Oct 01 Version

Oct-01 Version

Risk Analysis of Hedge Funds versus Long-Only Portfolios

Duen-Li Kao1

Correspondence: General Motors Asset Management 767 5th Avenue New York, N.Y. 10153 E-Mail: [email protected]

Current Draft: October 2001

1

Tony Kao is Managing Director of the Global Fixed Income Group at General Motors Asset Management. The author would like to thank Pengfei Xie and Kam Chang for their insightful research assistance. The author is grateful for many useful discussions with colleagues in the Global Fixed Income Group and constructive comments from Stan Kon, Eric Tang and participants at the “Q” Group Conference in spring 2001.

10/16/01 4:17 PM

-1

-

Oct-01 Version

Risk Analysis of Hedge Funds versus Long-Only Portfolios Introduction Despite the decade-long bull market in the 1990s and liquidity/credit crises in the late 90s, hedge fund investing has been gaining significant popularity among various types of investors. Total size of reported hedge funds increased four fold during the period 1994 to 20002. The Internet bubble and valuation concerns for global equity markets, especially among sectors such as telecommunications, media and technology, have provided additional catalysts for the soaring interest in hedge funds over the last two years. Institutional investors often use hedge funds as part of absolute return strategies in pursuing capital preservation while seeking high single to low double-digit returns. This strategy is primarily implemented by absolute return investors (e.g., endowments, foundations, high net-worth individuals). Allocations by corporate and public pension plans to hedge funds as a defined asset class is a recent phenomenon. A second application is to use hedge funds as an alternative to long-only investing through an alpha transfer process. This often involves combining hedge funds with various derivative overlays. The pension consulting and hedge fund communities have been advocating this application in view of long-only managers’ difficulty in achieving active returns over benchmarks. For example, pension plans can overlay an equity market neutral fund with equity index futures to create a synthetic equity long portfolio. To the extent the hedge fund component outperforms its funding cost (e.g., LIBOR), the alpha may be transferred back to a long equity portfolio via derivatives. In theory, one can reverse this process to form a pseudo-hedge fund. That is, an equity long-only manager’s alpha over an equity index can be transferred back to an absolute return fund by shorting equity futures. Most likely, 2

See TASS (2000). Estimated market size of hedge fund industry varies greatly. For example, Hennessee Hedge Fund Advisory puts it at $408 billion at the end of 2000 in contrast to $210 billion according to TASS.

10/16/01 4:17 PM

-2

-

Oct-01 Version

endowments and foundations would not pursue this fantasy strategy. Does a pure mathematical equivalence fail to convince these institutional investors to “expand” their hedge fund manager universe? Since theoretically one can transfer alphas from either long-only or long/short portfolios to a desired target investment, we can compare these two types of alphas over their respective benchmarks (index benchmark or LIBOR) on a common basis. It is a general perception that as a group, hedge fund managers produce just enough active return to earn their overall fees while long-only managers fail to do so. How different are these two types of alpha anyway? Do alphas from long-only and long/short investments present different return distributions? Do these alphas derive from different risk factors? This article examines these questions by examining empirical evidence of active performance differences in long-only versus long/short investing. It also provides potential explanations from the standpoint of compensation and investment constraints. To further gain insight of how hedge funds incur risks, the article reviews the evolution of methodologies for analyzing hedge fund risk. It first examines return/risk patterns of various hedge fund investments and issues related to data reliability. Risk factors related to market returns and financial markets are examined using performance indices of several popular hedge fund strategies. The article proposes an alternative method of analyzing “investment style” as applied to hedge fund investments. It also reviews the contingent claim approach to hedge fund risk analysis: replicating hedge fund’s optionlike payoffs or trading strategies. Classification of Hedge Funds Conventionally, hedge funds are classified into categories according to their trading strategies or styles. Sub-sectors of hedge funds include trend following, global/macro strategies, long-only, arbitrage, long-short, etc. Despite attempts by data vendors, practitioners and academics, no clear standard of classification currently exists as evident by diverse categories used by various data vendors. In addition, given a variety of

10/16/01 4:17 PM

-3

-

Oct-01 Version

dynamic investment strategies and multiple capital market instruments utilized within individual hedge funds, style classification of a hedge fund can be easily mishandled by data vendors or hedge funds themselves3. For a comprehensive discussion of the nature of these hedge fund strategies, see Fung and Hsieh (1999). In a broad sense, we can classify hedge fund styles according to how funds manage the first or second order of the distribution of systematic risk factors. From the viewpoint of the first order of factor distribution, hedge funds differ as to whether they are taking “market directional” bets. That is whether a fund is taking systematic versus idiosyncratic risk (e.g., credit, spread or event risks). On the other hand, we can examine how a hedge fund manages against the second order of factor distribution: volatility. For example, practitioners, for simplicity, often consider commodity trading advisors (CTA) long volatilities while arbitrageurs short volatilities. Thus, during extreme market volatilities, these two types of hedge funds tend to offset each other. Active Performance of Arbitrage Funds vs. Long-Only Portfolios Do hedge funds or active equity managers produce different types of alpha distributions? To isolate and compare these two types of alpha, we benchmark the funds’ performance versus their respective benchmarks. Arbitrage funds are measured against LIBOR and long-only portfolios against equity or bond market indices. We use the Frank Russell institutional long-only universe to represent long-only portfolios instead of a mutual fund universe as conventionally done by other studies. Arguably, the clientele of hedge funds is more likely to resemble institutional long-only portfolios than mutual funds. They both target more sophisticated and longer-term investors who may not require daily liquidity and thus, making it easier to pursue desired investment strategies. As for arbitrage funds, CSFB/Tremont hedge fund indices which are increasingly becoming the industry standard, are used. 3

In fact, this is one of the toughest problems in style classification. Most of data vendors use the category of multi-sector strategies to group those funds that are not easy to classify.

10/16/01 4:17 PM

-4

-

Oct-01 Version

It should be noted that the following simulation results make an implicit assumption of the alpha transfer process being perfect. That is, financing costs for both hedge funds and derivatives used in the transferring process are identical. As experienced by many practitioners in recent years, the violation of this assumption can introduce significant return variance to the transfer process. Exhibit 1 compares after-fee quarterly alphas of active U.S. long-only equity accounts versus the equity market neutral index for the period 1994 to 20004. The 45-degree line represents even performance of these two universes. Scatter points represent paired quarterly active performance under different equity market environments during the period. We use different types of points in the scatter plot to represent active performance under different states of equity markets. Solid points (diamond and square

Exhibit 1: Active U.S. Long-Only Equity vs. Equity Market Neutral for U.S. Equity Asset Class (Data Source: Frank Russell Company, CSFB/Tremont; All figures in %)

5

After-Fee Quarterly Excess Returns Over Respective Benchmarks Q1/94-Q4/00

4 3 2 1 0 -1 -2 -3

< -1 Std dev of S&P 500 > +1 Std dev of S&P 500

-4

+/- 1 Std dev of S&P 500 Even Performance Line

-5 -5

-4

-3

-2

-1

0

1

2

3

4

5

Market Neutral Excess Return

4

Spear and Wiltshire (2000) also investigate the return differences of equity market neutral managers and long-only equity universe and find similar results.

