Robust CAPM Estimation through Cross Validation

Abstract

Limitations of the capital asset pricing model (CAPM) continue to present inconsistent empirical results despite CAPM’s firm mathematical foundations provided in numerous studies. In this article, the authors examine how estimation errors of the CAPM can be minimized using the cross-validation technique, a concept that is widely applied in machine learning (ML-CAPM). They apply their approach to test the assumption that the CAPM is a well-diversified portfolio model by using data from the S&P 500 and the Dow Jones Industrial Average. The results from the ML-CAPM validate both market indexes as well diversified, with statistically insignificant variation in unsystematic risks during and after the 2007 financial crisis. Furthermore, the ML-CAPM provides smaller root mean square error and mean absolute deviations compared to the traditional CAPM.