基于GARCH族类模型的开放式基金收益率风险分析  

The Risk Analysis of Return Rate of ETF Based on GARCH Models

作　　者：林琦 欧思歆 LIN Qi;OU Si-xin(School of Mathematics and Finance,Putian University,Putian 351100,China;School of Mathematics,Xiamen University,Xiamen 361005,China)

机构地区：[1]莆田学院数学与金融学院,福建莆田351100 [2]厦门大学数学科学学院,福建厦门361005

出　　处：《数学的实践与认识》2019年第24期108-114,共7页Mathematics in Practice and Theory

基　　金：福建省自然科学基金项目(2019J01817);福建省社科规划项目(FJ2018B065);莆田市科技计划项目(2018RP4003)

摘　　要：基于中国上证50ETF的单位净值,实证分析了国内开放式基金的风险,并在此基础上研究了分位数回归改良情况.选取GARCH族类模型中的GARCH,TARCH和EGARCH,在95%和99%置信水平下计算VaR值,失败率检验结果表明:三种模型在同一置信水平和分布下估计结果差不多,t分布高估风险,而正态分布在99%置信水平下低估风险,GED是最理想的分布.将分位数0.01和0.05回归前后的VaR值作分析比较,发现0.01分位数回归后对风险的估计过于保守不可信,而0.05分位数回归则可以有效改善t分布高估VaR值现象.最后,提出了完善基金市场风险管理的政策建议.On the basis of reviewing literatures and based on the unit net value of Shangzheng 50 ETF,empirically analyzed the risks of domestic Exchange Traded Funds and the improvement of quantile regression.Selection of GARCH,TARCH and EGARCH models,calculating VaR under 95% and 99% confidence level values,the failure rate test results showed that the three models estimated similarly under the same confidence level and distribution,t-distribution overestimate VaR while normal distribution(under 99%confidence level)underpriced it,and the best distribution was GED.Finally,After compare VaR of quantile regressions of 0.01 and 0.05 before with later,it was found that the estimations under 0.01 quantile were too conservative and unbelievable,while the regression of the 0.05 quantile could effectively improve the overestimation of VaR undern the t-distribution.Therefore,the quantile regression can indeed meliorate the risk estimation,but it is prone to overestimation without appropriate quantiles.

关 键 词：开放式基金 VAR值 GARCH族类模型 分位数回归 

分 类 号：F832.51[经济管理—金融学] F224