The selection of individual managers is a key step in hedge fund investing. The risk and return dispersion in the same strategy highlights the heterogeneity among managers, to differing degrees among strategies. Consequently, strategy factor risk is not a sufficient basis to construct a portfolio of hedge funds. Moreover, a simple diversification of fund-specific risk via an increase in the number of funds can engender an increase in the exposure to market risk factors.

Instead of diversifying fund-specific risk, the authors advocate reducing it through manager selection. Here the authors focus on quantitative features, namely performance persistence, due diligence and ongoing monitoring.

The issue of performance persistence can be examined through two approaches. The first approach consists of measuring the persistence of relative returns (the returns of the individual funds are compared to a median return). In this study it is implemented through significance tests applied to a cross product ratio and regressions. Return, standard deviation and Sharpe ratio persistence is tested on seven strategies: convertible arbitrage, distressed securities, merger arbitrage, fixed income arbitrage, equity market neutral, equity long/short and global macro. The period is from January 1997 to December 2002. The sample contains 314 hedge funds provided by HFR. The two methods lead to the same conclusions. A lack of significant return and Sharpe ratio persistence is observed on all seven strategies, except for the tests on regressions, which exhibit significant Sharpe ratio persistence for two strategies. Conversely, standard deviation displays significant persistence, but this can be mitigated by the fact that the median is the median return for all strategies. That is why in an additional test, the returns of each fund are compared to a strategy index return. It gives similar results.

The second approach measures the persistence of individual returns directly, without a comparison to a median. Here it is done through the Hurst coefficient. An advantage of the Hurst exponent is that its efficiency is not related to an assumption on the return distribution. A Hurst exponent comprised between 0 and 0.5 indicates reverse persistence. An exponent of 0.5 indicates random performance. An exponent comprised between 0.5 and 1 indicates positive persistence. The period is divided into two sub-periods of three years. The first period is considered as the in-sample period, and the second period is considered as the out-of-sample period. The funds are divided into three groups according to the level of the Hurst exponent exhibited by each fund. To determine the weights of each manager in the groups, two methods are used. The simpler one consists of equal weighting. The second method introduces the notion of risk budgeting. Risk budgeting is "an asset allocation technique where […] capital is allocated to risk buckets with no consideration of associated returns." Here the authors use a quantitative algorithm.

The “low Hurst” group contains 105 managers, where exponents range from 0.32 to 0.58. The “medium Hurst” group contains 105 managers, where exponents range from 0.59 to 0.69. The “high Hurst” group contains 104 managers, where exponents range from 0.70 to 0.98. Except for the distressed securities strategy, which does not appear in the low Hurst group, all the strategies are represented in each group. Using an equal-weighting scheme, during the in-sample period returns, standard deviations and Sharpe ratios do not differ significantly among Hurst groups. During the out-of-sample period, the high Hurst group displays the highest rate of return, the lowest volatility (thus automatically the highest Sharpe ratio), the highest Calmar ratio and the highest number of months with consecutive gains. Using the risk budgeting approach, the high Hurst portfolio presents the best statistics too. In other words, persistent managers outperform non-persistent managers during the out-of-sample period. On the other hand, it is stated that risk budgeting permits the characteristics of the portfolio to be improved (in comparison with an equally-weighted portfolio).

Nevertheless, a high Hurst exponent does not indicate whether it is negative returns or positive returns that persist. Therefore a D-Statistic {1} is calculated, for the managers included in the high Hurst group only. Ranging from 0 to 1, the lower the D-statistic, the more favorable it is. The 104 managers of the high Hurst group are ranked into three portfolios according to their D-statistic calculated during the in-sample period. The equal-weighting scheme and risk budgeting are successively used to construct the three portfolios. During both periods, using either equally-weighted portfolios or a risk budgeting approach, the low D-statistic portfolio exhibits the lowest standard deviation, the highest Sharpe ratio, the highest Calmar ratio, and the highest number of months with consecutive gains. Moreover, it is confirmed that risk budgeting permits the characteristics of the portfolio to be improved in comparison with an equally-weighted portfolio.

Due diligence is defined by the authors as the process by which the operational risk is evaluated both qualitatively and quantitatively. In this paper it is tackled from the angle of the viability of the individual hedge funds. Firstly, 747 dead funds are compared to 2,298 active funds. In particular average assets managed, age, lockup period and average returns are significantly higher for the surviving funds as of December 2002.

Ongoing monitoring is the last step in the manager selection process. It corresponds to a complete analysis of risk. The authors use the Omega introduced by Keating and Shadwick (2002). They show that it permits changes in a manager’s risk profile and performance to be determined.

Footnotes

{1} D-statistic= sum lnegative returnsl / sum lall returnsl