INVESTING IN HEDGE FUNDS: RISK, RETURN AND PITFALLS

Francis Koh, Ph.D. 1

David Lee, Ph.D. 2

Phoon Kok Fai, Ph.D. 3

Abstract

Hedge funds are collective investment vehicles fast becoming popular with high net worth

individuals as well as institutional investors. These are funds that are often established

with a special legal status that allows their investment managers a free hand to use

derivatives, short sell, and exploit leverage to raise returns and cushion risk. Given that

that they have substantial latitude to invest, it is instructive to examine the performance of

hedge funds compared to other forms of managed funds. This paper provides an overview

of hedge funds and discusses their empirical risk and return profiles. It also poses some

concerns regarding the empirical measurements. Given the complexity of hedge fund

investments, meaningful analytical methods are required to provide greater risk

transparency and performance reporting. Hedge fund performance is also beset by a

number of practical issues generating “practical risks”. These risks are not fully addressed

by the usual risk-adjusted performance measures in the literature. A penalty function to

discount these extraneous risk dimensions is proposed. The paper concludes that further

empirical work is required to provide informative statistics about the risk and return of

hedge funds.

_______________________
1
Practice Associate Professor, School of Business, Singapore Management University ([email protected])
2
Managing Director and Chief Investment Officer, Ferrell Asset Management ([email protected])
3
Consultant, FDP Consultants ([email protected])

1
 INVESTING IN HEDGE FUNDS: RISK, RETURN AND PITFALLS

1. Background

Since the early 1990s, there has been a growing interest in the use of hedge funds

amongst both institutional and high net worth individuals. Due to their private nature, it is

difficult to obtain adequate information about the operations of individual hedge funds and

reliable summary statistics about the industry as a whole.

Hedge funds are known to be growing in size and diversity. As at the end of 1997, the

MAR/Hedge database recorded more than 700 hedge fund managing assets of US$ 90

billion. See Table 1. This is only a partial picture of the industry, as many funds are not

listed with MAR/Hedge. In practical terms, it is not easy to estimate the current size of the

hedge fund industry unless all funds are regulated or obligated to register their operations

with a common authority. Brooks and Kat (2001) estimated that, as at April 2001, there are

around 6000 hedge funds with an estimated US $400 billion in capital under management

and US $1 trillion in total assets.

While hedge funds are well established in the US and Europe, they have only begun to

grow aggressively in Asia. In the week ending April 30, 2002 alone, 8 new funds were

introduced or launched. These were fund initiated by familiar names like Merrill Lynch,

Lazard, Morley, Societe Generale, HSBC and Crosby. In year 2000, there were some 30

funds that were established. They attracted about US$ 600 million in capital. Another 20

hedge funds were set up in 2001. Many of these were start-ups and founded by talents

who were previously employed by large institutional fund management companies. In

1999, HedgeFund Intelligence launched a new publication, AsiaHedge. Recently,

AsiaHedge has also been set up 4 Asia-Pacific Hedge Fund Indices, covering funds in four

2
geographic areas: Japan, Asia ex-Japan, Asia inclusive of Japan and Australia.

AsiaHedge currently has the reported data of about 150 hedge funds operating in Asia and

Australia and/or investing in the Asia-Pacific markets. These hedge funds, like their

counterparts in the U.S., use long/short, global/macro, and event driven strategies.

Three interesting features differentiate hedge funds from other forms of managed funds.

Most hedge funds are small and organized around a few experienced investment

professionals. In fact, more than half of U.S Hedge Funds manage amounts of less than

US$25 million. Further, most hedge funds are leveraged. It is estimated that 70 per cent of

hedge funds use leverage and about 18% borrowed more than one dollar for every dollar

of capital1. Another peculiar feature is the short life span of hedge funds. Hedge funds

have an average life span of about 3.5 years.2 Very few have a track record of more than

10 years. These features lead many to view hedge funds, as “risky” and “opportunistic”.

2. What are hedge Funds?

“Hedge Funds” is a term coined by journalist Carol Loomis to describe an innovative

investment structure first created by Alfred Winslow Jones. Jones had established a fund

with unique features: (a) He set up “hedges” by investing in securities that he determined

as undervalued and funding these positions partly by taking short positions in overvalued

securities, creating a “market neutral” position; (b) He also designed an incentive fee

compensation arrangement in which he was paid a percentage of the profits realized from

his clients’ assets; and (c) He invested his own investment capital in the fund, ensuring

that his incentives and those of his investors were aligned and forming an investment

“partnership”.

3
Currently, most hedge funds retain these features. They are set up as limited partnerships

with a lucrative incentive-fee structure. Hedge fund managers also often have a significant

portion of their own capital invested in the partnerships.

However, the term “hedge fund” has also been generalized to describe investment

strategies that range from the original “market-neutral” style of Jones to many other

strategies and opportunistic situations, including global/macro investing, such as the

Quantum Fund of George Soros. Soros is famous for his “attack” on Sterling in September

1992, when he was reported to have a US$10 billion short position. He made $1 billion

when the British Pound subsequently devalued. Soros shorted the Pound but was long the

Yen.

