动态GARCH-EVT-Copula模型在金融序列局部分析中的应用  

Application of Dynamic GARCH-EVT-Copula Modelin Local Analysis of Financial Sequences

作　　者：徐刚刚 杜海霞 范雪芹 魏轩浩 谢佳洋 XU Gang-gang;DU Hai-xia;FAN Xue-qin;WEI Xuan-hao;XIE Jia-yang(College of Mathematics and Physics,Xinjiang Agricultural University,Urumqi 830052,China)

机构地区：[1]新疆农业大学数理学院,新疆乌鲁木齐830052

出　　处：《兰州文理学院学报（自然科学版）》2020年第4期10-16,共7页Journal of Lanzhou University of Arts and Science(Natural Sciences)

基　　金：新疆维吾尔自治区高校科研计划项目(XJEDU2018Y021);国家级大学生创新项目(dxscx2020518)。

摘　　要：随着经济的发展,金融数据的分布不仅表现出明显的波动集聚性,而且还具有时变性和局部联动性.首先选取GARCH-EVT模型对单一序列进行建模,通过模型对比分析选出合理的边缘分布.其次利用几类Copula函数分别从金融数据的整体、中间、上尾以及下尾分布特征建立GARCH-EVT-Copula模型,然后采用蒙特卡洛模拟方法计算多资产投资组合的VaR.最后结合Kupiec检验方法来验证模型的预测效果.实证研究表明:从整体数据拟合结果来看,T-copula模型比正态Copula模型更具说服力.由于Frank Copula模型的密度分布具有对称性,因此对金融序列中间数据的刻画也较为准确.利用Gumbel Copula与Clayton Copula模型分别对序列的上尾和下尾数据建模,结果也很理想.因此选用合理的GARCH-EVT Copula模型来分析金融序列的局部分布显得很有意义.With the development of economy,the distribution of financial data not only shows obvious volatility cluster,but also has temporal variability and local correlation.In this paper,GARCH-EVT model is firstly selected to fit a single sequence,and the reasonable marginal distribution is selected by model analysis.Secondly,several types of Copula functions are used to establish the GARCH-EVT-Copula model,which is respectively modeled and analyzed from the whole,middle,upper and lower tails of financial data,and then the Value at Risk of multi-asset portfolio is calculated by Monte Carlo simulation method.Finally,the prediction effect of the model is verified with Kupiec test.The empirical study shows that the T-copula model is more convincing than the normal Copula model in terms of the overall data fitting results,because the density distribution of the Frank Copula model is symmetrical,the description of intermediate data of financial sequence is also accurate,the Gumbel Copula model and the Clayton Copula model are used to model the upper and lower tails of the sequence respectively,and the results are also ideal.Therefore,it is significant to select a reasonable GARCH-EVT-Copula model for the local analysis of financial sequences.

关 键 词：两步法估计 VAR ARCH效应 蒙特卡洛模拟 局部分析 

分 类 号：O212[理学—概率论与数理统计]