0,02 0,01 0,89

D9 0,04 *** -0,08 0,07 *** 0,06 0,02 -0,02 0,87

D10 0,07 *** -0,13 0,07 0,07 0,04 0,00 0,78

D1a 0,08 ** -0,33 0,21 ** 0,13 -0,07 -1,57 0,76

D1b -0,02 -0,05 0,16 ** 0,12 -0,02 -0,62 0,76

D1c 0,06 0,04 0,12 *** 0,13 0,07 * -0,10 0,76

D10a 0,06 *** -0,12 0,04 0,11 0,02 -0,25 0,79

D10b 0,04 -0,12 0,06 0,13 0,04 0,15 0,77

D10c 0,12 ** -0,13 0,10 -0,02 0,06 0,19 0,64

Spread 1-10 -0,04 0,02 0,08 0,07 -0,05 -0,62 ** 0,41

Spread 1-2 -0,04 -0,11 0,06 0,02 -0,02 -0,37 ** 0,40

Spread 9-10 -0,03 0,09 0,00 0,00 -0,04 -0,05 0,31

Spread 1a-10c -0,01 -0,27 0,09 0,13 -0,15 -1,51 ** 0,28

This Table presents the results of the persistence in performance estimation of the combined Model for the
January 1994-December 2002 period. Each year, 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. S&P stands for the market factor,
MSWXUSt = return of the MSCI World Index excluding US, SMBt = the factor-mimicking portfolio for size
(‘small minus big’), HMLt = the factor-mimicking portfolio for book-to-market equity (‘high minus low’), PR1YRt
= the factor-mimicking portfolio for the momentum effect, EMBIt = return of the JP Morgan Emerging Market
Bond Index; SWGBIt = return of the Salomon World Government Bond Index; HYt = return of the Lehman
High Yield Credit Bond Index, GSCIt = return of the Goldman Sachs Commodity Index, DOLLAR = return of
the Federal Reserve Bank Trade Weighted Dollar Index. T-statistics are heteroskedasticity consistent. ***
Significant at the 1% level, ** Significant at the 5% level and * Significant at the 10% level.

244
Table 35: Persistence in Performance Analysis Based on Beta

(1/1997-12/2002)

Panel A: Multi-factor estimation - Global Funds classified on their beta (1/1997-12/2002)

MSCI w ex
Mean Std dev. Alpha S&P SMB HML
US
D1 0,56 6,24 0,30 1,06 *** -0,02 0,48 *** -0,19

D2 0,76 3,80 0,09 0,71 ** 0,02 0,28 ** 0,01

D3 0,61 3,34 0,24 0,49 *** 0,07 0,24 ** 0,08 *

D4 0,99 2,61 0,18 0,42 *** 0,05 0,19 *** 0,06

D5 0,74 2,31 0,39 ** 0,32 *** 0,05 0,11 *** 0,08 *

D6 0,84 1,64 0,31 * 0,21 *** 0,06 0,10 *** 0,03

D7 0,79 1,34 0,36 *** 0,11 *** 0,05 * 0,06 *** 0,03

D8 0,81 1,12 0,36 *** 0,08 *** 0,02 0,07 *** 0,06 ***

D9 0,69 0,88 0,39 *** 0,02 0,02 0,03 *** 0,02 *

D10 0,79 0,90 0,46 *** -0,18 0,02 -0,03 0,09 ***

D1a 1,14 7,55 0,30 1,36 *** -0,10 0,45 *** -0,18

D1b 0,27 6,07 0,36 0,94 *** 0,06 0,50 *** -0,17

D1c 0,29 5,04 0,07 0,91 ** -0,04 0,41 *** -0,13

D10a 0,69 0,85 0,30 *** 0,03 0,01 0,05 *** 0,05 **

D10b 0,75 1,13 0,44 *** -0,04 0,10 -0,01 0,08 *

D10c 1,03 2,35 0,72 *** -0,60 0,00 -0,13 0,13

Spread 1-10 -0,22 6,43 -0,92 *** 1,23 *** -0,04 0,51 *** -0,29 **

Spread 1-2 -0,19 3,05 -0,31 0,35 *** -0,04 0,21 *** -0,21 ***

Spread 9-10 -0,10 1,00 -0,52 *** 0,19 *** 0,00 0,06 *** -0,07 **

Spread 1a-10c 0,12 8,82 -1,21 1,94 *** -0,10 0,59 *** -0,32

245
Panel A (continued): Multi-factor estimation - Global Funds classified on their beta (1/1997-12/2002)

PR1YR WGBI EMBI HY GSCI DOLLAR R²

D1 0,11 ** -0,18 0,11 *** 0,04 0,02 -0,05 0,92

D2 0,07 ** -0,14 0,11 *** 0,07 0,05 * -0,28 0,94

D3 0,07 *** -0,18 0,08 ** 0,06 0,02 -0,25 0,91

D4 0,09 *** -0,14 0,09 *** 0,07 0,03 -0,19 0,87

D5 0,07 ** -0,12 0,08 *** 0,11 * 0,02 -0,18 0,78

D6 0,07 *** -0,12 0,09 *** 0,14 *** 0,03 -0,08 0,80

D7 0,05 *** -0,13 0,04 ** 0,10 *** 0,02 -0,13 0,74

D8 0,03 *** -0,08 0,06 *** 0,08 *** 0,01 -0,04 0,72

D9 0,02 ** -0,07 0,05 *** 0,11 *** 0,00 0,02 0,59

D10 0,01 -0,05 0,04 ** 0,09 * 0,01 -0,12 0,68

D1a 0,30 *** -0,36 0,13 0,04 -0,04 -0,28 0,84

D1b 0,11 *** -0,29 0,10 ** 0,05 0,04 -0,24 0,92

D1c 0,09 *** -0,05 0,11 *** 0,08 0,02 0,05 0,95

D10a 0,03 ** -0,08 0,02 * 0,06 0,01 -0,06 0,36

D10b 0,04 * -0,12 0,04 ** 0,09 0,00 -0,26 0,25

D10c -0,05 -0,02 0,03 0,13 * 0,00 -0,03 0,82

Spread 1-10 0,11 ** -0,12 0,08 * -0,04 0,01 0,01 0,92

Spread 1-2 0,05 -0,02 0,01 -0,02 -0,03 0,21 0,75

Spread 9-10 0,02 -0,03 0,01 0,02 -0,01 0,07 0,78

Spread 1a-10c 0,35 *** -0,32 0,11 -0,08 -0,04 -0,28 0,87

This Table presents the results of the persistence in performance estimation of the combined Model for the January
1994-December 2002 period. Each year, 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. S&P stands for the market factor, MSWXUSt = return of the MSCI
World Index excluding US, SMBt = the factor-mimicking portfolio for size (‘small minus big’), HMLt = the factor-
mimicking portfolio for book-to-market equity (‘high minus low’), PR1YRt = the factor-mimicking portfolio for the
momentum effect, EMBIt = return of the JP Morgan Emerging Market Bond Index; SWGBIt = return of the Salomon
World Government Bond Index; HYt = return of the Lehman High Yield Credit Bond Index, GSCIt = return of the
Goldman Sachs Commodity Index, DOLLAR = return of the Federal Reserve Bank Trade Weighted Dollar Index. T-
statistics are heteroskedasticity consistent. *** Significant at the 1% level, ** Significant at the 5% level and *
Significant at the 10% level.

246
Table 36: Persistence in Performance Analysis Based on the
Skewness and Kurtosis (1/1997-12/2002)

Panel A: Multi-factor estimation - Global Funds classified on their skewness (1/1997-12/2002)

MSCI w ex
Mean Std dev. Skewness Kurtosis Alpha S&P SMB
US

D1 0,79 1,98 -2,32 12,24 0,39 ** 0,16 *** 0,06 * 0,05 **

D2 0,75 2,26 -1,32 5,71 0,27 ** 0,31 *** 0,03 0,04 *

D3 0,96 2,65 -0,26 1,34 0,5 *** 0,37 *** -0,01 0,14 ***

D4 0,82 2,66 -0,54 2,51 0,28 0,41 *** -0,02 0,13 ***

D5 0,83 2,85 -0,03 2,36 0,38 ** 0,32 *** 0,09 0,15 ***

D6 0,94 2,9 0,24 1,59 0,39 * 0,38 *** 0,02 0,18 ***

D7 0,76 2,88 0,29 2,1 0,19 0,37 *** 0,03 0,22 ***

D8 0,76 2,65 0,35 1,02 0,25 0,32 *** 0,03 0,18 ***

D9 0,83 2,28 0,42 3,32 0,26 * 0,25 *** 0,03 0,19 ***

D10 0,92 2,47 0,89 3,31 0,33 ** 0,28 *** 0,04 0,21 ***

247
Panel A (continued): Multi-factor estimation - Global Funds classified on their skewness (1/1997-12/2002)

HML PR1YR WGBI EMBI HY GSCI DOLLAR R²

D1 0,05 0,01 -0,22 ** 0,07 *** 0,16 ** 0,06 ** -0,36 ** 0,68

D2 0,06 *** 0,04 -0,09 0,08 *** 0,15 *** 0,01 -0,23 *** 0,83

D3 0,01 0,02 -0,04 0,06 *** 0,11 * 0,01 -0,12 0,85

D4 0,03 0,05 ** -0,05 0,08 *** 0,09 0,01 -0,2 * 0,86

D5 0 0,06 * -0,16 *** 0,08 *** 0,11 * 0,03 -0,2 0,86

D6 0 0,09 *** -0,11 0,08 *** 0,06 0 -0,17 0,85

D7 0,03 0,09 *** -0,09 0,07 *** 0,02 0,03 -0,09 0,87

D8 -0,01 0,07 *** -0,05 0,1 *** 0,07 0,03 0,04 0,83

D9 0,05 0,09 *** -0,17 ** 0,05 *** 0,08 0,01 -0,1 0,85

D10 0,02 0,12 *** -0,16 ** 0,08 *** 0 0,02 0,04 0,87

248
Panel B: Multi-factor estimation - Global Funds classified on their kurtosis (1/1997-12/2002)

Mean Std dev. Skewness Kurtosis Alpha S&P MSCI w ex SMB

D1 0,96 2,78 0,16 1,41 0,44 *** 0,36 *** 0,03 0,2 ***

D2 0,77 2,76 0,25 1,84 0,24 0,37 *** 0,01 0,15 ***

D3 0,92 2,73 0,3 2,55 0,43 ** 0,33 *** 0,08 0,14 ***

D4 0,85 2,68 0,06 0,36 0,37 ** 0,36 *** 0,04 0,18 ***

D5 0,82 2,8 -0,2 1,86 0,3 * 0,32 *** 0,05 0,22 **

D6 0,75 2,62 -0,72 3,42 0,18 0,39 *** -0,05 0,16 ***

D7 0,87 2,47 0,39 2,1 0,37 ** 0,26 *** 0,06 0,15 ***

D8 0,77 2,26 -0,86 3,9 0,21 0,34 *** -0,01 0,11 ***

D9 0,88 2,14 -0,82 4,03 0,37 *** 0,27 *** 0,04 0,11 ***

D10 0,75 1,93 -0,89 4,21 0,27 * 0,19 *** 0,04 0,1 ***

249
Panel B (continued): Multi-factor estimation - Global Funds classified on their kurtosis (1/1997-12/2002)

HML PR1YR WGBI EMBI HY GSCI DOLLAR R²

D1 0 0,07 *** -0,04 0,06 *** 0,04 0,03 -0,02 0,86

D2 0 0,07 ** -0,09 0,09 *** 0,07 0,01 -0,19 0,84

D3 0,01 0,07 ** -0,12 * 0,08 *** 0,06 0,02 -0,09 0,84

D4 0,04 0,03 -0,06 0,06 *** 0,07 0,01 -0,02 0,88

D5 0,02 0,07 *** -0,13 0,07 *** 0,07 0,02 -0,18 0,88

D6 0,03 0,07 *** -0,08 0,09 *** 0,1 0,02 -0,24 * 0,86

D7 0 0,09 *** -0,18 *** 0,07 *** 0,08 0,02 -0,14 0,86

D8 0,05 ** 0,07 *** -0,07 0,09 *** 0,11 ** 0,02 -0,18 0,86

D9 0,06 *** 0,06 *** -0,13 * 0,06 *** 0,1 * 0,04 * -0,11 0,83

D10 0,03 * 0,06 *** -0,2 *** 0,09 *** 0,13 *** 0,03 -0,22 * 0,8

This Table presents the results of the persistence in performance estimation of the combined Model for the
January 1994-December 2002 period. We report the OLS estimators for equally weighted portfolios per
investment strategy, sub-strategy and for all funds. Each year, every fund is classified into a decile on the
basis of its skewness (Panel A) and kurtosis (Panel B) over the previous three years. S&P stands for the
market factor, MSWXUSt = return of the MSCI World Index excluding US, SMBt = the factor-mimicking
portfolio for size (‘small minus big’), HMLt = the factor-mimicking portfolio for book-to-market equity (‘high
minus low’), PR1YRt = the factor-mimicking portfolio for the momentum effect, EMBIt = return of the JP Morgan
Emerging Market Bond Index; SWGBIt = return of the Salomon World Government Bond Index; HYt = return of
the Lehman High Yield Credit Bond Index, GSCIt = return of the Goldman Sachs Commodity Index, DOLLAR =
return of the Federal Reserve Bank Trade Weighted Dollar Index. T-statistics are heteroskedasticity
consistent. *** Significant at the 1% level, ** Significant at the 5% level and * Significant at the 10% level.

250
 5. An Adapted Sharpe Ratio

As a final and combined risk measure, we use the new risk measure developed

by Capocci et al. (2007)61. This measure

R x is defined as:

1 1 1
R x = V x − CS x + C 2 K x
2 6 24 (11)

I
C≡c
Where
W0 is the product of the (intrinsic) risk perception coefficient, c,

and the (wealth-related) proportion of risky investment to initial wealth

I / W0 .

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. As stated previously

investors do not like volatility, negative skewness and positive kurtosis. In other

words, this risk measure accounts for volatility, asymmetry and the fat tails. The level

of C depends on the appetite for risk of the investor62. Capocci et al. (2007) consider

three representative levels. We focus on average investors but results are in line for

dynamic and protective investors.

61
For more details, please see Part 3 of the Thesis.
62
Capocci et al. (2006) estimate C as 10 for a dynamic investor that does not put much weight on

the extreme risks, 35 for an average investor and 60 for a very protective investor who mostly fears

the tail event of a distribution.

251
 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. Such an adapted Sharpe score would be defined as mean return divided by

the new risk measure: R x as defined in (11)63.

Results are reported in Table 37 (the adapted Sharpe Score are multiplied by

100 to be comparable). The five first columns report the mean return, standard

deviation, skewness, kurtosis and adapted Sharpe for each decile and each spread.

Fund with the highest Sharpe score have been classified in the bottom deciles while

those with the lowest one in the top deciles. There is no clear trend in the adapted

Sharpe reported. More interestingly, 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. The first columns indicates that these deciles have

had a limited volatility, depending on the decile considered negative (decile D1) or

positive skewness (decile D2 and D3) and positive kurtosis. 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, with no momentum or contrarian bets,

that they have not been exposed to emerging market bond and that more globally, the

R² reported for them are lower than for the other deciles. The spreads help us to

understand why decile D1 is so different from decile D2 and D3.

These results indicate that when higher moments are considered results are

mitigated. 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.

63
An adapted version of the Sharpe ratio would be defined as the excess mean return divided by

the adapted risk measure.

252
 Spread 1-2 and spread 9-10 are significantly negative. A negative spread 1-2

indicates that there is a significant difference between the least attractive funds in

terms of adapted Sharpe ratio. The worst ones (decile D1) are much less attractive

than the next to worst one (decile D2). On the other side, the next to top one (decile

D9) also significantly underperform the top one (decile D10). Such a result is

astonishing because the extremes (decile D1 and D10) outperform the neighbour at

the two end of Table 37.

6. Other Measures – Conclusion

Table 38 summarises the persistence analysis. Panel A reports the alpha

column and Panel B the U.S. equity exposure. 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. However, a deeper analysis

indicates that lower volatility funds (classified in the middle deciles) do significantly

and consistently outperform the indices considered. This result is directly confirmed by

the Sharpe score, the pure volatility analysis, the alpha and the beta classifications.

Moreover, the skewness and kurtosis analysis showed that when funds cannot be

differentiated by their volatility, pure alpha creators cannot be identified. These results

provide a simple way to create pure alpha using hedge funds. Volatility and returns

should be considered together to easily identify funds that should outperform classical

markets in the future. 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.

253
Table 37: Persistence in Performance Analysis Based an
adapted Sharpe score (1/1997-12/2002)

Adapted
Mean Std dev. Skew. Kurtosis
Sharpe

D1 1,08 2,40 -0,40 3,80 0,50

D2 1,04 2,40 0,40 1,20 1,70

D3 0,81 1,80 0,10 2,10 0,80

D4 0,65 1,70 -0,60 2,30 0,50

D5 0,59 2,00 -0,70 2,50 0,40

D6 0,73 2,50 -0,20 1,10 1,20

D7 0,85 3,00 -0,10 2,10 0,80

D8 0,85 3,20 -0,10 2,40 0,70

D9 0,71 3,30 -0,10 2,30 0,60

D10 1,04 3,10 0,40 2,00 1,00

D1a 1,05 2,60 0,50 1,60 1,30

D1b 1,14 2,70 -1,40 9,50 0,20

D1c 1,10 2,10 -0,20 1,90 1,10

D10a 1,12 3,10 0,50 1,40 1,50

D10b 0,97 3,50 0,00 1,80 0,01

D10c 1,02 2,90 0,90 3,80 0,01

Spread 1-10 0,05 1,10 -0,50 2,00 0,00

Spread 1-2 0,05 1,60 0,90 4,10 0,00

Spread 9-10 -0,32 0,70 -0,50 0,10 -3,80

Spread 1a-10c 0,03 1,10 0,10 1,70 0,00

254
Panel B (continued): Multi-factor estimation - Global Funds classified on their beta (1/1997-
12/2002)

WGBI EMBI HY GSCI DOLLAR R²

D1 -0,06 0,06 *** 0,10 * 0,02 -0,18 0,76

D2 -0,15 0,06 0,08 0,01 -0,09 0,52

D3 -0,23 *** 0,04 0,10 *** 0,02 -0,26 ** 0,65

D4 -0,18 *** 0,09 *** 0,13 *** 0,02 -0,16 0,76

D5 -0,07 0,07 *** 0,09 0,02 -0,13 0,82

D6 -0,08 0,10 *** 0,08 0,04 -0,04 0,82

D7 -0,10 0,08 *** 0,09 0,02 -0,04 0,91

D8 -0,11 0,08 *** 0,09 0,03 -0,09 0,88

D9 -0,09 0,08 *** 0,05 0,02 -0,22 0,87

D10 -0,03 0,08 *** 0,04 0,03 -0,13 0,86

D1a 0,00 0,06 ** 0,09 0,01 -0,07 0,78

D1b -0,12 0,08 *** 0,09 0,03 -0,35 * 0,63

D1c -0,06 0,04 ** 0,14 *** 0,02 -0,14 0,76

D10a -0,08 0,07 * 0,04 0,01 -0,17 0,84

D10b -0,02 0,10 *** 0,08 0,03 -0,17 0,85

D10c -0,01 0,07 ** 0,00 0,03 -0,06 0,80

Spread 1-10 -0,03 -0,01 0,06 -0,01 -0,09 0,61

Spread 1-2 0,08 0,01 0,02 0,01 -0,13 0,34

Spread 9-10 -0,06 0,01 0,02 -0,01 -0,12 0,07

Spread 1a-10c 0,02 -0,01 0,09 -0,03 -0,05 0,29

This Table presents the results of the persistence in performance estimation of the combined Model for
the January 1994-December 2002 period. We report the OLS estimators for equally weighted portfolios
per investment strategy, sub-strategy and for all funds. Each year, every fund is classified into a decile
on the basis of its adapted Sharpe score (Panel A) over the previous three years. S&P stands for the
market factor, MSWXUSt = return of the MSCI World Index excluding US, SMBt = the factor-mimicking
portfolio for size (‘small minus big’), HMLt = the factor-mimicking portfolio for book-to-market equity
(‘high minus low’), PR1YRt = the factor-mimicking portfolio for the momentum effect, EMBIt = return of
the JP Morgan Emerging Market Bond Index; SWGBIt = return of the Salomon World Government Bond
Index; HYt = return of the Lehman High Yield Credit Bond Index, GSCIt = return of the Goldman Sachs
Commodity Index, DOLLAR = return of the Federal Reserve Bank Trade Weighted Dollar Index. T-
statistics are heteroskedasticity consistent. *** Significant at the 1% level, ** Significant at the 5% level
and * Significant at the 10% level.

255
Table 38: Summary of Alpha Creation and Equity Exposure

Panel A: Alpha creation

Returns Sharpe Score Std dev. Alpha

D1 0,57 * 0,6 * 0,32 *** 0,19

D2 0,41 * 0,47 0,34 *** 0,43

D3 0,43 *** 0,15 0,33 *** 0,28

D4 0,35 *** 0,27 0,41 *** 0,5 **

D5 0,37 *** 0,25 0,41 *** 0,44 **

D6 0,38 *** 0,16 0,27 0,51 ***

D7 0,36 *** 0,23 * 0,29 * 0,36 ***

D8 0,29 * 0,25 ** 0,33 * 0,26 ***

D9 0,25 0,37 *** 0,22 0,17

D10 0,4 0,49 *** 0,19 -0,05

Kendall Tau -0,56 -0,07 -0,43 -0,35

256
Panel A (continued): Alpha creation

Beta Skewness Kurtosis Adapted Sharpe

D1 0,3 0,39 ** 0,44 *** 0,45 ***

D2 0,09 0,27 ** 0,24 0,7 **

D3 0,24 0,5 *** 0,43 ** 0,53 ***

D4 0,18 0,28 0,37 ** 0,16

D5 0,39 ** 0,38 ** 0,3 * 0,01

D6 0,31 * 0,39 * 0,18 0,08

D7 0,36 *** 0,19 0,37 ** 0,09

D8 0,36 *** 0,25 0,21 0,13

D9 0,39 *** 0,26 * 0,37 *** -0,03

D10 0,46 *** 0,33 ** 0,27 * 0,25

Kendall Tau 0,54 -0,31 -0,31 -0,47

257
Panel B: US equity exposure

Returns Sharpe Score Std dev. Alpha

D1 0,17 * 0,03 0,01 0,66 ***

D2 0,08 0,33 *** 0,07 *** 0,46 ***

D3 0,13 *** 0,51 *** 0,12 *** 0,36 ***

D4 0,18 *** 0,48 *** 0,2 *** 0,29 ***

D5 0,23 *** 0,41 *** 0,28 *** 0,23 ***

D6 0,26 *** 0,44 *** 0,36 *** 0,21 ***

D7 0,33 *** 0,39 ** 0,45 *** 0,25 ***

D8 0,48 *** 0,3 *** 0,46 *** 0,24 ***

D9 0,57 *** 0,16 *** 0,59 *** 0,3 ***

D10 0,56 *** 0,08 *** 0,77 *** 0,34 ***

Kendall Tau 0,78 -0,39 0,82 -0,39

258
Panel B (continued): US equity exposure

Beta Skewness Kurtosis Adapted Sharpe

D1 1,06 *** 0,16 *** 0,36 *** 0,33 ***

D2 0,71 ** 0,31 *** 0,37 *** 0,19 ***

D3 0,49 *** 0,37 *** 0,33 *** 0,11 ***

D4 0,42 *** 0,41 *** 0,36 *** 0,17 ***

D5 0,32 *** 0,32 *** 0,32 *** 0,27 ***

D6 0,21 *** 0,38 *** 0,39 *** 0,31 ***

D7 0,11 *** 0,37 *** 0,26 *** 0,39 ***

D8 0,08 *** 0,32 *** 0,34 *** 0,4 ***

D9 0,02 0,25 *** 0,27 *** 0,46 ***

D10 -0,18 0,28 *** 0,19 *** 0,43 ***

Kendall Tau -1 -0,11 -0,52 0,66

This Table summarises the persistence in performance analysis. Panel A reports the alpha, Panel B reports the
equity exposure. These results are based on the January 1994-December 2002 period. We report the OLS
estimators for equally weighted portfolios per investment strategy, sub-strategy and for all funds. Each year,
every fund is classified into a decile on the basis of its data over the past three years. T-statistics are
heteroskedasticity consistent. *** Significant at the 1% level, ** Significant at the 5% level and * Significant at the
10% level.

