Correcting Asset Pricing Models

Abstract

Equilibrium asset pricing models prescribe a correspondence between assets' risk exposures and premiums. Empirical factor models do not, however, satisfy this relationship. We show that a portfolio sorted on a multi-factor model's alphas is the optimal correction to this problem. This correction factor is a generalization of the betting-against-beta (BAB) factor. Whereas the BAB factor adjusts for the flatness in the security market line, the correction factor simultaneously adjusts for all such distortions. Augmenting the Fama-French five-factor model with its correction factor increases the model's out-of-sample Sharpe ratio by 40%, increases its power to explain the cross section of returns, and lowers its errors in pricing 208 anomalies.

Keywords: Beta, Betting against beta, Security market line, Factor model, Mean-variance efficiency

JEL Classification: G11, G12, G40