Sharpening the Sharpe Style Analysis with Machine Learning: Evidence of Mutual Fund Style-Shifting Skill

Abstract

We investigate the motivations behind mutual fund investment style adjustments and their consequences within a multi-style framework. Our approach involves a two-step machine learning procedure, integrated with the Sharpe (1992) style analysis, to identify tradable style sets. Our findings show that over 95% of mutual funds adopt multi-investment styles. We further develop a novel method for detecting style shifts and find that mutual funds actively adjust their investment styles. Notably, style-shifting funds not only show a capability to identify superior new styles but also outperform the benchmarks associated with new styles. In essence, our finding suggests that style-shifting fund managers possess both the ability to time style changes and the expertise to manage new styles, thereby supporting the hypothesis of style-shifting skills.

Keywords: Mutual fund style-shifting; Machine learning; LASSO; Sharpe style analysis; Style-timing; Style expertise

JEL Classification: G10, G11, G23