10/16/01 4:17 PM

-5

-

Oct-01 Version

shaped) are for large positive or negative equity market movements (observations outside of one standard deviation of the S&P 500 quarterly return distribution). Triangle/blank points represent normal equity market conditions. Below the 45-degree line, active return from equity market neutral strategy is greater than that of active U.S. equity accounts. Examining from the direction of x or y-axis, one can see that market neutral strategies had wider active return distributions than long-only accounts with a few observations at the extreme. Market neutral strategy outperformed its benchmark on an after-fee basis much more often than active long-only accounts did as indicated by more points below the 45-degree line. Furthermore, market neutral strategy performed better than the longonly accounts at extreme equity market conditions as also depicted by more solid points among them. Another interesting phenomenon is that long-only accounts produced negative active returns when equity markets are very strong. This is consistent with the findings of active performance of equity mutual funds from 1965 to 2000 by Mezrich et al. (2000). Conversely, market neutral funds generated positive alpha over LIBOR under these situations perhaps due to their positive exposures to the market risk factor (see the discussion in the later section). Turning to bond markets, Exhibit 2 shows similar results for fixed income arbitrage funds as compared with the active U.S. bond manager universe. However, active returns from bond portfolios produced a substantially narrower distribution as compared to fixed income arbitrage strategies. The most noticeable outliers for fixed income arbitrage performance are from the difficult periods for hedge funds: early 1994 and late 1998. High volatile outcomes should not surprise arbitrage fund investors since those funds tend to employ leverage that often averages five to ten times of the fund’s capital. In general, the investment objective of many fixed income arbitrage funds is to produce absolute returns comparable to equity markets with lower volatilities or higher return with comparable volatility. Since potential returns from relative value trades are often small, leverage is usually employed in order to achieve the return objective. However, this practice comes with a stiff price during credit or liquidity crises. As such, hedge funds often incur substantial losses from rapidly rising financial costs of leverage 10/16/01 4:17 PM

-6

-

Oct-01 Version

positions, forced liquidations stemming from margin calls at the worst market conditions and demands of true “marking-to-market” by brokers/dealers or from investors’ panic withdrawals.

Exhibit 2: Active U.S. Long-Only Bonds vs. Fixed Income Arbitrage for U.S. High Quality Bond Asset Class (Data Source: Frank Russell Company, CSFB/Tremon; All figures in %)

4

After-Fee Quarterly Excess Returns Over Respective Benchmarks Q1/94-Q4/00

3

2

1

0

-1

-2 < -1 Std dev of Leh Aggr > +1 Std dev of Leh Aggr +/- 1 Std dev of Leh Aggr

-3

Even Performance Line

-4 -7

-6

-5

-4

-3

-2

-1

0

1

2

3

4

Fixed Income Arbitrage Excess Return

Another possible reason for fixed income arbitrage funds having a more diverse active return distribution is attributed to differences in performance benchmarks. Fixed income arbitrage funds tend to stay within niche market segments where they have substantial expertise and devise various strategies to exploit investment opportunities. The performance index reflects various fixed income arbitrage funds employing a variety of fixed income relative value strategies. When they are measured against a simple and low volatile return benchmark (e.g., LIBOR, T-bills), the variance of alphas can easily be magnified. On the other hand, long-only managers tend to emphasize tracking errors when facing a more diversified and complex market benchmark. In measuring alpha, the return variance is largely offset by the market benchmark.

10/16/01 4:17 PM

-7

-

Oct-01 Version

One approach to deal with arbitrage funds’ volatility is to “de-lever” the investment. This can be accomplished by combining arbitrage investments with either short-term cash portfolios or a bond index fund depending on the objective of the overall portfolio in achieving absolute return or broad bond market exposures5. Exhibit 3 depicts the result of active returns of long-only bond portfolios versus the fixed income arbitrage index delevered by a ratio of one to ten. The “de-levered” bond portfolio would invest one-tenth of the asset in fixed income arbitrage fund with the remaining in a bond index fund. The hedge fund portion is further overlaid with bond derivatives to create synthetic bond exposures. As can be seen, a “de-levered” bond portfolio still offers higher alphas with comparable volatility. Moreover, negative active returns of this fund are generally not as severe as those of long-only portfolios during extreme bond market conditions. Exhibit 3: Active U.S. Long-Only Bonds vs. Fixed Income Arbitrage for U.S. High Quality Bond Asset Class: Risk “De-Levered” by Ratio of 10 to 1 (Data Source: Frank Russell Company, CSFB/Tremont; All figures in %) 1

After-Fee Quarterly Excess Returns Over Respective Benchmarks Q1/94-Q4/00

0.5

0

-0.5

-1

< -1 Std dev of Leh Aggr

-1.5

> +1 Std dev of Leh Aggr +/- 1 Std dev of Leh Aggr Even Performance Line

-2 -1

-0.5

0

0.5

1

Fixed Income Arbitrage Excess Return

5

If the investment objective of the “de-levered” portfolio is to achieve cash return, it implicitly assumes 90% of assets invests in LIBOR-based instruments.

10/16/01 4:17 PM

-8

-

Oct-01 Version

What type of hedge fund is a better source of alpha for a given asset class? Exhibit 4 compares excess returns of equity market neutral funds and fixed income arbitrage funds given equity market performance over the last seven years. The objective is to evaluate which is the better source of equity alpha if hedge fund’s alpha is transferred back to the equity asset class? It appears equity market neutral managers performed significantly better than fixed income arbitrage managers in most equity market conditions, even in extreme cases. They also had an active return distribution slightly tighter and less "fat tailed".

Exhibit 4: Equity Market Neutral vs. Fixed Income Arbitrage for U.S. Equity Asset Class (Data Source: CSFB/Tremont; All figures in %) 5

After-Fee Quarterly Excess Returns Over Respective Benchmarks Q1/94-Q4/00

4 3 2 1 0 -1 -2 -3 -4

< -1 Std dev of S&P 500 > +1 Std dev of S&P 500

-5

+/- 1 Std dev of S&P 500 Even Performance Line

-6 -7 -5

-4

-3

-2

-1

0

1

2

3

4

5

Market Neutral Excess Return

So what if alphas from these two types of hedge funds were transferred to the fixed income asset class? Exhibit 5 compares these alphas in different U.S. high quality bond market environments. Similar to the results in Exhibit 4, equity market neutral funds appear to provide more consistent sources of alpha to the U.S. bond asset class than a fixed income arbitrage strategy.

10/16/01 4:17 PM

-9

-

Oct-01 Version

Exhibit 5: Equity Market Neutral vs. Fixed Income Arbitrage for U.S. High Quality Bond Asset Class (Data Source: CSFB/Tremon; All figures in %.) 5

After-Fee Quarterly Excess Returns Over Respective Benchmarks Q1/94-Q4/00

4 3 2 1 0 -1 -2 -3 -4

< -1 Std dev of Leh Aggr > +1 Std dev of Leh Aggr

-5

+/- 1 Std dev of Leh Aggr Even Performance Line

-6 -7 -5

-4

-3

-2

-1

0

1

2

3

4

5

Market Neutral Excess Return

Based on previous exhibits, Exhibit 6 presents statistics of three different sources of afterfee active returns for equity and bond market asset classes over the last seven years. Market returns are divided into two states: the top half and bottom half among 28 quarters. A few observations are worth noting: •

Equity market neutral funds provided better and more consistent alphas for both equity and bond asset classes than other funds as evidenced by high average active returns and information ratios in all market conditions

•

Fixed income arbitrage funds seem more suitable for the bond asset class than for the equity asset class although information ratios were extremely low, especially without de-leveraging.

•

Both long-only equity and bond portfolios performed poorly compared with hedge funds, except for long-only bond accounts providing the most consistent alpha for the bond asset class when the market performed poorly (the bottom-half of market performance conditions).

10/16/01 4:17 PM

- 10

-

Oct-01 Version

Exhibit 6: Sources of After-Fee Active Returns, Q1/1994 to Q4/2000 (Data Source: CSFB/Tremont; All figures in %)

Equity Asset Class

Bond Asset Class

Equity

Fixed Inc.

Equity

Equity

Fixed Inc.

Bond

Mkt.Neutral

Arbitrage

Long-Only

Mkt.Neutral

Arbitrage

Long-Only

Avg. Excess Ret.

1.43

0.22

0.15

1.43

0.22

-0.11

Volatility

2.13

2.43

1.53

2.13

2.43

0.59

Info. Ratio

0.67

0.09

0.10

0.67

0.09

-0.19

Avg. Excess Ret.