Due to the diversity of the industry, there is no standard method to classify hedge funds

neatly. In the industry, there are at least 8 major databases set up by data vendors and

fund advisors. We follow the classification used by Eichengreen and Mathieson (1998),

who relied on the MAR/Hedge database. Under this classification, there are 8 categories

of hedge funds with 7 differentiated styles and a fund-of-funds category:

(a) Event driven funds. These are funds that take positions on corporate events,

taking an arbitraged position when companies are undergoing re-structuring or

mergers. For example, hedge funds would purchase bank debt or high yield

corporate bonds of companies undergoing re-organization (often referred to as

'distressed securities'). Another event-driven strategy is merger arbitrage. These

funds seize the opportunity invest just after a takeover has been announced. They

1
See Eichengreen and Mathieson, pp. 7
2 Stefano Lavinio, pp 128.

4
 purchase the shares of the target companies and short the shares of the acquiring

companies. Occasionally, they carried out the reverse if the deal would likely fail.

(b) Global funds is a catchall category of funds that invest in non-US stocks and

bonds with no specific strategy reference. It has the largest number of hedge

funds. It includes funds that specialize on the emerging markets

(c) Global/Macro funds refer to funds that rely on macroeconomic analysis to take

bets on major risk factors, such as currencies, interest rates, stock indices and

commodities.

(d) Market neutral funds refer to funds that bet on relative price movements utilizing

strategies such as long-short equity, stock index arbitrage, convertible bond

arbitrage and fixed income arbitrage. Long-short equity funds use the strategy of

Jones by taking long positions in selective stocks and going short on other stocks

to limit their exposure to the stock market. Stock index arbitrage funds trade on the

spread between index futures contracts and the underlying basket of equities.

Convertible bond arbitrage funds typically capitalize on the embedded option in

these bonds by purchasing them and shorting the equities. Fixed income arbitrage

bet on the convergence of prices of bonds from the same issuer but with different

maturities over time. This is the second largest grouping of hedge funds after the

Global category.

(e) Sector funds concentrate on selective sectors of the economy. For example, they

may focus on technology stocks if these are over-priced and rotate across to other

sectors.

5
(f) Short-sellers focus on engineering short positions in stocks with or without

matching long positions. They play on markets that have risen too fast and on

mean reversion strategies.

(g) Long-only funds take long equity positions typically with leverage. Emerging

market funds that do not have short-selling opportunities also fall under this

category.

(h) Fund of funds refer to funds that invests in a pool of hedge funds. They specialize

in identifying fund managers with good performance and rely on their good industry

relationships to gain entry into hedge funds with good track records.

Table 1 presents statistics about the various categories of hedge funds and past

performance. The sectoral hedge funds provided the best mean return over the period

studied, while the “market-neutral” funds had the lowest standard deviation of returns. On

a risk-adjusted basis (dividing the mean return by the standard deviation), the category of

fund that ranks highest is the market neutral funds followed by event-driven funds. But,

before this conclusion is valid, more discussion follows on the empirical problems using

the data obtained from incomplete databases.

___________________

See Table 1

___________________

6
3. Why invest in hedge funds?

Traditional asset allocation optimizes the use of equities, bonds, real estate and private

equity to invest in a portfolio that maximizes returns and minimizes the portfolio risk. With

this objective, hedge funds become a natural candidate for consideration. Firstly, it is

commonly believed that hedge funds may have superior returns. There are many

anecdotal stories about the stunning success of hedge fund managers and their skills.

Soros was reported to have obtained returns in excess of 30% p.a for a good number of

years. From Table 1, there is also apparent evidence that hedge funds, as a group, have

returns that are impressive. For example, over the period 1990-1997, all the hedge funds

had positive absolute returns. Global/Macro funds obtained mean returns of 28.1% p.a.

with a standard deviation that is comparable to equity funds.

Secondly, hedge funds have returns that are generally believed to be uncorrelated to the

traditional asset classes and may even have a lower risk profile. For example, Morgan

Stanley Dean Witter (November 2000, p 1) reported that hedge funds “exhibit a low

correlation with traditional asset classes, suggesting that hedge funds should play an

important role in strategic asset allocation”. Table 2 shows a common presentation of the

underlying relationships between hedge funds and the other assets.

___________________

See Table 2

___________________

7
Thirdly in a bear market, many investment mangers find it uninteresting to merely beat the

market index, which may have negative returns. They would have preferred to go short or

avoid long positions to have positive returns. Investing in appropriately chosen hedge

funds may provide the possibility of obtaining positive “absolute returns”.

The foregoing provides persuasive reasons to consider hedge funds as “alternative

investments”. However, relying on statistics culled from public databases is fraught with

data biases. An uninformed investor may be misled into common misperceptions about

the return and risk of hedge funds. There is now recent and definitive work by a number of

authors that have provided fascinating revelations about the risk and return profiles of

hedge funds which are discussed below.

