Searching the Factor Zoo

Abstract

We propose a Bayesian variable selection method to explore the space of possible factor models for alarge set of candidate factors identified in the asset pricing literature. Using thousands of individualstocks, we identify several parsimonious models which perform at least as well, and in some cases betterthan, widely used factor models such as those proposed by Fama and French (2015) and Hou, Xue andZhang (2015). We find that, in addition to the market return, factors that reflect investors’ behavioralbiases, market microstructure, and production-side risk, also matter to explain stock returns. The searchfor a single reduced-form model has limits in practice, as many combinations of characteristics-basedfactors achieve similar empirical performance.

Keywords: Multi-factor model, Factor zoo, Factor selection, Bayesian variable selection, SeeminglyUnrelated Regressions

JEL Classification: G12, C52