基于极值Copula的风险价值模型研究

利用Monte Carlo模拟对投资组合风险价值进行估计,假设风险资产的对数收益分布在中心服从边缘正态分布的多维Gaussian或Student-Copula分布,尾部服从极值分布。应用本模型对12只沪市股票计算持有期为1天的99%VaR,并执行4年时间窗的返回检验,证明了Copula-EVT方法好于正态对数收益分布的VaR模型。

A Value-at-Rish Model Based on Conpula-EVT

The purpose of this paper is to present a method based on Monte Carlo Simulation to estimate portfolio Value-at-Risk,assuming that the risk factor log-returns are generated from a multivariate distribution with such edge normal distribution as Gaussian or Student-Copula at the center and extreme value distribution at the tails.We apply the model to estimate the 99% VaR over a one-day holding period,and for a portfolio of twelve stocks to perform a back test procedure over a time window of four years,which proves that the Copula/EVT-based approach outperforms the traditional VaR models assuming a conditional normal multivariate distribution used for risk factor log-returns.