基于指标体系违约鉴别能力最大的小企业债信评级体系及实证

doi:10.16381/j.cnki.issn1003-207x.2020.06.004

中国管理科学 ›› 2020, Vol. 28 ›› Issue (6): 38-50.doi: 10.16381/j.cnki.issn1003-207x.2020.06.004

基于指标体系违约鉴别能力最大的小企业债信评级体系及实证

于善丽^1,2,3, 迟国泰¹, 姜欣¹

1. 大连理工大学经济管理学院, 辽宁大连 116024;
2. 中国人民银行金融研究所博士后科研流动站, 北京 100800;
3. 银行间市场清算所股份有限公司, 上海 200002

收稿日期:2018-03-07 修回日期:2019-03-19 出版日期:2020-06-20 发布日期:2020-06-29
作者简介:迟国泰(1955-),男(汉族),黑龙江海伦人,大连理工大学经济管理学院,金融学教授,博士,研究方向:信用风险管理、金融工程,E-mail:chigt@dlut.edu.cn.
基金资助:
国家自然科学基金资助重点项目（71731003，71431002）；国家自然科学基金资助面上项目（71873103，71971051，71971034）；国家自然科学基金资助项目（71901055，71903019）；爱德力智能科技（厦门）有限公司智能风险管控模型与算法项目（2019-01）；中国博士后科学基金资助项目（2018M641578）

Small Enterprise Facility Rating Based on the Maximum Discrimination of Indicator System

YU Shan-li^1,2,3, CHI Guo-tai¹, JIANG Xin¹

1. Faculty of Management and Economics, Dalian University of Technology, Dalian 116024, China;
2. Postdoctoral Research Station, Financial Research Institute, People's Bank of China, Beijing 100800, China;
3. Shanghai Clearing House, Shanghai 200002, China

Received:2018-03-07 Revised:2019-03-19 Online:2020-06-20 Published:2020-06-29

摘要/Abstract

摘要： 债信评级是通过评级体系确定债务违约的可能性大小。根据单个指标违约鉴别能力大小的遴选来建立评级指标体系看上去似乎是一个不错的选择，但事实并非如此。因为用违约鉴别能力强的单个指标组成的一组指标，其指标体系的违约鉴别能力却不一定强。本研究基于一组指标构成的指标体系的违约鉴别能力最大标准，构建了小企业债信评级体系。创新与特色：一是通过将单个违约鉴别力最强的指标组成的指标体系的b值，与本文建立的整体违约鉴别力最大的指标体系的b值对比，证明了单个违约鉴别力强的指标，组合起来的体系违约鉴别力不一定也强。在构建债信评级指标体系时，应关注指标体系整体的违约鉴别力，而非单个指标的违约鉴别力。二是在由n和n-1个指标构成的两组指标中，根据两组指标体系的非违约客户的综合得分S_i越高，则S_i偏离非违约状态（y_i=0）的距离越大；违约客户的综合得分S_i越低，S_i偏离违约状态（y_i=1）的距离越大；则布莱尔分数b越大，指标体系鉴别力越强的思路来遴选违约鉴别能力最大、而不是单个指标违约鉴别能力最大的一组指标体系，确保了评级体系具有最大的违约鉴别能力。三是在相关系数大于阈值的一对指标中，删除违约鉴别能力b值小、即区分违约状态能力弱的指标，既避免了指标体系的信息冗余、又避免了误删区分违约状态能力强的指标。实证表明：一是由于指标间的相互影响，单个违约鉴别力强的指标，组合起来的体系违约鉴别力不一定也强。二是非财务指标在小企业债信评级中的地位更加重要。

关键词: 信用评级, 债信评级, 指标体系遴选, 违约鉴别能力, 布莱尔分数

Abstract: Small enterprises are an important part of the national economy. However, because of their imperfect credit information and irregular management, it is difficult to grade their credit. Banks are limited by the control of credit risk and reluctant to lend or even not to lend to small enterprises, which leads to difficulties in financing and restricts the development of small enterprises. Therefore, it is very important to carry out research on credit rating of small enterprises to help alleviate the financing difficulties.Facility rating aims to estimate the default probability of a debt. It seems to be a good choice to build a rating indicator system according to discriminatory power of individual indicators, but it is not so. Because single indicator with strong discriminatory power cannot guarantee indicator system that they compose has strong discriminatory power, too.Brier score is combined with step-by-step backward elimination algorithm, and selected indicator system with strong discriminatory power between two indicator systems including n and n-1 indicators according to that the higher the scores S_i of non-default customers, the larger the distance between S_i and non-default state (y_i=0) and the lower the scores S_i of default customers, the larger the distance between S_i and default state (y_i=1), which means b value of scores will be bigger and the discriminatory power of the indicator system will be strong. What’s more, by comparing the b value of the indicator system composed of the indicator with strong discrimination with the b value of the indicator system established in this paper, it is proved that the indicator system whose indicator with strong discrimination may not have the strong discrimination, and when constructing the credit rating indicator system, we should pay more attention to the overall default discrimination of the indicator system, rather than the default discrimination of single indicator.Using the 3 045 small enterprises from a Chinese bank as empirical sample, it is shown that although the single indicator has the maximum discriminatory ability, the indicator system they compose may not have the most significant ability to discriminate default risk. And non-financial indicators play a more important role in the credit rating of small enterprises.The results of this study can be directly applied to credit rating, loan pricing and guarantee decision-making of small enterprises. Research methods and ideas can be extended to the credit rating of large and medium-sized enterprises, individuals and other objects.

