我国个人信用风险评估方法研究

发布时间：2018-09-04 09:56

【摘要】：在新中国成立后,我国建立了特色鲜明的计划经济体制,从而使得信用基础非常脆弱,个人信用体制的发展受到严重的阻碍。随着中国的改革开放,经济体制由计划经济转变为社会主义市场经济,消费信用得到很好的发展,从而信用体制以及其风险管理日益受到关注。近期《2014—2020年社会信用体系建设规划纲要》的颁布实施,作为中国第一部社会信用体系国家级建设专项规划,开启了中国社会信用体系规划建设的新篇章；同时在2015年“信用中国”网站的开通,国家平台先导工程已上线运行,接入了各省区市和37个部门,对社会信用体系发展做出阶段性成果。在国家开始高度重视信用体系发展的时期,更需要商业银行、学术界不断地开拓创新个人信用风险研究。促进利用个人信用进行消费是当今社会经济环境下扩大内需、促进经济发展的重要方法。当前中国首当其冲的任务就是大力发展经济,利用个人信用消费对国民经济的增长起到推动力的作用,但是中国经济还处于社会主义初级阶段,个人信用方面的发展会遇到很多困难阻碍,从而个人信用风险很难得到有效的控制。同时美国次贷危机,使得人们更加重视对个人信用风险的管理,因此本文研究个人信用风险评估方法更加具有现实意义。经过回顾相关的研究,对于个人信用风险评估的研究逐步从定性向定量方向发展,国内的文献往往仅局限于利用德国或澳大利亚公开信用数据库对国外研究过的信用风险评估方法进行实证改进,只是单纯的考虑信用评估方法,并没有将中国特有的国情特征作为评估指标,缺乏适合中国实际状况的评估指标体系。本文采用中国家庭金融调查中心的调研数据作为个人信用风险评估的样本数据,进一步对个人信用风险评估方法进行对比研究,从而发现更加有效的个人信用风险评估模型,促使中国个人信用风险评估指标体系更加健康快速的发展。本文主要从以下部分对国内个人信用风险评估方法进行研究。一、本文首先从三个方面介绍道德中的信用,法律中的信用,经济中的信用。对中国当前个人信用所面临的主要风险因素进行分析,即社会经济环境方面和放贷机构。从社会、经济环境方面看风险主要集中在系统性风险、利率风险、政策法律风险这三个方面。我们这里指的放贷机构主要是商业银行。从放贷机构看,目前的主要风险包括个人信用风险、流动性风险、操作风险等。放贷机构所面对的最重要的风险之一是个人信用风险。个人信用风险主要表现在债务人的违约、借款人信用等级变化等。当前个人信用风险主要表现如下：借款人的履约能力降低、借款人的还款意愿模糊、虚假按揭。当前的操作风险主要集中在：银行的贷款资格标准有所降低、银行管理体制不完善、技术水平相对落后、缺失法律依据。流动性风险主要指当前商业银行资产和负债“期限错搭”—“短存长贷”的现象,从而产生资金的流动性风险。二、本文主要研究个人信用风险,归纳了个人信用风险评估流程分为以下四个部分：(1)问题定义(2)样本数据收集及预处理；(3)建立个人信用风险评估模型；(4)模型的检验、解释及其应用；对主流的信用风险管理量化方法进行详细介绍如专家判别法、罗切斯特(logistic)回归、决策树、神经网络等方法,并比较其优缺点。三、本文根据中国国情及借鉴国内外商业银行的个人信用风险评估指标体系,最终初选出24项个人信用风险评估指标。我们将通过量化分析的方法对以上初选的24项指标进行个人信用风险识别能力的衡量,根据量化标准进一步对指标进行筛选,最终建立简单、有效的个人信用风险评估体系。对个人信用风险评估指标的识别能力进行判别：第一、通过独立样本t检验,得出5个评估指标识别个人信用风险的能力比较差,相对而言,其他的19个评估指标识别个人信用风险的能力比较强。所以我们需要将婚姻状况、其他非金融资产、活期账户存款总额、持有现金额、遵守交通规则这5个评估指标剔除出个人信用风险评估指标体系。第二、通过独立样本非参数统计检验得出,4个评估指标识别个人信用风险的能力比较差,相对而言,其他的20个评估指标识别个人信用风险的能力比较强。所以我们需要将婚姻状况、其他非金融资产、活期账户存款总额、遵守交通规则这4个评估指标剔除出个人信用风险评估指标体系。四、本文将罗切斯特(logistic)逐步回归统计方法进一步细分为Forward Stepwise罗切斯特(logistic)逐步回归和Backward Stepwise罗切斯特(logistic)逐步回归,将罗切斯特(logistic)逐步回归模型应用到个人信用风险评估。根据Forward Stepwise罗切斯特(logistic)逐步回归的结果,从个人信用风险管理的角度考虑,需要对个人信用风险评估指标体系中特别关注的是：年税后货币工资、信用卡记录、在银行已经申请的贷款项目数、住房情况、专业技术职称、政治面貌、股票账户。根据Backward Stepwise罗切斯特(logistic)逐步回归的结果,从个人信用风险管理的角度考虑,需要对个人信用风险评估指标体系中评估指标给予特别关注如下：年税后货币工资、信用卡记录、工作编制、在银行已经申请的贷款项目数、是否为农业户口、住房情况、专业技术职称、政治面貌、股票账户。五、为了更好的评估个人信用风险,我们尝试综合罗切斯特(logistic)回归分析方法和聚类分析方法的优势,本文采用了基于罗切斯特(logistic)逐步回归的聚类分析混合方法构造个人信用风险评估模型。首先运用罗切斯特(logistic)逐步回归模型进行回归来确认聚类成分,再者采用最近距离法对样本数据进行分类,最终实现个人信用的有效分类；在完成综合个人信用风险评估模型的建立后,运用ROC曲线对模型进行进一步检验。为了排除信用风险评估指标之间的含义重合对模型的不良影响,我们运用SPSS软件利用极大似然方法罗切斯特(logistic)逐步回归方法,最终经过筛选确定了9个评估指标,依次分别是政治面貌、文化程度、专业技术职称、住房情况、是否农业户口、股票账户、信用卡记录、在银行已经申请的贷款项目数、年税后货币工资。采用聚类分析进一步确定了4个聚类成分分别为政治面貌、文化程度、住房情况、信用记录。最终建立双边聚类模型,对罗切斯特(logistic)回归模型和双边聚类统计模型进行对比得出双边聚类统计模型更有效。六、结束语论述本文的主要结论及不足。
[Abstract]:After the founding of the People's Republic of China, China has established a distinctive planned economic system, which has made the credit foundation very fragile and seriously hindered the development of the personal credit system. Recently, the promulgation and implementation of the Outline of the Social Credit System Construction Plan for 2014-2020 has opened a new chapter in the planning and construction of China's social credit system as the first state-level special plan for the construction of China's social credit system, and at the same time, the opening of the "Credit China" website in 2015 has enacted the State. The platform pilot project has been