基于PROMETHEE-II的商户小额贷款信用评级模型及实证

截至2014年底,中国注册个体工商户为4984.06万户,个体私营经济吸纳社会从业人员已达2.5亿人,加上中国商户小额贷款对象的分散性、财务信息不健全等特点和难点,商户小额贷款信用评级体系极不完善,甚至绝大多数银行都没有建立这个体系。本文通过相关分析剔除反映信息重复的指标,通过显著性判别遴选对商户违约状态影响显著的指标,建立了能显著区分商户违约状态的小额贷款信用评级指标体系。在此基础上,结合PROMETHEE-II(偏好顺序结构)和聚类分析方法,构建了商户小额贷款信用评级模型,并对中国某国有商业银行2157个商户小额贷款样本进行了实证。本文创新与特色:一是通过将偏好顺序结构评估法(PROMETHEE-II)引入商户小额贷款信用评级,构建了基于PROMETHEE-II的小额贷款信用评分模型,求解商户的净流量信用得分Φ(a),揭示了商户a与其余商户、评价指标间的相互作用对评价结果的影响,避免了现有研究由于评价指标之间的相互替代性、严重影响评价结果可靠性的不足。二是借鉴模糊聚类“数据越集中、越应该被分为一类”的思想,采用R聚类对商户信用得分进行分类;进而采用K-W检验,对分类数目l进行非参数检验,确定商户的信用等级。既保证了不同等级商户在信用得分数值上存在显著差异,也确保了不同等级商户能反映不同的信用特征;同时,也避免了现有利用信用得分区间、违约概率阈值或客户数分布方法划分信用等级时,得分区间、违约概率阈值或客户数分布分位点人为主观确定的不足。三是实证研究表明,影响商户小额贷款信用风险的重要性排序依次为:X₃偿债能力>X₁基本情况>X₆宏观环境>X₅营运能力>X₂保证联保>X₄盈利能力。

A Credit Rating Model of Microfinance Loans for Small Private Business Based on PROMETHEE-II and Its Empirical Study

Until the end of year 2014, the number of privately-owned business had reached more than 49 million, and individual private business had absorbed 250 million social workers in China. Due to distribution sparsity and imperfect financial information of Chinese small private businesses, the credit rating system for small private business badly demands improvement. In this study, the credit rating index system of microfinance loans for small private business is established combining correlation analysis and significance discriminant. And then, a credit rating model of microfinance loans for small private business based on PROMETHEE-II and cluster analysis arithmetic is established using the data of 2157 customers in a Chinese national commercial bank. The contribution of this study is three-folded. Firstly, by introducing the method of PROMETHEE-II into credit rating of microfinance loans for small private business, a credit rating model based on PROMETHEE-II is constructed. Using the proposed model the net flow credit score Φ(a)for the small private business a is computed to reflect the impact of relative importance of the small private business a to the rest businesses and interaction among indicators on evaluation results. The trade-off of evaluation indicators which could seriously affect the reliability of results could be overcome. Secondly, by K-W non-parameter test of credit rating results, which are obtained from R cluster analysis, significant difference in the credit scoring values as well as credit characteristic difference of businesses in various levels could be guaranteed. The subjective determination of credit scoring interval and threshold of default probability could be eliminated. Thirdly, the empirical results show that the order of importance affecting the credit risk of microfinance loans for small private business is as follows: X₃ capacity of repaymentX₁ basic information X₆ microenvironment X₅ capacity of operation X₂ guarantee and joint guarantee X₄ capacity of profitability.