The Research On Fund Risk Measurement Of Time-varying Copula Model Based On MCMC Method

In order to accurately measure the risk of the financial random sequence,this thesis uses the ARMA model and the GARCH model to measure the single fund yield sequence the characteristics of the financial time series higher peak and fat tail and the volatility cluster.This paper estimates the model parameters based on the ML and MCMC method based on the Bayesian theory and compares the model parameters respectively.The corresponding AIC values of each model finally find that the MA(1)-GARCH(1,1)model based on the skewed t distribution is relatively optimal,and further combines the time-varying Copula model to measure the VaR of the fund portfolio to test the feasibility of the model through the back test.In this thesis,we first analyze the four open-end fund yield sequence,and find that each sequence has the volatility cluster and a high kurtosis value.It is suitable for modeling with GARCH model.In order to avoid the error caused by the unreasonable model hypothesis,four mean equation and five variance equations are modeled respectively.Using ML and MCMC method to estimate the model parameters and estimating the VaR values of each fund yield sequence,the conclusion is that MCMC method is superior to the ML method.Secondly,in order to describe the mutually dependent structure between each yield sequence,the double correlation relationship between the sequences is modeled by the time-varying Copula model,finally found that the time-varying Sjc-Copula function is optimal;then,taking the number of historical winners of each fund's return sequence as a priori information,using the M-H algorithm to simulate the portfolio ratio of the fund.Finally,using the fund portfolio ratio and the time varing Copula function to estimate the VaR of the portfolio and testing the VaR of portfolio by the common test.The results of the estimation are tested by the effective Kupiec method.The empirical study shows that the time-varying Copula model based on Bayesian theory does achieve very effective results in the measurement of the open-end fund portfolio VaR.The innovation of this paper is mainly reflected in the model enough to take full account of the higher peak and fat tail and volatility cluster of the single fund income sequence,and effectively measure the time variant dependence and the VaR value of the fund portfolio.