Biases in Hedge Funds Indices

Seminar Paper, 2005

33 Pages, Grade: 5,5 (1,5 in GER)

Free online reading

Table of contents

1. Introduction

2. Hedge Fund Industry

3. Indices

4. Biases
4.1. Attrition
4.2. Classification of Biases inherent to Hedge Funds Indices
4.3. Survivorship Bias
4.4. Self Selection Bias I
4.5. Self Selection Bias II
4.6. Selection Bias I
4.7. Selection Bias II
4.8. Backfilling Bias

5. Conclusions

6. References

7. Appendix

List of tables

Table 1: Overview of Hedge Funds database

Table 2: Comparison of TASS and HFR database

Table 3: Database overlap, Dead Funds

Table 4: Attrition Rates (TASS)

Table 5: Attrition Rates (TASS) –

Table 6: Overview of empirical studies

Table 7: Simulated Survivorship Bias

Table 8: Incubation Time

Table 9: Overview of Hedge Fund Indices

Table 10: Simulated Attrition

List of figures

Figure 1: Hedge Funds Industry (HFR)

Figure 2: Fund of Funds (HFR)

Figure 3: Classification of Biases

Figure 4: Simulated Attrition

Figure 5: Self Selection Bias I – Decision Tree

Figure 6: Self Selection Bias II – Decision Tree

List of Abbreviations

Abbildung in dieser Leseprobe nicht enthalten

1. Introduction

The Hedge Funds industry has grown rapidly over the last years. Especially in the last decade, there has been an extensive capital inflow in this industry. Although Hedge Funds exist for over 50 years now, the recent grow of amounts invested in this asset class has increasingly attracted the attention of financial markets and financial literature.[1] There are several reasons why Hedge Funds attracted interest, such as superior returns over other asset classes, high growth, and the lack of transparency. The performance of some highly successful Hedge Funds managers over the past two decades additionally fuelled the interest of both private and institutional investors in this field. The largest of these funds, namely the multibillion dollar Quantum Fund of George Soros, achieved an annual compound return over 30% for more than two decades.[2] Apart from this aspect, the difficulties of the LTCM Fund in 1998 caused attention.[3] It is also now commonly known that traditional active strategies, such as investing in mutual funds, significantly underperformed passive investment strategies. The only actively managed funds which have shown a higher performance on average have been the Hedge Funds.[4] Moreover, another reason for the capital inflow and the rising attention is that Hedge Funds performed quite well in the last baisse-years. Investors who invested in Equity Portfolios and Mutual Funds lost a lot of money.

Nowadays, modern investors are well informed by Hedge Funds managers who are not getting tired promoting the merit of investing in hedge funds. These advisers draw elaborated graphs showing the benefits of hedge funds to an active managed portfolio. Investors have to believe in the advantages of shifting a significant part of their portfolio to hedge funds. In terms of the classical risk and return measures the advisers are right, high returns, low volatility and above all low correlations to the other asset classes in the portfolio. But as we know only the half is true. The misleading picture of volatility if measured with the classical portfolio instruments and the correlation effects is not solved in this paper. The research interest in this short paper is the distorted picture of returns given by the Hedge Funds Indices because of biases inherent to those indices.

This paper gives an overview of the Hedge Funds Industry and the Hedge Funds Indices that are currently used by investors and highlights the differences between Hedge Funds and traditional Mutual Funds Indices. The problems of setting up those indices because of Hedge Fund idiosyncrasies are discussed. It is also shown why the performance of these indices is misleading due to construction problems. These systematic errors in the Indices are called biases. The paper provides an overview of the biases that can occur, when an Index is set up and why. We will introduce a classification of biases based on three phases. There will be an emphasis on the most popular bias, which is the survivorship bias. To support the existence of biases, the paper gives an overview of some empirical studies, which in general showed quite significant biases, ranging from 0.39 % to 3 % in performance misstating. Additionally the survivorship bias was simulated in this paper, using a methodology introduced in Zimmermann (2000).

Then we will provide some advice on how these biases can be reduced and how they can be avoided by investors.

2. Hedge Fund Industry

The class of Hedge Funds evolved out of the idea that investments could be either short or long-positions. This description is not valid any more and not suitable to define this class of Investments but there are at least some characteristics which describe Hedge Funds. These include largely unregulated organisational structure, flexible investment strategies, relatively sophisticated investors,[5] substantial managerial investments and strong managerial incentives. Furthermore another criterion, at least for the U.S., is that the fund does not confirm to the standards of the Investment Company Act of 1940 and it is therefore not regulated and limited for public offering. These criteria do not really describe the heterogeneous class of Hedge Funds, but provide an idea of what they could be. Hedge Funds could also be described as everything which is not an investment fund.[6]

Most Hedge Funds are typically established as limited partnerships with fewer than 500 investors and offshore funds which are not subject to any regulation.[7] This missing regulation allows Hedge Funds to be very flexible in their investment approach as they can use the whole spectrum of strategies.

The enduring equity downswing and the following sideward movement of the markets fuel the need of the investors for financial instruments that are also able to give a return in difficult economic environments.

