Development of an Analytical Framework for Hedge Fund Investment

Transcription

1 Development of an Analytical Framework for Hedge Fund Investment Nandita Das Assistant Professor of Finance Department of Finance and Legal Studies College of Business, Bloomsburg University 400 East Second Street, Bloomsburg, PA Tel: (570) web: Abstract This research attempts to develop an analytical framework for hedge fund investment. Various issues related to the hedge fund investment have been addressed. This paper focuses on three distinct areas of hedge fund research, namely, bias in hedge fund data, the classification of hedge funds, and performance attribution of hedge funds. All studies, reported in this paper, have been carried out using the ZCM/Hedge (formerly Mar/Hedge) database. The bias result varies from +0.13% to -0.44% for different categories. The bias study is carried out for three different lengths of study periods using two different measures of bias. The results show that the bias becomes more negative with the increase in the study period. The performance of disappeared and new portfolios is studied. The validity of the CAPM model is tested using the ZCM/Hedge database. The data does not support any of the implications of the CAPM. Hedge-fund databases vary with the types of funds included and their classifications. Research results on hedge-fund performance may then differ depending on the database, making them difficult to compare. This research uses a cluster-analysis approach to classify hedge funds. The results of the study are compared with the existing classifications of the ZCM/Hedge database. Analytical models to explain hedge fund return are developed using exact factor-pricing models: the macroeconomic factor model and the fundamental factor model. The results of the macroeconomic model show that only two state variables are statistically significant. This lends support to the similarity hypothesis that the macroeconomic factors that explain equity return also have explanatory power for hedge fund returns. Although not all the R-square values are impressive, it appears that the five-factor model does explain approximately 30% to 40% of the variation in hedge fund return. The fundamental factor model is developed using fund attributes that should have an effect on hedge fund return. The analysis is carried out using the OLS and WLS estimation procedures. The results are different for the different estimation procedures, leading to the conclusion that caution is appropriate when interpreting results from OLS coefficients because of presence of heteroscedasticity in the cross-sectional data. It appears from the analysis that the fundamental variables chosen for the model are successful in explaining hedge fund return for hedge funds domiciled in the US, but not for hedge funds domiciled outside the US. The author is thankful to the Foundation for Managed Derivatives Research for providing a research grant, to Mr. Richard E. Oberuc of LaPorte Asset Allocation System for providing the ZCM/Hedge database and to Lehigh University for the Warren-York Fellowship for this research project. The author is thankful to Dr. D.L. Muething and Dr. L.W. Taylor for their guidance of this research work. The author is also thankful to Dr. Richard J. Kish for his advice. The author is thankful to the participants at the 00 FMA Doctoral Student Consortium, the 003 Eastern Economic Association Annual Meeting, and the 9th International Conference on Computing in Economics and Finance for their helpful suggestions and comments.

2 1.0 Introduction Hedge funds, as an alternative investment vehicle, have enjoyed healthy growth in recent years and continue to increase in popularity. High net worth individuals have dominated the hedge fund industry for a long time. An increasing number of institutions are allocating a small portion of assets to alternative investments owing to the longterm success of some hedge funds. Hedge funds became popular for their philosophy of trying to outperform the overall market through individual stock and security selection and by taking market neutral positions in an effort to protect financial capital in times of market volatility. Today, the term Hedge Fund is used to describe a wide range of investment vehicles that can vary substantially in terms of size, strategy, and organizational structures. Work has been done on the benefits of adding hedge funds to the traditional investment portfolio, the performance characteristics of hedge funds, and the market impact of hedge funds. The study of performance persistence in the hedge fund industry is a recent phenomenon. Bias is closely linked to the issue of performance persistence; the direction of the bias is not clear even for traditional investments like mutual funds. For hedge funds, the issue becomes more complicated, because it is possible that hedge funds disappear from the database for various reasons. It is necessary to estimate bias to better measure performance and to get an idea of the relative performance. Hedge funds have different performance characteristics depending upon their investment strategy. It is important that bias be estimated for different categories to help measure the performance accurately. Hedge funds provide very limited information to investors, mainly periodic (monthly, quarterly, or annual) returns. Sources of data for the industry are the hedge fund database providers. Four main hedge fund databases are used in academics and industry. There is neither legal definition for hedge funds, nor any industry standard for their classification. The databases vary as to the type of funds to include and in their classification scheme. There appears to be a myriad of classifications in existence. There is a need for a unified approach to the classification of hedge funds. In this research project, an attempt is made to develop an analytical framework for hedge fund investment. This framework will enable the investor to estimate return of a hedge fund, based on the strategies and characteristics influencing the return. The number and diversity of hedge funds suggest that the decision to invest in hedge funds should be based on an analytical framework. This paper proceeds as follows: Section describes the Hedge Fund industry, Section 3 briefly describes the various classification schemes of the database providers, Section 4 describes the literature on hedge fund research, Section 5 measures bias in hedge fund data, Section 6 checks the validity of the CAPM model to explain hedge fund returns Section 7 develops a new approach to classify hedge funds, Section 8 models hedge fund return, and Section 9 concludes..0 Hedge Fund Industry In finance industry terminology, the meaning of hedge is the process of protecting oneself against unfavourable changes in prices. The term hedge fund is not defined or separately addressed in any securities or commodity laws. The term has undergone a considerable amount of mutation to represent what it means today compared to what it meant when it