4. Commercial databases and Statistical Inferences

Data Collation Issues

Organized as private limited partnerships, and frequently as offshore investment vehicles,

hedge funds generally do not disclose their activities to the public. This has resulted in

frequent complaints about the lack of transparency. Fortunately, many funds do release

selective information to publicize themselves and their performance to attract new

investors. These data are collected by a small number of data vendors and fund advisors.

A few large advisors and vendors are currently publishing performance data and

indices/sub-indices periodically corresponding to the various investment strategies. A

listing of Hedge Funds Databases and some descriptive details is provided in the

Appendix.

However, voluntary participation in performance reporting leads to incompleteness of

information regarding the hedge fund population as a whole. Thus, sampling biases are

8
present whenever an investor analyses a hedge fund database on a stand-alone basis.

Some of theses biases are briefly discussed below.

Survivorship Bias

Databases obviously only include hedge funds that submit information. Funds that perform

poorly often choose not to submit their performance. Thus, poorly performing funds are

likely to be missing in a database. A “survivorship bias” arises when a database includes

only the performance of funds that are alive and present at the end of the sample period. A

subset of survivorship bias, called liquidation bias, occurs when disappearing funds may

not report final periods leading up to and including their liquidation. If funds cease

operation due to poor performance, the historical returns of surviving funds in the

database is biased upward with risk biased downward relative to the population of hedge

funds.

Hedge funds may exit a database for other reasons than poor performance. Database

vendors often delist funds that do not provide reliable information. Some popular funds

also stop reporting their performance when they have reached a desired size, and do not

need to further solicit “new” money. Omissions of these funds would also severely bias a

database.

Brooks and Kat (2001) stated that around 30% of newly established funds do not survive

the first three years, primarily due to poor performance. Thus, not including defunct funds

is likely to lead to over-estimation of the returns and profile of hedge fund industry. Fung

and Hsieh (2001a) found that estimates of survivorship biases differed across two

commonly used databases, HFR and TASS. The survivorship bias (and attrition rate) was

9
much higher in TASS than that in HFR. They estimated that survivorship bias would over-

report hedge fund mean returns by about 1.5% to 3% per annum.

Selection Bias

Database vendors impose their own criteria before a hedge fund may enter their database.

The criteria would include the type of fund involved, track record and assets under

management. Databases may also exclude types of hedge funds whose trading activities

or instruments do not meet their criteria. Again, the result is a likely upward bias in the

database, which has become a biased sample belonging to the larger population.3

Data collation and statistical biases present problems when generalizations have to be

made about the returns and risk across the different categories of hedge funds. These

biases also affect the computation of hedge fund indices. Since this is so, statistical

inferences about the performance of hedge fund returns and the returns on hedge fund

indices may not be reliable.

Brooks and Kat (2001) provided evidence to support this view. They showed that different

databases have different sample statistics for similar categories of funds. Table 3 shows

that while the mean return for macro hedge funds computed by the various databases

ranges from 10.2% to 17.2%. Yet, this is a statistic for a common class of hedge funds

over the same time-period. More interestingly, the standard deviation ranges from 19.3%

to 50.2%. This is compelling evidence for the investor to be wary about obtaining statistics

from hedge fund databases and making statistical inferences.

3
Park (1995) analyze a subset of selection bias termed “instant history bias”. This bias arises because when a
new fund is first included, database managers often “back-fill” its performance history. Up to a year or more
of data may be added to the database. Again, another sampling bias is added onto the database.

10
 ____________

See Table 3

_____________

5. Nature of hedge funds, trading strategies and performance measurements

Mean, variance, skewness and kurtosis

It is clear that because of the method of collection and reporting of the hedge fund

databases, there are biases in the data collected. Some of the returns can be viewed as

the upper bound and the averages are likely to be smaller than actually reported. The wide

range in returns and dispersion indicates that mean and variance may not capture the full

picture regarding the activities of hedge funds. Indeed, the organization structure of hedge

funds, their investment objectives, trading strategies and managerial compensation

differentiate them significantly from the usual mutual fund. Most mutual funds are generally

engaged in “buy-and-hold activities” – acquiring and managing stocks and bonds over a

longer period of time. Although some mutual funds would engage in activities like leverage

or short-sell, most do not.

There is now increasing evidence that hedge fund returns and hedge fund indices returns

are not normally distributed. And, it is the strategies of hedge fund investments that have

directly contributed to this situation. Typically, hedge fund investments are based on

absolute return strategies. They are expected to deliver performance regardless of market

conditions. To do so, hedge fund managers use two main approaches to achieve absolute

return targets: (a) directional (or market timing) and (b) non-directional approaches.

11
The directional approach dynamically bets on the expected directions of the markets.

Funds will invest long or sell-short securities to capture gains from their advance and

decline. In contrast, the non-directional approach attempts to extract value from a set of

embedded arbitrage opportunities within and across securities. The non-directional

approach typically exploits structural anomalies in the financial market.

