Volatility Analysis In CSI 300 Index Based On Composite Quantile Regression Method

In recent years,China’s financial market has made rapid development,there exists instability in the development process,while volatility was usually used for measuring it.Due to its leverage and heterogeneity,the leverage heterogeneous autoregressive(LHAR)models based on high-frequency data was proposed,and the ordinary least squares(OLS)method is commonly used,but it assumes the error distribution strictly and it is susceptible to outliers.Quantile regression(QR)method has improved the OLS method to some extent,it is more robust.Composite quantile regression(CQR)is a method developed from the QR method.It preserves advantages of the QR method and depicts the influence of independent variables on the dependent variable more effectively.Therefore,this paper will use the CQR method to estimate the parameter of LHAR model,in order to capture the trend of volatility better,and hope to obtain the higher economic value.This paper takes the volatility of the CSI 300 Index as the study object,and uses QR method and CQR method to propose the LHAR-QR model and the LHAR-CQR model respectively based on the volatility theory and LHAR model,and carries the empirical analysis about the volatility.Firstly,we use the high frequency data of CSI300 Index to calculate the realized volatility of each time period,and make descriptive statistical analysis of it,and analyze the heterogeneity and leverage of the volatility of each time periods based on the LHAR model.The result shows that the volatility of CSI300 Index has the characteristics of non-normality,daily leverage and weekly memory.Then,the QR method and the CQR method are used to estimate the parameter of the LHAR model,and the LHAR-QR model and the LHAR-CQR model are proposed respectively.The LHAR model,the LHAR-QR model and the LHAR-CQR model are used to forecast the volatility out of the sample.The loss function method,the DieboldMariano test method and the Mincer-Zarnowitz regression method are used to compare the forecast ability of each models.The result shows that the LHAR-CQR model forecasts better,which contains information under each quantiles and it can capture the characteristics of the volatility more accurately.Finally,the Sharp ratio(SR)and Sortino ratio(So R)of these models are calculated via the out-of-sample forecast values of the LHAR model and the LHAR-CQR model to compare the economic value of these models.The result shows that the SR and the So R of the LHAR-CQR model are higher,indicating that the economic value of the LHAR-CQR model is higher,which can enable investors to obtain higher investment returns and provide the advice for financial institutions to improve service quality.