first originated in In 1949, A.W. Jones introduced the concept of the hedge fund. He combined a leveraged long stock position with a portfolio of short stocks in an investment fund with an incentive fee structure. Hedge fund investment practices and strategies have evolved and expanded since then. Some of today s hedge funds satisfy all criteria of Jones fund; namely long/short positions and incentive-based fees. With no legal definition of a hedge fund, any fund that satisfies two criteria of Jones fund is identified as a hedge fund. Some hedge funds do not hedge at all. While many hedge fund characteristics have changed significantly, many fundamental features have remained the same. Moreover, hedge funds are no longer unique to the U.S. markets, but exist in many areas around the world. In the United States, they normally offer their shares in private placements and have less than 100 high net-worth investors in order to make use of exemptions provided under the Securities Act of 1933, the Securities Exchange Act of 1934, and the Investment Company Act of l940. In the short history of fifty years, interest in hedge funds and their performance has waxed and waned. In recent years, however, hedge funds have enjoyed healthy growth and appear to have increased in popularity. In particular, the bull market of the late 1980s created more high-net-worth investors. These investors, looking for enhanced returns, started to invest in hedge funds. The renewed interest in hedge funds that began in the late 1980s has not vanished. In 1990, there were about 600 hedge funds worldwide with assets of approximately $38 billion. According to industry publications, at the end of 1998, despite the publicized collapse of Long Term Capital Management (LTCM), there were some 3,300 hedge funds with assets of approximately US$375 billion. The near failure of LTCM in 1998 does not appear to have slowed down the growth of and interest in hedge funds. The LTCM debacle has rightly led to more caution from regulatory authorities and investor interest groups. 1

3 All estimates suggest that the hedge fund industry has experienced tremendous growth since mid 1980s, measured either by the number of funds or by assets under management. Additional investments in the hedge fund industry in years 000 and 001 were US$40 billion and US$80 billion respectively, and the total industry size in the first quarter of the year 003 is between US$600 billion and US$700 billion. Hedge funds invest in a variety of liquid assets just like mutual funds, but are quite different from mutual funds. For example, under current federal law, hedge funds do not have any management limitations. There are virtually no limits on the composition of the portfolios and no mandatory disclosure of information about holdings and performance. Das et al. (00a) provides an overview of the hedge fund industry. 3.0 Hedge Fund Database Providers and Classification Four primary databases are popular among researchers and in the investment industry. Providers of these databases offer different services to the industry. The Zurich Capital Markets (ZCM/Hedge) database (formerly MAR/hedge) provides a comprehensive coverage of global hedge funds. The Hedge Fund Research (HFR) database contains more equity-based hedge funds. TASS is the information and research subsidiary of Credit Suisse First Boston Tremont Advisers. Various database providers classify hedge funds, but in different ways. All the four databases have their own indices based on the categories in the database. The index composition is also different for different databases. Hedge fund categories are based on the self-reported style classifications of hedge fund managers that are listed in a particular database. None of the database provides information on the complete hedge fund universe. The databases differ in the definition of the hedge fund. For example, TASS is the only database that includes the managed futures funds. Unlike hedge funds, managed futures funds limit their activities to the futures market. Following issues are observed about the performance data for various databases. A major limitation of most hedge fund databases is that they typically have data only on funds still in existence or that are new and growing. Most hedge fund indices do not include performance of closed funds. Only those funds that choose to report are included in the database. Not much can be done with this issue due to the industry structure. ZCM/Hedge and TASS have historical performances of all funds that are included in their database. Historical performances are not included (no backfiling) in index construction, but are available for fund analysis. HFR, ZCM/Hedge, and VanHedge have all inclusive selection criteria; they include all funds in their database that classify them as hedge funds. TASS has its own selection criteria. The classification method varies across databases making them difficult to compare. Hedge fund managers employ a diverse array of strategies. The database providers classify hedge funds based on the voluntary information that they collect from the hedge fund managers. Style definitions and the number of categories of hedge funds differ among the database providers. The classification of hedge funds by various database providers is briefly described here. The ZCM/Hedge database classifies hedge funds into four general classes and ten broad categories of investment styles, as reported by the managers of the hedge fund. The classes are onshore hedge funds (HF-US), offshore hedge funds (HF-NON), onshore fund-of-funds (FOF-US), and offshore fund-of-funds (FOF-NON). Some of the categories have further sub-classifications. TASS is the information and research subsidiary of Credit Suisse First Boston Tremont Advisers. It has nine categories of hedge funds, classified based on the investment styles of hedge fund managers. Figure 3.1 shows the classification of the ZCM/Hedge and TASS database. The Hedge Fund Research (HFR) has twenty-six categories of hedge funds. Some of these categories are merely a type of financial instrument or a geographic area for investment. This classification can be reorganized into eleven categories. Some of the categories have further sub-classifications. The VanHedge maintains an extensive database of hedge funds. It provides consultancy and detailed generic performance data on hedge fund styles. VanHedge database can be organized into thirteen categories and five subcategories. 3.1 Alternative Classification Requirement There exists a lot of variation in the definitions, calculation methodologies, assumptions, and data employed by the different managers and databases. It is necessary to benchmark hedge fund manager practices relative to their peers as hedge funds follow diverse strategies. Multiple peer groups may be relevant depending on the strategies employed by the manager, and it is important to clearly identify a peer for the various hedge fund strategies. This may not be an easy task since hedge fund managers refrain from disclosure.