2.08

1.03

0.10

1.32

0.19

-0.47

Volatility

2.41

2.27

1.29

1.76

2.59

0.55

Info. Ratio

0.86

0.45

0.08

0.75

0.07

-0.86

Avg. Excess Ret.

0.78

-0.60

0.19

1.54

0.24

0.25

Volatility

1.65

2.38

1.79

2.51

2.36

0.37

Info. Ratio

0.48

-0.25

0.11

0.62

0.10

0.67

Overall

0.50

0.23

-0.16

0.03

0.01

-0.73

Top Market Returns

0.50

-0.60

-0.30

0.13

0.01

-0.45

Bottom Market Returns

0.33

0.56

-0.14

0.13

0.07

-0.59

Statistics Overall

Top Market Returns

Bottom Market Returns

Correlation with Markets

•

Active returns of equity market neutral funds were positively correlated with the equity markets (about 0.5). It confirms the general perception of market neutral funds exhibiting some market directionality.

•

Fixed income arbitrage funds had higher correlations with equity markets than with bond markets. However, they performed poorly when equity market returns were high.

•

Active returns from equity and fixed income arbitrage funds were uncorrelated with bond markets.

•

Active returns of equity and bond long-only accounts showed negative correlations with their respective benchmarks in all market conditions, especially for long-only bond portfolios (-0.73).

10/16/01 4:17 PM

- 11

-

Oct-01 Version

In the sections that follow, we will examine potential explanations of market hedge and arbitrage funds appearing to be better sources of active returns than long-only portfolios. As for the comparison between hedge funds, why did equity market neutral funds have a more attractive active risk/return profile than fixed income arbitrage strategies? First of all, even though CSFB/Tremont indices used in this study are considered superior than most hedge fund data (Lhabitant, 2001), the time period covers only 1994 onward. The period examined here is not only short but generally regarded as a tough period for fixed income arbitrage strategies, e.g., 1994, 1997, 1998 and 1999. As shown above, positive exposures to market risk by equity market neutral funds further enhanced their performance advantages over fixed income arbitrage funds during equity bull markets. Furthermore, it should be noted that equity hedge funds (e.g., market neutral, convertible arbitrage, risk/merger arbitrage) have significant longer histories than fixed income funds. Many mistakes have been experienced by equity related hedge funds, especially during 1990 and 1991. Of course, fixed income related hedge funds learned an expensive lesson from the recent LTCM episode: the danger of accounting-based leverage, the power of margin calls, the importance of marking-to-market, and the unreliability of carry trades without proper downside risk hedges. Since then, fixed income hedge funds and the broker/dealer community have devised many remedies (willingly or unwillingly) in an attempt to avoid the same mistakes. For example, more fixed income arbitrage funds are employing leverage constraints, downside risk analytics, risk budgeting implementation and fund alliance6. Perhaps fixed income hedge funds will be able to reduce the performance gap versus equity hedge funds going forward.

6

I thank Eric Tang for pointing out these issues.

10/16/01 4:17 PM

- 12

-

Oct-01 Version

Hedge Funds versus Institutional Long-Only Portfolios Data Issues The previous section empirically compares active performance of hedge funds versus long-only portfolios. The conclusion should be taken carefully since hedge fund data by itself presents numerous debates among practitioners and academic researchers regarding its usefulness and reliability. Biases in the construction of hedge fund databases include survivorship, self-delist, selection and back filling7. On the other hand, even with the help of stringent disclosure requirements from investors and regulatory agencies, most performance databases of long-only portfolios also exhibit one or more of these database biases found in hedge funds8. The self-delist bias appears to be somewhat unique to hedge funds. Firms may stop reporting performance data to database vendors for a variety of reasons such as difficulty in executing trades due to asset capacity and potential liability in reporting errors. In addition, almost all the databases exhibit a significant selection bias. Most databases do not even include or gain the support of some large and preeminent hedge funds. Furthermore, many large hedge funds with impressive performance records catering primarily to financial institutions and institutional investors are not part of publicly available hedge fund databases. Perhaps the more significant issue of data reliability is the practice of stale pricing, questionable “mark-to-market” and “mark-to-model” employed by prime brokers and hedge funds (Asness et al., 2001; Tremont, 2000) 9. According to a recent survey of hedge fund valuation practices, price differences and valuation adjustments made by 7

For comprehensive discussions of issues related to hedge fund data, see Fung and Hsieh (2000a, 2001) and Brown et al. (1999). 8

For example, see Brown et al (1992) and Carhart (1997).

9

Pricing issues are even more severe and common before and during the LTCM debacle. However, we cannot untangle numerous other issues surrounding that market environment including price discovery process and the impact of dealer margin calls.

10/16/01 4:17 PM

- 13

-

Oct-01 Version

hedge funds can be substantial (30%-40%). This is especially problematic for illiquid or less liquid securities (e.g., high yield and distressed bonds, private securities, over-thecounter options, structured notes and mortgage derivatives)10. Stable pricing/modeling practice is essentially an artificial and costless process to smooth performance variation and “amortize” gains and losses11. It definitely contributes to hedge funds’ low return volatility, low correlation with other asset classes which in turn, enhances the notion of hedge funds being investment vehicles with high information ratios and great diversifiers. Stale pricing may well be the key factor underlying quarterly performance persistence of hedge funds found in Agarwal and Naik (2000). Despite the efforts by numerous studies in documenting and quantifying hedge fund data bias, conclusions based on the existing hedge fund databases were diverse and remain dubious. Thus, it is difficult to conclude the extent or even the direction of performance differentials between hedge funds and long-only accounts induced by database bias. Structural Differences One may argue that what lies beneath performance between hedge funds and long-only accounts are their differences in compensation structures, investment constraints from guidelines and regulations, and other structural factors12. These differences may allow hedge funds to: •

Focus on extracting returns related to idiosyncratic risks rather than relying primarily on taking systematic risks;

•

Serve as liquidity providers to hedgers;

10

Capital Market Risk Advisors, Inc. (2001)

11

Arguably, this is similar to the book value accounting used in insurance community.

12

Ackermann (2000) examine these issues as related to differences in performance persistence of mutual funds versus hedge funds.

10/16/01 4:17 PM

- 14

-

Oct-01 Version

•

Effectively execute certain investment strategies via various forms of derivatives; and

•

Customize investment/security structures to explore certain properties of return distributions.

The following table outlines various factors that may contribute the return differentials of these two types of alphas. Compensation

Investment Constraints

Structural Factors

Management fees

Leverage

Lockup period

Incentives

Short selling

Disclosure requirements

Hurdle rate

Use of derivatives

Asset capacity

High watermark

Concentrated positions

Simple benchmark

Management Capital

Investment guidelines

One of the common beliefs of hedge funds’ perceived outperformance is due to their unique compensation structure, which generally attracts supposedly more skillful professionals. Arguably, the most important factor is the setting of higher management fees in addition to potentially large payoffs from the incentive fee schedule (Ackermann et al., 1999)13. Furthermore, performance hurdle rate, high watermark14 and fund management contributing their own capital may provide hedge funds with additional drivers in achieving superior performance. Investment Constraints Another factor often cited for hedge funds’ outperformance is the flexibility they have in pursuing investment strategies. For example, short selling and the use of leverage are two of the trademarks of hedge fund management. Short selling allows fund managers to take advantage of their investment views on both sides of factor or security valuation. 13

It is often argued that this feature may encourage managers to take more risk. However, empirical studies indicate that this is not necessarily the case unless the implied option is deep out of money (Carpenter, 2000).

14

Incentive fees are earned only if cumulative performance recovers past shortfalls, if any.

