基于LM检验的小型工业企业债信评级模型及实证
|
摘要
债信评级是通过建立评级指标体系和评价方程来评价客户一笔债务信用等级的高低及债务违约损失的概率。它是信用风险管理极为重要的环节。小型工业企业财务信息更难获取,数据的真实性更难考证,导致小型工业企业债信评级更难。本文基于LM检验建立了小型工业企业债信评级模型。本文的创新与特色:一是通过把每一个待遴选的指标、与前面已选定的j(j=0,1,2,…m)个指标共同组成的j+1个指标组的所有数值、代入表示企业违约概率的拉格朗日函数,得到j个拉格朗日统计量LM_j(j=0,1,2,…m);然后对这j个拉格朗日统计量LM_j进行c~2检验,在所有通过c~2检验的统计量LM_j中、也即具有显著区分违约状态能力的指标中,遴选LM_j值最大的指标为评级指标,保证了遴选的指标对鉴别全部企业的违约状态具有显著影响。二是通过在相关系数大于0.7的一对指标中,删除均方差数值小的指标,既避免了指标体系的信息冗余、又避免了误删信息含量大的指标。三是对中国某商业银行1814笔小型工业企业贷款借据数据为样本的实证表明:行业景气指数等宏观因素;法人代表贷款违约记录等个人因素,未偿还贷款总额占资产总额比等信用参数组成的20个指标评级体系,能够显著区分小型工业企业的违约状态。
Facility rating evaluates the creditworthiness and the LGD(Loss Given Default) of debt by establishing indicator system and evaluation equation. It plays an extremely important role in credit risk management. However, the difficulty in getting financial information of small industrial enterprises and validating the authenticity of the data makes credit rating harder. Taking 1814 small industrial enterprises loans from a commercial bank of China as empirical samples, this paper establishes credit rating for small industrial enterprises based on LM test.First, by reviewing the international literature and credit rating system from various international financial agencies(Standard & Poor's etc.), we selected 107 small industrial enterprises credit rating indicators, and deleted 26 indicators from them and then standardize the remaining 81 indicators. Second, we removed indicators with smaller mean squared error between indicators whose the correlation coefficient is greater than 0.7, which reduces the information redundancy of facility rating indicator system and avoids eliminating indicators with more information content. Third, according to the criterion that the indicator with bigger discriminatory power should be reserved, we substituted j+1 indicators into Lagrangian function, which is composed of one undetermined indicator and the previous selected j(j = 0,..., m) numerical indicators, and selected the indicator xj whose LM_j is the maximum and passes c2 test. By this way, we can ensure that the selected indicator xj has a significant discriminatory power. Four, we weighed indicators by mean squared error method according to the principle that the indicator with more information should be endowed with a larger weight, and computed customer's facility score based on the linear weighted method. Last, according to the principle that customers with higher credit rating should correspond to lower LGD, this paper finally obtains 9 credit ratings and calculates the LGD of each credit rating. This paper constructs facility rating indicator system for small industrial enterprises. The system includes 20 indicators, including macroeconomic, personal and credit factors. Macroeconomic factors involve industry climate index, etc. Entrepreneurs' personal factors involve legal representative default record etc. Credit parameters involve outstanding loans accounted for total assets ratio. Furthermore, non-financial factors are more important than financial factors in the indicator system, which can be shown by the empirical result that total weight of financial factors is 0.188, whose quantity accounts for 30% of total quantity. The total weight of non-financial factors is 0.812, whose quantity accounts for 70%. What's more, this paper obtains 9 credit ratings, which satisfy the principle that customers with higher credit rating should correspond to lower LGD, and calculates the LGD of each credit rating to provide loan decision for banks. Facility rating indicator system must have significant discriminatory power. In addition, for small industrial enterprises' facility rating, we should pay more attention to non-financial factors. Moreover, the credit rating must satisfy the criterion that customer with higher credit rank should correspond to lower LGD.
