基于动态因子Copula模型的我国银行系统性风险度量

本文基于2007年至2019年我国14家上市银行的股票收益率,构建偏态t-分布动态因子Copula模型,利用时变荷载因子刻画单家银行与整个系统的相关性,计算联合风险概率作为系统性风险整体水平的度量,基于关联性视角提出了新的单家机构系统脆弱性和系统重要性度量指标——系统脆弱性程度和系统重要性程度。该方法充分考虑了银行个体差异性和系统的内在关联性以及收益率的厚尾性和非对称性,从而能够捕捉到更多的信息且兼具时效性。研究表明:银行机构在风险聚集时期相关程度更大,联合风险概率能够准确识别出系统性风险事件且在我国推行宏观审慎评估体系以后有明显降低;整体而言,大型商业银行系统重要性水平最高,同时风险抗压能力也最强;本文使用的度量方法降低了数据获取成本且更具时效性,有助于为宏观审慎差异化监管工作提供借鉴和参考。

Measuring Systemic Risk of China's Banking Based on the Time-Varying Factor Copula Model

The 2007 subprime crisis provides ample evidence of the inevitable consequences of systemic risk. The evidence has motivated researchers, academics, and regulators to recognize, measure, and prevent systemic risk. China's banking system occupies a very important place in its financial system. The banking system has a closer internal relationship and dependence structure than other financial sectors because of inter-bank borrowing, payment, and settlement. Therefore, studies that measure systemic risk in China's banking system, identify important and vulnerable systemic institutions, and prevent systemic financial risk are of great academic value and practical significance.An accurate model of institutional dependence structures is required for measuring systemic risk. The model captures the spillover effect between institutions. Studies have shown that the financial system's dependence structure is asymmetric and nonlinear, and that interaction increases during financial crises. Many studies have proposed indicators to measure systemic risk, but they have some shortcomings. First, classic indicators such as MES and CoVaR focus primarily on the relations between pairs of institutions or an individual firm and the market index. Consequently, they miss the dependency of the whole system. Second, network models based on tail risk can measure how institutions interact with each other in the system, but this kind of model is based on binary relations. Third, few studies focus on the balance of systemic importance and systemic vulnerability.We apply the time-varying factor copula model, which analyzes the banking system's idiosyncrasy and interconnectedness to 14 listed Chinese banks' return data from 2007 to 2019. This approach is suitable for high dimensions, and it can capture fat-tailed, time-varying, asymmetric, and nonlinear characteristics. It analyzes the dynamic dependence between the individual bank and the system according to dynamic factor loadings. The unified framework established by the joint distribution of the banking system, we propose indicators of systemic risk in China's banking system. First, the joint probability of distress (JPD) can be used as a measure for the probability that a majority of the financial institutions are in default. In addition, the Systemic Vulnerability Degree (SVD) and Systemic Importance Degree (SID) can identify systemically important institutions and systemically vulnerable institutions. The two categories account for the overall and local dependencies of the banking system. These indicators account for the individual bank's idiosyncrasy, local and overall dependence, and fat-tailed and asymmetric chrematistics of return data, capturing a range of information.This study's research results in two findings. First, the relationship between banks and the banking system increases as risk increases. The joint probability of distress accurately identifies the 2008 subprime crisis, the 2013 “money shortage,” and the 2015 stock market crash. The JPD shows that macro-prudential assessment lowers systemic risk and the 2018-2019 trade friction between China and US increases the risk. Second, big-five banks are most systemic stable and city commercial banks are most vulnerable in the sample period.The systemic importance indicator (SID) shows that big-five banks are most affected by spillover during the sample period, which implies that big-five banks are not only “too big to fail” but also “too connected to fail.”