Money on the Table: Testing Asset Pricing Models from Profitability of Pricing Error Information

Abstract

Based on the intuition that model pricing errors cannot generate any tradable profits if the model is true, we propose an asset pricing test and apply it to both six mainstream factor models and improved models recently developed with advanced machine learning tools. We find that all the models are rejected and their pricing errors share similar patterns that lead to significant trading profits, which cannot be explained by investor sentiment, limits-to-arbitrage, prospect theory, and expectation extrapolation. Our findings suggest that there is still a long way to go toward modeling well the cross section of stock returns.

Songrun He

Ph.D. Student in Finance

I am a Ph.D. student in finance at WUSTL with an interest in asset pricing, investment strategies, asset management, machine learning and deep learning in finance, textual analysis and high-frequency finance.