Taming China's Factor Zoo with a Bayesian Method

Abstract

In light of the superior advantages offered by the Bayesian Model Averaging (BMA) approach, this paper employs the BMA to analyse the Chinese stock market by constructing the Stochastic Discount Factor (SDF), which prices a comprehensive cross-section of anomalies across the space of 228 quadrillion models. First, the STD factor, known as the idiosyncratic volatility factor, is a significant contributor to priced risk within the SDF, accounting for approximately 60 percent of the posterior factor probability, due to underdevelopment of the Chinese’s financial system. Second, the SMB factor is more likely to be another source of priced risk due to the Chinese’s government intervention and the SOEs. Third, both in-sample and out-of-sample, the BMA SDF model outperforms both other well-known models and the KNS model. Fourth, there is no single model that stands out as the most effective in terms of cross-section and time series. Finally, the SDF is dense in the space of observable factors.

Keywords: Factor Zoo, Bayesian Model Averaging, Chinese Stock Market