Mean-variance analysis is appropriate when returns are normally distributed or investors’

preferences are quadratic. The reliability of mean-variance analysis therefore depends on

the degree of non-normality of the returns data and the nature of the (non-quadratic) utility

function. While the utility function may not be a serious problem, the non-normal

distribution of returns presents an issue.

According to Fung and Hsieh (1999a), “... when returns are not normally distributed (as it

is the case for hedge funds), the first two moments (i.e. mean and standard deviation) are

not sufficient to give an accurate probability.” Fung and Hsieh found that hedge fund

returns are leptokurtic or fat-tailed. One likely explanation is that net returns include

spreads that are distributed with fat tails.

Brooks and Kat (2001) found that hedge fund index returns are also not normally

distributed. Many hedge fund indices exhibit relatively low skewness and high kurtosis,

especially in the case of funds investing in convertible arbitrage, risk arbitrage and

distressed securities. These are non-normal profiles. Brooks and Kat argued that, while

hedge funds may offer relatively high means and low variances, such funds give investors

third and fourth moment attributes that are exactly the opposite to those that are desirable.

Investors obtained a better mean and a lower variance in return for more negative

skewness and higher kurtosis. There is no free lunch. These issues complicate a clear

12
conclusion on the return and risk of hedge funds as an asset class for investment by even

the most experienced investors.

Generally, the dynamic trading strategies of hedge funds render traditional mean-variance

measures relatively meaningless. While some hedge funds many have low standard

deviations, this does not mean they are relatively “riskless”. In fact, they harbor skewness

and kurtosis, which may be overwhelmingly “risky”.

Correlations of Returns

Fung and Hsieh (1997) examined the returns of hedge funds and commodity trading

advisers. They confirmed that hedge fund managers and commodity trading advisers

generate returns that have low correlations to the returns of mutual funds and standard

asset classes. This is the benefit often cited by portfolio managers in their choice of hedge

funds as an “alternative investment”. Having an additional asset with a low or negative

correlation permits the diversification of risk in a means-variance environment.

However, there are complications that arise in the case of hedge funds where correlation-

based diversification may not be valid. Lavino (2000, p177) argued that many hedge funds

are not consistently and continuously negatively or poorly correlated with other asset

classes over time. Hedge funds also may not have meaningful standard deviations. In

fact, many hedge funds have distributions with fat-tails, that is, exceptional events are

more frequent than would have been predicted based on normality assumptions. This

negates the use of correlation as a gauge to execute portfolio diversification.

Fung and Hsieh (2001) stated that “… Risk management in the presence of dynamic

trading strategies is also more complex.” Hedge fund managers have a great deal of

13
freedom to generate returns that are uncorrelated with those of other asset classes. But,

this freedom comes at a price. Dynamic trading strategies predispose hedge funds to

extreme or tail events. Thus, correlations may come at a cost. They cautioned that

“periodically the portfolio can become overly concentrated in a small number of markets”

and market exposures converge. This would lead to an “implosion” due to diversification.

Lo (2001) reinforced this view. He explained that many investors participate in hedge

funds to diversify their returns, as hedge fund returns seem uncorrelated with market

indexes such as S&P 500. However, uncorrelated events can become synchronized in a

crisis, with correlation changing from 0 to 1 overnight. These situations are examples of

“phase-locking” behavior encountered in physical and natural science.

We conclude that using means and standard deviations to report the returns and risks of

hedge funds is not adequate. Providing skewness and kurtosis statistics would be helpful.

Relying on simple correlation measures to diversify portfolio risks is not appropriate when

deciding to add hedge funds to a portfolio of other assets.

6. Some suggested measures to measure risk and return

Sortino and Price (1994) have proposed evaluating downside risks rather than total risks.

They defined a new measure and termed it the Sortino Ratio. This is similar to the Sharpe

Ratio, except that it uses 'downside deviation' instead of using standard deviation as the

denominator.

The Sortino Ratio was developed to differentiate between deviations on the upside and on

14
the downside and is more consistent with the investors' concern over risk of losses in their

investments. The Sortino ratio also allows for the setting of a user-defined return

benchmark where the numerator is the difference between the return on the portfolio and

the Minimum Acceptable Return (MAR). The MAR is usually the risk free rate, zero or

user-defined (for example, 5%)

We have earlier highlighted that the high skewness of a hedge fund's returns may be

connected to the hedge fund manager's selection of high-reward and low variance

opportunities. Lavinio (1999) has defined another measure to capture this, as follows:

d-Ratio = Abs (d/U)

where,

d = number of returns less than zero times their value

U = number of returns greater than zero times their value

Abs = absolute value.

The d-Ratio compares the value and frequency of a manager's winners to losers to

capture the skewness in returns. This statistic, which does not require any assumption of

the underlying distribution, may be used as a proxy for a fund's risk, with d=0 representing

a distribution with no downside, and d = infinity representing one in which the manager

does not make any positive returns.

