A study of cross-industry return predictability in the Chinese stock market

We investigate cross-industry return predictability for the Shanghai and Shenzhen stock exchanges, by constructing 6- and 26- industry portfolios. The dominance of retail investors in these markets, in conjunction with the gradual diffusion of information hypothesis provide the theoretical background that allows us to employ machine learning methods to test for cross-industry predictability. We find that Oil, Telecommunications and Finance industry portfolio returns are significant predictors of other industries. Our out-of-sample forecasting exercise shows that the OLS post-LASSO estimation outperforms a variety of benchmarks and a long–short trading strategy generates an average annual excess return of 13%. •We construct industry return portfolios for class A shares on the Chinese stock exchange.•The prominence of retail investors permits cross industry return predictability.•We use machine learning methods to identify the most relevant leading industry portfolio returns.•We find that Oil, Telecommunications, and Finance are leading indicators for other industries.•A long–short trading strategy generates an average annual excess return of 13%.