Does Unobservable Heterogeneity Matter for Portfolio-Based Asset Pricing Tests?
54 Pages Posted: 9 Apr 2020
Date Written: March 24, 2020
We show that portfolio sorts, as widely used in empirical asset pricing, tend to misattribute cross-sectional return predictability to the firm characteristic underlying the sort. Such misattribution arises if the sorting variable correlates with a firm-specific effect capturing unobservable heterogeneity across firms. We propose a new, firm-level regression approach that can reproduce the results from standard portfolio sorts. Besides, our method handles multivariate firm characteristics and, if firm fixed effects are included, is robust to misattributing cross-sectional return predictability. Our empirical results confirm that portfolio sorts have limited power in detecting abnormal returns: Several characteristics-based factors lose their predictive power when we control for unobservable heterogeneity across firms.
Keywords: Portfolio sorts, cross-section of expected returns, tests of asset pricing models, random effects assumption
JEL Classification: C21, G14, D1
Suggested Citation: Suggested Citation