In analyzing the performance of hedge fund managers, we also need to gain insights into

the permanence of a manager's skill. One way to examine if good performance is merely

transitory is to see if it is mean-reverting (i.e. whether the performance will reverse and

15
converge toward some predictable long-term value). We can capture this with the Hurst

Ratio4, which is defined as follows:

Hurst Ratio = log M / (log N - log a)

where

M(t) = (max(t) - min(t))/S(t)

N = length of shorter sub-periods into which a

manager's return record has been sub-divided

t = number of sub-periods into which a manager's

return record has been sub-divided

S(t) = standard deviation of data over sub-period t

a = constant term that is negligible if track record

is five years or less.

A Hurst Ratio between 0 and 0.5 means that a manager's return will tend to fluctuate

randomly, but converge to a stable value over time. With a Hurst Ratio around 0.5, a

hedge fund manager's track performance will be regarded as totally random, i.e. returns in

one period are not affected by returns in another period. Such hedge funds are deemed to

be “risky” as any stellar short-term gains may be accompanied by substantial losses in

another time period.

Hurst Ratios, which are between 0.5 and 1, describe returns that are persistent. These

fund managers have “hot” hands. We should, however, interpret such findings with care,

as there is a need to examine whether the same manager can maintain his fund’s Hurst

4
Lo (1991) applied the Hurst Ratio to stock returns and found that short-range dependence adequately
captured the time series behavior of stock returns.

16
Ratio in future time-periods that are beyond the chosen sampling periods. More rigorous

testing is required with out-of-sample data to provide meaningful conclusions.

Though the Sortino, d and Hurst Ratios would provide additional insights to the

performance and risk of hedge fund investing, further work is needed before these

analytical methods can be used to report on the risk and return performance of hedge

funds. In the next section, we examine some practical issues that complicate hedge fund

performance.

7. Practical Issues

We have seen that data issues may unwittingly lead to meaningless comparisons of hedge

fund performance. However, even if one possesses a set of clean and reliable data, it is

unlikely that there will be a statistically computed measure of risk-adjusted return, which

would satisfy a sophisticated investor. Hedge funds performance measures are beset by

many practical business issues, which make it extremely difficult to have a simple

measure to fully convey risk and return.

Specifically, hedge funds face many practical issues that increase their “riskiness.” For

ease of exposition, we have identified at least 6 types of practical issues that confound risk

and return measurements: style purity, consistency, fund size, use of leverage, liquidity

and asset concentration. We note that some of these problems are closely linked to one

another and create extraneous risks, which may not be correctly priced by the usual risk-

adjusted return measures.

17
Firstly, many hedge funds are assumed to have a pure and consistent style. This is rarely

the case. Many funds may be opportunistic and operate with more than one style. Thus,

many hedge funds do not always function exactly as their self-reported classifications

indicate. From the outside looking in, it is almost impossible to classify hedge funds neatly.

A hedge fund’s style purity over time is definitely less consistent when compared to Unit

Trusts (and mutual funds), which by nature are “buy-and-hold” accounts. Fung and Hsieh

(2001b) and others have suggested using factor analysis to discern the underlying

dimensions or “factors” that drive the returns for funds. This may, then, go below the

surface to determine unique hedge fund strategies that differentiate one fund from

another. Hopefully, this would enable an investor to detect style purity, style consistency

and most importantly, style deviations.

Till (2001a and 2001b) suggested that a number of hedge fund strategies might appear to

“earn their returns due to assuming risk positions in a risk-averse financial world, rather

than from inefficiencies in the market place.” In this sense, returns are made from a “risk

transfer”, and not due to managerial abilities per se. If indeed this is the case, then the

skill of selecting the appropriate hedge fund styles and the type of managers who can

execute the styles consistently, and how to allocate funds across these managers become

important to achieve superior returns. Viewed from this standpoint, style purity and

consistency are important attributes to measure exposure to hedge fund risks rather than

statistical measures like variance and skewness.

A hedge fund’s asset under management ("AUM") growth may be (a) internally generated

through performance, (b) externally induced because of inflows, or (c) magnified through

use of higher leverage. Hedge fund size is a dimension that has significant implications for

18
risk and return. A hedge fund’s risks increases proportionately with its AUM. This is

because the use of specialized strategies naturally limits a hedge fund to some “optimal

size” beyond which it becomes increasingly difficult to keep the same strategy or have the

opportunities for execution (often with leverage). We observe that hedge fund managers

are inclined to close their funds for further investments as soon as a target size is reached.

This is evidence that many managers understand the trade-offs between size and

performance. Yet, many often neglect to focus on the relationship between size and risks.

Hedge fund managers are drawn to the use of leverage to magnify potential returns from

small arbitrage opportunities. They are also inclined to concentrate their investable funds

in a small subset of potentially “rich” opportunities. Weisman and Abernathy (2000)

demonstrated the importance of guarding against excessive leverage, which is

compounded by a lack of liquidity when a disastrous event strikes. He pointed out that if

one were to construct a non-diversified, illiquid and/or leveraged portfolio and let it grow

over time, it would eventually lead to bankruptcy of the fund, if a misfortune strikes. The

potential risk is very high employing these strategies. The perceived risk may be low, as

a well-constructed downside-oriented measure using past data may not reveal the

potential risks from the occurrence of a future disastrous event. This is because a

misfortune has not yet struck. But the potential risks, which are usually unforeseen, are

large and threaten the eventual survival of the fund.

