基于极值理论的高频条件VaR动态区间估计模型  

Interval estimation of high frequency conditional VaR based on extreme value theory

作　　者：王春峰[1] 张亚楠[1] 房振明[1] 刘峥然[1] 

机构地区：[1]天津大学管理学院,天津300072

出　　处：《系统工程理论与实践》2010年第7期1162-1168,共7页Systems Engineering-Theory & Practice

基　　金：国家自然科学基金(70771076);国家杰出青年基金(70225002)

摘　　要：为了更加精确地度量在险值的估计精度,基于广义极值理论推导了条件极值VaR的动态区间估计模型,得到了条件极值VaR置信区间解析解的一般形式,对在险值的估计精度进行了实时度量.利用高频数据重点考察了不同置信水平和不同样本容量分块下的条件极值VaR区间估计结果的精度和模型的有效性.结果表明:条件极值VaR的动态区间估计模型与参数法、非参数法以及蒙特卡罗法区间估计模型相比,不仅能够更为有效地捕获极端条件下收益率时间序列的动态特征,而且具有更好的估计精度,精确和有效地描述VaR的估计风险.In order to catch the character of return series in extreme condition and improve VaR precision, a model of the conditional extreme value VaR was established and got the general form of analytical solution to the confidence interval of conditional extreme value VaR. Using high frequency data, the result precision of confidence interval of the conditional extreme value VaR and the validity of model were mainly studied under the different confidence level and block. The empirical results show that comparing our model with parametric method, non-parametric method and Monte Carlo method in estimation of the confidence interval of VaR, our model not only can catch the risk character of Chinese stock markets, but also has the better estimation accuracy and describe the estimation risk of the VaR more accurately.

关 键 词：置信区间 条件极值VaR 广义极值分布 高频数据 

分 类 号：F224.0[经济管理—国民经济]