经济政策不确定环境下银行系统性风险预警研究——基于SHAP可解释机器学习的应用

Research on the Early Warning of Banking Systemic Risk under the Uncertainty of Economic Policy——Application of SHAP Explainable Machine Learning

经济政策不确定性是影响银行系统性风险管理的重要因素。文章在传统银行系统性风险预警体系中嵌入经济政策不确定性,基于我国2012—2020年银行宏微观数据,构建银行系统性风险指数,采用逻辑回归、随机森林、XGBoost模型评估我国银行系统性风险。实证表明:一是嵌入经济政策不确定性因素的银行系统性风险指数反映了我国银行系统性风险走势;二是随机森林、XGBoost模型在银行系统性风险预警方面表现出较高的预测精度;三是通过SHAP可解释算法对指标体系中重要变量进行“黑箱”分解发现,广义货币量、拨备覆盖率以及经济政策不确定等因素对银行系统性风险有重要影响。最后,基于分解结果为我国防范化解银行系统性风险提出具体措施及相关监管建议。

The uncertainty of economic policy is an important factor affecting the systematic risk management of banks.This paper embeds the uncertainty of economic policy in the traditional banking systematic risk early warning system,builds a banking systematic risk index based on the macro and micro data of China's banks from 2012 to 2020,and uses logical regression,random forest,and XGBoost models to assess China's banking systematic risk.The empirical results show that:(1)The bank systematic risk index embedded with economic uncertainty factors reflects the trend of China's bank systematic risk;(2)Stochastic forest and XGBoost model show high prediction accuracy in risk early warning;(3)Through the“Black Box”decomposition of important variables in the indicator system using the SHAP interpretable algorithm,it is found that broad money volume,provision coverage,economic policy uncertainty and other factors have an important impact on the systemic risk of banks.Finally,based on the decomposition results,specific measures and relevant regulatory suggestions are proposed for China to prevent and resolve the systemic risks of banks.