put into operation on line and has been connected to provinces, municipalities and 37 departments, and has made periodic achievements in the development of social credit system. At present, China's primary task is to vigorously develop the economy and use personal credit consumption to promote the growth of the national economy. However, China's economy is still in the primary stage of socialism, and personal credit development will encounter many difficulties and obstacles. At the same time, the American subprime mortgage crisis makes people pay more attention to the management of personal credit risk. Therefore, it is more practical to study the methods of personal credit risk assessment. Domestic literatures are often limited to the empirical improvement of the credit risk assessment methods studied abroad by using German or Australian open credit databases. They only consider the credit assessment methods purely, and do not take the characteristics of China's unique national conditions as the evaluation index. They lack the assessment suitable for China's actual conditions. Indicator system. This paper uses the survey data of China Family Financial Survey Center as the sample data of personal credit risk assessment, and further makes a comparative study of personal credit risk assessment methods, so as to find a more effective personal credit risk assessment model, and promote the Chinese personal credit risk assessment index system to be healthier and faster. This paper mainly studies the domestic personal credit risk assessment methods from the following parts. First, this paper introduces the moral credit, the legal credit, and the economic credit from three aspects. From the perspective of social and economic environment, risk mainly concentrates on three aspects: systemic risk, interest rate risk, policy and legal risk. The lending institutions we refer to here are mainly commercial banks. One of the risks is personal credit risk. Personal credit risk is mainly manifested in the debtor's default, the change of the borrower's credit rating and so on. Liquidity risk mainly refers to the current phenomenon that the assets and liabilities of commercial banks are "mismatched in terms of maturity", "short-term deposit and long-term loan", thus resulting in liquidity risk of funds. Second, this paper mainly studies the personal credit risk and summarizes the personal credit risk. Risk assessment process is divided into the following four parts: (1) problem definition (2) sample data collection and preprocessing; (3) establishment of personal credit risk assessment model; (4) model testing, interpretation and application; detailed introduction of mainstream credit risk management quantitative methods such as expert discrimination, logistic regression, decision tree, God Thirdly, according to China's national conditions and the individual credit risk assessment index system of commercial banks at home and abroad, 24 individual credit risk assessment indicators are initially selected. We will weigh the individual credit risk identification ability of these 24 indicators through quantitative analysis. Quantity, according to the quantitative criteria for further screening indicators, and ultimately establish a simple and effective personal