Heterogeneous Response: An Extension of the Fama-MacBeth Regression
43 Pages Posted: 23 Mar 2022
Date Written: January 17, 2022
We propose an extension of the Fama-MacBeth regression by allowing stock returns to respond differently to firm characteristics. This heterogeneous model can potentially capture non-linearity and interactions. Empirical, applying it to a common set of fifteen firm characteristics, we find that the value-weighted long-short portfolio has an annualized Sharpe ratio of 0.97, doubling that of the usual homogenous model, and our model is not only easier to understand, but also performs better than existing machine learning models. We also propose a test detecting which risk exhibits heterogeneous reactions, and find that heterogeneity is significant for firm size, momentum, and stock volatility. Furthermore, we find that heterogeneous reactions are more pronounced during recession periods.
Keywords: Fama-MacBeth approach, cross-section of stock returns, long-short portfolio, out-of-sample prediction
JEL Classification: C53, C55, C58, G11, G12, G17
Suggested Citation: Suggested Citation