259
 In order to quantitatively determine if there is a correlation between the alphas

produced and/or the market exposures, that is to determine if some measures are

redundant, we report the Kendall Tau correlation coefficients in Table 39. Kendall’s

coefficient evaluates the degree of similarity between two sets of ranks given to a same

set of objects. Kendall Tau is used to measure the degree of correspondence between

two rankings and assessing the significance of this correspondence. In other words, it

measures the strength of association of the cross tabulations. This coefficient depends

upon the number of inversions of pairs of objects which would be needed to transform

one rank order into the other64. 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.

Table 39 reports the results. Panel A contains the Kendall coefficients for the

alphas Panel B contains the same factors obtained for the equity market exposure. Panel

A indicates that the only significant correlations exists between the alphas produced by

the standard deviation and alpha series (the coefficient is at 0.52 significantly positive)

and between returns and the adapted Sharpe (0.51 significantly positive). 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. 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. More globally, it is interesting to note that the Sharpe score, the

beta, skewness and kurtosis have not been correlated confirming that these ways of

ranking are completely independent from any other.

64
For more information please see Kendall (1938, 1948) or Abdi (2007).

260
Table 39: Kendall Tau Estimation

Panel A: Kendall Tau between the Alphas

[1] [2] [3] [4] [5] [6] [7] [8]

[1] Returns 1 *** 0,04 -0,02 -0,02 -0,25 0,49 * 0,14 0,51 **

[2] Sharpe Score 1 *** -0,09 -0,36 0,14 -0,02 0,12 0,22

[3] Std dev. 1 *** 0,52 ** -0,49 * 0,09 0,09 0,25

[4] Alpha 1 *** -0,39 0,13 -0,23 -0,07

[5] Beta 1 *** -0,14 -0,09 -0,52

[6] Skewness 1 *** 0,16 0,27

[7] Kurtosis 1 *** 0,14

[8] Adapted Sharpe 1 ***

Panel B: Kendall Tau between the US equity market betas

[1] [2] [3] [4] [5] [6] [7] [8]

[1] Returns 1 *** -0,38 0,87 *** -0,29 -0,87 *** -0,20 -0,45 * 0,78 ***

[2] Sharpe Score 1 *** -0,33 -0,33 0,33 0,75 *** 0,22 -0,60 **

[3] Std dev. 1 *** -0,33 -1,00 *** -0,07 -0,49 * 0,64 ***

[4] Alpha 1 *** 0,33 -0,39 0,00 -0,07

[5] Beta 1 *** 0,07 0,49 * -0,64 ***

[6] Skewness 1 *** 0,18 -0,43

[7] Kurtosis 1 *** -0,36

[8] Adapted Sharpe 1 ***

This Table reports that Kendall Tau between the alphas obtained usingthe various ways o ranking funds
into deciles (returns, Sharpe score, standard devition, alpa, beta, skewness, kurtosis and adapted
Sharpe). *** Significant at the 1% level, ** Significa

261
 As expected, 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, returns and beta, return and adapted Sharpe, Sharpe score and

skewness, Sharpe score and adapted Sharpe, standard deviation and beta, standard

deviation and adapted Sharpe and beta and adapted Sharpe).

This result can be explained by the fact that the beta exposures evolve linearly

(up-down, down-up, up-down-up or down-up-down) in several cases (Sharpe score,

alpha and skewness). In two cases they even evolve perfectly linearly (standard

deviation and beta with a Kendall Tau of 1).

It is interesting to note that the adapted Sharpe score is highly correlated with its

adapted version. 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. In this case, the Sharpe ratio is a good

approximation for one of its extended version.

The alpha and the kurtosis are the only series that are correlated with no other,

being the only to differentiate completely from all the others in terms of evolution of the

market exposure.

262
V Further Analysis

Our results may be due to specific aspects of our methodology, to the time period

covered or to the market conditions. To handle these issues, we perform several further

analyses that should confirm the stability of our findings. We perform two additional

checks. Firstly, we perform the same analysis but and divide the analysis period into sub-

periods in order to determine if the period covered or the market conditions can explain

at least partly our results. Secondly, we rank the funds based on past statistics each July

instead of each January.

5.1 Sub-Period Analysis

To determine if the results obtained are due to the analysis period, we divide this

period in two ways. Firstly, we will take March 2000 as the turning point between bull

and bear market conditions.65 Secondly, we will define bull market as positive months for

the S&P 500 and bear market as negative months for the S&P 500.66

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.
66
We do not report the results for the sake of brevity, but all the tables are available upon request.

263
 1. March 2000 as a Turning Point

March 2000 coincides with the top of the NASDAQ composite index. We perform

the same analysis as previously twice using on one side data before this date, on the

other side, data after this date. There are 39 months before the end of March 2000

(starting from January 1997) and 31 months after this date (starting from April 2000).

The results are clear: before 2000 almost half of the hedge funds significantly outperform

and create pure alpha. These funds have had a limited exposure to the equity markets

and a low volatility. After 2000, only 20% to 30% of the funds – with the same

characteristics – do outperform. These results confirm the results obtained by Capocci et

al. (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.

264
 2. 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. Over

the period covered, 39 months have been positive for the S&P500 while 33 months have

headed south. Our results indicate that no single decile significantly and consistently

outperforms the classical markets during bull market conditions. Top performers even

underperform (but not significantly) during market rally. These results confirm that

hedge funds face difficulties beating classical indices during bull markets. An investor

investing in the previous year’s best or worst performing funds, in funds offering the best

risk-return trade-off, in funds offering the best alpha, 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. During bear market conditions there is consistent alpha

creation for around one third of the deciles. These are another time the low volatile decile

confirming our previous results.

Interestingly, the standard deviation and the beta results suggest however, that

there are still possibilities to create pure alpha. Low volatility funds unexposed to the

equity markets have been able to create significant and consistent alpha over bull market

periods. These results confirm that integrating volatility in the way of classifying funds

helps investors in creating consistent and significant value over time.

265
5.2 Date Impact

To remove any effect that could arise from the January Effect (stock prices tend to

go up between December and January; see for example, Bhardwaj and Brooks, 1992 and

Kramer, 1994), we perform the same analysis by classifying the funds in July instead of

January.67 The first important point to stress is that the same low volatility middle deciles

create alpha. The differences with previous results are that deciles D8 and D9 offer

significantly positive alpha using the January classification date, whereas these alphas

are only marginally significant using the June date. Inversely, decile D10 alpha is

significantly positive using the June classification, but only marginally so using the

January classification. 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.

Factor exposure analysis indicates that classifying funds in January or in June

leads to very close results. Alpha creators are still mildly volatile, with a limited exposure

to the market, a small cap bias and value oriented. Moreover, there is a long exposure to

emerging market bonds and to the high yield market. The best measure remains the

Sharpe score, the standard deviation and the beta. The exposures to the majority of

factors do remain the same using these three best performing measures. There are some

marginal differences in exposures (from slightly exposed to significantly exposed and

inversely) but nothing major. The r-squared are high, ranging from 0.6 to 0.9.

Altogether, this section confirms that classifying funds in July rather than in January does

not change our results significantly.

67
The corresponding tables are available on request.

266
VI Frictions in the Real World

Hedge funds can offer persistence over time. This is a very attractive conclusion to

investors but, and there is naturally a but, 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. We call this issue “frictions the

real world”.

Why can investors not easily profit from our results? The first element to this

conclusion is liquidity. Hedge funds are not liquid investments even if they tend to

become more liquid (at least for equity listed related strategies). In the best case, hedge

funds offer monthly liquidity with a 30-day notice period meaning that it takes minimum

one month to get out of a fund. In this best case, if investors decide to close a position

based on ranking early in the year, it will take them at least one month before being able

to sell and maybe 2 or 3 weeks to get the cash, meaning that they cannot be invested

before the end of March. In this particular case, they get an estimation of performance at

the end of the by the middle of January, 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. In the worst

case, funds offer yearly liquidity with six months of notice. Another concern regarding the

liquidity of hedge funds is that, most hedge funds have lockup period in place meaning

that you cannot sell the position before the lockup period ends. This period lengths

usually minimum one year and goes up to three years. Finally, many funds have gate in

place. A gate is the maximum percentage of assets that can be sold by investors at any

single liquidity date. It ranges from 10% to 30% meaning that redemptions will be

blocked in case of large sell-off.

267
 Funds that report to the database have not to be open to investment or to have

minimum assets. In fact, 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). This means that at least some funds that are considered in studies like

ours cannot be considered for investments. Usually, the funds that are closed to

investments reach their target size because of their good past performance.

Finally, there is always a data problem with hedge fund databases. When a fund

frauds, it stops reporting to databases but its past performance remains in the database

with false figures. This may lead to a wrong choice. Moreover, as it has been mentioned

in many papers in the past. Hedge fund managers report their performance on a

voluntary basis meaning that fraud is even facilitated compared to mutual fund

databases. Moreover, even if most funds report real returns net of all fees, some of them

use pro-forma estimation or manage account performances that cannot be replicated

easily.

All these issues confirm the need to take all our results with care. These are

always based on underlying hypothesis regarding liquidity, the quality of data, aso.

268
VII Conclusion

This study brings new insights in the research analyzing the persistence in hedge

fund returns. Previous studies have all been focused on past performance as the unique

tool to analyse alpha creation. We go one step further in decomposing hedge funds

returns and analysing the persistence in hedge fund returns. Our study uses other

measures such as the Sharpe score, standard deviation, alpha, beta, skewness and

kurtosis.

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 next steps would be to focus on specific strategies and determine what way of

classification is most suitable for each hedge fund strategy. Another idea would be to

develop and use a measure combining the returns, volatility, 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.

269
Part two: Analysis of
Hedge Fund’s Market
Exposure
The Neutrality of Market Neutral Funds

Daniel P.J. 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, Daniel, 2006, The Neutrality of Market Neutral Funds, forthcoming Global

Finance Journal

272
The Neutrality of Market Neutral Funds

Abstract

Using an original database of 634 market neutral hedge funds, 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. We start by analysing this neutrality

using various market neutral indices before focusing on individual fund returns. Finally,

an analysis based on ex-post beta helps us explaining and confirming our previous

results. We perform this analysis over the global January 1993- December 2002 period

as well as on bull and bear market periods.

273
The Neutrality of market neutral funds

Introduction

Hedge funds considered as a whole have been studied since 1997. The precursors

of the academic world were Fung and Hsieh (1997) in their Review of Financial Studies

paper and Ackermann, McEnally and Ravenscraft (1999) in their Journal of Finance

article. Since then, the literature on the subject has expanded strongly and more and

more researchers are focusing on these investment products.

Hedge fund studies can be classified in four global categories. In the first one, we

report studies that are focused on hedge fund performance. This includes mainly studies

that compare the performance of hedge funds with equity and other indices (see for

example Ackermann, McEnally and Ravenscraft, 1999; Brown, Goetzmann and Ibbotson,

1999; Liang, 1999; Amin and Kat, 2001; Liang, 2001; Barès, Gibson and Gyger, 2002;

Liang, 2003; Agarwal and Naik, 2004; Capocci and Hübner, 2004). Results of such

studies are mitigated. The second field of hedge fund performance analysis compares the

performance of hedge funds with the one of mutual funds. In this context, Ackermann,

McEnally and Ravenscraft (1999) and Liang (1999) find that hedge funds constantly

obtain better performance than mutual funds, although lower and more volatile than the

reference market indices considered. Finally, performance analysis includes the study of

the persistence of hedge fund returns. Persistence is particularly important in the case of

hedge funds because, as suggested by Brown, Goetzmann and Ibbotson (1999) and

Liang (2000, 2001), the hedge fund industry has an attrition rate higher than mutual

funds. Brown, Goetzmann and Ibbotson (1999) are in this category. They prove that

offshore hedge funds have positive risk adjusted returns, but they input this result to

style effect and conclude that there is no proof of particular capacity of some fund

274
managers. Agarwal and Naik (2000) analyse the presence of persistence in hedge fund

returns using a one-year moving average period. They prove that there is proof of

persistence in hedge fund's performance, particularly for bad performing funds that

continue to underperform. Capocci and Hübner (2004) and Capocci, Corhay and Hübner

(2004) use the decile analysis developed in Carhart (1997) in order to determine if

persistence is present in hedge fund returns. The first concluded that some low-risk

managers have been able to consistently create alpha over the period studied (1/94-

6/00). 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, excess return creation was

present in most of the cases and there was a clear proof of persistency in hedge fund

returns.

It is important to stress that few authors have attempted to estimate the

behaviour of hedge funds in bear markets. The periods under review do not favour this

exercise, as periods of downward trends on the stock market were rare and

discontinuous before March 2000. For the period 1990-1998, Edwards and Caglayan

(2001) found that only three hedge fund strategies (Market Neutral, Event Driven and

Macro) provide protection to investors when stock markets head south. More recently,

Ennis and Sebastian (2003) contend that in general, hedge funds did not provide investor

protection after the market downturn of March 2000; rather, their superior performance

is mostly due to the good market timing of their managers.

On the other hand, 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, 1997; Brown, Goetzmann and Park,

1998; Brealy and Kaplanis, 2001, Brown and Goetzmann, 2001, Liang 2001; Liang

2003). In this context, Fung and Hsieh (1997) applied Sharpe's style analysis (see

Sharpe, 1992) to a large sample of hedge funds and CTA. They assumed that fund

275
returns are linearly related to the returns on a number of factors and measured the

factor by 8 mimicking portfolios, found that the regressions have little explanatory power

and suggested that the low R²adj is due to the fund's trading strategy. A particular aspect

that has been taken into account more recenthly is the style drift in hedge fund returns.

This effect comes from the fact that hedge fund managers are opportunity driven and

therefore change style over time. Brown, Goetzmann and Park (1998) analyse hedge

fund returns during the 1997-98 Asian crisis. They use rolling regression to take the style

drift into account. The methodology consists in realiazing a set of linear regressions and

moving the estimation perdio of each of them by one observation. This simple technique

enables to observe style variation of a manager over time. This methodology has one

major drawback: the choice of a number of observations used for the estimation.

McGuire, Remolona and Tsatsaronis (2005) apply the same methodology. To handle with

this issuer, 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). This

technique is adaptive in the sense that changes in th style exposures are piced up

automatically from the data. Unlike the ad hoc rolling regression approach, the time

variation in the exposures is explicitly modelled. No restrictions are imposed on the

betas. As stressed by Posthuma and Van der Sluis, this model is a state-space model and

can be estimated by using standard Kalman filter techniques68. No window size and ad

hoc chosen length need to be used. The Kalman filter procedure chooses the optimal

weigthing scheme directly from the data. The filter is an adaptive system based on the

measurement and updateing equations.

68
See Pollock (1999) for a detailed presentation of the Kalman filtering and space-state models.

276
 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. 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. This low correlation is also emphasised by

Liang (1999), 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, 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. They obtain

the best results when 10-20% of the portfolio value is invested in hedge funds. Taking all

these results into account, hedge funds seem to be a good investment tool. 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 means that a stronger risk

control does not correspond with a decrease in return.

Finally other authors have analysed various other aspect of the hedge fund

industry. This category includes the ”other studies”. Schneeweis and Spurgin (1999),

Amenc, Curtis and Martellini (2002), Amenc, Martellini and Vaissié (2002) and Berényi

(2002) have studied the risks involved in hedge fund investing. Schneeweis and Spurgin

and Amenc, Martellini and Vaissié (2002) proved that hedge fund returns are not only

exposed to the market risk, but that other risks like volatility risk, default risk or liquidity

risk have to be considered. Liang (2000) analysed the presence of survivorship bias in

hedge fund data and Fung and Hsieh (2000) included other biases in their analysis.

Ackermann and Ravenscraft (1998) emphasised that the stronger legal limitations for

mutual funds than for hedge funds hinder their performance. Various authors study

hedge fund indices. There are many different hedge funds’ indices providers like EACM,

277
HFR, CSFB/Tremont, Zurich Capital, Van Hedge, Hennessee Group, Hedgefund.net, LJH

Global Investment, Mar, Altvest and Magnum.69 Brooks and Kat (2001) Amenc and

Martellini (2002) studied this aspect in detail.

CTAs are a particular category in the hedge fund world. Some authors consider

them as part of the hedge fund world (Fung and Hsieh, 1997; Schneeweis and Spurgin,

2000), whereas others study them by separating them from hedge funds (Liang, 2003)

or study them on a stand-alone basis (Fung and Hsieh, 2000a; Gregoriou and Rouah,

2003; Capocci, 2004b). In this study, we only study market neutral funds and do not

include CTAs.

69
See Amenc and Martellini (2002) for a complete description of these hedge fund indices providers.

278
I Interest of the study

This study aims at analysing the exposure to the equity market for market neutral

funds. At first sight, it can seem quite reductive to focus on a particular strategy since

the bulk of the literature considers hedge funds as a whole. Three independent reasons

justify this choice. Firstly, market neutral funds are particular since their objective is to

create alpha while completely hedging the exposure to the market. 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. Recently, in their

study based on a very large database of a total of 2894 individual funds, Capocci, Corhay

and Hübner (2004) concluded that equity market neutral funds present interesting

characteristics that need further investigation. This is exactly the objective of the present

study. Secondly, as suggested in several studies, market neutral funds represent a high

percentage of the hedge fund industry. According to our database described hereafter, as

much as 28.3% of the global MAR/CISDM individual funds in the database are market

neutral funds. Other authors also report high numbers of market neutral funds in their

database (see for example Capocci and Hübner, 2004 and Capocci, Corhay and Hübner,

2004). A high number of market neutral funds allow to analyse them in more detail and

to obtain global results. Thirdly, more and more authors however consider individual

strategies to better understand their particularities. Mitchell and Pulvino (2002) only

focused on risk arbitrage, Fung and Hsieh (2002a) on fixed income arbitrage funds;

Capocci (2004a) has analysed the inclusion of convertible arbitrage funds in a classical

portfolio. Gatev, Goetzmann and Rouwenhorst (1999) study relative value funds on a

stand-alone basis. Market neutral funds should also be studied on a stand-alone basis to

understand their particularities. Navone’s (2001) look at the diversification benefits from

279
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.

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. Sub-strategies are

arbitrage market neutral funds, which are managers that apply market neutral arbitrage

strategies. This category includes, among other things, convertible arbitrage funds that

buy undervalued convertibles, while hedging all intrinsic value and fixed income arbitrage

funds that exploit pricing anomalies in the global fixed income. Another sub-strategy is

long/short market neutral funds also named equity market neutral funds. These funds

simultaneously take long and short positions of the same size within the same market.

Managers take advantage of relative price discrepancies. 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. In theory, market risk is greatly reduced but it is

very difficult to make a profit on a large diversified portfolio, so stock picking is critical.

The final sub-strategy is mortgage-backed funds that invest in mortgage-backed

securities but also in futures and options. Usually they focus on AAA-rated mortgage

bonds.

Market neutrality implies dollar neutral, beta neutral or both. Dollar neutral

strategies have zero net investment, i.e. equal dollar amount in long and short positions.

Beta neutral strategies target a zero total portfolio beta (i.e. beta of the long positions

equals the beta on the short side. While dollar neutrality has the virtue of neutrality, beta

neutrality better defines a strategy uncorrelated with the market return. Many managers

of such strategies balance their long and shorts in the same sector or industry. By being

sector neutral, they avoid the risk of market movements affecting some sectors or

industries differently than others.

280
 The study is organised as followed. Section III describes that database. In section

IV we report the descriptive statistics and analyse the attrition rate present in our

database. In section V we present our methodology. We report the global results in

section VI and the results obtained for individual funds in section VII. We perform a sub-

period analysis in section VIII. Section IX concludes the study.

281
II Database

Four main hedge fund databases are available for empirical studies, the Managed

Account Reports, Inc/Centre for International Securities Derivatives Markets, Hedge Fund

Research, Inc, TASS Management and the Barclays database. The first three ones are the

most used in academic studies. The Barclay database is currently under investigation by

various authors but it has not yet been used in published studies. The data provider

collects various information on the funds included. For a majority of funds, they record

other useful information such as company name, start and ending date, strategy

followed, assets under management, management and incentive fees, manager's name,

manager's address, etc. There is no consensus on the definition of the strategy followed

but there are similarities. MAR/CISDM defines 9 strategies with a total of 16 sub-

strategies. HFR defines 16 different strategies in 2 categories, 11 non-directional and 5

directional strategies, plus the Funds of Funds and the Sector categories. TASS defines

15 strategies. Finally, Barclays defines 20 individual strategies.

We use hedge fund data from MAR/CISDM, as in Fung and Hsieh (1997),

Schneeweis and Spurgin (1998), Amin and Kat (2001), and Capocci, Corhay and Hübner

(2004). The database gives monthly net-of-fee individual returns and other information

on individual funds and groups them in indices. The whole database consists of 634

individuals market neutral funds including 398 surviving funds (62.8%) and 236 (37.2%)

dissolved funds. We use 120 monthly returns between January 1993 and December

2002.

282
III Descriptive statistics and attrition rates

Before analyzing the presence of bias in hedge fund data, we present the

descriptive statistics, the correlation analysis and the attrition rates in this section.

3.1 Descriptive statistics

In Panel A of Table 41 contains descriptive statistics of the funds, whether alive or

dead, in our database. We report the statistics for the whole market neutral database

and we divide this database in arbitrage funds, long/short funds, mortgage backed

securities funds and non-classified funds.70 These hedge fund data are contrasted with

the descriptive statistics of the S&P 500 that represent the equity market. 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.

Panel A of Table 40 indicates that market neutral funds offered an average

monthly return of 1.08% over the period studied with a monthly standard deviation of

O.94%. Arbitrage funds offered almost the same return with the same volatility whereas

long/short funds offered 1.16% with a monthly volatility of 1.16%. Mortgage fund

returns are more volatile than the other market neutral funds. All average returns are

significantly positive. Median returns indicate the same pattern except that mortgage-

backed funds have an impressively high median return compared with the other funds

studied. This result can be explained but the low skewness and extremely large kurtosis

of mortgage-backed securities funds. Minimum monthly returns are relatively high with a

total of –2.54% for market neutral funds except for mortgage backed funds (-8.74%).

On the other side, the maximum is also relatively low for all strategies, particularly for

70
There are 3 funds with no sub-strategy. The data regarding these funds ended in January 2001.

283
arbitrage funds (2.93%) and mortgage backed funds (2.84%). Mortgage backed

securities have largely negatively skewed returns’ distribution (skewness of –4.11) with

fat tails (kurtosis of 26.45). All other sub-strategies except long/short funds and the

strategy considered as a whole all have significantly negative skewness, which leads to

negatively skewed return distribution. All classifications also have significantly positive

kurtosis which leads to the presence of fat tails (a least to some extent). 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,

which, in turn offer a better trade-off as mortgage-backed securities funds.

