藤Copula模型在我国股市多资产组合VaR预测中的应用

APPLICATION OF VINE-COPULA MODEL IN VAR FORECASTING OF MULTI-ASSET PORTFOLIO IN CHINA’S STOCK MARKET

选取我国六只股指数据为研究对象,基于几类藤Copula模型和VaR理论,运用滚动MonteCarlo技术及MST-PRIM算法确定各类模型的RVM结构,并在此基础上结合滚动时间窗口方法预测投资组合的动态VaR。为了进一步验证模型的拟合效果,采用返回值检验法测试模型的VaR预测结果。事实证明:在等权重与Mean-CVaR约束条件下,C-藤和R-藤对投资组合的VaR预测效果胜于D-藤;另外,R-Vine all Gumbel模型的VaR预测结果比R-Vine all t模型好,充分说明在面对非对称特点的金融数据时,Gumbel-Copula比t-Copula更具说服力。

Six stock markets indexes data of China are selected as the research object.Based on several types of Vine Copula model and VaR theory,rolling Monte Carlo technology and MST-PRIM algorithm are used to determine the RVM structure of various models.On this basis,combined with rolling time window method,the dynamic VaR of portfolio is predicted.In order to further verify the fitting effect of the model,the Kupiec’s test method is used to test the model VaR prediction.As a result,it has been proved that under equal weight and Mean-CVaR constraints,C-Vine and R-Vine have better predictive effect on portfolio VaR than D-Vine.In addition,the predictive result of R-Vine all Gumbel model is better than the predictive results of R-Vine all t model,which illustrates that Gumbel-Copula is more convincing than t-Copula for asymmetric financial data.