10/16/01 4:17 PM

- 15

-

Oct-01 Version

Grinold and Kahn (2000) develop an analytical framework to quantify the efficiency gain from loosening the short selling restriction. They find that it can have significant impact on active management especially if they deal with large sets of assets, low volatilities and high active risk. However, it is questionable whether this flexibility does generate double alpha. Alexander (2000) empirically shows that if one considers Regulation T restriction, liquidity haircut and derivatives availability in short selling, abnormal returns from popular pricing "anomalies" based on zero investment strategies may not be supportive. With regard to leverage, it is conventionally defined as a discrete, accounting-based measure and does not give a complete indication of the type or amount of risk taken. It does not consider market volatilities and possible diversification benefits within portfolios. In fact, a fund may be able to reduce its leverage while increasing portfolio risk. In addition to the lack of actual leverage information, researchers have difficulty in empirically analyzing whether and how leverage improves a hedge fund's risk-adjusted return. Recent advances in hedge fund risk management call for risk-based definitions of leverage instead of conventional accounting measures (even if they include on- and offbalance sheet items)15. Incorporating value-at-risk and scenario stress tests should help investors better evaluate the true impact of portfolio leverage. Further research is needed to understand (1) the relationship between hedge fund return distribution and leverage, (2) leverage limits and proper leverage for various hedge fund strategies, and (3) leverage dynamics: what factors influence hedge funds’ use of leverage over a market cycle. Since most hedge funds focus on generating absolute returns with a "below-market" volatility, they are often measured against a simple performance benchmark: the funding cost. As a result, unlike long-only fund managers, hedge fund managers do not have to deal with issues related to benchmark style drift (e.g., Brealy and Kaplanis, 2001) and investment style boxes. In fact, pension sponsors and the consulting community are increasingly relying on "style" indices to monitor long-only fund managers and to construct risk/return profiles of overall asset classes. There is a tendency for investors to 15

See Sound Practices (2000), the President’s Working Group (1999), Norland et al (2000) for excellent discussion on this subject.

10/16/01 4:17 PM

- 16

-

Oct-01 Version

end up with a locally optimized asset class since their focus is often a collection of “optimized” managers within individual investment styles. Perhaps in response to this trend in institutional investing, long-only fund managers have shown increasing concern with tracking error and maverick risk. They tend to stay around the given style benchmark rather than stay with their supposed investment conviction. Focusing on "style" products and benchmarking may prove to be detrimental to long-only asset management going forward. In sharp contrast to long-only portfolios, hedge funds face few, if any, investment guideline restrictions. They are not limited by capital markets they can trade, constraints imposed by the Investment Company Act of 1940, and investment guidelines (e.g., sector/security limits and duration/spread duration risk limits) often found in a long-only portfolio. This may account for the tendency of hedge funds' extensive use of exotic securities or derivatives, and holding concentrated positions of what is considered "the best ideas" rather than overly diversified positions often found in a long-only portfolio. Finally, most hedge funds have investment lockup periods that allow hedge funds to use illiquid and restricted securities16. Anecdotally, all the flexibility discussed above may contribute to seemingly better risk-adjusted returns earned by hedge funds versus longonly portfolios. Other Issues Recently, questions have been raised regarding practices supposedly used by some hedge funds and Wall Street that may distort the true picture of hedge fund performance. These practices include: •

Potential conflict of interest from trade allocation by a firm managing both long-only and hedge funds in view of compensation differentials. The possibility of allocating

16

Lockup period is the time restriction of redeeming hedge fund investments. Ackermann (2000) empirically showed that the provision of lockup period and incentive structure are two of the most important contributors to hedge funds’ superior performance.

10/16/01 4:17 PM

- 17

-

Oct-01 Version

profitable trades to funds with substantially more profitable compensation structures has caught regulatory attention17. •

Trader order-handling sequence by brokers/dealers for hedge funds versus long-only (Santini, 2001). The allegedly preferential treatment of hedge funds is perhaps due to the tendency of hedge funds to have higher portfolio turnover rates and their willingness to pay higher commissions in order to obtain information flows from Street traders.

Understanding Hedge Fund Risk As hedge funds employ diversified and dynamic trading strategies in a rather loosely defined operating environment, the return generating process of hedge funds can be complex and hard to analyze. Most studies show that factors based on market returns of standard asset classes are not sufficient to describe risk taken by hedge funds, especially those employing market neutral or arbitrage strategies. So, what are additional systematic risks that hedge funds incur? Hedge fund risk is a function of quantity (leverage), instruments/markets traded, market volatility, strategy diversification within the fund and liquidity. One may argue that investors can get a better understanding of risk exposures by a hedge fund from examining portfolio holdings and trades. Value added from this exercise is generally considered questionable. Hedge funds tend to dynamically and rapidly shift trading positions and exposures to risk factors daily or intra-day. Portfolio holdings or transactions are difficult to piece back together to their original tactical or strategic purposes. Perhaps the most important aspect of hedge fund risk analysis is to understand the nature of trading strategies and underlying risk elements of each strategy. By doing so, the

17

See Financial Times (2001), HedgeWorld (2001).

10/16/01 4:17 PM

- 18

-

Oct-01 Version

investor can develop a more reasonable expected risk/return of the fund. He or she will better understand how and when trading strategies and funds invested are correlated. Low correlation is also often found between hedge fund categories focusing on different "style” or “markets". However, within each hedge fund category, correlations vary. Individual funds within market directional hedge fund categories tend to have higher correlation while non-directional funds often exhibit lower correlations (Brealey and Kaplanis, 2001; Martin, 2001). Diversification of trading strategies within a hedge fund is also a powerful tool for delivering consistent performance in various market conditions. Exhibit 7 shows paired return correlations of six different investment strategies employed by a successful capital structure arbitrage fund. Monthly correlations ranged from -0.35 to 0.41 during the period 1998 to 2000. Notwithstanding, in order for hedge funds to be able to perform consistently and survive difficult market environments, it is important for a manager to dynamically manage the optimal mix of these lowly correlated strategies.

Exhibit 7: Strategy Diversification within a Fund (Monthly Return Correlations, 3/98-12/00)

Convert Arb.

Yld-To C/P

Capital Multi-C Str. Arb. Stk.Arb.

Yield-to-Call/Put

0.34

Capital Structure Arb.

0.11

0.41

Multiclass Stock Arb.

0.33

0.06

0.19

Paired Trades

0.23

0.10

-0.35

0.05

Special Situations

0.06

0.23

0.05

-0.14

10/16/01 4:17 PM

- 19

-

Paired Trades

-0.08

Oct-01 Version

Performance Measures Conventionally, the hedge fund community likes to use singular measures to describe hedge fund performance and risk. For example, hedge fund marketing materials often present the fund’s standard deviation of returns, maximum drawdowns (peak-to-trough performance) and percentage of negative months (or quarters). Various risk adjustment ratios are also popular -- Sharpe ratio, information ratio (excess returns divided by volatility of excess returns), efficiency ratio (ex ante risk divided by realized return volatility) and appraisal ratio (significance of the intercept of a CAPM-type regression). All these risk/return measures do not express the nature of a fat tail return distribution nor do they address investors' concern that under certain types of market condition, the “true” risk of hedge fund investment will appear. Risk Factors Exhibit 8 depicts returns of fixed income arbitrage funds under various bond market performance levels. Monthly returns of the Lehman Aggregate Bond Index from 1994 to 2000 were classified into seven buckets according to their return rankings. As shown, fixed income arbitrage funds earned positive active returns in all types of bond market conditions. Searching for methods to analyze hedge fund risk beyond exposures to various market/sector portfolios, researchers attempt to identify economic or financial market factors as additional systematic risk taken by hedge funds. Financial market factors are primarily based on publicly traded instruments (e.g., changes in levels and volatilities of market index, index futures, options, swaps and other forms of derivatives). Unlike information based on economic conditions (e.g., inflation, GDP growth and industrial production), financial market factors have advantages of higher pricing frequency and are directly related to trading strategies used. These factors combined with market factors provide investors with a better analytical framework and empirically explain higher

10/16/01 4:17 PM

- 20

-

Oct-01 Version

Exhibit 8: Performance of Fixed Income Arbitrage Funds vs. Bond Market Returns (Monthly, 1/94 to 12/00). All figures in %. 3

2

1

0

FI Arb.