By comparison, the equity index has a low average monthly return of 0.69%,

which is in line with the long-term return of the equity market of around 8%. This return

is weakly and significantly positive with a relatively high volatility of 4.47%. 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. 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, nor has fat tails.

This is in line with previous results indicating normality of returns for the stocks indices

(see for example Fung and Hsieh, 1999).

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. Decile 1

contains the worst performing funds over the period and decile 10 the best performing

ones. The dead fund column indicates that bad performing funds have more chance to be

dissolved. On the average, 52% of the funds in the worst 3 deciles are dissolved against

31% on the average for the top 3 deciles. Moreover 62% of the funds in decile 1 are

dissolved. Interestingly, fewer funds from deciles 5 to 9 have been dissolved compared

284
to decile 10 and middle decile fund returns are more stable than worst and top

performing decile (see standard deviation). 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. 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), but that high maximum does

not necessarily lead to higher decile (decile 1 maximum is the second highest maximum

over the 10 deciles). Skewness and kurtosis results do no lead to particular remarks. We

can however note that some skewness are significantly positive (deciles 2, 7 and 9)

whereas others are significantly negative (deciles 1, 4, 5 and 6) and that all kurtosis that

are significant are positive. There is only one insignificant negative kurtosis.

3.2 Correlation analysis

As suggested in the introduction, the traditional hedge funds literature contends

that, thanks to the weak correlation between hedge funds and other securities, hedge

funds are likely to improve the risk-return trade-off when added to a traditional portfolio

(see Fung and Hsieh, 1997; Schneeweis and Spurgin, 1997; Liang, 1999; Amin and Kat,

2001). 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. The correlation coefficients are reported in the Table 41.

285
 Table 40: Descriptive statistics and decile descriptive statistics

Panel A: Strategy, sub-strategies and index descriptive statistics

% of the Living
No of Fds Dead Funds Mean Return S. D.
strategy Funds

Market neutral
634 100% 398 236 1,08% *** 0,94%
funds

- Arbitrage funds 281 44,30% 175 106 1,03% *** 0,92%

- Long/short
306 48,30% 197 109 1,16% *** 1,16%
funds

- Mortgage
44 6,90% 26 18 1,06% *** 1,32%
backed funds

- No sub-strategy 3 0,50% 0 3 0,73% *** 1,11%

Index NA NA NA NA 0,69% * 4,47%

286
Panel A (continued): Strategy, sub-strategies and index descriptive statistics

% of the Living
No of Fds Dead Funds Mean Return S. D.
strategy Funds

Market neutral
634 100% 398 236 1,08% *** 0,94%
funds

- Arbitrage funds 281 44,30% 175 106 1,03% *** 0,92%

- Long/short
306 48,30% 197 109 1,16% *** 1,16%
funds

- Mortgage
44 6,90% 26 18 1,06% *** 1,32%
backed funds

- No sub-strategy 3 0,50% 0 3 0,73% *** 1,11%

Index NA NA NA NA 0,69% * 4,47%

287
Panel B: Decile descriptive statistics (120 months)

No of Fds Living Funds Dead Funds Mean Return S. D.

Decile 1 63 38% 62% -0,34% 3,53%

Decile 2 63 43% 57% 0,28% *** 1,21%

Decile 3 63 62% 38% 0,55% *** 0,59%

Decile 4 63 56% 44% 0,72% *** 0,95%

Decile 5 64 80% 20% 0,89% *** 0,84%

Decile 6 64 72% 28% 1,10% *** 1,04%

Decile 7 63 71% 29% 1,21% *** 1,28%

Decile 8 63 71% 29% 1,40% *** 1,10%

Decile 9 63 81% 19% 1,75% *** 1,43%

Decile 10 63 56% 44% 2,70% *** 2,41%

288
Panel B (continued): Decile descriptive statistics (120 months)

Sharpe
Median Min Max Skewness Kurtosis
score

Decile 1 -0,20% -17,50% 8,50% -1,28 *** 5,02 *** -0,1

Decile 2 0,30% -2,30% 4,40% 0,51 *** 0,78 *** 0,23

Decile 3 0,60% -1,00% 2,10% 0,01 -0,02 0,93

Decile 4 0,80% -2,40% 2,40% -1,11 *** 1,94 *** 0,76

Decile 5 1,00% -4,20% 2,70% -1,9 *** 10,31 *** 1,05

Decile 6 1,20% -2,70% 4,20% -0,27 *** 1,24 *** 1,07

Decile 7 1,20% -2,70% 5,90% 0,36 *** 2,05 *** 0,95

Decile 8 1,40% -2,50% 4,40% 0,01 1,02 *** 1,27

Decile 9 1,70% -2,40% 6,30% 0,3 ** 0,72 *** 1,23

Decile 10 2,50% -4,10% 9,20% 0,11 0,31 1,12

This Table shows the number of funds (No of Fds), percentage of the strategy (% of the strategy), the number of
living funds, the number of dead funds, the mean return, the standard deviation (S.D.), the median, the minimum
(min), the maximum (max), the skewness, the kurtosis and the Sharpe score for the individual hedge funds in our
MAR/CISDM database for the whole period 01:1993-12:2002. No sub-strategy ended in 01:12001. Sharpe score
is the ratio of return and standard deviation. Panel A focus on strategy, sub-strategy and index descriptive
statistics, Panel B on decile descriptive statistics. *** Significant at the 1% level, ** Significant at the 5% level and
* Significant at the 10% level.

289
Table 41: Correlation between market neutral strategies and

equity index

Market Mortgage
Arbitrage Long/short No sub-
neutral backed Index
funds funds strategy
funds funds

Market neutral
1
funds

- Arbitrage funds 0,91 1

- Long/short funds 0,92 0,7 1

- Mortgage backed
0,49 0,45 0,3 1
funds

- No sub-strategy 0,62 0,58 0,53 0,37 1

Index 0,54 0,46 0,54 0,19 0,3 1

This Table reports the correlation coefficients among market neutral sub-strategies and between
market neutral funds strategies and the index. The database consists of 634 individual market neutral
funds, including 398 survived funds and 236 dissolved funds as of December 2002. The period
covered is 01:1993-12/2002 period.

Table 41 indicates that the correlation between the market neutral index and its main

components, arbitrage and long/short funds is higher than 90%. This result is logical since

these 2 sub-strategies represent a total of more than 90% of the funds in the database. The

correlation between the market neutral index and the mortgage funds is significantly lower

at 0.49. The correlation between the market neutral sub-strategies ranges from 0.37

(between mortgage-backed funds and no sub-strategy funds) to 0.7 (between arbitrage

funds and long/short funds). These figures are reasonably high and suggest that the sub-

classification has a sense since no sub-strategies are completely correlated. The correlation

between market neutral funds (considered as a whole or per sub-strategies) ranges from

0.19 (between no sub-strategy funds and the index) to 0.54 (between market neutral funds

290
and the index). These figures are reasonable but relatively high for the index. These results

are in line with previous results obtained by Schneeweis and Spurgin, 1998; Liang, 1999;

Amin and Kat, 2001 and Capocci and Hübner, 2004).

Figure 9 reports the average, median, minimum and maximum individual correlation

between market neutral fund index and its sub-strategies and the equity index.71 We

describe Figure 9 using market neutral funds as an example. The bottom line of the block

represents the minimum individual correlation between market neutral funds and the equity

index, -0.29 in the first case. The upper part of the grey block ended at the average

correlation between market neutral funds and the index, 0.15 for market neutral funds. The

median is reported below the white part of the block at 0.3 for market neutral funds. Finally,

the top of the box is the maximum individual correlation between a single market neutral

fund and the equity index.

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. This result indicates that even

if individual market funds are not correlated with the equity index, it is important to consider

individual funds because index aggregation can increase the correlation and the exposure to

the equity market. This point is particularly important for the beta analysis. 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.

71
We removed 3 funds with few data to perform the correlation analysis at the individual fund level. We do

not report the result of the no sub-strategy index because it includes only 3 individual funds.

291
Figure 9: Minimum, maximum, median and mean individual

correlations between market neutral funds and the equity

index

80%

60%

40%

20%

0%

-20%

-40%

-60%
Market Neutral Arbitrage Funds Long/short Funds Mortgage backed
Funds Funds

This figure reports the minimum, maximum, median and mean individual correlations

between market neutral funds and the equity index between January 1993 and December

2002.

3.3 Birth and attrition rates

Attrition rates of hedge funds are largely publicized in academic studies (see for

example Fung and Hsieh, 1997; Liang, 2000; Liang 2003). This rate can be defined as the

percentage of funds in the database that are dissolved each year. Table 42 reports the birth

rate and the attrition rate of the fund in our database for each year under review.

292
Table 42: Birth and attrition rates

Year start New Dissolved Year end Birth rate Attrition rate

1993 63 27 1 89 42,90% 1,60%

1994 89 50 2 137 56,20% 2,20%

1995 137 44 18 163 32,10% 13,10%

1996 163 74 13 224 45,40% 8,00%

1997 224 80 19 285 35,70% 8,50%

1998 285 78 39 324 27,40% 13,70%

1999 324 58 36 346 17,90% 11,10%

2000 346 47 32 361 13,60% 9,20%

2001 361 69 34 396 19,10% 9,40%

2002 396 44 42 398 11,10% 10,60%

Total 571 236 Average 30,10% 8,70%

This Table reports the number of funds at year start, the number of new funds, the number dissolved
funds, the birth rate and the attrition rate for the market neutral funds in our database. Birth rate is
defined as the ratio between the number of new funds and the number of funds at year start. The
attrition rate is defined as the ratio of dissolved funds to number of funds at year start. The database
consists of 634 individual market neutral funds, including 398 survived funds and 236 dissolved funds
as of December 2002. The period covered is 01:1993-12/2002.

293
 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. On the other side,

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 13.7%. The high level of dissolution in 1998

should probably be due to the Russian, the Asian crisis and the bailout of Long Term Capital

Management, LTCM. The average birth rate in around 30% and the average attrition rate is

8.7% indicating that each year, on average around 8.7% of the fund in the industry are

dissolved. Table 42 also indicates that the total number of funds has increased linearly over

the period. These results are in line with the results obtained by Liang (2000) for the TASS

database but lower than the 14% obtained by Brown, Goetzmann and Ibbotson (1999).

294
IV Survivorship bias analysis

Survivorship bias is particularly important in the case of hedge funds (see Fung and

Hsieh, 1997; Fung and Hsieh, 2000; Ackermann, McEnally and Ravenscraft, 1999; Capocci

and Hübner, 2004). Usually this bias is studied on a global basis for full databases including

a variety of different strategies. In this study, since we focus on market neutral funds, we

will analyse the presence of survivorship bias for the funds in our database.

Survivorship bias can be defined in 2 ways: the performance difference between

surviving and dissolved funds (e.g. Ackermann et al., 1999) and the performance difference

between living and all funds (e.g. Fung and Hsieh, 2000). We report the bias using both

definitions.

Panel A of Table 43 indicates that the returns have been less interesting in recent

years compared to the past. The most difficult years for market neutral funds have been

1994 (bond market crisis), 1998 (Russian, Asian crisis and the LTCM bailout) and 2002 (3rd

year of bearish market). The average return for the funds that survived over the whole

period is 1.22%, higher than the whole funds in the database average return of 1.08% and

higher than the returns of funds that have been dissolved during the period studied with a

performance of 0.77%. Standard deviation is interesting because it indicates that in recent

years the volatility of the dissolved funds has increased dramatically compared to surviving

funds. This indicates that volatile funds have been more exposed to dissolution in 2001 and

2002.

295
 Panel B reports the difference of performance between living and dead funds (left

side) and living and all funds (right side). 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. The total bias is 5.4% using the first formula and 1.68% using the general one. The

latest figure is the generally used one in hedge fund studies and can be compared to past

results.

This latter value is much higher than the very low value obtained by Ackermann et al.

for the period 1988-1995. It is similar to the percentage of 1.5% from Fung and Hsieh

(1998), lower than the 0.30% monthly bias found by Fung and Hsieh (2000) and slightly

higher than the percentage of 1.2% found by Capocci and Hübner (2004) for the 1994-2000

period. It is however lower than the 3% bias found by Liang (2001), which is also the

industry consensus as stressed by Amin and Kat (2001).72

72
This consensus value is quite high when compared to the 0.8-1.5 bias reported by Malkiel (1995) and

Brown and Goetzmann (1995) for US mutual funds.

296
Table 43: Survivorship bias

Panel A: Annual performance (all funds, surviving funds and dissolved funds)

All Funds Surviving Funds Dissolved Funds

Year Return S.D. Return S. D. Return S. D.

1993 1,41% 0,43% 1,47% 0,47% 1,37% 0,47%

1994 0,25% 0,62% 0,45% 0,65% 0,10% 0,68%

1995 1,54% 0,55% 1,69% 0,58% 1,37% 0,75%

1996 1,56% 0,46% 1,88% 0,42% 1,16% 0,63%

1997 1,41% 0,92% 1,58% 0,94% 1,20% 1,02%

1998 0,58% 1,42% 0,64% 1,33% 0,48% 1,61%

1999 1,44% 0,80% 1,54% 0,78% 1,26% 0,91%

2000 1,41% 1,14% 1,64% 1,13% 0,70% 1,44%

2001 0,80% 0,74% 0,85% 0,65% 0,56% 1,39%

2002 0,36% 0,63% 0,43% 0,59% -0,55% 2,96%

Average 1,08% 0,77% 1,22% 0,76% 0,77% 1,19%

297
Panel B: Survivorship bias

Living - Dead Funds Living - All Funds

Year Return Return

1993 0,10% 0,06%

1994 0,35% 0,20%

1995 0,32% 0,15%

1996 0,73% 0,32%

1997 0,38% 0,17%

1998 0,16% 0,07%

1999 0,28% 0,09%

2000 0,94% 0,23%

2001 0,29% 0,05%

2002 0,97% 0,06%

Bias 93-02 0,45% 0,14% per Month

5,40% 1,68% per Year

This Table reports the survivorship bias of our database. Our MAR/CISDM database contains 634 individual
market neutral funds, including 398 survived funds and 236 dissolved funds as of December 2002. In Panel B
survivorship bias is calculated as the performance difference between surviving funds and dissolved funds (left-
hand side) and as the performance difference between surviving funds and all funds (right-hand side). All
returns are net of fees. Numbers in the table are yearly percentage unless otherwise indicated.

298
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. The methodology used in this study is

two-fold. 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. First, we

analyse market neutral funds grouped per sub-strategy. Secondly, 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. Thirdly, 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. Each year, we classify funds on the

basis of their previous year performance and we create return series. Then, 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.

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. For individual funds, we also perform an

ex-post beta analysis. This methodology uses the same results as previously but differently.

We classify the funds in decile on the basis of their ex-post betas and analyse their

descriptive statistics. 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. Does the beta

exposure explain better or worst returns?

299
 The model used is a single index model based on the classical CAPM developed by

Sharpe (1964) and Lintner (1965). Its equation to estimate is the following:

R Pt Ft = α P + β P * R Mt + ε Pt t = 1,2,..., T (12)

Where RPt = return of fund P in month t; RMt = return of the equity market portfolio, in our

case the S&P 500 on month t; εPt = error term; αP and βP are the intercept and the slope of

the regression, respectively. The intercept of this equation, αp commonly called Jensen’s

alpha (1968) is usually interpreted as a measure of out- or under-performance relative to

the market proxy used. The beta is interpreted as a measure of the dependence of a fund’s

return to the index. It is a relative measure. We compute all estimations by using Newey-

West (1987) standard errors to adjust for any autocorrelation in the returns

300
Table 44: Market exposure analysis

Panel A: Market exposure for market neutral funds

Average return S. D. Alpha Market R²adj

Market neutral funds 1,08% *** 0,94% 1,00% *** 0,11 *** 0,29

- Arbitrage funds 1,03% *** 0,92% 0,97% *** 0,1 *** 0,21

- Long/short funds 1,16% *** 1,16% 1,05% *** 0,14 *** 0,29

- Mortgage backed
1,06% *** 1,32% 1,02% *** 0,06 * 0,03
funds

- No sub-strategy 0,73% *** 1,11% 0,62% *** 0,08 *** 0,08

301
 Average return S. D. Alpha Market R²adj

Decile 1 -0,34% 3,53% -0,48% 0,21 *** 0,06

Decile 2 0,28% 1,21% 0,22% * 0,1 *** 0,12

Decile 3 0,55% *** 0,59% 0,51% *** 0,05 *** 0,12

Decile 4 0,72% *** 0,95% 0,64% *** 0,11 *** 0,29

Decile 5 0,89% *** 0,84% 0,85% *** 0,06 *** 0,08

Decile 6 1,10% *** 1,04% 1,01% *** 0,13 *** 0,31

Decile 7 1,21% *** 1,28% 1,15% *** 0,09 *** 0,08

Decile 8 1,40% *** 1,10% 1,32% *** 0,11 *** 0,19

Decile 9 1,75% *** 1,43% 1,66% *** 0,14 *** 0,19

Decile 10 2,70% *** 2,41% 2,56% *** 0,2 *** 0,13

The table reports the market exposure analysis. Panel A presents the average return, the standard deviation
(S.D), the alpha, the market exposure (the beta) and the R²adj. 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, including 398 survived funds and 236 dissolved funds
as of December 2002. *** Significant at the 1% level, ** Significant at the 5% level and * Significant at the 10%
level.. t-stats are heteroskedasticity consistent. The period covered is 01:1993-12:2002.

302
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. Panel A of Table 44 reports

the results for the market neutral strategy and various market neutral sub-strategies.

Panel A reports the average return and standard deviation of returns over the period

studied as well as the results of the regression analysis. A comparison between the average

return and the alpha indicates that they are relatively close to each other, which indicates

that only a small part of market neutral returns can be explained by the market. The R²adj

are relatively low, 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 betas indicate however that all

indices except mortgage-backed securities funds are significantly exposed to the equity

market at the 1% significance level. On an absolute term however, the beta obtained is low

between 0.06 for mortgage backed securities and 0.14 for long/short funds.

Panel B reports the results obtained using the decile classification. These results are

based on decile constructed on the basis of the performance over the period studied. As

previously, top performing funds are reported in Decile 10 and worst performing funds are

reported in Decile 1. Panel B indicates the same pattern as Panel A, the difference between

the alpha and the average return is relatively low, the R²adj are relatively low, but all

market betas are significantly positive even if they are all low in absolute term. Top and

worst performing funds have however the highest market exposure.

303
Figure 10: Lagged 12-month decile description

450 50
400 45
350 40
35
300
30
250
25
200
20
150
15
100 10
50 5
0 0

e
93

94

95

96

97

98

99

00

01

ag
19

19

19

19

19

19

19

20

20

er
Av
Funds per year Funds per decile

This figure reports the yearly lagged 12-month decile description. 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.

In a second step, we perform the same analysis on the basis of Carhart (1997)

methodology. The objective of this analysis is to determine if last year best or bad

performers are more exposed to the market. Each year, we classify the funds on the basis of

their previous year performance and we create return series. Then, 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.

304
 Before analysing the results of this analysis, we report the number of funds in each

decile in Figure 10. 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. Figure 10

indicates that the average number of funds per year is 264 with decile of 28 funds on

average. Over the whole period, decile 1 (worst performing funds) would have offered

0.72% per month against 1.26% for decile 10 (best performing funds) and 0.69% for the

index.

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. Moreover, the R²adj

obtained is very low, which confirms the results. On the other hand however, the market

column indicates that 6 out of 10 deciles are significantly exposed to the market. These

deciles are spread but the 3 top performing deciles are all significantly exposed to the

market. 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 results suggest that the exposure of market neutral funds to the equity market

is not clear. Even if the 6 out of 10deciles's exposure are significantly positive, in absolute

term, 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. In absolute term, the market factors are another time very low.

305
Table 45: Lagged 12-month decile exposure

Portfolio Average return S. D. Alpha Market R² adj

D1 0,72% *** 2,14% 0,70% *** 0,11 *** 0,06

D2 0,62% *** 1,32% 0,63% *** 0,07 * 0,05

D3 0,60% *** 0,83% 0,65% *** 0,04 ** 0,03

D4 0,48% *** 0,75% 0,49% *** 0,03 * 0,03

D5 0,61% *** 0,82% 0,61% *** 0,06 *** 0,11

D6 0,67% *** 0,95% 0,67% *** 0,03 0,01

D7 0,79% *** 1,22% 0,80% *** 0,06 * 0,03

D8 0,89% *** 1,09% 0,88% *** 0,08 *** 0,1

D9 1,11% *** 1,66% 1,04% *** 0,11 ** 0,08

D10 1,26% *** 2,38% 1,33% *** 0,11 ** 0,04

1-10 spread 0,54% ** 2,71% 0,63% *** 0 -0,01

1-2 spread -0,10% 1,69% -0,07% -0,04 0,01

9-10 spread 0,15% 2,03% 0,29% -0,01 -0,01

The table reports the market exposure using deciles constructed on the lagged-12 month performance
(see Carhart 1997). It reports the average return, the standard deviation (S.D), the alpha, the market
exposure (the beta) and the R²adj. Our MAR/CISDM database contains 634 individual market neutral
funds, including 398 survived funds and 236 dissolved funds as of December 2002. *** Significant at the
1% level, ** Significant at the 5% level and * Significant at the 10% level. t-stats are heteroskedasticity
consistent. The period covered is 01:1993-12:2002.

306
VII Individual fund analysis

In this section, 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. In the first sub-section, we analyse the market exposure. In the second one, we look

at the ex-post beta exposure.

7.1 Market exposure

As we suggested in the correlation analysis, it is important to check if the results

obtained on a strategy or sub-strategy basis are still valid on an individual fund basis. To

check this, we perform the same analysis using the returns of individual funds.73 Results are

reported in Table 46. Table 46 reports the average return, the return distribution which is

the percentage of average monthly returns that are significantly positive (+), significantly

negative (-) and not significant (0) on the one side, and the results of the regression of the

other. Like average return, individual alpha and beta are classified as significantly positive at

the 5% significance level, significantly negative or not significant.

73
We require all funds to have at least 24 months of data to perform this analysis. This lead us to remove

172 funds (74 arbitrage funds, 92 long/short funds, 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, 214 long/short funds, 39

mortgage backed securities funds and 2 no sub-strategy funds).

307
 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. Very few average returns are significantly negative over the period studied. The

average returns and alphas obtained for individual funds are higher compared to the ones

obtained while studying funds agglomerated in indices. Alphas are always lower than the

corresponding average return but the differences are relatively small. Nevertheless, much

more alphas are concentrated around 0, being not significantly different from it. 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, 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.

On an absolute term, the market exposure remains obviously the same here as it was

in previous analysis. 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. 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. First, the aggregation of funds in indices leads to an increase in the exposure to the

equity market. 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. 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, the funds

significantly exposed could bias the results. These two remarks are of particular interest.

