Nonlinear Factor Attribution

Abstract

Factor attribution based on linear regression often fails to satisfactorily explain the performance of systematic investment strategies. Sizeable attribution residuals that do not average out to zero over time suggest latent exposures to nonlinearities in factor returns. Our proposed adjustment takes a portfolio manager’s perspective in attributing the impact thereof, identifying which factor tilts were most responsible for the unexplained performance. The resulting nonlinear attribution better reconciles realized returns with the investment process and is testable for statistical significance, with R code provided for evaluation purposes. We illustrate how this deeper understanding of factor interactions may guide client discussions and point to strategy enhancements.

Keywords: Factor Attribution, Factor Tilts

JEL Classification: G10, G11