-1

Lehman Aggregate -2 State of Market Performance (lowest to highest)

portions of return variance than market risk factors alone18. Different hedge fund strategies may require different sets of factors to describe their risk propensity. Financial Market Risk Factors Continuing the example in Exhibit 8, we examine the performance of the fixed income arbitrage funds in different environments of fixed income volatilities during the period of 1995 to 2000. Volatility is represented by the changes in volatilities implied in the swaption market. As shown in Exhibit 9, the fixed income arbitrage strategy remarkably performed consistently in all but the highest volatility scenario. In fact, the only regime in which fixed income arbitrage funds averaged negative returns is when bond markets experienced their largest increases in implied volatilities (e.g., October 1997 and August to October 1998).

18

For example, see Martin (1999), Schneeweis and Spurgin (1998)

10/16/01 4:17 PM

- 21

-

Oct-01 Version

Exhibit 9: Performance of Fixed Income Arbitrage Funds vs. Bond Volatility Risk (Monthly, 1/95 to 12/00). All figures in %. 4

3

2

1

0

-1

FI Arb -2

3x10 Swaption Vol Change

-3

State of Factor Risk (lowest to highest)

Exhibit 10 presents other systematic risk factors critical to bond markets: the change in high yield spreads, Treasury volatility (implied volatility of Treasury options), swap volatility and equity volatility (implied volatility of the S&P 100 index options). Monthly excess returns of fixed income arbitrage funds over LIBOR show modest

Exhibit 10: Active Returns of Fixed Income Arbitrage Funds Under Different Risk Conditions (Monthly, 1/95 to 12/00). All figures in %. Most !

Ranks by Changes in Factors

Most "

1

2

3

4

5

6

Overall Correl.

HY Spreads

0.84

0.62

0.18

0.32

-0.04

-1.07

-0.46

Treasury Vol.

0.54

0.52

0.47

0.19

0.09

-0.95

-0.47

Swap Vol.

0.51

0.34

0.38

0.27

0.37

-1.01

-0.50

Whole Period

-0.72

0.02

0.57

0.56

0.28

0.15

0.23

Excl. 9,10/98

0.31

-

-

-

-

-

-0.27

Factors

Equity Vol.

10/16/01 4:17 PM

- 22

-

Oct-01 Version

negative correlations to the first three fixed income related systematic risk factors (about –0.5). The funds were most vulnerable when systematic risks drastically increased. High yield spread changes and Treasury volatility had a reasonably linear relationship with arbitrage funds’ active returns. As for the equity volatility factor, arbitrage funds performed the worst during extreme scenarios (both large declines and increases in the factor). However, excluding large decreases in equity volatilities following the LTCM episode (September and October of 1998), the correlation changed from a small positive to a small negative. This indicates that observations from that period (August to October 1998) have a critical impact on the analysis. Turning to convertible arbitrage funds, the same four systematic risk factors have similar impacts on active returns as shown in Exhibit 11. The underperformance of convertible arbitrage was most pronounced in regimes with the largest increases in three fixed income factors. At the first glance, the overall correlation of convertible funds and the changes in equity volatilities were virtually zero. At extreme market volatilities (the first and sixth states), the funds performed poorly as compared to more normal scenarios. Significantly negative performance from August to October 1998 (the impact is shown at

Exhibit 11: Active Returns of Convertible Arbitrage Funds Under Different Risk Conditions (Monthly, 1/95 to 12/00). All figures in %. Most !

Ranks by Changes in Factors

Most "

1

2

3

4

5

6

Overall Correl.

HY Spreads

0.95

1.03

0.55

0.94

0.80

-0.55

-0.46

Treasury Vol.

0.66

1.00

0.57

1.32

0.73

-0.56

-0.49

Swap Vol.

0.90

1.11

0.73

1.05

0.62

-0.69

-0.51

Whole Period

-0.34

0.59

1.30

0.88

1.20

0.09

0.00

Excl. 8/98

-0.34

-

-

-

-

0.56

0.41

Excl. 9,10/98

0.48

-

-

-

-

0.09

-0.39

Factors

Equity Vol.

10/16/01 4:17 PM

- 23

-

Oct-01 Version

the bottom of Exhibit 11) further demonstrates the vulnerability of convertible hedge funds during extremely volatile markets. During and after the LTCM debacle, convertible hedge funds are believed to have suffered significant "mark-to-market" issues that may have masked the extent of these relationships (Tremont, 2000). Exhibit 12 examines risk factor exposures of equity market neutral and long/short (directional) hedge funds. In addition to equity implied volatility, exposures to three Fama-French return factors are also analyzed. Market neutral funds show insignificant relationships to the changes in size and value factors. Their active performance was essentially flat when equity volatility increased the most.

Exhibit 12: Active Returns of Equity Hedge Funds Under Different Risk Conditions (Monthly, 1/95 to 12/00). All figures in %. Most !

Ranks by Changes in Factors

Most "

1

2

3

4

5

6

Overall Correl.

Equity Vol.

0.82

0.30

1.36

0.43

0.85

0.06

-0.29

Market Factor

-0.04

0.39

0.48

0.57

1.04

1.37

0.52

Size Factor (SML)

0.26

0.69

0.49

0.77

0.87

0.74

0.10

Value Factor (HML)

0.62

1.13

0.72

0.92

-0.20

0.62

-0.12

Equity Vol.

2.60

1.16

2.59

1.53

0.71

-1.62

-0.41

Market Factor

-3.33

-1.29

2.06

2.28

3.17

4.07

0.76

Size Factor (SML)

-2.67

-0.64

1.58

2.07

2.56

4.06

0.62

Value Factor (HML)

5.35

2.34

1.87

1.34

-0.70

-3.23

-0.77

Market Neutral

Long/Short

Despite what the name implies, the funds have positive directionality to the market factor (i.e., positive excess return increases as the equity market performs well). As for long/short hedge funds, they show strong correlations to all four systematic risk factors: short equity volatility and value factors while long market and size factors. Examining across six regimes, active returns of long/short funds had an almost perfect linear relationship to these factors.

10/16/01 4:17 PM

- 24

-

Oct-01 Version

Style Analysis of Hedge Fund Risk Style analysis, pioneered by Sharpe (1988 and 1992), uses market/sector benchmark portfolios as systematic factors to derive the asset mix implied in an active portfolio's return series. For a long-only portfolio, exposures to these market portfolios are constrained to be positive and summed to one. Many studies apply style analysis to analyzing hedge fund risk by relaxing these two constraints (e.g., Fung and Hsieh, 1997; Brown et al., 1999; Agarwal and Naik, 1999). While most studies employ capital market or style index portfolios as implied building blocks in style analysis, Lhabitant (2001) uses hedge fund style indices as risk factors in order to directly derive a fund’s implied exposures to conventional hedge fund styles/strategies. Brown and Goetzmann (2001) further extend hedge fund style analysis by allowing factor loadings on market portfolios (i.e., coefficients) to vary over time19. Using time varying factor loadings in style analysis is constructive since the method accommodates dynamic trading strategies with non-linear payoffs. All these studies found that individual fund returns have lower correlations to standard asset class returns as compared to mutual funds. Funds with styles of market neutrality, arbitrage or commodity have significantly low to nil exposures to these asset classes. Moreover, one of the criticisms of conventional style analysis is that investment risk as defined by these styles is too narrow and singular. It fails to recognize that investment risk is often multi-dimensional, asymmetrical and potentially correlated (Michaud, 1998). This problem becomes even more severe when analyzing hedge fund risk. Active returns of hedge funds generally exhibit asymmetric sensitivities to risk factors in different market environments. For example, it has been shown that hedge funds perform differently in positive versus negative equity markets (Lo, 2000) and in rising versus declining interest rate scenarios. Previous sections present empirical evidence of how the changes in implied volatilities in various capital markets may be of importance in 19

Brown and Goetzmann (1997) first present this methodology in studying mutual fund styles.

