Selection Of Innovation Distribution And VaR Prediction In Copula GARCH Model

Under the background of economic globalization,the rapid circulat ion of capital in the world promotes the correlation between financial markets,which makes the local risk fluctuation in the world affect other regions rapidly.Therefore,risk management has become the focus of relevant practitioners.Va R is a common method to measure the risk degree in the financial market,which is widely studied and used by major investment banks and securities investment companies.Its specific statistical meaning is the maximum economic loss that the asset value may suffer when the financial market operates normally under the given asset holding time period and confidence.Portfolio is an excellent way to disperse asset risk.Based on the volatility model of the portfolio,the risk Va R can be estimated by Monte Carlo method,which can accurately predict the risk situation of the portfolio.The financial time series has the characteristics of fluctuation cluster,time-varying fluctuation and peak thick tail distribution.Based on the nonlinear correlation between the volatility of asset return and portfolio assets,this paper uses Copula-GARCH model to describe the financial time series data.By using copula theorem,we can replace the hypothesis of multivariate normal dis tribution in the classical model,and change it to assume that the new information of the model obeys the multivariate joint distribution based on Copula function,so as to obtain more accurate results.Meta-ellipsoid distribution is a kind of joint distribution composed of several edge distribution and ellipsoid copula function,in which the ellipsoid copula function describes the correlation between multivariate variables.We generally assume that the new information of the model is subject to multivariat e normal distribution,and in this paper,it is extended to obey meta ellipsoid distribution,because the metaellipsoid distribution is more widely used and more suitable for fitting the actual data.In the past few literatures mentioned the test of the n ew distribution of Copula GARCH model.Based on the smooth test and feature test of uniform spherical distribution,the fitting goodness test of new interest distribution is carried out,and a series of sample conversion is carried out for the residual of the model,The fitting goodness test of meta ellipsoid distribution samples is transformed into the fitting goodness test of uniform spherical distribution,and then the model distribution can be tested to fit residual data better.In the empirical analysis,we collect several stock and bond data to obtain two different dimensions of portfolio,and then get the yield data of each asset by logarithmic difference processing;The time series of portfolio components are depicted by one-way GARCH-t model.The fitting goodness test of new interest obeys meta-t distribution is carried out based on the residual data of the model.The risk Va R is estimated by Monte Carlo method.The random simulation results show that the investment mode of portfolio can effectively reduce the risk of loss.