308
Table 46: Individual fund market exposure

Return distribution Alpha distribution

Beta distribution (5%)
Average (5%) (5%)
S.D. Alpha Market R²adj
return
+ 0 - + 0 - + 0 -

Market neutral funds 1,02% 78% 21% 1% 3,30% 0,95% 62% 38% 0% 0,13 32% 63% 5% 0,08

- Arbitrage funds 1,03% 84% 16% 0% 2,50% 0,94% 71% 29% 0% 0,09 33% 63% 3% 0,05

- Long/short funds 1,03% 71% 27% 2% 4,30% 1,00% 52% 47% 0% 0,16 36% 57% 7% 0,1

- Mortgage backed funds 0,89% 85% 13% 3% 3,00% 0,75% 67% 33% 0% 0,11 8% 92% 0% 0,03

- No sub-strategy 0,45% 50% 50% 0% 1,30% 0,36% 50% 50% 0% 0,18 50% 50% 0% 0,12

The table reports the individual fund market exposure. It reports the average return, the return distribution at the 5% level. (+ are significantly positive, 0 not
significantly different from 0 and – significantly negative), the standard deviation (S.D), the alpha, alpha distribution at the 5% level (+ are significantly positive, 0
not significantly different from 0 and – significantly negative), the market exposure (the beta), the beta distribution (+ are significantly positive, 0 not significantly
different from 0 and – significantly negative) and the R²adj. Our MAR/CISDM database contains 634 individual market neutral funds, including 398 survived funds
and 236 dissolved funds as of December 2002. T-stats are heteroskedasticity consistent. The period covered is 01:1993-12:2002.
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 1,21% 0% 35% 65% 0,13 5% 73% 22% -1,13 14% 82% 3% 2,51 9% 82% 9% 0,10

- Arbitrage funds 1,10% 0% 25% 75% 0,09 4% 77% 19% -0,69 12% 83% 5% -1,35 7% 82% 11% 0,09

- Long/short funds 1,27% 0% 50% 50% 0,17 7% 69% 24% -1,30 15% 83% 1% 6,16 11% 82% 7% 0,12

- Mortgage backed funds 1,47% 0% 8% 92% 0,08 5% 79% 15% -2,34 15% 85% 0% -4,47 10% 87% 3% 0,09

- No sub-strategy 0,77% 0% 50% 50% -0,01 0% 100% 0% -4,59 50% 50% 0% 148,98 0% 100% 0% 0,21

The table reports the individual fund market exposure. It reports the average return, the return distribution at the 5% level. (+ are significantly positive, 0 not
significantly different from 0 and – significantly negative), the standard deviation (S.D), the alpha, alpha distribution at the 5% level (+ are significantly positive, 0 not
significantly different from 0 and – significantly negative), the market exposure (the beta), the beta distribution (+ are significantly positive, 0 not significantly different
from 0 and – significantly negative) and the R²adj. Our MAR/CISDM database contains 634 individual market neutral funds, including 398 survived funds and 236
dissolved funds as of December 2002. T-stats are heteroskedasticity consistent. The period covered is 01:1993-12:2002.
Table 48: Bull and Bear Beta Estimation

Bear Bull
Alpha <0 0 >0 <0 0 >0 <0 0 >0 R² adj
Beta Beta

Market neutral funds 1,22% 1% 47% 47% 0,17 5% 62% 32% 0,05 8% 79% 13% 0,31

- Arbitrage funds 1,15% 1% 38% 38% 0,14 6% 60% 33% 0,04 7% 81% 12% 0,22

- Long/short funds 1,24% 1% 60% 60% 0,19 5% 64% 31% 0,09 8% 76% 14% 0,30

- Mortgage backed funds 1,58% 2% 23% 23% 0,20 2% 70% 27% -0,10 7% 86% 7% 0,11

- No sub-strategy 0,90% 0% 33% 33% 0,18 0% 33% 67% 0,01 0% 100% 0% 0,10

The table reports the individual fund market exposure. It reports the average return, the return distribution at the 5% level. (+ are significantly positive, 0 not
significantly different from 0 and – significantly negative), the standard deviation (S.D), the alpha, alpha distribution at the 5% level (+ are significantly positive, 0 not
significantly different from 0 and – significantly negative), 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, 0 not significantly different from 0 and – significantly negative) and the R²adj. Our MAR/CISDM database contains 634 individual market neutral funds,
including 398 survived funds and 236 dissolved funds as of December 2002. T-stats are heteroskedasticity consistent. The period covered is 01:1993-12:2002.
7.2 Non linearities in market exposure

In order to complete our analysis we perform two additional analysis. 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), Fang and Lai (1997), Harvey and Siddique (2000) and Dittmar

(2002). 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. 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. 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. The beta cubed reports the same

information for the co-kurtosis. The inclusion of the fund or the strategy may lead to

bigger tails (positive beta) or to lower tails (negative beta). Results are reported in Table

47. 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. Secondly, all the strategies considered have a negative impact on skewness

indicating that considered globally, market neutral strategies tend to lead to negatively

skewed distribution. Interestingly, the repartition of the beta squared shows that more

than 80% of the individual estimations have no significant impact on skewness. However,

the impact is negative for around 15% of the funds considered. 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. All in one, these results show

312
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). 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. Results are

reported in Table 48. 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). On the beta side, 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. 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. These results indicate some

presence of non-linearities even if there are not as high as one could expect.

7.2 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 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.

Table 49 indicates that there is no clear pattern in the average return column. The

figure indicates that the highest average returns are spread between decile 8, decile 1,

decile 10 and decile 6. 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
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,

decile 9 and decile 10. These results also indicate an interesting pattern. 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.

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.

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 distribution of the alpha indicates the same

pattern as the return distribution with higher figures for alpha’s not significantly different

from zero. This remark is particularly true, strongly exposed or negatively and

significantly exposed deciles (decile 1, 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, in other words that the market can explain a major part of their returns.

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. This result

is also true for other deciles but less precise.

314
 As we could expect, and by construction, the beta analysis indicates that the betas

are increasing over the decile. More interestingly, the beta distribution columns indicate a

clear pattern. No betas are significantly positive for the more weakly exposed deciles

(deciles 1 to decile 3). This logical result confirms our previous suggestion that low decile

funds profit from the bear market. Then, 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. 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. All over, this analysis suggests

that most market neutral funds are not significantly exposed to the market.

315
Table 49: Ex-post beta analysis

Return distribution Alpha distribution

Beta distribution (5%)
Average (5%) (5%)
S.D. Alpha Market R²adj
return
+ 0 - + 0 - + 0 -

Decile 1 1,14% 59% 39% 2% 4,90% 1,07% 39% 61% 0% -0,27 0% 54% 46% 0,1

Decile 2 0,89% 85% 15% 0% 2,60% 0,90% 70% 30% 0% -0,06 0% 96% 4% 0,01

Decile 3 0,96% 98% 2% 0% 1,60% 0,96% 83% 17% 0% -0,01 0% 100% 0% -0,01

Decile 4 0,81% 91% 9% 0% 1,20% 0,80% 83% 17% 0% 0,02 4% 96% 0% 0

Decile 5 1,00% 89% 9% 2% 1,90% 1,00% 81% 17% 2% 0,04 13% 87% 0% 0

Decile 6 1,07% 89% 9% 2% 2,00% 1,05% 83% 17% 0% 0,07 30% 70% 0% 0,03

Decile 7 0,96% 78% 20% 2% 2,80% 0,90% 59% 41% 0% 0,12 48% 52% 0% 0,07

Decile 8 1,24% 85% 13% 2% 3,90% 1,17% 67% 33% 0% 0,21 63% 37% 0% 0,1

Decile 9 1,00% 61% 39% 0% 4,70% 0,80% 26% 74% 0% 0,37 83% 17% 0% 0,16

Decile 10 1,11% 41% 57% 2% 8,00% 0,85% 26% 74% 0% 0,8 83% 17% 0% 0,31
The table reports the ex-post beta. Deciles are constructed on the basis of the market exposure of the individual funds. It reports the average return, the return
distribution at the 5% level. (+ are significantly positive, 0 not significantly different from 0 and – significantly negative), the standard deviation (S.D), the alpha,
alpha distribution at the 5% level (+ are significantly positive, 0 not significantly different from 0 and – significantly negative), the market exposure (the beta), the
beta distribution (+ are significantly positive, 0 not significantly different from 0 and – significantly negative) and the R²adj. Our MAR/CISDM database contains 634
individual market neutral funds, including 398 survived funds and 236 dissolved funds as of December 2002. t-stats are heteroskedasticity consistent. The period
covered is 01:1993-12:2002.
VIII Sub-period analysis

Since our analysis period covers a bull and a bear period, we will perform the same

analysis as in section VI in dividing the analysis period in 2 sub-periods. We chose the bull

and a bear market as defined by Capocci, Corhay and Hübner (2004). The cutting point

chosen for the identification of the up and down periods has been set at March 2000. 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

858.48 during the session of March 24, 2000. During the up period, the monthly index

return was positive in 70% of the months (52 out of 74) with an average yearly return of

19.4%. During the down period, the monthly index return was positive in 39% of the

months (12 out of 34) and the average yearly return was -16.9%. Those trends are

sufficiently strong to allow us to consider the whole sub-periods as, respectively, bullish and

bearish without having to use a complex rule to separate bullish, bearish and neutral months

since these rules would obviously not match the ones used by fund managers for their

market timing decisions.

8.1 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.

317
 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. 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. 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. Moreover, the

R²adj are very low indicating that the market factor cannot explain the returns. More

precisely, 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. 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 second part of Panel A of Table 50 reports the deciles’ regression analysis for the

bull period. 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. This means that

these funds do not offer absolute74 significantly positive returns over this period. On the

other side, the funds reported in the other deciles offered significantly positive absolute

returns over the period studied.

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.

318
 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. 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. Standard deviations are also in

almost all cases higher. Alphas indicate the same pattern. Interestingly, 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 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. On an absolute term however, the alphas generated by the market

neutral funds are significantly positive. The reasons explaining this result are however

completely different from the one explaining the alpha in the bull market. 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. On the other side, in the second case,

the bullish period, the funds significantly outperformed the market with a significant

exposure to it and comparatively high R²adj. This result is particularly important.

319
Table 50: Indices and deciles sub-period analysis

Panel A: 9/98-03/00

Average return S.D. Alpha Market R²adj

Market neutral
1,49% *** 1,11% 1,52% *** -0,01 -0,06
funds

- Arbitrage
1,40% *** 1,08% 1,48% *** -0,03 -0,04
funds

- Long/short
1,72% *** 1,34% 1,65% *** 0,03 -0,05
funds

- Mortgage
0,73% 2,52% 1,04% *** -0,12 -0,01
backed funds

- No sub-
0,64% *** 1,16% 0,74% *** -0,04 -0,03
strategy

Decile 1 -0,36% 0,03 -0,07% -0,12 -0,03

Decile 2 0,42% 1,87% 0,41% 0 -0,06

Decile 3 0,54% *** 0,60% 0,60% *** -0,03 -0,02

Decile 4 0,85% *** 1,01% 0,94% *** -0,03 -0,03

Decile 5 1,01% *** 1,41% 1,23% *** -0,09 0,03

Decile 6 1,24% *** 0,96% 1,18% *** 0,02 -0,04

Decile 7 1,46% *** 1,39% 1,52% *** -0,03 -0,05

Decile 8 1,62% *** 1,41% 1,58% *** 0,02 -0,06

Decile 9 2,60% *** 1,39% 2,53% *** 0,03 -0,05

Decile 10 4,27% *** 2,33% 4,04% *** 0,09 -0,03

320
Panel B: 4/0-12/02

Average
S.D. Alpha Market R²adj
return

Market neutral
0,71% *** 0,77% 0,84% *** 0,09 *** 0,35
funds

- Arbitrage
0,75% *** 0,70% 0,85% *** 0,07 *** 0,22
funds

- Long/short
0,65% *** 1,08% 0,82% *** 0,11 *** 0,29
funds

- Mortgage
0,87% *** 0,88% 0,95% *** 0,06 *** 0,1
backed funds
- No sub-
0,69% *** 1,31% 0,81% * 0,14 0,11
strategy

Decile 1 -0,59% 3,37% 0,03% 0,43 *** 0,43

Decile 2 -0,15% 0,99% -0,03% 0,08 *** 0,15

Decile 3 0,48% *** 0,67% 0,58% *** 0,07 *** 0,27

Decile 4 0,40% ** 1,06% 0,65% *** 0,17 *** 0,67

Decile 5 0,54% *** 0,52% 0,61% *** 0,05 *** 0,21

Decile 6 0,61% *** 0,92% 0,78% *** 0,11 *** 0,42

Decile 7 0,83% *** 0,67% 0,83% *** 0 -0,03

Decile 8 0,97% *** 0,80% 1,06% *** 0,06 ** 0,13

Decile 9 1,17% *** 1,06% 1,26% *** 0,07 * 0,08

Decile 10 1,97% *** 1,42% 2,06% *** 0,06 0,02

This table reports the market exposure analysis over sub-periods. Each Panel presents the average
return, the standard deviation (S.D), the alpha, the market exposure (the beta) and the R²adj. 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, including 398 survived funds and 236 dissolved funds as of December 2002. ***
Significant at the 1% level, ** Significant at the 5% level and * Significant at the 10% level. t-stats are
heteroskedasticity consistent. Panel A presents the bullish period (09:1998-03:2000), Panel B
presents the bearish period (04:2000-12:2002).

321
 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. 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. 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). 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.

Table 51 reports the same results based on the decile classification from Carhart’s

(1997). The average return and standard deviation column indicates that the top and bottom

decile funds, the most volatile one has not offered significantly positive return over the

period covered. This result is confirmed by the regression results. No alpha is significantly

positive when the market impact is taken into account. Interestingly, no deciles but decile 9

are exposed to the market. Most deciles are even negatively (but not significantly) exposed

to the equity market. The R²adj are low. 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. Moreover, this strong bull

market has a negative (but not significant) impact on the performance.

322
 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, the other deciles offer a

significantly positive alpha. These results are stronger than before, confirming that volatile

funds cannot create persistent value over time. In absolute term, top funds however offered

better (but not significantly positive) alphas than bad ones. Bottom and particularly top

deciles had the highest volatility confirming our previous results. Over the bearish period, no

decile but the one containing last year worst performing funds have been significantly

exposed to the market. Interestingly, the alpha of this decile is significantly positive whereas

its average return was not. This result can be explained by the fact that funds in this decile

were significantly exposed to the market. In fact, the average return is lower than the alpha.

Since the market went down over the period studied, 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. The spread between decile 1 and decile 2 is

significantly negative confirming this remark.

323
Table 51: Lagged 12-month decile sub-period analysis

Panel A: 9/98-03/00

Average return S.D. Alpha Market R²adj

Decile 1 0,40% 1,09% 0,17% 0,02 -0,05

Decile 2 0,56% * 1,28% 0,15% -0,01 -0,05

Decile 3 0,46% ** 0,88% 0,15% -0,03 -0,03

Decile 4 0,28% ** 0,57% 0,12% -0,03 -0,02

Decile 5 0,43% *** 0,70% 0,18% 0,04 0

Decile 6 0,35% ** 0,74% 0,15% -0,05 -0,01

Decile 7 0,44% * 1,12% 0,11% -0,05 -0,05

Decile 8 0,59% ** 1,14% 0,26% -0,01 -0,06

Decile 9 1,02% ** 1,86% 0,50% 0,11 ** 0,11

Decile 10 1,40% * 3,14% 0,57% 0 -0,06

1-10 spread 1,00% 2,93% 0,41% -0,02 -0,06

1-2 spread 0,16% 1,02% -0,02% -0,03 -0,01

9-10 spread 0,38% 2,00% 0,07% -0,11 0,1

324
Panel B: 04/00-12/02

Average return S.D. Alpha Market R²adj

Decile 1 0,65% 2,71% 0,83% *** 0,09 ** 0,11

Decile 2 0,53% ** 1,32% 0,92% *** 0 -0,03

Decile 3 0,37% *** 0,81% 0,76% *** 0,02 -0,02

Decile 4 0,45% *** 0,79% 0,53% *** 0,01 -0,03

Decile 5 0,56% *** 0,60% 0,67% *** 0,01 -0,03

Decile 6 0,67% *** 0,62% 0,73% *** 0,01 -0,02

Decile 7 0,72% *** 0,78% 0,93% *** 0,04 0,00

Decile 8 0,82% *** 0,64% 0,96% *** 0,03 -0,01

Decile 9 1,20% *** 0,83% 1,33% *** 0,01 -0,03

Decile 10 0,53% 2,50% 1,77% *** 0,13 0,01

1-10 spread -0,13% 3,31% 0,94% * 0,03 -0,03

1-2 spread -0,12% 2,05% 0,09% -0,09 ** 0,12

9-10 spread -0,67% 2,23% 0,44% 0,11 0,03

The table reports the market exposure using deciles constructed on the lagged-12 month performance (see
Carhart 1997). It reports the average return, the standard deviation (S.D), the alpha, the market exposure
(the beta) and the R²adj. Our MAR/CISDM database contains 634 individual market neutral funds, including
398 survived funds and 236 dissolved funds as of December 2002. *** Significant at the 1% level, **
Significant at the 5% level and * Significant at the 10% level. t-stats are heteroskedasticity consistent. Panel
A presents the bullish period (09:1998-03:2000), Panel B presents the bearish period (04:2000-12:2002).

325
8.2 Individual funds

Table 52 reports the sub-period analysis obtained for individual funds.75 As one could

expect, 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. The alpha column and distribution reports the same

pattern. The market exposure however indicates differences. Whereas it was on the average

positive over the whole period with one third of the beta being significantly positive, almost

all betas are not significantly different from zero over the bull market. This result indicates

that individual market neutral funds are not exposed to the market in bull markets. The

R²adj obtained confirms this result.

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. For

mortgage backed funds, 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).

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.

This lead us to remove 200 funds (90 arbitrage funds, 96 long/short funds, 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, 138

long/short funds, 29 mortgage backed securities funds and 2 no sub-strategy funds) for the 9/98-3/00

period analysis. This lead remove 125 funds (60 arbitrage funds, 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, 162 long/short funds and 14 mortgage backed securities funds) for the 4/00-12/02 period analysis.

326
for some strategy (mortgage backed funds for example) indicating that these funds did not

short the market. 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. On the other side when the market was down they were able

to create positive alpha but are more exposed to the market. On the other side, regarding

market exposure, in bear market, market neutral managers considered as a whole seemed

more to be exposed to the market, on the long on or the short side. Interestingly, market

neutral funds were significantly and negatively exposed to the market only during bear

market.

327
Table 52: Sub-period individual funds results

Return distribution Alpha distribution

Beta distribution (5%)
Average (5%) (5%)
Panel A: 9/98-03/00 S.D. Alpha Market R²adj
return
+ 0 - + 0 - + 0 -

Market neutral funds 1,57% 58% 42% 0% 3,76% 1,65% 61% 38% 0% -0,03 6% 92% 2% 0,01

- Arbitrage funds 1,52% 77% 23% 0% 2,38% 1,55% 84% 16% 0% -0,01 3% 94% 3% 0

- Long/short funds 1,81% 23% 77% 0% 4,55% 1,87% 21% 79% 0% -0,02 6% 94% 0% 0,02

- Mortgage backed funds 0,83% 38% 62% 0% 4,12% 1,10% 65% 35% 0% -0,11 0% 100% 0% 0,02

- No sub-strategy 0,64% 100% 0% 0% 1,16% 0,74% 100% 0% 0% -0,04 0% 100% 0% -0,03

 Return distribution Alpha distribution
Beta distribution (5%)
Average (5%) (5%)
Panel B: 4/00-12/02 S.D. Alpha Market R²adj
return
+ 0 - + 0 - + 0 -

Market neutral funds 1,57% 58% 42% 0% 3,76% 1,65% 61% 38% 0% -0,03 6% 92% 2% 0,01

- Arbitrage funds 1,52% 77% 23% 0% 2,38% 1,55% 84% 16% 0% -0,01 3% 94% 3% 0

- Long/short funds 1,81% 23% 77% 0% 4,55% 1,87% 21% 79% 0% -0,02 6% 94% 0% 0,02

- Mortgage backed funds 0,83% 38% 62% 0% 4,12% 1,10% 65% 35% 0% -0,11 0% 100% 0% 0,02

- No sub-strategy 0,64% 100% 0% 0% 1,16% 0,74% 100% 0% 0% -0,04 0% 100% 0% -0,03

The table reports the individual fund market exposure. It reports the average return, the return distribution at the 5% level. (+ are significantly positive, 0 not
significantly different from 0 and – significantly negative), the standard deviation (S.D), the alpha, alpha distribution at the 5% level (+ are significantly positive, 0 not
significantly different from 0 and – significantly negative), the market exposure (the beta), the beta distribution (+ are significantly positive, 0 not significantly different
from 0 and – significantly negative) and the R²adj. Our MAR/CISDM database contains 634 individual market neutral funds, including 398 survived funds and 236
dissolved funds as of December 2002. t-stats are heteroskedasticity consistent. Panel A presents the bullish period (09:1998-03:2000), Panel B presents the bearish
period (04:2000-12:2002).
8.3 Ex-post beta sub-period analysis

Table 53 reports the beta ex-post analysis over the 2 sub-periods considered.

Panel A reports higher average returns for the bull market period compared to the whole

period but fewer of these returns are significantly positive. The standard deviations of the

returns are lower for the bull market compared to the global period. The regression

results report higher alphas than previously for all deciles but fewer of them are

significantly positive. 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. Fewer individual betas are significantly positive over the bull market

conditions compared to the whole period studied. The R²adj are close to zero for all

deciles except the high beta decile. 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 B of Table 53 indicates that over the bear market period covered, 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. The standard deviation is close to

those obtained over the whole period and higher than those obtained over the bull

market period. Alphas are lower than over the whole period and fewer of them are

significantly positive. The market exposure increases logically by construction with

extreme’s beta value higher (respectively lower) for high (respectively low) beta funds.

This confirms our previous results that in the bear market period, market neutral fund’s

beta are higher in absolute term. The beta distribution confirms this result. 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.

330
Table 53: Ex-post beta sub-period analysis

Return distribution Alpha distribution

Beta distribution (5%)
Average (5%) (5%)
Panel A: 9/98-03/00 S.D. Alpha Market R²adj
return
+ 0 - + 0 - + 0 -

Decile 1 1,69% 15% 85% 0% 9,10% 3,03% 19% 81% 0% -0,54 0% 96% 4% 0,04

Decile 2 1,66% 27% 73% 0% 5,20% 2,24% 50% 50% 0% -0,23 0% 100% 0% 0,02

Decile 3 1,66% 50% 50% 0% 3,50% 1,99% 81% 19% 0% -0,13 0% 92% 8% 0,02

Decile 4 1,16% 62% 38% 0% 2,60% 1,36% 69% 31% 0% -0,08 0% 96% 4% 0

Decile 5 1,32% 96% 4% 0% 1,40% 1,42% 100% 0% 0% -0,04 0% 100% 0% -0,02

Decile 6 1,37% 85% 15% 0% 1,60% 1,40% 81% 19% 0% -0,01 0% 100% 0% -0,05

Decile 7 1,01% 69% 31% 0% 2,30% 0,96% 69% 31% 0% 0,02 0% 100% 0% -0,05

Decile 8 1,34% 69% 27% 4% 2,30% 1,20% 69% 27% 4% 0,06 4% 96% 0% -0,01

Decile 9 1,89% 50% 50% 0% 3,90% 1,47% 50% 50% 0% 0,17 12% 88% 0% 0,01

Decile 10 2,68% 50% 50% 0% 5,80% 1,44% 23% 77% 0% 0,5 46% 54% 0% 0,16
 Return distribution Alpha distribution
Beta distribution (5%)
Average (5%) (5%)
Panel B: 04/00-12/02 S.D. Alpha Market R²adj
return
+ 0 - + 0 - + 0 -

Decile 1 1,50% 36% 64% 0% 5,10% 0,95% 27% 73% 0% -0,37 0% 39% 61% 0,17

Decile 2 0,78% 61% 39% 0% 2,20% 0,63% 33% 67% 0% -0,11 0% 76% 24% 0,06

Decile 3 1,04% 91% 9% 0% 1,40% 0,98% 85% 15% 0% -0,04 0% 82% 18% 0,02

Decile 4 0,88% 91% 9% 0% 1,10% 0,87% 85% 15% 0% 0 0% 100% 0% -0,03

Decile 5 0,70% 91% 9% 0% 1,00% 0,73% 91% 9% 0% 0,01 0% 100% 0% -0,02

Decile 6 0,79% 76% 24% 0% 1,50% 0,84% 76% 24% 0% 0,03 9% 91% 0% 0

Decile 7 0,86% 76% 24% 0% 1,80% 0,96% 74% 26% 0% 0,07 50% 50% 0% 0,06

Decile 8 0,73% 53% 47% 0% 2,00% 0,93% 56% 44% 0% 0,14 71% 29% 0% 0,13

Decile 9 0,33% 3% 94% 3% 4,40% 0,75% 29% 68% 3% 0,28 47% 53% 0% 0,13

Decile 10 -0,55% 0% 94% 6% 8,10% 0,76% 24% 70% 6% 0,88 88% 12% 0% 0,36
The table reports the individual fund market exposure. It reports the average return, the return distribution at the 5% level. (+ are significantly positive, 0 not
significantly different from 0 and – significantly negative), the standard deviation (S.D), the alpha, alpha distribution at the 5% level (+ are significantly positive, 0 not
significantly different from 0 and – significantly negative), the market exposure (the beta), the beta distribution (+ are significantly positive, 0 not significantly different
from 0 and – significantly negative) and the R²adj. Our MAR/CISDM database contains 634 individual market neutral funds, including 398 survived funds and 236
dissolved funds as of December 2002. t-stats are heteroskedasticity consistent. Panel A presents the bullish period (09:1998-03:2000), Panel B presents the bearish
period (04:2000-12:2002).
IX Conclusion

This study focuses on the neutrality of market neutral funds. 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.

