Robust Estimation of Hedge Fund Performance

Abstract

Returns of hedge funds generally exhibit non-normality. It is well documented that if asset returns have systematic skewness, expected returns should include rewards for accepting this risk. This skewness risk premium should be controlled for in any estimate of performance. To investigate this issue we specify the Residual Augmented Least Squares (RALS) estimator, designed to exploit non-normality in a time series’ distribution. Specifying a linear factor model, we provide robust estimates of hedge fund performance, demonstrating the increase in efficiency of RALS relative to OLS estimation. Our evidence suggests that measuring performance using OLS alphas is inefficient, understating the performance of some hedge funds and overstating the performance of others. We then examine the source of the OLS mispricing. We find that the level of mispricing is positively related to estimates of skewness in the historical fund returns. We conclude that when estimated by OLS the performance of managers who pursue a strategy exhibiting positive skewness is understated and those whose strategy exhibits negative skewness is overstated. Estimation by RALS overcomes this. Our findings are robust to the biases in hedge fund databases.