Stock Portfolio Risk Measurement Used The High Frequency Data Based On Realized Garch-vine Copula Model

In recent years,with the deepening of economic globalization,the relationship between the financial markets has become increasingly close,and the linkage between the markets has become complicated and diversified.At the same time,with the frequent outbreak of financial crisis,the relationship between the volatility of return on financial assets and the dependence of various assets has become more and more complex.In the past,scholars usually use low-frequency data to study the volatility of return on financial assets.Although the GARCH Model and SV Model can reflect the volatility of return to a certain extent,the low-frequency data has some limitations and will lose a lot of intra-day volatility information.At the same time,a large number of studies show that the financial time series has the typical characteristics of "peak and thick tail",and there are significant tail correlation and nonlinear relationship between financial assets.It is not accurate to describe the correlation between financial assets with traditional methods.Based on the high-frequency trading data,this paper establishes the Realized GARCH-RV Model,Realized GARCH-RR Model,Realized GARCH-BV Model and HAR-RV Model,and proposes the HAR-RR Model and HAR-BV Model to describe the volatility of the yield of the SSE index.From the perspective of model construction and goodness of fit,Realized GARCH Model has better performance than the model,and the model with Realized Range as volatility measure performs better than Realized Volatility and bipower variation.Secondly,the fitting effect of Realized GARCH Model is better than that of normal distribution,t distribution and GED distribution.In view of the dependence relationship of China Securities Index,this paper establishes three models: C Vine Copula,D Vine Copula and R Vine Copula,and compares the performance of Realized GARCH(1,1)-RR-C Vine Copula Model,Realized GARCH(1,1)-RR-D Vine Copula Model and Realized GARCH(1,1)-RR-R Vine Copula Model from three aspects of model tree structure,goodness of fit and portfolio risk measurement.It is found that R Vine Copula model has better flexibility and better fitting effect.Based on this model,the static analysis of the dependence relationship among the five stock indexes shows that joining the conditional market can effectively reduce the rank correlation coefficient and tail correlation coefficient.In this paper,the rolling window Monte Carlo simulation method is used to predict the Va R and es of the portfolio out of the sample.Through the validity test,it is found that the risk prediction based on the Realized GARCH(1,1)-RR-R Vine Copula Model is the most accurate.It also further shows that the Realized GARCH(1,1)-RR-R Vine Copula Model is the most suitable model to describe the dependence relationship between the yield series of the SSE index.When the return series has extreme volatility,Va R method will underestimate the risk,ES can better reflect the extreme risk.