Replicating and Digesting Anomalies in the Chinese A-share Market

We replicate 469 anomaly variables similar to those studied by Hou, Xue, and Zhang (2020) using Chinese A-share data and a reliable testing procedure with mainboard breakpoints and value-weighted returns. We find that 83.37% of the anomaly variables do not generate significant high-minus-low quintile raw return spreads. Further adjusting risk increases the failure rate slightly to 84.22% based on CAPM alphas and 86.99% based on Fama–French 3-factor alphas. We show that the conventional procedure using all A-share breakpoints with equal-weighted returns for the anomaly test is indeed problematic, as it assigns too much weight to microcaps and has a very limited investment capacity. The CH3-factor, CH4-factor, and q-factor models show the best performance over the whole sample period. The q-factor model is the best performer in the post-2007 subsample period, after significant improvements occurred in China’s financial market environment, such as the completion of the split-share structure reform and the implementation of new accounting standards conforming to the IFRS. The non-SOE subsample in the post-2007 period is a cleaner sample, in which the CH4-factor and q-factor models are the best performers