19
8. Accounting for Various Sources of Risk

Assume we have two hedge funds with similar statistical attributes: the same average

holding period returns adjusted by its standard deviation. We want to know which fund has

a better "risk-adjusted" return. Let us further assume that the first fund (compared to the

second) is less leveraged, invests in more liquid assets, is less concentrated/ more

diversified, and more disciplined in its application of investment styles. We are, most likely,

very inclined to prefer the first fund to the second. That is because the second fund,

although it has the same average return adjusted by its standard deviation, has taken

extraneous risk to achieve the same results. This is especially more obvious if analyzed in

the context of possible disastrous events. Thus, depending on the strategy employed, it is

generally correct to say that a non-leveraged, more liquid, more diversified and more

disciplined fund has a better chance of survival in the long term.

Perhaps, the crucial question has now become more obvious: how to modify “risk-adjusted

returns” to account for the many other forms of risks not captured statistically. Generally,

“risk-adjusted return" is defined as:

(Observed Returns – Benchmark Returns)

Indicated Risk Measure

This measure assumes that all the named variables are observable, measurable and

reliable. The benchmark return may be a stock index, a contrived peer measure or the 90-

day Treasury Bill rate of interest. The risk measure may be the “tracking error“, “standard

deviation”, or some other measure.

20
From the foregoing, we are sanguine that this risk-adjusted measure will be able to tell the

whole story. We propose, instead, a new metric to account for the numerous risks faced

by a hedge fund investor. We define:

Risk Adjusted Return = (Observed Returns – Benchmark Returns) X Penalty Function

Indicated Risk Measure

Without delving into the statistical properties in this paper, we postulate that the Penalty

Function is a discount factor that takes into account various dimensions such as hedge

fund style (purity and consistency), size, leverage, liquidity and asset concentration. These

dimensions penalize the statistically measured risk-adjusted returns of hedge funds.

Table 4 itemizes the risk dimensions and suggests avenues to discount them in the

penalty function.

____________

See Table 4

_____________

It should be noted that the leverage, liquidity and concentration measures require

additional data supplied by hedge fund managers. This calls for more disclosure and

transparency from the hedge fund managers.

While no single performance measure can be complete, we argue that a properly

constructed "risk-adjusted return with penalty” that has accounted for practical business

risks is more meaningful to an investor. A return that is merely adjusted by standard

deviations cannot alert an investor to such risks as leverage or liquidity, which had been

21
undertaken (to achieve the returns). Using a penalty function would provide a handle to

scale the observed return for the many practical risks that had been assumed by the

hedge fund manager. Even identifying the components that will constitute the penalty

function would be a worthwhile exercise to avoid the pitfalls of investing in hedge funds.

9. Conclusions

The paper presented an overview of hedge funds, describing their development and

characteristics. It also surveys some of the pitfalls that investors face when they try to

make investment decisions using hedge fund data from commercial sources. Given the

dynamic trading strategies and the complexity of hedge fund investments, commonly used

statistics such as mean, standard deviation and correlations are not meaningful. These

statistics must be used with extreme caution as the underlying distribution of hedge fund

returns (and also the returns of hedge fund indices) is not normally distributed.

This paper has suggested 3 other metrics that may be useful: Sortino, d, and the Hurst

Ratios. However, more empirical work is needed before they are used. A future paper will

provide some empirical results relating to these measures.

Without specifying the mathematical form, we venture further to account for various other

sources of risk such as style purity, style consistency, size, leverage, liquidity, and asset

concentration. We also suggested a "penalty function" for the risks from these sources.

The statistical properties of this penalty as well as illustrations using real data are left for

future work.

22
We conclude by noting that many authors have pointed to the limited use of statistical

measures and have suggested option-based analytical approaches to evaluate hedge

fund performance. In particular, the works by Hsieh and Fung (1997, 1999b) and Agarwal

and Naik (1999) have discussed these avenues to provide insights into these complex but

crucial issues in hedge funds investing. In this paper, we have also suggested using more

dimensions to alert investors to the numerous sources of unseen risks when they invest in

hedge funds. This is a promising direction for more research.

Table 1
MAR Hedge Fund Categories: December 1997
Mean and Standard Deviation of Returns (1990-1997)

1990-1997
Category Number Assets Mean Standard Risk-
(US$ Return Deviation adjusted
billion) (%) (%) Returns

Event-driven 120 8.6 18.9 5.9 3.2

Global 334 30.9 17.7 9.4 1.9

Global/macro 61 29.8 28.1 16.3 1.7

Market neutral 201 18.0 8.6 2.1 4.1

Sectors 40 1.8 29.6 15.9 1.9

Short-sellers 12 0.5 7.0 15.2 0.5

Long-only 15 0.4 27.3 15.4 1.8

Source: Eichengreen et al. (1998, pp 37)

Notes:
(a) The mean returns are annually compounded returns over the period 1990 to 1997, except for the Long-
only Funds, which were computed from 1994 tom1997.
(b) The annualized standard deviations were computed from of the standard deviation of monthly returns for
each investment style.
(c) Risk-adjusted returns are obtained by dividing the mean return by the standard deviation.

