基于MCMC参数估计的POT极值理论度量影子银行与A股市场VaR-ES  

The Measurement of VaR-ES of A-share Market and Shadow Banking Based on MCMC Parameter Estimation of POT Extreme Value Theory

作　　者：李锦成[1] 

机构地区：[1]中国社会科学院数量经济与技术经济研究所

出　　处：《中央财经大学学报》2017年第5期37-47,共11页Journal of Central University of Finance & Economics

基　　金：国家自然科学基金项目"国际资本流动与宏观审慎性政策研究"(项目编号:71303044)

摘　　要：2008年次贷危机爆发后,美国学术界进行了大量研究检验影子银行与股票市场的极值情况。在此背景下,笔者研究中国影子银行的极值风险,对今后进一步研究极值影响性作出铺垫。对极值数据的分布函数,通常无法用高斯分布或者t分布来有效表达。金融时间序列的尖峰、拖尾、右偏情况,可以基于广义帕累托分布的POT模型进行对极值数据的拟合,确定超出安全阈值的极值数据的分布形式。利用Gibbs抽样的贝叶斯MCMC模拟方法来估计模型的参数,可以解决当样本数据不足时极大似然估计中误差增大的问题,提高数据的拟合效果。笔者首次基于POT模型对中国影子银行与A股市场的极值进行实证研究,确定阈值后分别测算了MCMC估计和MLE估计下的VaR和ES。结果发现:上证成交量极值风险更大,影子银行极值风险相对较小。After the subprime mortgage crisis in 2008, the United States academia has conducted a large number of studies to test the extreme value of shadow banking and stock market. In this context, studying the extreme value of China shadow banking can help the further study of extreme value influence on market. The distribution function of the extreme value data can not be effectively expressed by Gauss or t distribution. Especially for financial time series, it＇s extreme value can be fitted by POT model based on GPD. Bayesian MCMC simulation method is used to estimate the parameters of the model, which can solve the problem of increasing error in MLE when the sample data is insufficient, and fitting effect can be improved. The paper researched for the extreme value of shadow banking and A stock market based on the POT model, and calculated VaR and ES in terms of MCMC and MLE estimates. The results found that： the extreme value of Shanghai index turnover is greater risk, shadow banking is relatively small.

关 键 词：影子银行 A股市场 POT模型 MCMC估计 GPD VAR估计 ES估计 

分 类 号：F830.9[经济管理—金融学]