农村信用合作社信用卡评分模型构建及实证研究 ——以浙江某 .doc

农村信用合作社信用卡评分模型构建及实证研究 ——以浙江省某县农村信用合作社为例 吴茂国 黄少凯 摘要：信用卡是银行或其他金融机构发行的、并给予持卡人一定信用额度、且持卡人可在信用额度内先消费后还款的一种信用凭证。运作中若存在违约行为，则影响银行资金周转，也间接降低了正常信用卡申请人的通过率。因此通过必要的发卡审核来衡量申请人的信用等级，进行选择性发卡，可有效避免损失，提升整个金融体系资金利用效率。近年来，随着我国金融体制改革的深入，商业银行大量撤出在农村的金融机构，农村信用合作社为“三农”服务的任务更加艰巨，作用更加凸显。目前农村信用合作社信用卡业务作为一项扶持农业经济的金融服务，在我国并未得到足够的重视。农村信用合作社由于信用卡发卡失当出现的呆账、坏账等问题对农村金融体系的可持续运作造成了不利影响。 本文以浙江省某县农村信用合作社信用卡发卡失当为例，建立一个完整的信用卡发卡审核体系，以降低各类坏账问题的出现概率。本文首先介绍了发卡审核即信用评分的概念、运用过程及隐含假设、工作原理和工作假设，并分析了目前信用评分模型的关键所在和存在的一些问题。其次在国内外相关理论的基础上，总结了建立发卡审核模型即信用评分模型的主要方面。接着通过对借款人现状和XX农村信用合作社现状的分析，选取了16个特征大类48个具体的特征变量，据此进行样本选择、指标设计、描述性分析、回归模型建立及检验、信用评分模型建立及测试。该模型对预测“正常违约”客户的准确性可以分别达到80%和80.13%，并可以在拒绝16.67%正常客户的情况下将发卡风险控制在3.22%以内。最后为XX农村信用合作社设计了具体信用卡发卡模型，并为国家“三农”政策的实施提供了相应的政策建议。 关键词：农村信用合作社；信用卡发卡审核；信用评分；Logistic回归；权重分析 ABSTRACT This paper introduces the technique of evaluation of credit card issuing based on Logistic regression, its application process, implicit assumptions, working principle, and working hypothesis. It discusses some of the key issues prevalent in credit scoring models. Besides, it reviews a partial selection of related literature that pertains to evaluation of credit card issuing. It summarizes techniques of credit scoring models, elaborates the basic principles and assumptions of Logistic regression. Using data gathered from a countryside credit cooperative. Regression and robustness tests are conducted to predict the probability of paying back the debt on time and assess the overall significance of the regression function. This paper proposes a new credit scoring model. After probability is predicted, credit score of each customer is calculated. The predicted results are compared with the actual ones. It reveals that 80% of the customers who are issued credit cards pay back their debt on time. The risk of the credit scoring model is refined to be within 3.22%. Policy suggestions are put forward afterwards. This paper contributes to existing literature on credit card scoring models and credit evaluation. K