23
 Table 2
Performance Measures For Hedge Fund Indices
(Jan 1990 - April 2000)

Annualized Annualized Correlation with Correlation with

Return Std Dev S&P 500 Lehman Bro
(%) (%) Gov/Corp
EACM 1001 15.2 4.4 0.37 0.19
Eq Mkt Neutral 9.1 3.2 -0.11 0.15
Eq Hedged 20.6 10.3 0.20 0.00
Event 13.7 5.4 0.48 0.09
Global/Intl 20.8 11.5 0.61 0.15

Source: Lehman Brothers (2000)

Notes:
1. The EACM 100 is an index of hedge funds representing a wide range of strategies

24
 Table 3
Hedge Fund Indices from Different Databases
Mean and Standard Deviation of Returns (Jan 1995-Apr 2001)

Category/Database Mean (%) Standard

Deviation (%)
Risk Arbitrage1 14.13
Zurich2 13.2 12.8
Hennesse 13.0 11.8
Tuna 14.9 12.4
Altvest 15.6 13.4
HFR 13.6 12.7
Macro 13.3
Zurich 10.2 19.3
Hennesse 10.4 30.6
HFR 13.2 28.1
CSFB/Tremont 17.2 50.2
Tuna 15.6 33.8
Altvest 17.0 32.6
Van 9.4 41.8
Equity Market Neutral 12.8
Zurich 11.9 6.5
Hennesse 8.5 10.4
HFR 10.9 13.3
CSFB/Tremont 13.7 10.8
Tuna 15.2 19.2
HFR 16.8 17.2
Market Indices
S&P 500 18.6 54.4
DJIA 18.1 54.7
Russell 2000 13.7 69.1
NASDAQ 21.6 106.9
Lehman Government Bond 7.4 10.3

Source: Brooks & Kat (2001)

Notes:
1. The major databases are explained in the Appendix
2. Zurich Capital Markets computes the indices using the MAR/Hedge database that it acquired in
March 2001.
3. Simple Average of returns estimated using the different databases.

25
 Table 4
Discount to Risk-Adjusted Returns to Account for Various Types of Practical Risk

Sources Of To Penalise for Suggested Measurement Predicted Discount

Risks method to Returns

Style Purity Deviation from Self- Deviation from Style The higher the style
reported Investment Benchmark “impurity” the higher
Style the discount

Style Style Inconsistency Deviation from Factors Models The higher the style
Consistency “inconsistency” the
higher the discount

Asset Growth Unexpected Change in Fund Size The higher the

increases in Fund increase in fund size
Size (and Assets in the period under
Under Management) review, the higher the
discount

Leverage Excessive Leverage (a) Average gross exposure The higher the use of
leverage the higher
(b) Active Use of Leverage the discount.

(Computed from a comparison

of returns with and without the
use of leverage following the
standards recommended by
the Association for Investment
Management and Research)

Liquidity Low Asset Liquidity (a) Average Day to Complete The higher the threat
Sales of “illiquidity” the
higher the discount
(b) Ratio of Position to Trading
Volume

Asset (a) Single Security (a) Average Percentage of 10 The higher the asset
concentration Exposure Largest Holding over reporting concentration the
period higher the discount

