The Lasso and the Factor Zoo - Expected Returns in the Cross-Section

Abstract

We document that cross-sectional return predictions based on OLS and Lasso type linear methods contain no predictive power for large cap stocks over the last decades. Small and micro cap stocks are highly predictable throughout the entire sample. Based on the 68 firm characteristics (FC) included in our analysis, the variable selection step suggests a highly multi-dimensional return process. Additionally, our Monte Carlo simulations indicate advantages of Lasso type predictions over OLS in panel specifications with a low signal-to-noise ratio. The results are robust to various assumptions.