Hedge Funds: The Good, the (Not-So) Bad, and the Ugly

Abstract:     
This paper proposes a new methodology to evaluate the prevalence of skilled fund managers. We assume that each fund’s alpha is drawn from one of several distributions based on its skill level (e.g., good, neutral, or bad). For a sample of funds, the composite distribution of alpha is thus a mixture of the underlying distributions. We use the Expectation-Maximization algorithm to infer the proportion of funds of different skill levels and estimate the conditional probability each fund is of a skill type given estimated alpha. Applying our approach to hedge funds over 1994–2009, we find that about 50% of funds have positive skill. Funds identified by our approach as superior persistently deliver high out-of-sample alpha over the next three years. While investors chase past performance, inflows do not reduce fund performance in the near future.

Number of Pages in PDF File: 50

Keywords: Fund performance evaluation, hedge funds, mixture distribution, EM algorithm, performance persistence

JEL Classification: C13, G11, G23

working papers series