4 Category Event Driven Fund-of-Funds Diversified Niche Global Macro Opportunistic Long Only/ Leveraged 1a 1b 5a 5b 5c 8a Subcategory Risk Arbitrage Distressed Securities International Regional Established Regional Emerging Long/Short Category Event Driven Emerging Market Global Macro Long/Short Equity Convertible Arbitrage Equity Market Neutral 1a 1b 1c 1d Subcategory Risk Arbitrage Distressed Securities High Yield Regulation D 8 9 Market Neutral Sector 8b 8c Arbitrage Convertible Arbitrage Stocks 7 8 Fixed Income Arbitrage Dedicated Short Bias 10 Short Sellers. 8d Arbitrage Bonds 9 Managed Futures. (a) (b) Figure 3.1. Classification of Hedge Funds (a) in ZCM database and (b) in TASS database. Hedge funds are primarily distinguished by their use of short-selling, leverage, derivatives and portfolio concentration. Hedge fund manager refrains from disclosure for two reasons. They are not permitted by regulation to advertise to the public. Secondly, the proprietary nature of the trades may result in herding. Hedge fund managers profit by identifying arbitrage opportunities. These opportunities are based on very slim price differentials, but the manager hopes to profit by properly timing his trade and through portfolio concentration. There is a need for an alternative approach to hedge fund classifications given the lack of pure hedge fund types that exist in the industry. The hedge fund literature shows an almost complete reliance on the existing hedge fund classifications. Performance comparison of various hedge funds with the existing hedge fund indices return data is not appropriate as a particular hedge fund could be classified in two or more classes depending on the database. Table 3.1 compares the classifications of ZCM/Hedge, HFR, TASS and VanHedge databases. It appears from Table 3.1 that research on hedge fund performance may produce different results, based on the database used. There seems to be no common comparison basis for the existing literature on hedge funds. The disparity that is observed in the numbers produced between different organizations measuring hedge fund performance could be attributed to the varied classification of hedge funds. Goldman Sachs & Co. & FRM (1998) describe various methods used by hedge fund managers. The description of various hedge fund styles certainly does not cover all the permutations, but provides an overall idea of the various strategies used by the managers. 4.0 Literature on Hedge Funds The study of hedge funds in academics and in industry is a recent phenomenon. As such, most of the literature is less than a decade old. The study of hedge funds started receiving attention after the Asian and LTCM crisis. Work has been done as to the benefits of adding hedge funds to the traditional investment portfolio, the performance characteristics of hedge funds and the market impact of hedge funds. The empirical work on hedge fund can be divided into the following categories: 1. Performance Attribution (Modeling Returns),. Performance Evaluation and 3. Characteristics and the Impact on the Financial Markets. 3

5 Table 3.1. Comparison of ZCM/Hedge Classification with HFR, TASS and VanHedge Classifications. Item ZCM/Hedge Classification HFR Classification TASS Classification VanHedge Classification a. Event Driven: Risk Arbitrage Event Driven Event Driven: Special Situation 1 Merger Arbitrage Risk Arbitrage b. Event Driven: Distressed Securities Distressed Securities Event Driven: Distressed Securities Distressed Securities Fund of Funds Fund of Funds Not Applicable Fund of Funds 3 Diversified Fixed Income: Diversified Not Applicable Several Strategies Market Timing 4 Niche Fixed Income: High Yield Regulation D Fixed Income: Mortgage-Backed Fixed Income: Convertible Bonds Event Driven: High Yield Event Driven: Regulation D Aggressive Growth 5 Global Emerging Markets Emerging Markets Emerging Markets 6 Macro Opportunistic Macro Market Timing Relative Value Arbitrage Statistical Arbitrage Global Macro Macro Opportunistic Value Managers 7 Long Only / Leveraged Equity Nonhedge Not Applicable Not Applicable a. Market Neutral: Long/Short Equity Hedge Long/Short Equity Market Neutral: Securities Hedge 8 b. Market Neutral: Arbitrage Convertible Convertible Arbitrage Convertible Arbitrage Market Neutral: c. Market Neutral: Arbitrage Stocks Equity Market Neutral Equity Market Neutral Arbitrage d. Market Neutral: Arbitrage Bonds Fixed Income Arbitrage Fixed Income Arbitrage 9 Sector Sector: Energy, Sector: Financial, Sector: Health Care/Biotech, Sector: Metals/Mining, Sector: Real Estate Sector: Technology Not Applicable 10 Short Sellers Short Selling Dedicated Short Bias Short Selling Financial Services, Health Care, Income, Media/Communications Technology 4

6 4.1 Performance Attribution (Modeling Returns) Attribution analysis attempts to find the factors affecting hedge fund return. A limited number of academic researches have focussed on dissecting the sources of hedge fund return. Research has been done on the broader category of hedge fund performance and on particular hedge fund strategy. Research in this area can be divided into three groups: 1. Modeling hedge fund performance as a group: The researchers model hedge fund performance treating all of the hedge funds in a database as a group. No distinction is made between the different categories of hedge funds, and hence, hedge funds performance as a group compared to other asset classes is the result obtained from the studies.. Extracting strategies from observed returns: Different managers and databases classify hedge funds differently. One particular hedge fund could be grouped under one category (e.g. based on strategy) in one database, whereas the same hedge fund would be listed under a different category (e.g. based on investment sector) in some other database. The researchers extract strategies from observed returns and try to reclassify hedge funds based on observed return characteristics. 