As a consequence of the arguments mentioned above, there is an increasing interest in investing in Hedge Funds. Actual numbers of HFR (Hedge Fund Research) are estimating the total amount invested in Hedge Funds to be above 865 Billion USD at the end of 2004.[8]

Abbildung in dieser Leseprobe nicht enthalten

Figure 1: Hedge Funds Industry (HFR)

Source: HFR (2004)

Investments in Hedge Funds are becoming increasingly popular, although the required minimum investments are still quite high.[9] Over 75% of the Funds which are included in the HFR database are requiring a minimum investment of above $ 100.000.[10] This barrier is bypassed by Funds of Funds which are offering access to small investors at lower amounts. The idea behind Funds of Funds is to offer investors a “hassle free” alternative to constructing a basket of Hedge Funds.[11] This group which is now investing in Hedge Funds and additional institutional investors, such as insurance companies and pension funds, are asking for more transparency in the market.

Figure 2 shows the increasing interest in Fund of Funds. The database used is HFR, with the second quarter 2004 as closing date. It is obvious that Funds of Funds are relatively new compared to the whole Hedge Funds Industry and that the growth rates of this class are significantly higher than those of the whole Industry.

Abbildung in dieser Leseprobe nicht enthalten

Figure 2: Fund of Funds (HFR)

Source: HFR (2004)

Hedge Fund Managers are often unwilling to unveil the investments they have taken or the markets on which they activate. They are additionally, as mentioned before, not regulated in most cases.

A first step towards a transparent market is the possibility to compare different Hedge Funds strategies and different types of Hedge Funds. This comparison is eased by using indices.

The higher the transparency, the lower the searching costs for investors. Yet another point for the emergence of Hedge Funds Indices is the wish of investors to compare actual returns in relation to a benchmark. The investors also want to know how successful the manager of the Hedge Funds has been compared to his peers in the last period and how successful the whole Hedge Funds have performed against the traditional asset classes.

Assured by the massive capital inflow and the growing interest of a wider audience there was an expeditious formation of different Hedge Funds Indices.


There are several reasons for the existence of indices. In our opinion, the most important reasons is that investors, regardless in what asset classes they are investing, intend to compare the result of their portfolio. Although there are several absolute return strategies available, at the end of the day the investors want to know the relative return versus the market. Furthermore, they strongly rely on the past performance while they take their investment decision. Yet another reason for indices can be seen in the trend to passive investing. A series of studies showed that the most efficient way of investing is to invest in an index. Over the last years, a series of such products appeared in the market.[12] Although critics in the financial literature can be found, telling that passive investment in Hedge Fund Indices is simply not possible, the number of products is still growing.[13]

For the traditional asset classes like stocks, bonds, commodities etc. there is a wide range of indices that can be used by investors. The composition of most indices, especially for stocks and bonds, is highly transparent and can be easily replicated in the portfolio of an investor.[14] The assets which are the basis of indices are mostly traded on regulated markets and the prices of these assets are very transparent. For Hedge Funds which are not regulated and are not standardized, information about the current value is much more difficult to find in most cases.

Another very important difference is the economic situation and the business model of Indices Providers. The stock market indices are issued mainly by the stock exchanges where the securities are traded or specialized Indexproviders; the same is true for bonds where the indices are set up by big investment banks, which earn their money by trading. The Providers of Hedge Funds Indices are dependent on the disposal of data and on the listing fees they obtain from the Hedge Funds. Accordingly, the most important Hedge Funds Indices are from Data-Providers. The market is very concentrated and lies in the hands of a reduced number of individuals. The most important are HFR, CISDM and TASS.[15] The following table shows a comparison between the data-providers.

Abbildung in dieser Leseprobe nicht enthalten

Table 1: Overview of Hedge Funds database

Source: own figure, based on website information

HFR uses a subset of about 1500 Funds to calculate 33 indices. MARHedge, which was sold to Zurich Capital Markets in March 2001, calculates about 19 indices.[16] CSFB/Tremont is operating the TASS database and calculates 10 indices out of about 650 Hedge Funds.[17]

There are several differences in the databases; Liang (2000), for example, worked out the differences between the HFR and the TASS database. These differences are not only in the calculation, number of surviving Funds or number of dead Funds, but also in the data available for the same Funds. According to Liang (2000), the portion of dead Funds differs between the databases. TASS has a significant higher number of dead funds included in the database.

Abbildung in dieser Leseprobe nicht enthalten

Table 2: Comparison of TASS and HFR database

Source: Liang (2000), p. 317

As showed in table 2, green coloured box, the number of Funds included in both databases as of July 1997 and July 1998 is 465 and therefore quite small. It seems that both databases have different clients and only a very small number is reporting to both databases.

The overlap between the databases makes up only 38.7% for the TASS database and 44.2% for the HFR. This provides enough reason to say that assuming a single database as the “market” portfolio for Hedge Funds is just a very rough approximation for the market (considering that there are several databases on the market). This critic will be considered again in comparison with the empirical studies for the survivorship bias. Yet another point to mention and to be considered while comparing these studies is the difference in the portion of dead Funds in the databases. The portion of dead Funds is more than three times higher in the TASS database as it is in the HFR.

Abbildung in dieser Leseprobe nicht enthalten

Table 3: Database overlap, Dead Funds

Source: own calculation based on Liang (2000), p. 317

These databases were used for most empirical studies which were conducted in this field in the last years. Additionally, they are used as basis for a wide range of indices, also from third parties.

