Style Drift Analysis of Hedge Funds: with a K-means Clustering Algorithm

We investigate the existence of style drift within the hedge fund industry and examine the relationship between style drift and both the stages of the funds’ lives and the past returns. There are two key contributions made in this study. Firstly, we consider fund risk return profiles directly, rather than classifying funds by their self-described strategies. Secondly, we implement a K-Means clustering algorithm with correlation distance to classify strategy groups, unlike other studies which clustered on qualitative fund attributes. We report a number of interesting empirical findings. Style drift is present in the hedge fund industry, and certain groups are more prone to “drift” than others. Funds at the end of their lives display a significantly higher level of erratic behaviour compared to their behaviours at birth. Finally, poor past performance relative to peers induce funds to change their style more frequently.