3. Modeling particular hedge fund strategy: The researchers concentrate on modelling a particular hedge fund strategy. They do not reclassify hedge funds, nor do they treat all different categories of hedge funds as a group. They take the database classification as given and study only one strategy at a time. 4. Performance Evaluation Performance Evaluation is essentially concerned with comparing the return earned on a hedge fund with the return earned on some other standard investment asset. Research in this area can be divided into three groups. 1. Benchmarking: Webster s Dictionary defines a benchmark as a standard or point of reference in measuring or judging quality, value, etc. An investment benchmark is a passive representation of a manager s investment process. It represents the prominent financial characteristics that the investment would exhibit in absence of active investment judgment.. Performance persistence: The researchers examine whether hedge fund managers demonstrate persistence in their performance and how the survival rate affects performance persistence. 3. Performance in a portfolio context: The researchers study the performance of hedge funds in a portfolio context, i.e., the diversification benefits of including hedge funds in a traditional portfolio of stocks and bonds. 4.3 Characteristics and Impact on Financial Markets In this category, the researchers study the characteristics of hedge funds and their role in the financial markets. This category can be divided into two groups: 1. Hedge Fund Industry Characteristics: The researchers study the characteristics of the hedge fund industry including the fee structure, data conditioning biases, and the risk/return characteristic of fund strategies.. Role of Hedge Funds in the Financial Markets: The researchers study the role of hedge funds in the financial market crisis and the implications for policy. 4.4 Conclusions from Empirical Work on Hedge Funds The empirical work on hedge funds leads to the following conclusions: Hedge funds consistently outperform mutual funds but not standard market indices, Hedge funds returns are more volatile, Inclusion of hedge funds in diversified portfolios raises the efficiency of these portfolios, Hedge funds have low correlation with traditional asset classes, Fund of hedge funds offer diversification benefits to some extent, Hedge funds may have risk-adjusted performance persistence, There may be diminishing-return-to-scale in the hedge fund industry, Hedge funds did not have any direct role in precipitating risk in the financial market, Incentive fee structure does not lead hedge fund managers to take more risk because of the possibility of non-survival, and Hedge funds follow a very dynamic strategy. Hedge fund literature has been discussed under four categories. A summary of the literature on performance attribution and performance evaluation is provided in Table 4.1 and Table 4. respectively. Literature on hedge fund characteristics is summarized in Table 4.3. Table 4.4 summarizes the impact of hedge funds on financial markets. 5

7 Table.1. Performance Attribution (Modeling Returns). Author Issue Key Finding Time Period Database Modeling hedge fund performance as a group Use common set of factors to explain returns for Factors used explain the differences in investment return of these asset Jan 1990 to - active management of hedge funds, stock and bond classes. Dec 1995 mutual funds, and CTAs. Schneeweis and Spurgin (1998) Ackermann et al. (1999) Fung and Hsieh (1997) Brown and Goetzmann (001) Fung and Hsieh (001) Isolate hedge fund characteristics that explain the performance and volatility of hedge funds. Develop an integrated framework for analyzing traditional managers with relative return targets (mutual funds) as well as alternative managers with absolute return targets (hedge funds and commodity trading advisers). Study monthly return history of hedge funds. The authors use both return history and self-reported style information to characterize categories of hedge fund styles. Model the nonlinear relationships between style factors and the markets in which the hedge funds trade. Table.a. Performance Evaluation: Benchmarking. Ackermann et al. (1999) Incentive fees consistently explain risk-adjusted performance , monthly Extracting strategies from observed returns Sharpe s style regression not appropriate for performance attribution of , hedge funds and CTAs. monthly Hedge funds and CTAs have low correlation with returns on mutual funds and standard asset classes. Differences in investment style contribute about 0% of the cross sectional variability in hedge fund performance. The natural groupings like the global equity, US equity hedge and Global Macro styles take more risk than other hedge funds. Modeling particular hedge fund strategy The trend-following strategies can be modeled using look-back straddles to Jan 000, monthly MAR and HFR TASS - - Author Issue Key Finding Time Period Database Performance of hedge funds. Hedge funds consistently outperform mutual funds, but not standard , MAR and market indices. monthly HFR Hedge funds are more volatile than both mutual funds and market indices. Brown et al. (1999) Edwards and Liew (1999) Schneeweis and Spurgin (1999) Agarwal and Naik (000) Edwards & Caglayan (001) Fung and Hsieh (001) Performance of offshore hedge funds. Performance of managed future & hedge funds, both as stand-alone investments and as assets in diversified stock and bond portfolios. While offshore funds are perceived to post higher returns at considerable risk, the data indicate that returns as well as standard deviation were lower than the corresponding numbers for the S&P 500. Hedge funds and managed futures have provided attractive risk-adjusted returns and score high as stand-alone investments. Inclusion of these assets in diversified portfolio raises the Sharpe ratio of those portfolios , annual , monthly - Offshore Hedge Funds Directory Different ways for estimating alpha. - - The degree of out-performance of hedge fund strategies over a portfolio of passive strategies. Performance of hedge funds and commodity funds in bear versus bull stock markets. The need for a performance benchmark for hedge funds. Hedge fund managers exhibit superior market timing and security selection ability. Almost all hedge fund styles exhibit significantly higher positive correlation with stock returns in bear markets than in bull markets. Hedge fund categories should be reclassified into key hedge-fund styles, that is, pairs of strategy and location. MAR - HFR

