Peer group based style factors provide information not only on industry returns but also on risk characteristics and correlation structures with other investment styles. the average returns for the style factors are most often computed using equal weights for each peer group member.
Credit Suisse First Boston (CSFB)/Tremont is one of the few data providers that introduced in 2000 nine value-weighted indices (equity market neutral, long/short equity, dedicated short, managed futures, emerging markets, event driven, global macro, convertible arbitrage, and fixed income arbitrage).
Fung and Hsieh (2004) note that while value-weighted benchmarks are preferable as they take into consideration the disproportional allocation to large funds, computing the appropriate weights using assets under management may be problematic for highly levered investment strategies because their invested risk capital is substantially higher.
Some index providers require a minimum for assets under management or disclosure standards, such as audited financial statements, for a fund to be included in the index. Amenc and Martellini(2003) present an overview of different industry standards for peer group based benchmarks.
Given that hedge funds employ dynamic trading strategies, hold leveraged portfolios, and invest in derivative products makes it very challenging to determine a fair benchmark to assess the skills of a manager. Traditional performance benchmarks for mutual funds, such as the S&P 500 or the Russell style indices, are no longer adequate.
Therefore, investors turned to peer group averages. the advantage of using peer group based style factors is the comparison with strategies that have been implemented in practice and, thus, account for trading and transaction costs. On the downside, peer group based style factors typically rely on self-declared investment objectives and self-reported returns.
There exist neither accepted norms for classifying hedge funds nor standards for reporting realized returns. To remedy the issues with selfdeclared investment objectives, peer groups can be extracted using cluster analysis or (constrained) regressions on the return series of primitive trading strategies representing specific styles.
Another limitation to be considered is that, depending on the provider, indices are constructed from different databases. It is well documented in the literature that the most commonly used hedge fund data sources use a different nomenclature, are incomplete (selection bias, instant history bias) and exhibit major sampling differences.
In addition to the problem that omitting dead funds may lead to overestimated industry returns, survivorship bias is of concern as hedge funds on the decline become small and are eventually eliminated from the peer group. Agarwal et al. (2006) merge the CISDM, HFR, MSCI, and Tass databases and use the 3924 hedge funds in operation at the end of 2002 to document the disparity between different data sources.
For example, 27% of the funds are exclusively included in CISDM (23% in Tass, 20% in HFR) and a mere 3% of the hedge funds are included in all four databases. the low correlation between the HFR Composite Index and the CSFB/ Tremont Composite Index over the period 1994–2002 of 0.76 as reported by Fung and Hsieh (2004) illustrates the heterogeneity further.
In fact, Fung and Hsieh (2002) calculate a mean difference for annualized monthly returns of 1.5% between the HFR Performance Index and the CSFD/ Tremont Hedge Fund Index over the period 1994–1999 with even larger discrepancies on an annual basis. the difference (HFR – CSFD/Tremont) is 9.5% in 1994 and −9.1% in 1997.
They attribute a substantial fraction of these discrepancies to the weighting schemes used in computing monthly averages, that is, equally weighted versus value-weighted. Amenc and Martellini (2003) report that monthly returns of major hedge fund indices that are expected to represent the same investment style diverge substantially.
The monthly, nonannualized returns for a specific month differ by more than 20% for the relatively well-defined category long/short (Zurich Capital Markets and Evaluation Associates Capital Markets (EACM) indices).
They also report that the average pair-wise correlations among the ten index providers they study tend to be weak for nondirectional strategies; for example 0.43 for market neutral indices, 0.46 for long/short, or 0.54 for fixed-income arbitrage.
The lowest correlation is as low as −0.19 for the style long/short. On the other hand, well-defined strategies like merger arbitrage exhibit the highest homogeneity. The average correlation between all indices is 0.92 and the pair-wise correlation does not drop below 0.88.
Finally, peer group based style factors are typically not investable as they cannot be specified in advance, include closed funds, and equal weighting is not feasible due to minimal capital requirements or lockup periods.
The paper by Fung and Hsieh (2004) contains an extensive critique of peer group based style indices. Nevertheless, due to the lack of a generally accepted alternative, peer group based style factors are a popular tool to monitor the performance of hedge funds.
|Peer Group Based Style Factors|