|
Persistence in hedge fund performance: the true value of a track record Authors: H. M. Kat and F. Menexe Source: Journal of Alternative Investments Date: Spring 2003 |
The fact that many allocations to hedge funds are based on their track record implies that investors believe that performance persists.
This study, which considers two main aspects, is interesting. Firstly, the authors distinguish between two performance persistence approaches: the persistence in the risk-adjusted returns and the persistence in the risk profile. Kat and Menexe base their investigations on the second approach. Secondly, they distinguish between "persistence" and "predictability": the persistence of a risk profile does not imply that it can be predicted.
The database used, provided by Tremont TASS, covers the period from June 1994 to May 2001. 324 hedge funds are classified into 6 categories, according to their style: long/short equity (113 funds), event driven (59), global macro (17), emerging markets (26), relative value (31) and funds of funds (78).
The notion of "risk profile" corresponds to the second (standard deviation), third (skewness) and fourth (kurtosis) orders of a return distribution (the mean is order 1). Correlation with stocks (S&P 500 Index) and bonds (Salomon Brothers seven-year Government Bond index) are also computed.
In order to test the persistence of the risk profile, two methods are used: contingency tables associated with a Cross-Product Ratio (CPR), and a regression analysis. The initial period is separated into two sub-periods, from June 1994 to November 1997, and from December 1997 to May 2001. A hedge fund persists, as a winner or loser, if it is above or below the median of the considered parameters in the two sub-periods.
In terms of risk profile parameters, the CPR and regression analysis lead to a common conclusion: standard deviation and correlation with stocks show the strongest persistence.
For the mean, considering the CPR, no category exhibits a statistically significant persistence with a 5% significance level. Considering the regression analysis, only funds of funds and emerging markets show significant persistence.
For the standard deviation, considering the CPR, strong and significant persistence is found for each of the 6 categories. The regression analysis gives the same conclusion.
For the skewness, considering the CPR, only long/short equity shows significant persistence. Considering the regression analysis, only long/short equity and funds of funds have significant persistence.
For the kurtosis, considering the CPR, there are no categories with significant persistence. According to the regression analysis, only funds of funds have significant persistence.
For the correlation with stocks, considering the CPR, long/short equity, funds of funds and relative value funds have significant persistence. The regression analysis finds significant persistence for all 6 categories.
For the correlation with bonds, considering the CPR, only long/short equity has significant persistence. The regression analysis also finds significant persistence for funds of funds.
Nevertheless, Kat and Menexe suggest that the small size of the sample precludes an accurate measure of skewness and kurtosis.
The fact that some risk profile parameters persist does not imply that those parameters are predictable. Indeed, a risk profile can be predicted only if the median is predictable. The median allows the hedge funds to be sorted in the contingency tables and in the regression analysis.
Estimating biases in the predictors (i.e. the average value of each risk profile parameter in the first period for each strategy, compared to the average value in the second period) and estimating their accuracy reveals that a forecast cannot be made for each risk profile parameter. If the average values of the first period are used, this leads to an overestimation of the return mean and an underestimation of the standard deviation and the kurtosis in the second period.