In the first descriptive part of the analysis, we will focus on the literature, the

interest of the study, on the descriptive statistics, attrition and birth rates and on the

presence of survivorship bias in the data. The core of the study is based on a

methodology that use classical exposure measures like the beta in an original way. 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.

The results obtained using sub-strategy indices indicate the beta obtained are low

on an absolute term but significantly positive. Decile analysis indicates that top and worst

performing funds (over the whole period) have the highest market exposure. 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. In all cases, 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 individual fund analysis results indicate that on the average, 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). Then, the ex-post beta analysis indicates that negatively

exposed and strongly positive exposed funds are more volatile than market neutral

333
funds. Middle decile funds returns are the only real market neutral funds. Some funds are

strongly positively or negatively exposed to the market over the period studied, but most

market neutral funds are not significantly exposed to the market.

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. This study

stresses the importance of considering individual funds when performing market neutral

empirical analysis. This result could be explained by 2 reasons. First, the aggregation of

funds in indices leads to an increase in the exposure to the equity market. 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. 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, the funds

significantly exposed could bias the results.

The sub-period analysis also reports very interesting results. First, 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. Secondly, on a strategy and sub-strategy basis, over the bullish period no

index or decile has been significantly exposed to the market. Over the bearish period, all

but the best performing deciles have been significantly exposed to the market but they

all, except the best performing funds create significant alpha. 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.

Thirdly, arbitrage funds and mortgaged-backed funds perform better in bull market

334
without being exposed to the market. On the other side when the market is down they

are able to create positive alpha but are exposed to the market. On the other side,

regarding market exposure, in bear market, market neutral managers considered as a

whole tend more to be exposed to the market, on the long or the short side.

Interestingly, some market neutral funds are significantly negatively exposed to the

market only during bear market.

Our analysis leads to the conclusion that most market neutral funds are not

significantly exposed to the equity market, but that they tend to be more exposed during

bear market than during bull market without being negatively impacted.

335
Part 3: Hedge Funds as
Diversification Tools
Diversifying using Hedge Funds: A Utility-

Based Approach

Daniel P.J. 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,

Maastricht University, and Luxembourg School of Finance, University of

Luxembourg.

Capocci, Daniel, Frédéric Duquenne and Georges Hübner, 2006, Diversifying using Hedge

Funds: A Utility-Based Approach, Working paper, HEC-ULG Management School

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. This new efficient frontier is based on the Taylor’s

decomposition of Bell’s (1988, 1995) utility function. 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. Estimations

are performed for various levels of risk aversion, various levels of allocation to the risky

asset and by separating different kind of hedge fund investments.

339
Diversifying using Hedge Funds: A Utility-Based

Approach

Introduction

Hedge funds have been widely studied since 1997. Almost all areas of the

industry are under research. From performance to risk analysis, from style analysis to

their use as a diversification tool, the interest in hedge fund analysis is growing and the

literature more and more explain the specificities of these “dark angels”.76

Hedge fund studies can be classified in four global categories. The first one, which

is by large the richest in quantity, encompasses papers that focus on hedge fund

performance (see for example Ackermann et al., 1999; Agarwal and Naik, 2000; Amin

and Kat, 2003; Liang, 2001; Edwards and Caglayan, 2001; Capocci and Hübner, 2004;

Agarwal and Naik, 2004; Malkiel and Saha, 2005). 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, 1997; Brown et al., 2001;

Brealy and Kaplanis, 2001, Brown and Goetzmann, 2001, Liang 2001; Liang 2003). 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. “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.

340
example Fung and Hsieh, 1997; Schneeweis and Spurgin, 1998; Liang, 1999; Agarwal

and Naik; 2000). Finally, other studies analyse miscellaneous aspects of the hedge fund

industry. This category includes the “other studies” focusing on the risks involved in

hedge fund investing (see Jorion, 2000 or Lo, 2001); the presence of bias in the data

(see for example Liang, 2000; Fung and Hsieh, 2000 and Malkiel and Saha, 2005); or

the difference between the indices available (see for example Amenc and Martellini,

2002 and Fung and Hsieh, 2002a).

The current study belongs to the third stream. 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. The underlying idea of our

analysis is the same as the spanning tests developed by Huberman and Kandel (1987).

The main differences introduced by our methodology are twofold. First, 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, 1995) utility function, that combines

a linear and an exponential form. Second, 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, but study a wide range of

risk measures corresponding to different investor profiles. 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.

The paper is organised as following. 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. Section III

reports the results and Section IV concludes the paper.

341
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).

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. More recently Goetzmann et al. (2006)

show that statistics can yield attractive statistics in terms of classical measures like the

Jensen’s alpha using a simple rebalancing strategy. These authors conclude that

manipulation-proof performance measures can be completely characterized as the

weighted average of a utility-like function, which is exactly the focus of our study.

The utility function concept is widely used to model the process of decision

making. It offers a lot of applications in various disciplines including finance,

microeconomics and operational research, etc. In finance, portfolio selection theory

examines how a rational investor behaves when choosing his investments in a

framework of uncertainty about his futures outcomes. Many authors have devoted their

time searching for alternative ways of modelling an investor’s preference and deriving

the composition of his ideal portfolio. Based on its mean-variance approach, 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,

1997). In actual applications, and particularly in the case of hedge funds, returns are

typically not normally distributed and utility functions have a higher order than

quadratic. Levy and Markowitz (1979) justify the practice of using mean-variance

342
analysis by showing that mean-variance analysis can be regarded as a second-order

Taylor series approximation of standard utility functions.

The utility function concept used in the framework among uncertain investment

outcomes is sustained by a strong theoretical justification. Von Neuman and

Morgenstern (1947) proved that, when we accept some hypothesis about the concept of

rational choice, a decision maker can be thought of as comparing alternatives based on

his expected utility. 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.

There are various classes of utility functions. The most common are the

exponential, logarithmic, power and quadratic specifications. Each has specific

properties, drawbacks and advantages (see Spanier and Oldman, 1987 for a detailed

presentation of more than 400 utility functions). 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. 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. He will always

give more utility to a certain gain than to a random result with the

same expected return. This property involves that the second-order

derivative to be negative.

(c) Risk assets are not inferior goods.

343
 Classical utility functions like the exponential, logarithmic, power and quadratic

do not offer all these desirable properties. The main issue with the exponential utility

function is that the ARA77 is constant (and that its range of variation is the largest

possible, i.e. [0, +∞[) implying that investors invest constant amounts in risky assets as

their wealth increases, but in decreasing proportional amounts. The logarithmic utility

function and the power utility function have a constant RRA and are isoelastic, thus an

investor with a non-constant RRA cannot be characterized. Cohn et al. (1975), Friend

and Blume (1975) and Morin and Suarez (1983) find empirical support for decreasing

RRA. Finally, the quadratic utility function involves satiety and growing risk aversion.

Considering these elements, it seems that none of the popular utility functions

listed above has all the desirable properties. This is not the case with Bell’s function. Bell

(1988, 1995) introduces a new kind of utility function that displays appealing

characteristics and permits easy interpretation of alternatives in terms of risk and

return. In other words, Bell’s utility function is such that its expected value can be

expressed as a function of the wealth increase r, and the risk R, of an alternative x.

x ) ⎦ = f [r ( ~
E ⎣U (W0 , ~ x ), W0 ]
x ), R( ~ ( 13 )

77
ARA stands for Absolute Risk Aversion, RRA for Relative Risk Aversion. ARA is the ratio of the

second-order derivative to the first order one. See Arrow (1971) for more details.

344
 Bell’s function is a combination of a linear and an exponential utility function and

more commonly known as “linear plus exponential”, aka linex functions:

U (W ) = W − be − cW (14)

with b, c ≥ 0
~
The expected utility in the case of an additive random wealth x is then equal to

E ⎣U (W0 , ~ [
x ) ⎦ = E W0 + ~
~
] x ) − be − cW e − cr ( x ) − be − c [W0 + r ( x ) ] R( ~
x − e − aW e − ax = W(0 ) + r ( ~
~ ~
x)
(15)

~
where the measure of wealth increase r (x ) is the expected return of the

~
alternative E (x ) .

345
 The measure of risk can be isolated from this expression:

R( ~
1
[ ~
x ) = ln E (e −c ( x − x ) )
c
]
(16)

The measure of risk incorporates one parameter c, which is specific to the

investor. Parameter c depends on the distribution of return only, and particularly on the

tail of the distribution: investors with a high c place more emphasis on the possibility of

bad outcomes. On the other side, parameter b does control the degree of aversion to

risk, as a weight assigned to risk by the investor in his utility function. Thus, parameter

c constitutes relative riskiness and b aversiveness to that riskiness. Bell (1988, 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.

346
The function provides a new form of second order approximation that has very satisfying
properties.

(1) It is an increasing and concave function of wealth.

(2) The two parameters b and c leaves flexibility to match personal attitudes.

(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
(2)
largest possible.

(1) The RRA (relative risk aversion) is decreasing then increasing with respect to W.

With this function, if two alternatives are judged to be equivalent at two different
(2)
levels of wealth, they will be judged equivalent at all levels of wealth.

(1) It is a non-polynomial function as it includes an exponential.

(2) It is a proper utility function as the successive derivatives alternative in sign .

347
1.2 Spanning

Altough our methodology is not based on spanning, the underlying hypothesis is

similar, 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.

When investors condition their investment choices on the first two moments of

the distribution of returns, they want to know whether adding risky assets will improve

the minimum-variance frontier. This issue is addressed by the notion of spanning

introduced by Huberman and Kandel (1987). 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. Mean-variance spanning is attained if the new

frontier does not significantly differ from the old one.

There are two approaches to test for spanning: the regression approach and the

stochastic discount factor approach. Huberman and Kandel (1987) propose a

multivariate regression-based approach to test of this hypothesis. 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, “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. Ferson et al. (1993) generalise the test and integrate the issue of

heteroskedasticity in the testing while De Roon et al. (2001) test for spanning with

futures contracts and non-traded assets and Kan and Zhou (2001) study tests of mean-

variance spanning in detail.

One should estimate the linear relation that exists between the returns of the

existing portfolio and of the additional securities considered. The null hypothesis of

absence of additional diversification with the additional portfolio (spanning) involves that

348
the intercept of the regression and the slope coefficient are not significantly different

from 0 and 1 repsectively. 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. If the conditions are both satisfied, the minimum-variance

frontier that includes the added securities does not statistically differ, from a mean-

variance perspective, from the original portfolio the investors including new securities

cannot increase their utility by adding the new securities considered.

The null hypothesis can be tested in three ways. Huberman and Kandel (1987)

provide a likelihood ratio test of spanning and derive its finite sample distribution under

normality assumptions. The Wald test and the Lagrange multiplier test are two

alternative asymptotic tests described in detail by Kan and Zhou (2001). Note all three

tests are increasing transformation of each other. When the residuals exhibit conditional

heteroskedasticity, the three classical three tests will no longer be asymptotically Chi²

distributed under the null hypothesis. In this case, Hansen’s (1982) GMM is the common

viable alternative that relies on the moment conditions of the model. These results have

been stressed by Ferson et al. (1993) and Kan and Zhou (2001). 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.

Generalised methods of moments versions of the tests have been developed by

Ferson et al., 1993 by considering non-normality and heteroskedasticity. They consider

non-normality and heteroskedasticity. 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; two bootstrapped versions

of these Wald tests; a version of the two-step Wald test that uses a finite sample

corrected estimate of the variance of the two-step GMM estimator; the LM test; and

349
three criterion-based tests that have recently been proposed (see Kan and Zho, 2001).

The corrected two-step Wald test performs similarly to the standard one-step Wald test,

while the bootstrapped one-step Wald test, the LM test, and a simple criterion-difference

test can provide more reliable finite sample inference in some cases. Following these

results and Kan and Zho (2001) we will focus on the GMM Wald test.

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). 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, 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. If the rejection is due to the first test, then the two tangency portfolios are very

different. If the rejection is due to the second test, one can consider the two global

minimum variance portfolios are very different.

The second approach, the stochastic discount factor approach has been

developed by DeSantis (1993) and Bekaert and Urias (1996). Bekaert and Urias (1996)

exploit the duality of Hasen-Jagannathan (1991) bound and the mean-variance frontier.

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 null hypothesis of Huberman and Kandel (1987) can be

proved (see Kan and Zhou, 2001) and is equivalent to the null of Bekaert and Urias

(1996). 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, but the null hypothesis remains equivalent.

350
 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 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. The author further

observes that funds of hedge funds offer diversification for mean-variance investors.

351
II Methodology and data

Portfolio selection within a mean-variance-skewness framework has been

considered by Lai (1991), Chunhachinda et al. (1997), Sun and Yan (2003), Chan et al.

(2003), Harvey et al. (2004). Only recently, however, have studies focused on a four-

moment framework (see, Guidolin and Timmermann, 2002; Harvey et al., 2004).

Guidolin and Timmermann (2002) use a Taylor series expansion of the investors’

objective function to determine the optimal portfolios. They consider an m-th order

Taylor series expansion of a generic utility function

U (Wt +T ;θ ) .

Davies et al. (2004) analyse the impact of inserting/deleting hedge funds in a

hedge fund portfolio focusing on co-moments while Davies et al. (2005) incorporate

investor preferences for higher moments of a return distribution into a polynomial goal

programming optimization model. Jurczenko and Maillet (2006) build a four-moment

efficient frontier for portfolio including hedge funds using a non-parametric

methodology. These studies all focus on portfolio selection but do not specifically study

the spanning properties of a particular portfolio or a class of assets. These approaches to

multi-moment analysis are highlighted below before we present our utility-based

methodology using a Taylor expansion.

352
2.1 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. In our model, we combine and

adapt the tools presented in the previous section. First, we do not make the assumption

of quadratic or power utility function but consider the Bell (1988, 1995) utility function.

Second, 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. Moreover, 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.

Considering the set of Von Neumann-Morgenstern’s axioms about rationale

choice, we set that investors maximise their expected utility function at each period to

derive their optimal investment choice. This is the most general and commonly accepted

framework of portfolio selection theory referred as the

E (U ) approach. The problem with

this approach is that it is difficult to implement in practice because the information

requirements are strict. Investors must be able to provide a detailed specification of

their tastes, represented by a specific utility function as well as complete pattern of their

beliefs, described by a joint subjective probability distribution. As the returns may be

infinite, it is impossible to reliably define such a probability distribution and to calculate

the exact value of the expected utility. Moreover, maximizing expected utility among all

feasible portfolios is often a complex mathematical problem that requires a considerably

greater number of calculations.

353
 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. 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. These parameters correspond in most cases to the first k moments of

m1 , m 2, ,..., m k
the distribution: provided that they exist.

V k (m1 , m 2, ,..., 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. In this case there is

no approximation to be made. Amongst usual utility functions, the quadratic is the only

one satisfying this condition. The other more appropriate functions (exponential,

logarithmic, 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.

To isolate the impact of each moment, these non-polynomials are typically

expanded using a Taylor series approximation of the utility function U . This technique

consists in fitting a polynomial of degree k to U (W ) at one value (either 0 or W ).

354
 A first attempt to approximate E (U ) is to estimate the probability distribution by

its first two moments, leading to the mean-variance approximation (see for example

Tobin, 1969; Huberman and Kandel, 1987). However, 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., 1999; Fung and Hsieh, 2001; Agarwal and Naik, 2004;

Malkiel and Saha, 2005 and Kat and Miffre, 2005). Not only are hedge fund return

distributions asymmetric and leptokurtic, they also tend to display co-skewness and co-

kurtosis with the return of other assets classes, due to option-like features of alternative

investments (see Weisman, 2002, Goetzmann et al. 2003, Agarwal and Naik, 2004; and

Davies et al., 2004a).

The inclusion of higher moments in the efficient frontier analysis has been studied

for years. Several alternative approaches have been developed, yet no single conclusive

approach has emerged. As stated in Jurczenko and Maillet (2006): “we can distinguish

between primal and dual approaches for determining the mean-variance-skewness-

kurtosis efficient frontier.”

Primal approaches include the polynomial goal programming (PGP) approach to

determine the set of the mean-variance-skewness-kurtosis (see Davies et al., 2004).

There are two main shortcomings of this approach. The first one is that the allocation

problem solved cannot be precisely related to the expected utility function, since the

choice of the parameter used to weight the moment deviations is not related to the

parameters of the utility function. The second shortcoming is that the estimation does

not comply with the Pareto-optimal definition of an efficient portfolio frontier. Minimizing

deviations from the first four moments simultaneously only guarantees a solution that is

close to the mean-variance-skewness-kurtosis efficient frontier. Other authors tend to

analytically solve the mean-variance-skewness-kurtosis portfolio optimization problem

(see Athayde and Flôres, 2002 and Adcock, 2005). These approaches are partial since

355
they are mainly focused on one objective of the optimization program at the cost of the

others.

Dual approaches use Taylor series expansion of the investors’ objective functions

to determine the optimal portfolios (see Jondeau and Rockinger, 2005 and Jurczenko

and Maillet, 2006). 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). Moreover there is no rule in general for selecting the order of

truncation. Hlawitschka (1994) finds that adding terms to a Taylor-series expansion

does not necessarily improve the approximation, even when the series converges,

however, Ederington (1986) finds that extending the expansion to a fourth term

generally improves the approximation, and never worsens it.

We base our analysis on a dual approach since our utility function includes an

exponential, the first and main drawback of the approach does not apply and secondly,

we decide to select the mean-variance-skewness-kurtosis order of truncation as it seem

to be the most comprehensive.

2.2 Taylor approximation and risk measure

Following Loistl (1979) and Hlawitschka (1994) we approximate the distribution

on the basis of the first four moments. 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, as the descriptive statistics indicate, hedge fund returns may be

skewed and have fat tails. The integration of skewness and kurtosis will enable us to

take these statistics into account.

356
 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, then the Taylor series expansion is a finite series with n terms.

On the other hand, if f is non-polynomial, then the Taylor series expansion is an

infinite series (see Loistl, 1976):

∞
hj
f ( x + h) = ∑ f ( j)
( x)
j =0 j! (18)

When x is viewed as a random variable, the most common use of this formula is

the expansion of f around its mean argument x .

Now consider a utility function of the random variable U (W ) . The Taylor’s series

W = W0 (1 + R )
expansion of this function around the mean end-of-period wealth – where

R is the mean rate of return of the investment – equals:

(W − W ) 2 '' (W − W ) n n
U (W ) = U (W ) + (W − W )U ' (W ) + U (W ) + ... + U (W ) + ...
2 n! (19)

357
 An intuitive idea is that the more terms are kept in the expansion, the closer the

approximation will be to the expected utility value. If we truncate the series to keep the

first four terms, we have78:

E [U (W )] ≅ U (W ) + σ 2 (W )U '' (W ) + S (W )U ''' (W ) +
1 1 1
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:

c2 c3 c4
U (W ) = W + be − cW c(W − W ) − be − cW (W − W ) 2 + be − cW (W − W ) 3 − be − cW (W − W ) 4
2 6 24
(21)

with b, c >0

78
As stated by Guidolin and Timmermann (2002), 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.

358
 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 perform two sets of tests, 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, 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.

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. Practically, 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.

The difference between the capital market line determined in our case and the

classical one is the nature of the risk measure used. We can extract the risk measure

from equation (9).

Define
W0 as the initial value of the investors’ wealth and I as the global amount

invested in the risky asset. The total amount invested in the risk-free asset is
W0 − I . If

r
the risk-free rate gives a return equal to and the risky asset a return equal to m , as

we consider that everything that is not invested in the risky asset is invested in the risk-

free asset. If we define the risk premium as x = Θ − r , the utility function becomes:

359
 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 ) ) (24)

The Taylor series expansion of this expression around m yields:

− 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
 From (12), we get the risk measure
R x , for a standard unit of wealth (normalized

to)

1 1 1
R x = V x − CS x + C 2 K x
2 6 24 (26)

I
C≡c
where
W0 is the product of the (intrinsic) risk perception coefficient, c, and

the (wealth-related) proportion of risky investment

I /W0 .

This risk measure accounts for volatility, asymmetry and fat tails. The level of C

depends on the appetite for risk of the investor. Based on experiments on a sample of

40 students, we consider three representative levels. C=10 for a dynamic investor that

does not put much weight on the extreme risks, C=35 for an average investor and C=60

for a very protective investor who mostly fears the tail event of a distribution.

Our spanning test is performed as follows. For each frontier, we identify the

tangency portfolio, i.e. the efficient portfolio where the adapted Capital Market Line or

CML (defined, as in the CAPM, as the locus attainable combinations of the risk free rate

and this risky portfolio) and the efficient frontier are tangent, and draw the adapted CML

with the risk-free rate. 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
with hedge funds and one without) for that measure of risk. Our main challenge is to

estimate the slope of the adapted CML because, as risk is not measured with the

variance, this function is not a straight line.

The underlying idea is illustrated in Figure 11. 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.

Figure 12 illustrates the test performed. 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. Both called “average CML” are in dotted lines in

Figure 12. Secondly, 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. We plot point s C and D Plot points C and D at

another risk value Y.

In our example, 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, while at Y the test is partly

conclusive. The portfolio of hedge funds and mutual funds do significantly better than

mutual funds alone, but mutual funds alone do not significantly worse than hedge funds

plus mutual funds.

362
Figure 11: Illustration of the adapted Capital Market Line

Return
Capital Market
Line

Efficient Frontier

Risk Measure

2.4 Estimation methodology

The term “bootstrapping” due to Efron (1979), is an allusion to the expression

“pulling oneself up by one’s bootstraps” – in this case, 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. In the simpliest form of bootstrapping, one repeatedly

analyses subsamples of the data. Each subsample is a random sample with replacement

from the full sample.

363
Figure 12: Illustration of the test of the adapted Capital
Market Line

Return

B CML (HF+MF)
C CML95% (MF)

CML5% (HF+MF)

CML (MF)
D
A

Y X
Risk measure

Three sets of funds are taken from a database: hedge funds, equity mutual funds

and bond mutual funds described below. The algorithm that makes the calculations was

written in FORTRAN and C# and was compiled with the Microsoft.net framework.