Performance Measurement

The returns of the Hedge Funds in the databases are mostly defined as the monthly change in Net Asset Value (NAV) divided by the Net Asset Value at the beginning of the month.[18] These returns are net of management fees, incentive fees, and other fund expenses. In practice, these returns differ from the actual returns of the investor because of the spreads between bid- and ask-prices, sales, and redemption fees.[19]

Abbildung in dieser Leseprobe nicht enthalten [1]

In practice, there are several difficulties at deriving the actual and accurate Net Asset Value every month. Especially for Hedge Funds which invest in assets where no market values are available, alternative approaches have to be employed to derive a market value. In practice, four methods are used to determine the Net Asset Value:[20]

- If the securities/instruments are traded on an exchange, either market bid-, market offer- or mid-market price is taken;
- If there is a liquid OTC market, three or more quotes are used in order to calculate an average price;
- If there is no liquid market, models like the Black-Scholes are used;
- If there is no model a price committee may determine the price of the asset.

In addition to the problems of deriving market values, Hedge Fund Managers have the incentive to use the lack of transparency for smoothing the monthly returns.[21] Most empirical studies dealing with biases in Hedge Funds Indices calculate the yearly returns as the geometric mean of the monthly rate of returns.

4. Biases

4.1. Attrition

The funds which are included in the databases report on a voluntary basis, therefore no database is comprehensive and the use of the data is very critical for the investment and academic community.[22] The Hedge Fund Industry is rapidly changing over the time; there is an enormous growth in the number of new Hedge Funds in the databases, but at the same time, there is also a big number of funds leaving those databases.

“What many investors do not know, however, is that most of the Hedge Funds do not play the game for long.”[23]

Only 59.5% of the Hedge Funds which were active in 1996, for example, still existed in 2001. For this phenomenon, the notion of “attrition” is widely used in the literature.[24] The attrition rate describes which percentage of funds left the database, without giving exact reasons for the exit. Brown, Goetzmann and Park (2001) calculated a half-life for Hedge Funds of 30 month. This means that after 30 month, 50% of the Hedge Funds are not reporting any more.[25]

Attrition is important because of two reasons. First investors have to search for new investments if the Hedge Fund is closed down. This searching is costly for investors, they have to make efforts to find new opportunities and they often have to pay sale fees and the spread between the bid- and ask-price of the fund. The second reason is that with attrition, there is the possibility of survivorship bias, self-selection bias and selection bias.[26] Although it is not known what had happened to the Funds leaving the database, they are called “dead”.

In formulaic terms, the attrition rate can be written as:

Abbildung in dieser Leseprobe nicht enthalten [2]

InTable 4: Attrition Rates (TASS)the Attrition rates of the database TASS were tested by Amin and Kat (2003). The research considered the years from June to May, because they only had data available until May 2001. Amin and Kat eliminated 171 Hedge Funds from the TASS - database because of incomplete and ambiguous data. This leads to yearly attrition rates of 6.96% for Hedge Funds, 4.19% for Funds of Funds and 6.49% for the whole Hedge Fund Universe. For the attrition rates there is clear evidence for acceleration in the last years. The main reason for the growing attrition is a bigger number of small funds entering the market, if they do not reach enough capital or do have a bad performance they are more likely to close down than bigger funds. In Table 4 Amin and Kat (2003) showed attrition rates for the last sample year of above 12%, both for Hedge Funds and Fund of Funds, this is nearly twice the average of the whole sample period.

Similar research was conducted by Liang, for example. He calculated a yearly attrition rate for the TASS database of 8.30% for the years 1994 until July 1998.[27]

There is the main question why the results differ that heavily? In our opinion, the main reasons for the divergence are that Amin and Kat adjusted the database and they used a different definition of one year, in relation of the period. The differences become even bigger, if the researches are compared for the same time horizon (SeeTable 5: Attrition Rates (TASS) – 1998for details). This shows empirical research conducted in this area should be treated with caution.

Abbildung in dieser Leseprobe nicht enthalten

Table4: Attrition Rates (TASS)

Source: Amin and Kat (2003), p. 58, p.69 and own calculation

Abbildung in dieser Leseprobe nicht enthalten

Table5: Attrition Rates (TASS) – 1998

Source: Amin and Kat (2003), p.58 and p.69 and Liang (2000), p.25; own calculation

4.1.1. Reasons for Attrition

There are several reasons why funds leave the database, meaning that they are not reporting any more to the database or the database does not collect the data any more. The main reasons which can be found in the literature are:

1) Self Selection,
a. the Fund has collected enough money or
b. a lack of performance.

2) Selection of the database,
a. the Fund has a lack of size or
b. a lack of performance or the
c. fund is closed.

3) Liquidation of the Fund,
a. the Fund shuts down and there is no reporting any more.

Additionally to the reasons mentioned above, there are some factors driving the attrition rates. The age of the Funds and the size, for example, are factors that were empirically tested by Brown, Goetzmann and Park (2001).[28] Amin and Kat (2003) tested the size, the past performance, the leverage and the own money invested for their relevance for attrition. They found that for attrition, the size, the past performance and the own money invested are strengthening factors. The leverage, in contrast, did not fuel the attrition.

The attrition is the main reason for a number of biases in Hedge Funds Indices due to the fact that whenever Hedge Funds leave the index there is a chance for a survivorship bias, a self-selection bias or a selection bias.[29]

4.2. Classification of Biases inherent to Hedge Funds Indices

The Hedge Funds Industry is not regulated and therefore the databases consist only of Funds which report and are accepted by the database-provider. This leads to different returns of Hedge Funds in the database and Hedge Funds not captured by the database. The incompleteness of Hedge Fund data has several reasons. First, the reporting is voluntarily, which means that only a portion of the whole Universe of Hedge Funds is captured. Second, most of the commercially available databases were set up in the mid-1990s, as shown inTable 1: Overview of Hedge Funds database. The databases that are offering data for former periods, clearly misstate the performance of the funds, because the funds which closed operations before the databases where set up, are not included in the database. The information of the “dead” funds is lost forever. Third, the databases do have different design criteria regarding data-collection and performance calculation.

