基于最小显著差异法农户关键违约特征挖掘  

Mining the key default characteristics of loan farmers based on the least significant difference method

作　　者：赵志冲 迟国泰[2] 白雪鹏 ZHAO Zhichong;CHI Guotai;BAI Xuepeng(School of Management Science and Engineering,Dongbei University of Finance and Economics,Dalian 116025,China;School of Economics and Management,Dalian University of Technology,Dalian 116024,China)

机构地区：[1]东北财经大学管理科学与工程学院,大连116025 [2]大连理工大学经济管理学院,大连116024

出　　处：《系统工程理论与实践》2020年第9期2339-2351,共13页Systems Engineering-Theory & Practice

基　　金：国家自然科学基金青年基金(71901055);国家自然科学基金重点项目(71731003);辽宁省社科规划基金项目(L18DGL007);辽宁省教育厅科学研究项目(LN2019Q66)。

摘　　要：信用风险管理不仅是简单的对客户信用资质进行排序,而是识别出客户是否违约,如果违约将会给银行造成多大程度的损失.故信用风险管理问题可以归纳为:一是根据什么标准识别客户是否违约,这个标准也就是文中的关键指标;二是客户按照某一标准划分为多类,哪一类客户将给银行造成更大的损失,这类客户的识别也就是文中的关键特征识别.上述问题解决的是在信用风险管理中,具有哪些关键特征的贷款农户是造成银行较大违约损失的"害群之马".以中国某国有商业银行分布在28个省的农户贷款数据为实证样本,通过F检验的方法甄别出居住状况,年净收入/省人均GDP这2个指标是对中国农户贷款违约损失率有显著影响的关键指标.通过最小显著差异法确定"年净收入/省人均GDP"区间在10.02~19.24内,居住状况是"共有住房"特征的贷款农户的违约风险最大,是信用风险管理中的关键特征,具有这类特征的贷款农户的违约风险最大.In the process of credit risk management,it should not only determine the ranking of credit quality of customers,but also identify the source of credit risk.This involves two problems:One is determine which index is the key indicator of default risk.Two is the multiple characteristics of the same index,determine which customers'feature is the key default characteristics that will cause greater losses to the bank.Through the data of farmers'loans in 28 provinces in China,"living conditions"and"annual net income/GDP per capita of province"were selected using the F-test which have significant influence on LGD of peasant household loans.Further,through the least significant difference method to the two key characteristics,namely"living conditions"is"shared-ownership housing",and"annual net income/GDP per capita of province"between(10.02,19.24).Peasant households with these characteristics have the greatest default losses.

关 键 词：信用风险 关键指标 关键违约特征 违约损失率 农户贷款 

分 类 号：F832.43[经济管理—金融学]