基于VaR相关函数的金融风险度量的实证分析  

Positive analysis of the correlation function based on VaR in financial risk management

作　　者：陈刚[1,2] 袁其志 林金官[1] 

机构地区：[1]东南大学数学系,南京211189 [2]南通职业大学基础课部,江苏南通226007

出　　处：《扬州大学学报（自然科学版）》2014年第4期25-29,共5页Journal of Yangzhou University：Natural Science Edition

基　　金：江苏省自然科学基金资助项目(BK20141326);教育部高等学校博士学科点专项科研基金(博导类)资助课题(20120092110021);江苏省高等教育教学改革研究课题重点项目(2011JSJG085)

摘　　要：以风险价值(value at risk,VaR)为金融风险度量,结合Copula函数及其相关函数建立金融风险模型.考虑到金融时间序列的时变性和厚尾特性,根据GARCH(generalized autoregressive conditional heteroscedasticity)模型和极值理论的POT(peak over threshold)模型,运用Copula方法来估计VaR的值.给出实例验证,将上述方法用于刻画美国纳斯达克指数和标准普尔指数的相关性,并计算了等权重下资产组合的VaR估计值.结果表明:VaR估计值的大小与所取的置信水平以及持有期有关;t-Copula和Clayton Copula方法较其他方法能更好地捕捉资产组合的相关关系,从而可以得到更好的VaR估计值.Using VaR （value at risk） as the measure of financial risk, this thesis applies Copula function and its relevant functions to establish a financial risk module. Considering the time-varying and fat-tail features of the financial time series, it uses Copula method to estimate the value of VaR, according to GARCH （generalized autoregressive conditional heteroscedasticity） model and POT （peak over threshold） module in extreme value theory. The method is illustrated with an example that the correlativity of American NASDAQ index and Standard ＆ Poor＇s indices are indicated and the estimated VaR of the equal weight portfolio is calculated. The result shows that the estimated value of VaR is related to confidence level and holding period, t Copula and Clayton Copula methods are superior to others in capturing the correlativity of portfolio, hence the estimated VaR is better.

关 键 词：风险价值 金融时间序列 广义自回归条件异方差 极值理论 COPULA函数 

分 类 号：O213.9[理学—概率论与数理统计] F830.9[理学—数学]