Volatility forecasting revisited using Markov‐switching with time‐varying probability transition

Abstract

AbstractThis study proposes a novel model, Markov‐switching Heterogeneous Autoregressive (MS‐HAR) model with jump‐driven time‐varying transition probabilities (TVTP), to forecast the future volatility in Chinese stock market. The in‐sample results show that MS‐HAR models are more powerful than HAR‐RV‐type models; furthermore, the high‐volatility regime is short‐lived. Moreover, the out‐of‐sample results indicate that the MS‐HAR with TVTP model can achieve a superior forecasting performance and increase the economic value than the competing models including the simple HAR model and the MS‐HAR with fixed transition probabilities (FTP) model. The results are robust to several robustness checks including alternative forecast window, alternative evaluation method, alternative predictive model, sub‐sample analysis and alternative representative index.