Hedge Fund Performance, Classification With Machine Learning, and Managerial Implications

Abstract

Prior academic research on hedge funds focuses predominately on fund strategies in relation to market timing, stock picking, and performance persistence, among others. However, the hedge fund industry lacks a universal classification scheme for strategies, leading to subjective fund classifications and inaccurate expectations of hedge fund performance. We use machine learning techniques to address this issue. First, it examines whether the reported fund strategies are consistent with their performance. Second, it examines the potential impact of hedge fund classification on managerial decision making. Our results suggest that for most reported strategies there is no alignment with fund performance. Classification matters in terms of abnormal returns and risk exposures, although the market factor remains consistently the most important exposure for most clusters and strategies. An important policy implication of our study is that the classification of hedge funds affects asset and portfolio allocation decisions, and the construction of the benchmarks against which performance is judged.

Keywords: Hedge funds, classification, machine learning, alternative investments

JEL Classification: C38, C58, G23