Modeling and managing stock market volatility using MRS-MIDAS model

This paper adds the decomposed components of realized volatility to investigate China's stock market volatility based on the mixed data sampling (MIDAS) framework. Empirical results show that considering extreme negative volatility and extreme positive volatility, and moderate volatility can have a significantly better performance than the benchmark model for predicting the Chinese stock market volatility. Importantly, we find that taking the regime switching into consideration can further improve forecasting accuracy. The results are robust in alternative evaluation method, different forecasting windows, the direction-of-change test, and China's stock bubble period, showing decomposing the realized volatility into extreme negative volatility and extreme positive volatility, and moderate volatility can further improve the forecasting accuracy of the models. This paper tries to give new evidence to stock market volatility prediction.