10/16/01 4:17 PM

- 25

-

Oct-01 Version

evaluating hedge fund strategies. In summary, to analyze hedge fund risks, we not only have to incorporate various systematic risk factors beyond conventional market return factors but also employ a multi-dimensional framework. Risk Style Analysis Under A Long-Only Framework Kao (2000a) presents a return-based approach to analyze investment styles of fixed income managers. It involves identifying several systematic risk factors important to active performance of a fixed income portfolio; e.g., changes in 10-year Treasury rate, implied volatility of interest rate options, swap spreads, swap volatility and systematic risks in equity markets20. Exposures to these risk factors in relation to a bond benchmark are grouped and summarized in two dimensions: interest rate risk and spread risk. Exhibit 13 compares two distinct long-only fixed income investment styles. The construction of this risk factor model follows Kon (1999) in which factors are adjusted for the variable dependence of prominent risk factors such as the level of interest rates. For example, the changes in implied volatilities are adjusted for directionality of ten-year Treasury rates. The changes in swap spreads are adjusted for both the changes in interest rates and the adjusted changes in volatility. The exhibit shows how portfolios managed active exposures to two risk dimensions differently with each point covering a rolling 36-month period. The center point represents a neutral position of risk exposures versus the benchmark. To illustrate the changes in exposures over time, the largest point is the most recent observation and the smallest the earliest. Manager A is a highly risk controlled bond fund of funds (diversified multiple advisors) as evident by its stable exposures to both risk dimensions. On the other hand, viewing from both interest rate and spread risk related to their benchmark, Manager B took more

20

See Kao (2000b) for an application to analyzing determinants of the changes in corporate credit spreads.

10/16/01 4:17 PM

- 26

-

Oct-01 Version

active risks with drastic shifts in exposures than Manager A. The bottom table of Exhibit 13 presents average statistics of risk exposures of the bond index and these two portfolios according to this risk style model.

Exhibit 13: Risk Styles of U.S. High Quality Core Bond Managers (8/99-6/00) 2.25

Benchmark: Salomon BIG Index Monthly Exposures: 8/99-6/00

2

Manager A 1.75

1.5

1.25

1

0.75

Manager B 0.5

0.25

0 0.9

1

1.1

Relative Interest Rate Risk

Average Statistics

R-Square

Ten-Year Rate

Int. Rate Volatility

Swap Spread

Equity Risk (t-1)

10-Year Rate

Int.Rate Risk

Four Facotrs

Coef.

-3.92

-0.77

-2.41

1.42

0.91

0.94

0.97

T-Stat

-31.65

-6.54

-5.48

2.81

Coef.

-3.80

-1.40

-2.97

2.80

0.68

0.77

0.83

T-Stat

-12.16

-4.71

-2.59

2.25

Coef.

-3.87

-1.09

-2.01

1.58

0.87

0.93

0.96

T-Stat

-24.62

-7.33

-3.62

2.47

Bond Index

Manager A

Manager B

The risk factor model explains return variances of these three portfolios very well as evidenced by the significance of T-statistics and R-squares. Manager B had larger exposures to all risk factors than Manager A and the benchmark except for exposure to the changes in 10-year Treasury rates. Obviously, two spread risk factors are very

10/16/01 4:17 PM

- 27

-

Oct-01 Version

important in explaining the return volatility of Manager B’s performance as R-square increases from 0.77 to 0.83. We apply the same risk factor model to examine relative risk exposures of fixed income arbitrage funds versus the long-only bond fund of funds from June 1998 to December 2000 in Exhibit 14. Monthly excess returns of fixed income arbitrage funds over LIBOR are assumed to transfer to a bond market index in order to make it comparable to the

Exhibit 14: Risk Styles of Fixed Income Arbitrage Overlay Versus Long-Only Bond Fund (6/98-12/00): Fixed Income Arbitrage de-levered by 10:1 1.5

Benchmark: Salomon BIG Index Quarterly Exposures: 6/98-6/00

Long-Only Bond Fund 1.25

1

Fixed Inc. Arb. Overlay

0.75

0.5 0.8

0.9

1

1.1

1.2

Relative Interest Rate Risk

Average Statistics

R-Square

Ten-Year Rate

Int. Rate Volatility

Swap Spread

Equity Risk (t-1)

(b.p.)

(10 b.p.)

(b.p.)

(%)

Coef.

-3.95

-0.71

-2.42

1.25

T-Stat

-34.98

-5.59

-4.91

2.58

Coef.

-3.80

-1.05

-2.01

1.58

T-Stat

-22.00

-5.40

-2.73

2.11

Coef.

-3.94

-1.04

-2.09

1.66

T-Stat

-25.45

-6.36

-3.16

2.58

10-Year Rate

Int.Rate Risk

Four Facotrs

0.92

0.95

0.97

0.86

0.91

0.94

0.88

0.93

0.96

Bond Index

Fixed Inc. Arb.

Long-Only Fund

10/16/01 4:17 PM

- 28

-

Oct-01 Version

long-only portfolio. Furthermore, as in the case of Exhibit 3, to make return volatilities of these two portfolios more comparable, the performance of the arbitrage fund index was “de-levered” by investing one-tenth of assets in hedge funds and the remainder in a bond index fund. Again, the analysis is done on a 36-month rolling basis to explore the funds’ changes in risk exposures (only quarter-end observations are displayed). Exhibit 14 shows these two investments possess similar and rather consistent exposures to both directional and volatility risks. The “de-levered” fixed income arbitrage overlay portfolio had slightly lower relative interest rate and spread risks than the long-only bond fund. This was achieved through having lower exposures to ten-year interest rate and equity risk factors. Comparing with the bond market index, however, this portfolio still had higher exposures to volatility and equity risk factors. Remarkably, hedge fund overlay and long-only portfolios also changed their exposures over time in a similar pattern. After the LTCM debacle, both portfolios became more risk neutral versus the benchmark. Risk Style Analysis Under A Hedge Fund Framework If we were to analyze the source of active risk of hedge funds on a stand-alone basis (i.e., without an overlay process), the risk factor model requires some modifications. First, we define risk style dimensions relevant to hedge fund investment risks: directional risk (first order) and volatility risk (second order). Continuing the example in Exhibit 14, systematic risk factors important to fixed income arbitrage funds as discussed in previous sections are categorized into these two dimensions. For example, exposures to the changes in interest rates and credit spreads are jointly formed to measure directional risk (correlation of these two factors is considered). Volatility risk combines the changes in implied volatilities of equity and interest rate options. Again, the model construction requires the adjustment of variable dependence. Exhibit 15 compares risk exposures of after-fee active returns of fixed income arbitrage funds and a long-only bond fund over their respective benchmarks. Fixed income

10/16/01 4:17 PM

- 29

-

Oct-01 Version

Exhibit 15: Risk Style of Active Quarterly Returns of Fixed Income Arbitrage Versus Long-Only Bond Fund (12/98-12/00) 15

Benchmarks: 3-Mon. LIBOR and Salomon BIG Index Quarterly Exposures: 12/98-12/00

Fixed Income Arb Index

12.5

10

7.5

Active US Bond FoF

5 0

1

2

3

4

5

6

7

8

9

10

Directional Risk

Average Statistics

R-Square

10-Year Rate

HighYld Spread

Int.Rate Volatility

Equity Volatility

(b.p.)

(b.p.)

(10 b.p.)