8 Table.b. Performance Evaluation: Performance Persistence. Author Issue Key Finding Time Period Database Park and Staum Whether there is evidence of skill persistence in the Evidence shows that hedge funds have risk-adjusted - - (1998) alternative investment industry. performance persistence. Brown et al. The evidence for performance persistence of offshore hedge Performance persistence not found on a pre-fee basis, that - - (1998, 1999) funds. is, performance fees are unrelated to future performance. Agarwal and Performance persistence within individual hedge fund Results indicate a reasonable amount of performance - HFR Naik (000) strategies using both parametric and non-parametric methods. persistence but more so for losers. Agarwal and Performance of hedge funds using a multi-period The extent of persistence decreases as the return - HFR Naik (000) Lamm and Ghaleg-Harter (000) framework. Whether the hedge fund managers who outperform tend to repeat. Table.c. Performance Evaluation: Performance in a Portfolio Context. measurement interval increases. Winning hedge fund managers tend to repeat during subsequent periods varying from one quarter to two years. Outperformance persistence exists when returns are adjusted for style exposure to traditional assets , monthly Evaluation Associates Inc. (EAI) Author Issue Key Finding Time Period Database Goldman Sachs Potential benefits of including hedge funds in plan Plan sponsors may be able to utilize certain hedge fund - HFR and Co. (1998) sponsors portfolios. strategies to broaden their sources of return and improve the risk-adjusted returns. Edwards and Performance of hedge funds as assets in diversified stock Inclusion of hedge funds in diversified portfolio raises the , MAR Liew (1999) Lamm and Ghaleg-Harter (1999) Agarwal and Naik (000) Goldman Sachs and Co. (000) Lamm and Ghaleg-Harter (000) and bond portfolios. Appropriateness of including fund of hedge funds and principal-protected versions of hedge funds in portfolios and the allocation that these products should receive. The risk-return trade-off observed by including hedge funds in the portfolio. The theoretical impact of allocating 10% of the assets in pension plan to a portfolio of absolute return funds. Design a portfolio of hedge funds that possess the desired alpha and beta characteristics. Sharpe ratio of the portfolios. Hedge funds enter efficient frontiers across virtually all risk levels, even when relatively low returns are assumed. monthly , monthly FRM Hedge funds provide better opportunities for diversification. - - Hedge funds have a very good diversification benefit with low correlation with common asset class benchmarks. The efficient frontier shifts downward as restrictions are progressively tightened , monthly , monthly FRM Evaluation Associates Inc. 7

9 Table.3. Hedge Fund Characteristics. Author Issue Key Finding Time Period The effect of high water-mark compensation High water-mark lend to managers having an incentive to taking risk. scheme. Existence of high water-mark is due to diminishing returns to scale in the hedge Goetzmann et al. (1998) Database annual fund industry. Skill statistic is leverage-invariant Offshore Funds Directory Park and Staum (1998) The issue of skill persistence and the shortcomings of general risk measures. Schneeweis and Misconceptions about hedge funds. Not all hedge funds use derivatives. The principal economic benefit of hedge - - Spurgin (1998) funds is to provide capital to relatively illiquid investment markets. Ackermann et Characteristics of hedge fund industry and the Positive relationship between the life of funds & size, and negative relationship , MAR and HFR al.(1999) impact of data conditioning biases. between life of funds & incentive fee. Termination and self-selection biases are monthly most powerful. Edwards (1999) Hedge fund industry study. High returns of hedge funds reflect the high risk that hedge fund manager takes MAR Fung & Hsieh Return characteristics of different styles of Global/Macro fund is positively correlated with stocks. - - (1999) hedge funds. Fixed Income Arbitrage return is insensitive to US equities. Lamm et al. (1999) Reasons for superior returns of hedge funds. Lack of transparency and illiquidity contribute to superior performance EAI Purcell & Crowley Hedge fund structures and strategies, and Hedge funds are riskier than the traditional accounts , MAR (1999) analyze hedge fund performance. Hedge fund risk-return characteristics and correlation has diversification benefits. monthly Fung and Hsieh Different types of biases present in the hedge Suggest fund-of-funds as a better proxy for market portfolio based on the smaller - TASS (000) fund performance data. impact of biases inherent to individual hedge fund returns. Goldman Sachs Trends in hedge fund industry. The average equity-oriented hedge funds use less leverage than a fixed -incomeoriented , FRM (000) hedge fund even for comparable investment strategies. monthly Lamm et al. (000) Performance of hedge funds over a five-year Individual hedge fund behavior differs significantly by the type of strategy , EAI period. employed. Hedge funds are highly correlated with each other. monthly Brown et al , (001) annual Whether hedge fund and CTA return variance depends