基于网络传导的金融系统风险度量：理论及其应用

In this project we present some new measures of the systemic risks of financial systems globally. In particular, we consider the interdependence of individual countries in times of crisis and explore to which extent the crisis of a single country will affect the global market through the channel of chain reactions. Specifically, we quantify this interconnectedness between financial individual entities by employing a centrality measure originally proposed by Bonacich (1987) to capture the importance of a particular country relative to other countries within the system/network. A country with higher centrality score contributes more to the global systemic risk compared to a country with a lower centrality score (CRISK). We also present a country-beta weighted average of CRISK's to measure the global systemic risk - GRISK. We shall apply these measures to three levels of empirical studies: first, at the national level we will examine the order importance of different nations in terms of their contributions to the global systemic risk and construct a daily time series of global systemic risk; second, we examine the order importance of major global companies in the financial systemic risk; lastly, we investigate the systemic risk in China using Chinese banks and other financial firms as a sample. Lastly, we model the time seris dynamics of systemic risk and implement out of sample forecasts. Macroeconomic and financial factors and firm idiosyncratic characteristics that can potentially affect the systemic risk are identified through regression analysis.

本课题研究度量金融系统风险的新方法。之前各国央行和学术界提出的度量指标都忽视了系统内个体之间的风险传导机制，即不同的系统内某个体的破产会在不同程度上导致其他个体的破产，而个体间的这种关联恰恰是金融危机的形成渠道。我们运用现代网络理论和计量理论将这种风险传导机制量化。具体地，我们对金融系统内个体遭遇危机后给其他个体带来的链式反应两两进行时间序列建模，形成传导网络，然后结合网络理论中的集中度概念提出CRISK（链式风险）指标，以此衡量个体在金融系统风险中的重要性。在此基础上，我们将个体的链式风险进行贝塔加权，提出整个系统风险的度量指标（GRISK）。此外，本课题将金融系统风险度量方法运用到国家层面和企业层面，对全球各国各地区及我国金融企业在系统风险上的重要性进行评估。最后，本课题对金融系统风险时间序列进行建模和样本外预测，并分析影响金融系统风险的宏观经济、金融市场和个体特征等因素。

本项目研究计划产生于2008年全球金融危机后引发的对系统风险的反思。如何刻画金融系统内个体间的风险传导是本项目的重点研究内容。我们的研究思路是通过网络理论给风险传导进行建模，并构建系统风险指标。具体地，我们采用了有向无环图网络理论（DAC），并结合波动溢出模型对系统内个体之间的风险传导进行建模，在此基础上推导出全球系统风险指标（GRISK）。我们将该模型应用到以下四个方面的子课题：第一，我们考察了全球主要金融市场的隐含波动率的网络传导行为，得到了一些有趣的发现；第二，我们考察了全球主要货币对之间的传染行为；第三，我们研究了期权隐含波动率期限结构的全球传染特征，并建立模型进行样本外的预测，取得了很好的效果；第四，我们研究了2015年我国股市异常波动期间股指期货与现货之间的价格发现和波动率传导关系。