Scaledt分布、杠杆效应和上证综指的VaR风险

与正态分布相比，上证指数收益率的经验分布具有尖峰厚尾特征，但用Scaled t-分布比正态分布可以更好地拟合上证指数收益率的经验分布。本文以Scaled t-分布假设下的GJR模型为基础，测量了上证指数收益率波动性的杠杆效应，即信息对波动性的不对称影响：并根据GJR模型应用Monte Carlo模拟方法，测定上证指数日收益率和持有期收益率的风险价值（VaR）。根据GJR模型提供的结果，上证指数30天、60天和90天持有期收益率的风险值分别为12．1％、17．8％、22．0％。用GJR模型比均值-方差模型和历史模拟方法计算的5％显著性水平VaR值更接近实际收益率。

Scaled t-Distribution、Leverage Effect and the VaRs of the CISSE

The daily rates of return from the Composite Index of Shanghai Stock Exchange have more leptokurtosis and fatter tails than are compatible with the normal distribution. Instead, the scaled t-distributlon fits the data series quite well. The GJR model allowing for conditionally t-distributed errors is presented to measuring the leverage effect, i.e. the asymmetry of the volatility response to news. Meanwhile, the GJR model is used to estimate the VaRs of the daily and holding-period rates of return via Monte Carlo simulation method. Our results from the GJR model suggest that the VaRs of CISSE for 30,60 and 90 trading dates are 12.1%,17.8%,22.0% respectively. The VaRs from the GJR model are closer to the actual rates of return than those from the mean-variance method and the historical simulation method.