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, 40 % equity mutual funds and 40 % bond mutual funds). All the funds

from the three sets are randomly selected, constructing portfolios each with ten funds

and we repeat the process as many times as required. For each portfolio the algorithm

determines two elements: 1) as a first step the CML generated from the funds included

364
in this portfolio and 2) a series of ten slopes for each CML, each slope corresponding to a

particular value of
Rt .

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.

At each step the algorithm implements an optimization problem with the required return

as objective function and the risk function as constraint. The curve is then estimated

with the locus of these risk/return pairs79.

The calculation of the objective is worked out in accordance with the classical

CAPM. The benchmark index return

rb for the beta calculations is S&P 500 and the risk

free rate Rf is 2.78%.80

E ( x0 ,K , x n −1 ) = R f + ( x0 β 0 + K x n −1 β n −1 )( Rm − R f )
(27)

n
rp = ∑ wi ri
for a portfolio i =1

The calculation of the various centred moments m p ,l appearing in the risk

function is carried out using the following general formula:

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.

80
This 2.78% 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).

365
 l! 1 T k1
m p ,l = ∑( k1 ,...,k n s.t . ∑ w1 ...wnk n (r1t − r1 ) k1 ...(rnt − rn ) k n
k1 +...+ k n =l ) k + ... + k T
t =1
1 n (28)

And the optimization problem is:

max E (rp )
w1,..., wn

1 1 1
s.t. R p ≡ m p , 2 − Cm p ,3 + C 2 m p , 4 = λRb
2 6 24
w1 + ... + wn = 1, wi ≥ 0∀i (29)

where λ is the target proportion of the (fixed) risk measure of the benchmark

portfolio Rb .

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, see Schittkowski, 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, and the

366
linearization of the constraints. The resulting quadratic problem is then solved for every

iteration.

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 “non-

feasible” intermediate points near the field of the constraints. Therfore, contrarily to

many methods where the constraints are verified at each iteration, the SQP method

requires only the final solution to respect the constraints.

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. Thus a set of risk values, summarized with values of

parameter λ, is selected within the whole range of the risk values. 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 )
∆R p
R p = λRb
in accordance with the above procedure. The slope is then given by .

In order to determine the reference p-value at the any confidence level, we

estimate the returns and slope of 2,000 CML (using bootstrapping), rank the slopes or

the returns depending on the case considered by decreasing values for each value of

risk, and select the level of confidence corresponding to the numerical p-values. We

perform 2,000 estimations for 5 different levels of lambda (5 different levels of allocation

to the risk free rate - 20%, 40%, 60%, 80% and 100%). Then, we perform the same

algorithm including various levels of hedge funds (10%, 20%, 50% and 100%) and

separate directional, undirectional and fund of funds that do not have the same

objectives and distribution characteristics. Practically, we estimate the 95% percentile

367
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.

81
We made the estimation for the 90%, 95% and 99% confidence levels but we only report the 95%

level. Complete tables are available.

82
We make the same estimation using the median. The results are not reported because they are

perfectly in line. Tables are available upon request.

368
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. For

hedge funds, we use the Hedge Fund Research, Inc database. Following Amin and Kat

(2003), we concentrate on the 431 funds that exist during the entire January 1996-May
83
2006 period. Note that our sample includes the period 1997-1998 which includes

crises in Asia and Russia and the subsequent collapse of LTCM. 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, we use data from Micropal. We

use monthly performance data from 2,418 surviving equity mutual funds and 1,308

surviving bond mutual funds. We report the descriptive statistics of our databases in

Table 54.

Table 54 indicates that the bulk of the hedge funds in the database are funds of

funds, equity hedge and CTAs (almost 60% on aggregate). The best performers over the

period under review have been directional funds led by emerging markets, equity non-

hedge and equity hedge managers providing net annualized returns of 21.8%, 18.2%

and 15.5% respectively. The lower return providers have been equity market neutral,

83
Obviously, this introduces survivorship bias. As this bias favors the performance of the surviving

funds, our conclusions are very likely to embellish the risk-return trade-off of hedge funds, and should

be considered with caution. There is no simple way to mitigate this problem though. Note, also that a

similar bias is likely to hold for our sample of mutual funds as well. 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. This would be the next step in such an analysis. A large number of

estimations are required and we therefore postpone this step.

369
merger arbitrage and convertible arbitrageurs that provided net returns of 7.85%,

9.34% and 9.47% respectively. In terms of volatility, as one could expect, a large

majority of the directional strategies have been the most volatile. Pure arbitrage

strategies have offered the most stable returns.

The minimum and maximum monthly returns range range from -25.4% for

emerging markets to -2.1% for convertible arbitrage and from 2.5% for relative value to

33.4% for short sellers respectively. Next, 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. The results interestingly show that even if

skewness reported per strategy is close to zero, there is a diversification effect and the

individual skewness reported has a large range, which is illustrated in Figure 13. These

results confirm that at least for some strategies (event driven, equity market neutral,

fixed income arbitrage, fund of funds and relative value) skewness tends to be negative.

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
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. Return Std. Dev. Sharpe Ratio

DIRECT. 245

DS 11 2,6% 12,90% 5,30% 2,07

EM 19 4,4% 21,90% 20,60% 0,96

ENH 20 4,6% 18,20% 12,70% 1,27

EH 81 18,8% 15,50% 9,40% 1,44

ED 25 5,8% 11,70% 7,80% 1,24

MAC 18 4,2% 13,20% 7,10% 1,59

SEC 13 3,0% 13,80% 10,60% 1,12

SS 1 0,2% 10,40% 36,10% 0,23

CTA 54 12,5% 13,30% 11,60% 0,97

MT 3 0,7% 14,20% 11,20% 1,09

UNDIR. 70

MA 8 1,9% 9,30% 3,70% 1,96

RV 12 2,8% 11,00% 3,10% 2,94

EMN 14 3,2% 7,90% 3,10% 1,91

FIA 17 3,9% 10,20% 5,10% 1,6

CB 19 4,4% 9,50% 3,30% 2,25

FOF 116 26,9% 11,70% 5,10% 1,88

GLB 431 100% 13,10% 6,00% 1,86

Panel A (continued): Hedge Fund Descriptive Statistics (1/1996-5/2006)

Min Max Skewness Kurtosis

DIRECT.

DS -4,80% 7,80% 0,01 0,01

EM -25,40% 19,00% 0,09 0,06

ENH -13,70% 8,00% 0,03 0,02

EH -9,50% 9,60% 0,04 -0,01

ED -10,20% 7,60% 0,02 0

MAC -3,40% 7,70% 0,01 0,01

SEC -14,90% 7,50% 0,04 -0,02

SS -21,80% 33,20% -0,06 0,19

CTA -6,90% 9,20% -0,02 0,02

MT -6,60% 11,60% 0,02 -0,01

UNDIR.

MA -5,20% 3,90% 0,02 0,02

RV -3,10% 2,60% 0,02 0,01

EMN -2,80% 3,10% 0,02 0,01

FIA -4,70% 4,80% 0,01 0,01

CB -2,10% 3,40% 0,01 0,01

FOF -5,20% 6,60% 0,02 0,01

GLB -5,70% 5,80% -0,16 1,02

372
Panel B: Mutual Fund Descriptive Statistics (1/1996-5/2006)

N Mean Ann. Return Ann. Std. Dev. Sharpe Ratio (2%)

Eq MF 2 418 9,60% 19,30% 0,39

Bd MF 1 308 4,85% 7,68% 0,37

Min Max Skewness Kurtosis

Eq MF -16,5% 16,5% -0,22 0,98

Bd MF -4,2% 7,2% 0;62 0,51

This table reports the descriptive statistics of our database. Our database covers 431 hedge funds, 2418
equity mutual funds and 1308 bond funds over the January 1996 to May 2006 period. CB = convertible
arbitrage, DS = distressed securities, EM = emerging markets,EqLong = equity long biaised and equity long
only, ELS= equity long/short, EqShort = equity short, ED = event driven, MAC = macro, Multi = mutli-strategy,
SEC = sector funds, Others = other strategies, EMN = equity market neutral, MA = merger arbitrage, StatArb =
statistical arbitrage, FI = fixed income strategies, FoF = fund of funds and HF = global database. DIR.=
directional strategies; UNDIR.=undirectional strategies.

373
 On the mutual fund side, the net mean annualized return is at 9.6% for equity

funds with volatility at 19.3%. Such returns are relatively low when compared to hedge

funds with a high volatility. The skewness figures (more particularly the minimum and

maximum skewness) indicate a positive skewness trend. Kurtosis column indicates the

presence of fat tails. Bond mutual funds offer much lower returns with a relatively limited

volatility, positive skewness and positive kurtosis.

These results indicate that hedge funds seems to offer attractive risk-return trade-

offs 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.

374
Figure 13: Repartition of the skewness of individual funds

0
ng

ac

A
EM

ti

F
b
S

ED
F

FI
B

C
or

er
ul

Ar

Fo
H

M
EL
C

EM
SE
M
Lo

Sh

th

at
Eq

O
Eq

St
-1

-2

-3

-4

This figure reports the repartition of the skewness of individual funds. Our database
covers 431 hedge funds, 2,418 equity mutual funds and 1,318 bond funds over the
January 1996 to May 2006 period. CB = convertible arbitrage, DS = distressed securities,
EM = emerging markets,EqLong = equity long biaised and equity long only, ELS= equity
long/short, EqShort = equity short, ED = event driven, MAC = macro, Multi = mutli-
strategy, SEC = sector funds, Others = other strategies, EMN = equity market neutral,
MA = merger arbitrage, StatArb = statistical arbitrage, FI = fixed income strategies, FoF
= fund of funds and HF = global database. DIR.= directional strategies;
UNDIR.=undirectional strategies.

375
Figure 14: Repartition of the kurtosis of individual funds

12

10

0
s

N
ti
ng

ac

A
F

EM

F
b
ED
B

FI
or

er
ul

Ar

Fo
H

M
C

EL

EM
SE
M
Lo

Sh

th

at
-2
Eq

O
Eq

St
-4

-6

This figure reports the repartition of the kurtosis of individual funds. Our database covers
431 hedge funds, 2,418 equity mutual funds and 1,318 bond funds over the January
1996 to May 2006 period. CB = convertible arbitrage, DS = distressed securities, EM =
emerging markets,EqLong = equity long biaised and equity long only, ELS= equity
long/short, EqShort = equity short, ED = event driven, MAC = macro, Multi = mutli-
strategy, SEC = sector funds, Others = other strategies, EMN = equity market neutral,
MA = merger arbitrage, StatArb = statistical arbitrage, FI = fixed income strategies, FoF
= fund of funds and HF = global database. DIR.= directional strategies; UNDIR.=
undirectional strategies.

376
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.

We start this section by explaining how we determine the levels of the p-values,

that is, the level over which the inclusion of hedge funds in a portfolio enables the

investor to significantly improve his satisfaction. 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.

4.1 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 first have to determine the significance levels,

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 2,000 capital market line for each level of lambda, by ranking them to

extract the level of confidence (95% corresponding to the 1900th estimation).

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. 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.

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. The p-values

377
for the slope analysis are estimated for mutual fund portfolios only and are then ranked.

For inverse relations, 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.

4.2 Including hedge funds in mutual funds portfolio

As previously stated, we consider three categories of investors: average,

protective and progressive. We report the results obtained for the three categories of

investors separately. For each estimation we report a Table that provides comparison and

figures that illustrate the results.

a. Average investor

The first line of Table 56 reports the p-value for the 95% confidence level for a

portfolio of mutual funds only. Panel A reports the mean returns when only directional

hedge funds are added; Panel B the same results when undirectional hedge funds are

added. Finally Panel C the mean returns when funds of hedge funds are added to the

portfolio.

The p-value line indicates that 95% of the time, a portfolio consisting of equity

mutual funds and bond mutual funds would have offered a return lower than 5.4%

annualized if 20% of the assets were allocated to the risky asset. For an allocation of

40% to the risky asset, the mean annualized return would have been lower than 8.3%

annualized 95% of the time. A full exposure to the risky asset would have yield returns

as high as 16%, 5% of the time.

378
 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. This result holds for

any level of 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.

The second line of Panel A shows that results are mitigated for an allocation of

20% to directional hedge fund strategies. For low levels of allocation to the risky asset

and for high allocation to the risky asset, the levels of return achievable are lower than

the corresponding p-values. For allocation to the risky asset between 40% and 60%

however, the inclusion of hedge funds helps for medium risks levels. This indicates a first

inflection point whereby depending on the risk appetite and the allocation to the risky

asset, 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. 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. 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.

The first column of Panel A indicates that as they tend to be risky, directional

hedge funds do not enable the investor to attain the risk level corresponding to the 20%

lambda.

379
Table 55: Mean return estimation for average investors

Lambdas
EMF BMF HF
20% 40% 60% 80% 100% All

P-value (95%) 50% 50% 0% 5,4% 8,3% 11,5% 14,4% 16,0% 14,6%

PANEL A: DHF

Mean Return 45% 45% 10% 4,7% 7,9% 10,4% 12,5% 13,7% 11,5%

Mean Return 40% 40% 20% 5,1% 9,2% 12,0% 14,2% 15,6% 13,0%

Mean Return 25% 25% 50% 5,3% 10,9% 14,1% 16,5% 18,3% 15,3%

Mean Return 0% 0% 100% NA 12,7% 16,3% 18,7% 20,6% 17,5%

PANEL B: UHF

Mean Return 45% 45% 10% 5,9% 8,7% 10,4% 11,7% 12,6% 10,8%

Mean Return 40% 40% 20% 6,8% 9,8% 11,5% 12,4% 13,2% 11,5%

Mean Return 25% 25% 50% 8,2% 11,6% 12,9% 13,6% 14,1% 12,4%

Mean Return 0% 0% 100% 9,1% 13,0% 14,6% 15,4% 16,3% 13,6%

 Lambdas
EMF BMF HF
20% 40% 60% 80% 100% All

PANEL C: FOF

Mean Return 45% 45% 10% 5,4% 9,5% 11,1% 12,3% 13,2% 11,5%

Mean Return 40% 40% 20% 5,9% 10,5% 12,2% 13,3% 14,2% 12,5%

Mean Return 25% 25% 50% 6,3% 12,0% 13,6% 14,6% 15,7% 13,5%

Mean Return 0% 0% 100% 8,2% 13,3% 14,8% 16,0% 17,6% 14,9%

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. 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; S x = the skewness of the return distribution and x = the kurtosis of the
return distribution. P-values are obtained by estimating 2,000 CML for a portfolio of mutual funds for any level of lambda and taking the
95% percentage level value. 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, 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. Lambda is the level of allocation to the risk free rate (20%, 40%, 60%, 80% and 100%). Our
database covers 431 hedge funds, 2,418 equity mutual funds and 1,318 bond funds over the January 1996 to May 2006 period. Numbers in
the table are annualized mean returns.
 Panel B reports a different pattern. 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%. For higher risk

allocation, undirectional strategies do not help diversifying. 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.

Panel C reports exactly the same pattern as Panel B. 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%). These results are in line with our

expectations. 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. 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, 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. 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.

Figures 15 to 17 report the same results visually. 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, particularly for higher risk level.

382
Figure 15: Directional Hedge Funds - Mean

Percentile Return

0,05

0,04

0,03

0,02

0,01

0,00

-0,01

-0,02

-0,03

-0,04
100%DHF
All lambdas

50%DHF
Lambda 100%

Lambda 80%

20%DHF
Lambda 60%

10%DHF
Lambda 40%

Lambda 20%

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%, 10%, 20%,
50% and 100%) and for various levels of allocation to the risk-free asset (20%, 40%,
60%, 80% and 100%) for an average investor. 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; S x = the skewness of

the return distribution and

K x = the kurtosis of the return distribution. P-values are
obtained by estimating 2,000 CML for a portfolio of mutual funds for any level of lambda
and taking the 95% percentage level value. P-values estimated with a mutual fund
portfolio divided equally between equity mutual funds (EMF) and bond mutual funds
(BMF). P-values are reported in the 0% allocation to hedge funds part of the Figure.
Lambda is the level of allocation to the risk free rate (20%, 40%, 60%, 80% and 100%).
The 0% hedge fund portfolio is divided equally between equity mutual funds (EMF) and
bond mutual funds (BMF). Our database covers 431 hedge funds, 2,418 equity mutual
funds and 1,318 bond funds over the January 1996 to May 2006 period. Numbers in the
table are annualized mean returns.

383
 Figure 16 shows that including undirectional hedge funds enable investors to earn

higher returns, particularly for low values of lambda. 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. 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.

As indicated by Figure 17, 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. For higher allocation to the risky asset,

integrating 10% or 20% of funds of funds does not help the investor to attain higher

returns. More than 20% of the portfolio has to be allocated to hedge funds.

b. Protective & progressive investors

By definition, 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. A progressive

investor does not care as much about the risk but is focused on returns that can be

attained. Results are reported in Table 57 and Table 5884.

84
We do not report the figures for the sake of brevity but they are available.

384
Figure 16: Undirectional Hedge Funds - Mean Percentile

Return

0,05

0,04

0,03

0,02
0,01
0,00
-0,01
-0,02
-0,03
-0,04
100%UHF
All lambdas

50%UHF
Lambda 100%

Lambda 80%

20%UHF
Lambda 60%

10%UHF
Lambda 40%

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%, 10%,
20%, 50% and 100%) and for various levels of allocation to the risk-free asset (20%,
40%, 60%, 80% and 100%) for an average investor. 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; S x = the skewness of

the return distribution and

K x = the kurtosis of the return distribution. P-values are
obtained by estimating 2,000 CML for a portfolio of mutual funds for any level of lambda
and taking the 95% percentage level value. P-values estimated with a mutual fund
portfolio divided equally between equity mutual funds (EMF) and bond mutual funds
(BMF). P-values are reported in the 0% allocation to hedge funds part of the Figure.
Lambda is the level of allocation to the risk free rate (20%, 40%, 60%, 80% and 100%).
The 0% hedge fund portfolio is divided equally between equity mutual funds (EMF) and
bond mutual funds (BMF). Our database covers 431 hedge funds, 2,418 equity mutual
funds and 1,318 bond funds over the January 1996 to May 2006 period. Numbers in the
table are annualized mean returns.

385
Figure 17: Funds of Hedge Funds - Mean Percentile Return

0,06
0,05
0,04
0,03
0,02
0,01
0,00
-0,01
-0,02
-0,03
-0,04
100%FOF
S6

50%FOF
All lambdas

Lambda 80%

20%FOF
Lambda 60%

10%FOF
Lambda 40%

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%, 10%, 20%,
50% and 100%) and for various levels of allocation to the risk-free asset (20%, 40%,
60%, 80% and 100%) for an average investor. 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; S x = the skewness of

the return distribution and

K x = the kurtosis of the return distribution. P-values are
obtained by estimating 2,000 CML for a portfolio of mutual funds for any level of lambda
and taking the 95% percentage level value. P-values estimated with a mutual fund
portfolio divided equally between equity mutual funds (EMF) and bond mutual funds
(BMF). P-values are reported in the 0% allocation to hedge funds part of the figure.
Lambda is the level of allocation to the risk free rate (20%, 40%, 60%, 80% and 100%).
The 0% hedge fund portfolio is divided equally between equity mutual funds (EMF) and
bond mutual funds (BMF). Our database covers 431 hedge funds, 2,418 equity mutual
funds and 1,318 bond funds over the January 1996 to May 2006 period. Numbers in the
table are annualized mean returns.

386
Table 56: Mean return estimation for protective investors

Lambdas
EMF BMF HF
20% 40% 60% 80% 100% All

P-value (95%) 50% 50% 0% 5,3% 8,3% 11,5% 14,5% 16,3% 14,7%

PANEL A: DHF

Mean Return 45% 45% 10% 4,7% 8,1% 10,5% 12,6% 13,9% 11,7%

Mean Return 40% 40% 20% 5,0% 9,1% 11,9% 14,0% 15,4% 12,9%

Mean Return 25% 25% 50% 5,5% 10,9% 14,1% 16,4% 18,0% 15,1%

Mean Return 0% 0% 100% NA 12,7% 16,2% 18,7% 20,5% 17,5%

PANEL B: UHF

Mean Return 45% 45% 10% 6,0% 8,6% 10,4% 11,6% 12,6% 10,7%

Mean Return 40% 40% 20% 6,9% 9,8% 11,3% 12,4% 13,2% 11,4%

Mean Return 25% 25% 50% 8,0% 11,5% 12,8% 13,5% 14,1% 12,3%

Mean Return 0% 0% 100% 9,2% 13,0% 14,5% 15,2% 16,1% 13,4%

 Lambdas
EMF BMF HF
20% 40% 60% 80% 100% All

PANEL C: FOF

Mean Return 45% 45% 10% 5,3% 9,4% 11,1% 12,4% 13,4% 11,6%

Mean Return 40% 40% 20% 5,7% 10,5% 12,1% 13,3% 14,2% 12,4%

Mean Return 25% 25% 50% 6,8% 12,0% 13,5% 14,5% 15,5% 13,5%

Mean Return 0% 0% 100% 8,5% 13,1% 14,6% 15,8% 16,9% 14,7%

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. 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
measure defined as 2 6 24 with V x =the variance of the return; S x = the skewness of the return

distribution and
K x = the kurtosis of the return distribution. P-values are obtained by estimating 2,000 CML for a portfolio
of mutual funds for any level of lambda and taking the 95% percentage level value. 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, 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. Lambda is the level of allocation to the risk free rate (20%, 40%, 60%, 80% and 100%). Our
database covers 431 hedge funds, 2,418 equity mutual funds and 1,318 bond funds over the January 1996 to May 2006
period. Numbers in the table are annualized mean returns.
Table 57: Mean return estimation for progressive investors

Lambdas
EMF BMF HF
20% 40% 60% 80% 100% All

P-value (95%) 50% 50% 0% 5,4% 8,2% 11,2% 14,3% 16,9% 14,6%

PANEL A: DHF

Mean Return 45% 45% 10% 4,7% 7,9% 10,3% 12,5% 13,9% 11,5%

Mean Return 40% 40% 20% 5,1% 9,1% 11,7% 14,0% 15,6% 12,9%

Mean Return 25% 25% 50% 5,7% 11,0% 14,1% 16,4% 18,3% 15,2%

Mean Return 0% 0% 100% NA 12,7% 16,3% 18,8% 20,8% 17,6%

PANEL B: UHF

Mean Return 45% 45% 10% 6,1% 8,7% 10,4% 11,7% 12,7% 10,8%

Mean Return 40% 40% 20% 7,0% 9,8% 11,4% 12,4% 13,4% 11,5%

Mean Return 25% 25% 50% 8,3% 11,6% 13,1% 13,6% 14,5% 12,5%

Mean Return 0% 0% 100% 9,2% 13,0% 14,7% 15,5% 16,7% 13,6%

 Lambdas
EMF BMF HF
20% 40% 60% 80% 100% All

PANEL C: FOF

Mean Return 45% 45% 10% 5,2% 9,4% 11,0% 12,3% 13,2% 11,5%

Mean Return 40% 40% 20% 5,7% 10,5% 12,2% 13,3% 14,1% 12,3%

Mean Return 25% 25% 50% 6,5% 11,9% 13,6% 14,8% 15,7% 13,6%

Mean Return 0% 0% 100% 8,4% 13,2% 14,8% 16,1% 17,8% 14,9%

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. A progressive investor is defined as an investor with a C value of 10 in our extended
1 1 1
R x = V x − CS x + C 2 K x
risk measure defined as 2 6 24 with V x =the variance of the return; S x = the skewness of the

return distribution and

K
x = the kurtosis of the return distribution. P-values are obtained by estimating 2,000 CML for a

portfolio of mutual funds for any level of lambda and taking the 95% percentage level value. 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, 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. Lambda is the level of allocation to the risk free rate (20%, 40%, 60%, 80% and
100%). Our database covers 431 hedge funds, 2,418 equity mutual funds and 1,318 bond funds over the January 1996 to
May 2006 period. Numbers in the table are annualized mean returns.
 The results obtained for progressive and protective investors are very close to

those obtained for an average investor. 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.

