极值风险价值和预期损失的估计应用：以影子银行和A股市场为例

The Va R and ES of Extreme Value by MCMC and MLE Estimation: A Case Study of Shadow Bank and A-share Market

在检测金融时间序列极值的风险价值和预期损失时,由于金融时间序列具有尖峰和厚尾的特征,首先利用广义帕累托分布的安全阈值模型Hill图观测。当样本数据不多时,利用Gibbs抽样和蒙特卡洛模拟马尔科夫链来检验参数的拟合效果。最后,对从Hill图观测出的极值情况基于MCMC和MLE方法估计其风险价值和预期损失。本文以中国的影子银行规模、上证指数、上证成交量时间序列为例,检测了三者的极值风险值和预期损失,经过比较发现:上证成交量极值风险更大,影子银行极值风险相对较小。

In the case of measuring the Va R and ES of the extreme value of the financial time series, the financial time series has the characteristics of spikes and thick tails, first using the GPD of the POT model Hill diagram observation. When the sample data is not large, the Gibbs sampling and MCMC are used to test the fitting effect of the parameters. Finally, the MCMC and MLE methods are used to estimate the Va R and the ES for the extreme values observed from the Hill chart. This paper uses China＇s shadow bank size, the Shanghai index, the Shanghai trading volume time series as an example, measured the three extreme Va R and the ES, after comparison found： the Shanghai trading volume has a greater risk, shadow bank has a smaller risk.