基于Expectile-based VaR变点检测的金融传染分析  

Analysis of Financial Contagion Based on Change point Testing of Expectile-based VaR

作　　者：谭常春 操毅文 叶五一[3] 

机构地区：[1]合肥工业大学经济学院,安徽合肥230601 [2]合肥工业大学数学学院,安徽合肥230601 [3]中国科学技术大学统计与金融系,安徽合肥230026

出　　处：《数理统计与管理》2018年第2期371-380,共10页Journal of Applied Statistics and Management

基　　金：国家社会科学基金(16BTJ023);全国统计规划重点项目(2012LZ009);安徽省自然科学重点项目(KJ2017A912)

摘　　要：本文在分位点回归的基础上采用ALS方法，建立了线性Expectile模型，并以此计算美国金融危机期间有关经济体的Expectile-based VaR（EVaR），对EVaR进行变点检测。根据变点的位置来估计危机发生的时间，并通过比较模型的系数在危机前后的变化情况，分析了美国金融危机对中国、香港、英国、德国和日本造成的传染情况。同时结合实际情况分析线性Expectile模型所确定的危机发生时间和系数变化所反映的危机传染性强弱程度，并与斜率模型得到的结论进行比较，体现了线性Expectile模型的准确性。Abstract： In this paper, the coefficient of the linear Expectile model is estimated by using asymmetric least squares criterion function in the Quantile Regression Model. We compute the value at the risk with the Expectile function （EVaR） of some economies, such as China, Hong Kong, UK, German and Japan during the USA financial crisis, detect and estimate the position of change points of the EVaR. According to the change time of EVaR, comparing the estimated coefficient of the linear Expectile model before and after the crisis, we analysis the contagion among USA and other countries. We measure the degree of the financial contagious with the change of the coefficients and the situation of the finical crisis. The results of the linear Expectile model compared with the slope model show that the linear Expectile model is more accuracy than the slope model.

关 键 词：分位点回归模型 预期位在险价值 变点 金融传染 

分 类 号：F224[经济管理—国民经济] F830