upon the manager s performance. The factors contributing to fund disappearance. Table.4. Hedge Funds and the Financial Markets. Trade-off between maximizing single-period fee option and survival. Survival depends on volatility, age and both absolute and relative performance of the fund. Offshore Hedge Funds Directory Author Issue Key Finding Time Period Database Yago et al. Role of hedge funds in Asian crisis. Hedge funds were at the rear end in liquidating their forward contracts - - (1998, 1999) on Asian currencies. Eichengreen and The Asian currency trade. Evidence shows that hedge funds were not the first to liquidate Mathieson (1998) contracts. Edwards (1999) The policy implications of the collapse of LTCM. Need for better risk management technique MAR Brown et al. Testing of the hypothesis that hedge funds were Hedge fund managers as a group did not cause the crash. - (000, 001) responsible for the 1997 crash in the Asian currencies. Fung and Hsieh (000) Hedge fund exposures during a number of major market events. No evidence of hedge funds using positive feedback trading strategies. - MAR, TASS 8

10 5.0 Analysis of Hedge Fund Data Bias The study of performance persistence in the hedge fund industry is a recent phenomenon. There appears to be performance persistence in hedge funds although there is still a debate as to whether the persistence is more among losers than among winners. Bias is closely linked to the issue of performance persistence; the direction of the bias is not clear even for traditional investments like mutual funds. For hedge funds, the issue becomes more complicated, because it is possible that hedge funds disappear from the database for various reasons. There is no way to track the disappeared funds. It is necessary to estimate bias to better measure performance and to get an idea of the relative performance. Hedge funds have different performance characteristics depending upon their investment strategy. 5.1 Literature on Bias Survivorship bias is the effect of considering only the performance of funds that are present in the database at a given time. Since investors are interested only in the funds that are available to them, most databases do not provide the performance of the defunct funds. Performance studies that use only surviving funds will result in biased measures. Much work has been done in providing estimates of survivorship bias for traditional investments. All these studies have documented an upward bias in measures of performance. For bond funds, Blake, Elton and Gruber (1993) find an upward survivorship bias of 7 basis points per annum. For equity funds, various researchers have come up with different estimates of survivorship bias. This apparent difference in estimates of survivorship bias for the same investment class is attributable to differences in methodology used and on the length of the study period. For Commodity Trading Funds (CTAs), researchers have estimated an upward survivorship bias to range from 350 to 470 basis points. The estimates of survivorship bias for hedge funds range from 16 basis points to 300 basis points. Brown, Goetzmann, and Park (1997), Fung and Hsieh (1998), Brown et al. (1999), and Liang (000) calculate survivorship bias as the performance difference in the equal-weighted portfolios of surviving funds and that of all funds existing in the database. Ackermann, McEnally, and Ravenscraft (1997) calculate survivorship bias as the performance difference between surviving funds and disappeared funds. All the studies in hedge fund survivorship bias have mostly concentrated on estimating bias for the complete database, except for Liang (000). The databases used for all these studies are different. Fung and Hsieh (1997) used combined data from Paradigm LDC and TASS; Brown et al. (1999) used the US Offshore Funds Directory; Ackermann, McEnally, and Ravenscraft (1999) used combined data from the HFR and Managed Account Reports Inc. (Mar/Hedge) databases; and Liang (1999 and 000) used the TASS and HFR databases. The period of study also differs, as does the procedure of calculating the survivorship bias. However, all these studies have same conclusion that hedge fund survivorship study is different from other survivorship studies. It is clear that there is an impact of survivorship bias on performance measures, but the direction of the bias is not clear even for traditional investments like mutual funds. There are two views regarding the impact of survivorship bias. One view, originating from the work of Brown et al. (199), is that survivorship bias will result in induced persistence in fund performance. The alternative view (Grinblatt and Titman, 199) relies on a market correction mechanism. It states that poor performers are the most likely candidates for disappearance from the dataset and the proportion of funds with an inconsistent performance record will increase, leading to induced non-persistence or performance reversals. These two views differ on the direction of the impact of survivorship bias, but are in agreement with the source of bias, namely poor performers. The direction of the bias is more complicated for hedge funds, as they presumably disappear from the database for two opposite reasons. Some funds disappear because of poor performance and others disappear because they no longer need new money or are closed to the general investors who have access to the database. This is because, unlike traditional investments, hedge funds are not required to disclose any information, and they report to databases with the sole purpose of attracting new investors. There is bias due to the disappearance of funds from the database. There could be bias also in the performance because funds presumably enter the database with positive performance experience. These three factors that lead to bias in the performance measurement could counteract each other depending on the magnitude and direction of bias. This study thus uses the term bias instead of survivorship bias since survivorship bias in mutual funds presumably leads to an upward bias in performance measurement. In this research, the bias is estimated for each class and category of hedge funds and for the complete database, including and excluding fund-of funds. This study is comprehensive and allows for easy comparison of the results with the results obtained by other researchers. Most researchers have estimated bias for the complete database and differ only as to including fund-of funds in their study. The present study also varies the period under study to observe the impact of the length of the study period on bias. 9