These reasons can lead to significant differences between the Hedge Funds in the database and the whole Hedge Funds Universe.[30]

In the following section we propose a classification of the biases under five major headings. In this classification, the liquidation bias, the self selection bias II and the selection bias II are subgroups for the survivorship bias.

- Survivorship Bias
- Liquidation Bias
- Self Selection Bias I+II
- Selection Bias I+II
- Backfilling Bias

Graphically, the context of these biases can be shown as following:

Abbildung in dieser Leseprobe nicht enthalten

Figure 3: Classification of Biases

Source: own chart

As shown in Figure 3 we classify the biases of Hedge Funds Indices into three phases of their occurrence.

Phase I describes the Decision of the Hedge Fund Management (Self Selection) and the Decision of the Index-Provider (Selection), if the Hedge Fund should be captured or not.

In Phase II the Hedge Fund Management decided to provide data and the database provider decided to collect the provided data. If the Fund is included in the database and the data is filled in, since the Hedge Fund was launched, there is a backfilling bias (instant history bias).

The third Phase describes the situation when a Hedge Fund does no longer deliver data. This can happen due to several reasons, such as Liquidation of the Fund, by Decision of the Hedge Fund Management (Self Selection) or by Decision of the database-Provider. The existence of the survivorship bias is reflected in attrition per year.

4.3. Survivorship Bias

The term survivorship bias refers to the conceptual incorrectness of measuring the performance of a portfolio that could have been defined at some time in the past only with a crystal ball, otherwise said.[31] The survivorship bias describes the difference of return in funds which are dead and such which survived. This can be written as:

Abbildung in dieser Leseprobe nicht enthalten [3]

This bias can also be seen in connection with mutual funds. Empirical research from Grinblatt and Titman (1989), Brown and Goetzmann (1995), Malkiel (1995) and Elton, Gruber and Blake (1996), for instance, has shown the existence of this bias. The bias in these studies, however, is relatively small compared to those in Hedge Funds Indices.

The commonly employed data sets of Hedge Funds typically show the past records of all funds currently in existence. An Investor is clearly not interested in the record of a fund, where there is no possibility to invest. A Fund which accepts high risks will have a higher probability of failure and will be excluded of the database. If, however, the Fund performs well the record of this fund will be included and will rise the overall performance of the sample.[32]

The average annual return of the portfolio is calculated by deriving the geometric mean of the monthly rates of return. However, another problem of databases could be that the sample portfolios do not consider the size of the funds and are therefore assuming that at the beginning of every period (e.g. each month) the portfolio is equally invested in all funds containing the database. The more accurate manner for calculating the performance would be to weight the investment by the value of the Funds. However, assets under management are often incomplete or simply not available in the databases. Therefore, the equally weighted portfolio is a widely approximation.[33]

In the literature there are two methods of calculating the actual survivorship bias in terms of performance. One is to calculate the survivorship bias as the difference between the Portfolio of surviving funds at the end of the sample period and a Portfolio of all funds used in this sample (definition A). The second is more restrictive and starts with a surviving portfolio that only includes the funds which survived the entire sample period and deducts the average portfolio return of all funds used in the sample (definition B).[34] The survivorship bias is the most often empirically tested bias. The problem with the results of the different studies is the fact that they deliver a wide range of a possible survivorship bias ranging from 0.39 % to 3.0 % in terms of annual returns.

Table 6:offers an overview of the empirical studies conducted in the last years concerning survivorship bias.

Abbildung in dieser Leseprobe nicht enthalten

Table6: Overview of empirical studies

Source: own chart

4.3.1. Differences of the studies and shortcomings

The empirical studies differ quite heavily in the following testing methodology:

Database: The available databases differ quite heavily in their composition and characteristics. Additionally, the data included in the databases is not even the same for the same funds. Liang (2000) found that the databases of HFR and TASS differ in 5% of the Funds quite heavily in terms of the monthly performance. Furthermore, also in the composition and the status of dead and surviving there is sometimes a difference between these two databases.[35] Moreover, another problem of the databases is the date of the study, because at least the two most important databases are overwriting their historically data and two studies for the same time horizon and database may deliver different results.[36]

Sampling Period: The studies differ in the chosen time horizon. Some studies are considering time horizons that begin in the 1980s. This however is not useful at all, because as mentioned in the section of the databases, the providers started collecting data in the mid-1990s while the data of dead funds is not available for the time horizons before this date. We can consider this aspect as main reason why the study of Ackermann, McEnally and Ravenscraft comes to the conclusion that there is no significant bias in the Hedge Funds Indices. These authors employed only a time horizon from 1988 until 1995. Furthermore, Brown, Goetzmann and Ibbotson (1999) are working with a similar time horizon, but they face fewer problems as they use a different database, where this problem is partly solved.

Selection Criteria: To test the databases on the survivorship bias, researchers employ different sample definitions and criteria. Some researchers exclude Funds with a history of lower than for example two years while others test only a certain group of Funds, e.g. Fund of Funds or CTAs.

Definition of the survivorship bias: As already mentioned before, there are two methods for calculating the survivorship bias: Definition A and B. Brown, Goetzmann and Ibbotson (1999) find large differences between these two definitions, resulting in a survivorship bias (A) of 2.71% and a survivorship bias (B) of 0.69%.

