基于贝叶斯方法的中国商业银行同业借贷网络中系统风险研究

Systemic Risk in China’s Interbank Liability Networks Based on the Bayesian Methodology

本文针对银行双边风险敞口不可得的现实情况,利用贝叶斯方法,基于185家商业银行在2013年至2017年的资产负债表数据,在不同的网络结构设定下构建吉布斯抽样器,根据大量银行间同业资产及同业负债分布矩阵的样本,考察了每个商业银行在负面冲击后违约的概率及其分布。研究结果表明,银行同业借贷网络的结构能够显著影响银行的系统风险和违约概率。当网络连接概率处于中等水平时,冲击影响的范围最广;在完全网络结构下,风险分担的作用大于风险传染。总之,银行同业借贷既可以分担风险,也成为了风险传染的渠道,这种功能的转换取决于以下几类因素的相互作用:冲击的性质,例如冲击的规模,受冲击银行的数量以及冲击涉及的银行类型;清算时资产的贬值程度;银行自身资产负债表的特征。如果仅考虑银行同业借贷渠道,样本期内最稳健的银行系统是在2017年,而2014年的银行系统最脆弱。

Interbank exposure is one of the main channels of risk contagion especially in terms of systemic risks.However,the theoretical models in this field have standardized and generalized specifications and don’t capture banks’characteristics in reality.In this sense,this study is of practical significance as it is based onthe balance sheet data of real players in the China’s financial system.Due to the lack of bilateral exposures between banks,interbank liability matrices must be estimated.The most widely used estimation method,such as entropy maximization,gives point estimator that underestimates systemic risks,while a large number of interbank liability matrices are sampled with the application of Bayesian methodology.The Gibbs samplers of three different network specifications are constructed.Moreover,most of the previous literature only has listed banksor a dozen of largest banks as samples,ignoring more than one hundred smaller banks which are also indispensable for the whole system.The aim of this study is to give a better understanding of the systemic risk in China’s interbank market and to identify the susceptible or robust institutions.Based on the balance sheet data of 185 banks in China,including the"Big Six";12 joint-equity banks;98 city banks;42 rural banks;and 27 foreign banks,the systemic risk arising from interbank liabilities is examined by the Bayesian methodology which samples a large number of interbank liability matrices.Then the default probabilities of individual banks are generated after negative loss shocks.To the best of our knowledge,it might be the first paper in the Chinese context to measure systemic risks based on the simulation of interbank liability networks.The results show that the structure of interbank liability matrices can remarkably impact on the default probabilities of financial institutions.The scope of being influenced is wider if the network connection probability is in the intermediate level.In a complete network,the effect of risk sharing exceeds that of risk contagion.To conclude,the interbank exposure not only enables the institutions to share risks but also provides the channel for spreading risks.This role transformation depends on the interaction of the following factors:the characteristics of shocks,e.g.the shock size,the number of attacked banks as well as the category of attacked banks;the recovery rate of liquidation;the structure of banks’balance sheets.