Depending on the global risk level and the kind of investor considered, the investor

should decide if he will use directional, undirectional hedge funds or funds of funds, and

to what extent they should be used to achieve higher levels of satisfaction.

d. 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. 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
With the risk measure defined as 2 6 24 (31)

85
We report the results for average investors. Corresponding Tables for protective and progressive

investors are available upon request.

391
Table 58: Slope estimation for average investors

Lambdas
EMF BMF HF
20% 40% 60% 80% 100% All

P-value (95%) 50% 50% 0% 0,078 0,038 0,020 0,004 0,002 0,022

PANEL A: DHF

Mean Return 45% 45% 10% 0,068 0,024 0,009 0,002 0,001 0,007

Mean Return 40% 40% 20% 0,079 0,028 0,008 0,003 0,001 0,008

Mean Return 25% 25% 50% 0,118 0,038 0,008 0,003 0,001 0,010

Mean Return 0% 0% 100% NA 0,033 0,008 0,003 0,001 0,009

PANEL B: UHF

Mean Return 45% 45% 10% 0,088 0,017 0,005 0,002 0,001 0,008

Mean Return 40% 40% 20% 0,006 0,008 0,009 0,010 0,010 0,009

Mean Return 25% 25% 50% 0,155 0,013 0,003 0,001 0,001 0,020

Mean Return 0% 0% 100% 0,161 0,012 0,003 0,001 0,001 0,031

 Lambdas
EMF BMF HF
20% 40% 60% 80% 100% All

PANEL C: FOF

Mean Return 45% 45% 10% 0,099 0,019 0,005 0,002 0,001 0,007

Mean Return 40% 40% 20% 0,160 0,020 0,004 0,001 0,001 0,008

Mean Return 25% 25% 50% 0,286 0,016 0,003 0,001 0,001 0,009

Mean Return 0% 0% 100% 0,507 0,014 0,003 0,001 0,001 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. 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
defined as 2 6 24 with V x =the variance of the return; S x = the skewness of the return distribution and
K x = the kurtosis of the return distribution. P-values are obtained by estimating 2,000 CML for a portfolio of mutual funds for
any level of lambda and taking the 95% percentage level value. 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, 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. Lambda
is the level of allocation to the risk free rate (20%, 40%, 60%, 80% and 100%). Our database covers 431 hedge funds, 2,418
equity mutual funds and 1,318 bond funds over the January 1996 to May 2006 period. Numbers in the table are annualized
mean returns.
 Table 59 reports mitigated results. 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. Two remarks, first, the slope decreases as we increase the risk level. This is

perfectly in line with the visual aspect of an efficient frontier. 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, the curve drawn will be concave. Second, the

lambda 100% column indicates that for high allocation to the risky asset, the slope is low

on absolute terms and close to zero indicating that we are close to the flat tangent.

Panel A reports the slopes of the capital market line obtained when directional

hedge funds are added to the portfolio. 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%). Results are however

completely different when we go further to the right and consider higher risk allocation.

Directional hedge funds do not enable investors to reach higher levels of satisfaction.

These results have to be mitigated because we are looking at very small numbers on

absolute terms.

Panel B and C report the same pattern for undirectional hedge funds and for funds

of hedge funds. 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.

We obtain exactly the same pattern for protective and progressive investors,

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, but the

results are unclear for higher levels of risk since figures reported are very low in absolute

terms.

394
4.3 Including mutual funds in hedge fund portfolios

We perform the inverse test as a complementary analysis, 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. We perform this test for

average, protective and progressive investors as well as use mean returns as a measure

of satisfaction and slope.

a. 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. In the case of

directional hedge funds, 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). When higher allocation are made to 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.

For directional hedge funds, the results are even more interesting. Adding mutual

funds to hedge fund portfolio is attractive. 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%). For higher level of risks, the returns offered remains attractive even

for low allocation to hedge funds. This result is in line with the underlying characteristics

of undirectional strategies. Such strategies are low risk strategies by definition. To offer

riskier returns, such strategies should be diversified with directional high risk strategies.

Results are more controversial with funds of funds. 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%). Over this level, the returns offered for each level of risk are

more attractive when only funds of funds are considered. For high risk portfolio however,

395
this is not true. This result is logical since funds of hedge funds are low risk investments

that should be diversified to be considered risky.

b. Protective and progressive investors86

The results obtained for protective and progressive investors are the same as

those obtained for average investors. 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.

86
Complete results are available upon request.

396
Table 59: Mean return estimation for average investors (including mutual funds in a
hedge fund portfolio)
Lambda
Panel A: DHF EMF BMF HF
20% 40% 60% 80% 100% All

Percentile 5% 0% 0% 100% NA 9,6% 13,2% 14,9% 15,8% 11,8%

Mean Return 25% 25% 50% 5,3% 10,9% 14,1% 16,5% 18,3% 15,3%

Mean Return 40% 40% 20% 5,1% 9,2% 12,0% 14,2% 15,6% 13,0%

Mean Return 45% 45% 10% 4,7% 7,9% 10,4% 12,5% 13,7% 11,5%

Mean Return 50% 50% 0% 4,5% 6,8% 8,9% 11,0% 12,3% 10,2%

Lambda
Panel B: UHF EMF BMF UHF
20% 40% 60% 80% 100% All

Percentile 5% 0% 0% 100% 6,3% 10,0% 10,5% 10,7% 11,1% 7,7%

Mean Return 25% 25% 50% 8,2% 11,6% 12,9% 13,6% 14,1% 12,4%

Mean Return 40% 40% 20% 6,8% 9,8% 11,5% 12,4% 13,2% 11,5%

Mean Return 45% 45% 10% 5,9% 8,7% 10,4% 11,7% 12,6% 10,8%

Mean Return 50% 50% 0% 5,9% 8,7% 10,4% 11,7% 12,6% 10,8%
 Lambda
Panel C: FoF EMF BMF FOF
20% 40% 60% 80% 100% All

Percentile 5% 0% 0% 100% 6,7% 11,3% 12,6% 13,2% 13,5% 11,7%

Mean Return 25% 25% 50% 6,3% 12,0% 13,6% 14,6% 15,7% 13,5%

Mean Return 40% 40% 20% 5,9% 10,5% 12,2% 13,3% 14,2% 12,5%

Mean Return 45% 45% 10% 5,4% 9,5% 11,1% 12,3% 13,2% 11,5%

Mean Return 50% 50% 0% 4,5% 6,8% 8,9% 11,0% 12,3% 10,2%

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. 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
defined as 2 6 24 with V x =the variance of the return; S x = the skewness of the return distribution and
K x = the kurtosis of the return distribution. P-values are obtained by estimating 2,000 CML for a portfolio of hedge fund for any
level of lambda and taking the 95% percentage level value. The allocation to mutual fund is equally weighted between equity
mutual funds (EMF) and bond mutual funds (BMF). Panel A reports the result obtained when the initial portfolio consists of
directional hedge funds, 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. Lambda is the level of allocation to
the risk free rate (20%, 40%, 60%, 80% and 100%). Our database covers 431 hedge funds, 2,418 equity mutual funds and
1,318 bond funds over the January 1996 to May 2006 period. Numbers in the table are annualized mean returns.
 c. 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. Mutual funds may enable hedge fund investors to significantly increase their
level of satisfaction.

399
Table 60: Slope estimation for average investors (including mutual funds in a hedge
fund portfolio)
Lambda
Panel A: DHF EMF BMF HF
20% 40% 60% 80% 100% All

Percentile 5% 0% 0% 100% NA 0,0694 0,0138 0,0053 0,0027 0,0335

Mean Return 25% 25% 50% 0,1178 0,0375 0,0077 0,0026 0,0012 0,0099

Mean Return 40% 40% 20% 0,0791 0,0284 0,0079 0,0025 0,0011 0,0082

Mean Return 45% 45% 10% 0,0684 0,0235 0,0088 0,0024 0,0011 0,0074

Mean Return 50% 50% 0% 0,0520 0,0197 0,0095 0,0022 0,0011 0,0065

Lambda
Panel B: UHF EMF BMF HF
20% 40% 60% 80% 100% All

Percentile 5% 0% 0% 100% 0,3578 0,0273 0,0067 0,0023 0,0013 0,1807

Mean Return 25% 25% 50% 0,1551 0,0131 0,0028 0,0011 0,0006 0,0197

Mean Return 40% 40% 20% 0,0055 0,0078 0,0091 0,0098 0,0104 0,0091

Mean Return 45% 45% 10% 0,0884 0,0170 0,0046 0,0017 0,0008 0,0079

Mean Return 50% 50% 0% 0,0520 0,0197 0,0095 0,0022 0,0011 0,0065
 Lambda
Panel C: FoF EMF BMF FOF
20% 40% 60% 80% 100% All

Percentile 5% 0% 0% 100% 1,6006 0,0315 0,0059 0,0025 0,0014 0,0238

Mean Return 25% 25% 50% 0,2858 0,0158 0,0033 0,0014 0,0007 0,0093

Mean Return 40% 40% 20% 0,1598 0,0198 0,0038 0,0014 0,0007 0,0077

Mean Return 45% 45% 10% 0,0985 0,0190 0,0048 0,0015 0,0008 0,0068

Mean Return 50% 50% 0% 0,0520 0,0197 0,0095 0,0022 0,0011 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. 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
defined as 2 6 24 with V x =the variance of the return; S x = the skewness of the return distribution and
K x = the kurtosis of the return distribution. P-values are obtained by estimating 2,000 CML for a portfolio of hedge fund for any
level of lambda and taking the 95% percentage level value. The allocation to mutual fund is equally weighted between equity
mutual funds (EMF) and bond mutual funds (BMF). Panel A reports the result obtained when the initial portfolio consists of
directional hedge funds, 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. Lambda is the level of allocation to
the risk free rate (20%, 40%, 60%, 80% and 100%). Our database covers 431 hedge funds, 2,418 equity mutual funds and 1,318
bond funds over the January 1996 to May 2006 period. Numbers in the table are annualized mean returns.
V Conclusion

Hedge funds exhibit abnormal returns. 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. 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. Our methodology is based on the Taylor’s

extension of the linex utility function developed by Bell (1988, 1995).

This new methodology opens a new field in portfolio analysis research. It enables

investors to determine if they should include alternative investments in their portfolio

depending on their level of risk aversion. When the risk measure is limited to volatility,

hedge funds exhibit an attractive risk/return profile, however this result needs careful

care when higher moments are considered.

Our results indicate that directional hedge funds should be considered separately

from undirectional hedge funds and fund of hedge funds. 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. This result in a significant improvement of satisfaction for

diversified portfolios. Over a 50% allocation to directional hedge funds provides a

significantly more attractive returns in all cases.

On the other hand, undirectional hedge funds and fund of funds can be grouped

together because they have the same pattern. 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. For high allocation

to the risky asset undirectional strategies do not help diversifying and reaching higher

return levels. This result is in contrast with the one obtained for directional hedge funds

402
and confirms the need to differentiate directional from unidirectional hedge funds.

Undirectional strategies and funds of funds are low risk profile investments and should be

used as such.

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.

The new adapted efficient frontier opens new doors for asset allocators. Based on

the clients’ objective and the market conditions, 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.

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.

Second, the identical tests should be performed on various analysis period to determine

what kind of fund should be favoured in specific market conditions. 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.

403
General Conclusion
General Conclusion

Hedge funds have attracted the attention of many academicians over the last

decade. Almost every aspect of the hedge fund world has been studied from performance

decomposition, to risk analysis, strategy explanation, regulatory aspects, and so on. The

purpose of this doctoral thesis is clearly established: to understand hedge fund strategies

by looking at the number produced. Our contribution is on the quantitative analysis of

hedge fund strategies. We divide this Thesis in three complementary parts.

Our first objective is to clearly understand hedge fund managers and to explain

how they create alpha over time. To extract pure alpha we develop, test and improve an

extended multi-factor performance analysis model in order to understand hedge fund

performance on the one hand, while on the other to develop and adapt a methodology to

determine whether there is any persistence in hedge fund returns. 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).

We analyse hedge fund returns, understand how manager create performance

over time and if the extracted is consistent over time. Globally, 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. Depending on the period under review and on the strategy considered

between 20% and 50% of the individual managers do outperform classical indices. This

result indicates that hedge funds have been attractive investments in various market

conditions. With these results in hand, we looked for a systematic way of investing in

405
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 Skew-

Student 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, 251-

274.

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, 129-

156.

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, 1035-

1058.

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.
Efron, B., 1979, Bootstrap Methods: Another Look at the Jackknife, The Annals of

Statistics 7, n°1, 1-26.

Efron, B., 1983, Estimating the Error Rate of a Prediction Rule: Improvement on Cross-

validation, Journal of the American Statistical Association, 78, n°382, 316-331.

Efron, B. and Tibshirani, R.J., 1993, An Introduction to the Bootstrap, London: Chapman

& Hall.

Elton, E., Gruber, M., Das, S., Blake, C., 1996, The Persistence of Risk-adjusted Mutual

Fund Performance, Journal of Business 69, n°2, 133-157.

Elton, E., Gruber, M., Das, S., Hlavka, M., 1993, Efficiency with costly Information: A Re-

interpretation of Evidence from Managed Portfolios, Review of Financial Studies 6, n°1,

1-21.

Ennis, R. M., Sebastian, M. D., 2003, A Critical Look at the Case for Hedge Funds, Journal

of Portfolio Management 29, n°4, 103-112.

Fabozzi, F. J. and Francis J. C., 1979, Mutual Fund Systematic Risk for Bull and Bear

Markets: An Empirical Examination, Journal of Finance 34, n°5, 1243-1250.

Faff, R., 2001, A Multivariate Test of a Dual-Beta CAPM: Australian Evidence, Financial

Review 36, 157-174.

Fama, E.F., French, K.R., 1992, The Cross-section of expected Returns, Journal of

Finance 47, n°2, 427-465.

Fama, E.F., French, K.R., 1993, Common risk Factors in the Returns on Stocks and

Bonds, Journal of Financial Economics 33, n°1, 3-56.

Fama, E.F., French, K.R., 1996, Multifactor Explanations of Asset Pricing Anomalies,

Journal of Finance 51, n°1, 55-84.

Fama, E.F., French, K.R., 1998, Value versus Growth: The International Evidence,

Journal of Finance 53, n°6, 1975-1999.

Fama, E.F., French, K.R., 2000, Characteristics, Covariances, and Average Returns:

1929-1997. Journal of Finance 55, n°1, 389-406.

Fang, H., T. Lai, 1997, Co-Kurtosis and Capital Aseet Pricing, 1997, The Financial Review

32, n°2, 293-307.

Friend, I. and M. E. Blume, 1975, The Demand for Risky Assets, The American Economic

Review 65, n°5, 900-922.

Fung, W., Hsieh, D.A., 1997, Empirical Characteristics of Dynamic Trading Strategies:

The Case of Hedge Funds, Review of Financial Studies 10, n°2, 275-302.

Fung, W., Hsieh, D.A., 1998, A Risk Neutral Approach to Valuing Trend Following Trading

Strategies, Working Paper. Duke University, Durham, NC.

Fung, W., Hsieh, D.A., 1999, Is Mean-Variance Analysis applicable to Hedge Funds?,

Economics Letters 62, n°1, 53-58.

Fung, W., Hsieh, D.A., 2000, Performance Characteristics of Hedge Funds and

Commodity Funds: Natural vs. spurious Biases, Journal of Quantitative and Financial

Analysis 35, n°3, 291-307.

Fung, W., Hsieh, D.A., 2001, The Risk of Hedge Funds Strategies: Theory and Evidence

from Trend Followers, Review of Financial Studies 14, n°2, 313-341.

Fung, W., Hsieh, D.A., 2002a, The Risks in Fixed Income Hedge Fund Styles, Journal of

Fixed Income 12, n°2, 6-27.

Fung, W., Hsieh, D.A., 2002b, Benchmarks of Hedge Fund Performance: Information

Content and Measurement Biases, Financial Analyst Journal 58, 22-34.

Fung, W., Hsieh, D.A., 2006, Hedge Funds: An Industry in its Adolescence, Economic

Review 91, n°4

Galadera, D.U.A, Faff, R., 2004, Modelling the Risk and Return Relation Conditional on

Market Volatility and market Conditions, Working Paper 08/04, Monash University.
Gatev E., Goetzmann, W.N., Rouwenhorst, K., 1999, Pairs Trading: Performance of a

Relative Value Rule, working paper, Yale School of Management.

Goetzmann, W.N., Ibbotson, R.G., 1994, Do Winners Repeat? Patterns in Mutual

Performance, Journal of Portfolio Management 20, n°2, 9-18.

Goetzmann, W.N., Ingersoll, J.E., Spiegel, M.I., Welch, I., 2006, Portfolio Performance

Manipulation and Manipulation-Proof Performance Measures, Yale ICF Working Paper

No. 02-08.

Gregoriou, G., F. Rouah, 2003, Random Walk Behavior of CTAs Returns, Journal of

Alternative Investments 6, n° 2, 51-56

Grinblatt, M., Titman, S., 1989, Mutual Fund Performance: An Analysis of Quarterly

Portfolio Holdings, Journal of Business 62, n°3, 393-416.

Grinblatt, M., Titman, S. 1992, The Persistence of Mutual Fund Performance, Journal of

Finance 47, n°5, 1977-1984.

Grinblatt, M., Titman, S., 1994, A Study of Monthly Mutual Fund Returns and

Performance Evaluation Techniques, Journal of Financial and Quantitative Analysis 29,

n°3, 419-444.

Grinblatt, M., Titman, S., Wermers, R., 1995, Momentum Investment Strategies, Portfolio

Performance and Herding: A Study of Mutual Fund Behaviour, American Economic

Review 85, n°5, 1088-1105.

Gruber, M. J., 1996, Another Puzzle: the Growth in Actively Managed Mutual Funds,

Journal of Finance 51, n°3, 783-810.

Hansen, L, 1982, Large Sample Properties of the Generalised Method of Moments

Estimators, Econometrica 50, n°4, 1029-1054.

Harvey, C., J., Liechty, M., Liechty, and P., Müller, 2002, Portfolio Selection with Higher

Moments, mimeo, Duke University.

Harvey, C., A. Siddique, 2000, Conditional Skewness in Asset Pricing Tests, Journal of

Finance 55, n°1, 1263-1295.

Hendricks, D., Patel, J., Zeckhauser, R., 1993, Hot Hands in Mutual Funds: Short-run

Persistence of Performance: 1974-88, Journal of Finance 48, n°1, 93-130.

Investment Company Institute, 2003, Worldwide Mutual Funds Assets and Flows, Second

Quarter 2003, ICI Statistics and Research quarterly report, www.ici.org/stats.

Ippolito, R. A., 1989, Efficiency with Costly Information: A Study of Mutual Fund

Performance, Quarterly Journal of Economics 104, n°1, 1-24.

Jagannathan, R., Korajczyk, R.A., 1986, Assessing the Market Training Performance of

Managed Portfolios, The Journal of Business 59, n°2, 217-235.

Jegadeesh, N., Titman, S., 1993, Returns to Buying Winners and Selling Losers:

Implications for Stock Market Efficiency, Journal of Finance 48, n°1, 93-130.

Jensen, M.C., 1968, The Performance of Mutual Funds in the Period 1945-1964, Journal

of Finance 23, n°2, 389-416.

Jensen, M.C., 1969, Risk, the Pricing of Capital Assets and Evaluation of Investment

Portfolios, Journal of Business 42, n°2, 167-247.

Jondeau, E., Rockinger, M., 2006, Optimal Portfolio Allocation Under Higher Moments,

European Financial Management 12, n°1, 29-55.

Jurczenko, E., Maillet, B., 2006, The Four-Moment Capital Asset Pricing Model: between

Asset Pricing and Asset Allocation, in Multi-moment Asset Pricing Models and Related

Topics, ed. Bertrand Maillet and Emmanuel Jurczenko, Wiley.

Kan, R., Zhou, G., 2001, Tests of Mean-Variance Spanning, Washington University

Working Paper.

Kao, G. W., Cheng, L.T.W., Chan K.C., 1998, International Mutual Fund Selectivity and

Market Timing during Up and Down Markets, Financial Review 33, n°2, 127-144.
Kat, H. M., Amin, G. S., 2003, Stocks, Bonds, and Hedge Funds, Journal of Portfolio

Management 29, n°4, 113-120.

Kat, Harry, Joëlle Miffre, 2005, Hedge Fund Performance: The Role of Non-Normality

Risks and Conditional Asset Allocation, Working Paper WP-FF-21-2005, Cass Business

School.

Keating, C., Shadwick, W., 2002, A Universal Performance Measure, Journal of

Performance Measurement 6, n°3, 59-84

Kendall, M., 1938, A New Measure of Rank Correlation, Biometrica, 30, 81-89.

Kendall, M., 1948, Rank Correlation Methods, Charles Griffin & Company Limited.

Kimball, M., 1993, Standard Risk Aversion, Econometrica 61, n°3, 589-611.

Kooli, M., 2007, The diversification benefits of hedge funds and funds of hedge funds,

Derivatives Use, Trading & Regulation 12, n°4, 290-300.

Kothari, S.P., Warner, J.B., 2001, Evaluating Mutual Fund Performance, Journal of

Finance 56, n°5, 1985-2010.

Kramer, C., 1994, Macroeconomic Seasonality and the January Effect, Journal of Finance

49, n°5, 1883-1891.

Kraus, A., R. H. Lintzenberger, 1976, Skewness Preference and the Valuation of Risk

Assets, Journal of Finance 31, n°4, 1085-1100.

Lai, T.Y., 1991, Portfolio Selection with Skewness: A Multiple-objective Approach, Review

of Quantitative Finance and Accounting 1, n°1, 293-305.

Lehmann, B.N., Modest, D.M., 1987, Mutual Fund Performance Evaluation: A Comparison

of Benchmarks and Benchmarks Comparisons, Journal of Finance 42, n°2, 233-245.

Levy, H., Markowitz, H. M., 1979, Approximating Expected Utility by a Function of Mean

and Variance, American Economic Review 69, n°3, 308-317.

Liang, B., 1999, On the Performance of Hedge Funds, Financial Analysts Journal 55, n°4,

72-85.
Liang, B., 2000, Hedge Funds: The Living and the Dead, Journal of Financial and

Quantitative Analysis 35, n°3, 309-325.

Liang, B., 2001, Hedge Funds Performance: 1990-1999, Financial Analysts Journal 57,

n°1, 11-18.

Liang, B., 2003, On the Performance of Alternative Investments: CTAs, Hedge Funds,

and Funds-of-Funds. Working Paper, Case Western Reserve University, Cleveland.

Liew, J., 2003, Hedge Fund Index Investing Examined, Journal of Portfolio Management

29, n°2, 113-123.

Lintner, J., 1965, The Valuation of Risk Assets and the Selection of Risky Investments in

Stock Portfolio and Capital Budgets, Review of Economics and Statistics 47, n°1, 13-37.

Litzenberger R., Ramaswamy, K., 1979, The Effects of Personal Taxes and Dividends on

Capital Asset Prices: Theory and Empirical Evidence, Journal of Financial Economics 7,

n°2, 163-195.

Litzenberger R., Ramaswamy, K., 1982, The Effects of Dividends on Common Stocks

Returns: Tax Effects or Information's Effects ?, Journal of Finance 37, n°2, 429-443.

Lo, A., 2001, Risk Management for Hedge Funds: Introduction and Overview, Financial

Analyst Journal 57, n°6, 16-33.

