Forecasting China's stock market variance

Abstract

We conduct a comprehensive study on the prediction of China's stock market variance using 24 commonly used forecasting variables over the 1995 to 2018 period. The empirical evidence from both the formal variable selection procedure, LASSO, and the out-of-sample test indicates that lagged stock market variance, the Pástor and Stambaugh (2003) illiquidity measure, and aggregate turnover have statistically significant predictive power, while the other variables contain little additional information about future market variance. In contrast with the conventional wisdom, we document an unstable relation between the scaled stock market price and market variance.