(10 b.p.)

Coef.

0.18

-0.21

-0.24

0.14

T-Stat

2.28

-2.21

-2.04

3.96

Coef.

0.04

-0.05

-0.33

0.09

T-Stat

0.56

-0.51

-2.76

2.41

10-Year Rate

Direct'l Risk

Four Facotrs

0.08

0.30

0.58

0.01

0.16

0.42

Fixed Inc. Arb.

Long-Only Fund

arbitrage funds have large and statistically significant active risk exposures to all four factors, especially to the changes in high yield spreads and equity volatilities as compared to the long-only fund. Viewing from risk factors important to fixed income arbitrage funds, volatility risk had significant impact on active returns of the long-only bond fund. In contrast to fixed income arbitrage funds, the directional risk factor (ten-year rate) had nil effect on active return variance of the long-only fund. As indicated by R-square measures in the last three columns of the exhibit, factors related to directional risk explain about 30% of active return variance of fixed income arbitrage

10/16/01 4:17 PM

- 30

-

Oct-01 Version

portfolios (ranging from 18% to 41% during the period). This is substantially higher than the 8% achieved if only the changes in interest rate levels (10-year rates) is used21. Adding volatility risk factors, average explanatory power increases to 58% for the arbitrage index. As for the long-only bond fund, directional risk explains 16% of active return variance and volatility risk factors add another 26%. During this period, hedge funds as well as the long-only fund generally maintained their directional risk but decreased exposures to volatility risk.

Mimicking Portfolio/Strategy Approach to Risk Analysis Recently, several researchers have taken a more direct approach to analyze hedge fund’s systematic risk beyond market returns. We call this the mimicking portfolio/strategy approach since it attempts to replicate either the payoff pattern or explicit trading strategies of hedge fund activities22. Fung and Hsieh (1997) apply principal component analysis to extract benchmarks for various trading strategies as implied in hedge fund return series. When combined with conventional asset class factors, it can effectively capture the essence of hedge funds' extreme outcomes. Following the contingent claim concept of performance measurement advocated by Glosten and Jagannathan (1994), several studies use a series of financial options to directly replicate the option-like pattern which existed in hedge fund data23. Other methods involve constructing naive trading strategies actually employed by hedge funds

21

As a reference, if one follows the conventional approach of using a bond market index as the risk factor (e.g., Lehman Aggregate Index), the R-square is only 3%.

22

Broadly speaking, style analysis approach using market/factor portfolios or risk factors can be considered a mimicking portfolio/strategy method for analyzing a fund’s risk and return.

23

Fung and Hsieh (2000b) construct five trend-following mimicking benchmarks that produce straddle option payoffs commonly observed in hedge fund returns. R-squares were about 48% versus average 7% with standard asset return factors. Agarwal and Naik (2001) also employ a similar methodology to studying Event Driven and Relative Value Arbitrage funds. Lo (2000) uses a trading strategy of selling out-of-the-money puts on equity index to demonstrate the illusion of a hypothetical hedge fund’s super “risk”-adjusted performance.

10/16/01 4:17 PM

- 31

-

Oct-01 Version

and thus, provide a more direct and realistic evaluation framework24. Tang (1999) extends the framework by simulating hypothetical investment opportunities available to hedge fund managers, rather than replicating hedge funds’ trading strategies and instruments used. The approach attempts to address a difficult task in hedge fund research: hedge funds (especially arbitrageurs) generally employ multiple investment strategies within a fund that are seemingly uncorrelated and hard to replicate by trading unified instruments25. In all these studies, they found that return patterns from these simulated passive trading strategies resemble those of actual hedge funds or CTAs. Risk attributes detected from these time series are generally consistent with what we would expect from specific trading strategies employed by hedge funds. Return series obtained from this analytical approach can be used to: •

Evaluate and extract various systematic risks not observed by return series of conventional asset classes. In the spirit of Sharpe’s style analysis framework, mimicking portfolios can be viewed as alternative or additional asset/benchmark style factors.

•

Directly model hedge fund's asymmetric return distributions.

•

Examine how hedge funds manage their risk exposures in extreme market conditions.

•

Serve as a “true” hedge fund benchmark26. Performance in excess of these benchmark portfolios is considered a better indication of the manager's skill.

24

See Gatev et al (1999) on paired trading (a convergence strategy used to explore relative pricing of close substitutes of financial instruments), Mitchell and Pulvino (2000) on risk arbitrage strategy and Richards (1999) on relative value trades, and Liew (1999) on equity long and short of equity risk factors. Return indices (e.g., Mount Lucas Management Index) based on naive trading strategies in active commodity and financial futures is used in analyzing CTA investment risks (see Schneeweis and Spurgin, 1998; Spurgin, 1999). 25

Under this approach composite relative value indices for capital market segments in which hedge funds operate are constructed. Each relative value index combines factors related to rich/cheap valuation and technical indicators for the market at a given point of time. For example, for yield curve trades, it calculates relative value opportunities available to carry, butterfly and basis trades. As for technical factors, it uses measures such as spreads versus their historical averages popular among practitioners.

26

In fact, recently a few hedge funds replicating index or naïve trading strategies are being publicized as passive alternatives to active hedge fund investing.

10/16/01 4:17 PM

- 32

-

Oct-01 Version

•

Avoid biases found in most hedge fund databases as discussed previously.

The replication approach to studying hedge fund performance is expected to extend to other types of trading strategies. This should shed light on the myth surrounding hedge fund activities. Investment Style and Performance Evaluation Both risk style factors and mimicking portfolios can be useful in understanding hedge fund risk. They serve as better yardsticks for measuring hedge funds’ performance and their true active skill beyond naïve trading strategies. However, the hedge fund investment community should keep in mind the experience of improving methods of measuring long-only portfolio performance in recent years. Style analysis was originally designed to facilitate the evaluation of a money manager’s active skill in view of their exposures to some systematic risks. Style indices created from this analytical framework are not intended to be a primary tool for managing money managers. Investors and consultants tend to put too much emphasis on the performance tracking error versus a style benchmark or a customized benchmark based on a set of systematic risk factors. By doing so, they “delegate” the responsibility of understanding managers’ investment process and what truly drives active performance to a classification scheme based on singular factor measures. The end result is the danger of further restricting (implicitly or explicitly) an investment manager in expressing his/her true convictions. This would be especially troublesome for hedge funds whose active returns rely on multiple, complex and dynamic trading strategies that may not be easily classified into one particular style box. Conclusion The option-like return pattern of hedge funds presents a challenge for investors in analyzing risk exposures. Singular measures of risk and return can be misleading especially in analyzing hedge fund risk. Investors should carefully examine the return

10/16/01 4:17 PM

- 33

-

Oct-01 Version

patterns under various market conditions and other systematic risk factor exposures. Due to the investor’s ability of transferring alpha to a desired asset class, it is more appropriate to evaluate hedge funds and long-only portfolios by comparing them against respective benchmarks. Hedge funds, especially equity market neutral strategies, seem to provide more consistent alpha than long-only portfolios for different asset classes under various market environments. The qualitative assessments of possible explanations are reviewed here. Factors derived from asset prices in financial markets are timely and useful for hedge fund risk analysis. These risk factors depict exposures to market direction, volatility and valuation that are most relevant to hedge fund’s risk profiles. This article shows that how a hedge fund manages its exposures to implied volatilities at extreme market conditions can be the key to consistent performance. The results highlight the importance of strategy diversification between funds as well as within a fund in achieving consistent performance. An analytical framework incorporating multiple risk factors gives investors a more complete picture of hedge fund risk taking. In the spirit of equity style analysis popular among practitioners, this article presents an approach of risk style analysis to evaluate common risk factors driving the performance of hedge funds and long-only portfolios. Various financial market risk indicators can be categorized into directional and volatility risk dimensions to provide a more concise assessment of risk exposures over time. Another approach to analyze hedge fund risk is to directly replicate the hedge fund’s option payoff profile, trading strategies employed or arbitrage opportunities available. Return series derived from this mimicking approach is particular useful in studying risk factors and performance attributes underlying hedge fund investing. It also provides a promising direction for future research of hedge fund asset pricing.

