基于厚尾分布模型在上证股指波动趋势上的应用

Application of Models Based on Fat-tail Distribution in Fluctuation Tendency of Shanghai Composite Index

高风险高波动时期的股指收益率序列存在更为明显的厚尾现象,将残差设定为偏t分布的GARCH族模型要优于普通的正态分布以及学生t分布模型。以五种损失函数作为评价准则,在残差序列偏t分布前提下比较了GARCH模型、EGARCH模型和GJR-GARCH模型,最后得到在高风险时期结构更为简洁的GARCH(1,1)模型具有更好的预测效果。

During the high risk period,the yield rate series of stock of index has more obvious fat-tail phenomenon,Setting the residuals for fattail distribution of these GARCH models is superior to the models with normal distribution and t distribution.According to the five kinds of loss functions,GARCH model、EGARCH model and GJR-GARCH model are compared based on fat-tail distribution,the conclusion is that GARCH（1,1）model with briefer structure has a better prediction ability in the high risk period.