Loistl, O., 1976, The Erroneous Approximation of Expected Utility by Means of a Taylor’s

Series Expansion Analytic and Computational Results, The American Economic Review

66, n°5, 904-910.

Malkiel, B.G., 1995, Returns from Investing in Equity Mutual Funds 1971 to 1991, Journal

of Finance 50, n°2, 549-572.

Malkiel, B., Saha, A., 2005, Hedge Funds: Risk and Return, Financial Analyst Journal 61,

n°6, 80-88.

McGuire, P., Remolona E. Tsatsaronis, K., 2005, Time-Varying Exposures and Leverage in

Hedge Funds , BIS Quarterly Review, 59-72.

Metrick E., 1999, Performance Evaluation with Transactions Data: The Stock Selection of

Investment Newsletters, Journal of Finance 54, n°5, 1743-1775.

Mitchell, M., Pulvino, T., 2001, Characteristics of Risk and Return in Risk Arbitrage,

Journal of Finance 56, n°6, 2135-2175.

Morin, R. A., Fernandez Suarez, A., 1983, Risk Aversion Revisited, The Journal of Finance

38, n°4, 1201-1216.

Navone, M., 2001, Diversifying Market Risk through Market Neutral Strategies, working

paper, IEIF Bocconi University.

Newey, W., West, K., 1987, A Simple Positive-definite Heteroskedasticity and

Autocorrelation Consistent Covariance Matrix, Econometrica 55, n°3, 703-708.

Park, J., 1995, Managed Futures as an Investment Set, Ph.D. Dissertation, Columbia

University.

Pollock, D.S.G., 1999, A Handbook of Time-series Analysis, Signal Processing and

Dynamics, Academic Press, 800pages.

Posthuma, N., Van der Sluis, P. J., 2005, Analyzing Style Drift in Hedge Funds », in

Hedge Funds, G. Gregoriou, G. Hübner, N. Papageorgiou, F. Rouah, Ed. Wiley.

Posthuma, N., Van der Sluis, P. J., 2003, A Reality Check on Hedge Fund returns,

working paper ABP Invstments.

Rao, U., 2001, Mutual Fund Performance during Up and Down Market Conditions, Review

of Business 22, n°2, 62-65.

Reinganum, M.R., 1981. A new Empirical Perspective on the CAPM, Journal of Financial

and Quantitative Analysis 16, n°4, 439-462.

Roon, F., Nijman, T., Werker, B., 2001, Testing for Mean-Variance Spanning with Short

Sales Constraints and Transaction Costs: The Case of Emerging Markets, The Journal of

Finance 56, n°2, 721-742.

Rosenberg, B., Reid, K. Lanstein, R., 1985, Persuasive Evidence of Market Inefficiency,

Journal of Portfolio Management 11, n°3, 9-17.

Schittkowski, K, 2002, Implementation of a Sequential Quadratic Programming Algorithm

for Parallel Computing, Report, Department of Mathematics, University of Bayreuth.

Schneeweis, T., Spurgin, R., 1998, Multifactor Analysis of hedge funds, Managed Futures

and Mutual Fund Return and Risk Characteristics, Journal of Alternative Investments 1,

n°1, 1-24.

Schneeweis, T., Spurgin, R., 2000, Quantitative Analysis of Hedge Funds and Managed

Futures Return and Risks Characteristics, in Evaluating and Implementing Hedge Fund

Strategies, 3rd Edition: The Experience of Managers and Investors, Institutional

Investor Publishing.

Sharpe, W.F., 1964, Capital Asset Prices: A Theory of Market Equilibrium under

Conditions of Risk, Journal of Finance 19, n°3, 425-442.

Sharpe, W.F., 1992, Asset Allocation: Management Style and Performance Measurement,

Journal of Portfolio Management 18, n°2, 7-19.

Sirri, E.R., Tufano, P., 1998, Costly Search and Mutual Fund Flows, Journal of Finance 53,

n°5, 1589-1622.

Spanier, J, Oldham, K., 1987, An Atlas of Functions, Hemisphere Publishing Corporation.

Sun, Q. Yan, Y., 2003, Skewness Persistence with Optimal Portfolio Selection, Journal of

Banking and Finance 27, n°6, 1111-1121.

Swinkels, L.A.P, Van der Sluis, P.J., 2001, Return-based Style Analysis with Time-

varying Exposures, discussion paper 96, Center for Economic Research, Tiburg

University.

Van Hedge Funds Advisors International, 2002, Size of the Hedge Fund Universe.

www.hedgefund.com/abouthfs/universe/universe.htm.
Weisman, A., 2002, Informationless Investing and Hedge Fund Performance

Measurement Bias, Journal of Portfolio Management 28, n°4, 80-91.

Wermers, R., 1996, Momentum Investment Strategies of Mutual Funds, Performance

Persistence, and Survivorship Bias, Working paper, Graduate School of Business and

Administration, University of Colorado.

Zheng, L., 1999, Is Money Smart? A Study of Mutual Fund Investor’ fund Selection

Ability, Journal of Finance 54, n°3, 901-933.

An Analysis of Hedge Fund Strategies – List of Figures

FIGURE 1: FOUR GLOBAL CATEGORIES OF HEDGE FUND ACADEMIC STUDIES .................................. 7

FIGURE 2: HEDGE FUND DATABASE UNIVERSE REPARTITION .................................................. 17

FIGURE 3: MONTHLY RETURNS FOR DEAD FUNDS TOWARDS THE DISSOLUTION DATE ...................... 65

FIGURE 4: HEDGE FUNDS DISSOLUTION FREQUENCIES ...................................................... 108

FIGURE 5: DECILE REPARTITION PER STRATEGY (RANKING BASE ON ALPHA) ............................ 226

FIGURE 6: STRATEGY REPARTITION IN THE ALPHA RANKING ............................................... 227

FIGURE 7: DECILE REPARTITION PER STRATEGY BASED ON BETA .......................................... 228

FIGURE 8: STRATEGY REPARTITION IN THE BETA RANKING ................................................. 229

FIGURE 9: MINIMUM, MAXIMUM, MEDIAN AND MEAN INDIVIDUAL CORRELATIONS BETWEEN MARKET

NEUTRAL FUNDS AND THE EQUITY INDEX ................................................................ 292

FIGURE 10: LAGGED 12-MONTH DECILE DESCRIPTION ...................................................... 304

FIGURE 11: ILLUSTRATION OF THE ADAPTED CAPITAL MARKET LINE ...................................... 362
An Analysis of Hedge Fund Strategies – List of Figures

FIGURE 12: ILLUSTRATION OF THE TEST OF THE ADAPTED CAPITAL MARKET LINE ....................... 363

FIGURE 13: REPARTITION OF THE SKEWNESS OF INDIVIDUAL FUNDS ...................................... 375

FIGURE 14: REPARTITION OF THE KURTOSIS OF INDIVIDUAL FUNDS ....................................... 376

FIGURE 15: DIRECTIONAL HEDGE FUNDS - MEAN PERCENTILE RETURN .................................. 383

FIGURE 16: UNDIRECTIONAL HEDGE FUNDS - MEAN PERCENTILE RETURN ............................... 385

FIGURE 17: FUNDS OF HEDGE FUNDS - MEAN PERCENTILE RETURN ...................................... 386
An Analysis of Hedge Fund Strategies – List of Tables

TABLE 1: STUDIES SPECIFICITIES AND OBJECTIVES ............................................................. 4

TABLE 2: DATABASE COMPARISON ............................................................................... 18

TABLE 3: BACKFILL BIAS AND SURVIVORSHIP BIAS ESTIMATION .............................................. 20

TABLE 4: HEDGE FUND INDICES COMPARISON ................................................................. 24

TABLE 5: MULTI-FACTOR PERFORMANCE DECOMPOSITION MODEL ........................................... 31

TABLE 6: PERSISTENCE ANALYSIS STUDIES SPECIFICITIES, OBJECTIVES AND CONCLUSIONS ............ 35

TABLE 7: NEUTRALITY OF MARKET NEUTRAL FUND SPECIFICITIES, OBJECTIVES AND CONCLUSIONS .... 37

TABLE 8: HEDGE FUND AS DIVERSIFICATION TOOLS SPECIFICITIES, OBJECTIVES AND CONCLUSIONS . 40

TABLE 9: SURVIVORSHIP BIAS IN HEDGE FUNDS .............................................................. 63

TABLE 11: ESTIMATION OF INSTANT RETURN HISTORY BIAS ................................................. 69

TABLE 12: DESCRIPTIVE STATISTICS OF HEDGE FUNDS STRATEGIES AND PASSIVE INVESTMENT

STRATEGIES .................................................................................................. 71

TABLE 13: CORRELATION BETWEEN HEDGE FUNDS AND PASSIVE INVESTMENT STRATEGIES ............. 77

TABLE 14: PERFORMANCE MEASUREMENT USING THE CAPM, CARHART’S 4-FACTOR MODEL AND THE

COMBINED MODEL. ........................................................................................... 82

An Analysis of Hedge Fund Strategies – List of Tables

TABLE 15: PERFORMANCE OF HEDGE FUNDS IN DIFFERENT SUB-PERIODS ................................. 97

TABLE 16: HEDGE FUNDS PERSISTENCE BASED ON 12 MONTH LAGGED RETURNS....................... 101

TABLE 17: HEDGE FUNDS STRATEGY PERSISTENCE BASED ON 12 MONTH LAGGED RETURNS .......... 110

TABLE 18: DESCRIPTIVE STATISTICS OF HEDGE FUNDS STRATEGIES AND PASSIVE INVESTMENT

STRATEGIES ................................................................................................ 133

TABLE 19: DESCRIPTIVE STATISTICS OF HEDGE FUNDS STRATEGIES FOR THE BULLISH AND BEARISH SUB-

PERIODS ..................................................................................................... 138

TABLE 20: CORRELATION AMONG HEDGE FUNDS, BETWEEN HEDGE FUNDS AND PASSIVE INVESTMENT

STRATEGIES, AND AMONG PASSIVE INVESTMENT STRATEGIES ........................................ 144

TABLE 21: SURVIVORSHIP BIAS IN HEDGE FUNDS ............................................................ 149

TABLE 22: ESTIMATION OF INSTANT RETURN HISTORY BIAS ................................................ 152

TABLE 23: PERFORMANCE MEASUREMENT USING THE CAPM, CARHART’S 4-FACTOR MODEL AND THE

COMBINED MODEL .......................................................................................... 156

TABLE 24: PERFORMANCE OF HEDGE FUNDS DURING THE BULLISH AND BEARISH SUB-PERIODS ........ 162

TABLE 25: HEDGE FUNDS PERSISTENCE BASED ON 12 MONTH LAGGED RETURNS ........................ 172
An Analysis of Hedge Fund Strategies – List of Tables

TABLE 26: HEDGE FUNDS PERSISTENCE DURING THE BULLISH AND BEARISH SUB-PERIODS............. 175

TABLE 27: HEDGE FUNDS PERSISTENCE FOR THE MARKET NEUTRAL STRATEGY .......................... 180

TABLE 28: CORRELATION BETWEEN THE OPTION FACTORS AND THE MARKET FACTOR .................. 198

TABLE 29: DESCRIPTIVE STATISTICS .......................................................................... 204

TABLE 30: HEDGE FUND STRATEGIES PERFORMANCE ANALYSIS (1/1994-12/2002) ................. 212

TABLE 31: HEDGE FUND PERSISTENCE IN PERFORMANCE (1/1995-12/2002)......................... 218

TABLE 32: PERSISTENCE IN PERFORMANCE ANALYSIS BASED ON THREE-YEAR DATA (1/1997-

12/2002).................................................................................................. 221

TABLE 33: PERSISTENCE IN PERFORMANCE ANALYSIS BASED ON THE SHARPE SCORE (1/1997-

12/2002).................................................................................................. 236

TABLE 34: PERSISTENCE IN PERFORMANCE ANALYSIS BASED ON THE STANDARD DEVIATION (1/1997-

12/2002).................................................................................................. 238

TABLE 35: PERSISTENCE IN PERFORMANCE ANALYSIS BASED ON THE ALPHA (1/1997-12/2002) .. 243

TABLE 36: PERSISTENCE IN PERFORMANCE ANALYSIS BASED ON BETA (1/1997-12/2002) ......... 245

TABLE 37: PERSISTENCE IN PERFORMANCE ANALYSIS BASED ON THE SKEWNESS AND KURTOSIS

(1/1997-12/2002) ..................................................................................... 247

An Analysis of Hedge Fund Strategies – List of Tables

TABLE 38: PERSISTENCE IN PERFORMANCE ANALYSIS BASED AN ADAPTED SHARPE SCORE (1/1997-

12/2002).................................................................................................. 254

TABLE 39: SUMMARY OF ALPHA CREATION AND EQUITY EXPOSURE ....................................... 256

TABLE 40: KENDALL TAU ESTIMATION ........................................................................ 261

TABLE 41: DESCRIPTIVE STATISTICS AND DECILE DESCRIPTIVE STATISTICS .............................. 286

TABLE 42: CORRELATION BETWEEN MARKET NEUTRAL STRATEGIES AND EQUITY INDEX ................. 290

TABLE 43: BIRTH AND ATTRITION RATES ...................................................................... 293

TABLE 44: SURVIVORSHIP BIAS ................................................................................ 297

TABLE 45: MARKET EXPOSURE ANALYSIS...................................................................... 301

TABLE 46: LAGGED 12-MONTH DECILE EXPOSURE ........................................................... 306

TABLE 47: INDIVIDUAL FUND MARKET EXPOSURE ............................................................ 309

TABLE 48 : INDIVIDUAL FUND MARKET EXPOSURE – MULTI FACTOR MODEL ............................. 310

TABLE 49: BULL AND BEAR BETA ESTIMATION ............................................................... 311

TABLE 50: EX-POST BETA ANALYSIS ........................................................................... 316

An Analysis of Hedge Fund Strategies – List of Tables

TABLE 51: INDICES AND DECILES SUB-PERIOD ANALYSIS ................................................... 320

TABLE 52: LAGGED 12-MONTH DECILE SUB-PERIOD ANALYSIS............................................. 324

TABLE 53: SUB-PERIOD INDIVIDUAL FUNDS RESULTS ....................................................... 328

TABLE 54: EX-POST BETA SUB-PERIOD ANALYSIS ............................................................ 331

TABLE 55 : DESCRIPTIVE STATISTICS OF THE HEDGE FUND (PANEL A) AND OF THE EQUITY AND BOND

MUTUAL FUNDS (PANEL B) ............................................................................... 371

TABLE 56: MEAN RETURN ESTIMATION FOR AVERAGE INVESTORS .......................................... 380

TABLE 57: MEAN RETURN ESTIMATION FOR PROTECTIVE INVESTORS ...................................... 387

TABLE 58: MEAN RETURN ESTIMATION FOR PROGRESSIVE INVESTORS .................................... 389

TABLE 59: SLOPE ESTIMATION FOR AVERAGE INVESTORS ................................................... 392

TABLE 60: MEAN RETURN ESTIMATION FOR AVERAGE INVESTORS (INCLUDING MUTUAL FUNDS IN A HEDGE

FUND PORTFOLIO).......................................................................................... 397

TABLE 61: SLOPE ESTIMATION FOR AVERAGE INVESTORS (INCLUDING MUTUAL FUNDS IN A HEDGE FUND

PORTFOLIO)................................................................................................. 400
An Analysis of Hedge Fund Strategies – Detailed Table of

Contents

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

An Analysis of Hedge Fund Strategies – Table of Contents ........................................... i

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

Preface.................................................................................................................. i

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

Global Literature Review......................................................................................... 5

The Data Issue.................................................................................................... 14

Investing in Hedge Funds ..................................................................................... 26

Abstract Part One: The Persistence in Hedge Fund Performance ................................ 29

Abstract Part Two: The Neutrality of Market Neutral Funds ....................................... 36

Abstract Part Three: Hedge Funds as Diversification Tools ........................................ 38

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

I Introduction...................................................................................................... 46

II Literature Review ............................................................................................. 47

2 . 1 P e rf o rm a nce Stu d ies . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4 7

2 . 2 Ev o lu tion in P erf o rma nc e M e asu rement . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4 9

III Performance Measurement Models .................................................................... 52

3 . 1 Th e Ca p it a l A sset Pr ic ing M o d e l . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5 2

3 . 2 Th e 3- 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 3

3 . 3 Th e 4- F act o r Mod e l of C ar h ar t ( 1 99 7 ) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5 4

3 . 4 A n Ext en d ed Mu lt i-F ac tor M od e l . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5 5

IV Data .............................................................................................................. 57

4 . 1 D at a P rov id e rs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5 7

4 . 2 H ed g e Fund s . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5 8

4 . 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 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5 9

4 . 4 B ias es in H edge Fund s Da t a . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6 0

V Data analysis ................................................................................................... 61

5 . 1 S ur v iv o rs h ip b ia s . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6 1

5 . 2 Ins tant R e tur n H is t o ry Bias . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6 4

5 . 3 Bas ic P e rf o r manc e . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6 8

5 . 4 C or r e la t ion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7 4

VI Hedge Funds Performance ................................................................................ 76

6 . 1 P e rf o rm a nc e M ea su r e ment us ing the C APM . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7 6

6 . 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 . . . . . . . . . . . . . . . . . . . . 8 0

6 . 3 P e rf o rm a nc e ove r Sh or t er P er io ds . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9 6

6 . 4 C om pa r is on w ith o th er S tu d ies . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9 9

VII Persistence in Performance............................................................................ 100

7 . 1 P e rs ist en c e in On e- 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 . . . . . . 1 0 0

7 . 2 P e rs ist en c e ov er the Asian c r is is . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1 0 6

7 . 3 D iss o lu t ion Fr e qu en c ies . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1 0 7

7 . 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 . . . . . . . . . . . . . . . . . . . . . . . . . 1 0 7

VIII Conclusion ................................................................................................. 115

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

Abstract ........................................................................................................... 120

I Introduction.................................................................................................... 121

II Performance Measurement Models ................................................................... 125

2 . 1 Th e C a p it a l A sset P r ic ing M o d e l . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1 2 5

2 . 2 Th e 4- F act o r M o d e l of C ar h ar t ( 1 99 7 ) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1 2 6

2 . 3 Th e C o m p osit e M od e l . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1 2 7

III Data............................................................................................................ 129

3 . 1 D at aba s e. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1 2 9

3 . 2 Ba s ic P e rf o r man c e . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1 3 0

3 . 3 A n a lys is p e r Sub -p e r io ds . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1 3 2
3 . 4 C or r e lat ion s . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1 4 2

IV Analysis of biases .......................................................................................... 147

4 . 1 S u r v iv o rs h ip b ia s . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1 4 7

4 . 2 Ins t ant R e t u r n History Bias . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1 5 1

4 . 3 C on c lu s ion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1 5 4

V Hedge Funds Performance ............................................................................... 155

5 . 1 P e rf o rm a n c e M ea su r e ment u s ing t he C AP M . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1 5 5

5 . 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 . . . . . . . . . . . . . . . . . . . 1 6 0

5. 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 . . . . . . . . . . . . . . . . . . . . . . . 1 6 1

VI Persistence in Performance ............................................................................. 168

6 . 1 P e rs ist en c e ov er t h e t ot a l p er io d . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1 6 8

6 . 2 P e rs ist en c e ov er t h e su b-p er io ds . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1 7 0

6.3 Analys is of the Mark et N eu t r a l strateg y . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1 7 4

VII Conclusion................................................................................................... 186

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

Abstract ........................................................................................................... 190

Introduction...................................................................................................... 191

I Methodology ................................................................................................... 195

II Database ...................................................................................................... 200

III Preliminary Analysis ...................................................................................... 203

3 . 1 D es c r ip t ive S t a t is t ic s . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2 0 3

3 . 2 C or r e la t ion An a ly s is . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2 0 8

3 . 2 S u r v iv o rs h ip B ia s . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2 0 9

IV Global Results: Market Analysis....................................................................... 210

4 . 1 P e rf o rm a n c e Ana ly s is . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2 1 0

4 . 2 P e rs ist en c e in P e r f o rm an ce - R et u rns . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2 1 6

4 . 3 P e rs ist en c e in P e r f o rm an ce – Three Years of D at a . . . . . . . . . . . . . . . . . . . . . . . 2 2 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 . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2 2 3

V Further Analysis ............................................................................................. 263

5.1 Sub- Period An a ly s is . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2 6 3

5 . 2 D at e Imp a ct . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2 6 6

VI Frictions in the Real World .............................................................................. 267

VII Conclusion................................................................................................... 269

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

Abstract ........................................................................................................... 273

Introduction...................................................................................................... 274

I Interest of the study ........................................................................................ 279

II Database ...................................................................................................... 282

III Descriptive statistics and attrition rates ........................................................... 283

3 . 1 D es c r ip t ive s t a t is t ics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2 8 3

3 . 2 C or r e la t ion a n a ly s is . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2 9 0

3 . 3 B irth a n d a t t rit ion rat e s . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2 9 2

IV Survivorship bias analysis............................................................................... 295

V Methodology .................................................................................................. 299

VI Strategy and decile analysis............................................................................ 303

VII Individual fund analysis................................................................................. 307

VIII Sub-period analysis ..................................................................................... 317

8 . 1 M ar k et exp o su re . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3 1 7

8 . 2 Ind iv id u a l funds . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3 2 6

8 . 3 E x- p os t b et a s u b -period an a ly s is . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3 3 0

IX Conclusion.................................................................................................... 333

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

Abstract ........................................................................................................... 339

Introduction...................................................................................................... 340

I Utility functions and Spanning........................................................................... 342

1.1 Utility functions an d the Bell function . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3 4 2

1 . 2 S p an n in g. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3 4 8

II Methodology and data .................................................................................... 352

2 . 1 C las s ic a l p o rt f o lio select ion app r o ach e s . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3 5 3

2 . 2 T ay lo r a pp r ox ima t ion a n d r is k m e asu r e . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3 5 6

2 . 3 R is k M easu r e . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3 5 9

2 . 4 E s t ima t ion method olo gy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3 6 3

III The database................................................................................................ 369

IV Results......................................................................................................... 377

4 . 1 S ign ific a nc e level . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3 7 7

4. 2 Inclu din g h edg e f u n d s in mutual funds p or t f olio . . . . . . . . . . . . . . . . . . . . . . . . . 3 7 8

4 . 3 Inc lu din g mu t u al f u n ds in hedg e fund p o rtf o lios . . . . . . . . . . . . . . . . . . . . . . . . . 3 9 5

V Conclusion ..................................................................................................... 402

General Conclusion ............................................................................................ 404

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

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

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

 © Copyright by Daniel Capocci 2007
All rights reserved

© Copyright by www.jlbphoto.net for Daniel Capocci’s picture on the back

440
Hedge Funds are relatively unknown investment vehicles. This PhD Thesis focuses on

analyzing hedge fund strategies and focuses more precisely on hedge fund returns. The

Thesis is divided in three parts. Part 1: The Persistence in Hedge Fund Performance

focuses on the explanation of 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. Finally, Part 3: Hedge Funds as Diversification Tools aims at determining if

investors should include hedge funds in a classical portfolio of stocks and bonds. We

develop an original methodology and present it and its application in details. Figures are

added when more illustrations are needed. This Thesis covers the whole analysis of

hedge fund returns.

Daniel Capocci, Ph.D.- Chartered Alternative Investment

Analyst, 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.
Daniel has published several papers in academic and
business journals such as the Journal of Empirical Finance,
the European Journal of Finance, Global Finance Journal,
l’Echo, Agefi, Haute Finance, aso. and he has contributed
to several hedge fund readers and presented his research
to several conferences. This Thesis is an aggregation of
his main research in the hedge fund area.

441