However, there are other difficulties for testing the survivorship bias empirically. Fung and Hsieh (2002) describe some general shortcomings of these studies. First, the Hedge Funds which stopped to exist before the database providers set up their databases are missing in the available data. Consequently, there is clearly a survivorship bias inherent in the databases, which cannot be measured due to the lack of information on defunct Funds.[37] The same holds for the backfilling bias. Second, there is no real “market” portfolio. The only available data are in the databases. There is no distinction whether the Fund decides to stop reporting or whether it is in liquidation. Thus, there is no “market” portfolio which describes the Hedge Fund Universe, but an “observable” portfolio, including the Funds reporting to the database, and a “survivor” portfolio. The incomplete portfolio can give a misleading picture of the real survivorship bias.[38]

4.3.2. Simulation of Survivorship Bias

Another methodology for measuring the survivorship bias is to simulate the effects. This has the advantage that the other biases and data problems of the databases can be neglected. Brown, Goetzmann, Ibbotson and Ross (1992) applied this methodology for Mutual Funds, Fung and Hsieh (1997) proposed it for further research for Hedge Funds and finally Zimmermann (2000) simulated the effect for Hedge Funds. We followed the ideas of this study and simulated the process for Hedge Funds. We considered a starting volume of 10.000 Funds. The Net Asset Value of the Funds follows a geometric Brownian motion, with a drift and certain volatility. There is one Survivorship condition, namely if the NAV of the Fund exceeds a closing level (between 80% and 100%), it continues to exist. If the NAV falls below the closing level, the Fund is seen as “dead”. The survivorship bias is calculated as the difference in returns from the Portfolio of all funds and the surviving funds. The returns of the Hedge Funds follow the normal distribution. The Attrition rates of the simulation can be found in the appendix.

For the closing levels 80%, 90% and 100%, the expected returns 7%, 10% and 15% and the volatilities of 10%, 15% and 20%, the following survivorship bias was reached.

Abbildung in dieser Leseprobe nicht enthalten

Table7: Simulated Survivorship Bias

Source: own simulation, following Zimmermann (2000).

It is evident that the higher the closing-level, the volatility and the lower the expected return, the bigger the survivorship bias. As depicted in the following graph, the Attrition-Level is very high at the beginning of the Fund, because the NAV is very close to the Closing level (Simulation Specification: Closing Level 100%, Expected return 7%, Volatility 10%).

Abbildung in dieser Leseprobe nicht enthaltenAbbildung in dieser Leseprobe nicht enthalten

Figure4: Simulated Attrition

Source: own simulation, following Zimmermann (2000).

This very basic simulation has some shortcomings too. There is, for instance, only one starting date for Funds and therefore new Funds are not considered. Moreover, another aspect is the dependency on the first period. In practice, a Fund which performs poorly in the first period will never be included in the database.

4.4. Self Selection Bias I

The self selection bias I describes the situation before a Hedge Fund is included in the database. The Hedge Fund Management has the power to decide after some period, whether it would report the Fund to the database or not. The time lag between the launch of the fund and the reporting is called “incubation period”. (SeeTable 8. for Details)

The self selection bias I describes the situation where the Funds included in the database are not reflecting the Hedge Fund Universe.

The basis for the decision of the Hedge Fund Management whether to report or not can be illustrated using a decision tree. The first question is: “Has the fund performed outstanding since the launch?” If this question can be answered with “yes”, there may be a need of the Fund for additional capital. This holds for the most cases, because the amount needed for a Hedge Fund is relatively high. The decision in this case would be to report to the database. If this particular is specified as a “closed Fund” and does not accept capital, there may be the need for the Hedge Fund Company to promote its name and it would therefore report to the database. In all other cases a reporting does not make sense for the Hedge Fund, because a Fund which has a bad Track record is not likely to attract new investors and if no advertising is needed, the costly reporting could be avoided. There can be two effects of this bias. First, if a fund is just reporting while obtaining good results it is likely that the Performance of the Funds in the database is overstated. The second effect can be that already closed funds offer a higher performance and the Funds in the database are showing a minor performance. Fung and Hsieh offer some examples for this aspect. The Quantum Fund of George Soros, for example, was already closed in 1992 and never reported to database vendors. Similarly, LTCM never reported their data.[39] These contrarily effects limit the magnitude of the self selection bias I.

Abbildung in dieser Leseprobe nicht enthalten

Figure5: Self Selection Bias I – Decision Tree

Source: own graph

4.5. Self Selection Bias II

The self selection bias II is part of the survivorship bias, because if the decision of the Hedge Fund Management is to stop reporting data to the Hedge Fund, this is labelled as “dead” in the database. The decision tree of the management whether to report further after the Fund is listed can be illustrated as follows. Has the Fund gathered enough money in the last period, and therefore does not accept fresh money, it could decide to stop reporting. If the performance was poor in the last period and the Fund is in Liquidation, the management can decide to stop reporting. In all other cases, reporting to the database is more favourable for the Hedge Fund.

Abbildung in dieser Leseprobe nicht enthalten

Figure6: Self Selection Bias II – Decision Tree

Source: own graph

4.6. Selection Bias I

The selection bias I can be caused by the database-vendors. If for instance a Hedge Fund decides to report to the database, but the data will not be captured because of some design or compliance by the database-Provider, a bias can occur. These criteria can be a minimum size, some ratios, or a certain strategy of the fund. We do not rate this bias as very important, because the minimum size for the databases is not very high and there is a vast variety of different Hedge Funds Indices Strategies which should include all available strategies.

