基于半参数方法进行拒绝推断的信用评级模型

›› 2018, Vol. 30 ›› Issue (10): 40-48.

基于半参数方法进行拒绝推断的信用评级模型

夏利宇^1,2, 何晓群²

1. 国网能源研究院有限公司, 北京 102209;
2. 中国人民大学应用统计科学研究中心, 北京 100872

收稿日期:2016-06-17 出版日期:2018-10-28 发布日期:2018-10-23
通讯作者: 何晓群(通讯作者),中国人民大学应用统计科学研究中心教授,博士生导师
作者简介:夏利宇,国网能源研究院有限公司研究员,中国人民大学应用统计科学研究中心,博士
基金资助:
教育部人文社会科学重点研究基地重大项目（15JJD910002）。

Reject Inference in Credit Score Model Based on Semi-parametric Method

Xia Liyu^1,2, He Xiaoqun²

1. State Grid Energy Research Institute, Beijing 102209;
2. Center for Applied Statistics, Renmin University of China, Beijing 100872

Received:2016-06-17 Online:2018-10-28 Published:2018-10-23

摘要/Abstract

摘要：

拒绝推断可视为因变量非随机缺失问题的特例，它处理信用评级建模中由于被拒客户的信用表现未知，样本偏差导致的参数估计有偏问题。本文基于Kim和Yu 2011年提出的非随机缺失下均值泛函的半参数估计模型，提出处理拒绝推断的迭代半参数法。运用此方法在5类缺失情形下进行模拟研究，并对Australian数据和中国某银行的征信数据进行实证研究。结果表明，与常用方法相比，迭代半参数法可以有效地识别被拒绝申请者中的"坏"客户，降低金融机构的违约风险，是一种相对保守的方法。

关键词: 信用评级模型, 拒绝推断, 半参数估计, 非随机缺失

Abstract:

Reject inference can be considered a specific case of non-ignorable missing data analysis. It can deal with the biased estimation of model parameters caused by sample bias in credit score model resulting from absence of the credit quality of rejected applicants. Based on the semi-parametric regression model of mean function with non-ignorable missing response proposed in Kim and Yu (2011), we propose an iterative semi-parametric method to infer credit-quality data with non-ignorable missing mechanism. Simulation studies in 5 missing scenarios are implemented and empirical studies with Australian dataset and bank of C dataset are conducted. The results demonstrate that our method is a relatively conservative approach that can effectively identify the "bad" applicants and reduces credit risk faced by financial institutions.

Key words: credit score model, reject inference, semi-parametric estimation, non-ignorable missing

夏利宇, 何晓群. 基于半参数方法进行拒绝推断的信用评级模型[J]. , 2018, 30(10): 40-48.

Xia Liyu, He Xiaoqun. Reject Inference in Credit Score Model Based on Semi-parametric Method[J]. , 2018, 30(10): 40-48.

参考文献

[1] Rubin D. B. Inference and Missing Data[J]. Biometrika, 1976,63(3):581-592
[2] Boyes W., Hoffman D., Low S. An Econometric Analysis of the Bank Credit Scoring Problem[J]. Journal of Econometric, 1989,40(1):3-14
[3] Greene W. Sample Selection in Credit-scoring Models[J]. Japan and the World Economy, 1998,10(3):299-316
[4] Jacobson T., Roszbach K. F. Bank Lending Policy, Credit Scoring and Value at Risk[J]. Journal of Banking Finance, 2003,27(4):615-633
[5] Crook J., Banasik J. Does Reject Inference Really Improve the Performance of Application Scoring Models[J]. Journal of Banking and Finance, 2004,28(4):857-874
[6] Banasik J., Crook J. Reject Inference, Augmentation and Sample Selection[J]. European Journal of Operational Research, 2007,183(3):1582-1594
[7] Banasik J., Crook J. Reject Inference in Survival Analysis by Augmentation[J]. Journal of the Operational Research Society, 2010,61(3):473-485
[8] Heckman J. J. Sample Selection Bias as a Specification Error[J]. Econometrica, 1979,47(1):153-161
[9] 张景肖,魏秋萍,姜玉霞,等. 基于两阶段思想处理拒绝推断的信用评分模型[J]. 数理统计与管理, 2012,31(6):1049-1060
[10] Sohn S. Y., Shin H. W. Reject Inference in Credit Operations Based on Survival Analysis[J]. Expert Systems with Applications, 2006,31(1):26-29
[11] 魏秋萍,张景肖,张波. 基于核函数法进行拒绝推断的信用评分模型[J]. 统计与决策, 2012,12(12):4-8
[12] Chen G. G., Astebro T. Bound and Collapse Bayesian Reject Inference for Credit Scoring[J]. Journal of the Operational Research Society, 2012,63(10):1374-1387
[13] Kim J. K., Yu L. C. A Semi-parametric Estimation of Mean Functionals with Non-ignorable Missing Data[J]. Journal of the American Statistical Association, 2011,106(493):157-165
[14] Zhao P. Y., Tang M. L., Tang N. S. Robust Estimation of Distribution Functions and Quantiles with Non-ignorable Missing Data[J]. The Canadian Journal of Statistics, 2013,41(4):575-595