基于风险等级匹配和普惠金融双重约束的农户小额贷款信用评级研究

Credit rating of farmers’ microfinance is an issue of credit risk management, and is also related to the economic and social stability. The existing researches divided customers’ credit ratings using loan customers’ credit scores or their probability defaults (PD). These credit rating results did not consider the real loss of loan customers, nor did they possess the function of lending decision. In order to overcome the above deficiencies, this project proposes a novel credit rating method based on double constraints of credit risk-rating match-up and the inclusive finance. First of all, the approximate accuracy classification coefficients between the evaluation indicators and the customers’ default status can be obtained utilizing neighborhood rough set method. These coefficients would help deleting the indicators which can not divide the default farmers and the non-default farmers. Secondly, according to the idea that the stronger of the indicator’s default status identification ability, the bigger the weight would be, the credit scoring model which can significantly distinguish farmers’ default status can be constructed utilizing F-test of variance homogeneity. Lastly, the study will establish a nonlinear programming credit rating model consisting of the credit risk-rating match-up constraint condition that the credit rating increases with the decreasing Loss Given Default (LGD) and the objective function, i.e. the maximum number of farmers to get loans on the premise of achieving banks’ goals profits. This research can not only ensure that the credit rating result meets the credit risk-rating match-up standard, but also can guarantee that the number of farmers to get loans on the premise of achieving banks’ goals profits is maximal. It reflects the inclusive finance concept that credit funds benefit more farmers. And, the project will open up a novel idea for farmers’ credit risk management.

农户小额贷款信用评级既是信用风险管理问题，也事关经济社会稳定。本项目针对现有研究仅立足于贷款客户信用得分或违约概率来划分其信用等级，评级结果没有考虑客户真实的违约损失、也不具有贷款决策功能的弊端，提出了基于风险等级匹配和普惠金融双重约束的农户信用评级方法。通过邻域粗糙集求解评价指标与违约状态之间的近似精度分类系数，剔除对农户违约、非违约两种状态分类不显著的指标。根据指标对农户违约状态鉴别能力越强、权重越大的思路，利用方差齐性检验构建能显著区分农户违约状态的信用评分模型。以信用等级越高、违约损失率越低的风险等级匹配标准为约束，以银行目标利润临界点以上贷款农户数最多为目标函数，构建信用评级模型，确定农户的信用等级。既可保证信用等级划分结果满足风险等级匹配标准，也能确保银行在实现目标利润的前提下、发放贷款的农户数最多，体现信贷资金惠及更多农户的“普惠金融”理念，开拓农户信用风险管理的新思路。

农户小额贷款信用评级既是信用风险管理问题，也事关我国经济社会稳定。本项目针对现有研究仅立足于贷款农户信用得分或违约概率来划分其信用等级，评级结果没有考虑客户真实的违约损失、也不具有贷款决策功能的弊端，从农户信用等级越高、违约损失率越低的风险等级匹配标准和目标利润临界点之上贷款农户数最多两个视角出发，对农户小额贷款信用评级问题进行了深入研究。通过小额贷款信用风险评价指标遴选→信用得分求解→信用等级划分，构建了农户小额贷款信用评级体系，揭示了评价指标与农户违约状态之间的规律性联系。一是利用利用模糊C均值-模糊粗糙集相结合的方法，构建了农户信用评级指标约简模型，剔除了给定精度下，对农户信用评级分类影响较小的指标，保证筛选后保留的指标能显著判别农户的违约状态。二是由于评价指标之间存在相互替代性，即权重大、得分低的指标和权重小、得分高的指标具有相同的信用得分，这无疑会对评价结果的可靠性造成影响。借鉴选择消去与选择转换评价(ELECTRE III)和偏好顺序结构评估方法(PROMETHEE-II)，构建小额贷款信用评分模型，测度贷款客户的信用状况。三是对已结清债务的贷款农户(存在应收未收本息、应收本息数据)，以风险等级匹配标准和农户违约损失率LGDk求解的等式为约束条件，以银行目标利润临界点以上等级的贷款农户最多为目标函数，构建小额贷款信用等级划分模型，划分贷款客农户的信用等级，体现了信贷资金惠及更多农户的普惠金融理念。在理论上，本项目同时将普惠金融、风险等级匹配、违约损失显著判别纳入信用评级体系中，为小额贷款信用评级开拓了新思路。在实践中，本项目研究成果获批国家发明专利，先后被西安微电机研究所、中国普惠金融研究院采纳，实现了从理论到实践的推广转化。