基于vine copula方法的股市组合动态VaR测度及预测模型研究

Measurement of dynamic stocks portfolio VaR and its forecasting model based on vine copula

以世界十大股票市场指数为例,运用滚动Monte Carlo模拟技术,实证计算了R-vine、Dvine、C-vine及R-vine all t四种vine copula结构对投资组合的动态VaR预测值,并进一步运用严谨的Back-testing检验方法,实证对比了上述四种vine copula结构对投资组合的VaR预测能力的优劣.实证结果显示:不论是在等权重还是在mean-CVaR约束条件下,R-vine对投资组合的VaR预测效果是最好的.特别在高分位数水平下,其表现得更为突出.另外,D-vine的预测精度总体上要高于C-vine和R-vine all t的,而节点间全为t copula的R-vine all t表现相对较差.

Taking the ten most important stock markets indexes as examples, this paper calculated the predicted VaR value of the portfolio with the four vine copula models, R-vine, D-vine, C-vine and R-vine all t, based on the rolling technique of monte carlo simulation. Furthermore, using rigorous Backtesting method, we compared the performance of forecasting VaR of the four vine copula models. The empirical results show that, the R-vine is the best model in measuring and predicting of the portfolio of the VaR no matter under the condition of equal weight or mean-CVaR. Under high quantile condition, the performance is better. Moreover, the overall prediction accuracy of D-vine is higher than that of C-vine and R-vine all t, however, the performance of R-vine all t is relatively worse when there is only t copula between the nodes.