利用中国资产管理公司的不良贷款数据库,对影响我国不良贷款回收率的因素:风险暴露规模、地区、行业、担保方式、五级分类、逾期时间等,进行了统计分析;在此基础上,建立了单户处置企业的不良贷款回收率预测模型,并且利用模型的各个影响因素对回收率的贡献程度进行了测算.以单户预测模型为基础,结合打包处置的处置策略,利用十折交叉验证和组合预测的思想,建立了打包处置的回收率预测模型.实证结果表明:无论是单户预测模型还是打包预测模型,预测结果均达到了较高的精度.

Based on NPL database of AMC in China, this paper gives a comprehensive investigation on influencing factors for recovery rate of China’s non-performing loans. These factors include risk exposure, area, industry, collateral type, 5-category loan classification, and time between default and clearing, etc. A recovery rate prediction model is built for the individual company clearing way. This model is used to analyze the recovery contribution of each influencing factor. Moreover, based on this model, this paper further builds a prediction model for package clearing way by incorporating with the techniques of 10-fold cross-validation and combination forecasting. Empirical results show that both models have relatively high precision.