Facility rating evaluates the creditworthiness and the LGD(Loss Given Default) of debt by establishing indicator system and evaluation equation. It plays an extremely important role in credit risk management. However, the difficulty in getting financial information of small industrial enterprises and validating the authenticity of the data makes credit rating harder. Taking 1814 small industrial enterprises loans from a commercial bank of China as empirical samples, this paper establishes credit rating for small industrial enterprises based on LM test.First, by reviewing the international literature and credit rating system from various international financial agencies(Standard & Poor's etc.), we selected 107 small industrial enterprises credit rating indicators, and deleted 26 indicators from them and then standardize the remaining 81 indicators. Second, we removed indicators with smaller mean squared error between indicators whose the correlation coefficient is greater than 0.7, which reduces the information redundancy of facility rating indicator system and avoids eliminating indicators with more information content. Third, according to the criterion that the indicator with bigger discriminatory power should be reserved, we substituted j+1 indicators into Lagrangian function, which is composed of one undetermined indicator and the previous selected j(j = 0,..., m) numerical indicators, and selected the indicator xj whose LM_j is the maximum and passes c2 test. By this way, we can ensure that the selected indicator xj has a significant discriminatory power. Four, we weighed indicators by mean squared error method according to the principle that the indicator with more information should be endowed with a larger weight, and computed customer's facility score based on the linear weighted method. Last, according to the principle that customers with higher credit rating should correspond to lower LGD, this paper finally obtains 9 credit ratings and calculates the LGD of each credit rating. This paper constructs facility rating indicator system for small industrial enterprises. The system includes 20 indicators, including macroeconomic, personal and credit factors. Macroeconomic factors involve industry climate index, etc. Entrepreneurs' personal factors involve legal representative default record etc. Credit parameters involve outstanding loans accounted for total assets ratio. Furthermore, non-financial factors are more important than financial factors in the indicator system, which can be shown by the empirical result that total weight of financial factors is 0.188, whose quantity accounts for 30% of total quantity. The total weight of non-financial factors is 0.812, whose quantity accounts for 70%. What's more, this paper obtains 9 credit ratings, which satisfy the principle that customers with higher credit rating should correspond to lower LGD, and calculates the LGD of each credit rating to provide loan decision for banks. Facility rating indicator system must have significant discriminatory power. In addition, for small industrial enterprises' facility rating, we should pay more attention to non-financial factors. Moreover, the credit rating must satisfy the criterion that customer with higher credit rank should correspond to lower LGD.
引文
[1]工业和信息化部,国家统计局,发展改革委,财政部.中小企业划型标准规定[EB/OL].工业和信息化部网站,http://www.gov.cn/zwgk/2011-07/04/content_1898747.htm,2011-6-18.
[2]Dun and Bradstreet.Dun and Bradstreet Rating[EB/OL].http://www.dandb.com/glossary/d-b-rating/,2014.
[3]徐广军,倪晓华,肖运香.标普、穆迪、邓白氏企业信用评级指标体系比较研究[J].浙江金融,2007,(3):51-52.
[4]中国工商银行.关于印发《中国工商银行小企业法人客户信用等级评定办法》[R].中国工商银行,工银甬[2005]185号.
[5]中国建设银行.关于印发《中国建设银行小企业客户信用评级办法(试行)》[R].中国建设银行,建总发[2009]101号.
[6]中国银行.关于印发《国内机构公司客户信用评级管理办法》[R].中国银行,中银险[2010].
[7]中国人民银行征信中心与金融研究所联合课题组,纪志宏,王晓明,曹凝蓉等.互联网信贷、信用风险管理与征信[J].金融研究,2014,10:133-147.
[8]黄飙.网络时代的产业链中小贷-网络新金融系列专题报告之二:阿里小贷与线上供应链金融[R].北京:长城证券,2013.4-8.
[9]Maik Dierkes,Carsten Erner,Thomas Langer,et al.Business credit information sharing and default risk of private firms[J].Journal of Banking&Finance,2013,37(8):2867-2878.
[10]迟国泰,潘明道,程砚秋.基于综合判别能力的农户小额贷款信用评价模型[J].管理评论,2015,06:42-57.
[11]邓超,胡梅梅,曾文潮.基于贝叶斯界定折叠法的小企业信用评分模型研究[J].管理工程学报,2015,04:162-170.
[12]范柏乃,朱文斌.中小企业信用评级指标的理论遴选与实证分析[J].科研管理,2003,06:83-88.
[13]Mu Zhang,Ying He,Zong-fang Zhou.Study on the Influence Factors of High-Tech Enterprise Credit Risk:Empirical Evidence from China's Listed Companies[J].Procedia Computer Science,2013,17:901-910.
[14]刘澄,胡巧红,孙莹,武鹏,黄翔.基于Vague集理论的商业银行客户信用等级测定[J].统计与决策,2013,04:43-45.
[15]张目,周宗放.改进的云重心评判法在高技术企业信用评级中的应用[J].数学的实践与认识,2010,19:37-44.
[16]Dimitrios Niklis,Michael Doumpos,Constantin Zopounidis.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(15):69-81.
