New evidence on Bayesian tests of global factor pricing models

Abstract

To identify the best global factor pricing model is crucial in international asset pricing literature. This study adopts the Bayesian methods of Chib et al. (2020b) and Chib and Zeng (2020) to estimate and compare 14,322 Gaussian and Student-t distributed global factor pricing models. We find strong evidence that Student-t distributed models significantly outperform Gaussian distributed models in both in-sample and out-of-sample tests. This finding highlights the importance of using the Student-t distributions to model the fat tails in global risk factor data. Analysis reveals that the best global factor pricing model is a Student-t distributed factor model with three degrees of freedom and seven risk factors including the six factors of Fama and French (2018) and the betting against beta (BAB) factor of Frazzini and Pedersen (2014). Our results are robust for different estimation samples and in both relative and absolute pricing performance tests.