4.7. Selection Bias II

The selection bias II is also caused by the database-vendors. This bias occurs if an already reporting Hedge Fund is eliminated in the database because it no longer fulfils certain criteria. These criteria can be that the Hedge Fund is no longer accepting money and the database-vendors are asking for this speciality because they are creating an investable index, or that the Fund showed a lack of size, lack of performance etc. The selection bias II is included in the survivorship bias and it is reflected in the attrition rates.

4.7.1. Summary

The biases which are occurring in the first phase; self selection bias I and the selection bias I cannot be quantified, because there is no data available of Funds which are not reporting or rejected by the database-Provider. The effects of Funds not reporting because they do not need money and the Funds which stop reporting because of quitting operations are contrarily and therefore limit the effect of these biases.[40]

4.8. Backfilling Bias

When Hedge Funds are added to the databases, it is favourable for the Hedge Funds to include the complete history of the Fund. The historical returns are backfilled. This happens because it is much easier for Hedge Funds to advertise with good track records. When Hedge Funds are launched, there is a period where the Fund has an incubation period. In this period, Funds are investing money from managers, friends and relatives. If the performance in this incubation period is good, the Fund starts to promote itself trough a listing in the database, with the complete history of the Funds. When database-providers are putting the funds into their databases, they are filling back the history during the incubation period.Table 8: Incubation Timegives an overview of how long it takes for a Hedge Fund and a Fund of Hedge Fund to be captured within the TASS database. It can be seen that a very small number of Funds are reporting from the first day of issue and some funds even wait years until they are captured by the databases.

Abbildung in dieser Leseprobe nicht enthalten

Table8: Incubation Time

Source: Fung and Hsieh (2000), p.299

Fung and Hsieh (2000) found a median incubation period for Hedge Funds of 343 days in the TASS database. Therefore they deleted the first 12 month of the performance of the Hedge Funds and created an adjusted observable portfolio for their study. The bias was calculated with the same methodology as the survivorship bias. The bias is the difference in returns of two portfolios. One portfolio consists of all Hedge Funds and one is the adjusted portfolio, where for every Hedge Fund the first 12 months of the performance are eliminated. Finally, Fung and Hsieh come to the result of an annual backfilling bias of 1.4 % for the period 1994-1998. For the Fund of Funds, they deliver a very similar result of 1.3 % in return differences, although the incubation period is significantly higher in the median for Fund of Funds.[41] If this difference in the return is not known, it leads to a significant overstating of the annual return of Hedge Funds.

5. Conclusions

Hedge Funds are clearly an interesting alternative for investors to supplement their portfolio. Investors use Hedge Fund Indices to evaluate the Performance, on the one hand, and as a Benchmark in their Asset Allocation Decision, on the other hand. However the Decision which Index to consider as well as the construction of this Index is a quite demanding task. Prima facie it seems nearly impossible to gain accurate information on the history of the Industry, because of the missing oversight in this market and the construction problems of the Indices. The Investors have to be aware of the biases inherent to the Indices and take them into account for their Investment Decision. The biases discussed in this paper can be summarised as follows:

Self selection I bias and selection I bias are assessed as not having a strong implication on the performance of Hedge Funds. They are difficult to quantify and the contrarily effects in the biases can equal each other.

The self selection bias II, the selection bias II and the liquidation bias, which can be summarized under the survivorship bias clearly have an effect on the Performance of Hedge Funds, all empirical studies in this field delivering a significant difference in the return of the Surviving Portfolio and the Portfolio including the dead funds. The only exemption was the study of Ackermann, McEnally and Ravenscraft (1999), but their work has to be questioned because of the time horizon of the study.

The Simulation, which is undiluted from other biases and database shortcomings also showed the existence of the survivorship bias. The absolute number of the bias is strongly dependent on the level of volatility, the closing level and the expected return.

Investors should pay serious attention to this bias, as not accounting for this bias will lead to an overestimation of the Hedge Fund return and therefore to a significant overweighting in the Asset Allocation.

The same is true for the backfilling bias. There is empirical evidence for the existence of this bias and it can be considered at about 1.4 % per annum in overestimating the return of Hedge Funds.

In the last years, a vast number of Hedge Funds Indices appeared in the market, some are investable while some are not. The evaluation all the Indices for inherent biases may prove to be a never ending task. The existence of the biases in the databases was demonstrated by empirical research. Additionally, we proved the existence of the survivorship bias at an ex ante consideration, employing simulation methods. The challenge for Index Providers is to avoid such biases in the construction of an Index. A Hedge Fund Index must therefore take into account certain idiosyncrasies that characterize the Hedge Fund Industry.

The Investors should follow some principles in choosing a Hedge Fund Index, as benchmark or investment, in order to find out the indices who are avoiding the biases at the best.

- The composition of the Index should be transparent;
- It should be clear that Backfilling is avoided in the Index;
- The data should be verified by the Index Provider;
- There should be a clear policy for defunct Funds;
- The Index should be representative for the Hedge Fund Universe;
- The Index should be asset weighted.

Funds of Funds are typically not so vulnerable to biases. The Indices for Funds of Funds are therefore a good approximation for the Hedge Fund Universe.