11 5. Data Organization and Methodology The ZCM/Hedge database provides monthly returns for all the funds. The database began collecting data from January 1994 and there is no performance record of funds that disappeared prior to January Table 5.1 describes the data used for this study. Four year and three year dataset is used to observe the effect of the study period on bias. Table 5.1. Available Data and Study Period Dataset Composition. Category/Class/Total Total Database Study Period Study period Observations Funds Observations Funds Panel A. Category Event-Driven 19 1, ,69 6,699 5,106 FOF-US 17 14, ,407 7,597 5,807 FOF-NON 9 16, ,033 9,048 7,101 Global International 70 4, ,356,518,009 Global Regional Established 459 8, ,38 16,87 1,861 Global Regional Emerging 135 6, ,858 4,63 3,737 Global US 178 8, , Global Macro 150 8, ,10 3,56,709 US Opportunistic 40 1, Long Only/Leveraged 35 1, ,388 1, Market Neutral 483 6, ,183 15,107 11,81 Sector 179 7, ,715 5,368 4,419 Short Sellers 37, ,619 1, Total, ,, ,7 73,001 57,7 Panel B. Class HF-US , ,665 34,171 6,16 HF-NON 85 44, ,167 3,841 18,567 FOF-US 17 14, ,407 7,597 5,807 FOF-NON 9 16, ,033 9,048 7,101 Total, ,, ,7 74,657 57,691 The bias can be calculated in two different ways, as in Brown et al. (1999). A surviving portfolio in a particular month during the study period consists of all funds that have reported return until the end of the study period. An observed portfolio consists of funds that are in the database for that particular month irrespective of their start and end date. A complete portfolio consists of funds that have reported returns for the complete study period. Figure 5.1 illustrates the portfolio construction process. For the month of October 000, the surviving portfolio consists of B, C, D, E, and F funds; the observed portfolio consists of B, C, D, E, F, and G funds; and the complete portfolio consists of funds C, D, E, and F. It can be seen that the complete portfolio consists of funds C, D, E, and F in each month. January 1994 October 000 November 000 December A A - B B B C C C C D D D D E... E E E F F F F G G H Figure 5.1. Portfolio Construction for the Study of Bias. 10

12 The portfolio returns are calculated using an equal-weighted and value-weighted approach. An equal-weighted portfolio invests equal amounts in each hedge fund irrespective of the size of the hedge fund. A value-weighted portfolio invests in hedge funds based on the market value of the hedge fund and thus gives more weight to larger hedge funds than smaller ones. Bias is calculated in two ways: the difference between the return of the surviving portfolio and the observed portfolio (SP-OP), and the difference between the return of the complete portfolio and the observed portfolio (CP-OP). The return data is available for each month. Bias is calculated on a monthly basis and is reported as average monthly bias for each year of study. Bias is calculated for each category, each class, total database including fund-of-funds, and total database excluding fund-of-funds. Bias results for the 84-month study period ( ) are compared with that for the 48-month ( ) and 36-month ( ) study period. 5.3 Bias Study Results This study considers both after-fee returns and before-fee returns. A before-fee return is more robust than the after-fee-return, because of the vagaries of the fee structure and the complexities of calculation. The investors are concerned with the after-fee return. The bias study is carried out for each class, category, and total dataset separately. The bias is calculated both for after-fee return and before-fee return and for equal-weighted and value-weighted returns. The bias is calculated based on the complete and observed portfolios. The complete portfolio is the portfolio of hedge funds that have survived for the entire study period. By definition, the number of funds in the portfolio remains fixed for each year of study for the complete portfolio. The surviving portfolio consists of funds that have survived until the end of the study period (December 000), irrespective of the starting date. The observed portfolio consists of all hedge funds that are in the database when the monthly return is calculated. The bias study result is shown in Table 5.. The bias is calculated for the equal-weighted, for the value-weighted return, and for different study periods. Table 5. also shows the corresponding t-statistic for the bias calculations. Approximately 45% of the categories have significant t-statistic for bias calculated as CP-OP, and approximately 55% have a significant t-statistic for bias calculated as SP-OP. In general, all the classes appear to have a significant t- statistic for bias calculated as CP-OP and for total hedge funds both including and excluding fund-of funds. However, bias for the total class does not give an idea about the bias for different categories as is clear from Table5.. The bias study is done for 48-month and 36-month study period also. The results are not reported here. The focus for studying three different study periods is to observe the effect of length of study period on the bias results. Since monthly hedge fund returns are available, all bias calculations are done at a monthly level and then average bias results are