We use heterogeneous autoregression (HAR) and two related HAR extension models to examine volatility forecasting performances before and after the launch of night trading sessions in the Shanghai Futures Exchange (SHFE) gold futures market. To capture fluctuations from external information and volatility of realized volatility (RV), we incorporate the trading volume and jumping into the HAR-V-J model in the first place and then incorporate a GARCH specification into the HAR-GARCH model. Results showed that there were large fluctuations in SHFE gold futures market before the launch of night trading sessions and mostly stemmed from overnight fluctuation in the international gold futures market. After the launch of night trading sessions, the realized volatility has a clear trend of moderation. In the in-sample estimation, both jump and external information are found to have significant explanatory power with the HAR-V-J model. Additionally, the volatility clustering and high persistence of the realized volatility were confirmed by the GARCH coefficients. Last but not the least, night trading sessions have significantly improved the out-of-sample forecasting performances of realized volatility models. Among them, the HAR-V-J model is the best-performing model. This conclusion holds for various prediction horizons and has great practical values for investors and policymakers.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.