Study On Risk Measurement Of Chinese Stock Market Based On Range GARCH-MIDAS Model

The macro economy can be reflected by its stock market.The fluctuation of the stock market is closely related to the development of the Chinese economy.Especially,under the era of ‘Internet Finance',as the world economy and financial openness increasing,and the innovation of financial products accelerating,any changes in the political and economic environment from both domestic and abroad will cause fluctuations in the stock market,causing the risk of the stock market increasing and complicated.As in the second half of 2015,Chinese stock market has experienced a sharp rise and fall.Nearly a thousand stocks have fallen and stopped,which has damaged investors' wealth and confidence,and investors' sentiment has been depressed for a time.Therefore,accurate measurement of stock market risk is of great significance to the risk management of Chinese stock market as well as the correct decision-making of investors.With the development of information technology and the maturity of high frequency data acquisition technology,the GARCH model,which based on mixing data(GARCHMIDAS),has been gradually applied to risk measurement,and achieved good results.The model divides the conditional variance into long-term and short-term components.However,the long-term component of the traditional GARCH-MIDAS model only considers the realized volatility.Thus causing the loss of information due to lacking of consideration of the influence of multiple measurement.Therefore,this paper introduces price range into the long-term components of the model,establishing the Range GARCH-MIDAS model,so the market information can be fully utilized.In this paper,daily closing price,daily maximum price,daily minimum price,and the realized volatility data which based on 5-minute high-frequency data of Shanghai Stock Exchange Composite are used as the sample data for empirical research.Based on GARCH model,GARCH-MIDAS model and Range GARCH-MIDAS model,the rolling time windows method is used to make out-of-sample prediction of volatility.Then,four loss functions(RMSE,MAE,MAPE,QLIKE)and M-Z regression tests are used to compare the precision difference of volatility prediction of the three volatility models.VaR calculation and Kupiec testing are further used based on the prediction results of volatility to compare the precision difference of VaR prediction which based on difference volatility models.The empirical results based on the Shanghai Stock Exchange Composite suggest that the results of parameter estimation,volatility prediction,and VaR calculation of different volatility models turn out to be quite different.It can be seen from the parameter estimation results that the introduction of realized measures(realized volatility and price range)can disperse the dependence of the conditional variance of the model on the first-order term of its own lag.According to the results of volatility prediction and the check of loss function,the Range GARCH-MIDAS model has the highest volatility prediction precision,followed by GARCH-MIDAS model and GARCH model.The VaR calculation and test results show that whether the confidence level is at 95%?97.5% or 99%,the VaR of the Range GARCHMIDAS model has the highest prediction accuracy.The result clearly shows that the introduction of the price range can indeed improve the volatility prediction ability of GARCH-MIDAS model as well as the precision of VaR prediction,which have profound practical implication in risk management.