诉讼处置不良贷款违约损失率估计的模型簇

发布时间：2018-04-09 12:41

本文选题：不良贷款　切入点：诉讼处置　出处：《系统工程》2015年08期

【摘要】：违约损失率(LGD)是内部评级高级法要求的重要参数之一,已成为商业银行风险管理的重要手段。由于受数据等多方面的限制,国内外尚无对我国大型商业银行诉讼处置不良贷款违约损失率估计的研究。本文在对我国某大型商业银行诉讼处置不良贷款违约损失率全面统计分析的基础上,找出回收率的影响因素,采用决策树模型,判别出极端回收和非极端回收。针对非极端回收的情况综合运用Logit变换、Beta-正态逆变换、WOE变换等方法,建立点估计模型;随后利用广义Beta回归给出了LGD的分布模型;在对四个模型进行相关性分析的基础上用最小误差平方和的方法建立了组合模型,由此形成由判别模型与组合模型构成的模型簇。实证结果表明,极端回收的判别准确率高达77.6%,组合模型的均方误差低于5%;模型簇在极端回收和非极端回收两类表现出很好的一致性。
[Abstract]:The default loss rate (LGD) is one of the important parameters required by the advanced internal rating law and has become an important means of risk management in commercial banks.Due to the limitation of data and other aspects, there is no research on the estimation of default loss rate of non-performing loans in litigation disposal of large commercial banks in China at home and abroad.Based on the comprehensive statistical analysis of the default loss rate of a large commercial bank in China, the factors affecting the recovery rate are found out, and the decision tree model is used to distinguish extreme and non-extreme recovery.The point estimation model is established by using Logit transform and Logit transform, then the distribution model of LGD is given by generalized Beta regression.On the basis of the correlation analysis of the four models, the combined model is established by the method of least error square sum, and the model cluster composed of the discriminant model and the combined model is formed.The empirical results show that the accuracy of extreme recovery is as high as 77.6, the mean square error of combination model is lower than 5, and the model cluster shows good consistency between extreme recovery and non-extreme recovery.
【作者单位】：天津大学管理与经济学部;中国人民大学财政金融学院;中央财经大学管理科学与工程学院;中国科学院数学与系统科学研究院;
【基金】：国家自然基金重点资助项目(70933003);国家自然基金资助项目(70871109) 国家自然科学基金青年项目(71203247) 教育部人文社科青年基金资助项目(11YJC790015)
【分类号】：D922.281;F832.4

【参考文献】