Abstract:

Modelling and forecasting volatility has attracted a great deal of attention in financial econometrics literature due to the fact that volatility plays an important role in many financial applications， such as portfolio allocation， risk measurement and derivative pricing. It is well known that volatility is time-varying and highly persistent， and many models have been proposed to capture these stylized facts. The stochastic volatility （SV） model is among the most popular model. However， the standard SV model is a single factor model， which ignores important information contained in the high-frequency data and macroeconomic variables.

An alternative approach for modelling volatility is the realized stochastic volatility （RSV） model， which extends the SV model by incorporating the realized volatility measures and produces more accurate volatility forecasts than the SV model. Despite the empirical success of the RSV model， it still fails to capture the impact of macroeconomic variables on stock market volatility. In recent years， the level of economic policy uncertainty （EPU） keeps rising due to a series of major events， such as the Sino-US trade war， the COVID-19 pandemic and the Russia-Ukraine conflict. Chinese stock market is an emerging stock market， which is affected greatly by policy. As a consequence， it can be argued that Chinese stock market is closely related to EPU. The existing literature on the impact of EPU on the stock market volatility is extensive and has not reached consistent conclusions. Moreover， the existing research is mainly based on the GARCH-MIDAS model， which lacks flexibility compared to the SV framework.

Motivated by the above interpretation， the RSV-MIDAS model framework is proposed， which combines the insights of the SV-MIDAS model and the RSV model. The proposed framework exploits the high-frequency intraday information and allows to link macroeconomic variables （such as the EPU） directly to the long-term volatility via the flexible MIDAS structure. By incorporating the EPU into the framework， it aims to investigate the impact and predictive value of the EPU on Chinese stock market volatility. The model is flexible， but has challenges in estimation owing to the lack of a closed-form expression for the likelihood function. To address this issue， a continuous particle filters-based maximum likelihood method is proposed. Monte Carlo simulation results show that the estimation method performs well.

The RSV-MIDAS model incorporating the EPU （hereafter the RSV-MIDAS-EPU model） is applied to the monthly Chinese EPU index and the high-frequency intraday data of the Shanghai Stock Exchange Composite Index and Shenzhen Stock Exchange Component Index. The empirical results show that the EPU has a significantly negative impact on the long-term volatility of Chinese stock market. That is， an increase in the EPU level predicts lower level of the long-run volatility of Chinese stock market. The impact of the EPU on the long-run volatility of Chinese stock market is more persistent than that of the monthly realized volatility （RV）. Using various loss functions and the model confidence set （MCS） test， the out-of-sample forecasting ability of the RSV-MIDAS-EPU model and the competitor models for Chinese stock market volatility are compared. It is observed that the realized measure and EPU play an important role in forecasting Chinese stock market volatility， and the RSV-MIDAS-EPU model achieves the best forecasting performance. Further， extensive robustness analysis shows that the superior predictive ability of the RSV-MIDAS-EPU model is robust. Finally， a volatility timing strategy shows that the RSV-MIDAS-EPU model yields more significant economic value of portfolio compared to other models.

Key words: economic policy uncertainty, volatility forecasting, realized SV-MIDAS, continuous particle filters, volatility timing