基于C藤copula的收益率自相依结构估计以及条件VaR计算  

Estimating auto-dependence structure and conditional VaR based on canonical vine copula

作　　者：李磊[1] 叶五一[1] 缪柏其[1] 

机构地区：[1]中国科学技术大学管理学院统计与金融系,安徽合肥230026

出　　处：《中国科学技术大学学报》2013年第9期745-753,共9页JUSTC

基　　金：国家自然科学基金青年科学基金(71001095);国家自然科学基金青年面上连续项目(71371007);高等学校博士学科点专项科研基金(20103402120010)资助

摘　　要：风险价值VaR是一种非常重要的风险度量方法,基于C藤copula(canonical vine copula)给出了条件VaR的一种新的估计方法.首先基于C藤copula对连续几个交易日收益率之间的自相依结构进行了估计,进而给出了在前n个交易日收益率条件下,下一交易日收益率密度函数的估计方法,并对下一交易日的VaR值进行估计.C藤copula的引入使我们能更准确地描述收益率序列中的相依结构,从而能够更加准确地预测市场风险.最后分别对沪深300指数、上证180指数和上海黄金交易所贵金属价格进行了CVaR估计以及预测效果检验实证分析,实证结果表明所提出的模型在VaR值预测方面的表现要远远优于历史模拟法以及方差-协方差法等.Value at Risk is a very important method for measuring risk. A new VaR estimating method based on canonical vine copula was presented. The auto-dependence structure of several consecutive trading days＇ yield was estimated based on canonical vine copula. Then, the conditional density function of the next trading day under the condition of previous trading days＇ yield was calculated, and the VaR of the next trading day could be estimated from the conditional density function. The dependence structure among return series could be described more accurately by canonical vine copula, and the market risk could be predicted with greater accuracy. At last, an empirical analysis of CSI 300, SSE 180 and precious metal prices listed on Shanghai Gold Exchange was presented. The empirical results show that performance of the proposed model is far superior to historical simulation and variance covariance methods when forecasting VaR.

关 键 词：canonical藤copula 自相依结构 条件VAR 

分 类 号：F830.9[经济管理—金融学]