深证100指数的风险值估计方法

An Estimate Method for Value-at-Risk Incorporating Extreme Value Theory into GARCH Model

实证研究表明,用VaR-X修正模型中的残差分布(t分布)尾部指数的简化估计方法得到的尾部指数估计值与最大观察数目选择有关,具有不稳定性.当最大观察数目足够大时,尾部指数序列的折线图是曲线,而不是直线,且采用普通最小二乘法估计时,模型存在条件异方差,会导致估计失效.采用ARCH模型或GARCH模型估计法不仅可以克服模型的条件异方差性,同时可解决最大观察数目选择难的问题.

Empirical research evinces that the tail - index estimate valuable by using the tail - index estimate method of residual distributing（ t - distributing ） of Var - X model has something to do with the largest observation number of choice and is unsteady . When the largest observation number is large enough, the line graph of the tail - index series is bend graph not line graph. And when adopting Ordinary Least Square estimate, model exists Heteroskedasticity. Least Squares estimate may be invalid. However,adopting ARCH or GARCH model estimate not only overcomes model Heteroskedasticity but also resolves the choice problem of the largest observation number.