Key words: credit rating, facility rating, indicator system selection, discriminatory power, Brier Score

中图分类号:

F830.56/N945.16

于善丽, 迟国泰, 姜欣. 基于指标体系违约鉴别能力最大的小企业债信评级体系及实证[J]. 中国管理科学, 2020, 28(6): 38-50.

YU Shan-li, CHI Guo-tai, JIANG Xin. Small Enterprise Facility Rating Based on the Maximum Discrimination of Indicator System[J]. Chinese Journal of Management Science, 2020, 28(6): 38-50.

参考文献

[1] 迟国泰, 张亚京, 石宝峰. 基于Probit回归的小企业债信评级模型及实证[J]. 管理科学学报, 2016, 19(6):136-156.
[2] 尹力行. 小企业数量占全球95%以上,技术将扩大普惠金融[EB/OL]. (2016-09-23)[2017-03-20]. http://www.ccstock.cn/money/licai/2016-09-23/A1474634509323.html.
[3] 赵志冲, 迟国泰. 基于似然比检验的工业小企业债信评级研究[J]. 中国管理科学, 2017, 25(1):45-56.
[4] Kruppa J,Schwarz A,Arminger G,et al. Consumer credit risk:Individual probability estimates using machine learning[J]. Expert Systems with Applications, 2013, 40(13):5125-5131.
[5] Mileris R. Macroeconomic determinants of loan portfolio credit risk in banks[J]. Engineering Economics, 2012, 23(5):496-504.
[6] Nandi J K, Choudhary N K. Credit risk management of loan portfolios by Indian banks:Some empirical evidence[J]. IUP Journal of Bank Management, 2011, 10(2):32-42.
[7] Niklis D, Doumpos M, Zopounidis C. Combining market and accounting-based models for credit scoring using a classification scheme based on support vector machines[J]. Applied Mathematics and Computation, 2014, 234(C):69-81.
[8] 沈沛龙, 周浩. 基于支持向量机理论的中小企业信用风险预测研究[J]. 国际金融研究, 2010, 2010(8):77-85.
[9] Tsai C F, Hung C. Modeling credit scoring using neural network ensembles[J]. Kybernetes, 2014, 43(7):1114-1123.
[10] 迟国泰, 陈洪海. 基于信息敏感性的指标筛选与赋权方法研究[J]. 科研管理, 2016, 37(1):153-160.
[11] Ono A, Hasumi R, Hirata H. Differentiated use of small business credit scoring by relationship lenders and transactional lenders:Evidence from firm-bank matched data in Japan[J]. Journal of Banking and Finance, 2014, 42(1):371-380.
[12] 于亮, 方志耕, 吴利丰,等. 基于灰色类别差异特性的评价指标客观权重极大熵配置模型[J]. 系统工程理论与实践, 2014, 34(8):2065-2070.
[13] Chen Yibing, Shi Yong, Wei Xianhua, et al. How does credit portfolio diversification affect banks' return and risk? Evidence from Chinese listed commercial banks[J]. Technological and Economic Development of Economy, 2014, 20(2):332-352.
[14] 石宝峰, 王静, 迟国泰. 普惠金融、银行信贷与商户小额贷款融资——基于风险等级匹配视角[J]. 中国管理科学, 2017, 25(9):28-36.
[15] Lyra M, Onwunta A, Winker P. Threshold accepting for credit risk assessment and validation[J]. Journal of Banking Regulation, 2015, 16(2):130-145.
[16] Zhi Hongyan,Yang Zhongyuan. Research on credit rating of SMEs based on combination evaluation[C]//Proceedings of 2011 International Conference on Business Management and Electronic Information, Guangzhou, China, May 13, 2011.
[17] 张洪祥, 毛志忠. 基于多维时间序列的灰色模糊信用评价研究[J]. 管理科学学报, 2011, 14(1):28-37.
[18] 葛林, 黄海峰, 王美昌. "两高"和新能源产业的绿色信贷信用风险研究——基于KMV模型的经验数据检验[J]. 数学的实践与认识, 2016, 46(1):18-26.
[19] 迟国泰, 潘明道, 齐菲. 一个基于小样本的银行信用风险评级模型的设计及应用[J]. 数量经济技术经济研究, 2014, 2014(6):102-116.
[20] 程建. 信用评级体系的定量验证研究[J]. 经济问题, 2009, 2009(1):17-21.
[21] 尼尔.J.萨尔金德. 爱上统计学[M]. 重庆:重庆大学出版社, 2010.
[22] Hernández-Orallo J, Flach P, Ferri C. A unified view of performance metrics:Translating threshold choice into expected classification loss[J]. Journal of Machine Learning Research, 2012, 13(4):2813-2869.
[23] 迟国泰, 石宝峰. 基于信用等级与违约损失率匹配的信用评级系统与方法[P]. 中国:ZL201210201461.6. 2015-08-19.
[24] 工业和信息化部、国家统计局、国家发展和改革委员会、中华人民共和国财政部. 关于印发中小企业划型标准规定的通知(工信部联企业
[2011] 300号)[R]. 工业和信息化部、国家统计局、国家发展和改革委员会、中华人民共和国财政部, 2011-06-18.
[25] 大连理工大学迟国泰课题组. 大连银行小企业贷款决策评价系统与信用等级评级系统研究[R]. 大连理工大学, 2008.

基于指标体系违约鉴别能力最大的小企业债信评级体系及实证

Small Enterprise Facility Rating Based on the Maximum Discrimination of Indicator System

PDF (PC)

可视化

摘要/Abstract

引用本文

使用本文

参考文献

相关文章 1

Metrics

本文评价

推荐阅读 10