[17]Nebojsa Nikolic,Nevenka,Iva Joksimovic.The application of brute force logistic regression to corporate credit scoring models:Evidence from Serbian financial statements[J].Expert Systems with Applications,2013,15(40):5932-5944.
[18]Zarkic-Joksimovic,Djordje Stojanovski,Antonio Blanco,et al.Credit scoring models for the microfinance industry using neural networks:Evidence from Peru[J].Expert Systems with Applications,2013,40(1):356-364.
[19]尼尔.J.萨尔金德.爱上统计学[M].重庆:重庆大学出版社,2010.
[20]巴克豪斯等.多元统计分析方法[M].上海:格致出版社,2009.
[21]张定胜.计量经济学[M].武汉:武汉大学出版社,2000.
[22]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:CN102779317B,2015-08-19.
[23]黄海等.spss10.0 for Windows统计分析[M].北京:人民邮电出版社,2001.
[24]徐秀渠.Altman’s Z-Score模型在企业风险管理中的应用研究[J].经济经纬,2010,04:103-106.
[2]Dun and Bradstreet.Dun and Bradstreet Rating[EB/OL].http://www.dandb.com/glossary/d-b-rating/,2014.
[3]徐广军,倪晓华,肖运香.标普、穆迪、邓白氏企业信用评级指标体系比较研究[J].浙江金融,2007,(3):51-52.
[4]中国工商银行.关于印发《中国工商银行小企业法人客户信用等级评定办法》[R].中国工商银行,工银甬[2005]185号.
[5]中国建设银行.关于印发《中国建设银行小企业客户信用评级办法(试行)》[R].中国建设银行,建总发[2009]101号.
[6]中国银行.关于印发《国内机构公司客户信用评级管理办法》[R].中国银行,中银险[2010].
[7]中国人民银行征信中心与金融研究所联合课题组,纪志宏,王晓明,曹凝蓉等.互联网信贷、信用风险管理与征信[J].金融研究,2014,10:133-147.
[8]黄飙.网络时代的产业链中小贷-网络新金融系列专题报告之二:阿里小贷与线上供应链金融[R].北京:长城证券,2013.4-8.
[9]Maik Dierkes,Carsten Erner,Thomas Langer,et al.Business credit information sharing and default risk of private firms[J].Journal of Banking&Finance,2013,37(8):2867-2878.
[10]迟国泰,潘明道,程砚秋.基于综合判别能力的农户小额贷款信用评价模型[J].管理评论,2015,06:42-57.
[11]邓超,胡梅梅,曾文潮.基于贝叶斯界定折叠法的小企业信用评分模型研究[J].管理工程学报,2015,04:162-170.
[12]范柏乃,朱文斌.中小企业信用评级指标的理论遴选与实证分析[J].科研管理,2003,06:83-88.
[13]Mu Zhang,Ying He,Zong-fang Zhou.Study on the Influence Factors of High-Tech Enterprise Credit Risk:Empirical Evidence from China's Listed Companies[J].Procedia Computer Science,2013,17:901-910.
[14]刘澄,胡巧红,孙莹,武鹏,黄翔.基于Vague集理论的商业银行客户信用等级测定[J].统计与决策,2013,04:43-45.
[15]张目,周宗放.改进的云重心评判法在高技术企业信用评级中的应用[J].数学的实践与认识,2010,19:37-44.
[16]Dimitrios Niklis,Michael Doumpos,Constantin Zopounidis.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(15):69-81.
[17]Nebojsa Nikolic,Nevenka,Iva Joksimovic.The application of brute force logistic regression to corporate credit scoring models:Evidence from Serbian financial statements[J].Expert Systems with Applications,2013,15(40):5932-5944.
[18]Zarkic-Joksimovic,Djordje Stojanovski,Antonio Blanco,et al.Credit scoring models for the microfinance industry using neural networks:Evidence from Peru[J].Expert Systems with Applications,2013,40(1):356-364.
[19]尼尔.J.萨尔金德.爱上统计学[M].重庆:重庆大学出版社,2010.
[20]巴克豪斯等.多元统计分析方法[M].上海:格致出版社,2009.
[21]张定胜.计量经济学[M].武汉:武汉大学出版社,2000.
[22]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:CN102779317B,2015-08-19.
[23]黄海等.spss10.0 for Windows统计分析[M].北京:人民邮电出版社,2001.
[24]徐秀渠.Altman’s Z-Score模型在企业风险管理中的应用研究[J].经济经纬,2010,04:103-106.