基于面板Logit模型的银行客户贷款违约风险预警研究

A study on early-warning of bank customer loan defaults based on panel Logit model

采用我国主要银行业金融机构的客户大额授信月度微观面板数据,对银行客户贷款违约风险进行预警.从客户外部因素、客户经营水平、客户交易水平三个维度,建立了包含56个因素的客户贷款违约预警三级指标体系.基于这些指标,构建了银行客户贷款违约风险预警的面板Logit模型;并提出了基于模拟退火算法的面板Logit模型变量选取方法,最终选出24个预警指标.根据这些指标进行银行客户贷款违约预警,达到了较高的预警精度.最后,对我国银行客户风险管理提出了相应的政策建议.

In this paper, monthly micro panel data of customers＇ big credit for main banking financial institutions in China is used to study early-warning of loan defaults. First, from the dimension of customer external factors, customer managerial level, and customer transaction level, we constructed a three-level index system for early-warning of customer loan defaults, which include 56 factors. Second, based on these indicators, a panel Logit model for early-warning of bank customer loan defaults is established; and a variable selection method based on simulated annealing （SA） algorithm for panel Logit model is proposed, and 24 indicators are chosen using the proposed SA algorithm. Using these 24 indicators for early-warning of bank customer loan defaults, and the warning reached a high precision. Finally, we proposed the corresponding policy recommendations for the bank customer risk management.