我国商业银行系统性风险的识别与预警研究  

Identification and Early Warning of Systemic Risk of Chinese Commercial Banks

作　　者：李辰颖 LI Chen-ying

机构地区：[1]上海财经大学会计学院 [2]北京林业大学经济管理学院

出　　处：《中央财经大学学报》2020年第10期36-53,共18页Journal of Central University of Finance & Economics

基　　金：国家社会科学基金一般项目“基于影子银行渠道的房地产泡沫与系统性金融风险关系研究”(项目编号:14BJY172)。

摘　　要：防范金融系统性风险是十九大提出的重要任务,而银行是金融系统非常重要的子系统,识别银行系统性风险因素、构建风险预警模型是防范风险的重要手段。首先,从宏观经济和银行系统两方面建立评估银行系统性风险的指标体系,并采用方差阈值法对指标进行初步筛选,接着提出了基于孤立森林的指标筛选法进一步筛选出与银行系统性风险高度相关的4个指标。其次,为了识别银行系统性风险,提出了包含孤立森林法、基于角度的离群点检测法和局部异常因子法的多种异常检测算法相结合的银行系统性风险识别模型,并选取2012年1月到2018年12月的数据进行风险识别。从识别结果来看,所构建的风险识别模型基本准确识别出了历史风险时点,与历史事件切合度高,结果具有可解释性。最后,采用DeepAR方法分别预测入选的4个指标,并将预测结果输入到银行系统性风险识别模型,以此构建银行系统性风险预警模型,预警结果表明该模型在一定程度上能够对银行系统性风险进行预警。Preventing financial systemic risks is an important task proposed by the 19th National Congress of the Communist Party of China.As banks are very important subsystems of the financial system,identifying bank systemic risk factors and building risk early warning models are critical tools for preventing risks.Firstly,an index system that including macroeconomic and banking systems aspects was established to evaluate the systemic risk of banks.Then,11 indicators were extracted using a variance threshold method,in order to further extract the indicators.A method based on Isolation Forest was proposed and only four indicators left,which are highly relevant to the systemic risk of banks.Secondly,a bank systemic risk identification model compounded from Isolation Forest,Angle-based Outlier Detector and Local Outlier Factor was proposed.Moreover,the data from 2012 to 2018 were selected for risk identification.The results showed that the model could identify historical risk time points which were highly consistent with the historical events.Finally,a bank systemic risk early warning model was proposed based on DeepAR.First,four indicators were predicted respectively by DeepAR,then the prediction results were input into the bank systemic risk identification model,and the identification(warnning)results were calculated.The results showed that the model could give a technical support for the early warning of the bank's systemic risk to a certain extent.

关 键 词：商业银行 系统性风险 识别 预警 DeepAR 

分 类 号：F830.593[经济管理—金融学]