Investors should also consider these biases in their investment decision. This can be done, for example, by adjusting the expected return for the class of Hedge Funds by the survivorship and the backfilling bias respectively.[42] To use the words of Fung and Hsieh (1997b) “Historical Performance of Hedge Funds must be adjusted for survivorship bias before portfolio managers can make intelligent portfolio allocation decision”.[43]

6. References

Ackermann, C., McEnally, R., Ravenscraft, D. (1999).. The Performance of Hedge Funds: Risk, Return and Incentives. Journal of Finance, 54(3), 833-874

Agarwal, V., Naik, N.Y. (2000). Multi-Period Performance Persistence Analysis Hedge Funds. Journal of Financial and Quantitative Analysis. 35(3), 327-341

Amin, G., S., Kat, H., M. (2003). Welcome to the Dark Side: Hedge Fund Attrition and Survivorship Bias over the period 1994-2001. The Journal of Alternative Investments, Summer 2003, 57-73

Amin, G., S., Kat, H., M. (2003b). Hedge Fund Performance 1990-2000: Do the “Money machines” really add value? Journal of Financial and Quantitative Analysis, 38(2), 251-274

Bookstaber, R. (2003). Hedge Fund Existential. Financial Analyst Journal, 59(5), 19-23

Brooks, C., Kat, H., M. (2001). The statistical properties of Hedge Fund Index returns and their implications for Investors. Working Paper, Version. 31.10.2001

Brown, S., Goetzmann, W., Ibbotson, R., Ross, S. (1992): Survivorship Bias in performance studies. Review of Financial Studies, 5 (4), 553-580

Brown, S., Goetzmann, W. (1995): Performance Persistence. Journal of Finance 50 (2), 679-698

Brown, S., Goetzmann, W., Ibbotson, R. (1999), Offshore Hedge Funds: Survival and Performance, 1989-1995. Journal of Business, 72, 91-118

Brown, S., Goetzmann, W., Park, J. (2001). Careers and Survival: Competition and Risk in the Hedge Funds and CTA Industry. Journal of Finance, 56 (5), 1869-1886

Cottier, P. (1997). Hedge funds and managed futures – Performance, risks, strategies, and use in investment portfolios. Doctoral Thesis, University of St.Gallen

Elton, E., Gruber, M., Blake, C. (1996): Survivorship Bias and mutual fund performance. Review of Financial Studies 9, 1097-1120

Fano-Leszczynski, U. (2002): Hedgefonds: Erfolgreich investieren, Risiko minimieren. Manz-Verlag, Wien

Fung, W., Hsieh, D. (1997): Empirical characteristics of dynamic trading strategies: the case of Hedge funds. Review of Financial studies. 10, 275-302

Fung, W., Hsieh, D. (1997b): Survivorship Bias and investment style in the returns of CTAs: Journal of Portfolio Management. 24, 30-41

Fung, W., Hsieh, D. (2000). Performance characteristics of hedge funds: Natural versus spurious biases. Journal of Financial and Quantitative Analysis, 35 (3), 291-307

Fung, W., Hsieh, D. (2002). Hedge-Fund Benchmarks. Information Content and Biases, Financial Analyst Journal, 22-34

Fung, W., Xu, E. X., Yau, J. (2004). Do Hedge-Fund Managers Display Skill? The Journal of Alternative Investment, 22-31

Hedge Fund Research (2004): HFR Q2 Industry Report .

Garcia C., Gould, F. (1993): Survivorship Bias. The Journal of Portfolio Management 19, 52-56

Grinblatt, M., Titman, S. (1989): Mutual Fund performance: An analysis of quaterly portfolio holdings. Journal of Business. 62, 393-416

Grünbichler, A., Graf, S., Gruber, A. (2001): Private Equity und Hedge Funds: Alternative Anlagekategorien im Überblick. Verlag NZZ, Zürich

Jaeger, R., A., (2000): All about Hedge Funds. Mc-Graw Hill, New York

Liang, B. (2000). Hedge Funds: The Living and the Dead. Journal of Financial and Quantitative Analysis, 35 (3), 309-326

Liang, B. (2001). Hedge Fund Performance: 1990-1999. Financial Analyst Journal, 11-18

Malkiel, B., (1995): Returns from investing in equity mutual funds 1971-1991. Journal of Finance 50.549-572

McFall, L., R. (2005). The Answers to your dreams? Investment Implications of positive asymmetry in CTA returns. The Journal of Alternative Investments, p. 22-32

Muhtaseb, M., R. (2003). Hedge Funds Asset Allocation and Investable Benchmarks. The Journal of Wealth Management. 64-67

Schneeweis, T., Kazemi, H., Martin, G. (2003). Understanding Hedge Fund Performance: Research Issues Revisited - Part II. The Journal of Alternative Investments, 5(4), 8-30

Zimmermann, H. (2000): “Survivorship” die verzerrte Wahrnehmung von Chancen und Risiken. Finanzmarkt und Portfolio Management, 14 (1), 1-6

7. Appendix

Overview of Hedge Fund Indices and useful websites

Abbildung in dieser Leseprobe nicht enthalten

Table9: Overview of Hedge Fund Indices

Source: (2005)

Abbildung in dieser Leseprobe nicht enthalten

Table10: Simulated Attrition

Source: own simulation based on Zimmermann (2000)


[1] Ackermann, McEnally and Ravenscraft (1999), p.833, Liang (2000), p.2, Jaeger (2000), p. 32, and Cottier (1997), p. 13

[2] Brown, Goetzmann and Ibbotson, (1999), p. 91

[3] See Fano-Lescczynski (2002) p. 143ff for details on the rise and fall of LTCM.

