基于Copula和极值理论的在险价值度量  

VaR Measure Based on Copula and Extreme Value Theory

作　　者：冯烽[1,2] 

机构地区：[1]广西财经学院信息与统计学院,广西南宁530003 [2]福州大学管理学院,福建福州350002

出　　处：《数学的实践与认识》2011年第15期97-107,共11页Mathematics in Practice and Theory

基　　金：广西教育厅资助项目(200802LX233);广西财经学院资助项目(2011A01)

摘　　要：针对传统孤立使用GJR模型、极值理论、Copula理论进行风险分析的不足,把GJR模型、极值理论和Copula理论有机的结合起来,给出了基于Copula和极值理论的投资组合VaR的测度方法.首先利用GJR模型刻画单个资产收益率中的自相关和异方差现象,获得近似独立同分布的新息序列,再分别应用高斯核估计的方法、极值理论拟合新息序列的分布函数的内部和两尾,利用Copula函数有效捕抓了市场之间的波动溢出效应,最后使用Monte Carlo模拟法,计算出投资组合的VaR值.实证结果表明,基于Copula和极值理论的VaR度量方法比历史模拟法更有效.Instead of using G JR Model, Extreme Value Theory or Copula Theory alone, G JR Model, Extreme Value Theory are combined with Copula Theory for providing a measure method of the portfolio VaR based on the Copula Theory and Extreme Value Theory. In first, approximate i.i.d, innovations sequence is obtained by using G JR model to fit the autocorrelation and heteroskedasticity of return series of single asset. Secondly, Estimating the distribution function of the innovations sequence by fitting the internal and two tails of the distribution function of the innovations sequence using Gaussian kernel estimation method and Extreme Value Theory especially. Thirdly, the Copula function be use to capture volatility spillover effects between markets effectively. Finally, the portfolio VaR can be calculating by Monte Carlo simulation. The empirically study show that the VaR measure method based on Copula and Extreme Value Theory is more effective than History Simulation Method.

关 键 词：GJR模型 极值理论 COPULA函数 在险价值 

分 类 号：O211.67[理学—概率论与数理统计]