相依结构、动态系统性风险测度与后验分析  

Dependence Structure,Dynamic Systemic Risk Measurement and Backtesting Analysis

作　　者：王锦阳 刘锡良 杜在超 Wang Jinyang,Liu Xiliang, Du Zaichao

机构地区：[1]西南财经大学中国金融研究中心 [2]西南财经大学经济与管理研究院

出　　处：《统计研究》2018年第3期3-13,共11页Statistical Research

基　　金：中国博士后科学基金面上资助项目“中国银行业系统性金融风险：生成机理、测度与宏观审慎监管”(2016M590077);教育部人文社会科学重点研究基地重大项目“我国金融安全影响机制研究”(17JJD790024)、“基于金融稳定的货币政策与宏观审慎监管协调配合研究”(15JJD790027);西南财经大学一流学科建设项目“经济结构转型中金融风险与金融安全研究”;西南财经大学金融安全协同创新中心年度项目“基于医学传染原理的金融危机传染机制研究”(JRXT201607)的阶段性成果.

摘　　要：本文利用Copula相依结构理论扩展和求解了现有的系统性风险测度Co VaR,以得到适用于不同类型常参数和时变参数Copula函数及不同分布假设的动态系统性风险测度。为了验证和评估模型设定的准确性与应用价值,本文构建了适用于该动态系统性风险测度Co VaR的严谨后验分析工具。除"无条件覆盖性"、"独立性"和"条件覆盖性"外,本文首次提出了"混合独立性"检验。基于我国14家上市商业银行的实证分析表明:我国上市商业银行与我国银行业之间的相依结构呈现多样化特征;无论是样本内还是样本外预测区间,本文的动态Copula-Co VaR模型能够有效地捕捉典型系统性风险事件;严谨的后验分析不仅需要检验系统性风险测度Co VaR,也需要检验条件事件的临界值VaR。Using the the Copula dependence structure theory, the paper extends the existing systemic risk measurement CoVaR, and obtains the dynamic systemic risk measurement which is suitable to different Copula functions and different distribution assumptions. At the same time, we construct a rigorous backtesting analysis tools for the dynamic systemic risk measurement CoVaR. In addition to the unconditional coverage property, independence property, and conditional coverage property, we first propose the cross independence property for backtesting the systemic risk measurement CoVaR. In the empirical study of 14 Chinese listed commercial banks, we show that： The dependence structure between the listed bank and bank system is diversified; Our dynamic copula CoVaR model can effectively capture the systemic risk events in the both in-sample period and out-of-sample period; We argue that one needs to apply the rigorous backtesting analyses not only to the systemic risk measurement CoVaR, but also to the critical value of the conditional event, VaR.

关 键 词：相依结构 条件在险价值CoVaR 系统性风险 后验分析 

分 类 号：C812[社会学—统计学]