[4] Agarwal and Naik, (2000), p. 327

[5] The ‘relatively sophisticated investors` feature must not hold in these days, when funds of funds are offering the access to Hedge Funds also to small investors. These small investors often do not have the possibility to look through the Fund of Funds to identify the underlying products. This is done by the manager of the umbrella product.

[6] See Bookstaber, (2003), p. 19

[7] Brown, Goetzmann and Ibbotson, (1999), p. 93 and Ackermann, McEnally and Ravenscraft (1999), p.834 for the US. The same problem can be seen outside the US. There are currently discussions about the regulation of Hedge Funds, which will be also political issues in the next years. The limit for the numbers of investors was 100 until 1996 and is regulated by the Investment Company Act of 1940. These provisions prevent funds that are not confirming to the regulation of this act from a public offering; see Brown, Goetzmann, and Park, (2001), p. 1872

[8] The Estimation of HFR, was made in the second quarter of 2004. HFR (2004), p.6; MSCI (2005) p. 2 estimates the total amount invested in Hedge Funds between 850-950 Billion USD with nearly 8.000 funds at year end 2004.

[9] The high limits result of the limitation of a maximum of 499 US-Investors to a single Hedge Fund; see Brown, Goetzmann, and Park, (2001), p. 1872

[10] HFR, (2004), p. 14

[11] Brooks and Kat (2001, p. 3

[12] The first mutual fund which tracked an index and signed the advent of passive investment strategies was managed by Wells Fargo and was established in 1970.

[13] See, for example, Muhtaseb (2003), p. 64ff.

[14] In reality, there are several difficulties in replicating indices. For instance, there are problems like the minimum size of transactions, the transaction costs, the treatment of dividends etc.

[15] See, for example, Brown, Goetzmann and Ibbotson, (1999); Brown, Goetzmann, and Park, (1999); Ackermann, McEnally and Ravenscraft (1999). Liang (2000). Brooks and Kat (2001). CISDM was formerly known as MARHedge, but was transferred to the Center of International Securities and Derivatives Markets. Just recently MSCI and S&P entered the Hedge Fund Market with some indices. In the Appendix an exhaustive list of Hedge Fund Indices can be found.

[16] In May 2005, MARHedge stopped the data collection. The managing and distribution is currently done by AIRT (Alternative Investment Research & Technologies). The database itself is now owned by the University of Massachusetts, which received it as donation from Zurich Capital Markets.

[17] See Brooks and Kat (2001), p.4f

[18] See Schneeweiss, Kazemi, and Martin (2003), p. 26

[19] See for example, Ackermann, McEnally and Ravenscraft (1999), p.839 and Cottier (1997), p. 154

[20] See Cottier (1997), p.153f and Grünbichler, Graf and Gruber (2001), p. 256.

[21] See Fung, Xu and Yau (2004), p. 23.

[22] For example Ackermann, McEnally and Ravenscraft (1999), p.838 and Liang (2000), p. 2

[23] Amin and Kat (2003), p. 57

[24] See, for example, Amin and Kat (2003), p. 57, and Liang (2000)

[25] See Brown, Goetzmann and Park (2001), p. 1875

[26] see Amin and Kat (2003), p. 57

[27] See Liang (2000), p. 25; Liang (2001) calculated an attrition rate for the years 1993-1999 of 8.54%. Brown, Goetzmann and Park (2001) calculated an attrition rate of less than 15% per year from 1994 to 1998. p. 1875

[28] In their earlier research on Mutual Funds, Brown and Goetzmann (1995) found out that the age and the size of the Fund are the main drivers for attrition.

[29] In the following we are referring to databases as a source for the composition of Hedge Fund Indices. There is a wide variety of Hedge Fund Indices, which obtain the input data from some data-providers. Therefore this basis should be tested on Biases.

[30] See Fung and Hsieh (2002), p. 23

[31] Garcia and Gould (1993), p. 53

[32] See Makiel (1995), p. 551 for Mutual Funds.

[33] See Fung and Hsieh (2000), p.295

[34] See Amin and Kat (2003); p.65

[35] See Liang (2000), p.319

[36] See Liang (2000), p.317

[37] See Fund and Hsieh (2002), p.24

[38] See Fund and Hsieh (2002), p.24

[39] See Fung and Hsieh (2000), p.299. The Quantum Fund never reported data, but the databases obtained the data from public sources, as noted by Fung and Hsieh.

[40] See Fung and Hsieh (2002), p. 24f.

[41] See Fung and Hsieh (2000b), p. 298

[42] See McFall (2005) for an adjustement in CTAs returns.

[43] Fung, and Hsieh (1997b), p. 31

33 of 33 pages


Biases in Hedge Funds Indices
University of St. Gallen
Doktorandenseminar; Corporate Finance
5,5 (1,5 in GER)
Catalog Number
ISBN (eBook)
ISBN (Book)
File size
592 KB
Biases, Hedge, Funds, Indices, Doktorandenseminar, Corporate, Finance
Quote paper
Vinzenz Benedikt (Author), 2005, Biases in Hedge Funds Indices, Munich, GRIN Verlag,


  • No comments yet.
Read the ebook
Title: Biases in Hedge Funds Indices

Upload papers

Your term paper / thesis:

- Publication as eBook and book
- High royalties for the sales
- Completely free - with ISBN
- It only takes five minutes
- Every paper finds readers

Publish now - it's free