computed for the study period. This method of computation should result in a better estimate of bias as compared to the result obtained as the difference in annual return because the number of funds in the database changes from month -to-month and so does the performance of the portfolio. The direction of the bias differs according to category. It can be said that if bias is a negative number, then the average performance of survived new funds and disappeared funds (new and old) is greater than the average performance of funds that have survived for the complete study period. It then becomes important to analyze the performance of survived new and disappeared funds. The reason for the disappearance of hedge funds from the database is not straightforward as it is for mutual funds. Hedge funds drop out of the database for two entirely opposite reasons, poor performance and probably also because of limitations in the arbitrage opportunities in the investment strategy. If funds drop out because of poor performance, this would, in general impart an upward bias in performance measures (performance persistence), if there is no impact from survived new funds. The analysis becomes more complicated when survived new funds also have an effect on the bias. The bias results are different for different hedge fund categories and for different ways of calculating bias (equal-weighted versus value-weighted portfolio). The bias (calculated as CP-OP) result varies from +0.13% to -0.44% for different categories for the 84 month study period, and the overall bias of hedge funds excluding fund-of-funds is % per month using value-weighted method. The corresponding results for the 48-month and 36-month study periods are +0.33% to -0.44% and +0.4% to -0.14%, respectively. The bias result of +156 to -58 basis points per year for different categories is different from the results obtained by other researchers. It is important to note that the database and the methodology used in this study are different from other researchers. This study calculates bias on a monthly basis and then average bias is calculated for the study period. This should definitely result in a better estimate because the monthly calculations are a better representation of the inflow and outflow of funds from the database, and of the monthly fund return data. The other researchers used a different database, calculated bias only for equal-weighted portfolios and for the complete database only. Liang (000) reported a bias of 4 basis points using the TASS database and has reported results for different categories of hedge funds but only for equal-weighted portfolios. The present study is comprehensive as the bias is calculated for both equal-weighed and value-weighted portfolios and also using both after-fee and before-fee returns, for individual 11

13 category class and total hedge funds (both excluding and including FOF). The bias numbers are smaller by 30 basis points per month on an after-fee basis. Fund-of-funds have a slightly lower bias than hedge funds alone. Results are provided for hedge funds, both including fund-of funds and excluding fund-of-funds. There does not appear to be much difference in the results. Composite bias for all categories can be inaccurate as the bias results of different categories may cancel out. The result for the complete hedge-fund database is provided for comparison with the results obtained by other researchers. In addition, the present study is the first survivorship study done on ZCM/Hedge database. Table 5.. Average Monthly Bias (before-fee), Category/Class/T otal Average Monthly Equal Weighted EW Bias (%) Value Weighted VW Bias (%) OP SP CP SP-OP t-stat CP-OP t-stat OP SP CP SP-OP t-stat CP-OP t-stat Panel A. Category Event Driven Global International Global Regional Established Global Regional Emerging Global Macro Long Only/ Leveraged Market Neutral Sector Short Sellers Panel B. Class HF-US HF-NON FOF-US FOF-NON Panel C. Total Excluding FOF Including FOF The bias is also calculated as the difference between the surviving portfolio and the observed portfolio (SP-OP). The surviving portfolio considers new funds coming in, but funds that drop out before the end of the study period are not part of the portfolio. In constructing the portfolio, there is an implied assumption that funds that come in anytime during the study period will continue in the database. For example, if a fund enters the database in November 000 it is considered a survivor even though it has an age of only one month. It is possible that the fund will survive, but it is equally possible that it will not survive if the end of the study period is changed to some other time in the future. Calculating survivorship bias, as the difference in the surviving and observed portfolio is still useful as it helps us to understand the performance characteristics of incoming funds. Table 5.3 compares the category and class bias results for different study periods of 84, 48, and 36 months. It can be seen that as the length of study period increases, the CP-OP bias becomes more negative. This can be explained by the fact that the complete portfolio (CP) depends on the length of the study period. The number of funds in the CP portfolio decreases as the length of study period increases. Presumably, these funds have lower risk and possibly lower returns leading to a negative bias. The same analysis can be done for the bias calculated as SP-OP. In this case, the magnitude of the change in bias result as a factor of study period length is much smaller than that of CP-OP bias. It appears that SP-OP is a more stable measure of bias, specifically with the varying lengths of the study period. However, as mentioned earlier this measure of bias has its limitation because of the way the surviving portfolio is constructed. The advantage of this measure of bias is in its robustness not its accuracy, compared to the CP-OP measure of bias. 1