藤Copula模型与多资产投资组合VaR预测  

Vine Copula Model and VaR Forecast for Multi-Asset Portfolio

作　　者：高江[1] 

机构地区：[1]上海财经大学金融学院,上海200433

出　　处：《数理统计与管理》2013年第2期247-258,共12页Journal of Applied Statistics and Management

基　　金：上海财经大学研究生科研创新基金(CXJJ-2010-335)

摘　　要：投资组合风险管理往往涉及多个资产,在传统的二元Copula函数面临"维度诅咒"问题及多元Copula函数刻画多变量联合分布时其精确性和灵活性存在各种局限性的情况下,引入藤Copula刻画多个资产收益的联合分布,基于不同的Pair-Copula类别构建藤Copula,运用蒙特卡罗模拟方法计算多资产投资组合的VaR,通过Kupiec和Christoffersen返回检验方法测试藤Copula模型的VaR预测效果,并与传统方差-协方差风险管理方法做比较。实证分析表明,传统的方差-协方差风险管理方法和基于正态Pair-Copula作为藤Copula构建模块的方法不能通过多资产投资组合的VaR预测返回检验;而基于student-t Copula、Clayton Copula具有尾部分布特征的Copula作为构建模块的藤Copula模型能够有效地用于多资产投资组合VaR预测,从而更好的用于指导实践。The risk management of portiolio usually revolved m more than two assets,unaer ne conm tion that traditional bivariate Copula faces the problem of ＇dimensional curse＇ and multivariate Copula functions have limitation on precise and flexibility when they are used to characterize the multivariate joint distribution. VineCopula is introduced to characterize multiasset returns＇ joint distribution, based on different PairCopula decomposition, VineCopula is constructed. Then VaR of the portfolio is cal culated by Monte Carlo method and the VaR forecast efficiency of Vine Copula model is backtested through Kupiec and Christoffersen backtesting and compared with traditional variancecovariance risk management method. Conclusion could be obtained that traditional variancecovariance method and Vine Copula based on normal Pair Copula decomposition method can not forecast VaR of multiasset portfolio successfully. The VineCopula based on studentt pair Copula and Clayton Copula which show tail distribution feature can pass the backtesting. The conclusion can be used in guiding practice better.

关 键 词：PAIR-COPULA 藤Copula VAR 投资组合 

分 类 号：F830.9[经济管理—金融学] O212[理学—概率论与数理统计]