Forecasting volatility of the Chinese stock markets using TVP HAR-type models

Abstract

In this paper, we use a novel Heterogeneous Autoregressive Model (HAR) with time-varying parameters (TVP) to forecast China’s stock market volatility. Many traditional constant coefficient (CC) HAR-type models, incorporating signed variance, jump and volatility leverage effect, are extended to be TVP models. The empirical results show that the extended TVP HAR-type models can beat those CC HAR-type ones in both in-sample estimation and out-of-sample prediction perspective. Moreover, the TVP HAR model that can describe continuous volatility component, signed jump and leverage effect is superior to other CC or TVP HAR-type models in forecasting the volatilities of China’s stock market.