10/16/01 4:17 PM

- 34

-

Oct-01 Version

References Ackermann, Carl, R. McEnally and D. Ravenscraft. 1999. “The Performance of Hedge Funds: Risk, Return, and Incentives.” Journal of Finance, 54:833-874. Ackermann, Carl. 2000. “Essays on Hedge Funds.” Ph.D. Dissertation, University of North Carolina, Chapel Hill. Agarwal, Vikas and Narayan Naik. 1999. “On Taking the ‘Alternative’ Route: Risks, Rewards Style and Performance Persistence of Hedge Funds.” Working paper, London Business School. __________. 2000. “Multi-Period Performance Persistence Analysis of Hedge Funds.” Journal of Financial and Quantitative Analysis, 35:327-342. __________. 2001. “Characterizing Hedge Fund Risks with Buy-and-Hold and OptionBased Strategies.” Working paper, London Business School. Alexander, Gordon. 2000. “On Back-Testing ‘Zero Investment’ Strategies.” Journal of Business, vol. 73, no. 2:255-278. Asness, Clifford, Robert Krail and John Liew. 2000. “Do Hedge Funds Hedge?” Working paper, AQR Capital Management. Brealey, Richard and Evi Kaplanis. 2001. “Changes in the Factor Exposures of Hedge Funds.” Working paper, London Business School. Brown, Stephen and William Goetzmann. 1997. “Mutual Fund Styles.” Journal of Financial Economics, 43:373-399. __________. 2001. “Hedge Funds with Style.” Working paper 00-29, Yale University. Brown, Stephen, William Goetzmann and Roger Ibbotson. 1999. “Offshore Hedge Funds: Survival and Performance1989-1995.” Journal of Business, 72:91-118. Brown, Stephen, William Goetzmann, Roger Ibbotson and Stephen Ross. 1992. “Survivorship Bias in Performance Studies.” Review of Financial Studies, 5:553580. Capital Market Risk Advisors, Inc. 2001. “NAV/Fair Value Practices Survey Results.” July 9. Carhart, Mark. 1997. “Mutual Fund Survivorship.” Working paper, University of Southern California. Carpenter, Jennifer. 2000. “Does Option Compensation Increase Managerial Risk Appetite?” Journal of Finance, 55:2311-2331 10/16/01 4:17 PM

- 35

-

Oct-01 Version

Financial Times. 2001. “More Scrutiny for Hedge Fund and Mutual Fund Managers.” March 27. Fung, William and David Hsieh. 1997. “Empirical Characteristics of Dynamic Trading Strategies: The Case of Hedge Funds.” Review of Financial Studies, 10:275-302. __________. 1999. “A Primer for Hedge Funds.” Journal of Empirical Finance, __________. 2000a. “Performance Characteristics of Hedge Funds and CTA Funds: Natural versus Spurious Biases.” Journal of Quantitative and Financial Analysis. __________. 2000b. “The Risk in Hedge Fund Strategies: Theory and Evidence from Trend Followers.” Review of Financial Studies, . __________. 2001. “Benchmarks for Hedge Fund Performance: Information Content and Measurement Biases.” Financial Analysts Journal, Forthcoming. Gatev, Evan, William Goetzmann and Geert Rouweinhorst. 1999. “Pairs Trading: Performance of a Relative Value Arbitrage Rule.” Working paper 7032, National Bureau of Economic Research. Glosten, Lawrence and Ravi Jagannathan. 1994. “A Contingent Claim Approach to Performance Evaluation.” Journal of Empirical Finance, 1:133-160. Grinold, Richard and Ronald Kahn. 2000. “Efficiency Gains of Long-Short Investing.” Financial Analysts Journal, November/December:40-53. HedgeWorld. 2001. “Rivalry Gets Uglier: Hedge Funds vs. Mutual Funds.” The Alternative Edge Newsletter, June 4. Lhabitant, Francois-Serge. 2001. “Assessing Market Risk for Hedge Funds and Hedge Funds Portfolios.” Research Paper No. 24, International Center for Financial Asset Management and Engineering. Liew, Jimmy Kyung-Soo. 1999. “Essays in Investment Strategies.” Ph.D. Dissertation, Columbia University. Lo, Andew. 2001. “Risk Management for Hedge Funds: Introduction and Overview.” Working paper, presented at the PricewaterhouseCoopers Risk Institute 2000 Conference. Kao, Duen-Li. 2000a. “Analyzing Investment Styles of Fixed Income Managers.” Journal of Investment Consulting, December. __________. 2000b. “Estimating and Pricing Credit Risk: An Overview.” Financial Analysts Journal, 56 (July/August): 50-66.

10/16/01 4:17 PM

- 36

-

Oct-01 Version

Kon, Stanley. 1999. “Mortgage Portfolio Performance Attribution.” Working paper, Smith Breeden Associates. Martin, George. 2001. “Making Sense of Hedge Fund Returns: What Matters and What Doesn’t.” Derivative Strategy. Mezrich, Jpseph, Qi Zeng and Matthew Rothman. 2000. “The Return of Stock Picker”. Quantitative Strategies, Morgan Stanley Dean Witter, December. Michaud, Richard. 1998. “Is Value Multidimensional? Implications for Style Management and Global Stock Selection.” The Journal of Investing, Spring:6165. Mitchell, Mark and Todd Pulvino. 2000. “Characteristics of Risk and Return in Risk Arbitrage.” Working paper, Harvard University. Norland, Erik, Jose Maio Quintana and Sykes Wilford. 2000. “Leverage: A Very Misleading Way to Measure Risk.” Derivatives Quarterly, Fall:3-15. The President’s Working Group on Financial Markets. 1999. Hedge Funds, Leverage and the Lessons of Long-Term Capital Management. Report of the President’s Working Group on Financial Markets, April 28. Richards, Anthony. 1999. “Idiosyncratic Risk: An Empirical Analysis, with Implications for the Risk of Relative-Value Trading Strategies.” Working paper 99-148, International Monetary Funds. Santini, Laura. 2001. “Trading’s Hidden Costs Hedge Funds, Technophobia and Benchmarking Games Eat Into the Performance of Mutual Funds.” The Investment Dealers’ Digest, August 14. Schneeweis, Thomas and Richard Spurgin. 1998. “Multifactor Analysis of Hedge Fund, Managed Futures, and Mutual Fund Return and Risk Characteristics.” Journal of Alternative Investments, 1 (Fall):1-24. Sharpe, William. 1988. “Determining a Fund’s Effective Asset Mix.” Investment Management Review, December:59-69. Sharpe, William. 1992. “Asset Allocation: Management Style and Performance Measurement.” Journal of Portfolio Management, Winter:7-19. Spear Sheena and Steve Wiltshire. 2000. “The Long and the Short of Market Neutral Investing.” Russell Research Commentary. Suprgin, Richard. 1999. “A Benchmark for Commodity Trading Advisor Performance.” Journal of Alternative Investments, 2 (Summer):11-21.

10/16/01 4:17 PM

- 37

-

Oct-01 Version

Sound Practices for Hedge Fund Managers. 200027. Tang, Eric. 1999. AlphaTrack. Working paper, Portfolio Management Technology. TASS. 2000. The TASS Asset Flows Report: Fourth Quarter 2000. Tremont. 2000. “Convertible Arbitrage: Opportunity & Risk.” White paper, September.

27

An excellent report on risk disclosure and management of hedge funds that was jointly developed by several well-known large hedge funds. The report can be accessed at www.hfinsoundpractices.com

10/16/01 4:17 PM

- 38

-