credit risk assessment system. Individual credit risk assessment indicators to identify the ability to distinguish: First, through independent sample t test, five evaluation indicators identified the ability of individual credit risk is relatively poor, relative to the other 19 So we need to exclude the personal credit risk assessment index system from the five evaluation indicators: marital status, other non-financial assets, total current account deposits, cash holdings, compliance with traffic rules. Second, through independent sample non-parametric statistical test, we get four evaluations. Assessment indicators have a poor ability to identify individual credit risk. Relatively speaking, the other 20 indicators have a strong ability to identify individual credit risk. Therefore, we need to exclude the individual credit risk assessment index system from the four evaluation indicators: marital status, other non-financial assets, total current account deposits and compliance with traffic rules. Fourthly, this paper subdivides the logistic stepwise regression method into Forward Stepwise Rochester stepwise regression and Backward Stepwise Rochester stepwise regression, and applies the logistic stepwise regression model to personal credit risk assessment. According to Backward Stepwise As a result of logistic regression, from the perspective of personal credit risk management, it is necessary to pay special attention to the evaluation index system of personal credit risk assessment as follows: annual monetary salary after tax, credit card records, work preparation, the number of loan items that have been applied for in banks, whether they are agricultural accounts, housing conditions Fifth, in order to better evaluate personal credit risk, we try to integrate the advantages of logistic regression analysis and clustering analysis. In this paper, we construct a personal credit risk assessment model based on logistic stepwise regression. Firstly, the logistic stepwise regression model is used to confirm the clustering components, and then the nearest distance method is used to classify the sample data to realize the effective classification of personal credit. After the establishment of comprehensive personal credit risk assessment model, the model is further tested by ROC curve. We use SPSS software to use the maximum likelihood method Rochester (logistic) stepwise regression method, and finally through screening to determine nine evaluation indicators, respectively, the political outlook, education level, professional titles, housing conditions, whether agricultural household registration, stocks. Accounts, credit card records, the number of loans that have been applied for in the bank, annual monetary wages after tax. Cluster analysis was used to further determine the four clustering components are political outlook, education level, housing situation, credit records. Finally, a bilateral clustering model was established, and a logistic regression model and a bilateral clustering statistical model were used. The comparison shows that the bilateral clustering statistical model is more effective. Six, concluding remarks discuss the main conclusions and shortcomings of this paper.
【学位授予单位】：西南财经大学
【学位级别】：硕士
【学位授予年份】：2016
【分类号】：F832.4

【相似文献】