The SZSE Component Index Volatility Analysis And Risk Prediction

In financial markets,volatility plays an important role in Financial asset pricing and risk management-The 5-minute high-frequency trading data of Shenzhen Component Index is taken as the research object,and the Realized GARCH model is established to estimate and predict the volatility and VaR of Shenzhen Stock Index.The fit data for the volatility model uses 5-minute high-frequency trading data for each trading day from January 4,2012 to November 23,2018.According to the fluctuation of the daily yield of Shenzhen Stock Exchange,the improved Realized GARCH model is constructed.Through the comparison of the fitting effects of the volatility of each model,the model that can better fit the volatility of the Shenzhen Stock Index and the volatility is selected.Make predictions.At the same time,the risk value model(VaR)will be used to conduct a risk analysis of the Shenzhen Stock Exchange and predict the downside risks of the future stock market.In terms of volatility estimation,the main three factors affect the volatility estimation,including RV,RRV,RBV three implemented measures,Skewed-t distribution,t distribution,and the distribution of three residual distributions of normal distribution.Whether to consider the long memory of volatility to introduce the ARFIMA model,compare and select the model with the best fitting effect,and predict the 500 out-of-sample data based on the forward one-step prediction.For the calculation of VaR,the rolling window one-step prediction method is used to predict and draw the curve for the future VaR.At the same time,the Kuipice test method is used to test the prediction results of each model and compare the results to select the model with the best prediction effect.The first is to model and analyze the stock market volatility.The empirical results show that the SZSE's yield series has the characteristics of sharp peak and thick tail and skewed distribution.The different implemented measures,different residual distributions,and whether or not to join the ARFIMA model are analyzed separately.It is found by comparison of the fitting effects:(1)The mean value equation adopts the ARFIMA model considering long memory and the fitting effect is slightly improved;2)The model has the best fitting effect when the input variable is RRV,and RBV is the second;(3)Compared with the t-distribution model,the log-likelihood function is greatly improved compared with the t-distribution model.Compared with the t-distribution model,the Skewed-t distribution model also slightly increases the log-likelihood function value,indicating that the Skewed-t distribution fits the residual distribution of the model better than the t-distribution and the normal distribution.Followed by VaR prediction,the corresponding Realized GARCH model was constructed by previous analysis and the rolling window prediction was performed on VaR.The empirical results show that the predicted effects of RV on VaR are significantly improved in RBV and RRV,while RRV and RBV have similar performance in Realized GARCH model,and RBV may be slightly better than RRV.For the distribution of model residuals,the prediction effect of the Skewed-t distribution model is not significantly improved compared to the t-distribution model.Because the distribution of volatility has a sharp tail tail,the Skewed-t model with skewed and sharp tails can better predict the extreme risk value,while the Realized GARCH model based on normal distribution has the lowest prediction effect..