(b) Erratic Returns (b) Fractal Dimension or

Inverse of Hurst Ratio

26
 APPENDIX
List of Commercial Hedge Fund Databases

Name Description Features of Indices

HFR (www.hfr.com) Hedge Fund Research (HFR) is a Around 1,500 funds are used to
hedge fund research and calculate 33 indices that reflect the
consulting firm that has collected monthly net of fee returns on
data on around 4,000 different equally weighted baskets of funds.
hedge funds.
Zurich Capital Markets Originally developed by Managed Database contains 1,500 hedge
(www.marhedge.com) Accounts Reports (MAR) but it was funds, which are used to calculate
sold to Zurich Capital Markets in 19 indices that reflect median
Mar 2001. monthly net of fee returns.
CSFB/Tremont The TASS database tracks around Using a subset of around 650
(www.hedgeindex.com) 2,600 funds. There are strict rules funds, CSFB/Tremont calculates 10
for fund selection. The universe indices that the monthly net of fee
consists only of funds with a returns on an asset-weighted
minimum of USD 10m under basket of funds. Large fund have a
management and a current audited larger influence in these indices.
financial statement. Funds are re-
selected quarterly as necessary.
Hennesse The Hennesse Group is a hedge Based of subset of about 500
(www.henessegroup.com) fund advisory firm that maintains a funds, Hennessee calculates 23
database of around 3000 funds. indices that reflect the monthly net
of fee returns on equally weighted
basket of funds.
Van (www.vanhedge.com) Van Hedge Fund Advisors is a Using a subset of around 500
hedge fund advisory firm with a funds, Van calculates 15 indices
database of about 3,400 funds. that reflect the monthly net of fees
returns on equally-weighted
baskets of funds
Altvest (www.altvest.com) Altvest is hedge fund website that Altvest calculates 14 equally
provides information on alternative weighted indices from the monthly
investments. The Altvest database net of fee returns of the funds in its
contains information on around database.
2000 hedge funds.
TUNA Hedgefund.net is a website Hedgefund.net calculates 35
(www.hedgefund.net) providing free hedge fund equally weighted indices from the
information and performance data. monthly net of fee returns of the
Its database covers 1,800 hedge funds in its database. In Tuna's
funds. case, if a fund shuts down, it is
completely removed from the
5
indices
AsiaHedge AsiaHedge is a subscription AsiaHedge establish the Bank of
(www.hedgefundintelligence.com) database that provides information Bermuda AsiaHedge indices. There
on hedge fund industry in the Asia are 4 indices to measure the
Pacific Region. Publishes a league performance of hedge funds in 4
table of 156 funds. geographies based on the median
net of fee returns of funds in its
league table.
Source: Brooks and Kat (2001), Hedge Fund Intelligence

5
Estimated returns may suffer from survivor bias (ranging from 1.5-3%). Around 30% of newly established
funds do not survive beyond 3 years. Most data vendors (with the exception of TUNA) do incorporate funds
that have ceased to exist in their index to avoid this.

27
 References

Agarwal, V., and N.Y. Naik (1999) ‘Performance Evaluation of Hedge Funds with
Option-based and Buy-and-Hold Strategies, "A Primer on Hedge Funds,” Journal of
Empirical Finance, 6, pp 309-331.

AMIR Performance Presentation Standards ™ Handbook, Association for Investment

Management and Research, 2nd Edition, 1997.

Brooks, C. and H.M. Kat (2001), "The Statistical Properties of Hedge Fund Index
Returns and their Implications for Investors," Working Paper, ISMA Centre.

Brown, Stephen J. and William N. Goetzmann (2001), “Hedge Funds with Style,”
Working Paper 8173. National Bureau of Economic Research.

Eichengreen, Barry and Donald Mathieson (1998), Hedge Funds and Financial
Market Dynamics. International Monetary Fund, Washington, D.C.

Fung, W., and D.A. Hsieh (1997), "Empirical characteristics of dynamic trading
strategies: the case of hedge funds.” Review of Financial Studies, 10, 275-302.

Fung, W., and D.A. Hsieh (1999a), "Is Mean-Variance Analysis Applicable to Hedge
Funds,” Economic Letters, 62 pp. 53-58.

Fung, W., and D.A. Hsieh (1999b), "A Primer on Hedge Funds,” Journal of Empirical
Finance, 6, pp 309-331.

Fung, W., and D.A. Hsieh (2000), ‘Performance Characteristics of Hedge Funds and
CTA Funds: Natural Versus Spurious Biases,” Journal of Financial and Quantitative
Analysis, Vol.35 No.3, pp 291-307.

Fung, W., and D.A. Hsieh (2001a), "The Risk in Hedge Fund Strategies: Theory and
Evidence from Trend Followers,” Review of Financial Studies, 14, pp. 313-341.

Fung, W., and D.A. Hsieh (2001b), "Asset-Based Hedge-Fund Styles and Portfolio
Diversification,” Financial Analyst Journal.

Lavinio, Stefano, The Hedge Fund Handbook, Irwin Library of Investment & Finance,
McGraw-Hill 2000.

Lehman Brothers (2000), "The Benefits of Hedge Funds: Asset Allocation for the
Institutional Investor."

Lo, Andrew W. (1991), "Long-Term Memory in Stock Prices," Econometrica, 59, pp

1279-1313.

Lo, Andrew W. (2001), "Risk Management for Hedge Funds: Introduction and
Overview," Financial Analyst Journal, 57, pp 16-33.

Morgan Stanley Dean Witter (2000), "Why Hedge Funds Make Sense."

Park, J. (1995), "Managed Futures as an Investment Set," Doctoral dissertation,

Columbia University.

28
Sortino, Frank A. and Lee N. Price (1994), “Performance Measurement in a
Downside Risk Framework,” The Journal of Investing, Fall.

Till, Hilary (2001a), “Life at Sharpe’s End”, Sep Issue, Risk & Reward.

Till, Hilary (2001b), “Measure for Measure”, Oct Issue, Risk & Reward.

UBS Warburg (2000), ”The Search for Alpha Continues.”

Weisman, Andrew and J. Abernathy (2000), "The Dangers of Historical Hedge Fund
Data," in L. Rahl, ed. Risk Budgeting, Capital Market Risk Advisors.

29