混合分布的VaR非参数估计:对期货市场的实证分析  

Estimating VaR by nonparametric estimation with mixed distribution:Empirical investigation of futures market

作　　者：肖佳文[1] 杨政[1] 

机构地区：[1]电子科技大学经济与管理学院,四川成都611731

出　　处：《系统工程学报》2016年第4期471-480,共10页Journal of Systems Engineering

基　　金：国家自然科学基金资助项目(70901013);教育部人文社会科学研究青年基金资助项目(15YJC790132);中央高校基本科研业务费资助项目(ZYGX2015J159)

摘　　要：基于价格走势的不同将金融市场分为平稳行情,趋势与震荡行情,用混合分布刻画资产收益的母体分布特征,提出一种新的VaR非参数估计方法.利用高频数据对市场状态分类,基于Bootstrap的核密度方法估计子分布,最后计算总体分布的VaR.选取股指期货,铜期货,橡胶期货,豆粕期货等四个期货品种进行实证研究,估计了不同置信水平下的多头和空头VaR.实证结果表明:新方法相较传统的GARCH族模型,多种基于犀尾分布的GARCH模型,混合正态GARCH模型以及方差一协方差法能够更准确地估计空头VaR与尾部风险.This paper proposes a new nonparametric estimation method for the VaR （value at risk） based on a mixed distribution. This method classifies market states as the stationary state and trend or fluctuating state via market changes and assumes the distribution of the asset return to be a mixed distribution. It classifies the market states using high-frequency daily data, estimates sub-distributions using kernel density estimation based on Bootstrap method, and computes the total VaR at last. CSI300, copper, rubber and soybean meal futures were selected to estimate the VaR of both long and short positions at different confidence levels in the empirical study. The empirical results show that the new method can estimate the VaR of short positions and tail risks more accurately than GARCH-type models, GARCH models based on fat tail assumptions, NM-GARCH model and variance-covariance method.

关 键 词：VAR 混合分布 核密度估计 BOOTSTRAP 

分 类 号：F830.91[经济管理—金融学]