EGARCH模型和分位数自回归类模型在风险预测中的比较  

Comparison of EGARCH Model and Quantile Autoregressive Model in Risk Prediction

作　　者：李月鲜[1] 刘媛媛[1] 张巍巍 LI Yuexian;LIU Yuanyuan;ZHANG Weiwei(College of Science,Inner Mongolia Agricultural University,Hohhot O10018,China)

机构地区：[1]内蒙古农业大学理学院,呼和浩特010018

出　　处：《内蒙古大学学报（自然科学版）》2022年第6期660-666,共7页Journal of Inner Mongolia University：Natural Science Edition

基　　金：内蒙古自治区留学回区人员创新启动项目(DC1900004065);内蒙古自治区高等学校科学技术研究项目(NJZY22494)。

摘　　要：利用了三种时间序列模型对中美汇率的风险进行一步向前滚动预测。(1)EGARCH模型,它可以刻画金融资产收益率的尖峰厚尾现象,异方差性,波动集聚性和杠杆效应。(2)非常适合具有异方差性的QAR模型,该模型不需要假定误差项的分布,可以很好地刻画条件异质性,在模拟具有尖峰厚尾的金融数据时,得到了广泛的应用。(3)T-QAR模型,该模型可以通过门限变量控制分段线性函数,进而刻画金融资产具有的非线性性和非对称性。最后,对三种模型的预测效果进行了VaR回溯检验,结论是T-QAR模型的预测效果最佳,而EGARCH模型的预测效果最差。Three classical time series models are used to predict the risk of Sino-US exchange rate.First, The EGARCH model can describe the phenomenon of Leptokurtosis, heteroscedasticity, fluctuation clustering and leverage effect of financial asset return rate.Second, The QAR model does not need to assume the distribution of error terms, and can well describe Conditional heterogeneity, heteroscedasticity and Leptokurtosis.Third, T-QAR model, which can control piecewise linear function through threshold variables, and then describe the nonlinearity and asymmetry of financial assets, was studied.Finally, VaR back testing was carried out on the prediction effect of the three models, and the conclusion was that the T-QAR model had the best prediction effect, while the EGARCH model had the worst prediction effect.

关 键 词：EGARCH模型 QAR T-QAR VAR 回溯检验 

分 类 号：F830[经济管理—金融学]