Hedge Funds*
2. Performance of Hedge Funds
2.2 Performance Measurement Biases 20 .1 Survivor Bias
Survivor bias results when managers with poor track records exit the business, while managers with good records remain. If the survivor bias is large, then the historical record of the average return of sur- viving managers is higher than the average return of all managers over the test period. Since a diversified portfolio would likely consist of funds that are destined to fail as well as funds destined to succeed,
20Other biases discussed in the literature but not here are those found to have little impact on performance. For example, Fung and Hsieh (2000) found the self-selection and multi-period sampling biases small and negligible.
studying only survivors results in overestimation of the historical return. The extant literature reports the estimate of this bias to range from 0.16% (Ackerman et al., 1999) to 3.0% (Brown et al., 1999;
Fung and Hsieh, 2000).
Schneeweis, et al. (2002) point out that different market factors produce different rates of survival for different styles. Their results show that survivor bias is minor for event-driven strategies but is higher for hedged equity, and is considerable for currency funds.
Moreover, they show that survivor bias varies by style, time period, and economic conditions. They argue that the problem of survivor bias may be exaggerated if one assumes that current conditions do not evoke a market factor leading to increased probability of funds being driven out of business. Furthermore, they argue that the problem of survivor bias may be reduced by conducting superior due diligence or simply focusing on funds for which survivor bias may be reduced.
2.2.2 Stale Price Bias
Hedge fund managers have considerable flexibility both in valuing portfolios at month-end to determine the fund’s net asset value and in delaying reporting because of lax regulation by the authorities. It is not unreasonable to presume that hedge fund managers have an incentive to smooth income such that the fund asset value does not necessarily reflect the true market value. In addition, when securities held by hedge funds are thinly or infrequently traded, stale prices will induce an autocorrelated portfolio return pattern similar to patterns observed in the case of income smoothing (e.g., Campbell et al., 1997;
Chalmers et al., 2001; Lo, 2001). Thus, for securities with stale prices, whether induced by the manager or naturally arising from illiquidity (or thin trading), measured correlation may be lower than expected, and depending on the time period chosen, measured standard devia- tion may be higher or lower than would exist if actual prices existed.
Asness et al. (2001) argue that stale price bias can artificially reduce estimates of beta, volatility, and correlation with traditional indices.
They present evidence of significant lagged relations between monthly market returns (based on S&P 500) and reported hedge fund returns.
Kazemi and Schneeweis (2004) disagree and question the empirical validity of Asness et al.’s argument. They argue that there is no sig- nificant stale price bias effect on hedge fund returns for two reasons.
First, many hedge funds do not contain equity issues such that evi- dence of a correlation with lagged equity returns is not necessarily indicative of stale prices. Second, unlike tests of stale prices in tradi- tional markets, most research in hedge funds used monthly data. It is unlikely that monthly data would capture stale price effects over lengthy time period especially since for many hedge fund strategies, the under- lying holdings are relatively liquid compared to many traditional assets (e.g., real estate) or traditional alternatives such as private equity for which appraisal values are used. By replicating Asness et al.’s tests, Kazemi and Schneeweis (2004) empirically show that stale price bias was not significant and Asness et al.’s results were just a reflection of the historical anomaly due to the LTCM crisis in 1998.
2.2.3 Backfill Bias
Backfill (or instant history) bias happens when hedge funds choose to enter the database after achieving good performance with earlier good returns being backfilled between the fund’s inception date and the date of fund’s entry into the database. In other words, instant history of past good performance of a hedge fund will become part of the history of the hedge fund database. This tends to introduce an upward bias in the reported returns. Previous studies estimate this bias to be about 1.2–1.4% per year (Fung and Hsieh, 2000; Edward and Caglayan, 2001).
In summary, these biases overstate the return and understate the risk estimates of hedge funds. Thus, biases should be taken into con- sideration when the performance of hedge funds is being evaluated.
Also, because these biases are inherent in the databases, their effect will be escalated to the indices which are sometimes used as bench- marks. Hence, caution is called for when indices are used as bench- marks. Fung and Hsieh (2004) show that using funds of funds (FOF) as a benchmark for hedge fund performance mitigates some of the problems resulting from the biases mentioned above. However, it still
suffers from the problem of heterogeneity because the FOF indices are basically multi-strategy and they do not represent a specific strat- egy (Amenc and Martellini, 2002). It is thus not possible to use these FOF indices as benchmarks for individual hedge fund performance.
2.3 Hedge Fund Indices and Benchmarks
Many performance indices can be used as benchmarks for hedge fund performance. An appropriate benchmark shall reflect the particular style of an investment manager and serve as a surrogate for the man- ager in the studies of risk and return performance and asset allocation.
Earlier studies on the performance of hedge funds were often based on the use of various existing active manager-based hedge fund indices and subindices (Table 2).21 The use of manager-based hedge fund indices in performance and asset allocation is based on the premise that the indices adequately reflect the underlying performance
Table 2. Sample of Major Non-Investable Hedge Fund Indices
Index Launch Base Index Rebalancing
Provider Date Date Weighting Frequency
Barclay Group 2003 1997 EW Monthly
CISDM 1994 1990 Median Monthly
CS/Tremont 1999 1994 Value wt. Quarterly
Dow Jones 2004 2002 EW Quarterly
EACM 1996 1990 EW Annual
Edhec 2003 1997 PCA Quarterly
HedgeFund.net 1998 1976–1995♣ EW Continual
HFR 1994 1990 EW Monthly
MSCI 2002 2002 EW/VW* Quarterly**
Note: EW — Equal-weighted; VW — Value-weighted; PCA — Principal component analysis.
