信用风险评级与商业银行信用风险管理

【作者】 张贵清；

【导师】 刘树林；

【作者基本信息】 对外经济贸易大学 ， 国际贸易学， 2005， 博士

【摘要】 信用风险是金融机构面临的最主要的风险,随着经济全球一体化进程的加快,信用风险管理方法在最近20年取得了巨大进展。发达国家商业银行对信用风险的管理比较成熟,在实践和理论上已经形成相应的体系,信用风险分析不断尝试采用新的技术方法。相比之下,我国商业银行的信用风险管理体系不健全,管理技术较简单,远不能满足商业银行对信用风险管理的要求。而信用风险管理中信用风险评级是银行信用风险管理的核心。2002年发布的《新巴塞尔资本协议》要求银行首先建立和完善信用风险评级制度。 基于这些,本文首先系统介绍了信用风险管理的最新研究进展及国际活跃银行的管理状况,对比我国商业银行的实践,分析得出我国商业银行应首先建立信用风险评级系统。 本文以某商业银行1900个贷款样本数据为基础,分别采用聚类分析方法、Logistic回归方法和多元判别方法构建了我国商业银行的信用风险评级模型。并通过实证检验,表明三种模型都能较准确地判别信用风险等级,比目前我国商业银行实行的五级分类法能更准确地反映客户的信用风险水平,对提高我国商业银行的信用风险分析能力和信用风险管理水平提供有益的借鉴。 三种模型对全部样本和各类样本的判别准确率中,都是Logistic回归模型最高,对建模样本和检验样本的总体判别准确率分别为83.4%和75.53%,聚类方法次之,对建模样本和检验样本的总体判别准确率分别为72.05%和71.94%,多元判别方法的预测准确率最低,对建模样本和检验样本的总体判别准确率分别为68.14%和62.39%。三种模型对各类样本的判别准确趋势相同,即都是对正常类和损失类样本的判别准确率较高,对中间的三类样本判别准确率较低。 本文还根据国际上现代商业银行信用风险评级的最新理论、技术和方法,结合我国商业银行的实际状况,构建了适合我国商业银行风险管理特点的商业银行信用风险评级体系和信用风险管理体系。 与以往相关研究相比,本文有如下特点: 一 选用了1900个样本数据,大大超过其他国内相关研究选用的样本数。 二 所选样本贷款额都高于一千万元人民币,而且包括了该银行发放贷款的所有行业,银行关于大规模贷款的评估、审查工作更加详细、严格,而需要较大规模贷款的企业管理也更符合各项标准,因此,所选样本能客观反映中国商业银行客户信用的状况。 三 本论文在信用风险评级模型构建的思路、指标选取、数据选择、体系设计等方面,是基于我国商业银行的实际状况,而不仅在理论上可行。更多还原

【Abstract】 Credit risk is one of the most essential risks in financial institution. Credit risk management has evolved dramatically over the last 20 years in response to the speed-up of globalization of economy and finance. The management system of credit risk develops quickly and many kinds of new technique are used to analyze the credit risk in some developed countries .In China the simple technique is still used to manage the credit, which is not suitable for the fast-developing commercial bank. In the credit management system the key problem is the credit rating. The new Basel Accord issued in 2002 requires the bank to establish the credit rating system firstly.The main motivation of this paper is to give a systematic introduction to the latest research result in credit risk management and the management situation in the active banks, compare the practice in the Chinese banks, and propose to build the credit rating system first.Based on the data set of 1900 loans this paper builds three models of clustering, logistic regression and discrimination analysis to classify the credit risk rate. The empirical test result shows that the logistic regression model, the clustering model, and the discrimination analysis model can predict the credit rate with the overall accuracy percentage of 83.4%, 72.05%, and 68.14%, respectively. The three models all predict the best and the worst sample data more correctly than the middle three kinds of sample data. Thus, the models built in this paper can be used in the practice of the Chinese commercial banks. Based on the models this paper also designs the credit rating and the credit management system for the Chinese commercial banksThis paper differs from other research about this topic as shown below:(1) The number of the sample data is 1900 loans which is much bigger compared with other research about the same topic. (2) The loans are all more than 10 million RMB and include all the industries. The data can represent the true situation of the loans since the banks evaluate these loans more strictly and the management in the enterprises borrowing these loans is more standard. (3) This paper is based on the practice of the Chinese commercial banks in choosing the variables and the data, building the forecasting models, designing the credit rating and managing system. Thus, this paper combines the practice with the theory.更多还原