Forecasting Stock Volatility Of Chinese CSI300 Companies With HAR-type Models

For a long time,forecasting the volatility of financial assets is a hot area in academia because volatility is applied in many aspects of the financial sector.The study of volatility involves the definition of volatility expression,the construction of capture model,and the prediction of future volatility.It has important theoretical and practical significance.With the increasing convenience of high-frequency data acquisition and processing,the academic community has put forward higher requirements for volatility prediction.In order to better predict volatility based on high frequency data,the realized volatility(RV)was proposed by Andersen and Bollerslev.The advantage of RV is that even if the closing price of the adjacent trading day is close,the realized volatility can still contain more intraday trading information in the market.RV has better unbiasedness and less market noise.The HAR-type model proposed by Corsi is verified to have better prediction of high-frequency data volatility than the traditional prediction model.Subsequent HAR-type model studies again considered the leverage effect and introduced different jump variables.The Hawkes process jump component has become a hot topic in recent years for the HAR-type model.The Hawkes process can abstract time series data into a point process and identify the intensity of the process.The model has been successfully applied in the field of high frequency financial data.This paper will use the HAR-type model as the benchmark model for predictive empirical analysis,and introduce the Hawkes process jump components to apply to the volatility forecast of the constituent stocks of the Chinese stock market index.We try to improve the accuracy of volatility prediction,and try to explain the principle of the introduced variables from the perspective of market investors’ performance,and provide a new research perspective for volatility prediction based on high frequency data.In the past literature,the advantage of predicting the volatility of a model may often stay at the level of measurement and statistics,and has not yet made clear application and contribution to the actual market.Therefore,this paper not only verifies the HAR-type models after the variables introduced in this paper from the measurement point of view,but also tries to apply the research results to the construction of stock market trading strategies,which brings economic benefits to investors.