♣Depends on strategy.
*For the global indices.
**For inclusion and monthly for the reranking of funds.
Source: Géhin and Vaissié (2004), Yau et al. (2007).
21See Yau et al. (2007) for a description of the major hedge fund indices.
of the strategy used by investors for actual investment. It is also assumed that investors have access to these managers whose perform- ance is represented in the indices, i.e., funds managed by these man- agers are still open to new investors.
One of the primary concerns over the use of hedge fund indices is that since most databases are self-reported, indices may not reflect the performance of many individual managers because of its lack of representation. In addition, indices themselves may not be directly comparable in terms of expected return and risk.22 For example, Schneeweis et al. (2002) show that the actual historical returns for various well-known hedge fund indices vary widely over a five-year period as well as across “similar strategy” indices. They suggest that the fund size and the fund age restrictions on the membership of a hedge fund index are one of the causes of such variations.
Previous studies have also analyzed the actual tracking error between various hedge fund indices as well as various weighting schemes (e.g., value-weighted versus equal-weighted).23 Fung and Hsieh (2002) point out that indices that are value-weighted reflect the weights of pop- ular bets by investors since the asset value of the various funds change due to asset purchases as well as price. As such, the ability of an investor to track such an index based on a market momentum strategy is prob- lematic. Equal-weighted indices may better reflect potential diversifica- tion of hedge funds and funds designed to track such indices. However, the cost of rebalancing may make these indices likewise difficult to create in an investable form. Investable hedge fund indices have been created recently, which are themselves investable or created with the express goal of tracking a comparative non-investable index.
2.3.1 Investable Hedge Fund Indices
Many investable indices are based on either managed account plat- forms or direct investment in fund-based products. Examples of investable indices are presented in Table 3.
22Previous studies find that heterogeneity among hedge funds renders low correlation among different indices for the same universe (e.g., Amenc and Martellini, 2002).
23For example, McCarthy and Spurgin (1998) and Fung and Hsieh (2002).
Conceptually, the investable indices are passively managed funds of hedge funds. Géhin and Vassié (2004) have identified three common attributes of investable indices: 1) full transparency; 2) initial and ongo- ing due diligence; and 3) investability in terms of low capital entry level and high redemption frequency. The index providers basically furnish investors with a low-cost approach to selecting hedge fund managers and strategy, allocating assets (by fixing the index weighting), and elim- inating funds that are prone to operational risks in the future. They do not, however, attempt to allocate funds dynamically. Both management and incentive fees tend to be significantly lower for investable hedge fund indices than the funds of hedge funds (Géhin and Vassié, 2004).
Given that indices differ in a number of characteristics (e.g., strat- egy, availability, construction etc.), the common concern is that the number of managers/funds within the representative index is too small to represent the performance of the overall hedge fund universe or strategy. In brief, if the investable indices are not representative of the strategies, they are no more than funds of funds. Schneeweis and Remillard (2007) examine the Royal Bank of Canada (RBC) Hedge 250 Index which comprised approximately 250 funds,24and its
Table 3. Sample of Major Investable Hedge Fund Indices
Index Launch Base Strategy/Fund Rebalancing
Date Date Weighting Frequency
CS/Tremont 2003 2000 VW / VW Semi-annual
Investable
DJ BPI 2003 2002 NA/EW Quarterly
FTSE 2004 2004 IW/IW Annual
HFRX 2003 2003 VW/* Quarterly
MSCI Hedge 2003 2003 Adj. median Quarterly
Invest asset weighted/EW
RBC Hedge 250 2006 2005 TW/VW Monthly
Note: EW — Equal-weighted; VW — Value-weighted; TW — Target-weighted; NA — Not mentioned; IW — Investability-weighted.
*Fund-weighted optimizing correlation within group.
24For details on the characteristics of the RBC 250 Hedge Index, see www.rbchedge250.com.
constituent strategy-based sub-indices relative to other investable indices and non-investable and strategy-based indices in terms of risk and return to determine whether the investable index created based on a larger sample size is a better representation of the performance of hedge funds.25Their results show that for many hedge fund strategies, the RBC indices perform either markedly superior or inferior to other comparable indices. In addition, the RBC indices generally have higher volatility at the sub-index level and higher equity betas. However, the index history is still too short for a definitive conclusion.
Although the proliferation of investable indices is one step closer to benchmarking the hedge fund performance precisely, Schneeweis et al. (2003) note that one should not expect any one single hedge fund index to track the performance of hedge fund managers even within the same strategy. However, as for traditional securities, for a strategy-pure index of hedge funds, a portfolio of similar funds should have performance similar to that of the representative hedge fund index. The lack of a clear hedge fund benchmark, however, is not indicative of an inability to determine a comparable return for a hedge fund strategy. Hedge fund strategies within a particular style often trade similar assets with similar methodologies and are sensitive to similar market factors. Thus, replication of hedge fund returns is feasible.