Risk Measurement And Prediction Of Chinese Stock Market Based On Realized Volatility Method

With the rapid development of financial market,financial risk has become the focus of financial market investors,regulatory authorities and financial economic participants,because the profit and loss of financial assets are closely related to financial risk.The measurement of financial risk is related to volatility.If the price of financial assets fluctuates more,the volatility will be higher,and the rate of return on financial assets will be more unstable,and the risk will be greater.It can be seen that the measurement of volatility affects the estimate of risk.Parametric method based on volatility model is the most widely used volatility measurement method at present.This method includes volatility in the model and uses parametric model to estimate volatility.For volatility modeling,Engle proposed the autoregressive conditional heteroskedasticity(ARCH)model,but it has not been widely used due to its great defects.Bollerslev proposed a generalized autoregressive conditional heteroskedasticity(GARCH)model on the basis of ARCH model,which has been widely used.Scholars continue to expand the GARCH model and propose various GARCH models based on the GARCH model.For low frequency data,GARCH model can be used for volatility modeling.With the rapid development of computer technology,it becomes easy to obtain high frequency data.Based on high frequency data,Andersen et al.proposed Realized Volatility(RV)method to estimate volatility.Realized measures based on high frequency data models used in volatility modeling have the Realized GARCH model.Since the realized volatility method has many advantages,this paper will measure and forecast the risks of China's stock market based on the realized volatility method.Financial risk measurement has always been the core problem of financial mark et risk management.This paper selects Shanghai Composite Index,Shenzhen Compo nent Index and Hengsheng Index of Shanghai,Shenzhen and Hong Kong stock market s as the research objects for empirical analysis.The obtained data were processed,a nd descriptive statistical analysis and stationarity test were conducted.But the Skewe d Skewed Skewed Skewed model uses five-minute high-frequency data to model the Realized volatility of the Skewed Skewed distribution.The GARCH model based on no rmal distribution is modeled with daily low frequency data.The maximum likelihood estimation method is used to estimate the model parameters at one time,and the pa rameter estimates of each model are obtained.Then,the Kupiec failure rate test met hod was used to test the estimated VaR value,and the test results were obtained.Th e degree of fit and the accuracy of VaR prediction of each model were evaluated,and the conclusion was drawn.The empirical analysis results show that the return series of the three stock inde xes all have the characteristics of sharp peak,thick tail and skewness.Using high-freq uency data for modeling can make full use of intraday trading data and improve the a ccuracy of estimation,while using low-frequency data will cause information waste a nd accuracy will be reduced.With the combination of Realized measures and conditi onal volatility,the Realized GARCH model is superior to the traditional GARCH model.But compared with the GARCH model based on the normal distribution,the Skewed model has better fitting effect and higher VaR prediction accuracy.