基于偏最小二乘法的商业银行信用风险压力测试研究

发布时间：2018-06-26 09:47

本文选题：商业银行 + 信用风险　；参考：《湖南大学》2012年硕士论文

【摘要】：已有的商业银行信用风险压力测试模型存在设计缺陷，变量选取上受到诸多限制，只能选取较少的变量，对于影响商业银行信用风险的宏观因素考虑不足。事实上，商业银行的信用风险通常受多个变量共同影响，为突破原有模型对选取变量的限制，本文采用偏最小二乘法替代了原似无关回归法，对信用风险压力测试模型中的传导模型进行了改进。本文将我国股份制商业银行2005~2010年季度不良贷款率作为信用风险压力测试模型的承压指标，选取了国内生产总值增长率、固定投资增长率、进出口贸易总额增长率等十个宏观因素作为压力指标。利用统计分析等方法对这些指标进行处理，构建不良贷款率与宏观因素之间的压力传导模型。分别利用似无关回归法与偏最小二乘法对传导模型进行估计，对比结果发现：在变量间存在相关性时，使用似无关回归法会出现系数检验无法通过、符号异常等问题，而选用偏最小二乘法则很好地解决了这一问题。传导模型构建完成后，本文采用蒙特卡洛模拟法，模拟宏观经济变量自回归方程的残差项变动，通过残差项求得压力冲击情景，构建商业银行信用风险压力测试情景模型。模型结果显示，在实际国民生产总值下降的冲击情景下，银行不良贷款率显著上升。我国商业银行运用基于偏最小二乘法的压力测试模型开展信用风险压力测试，，需要加强对相关宏观经济变量的监测，建立及时有效的反馈报告机制。
[Abstract]:The existing commercial bank credit risk stress test model has the design flaw, the variable selection is subjected to many restrictions, can only select the relatively few variables, for the commercial bank credit risk macroscopic factor consideration is insufficient. In fact, the credit risk of commercial banks is usually affected by multiple variables. In order to break through the limitation of the original model on the selection of variables, the partial least square method is used to replace the prima facie independent regression method. The conduction model of credit risk stress test model is improved. In this paper, the non-performing loan ratio of China's joint-stock commercial banks during the period of 2005 ~ 2010 is regarded as the bearing index of credit risk stress test model, and the GDP growth rate and the fixed investment growth rate are selected. Total growth rate of import and export trade and other ten macro factors as pressure indicators. By means of statistical analysis and other methods, the pressure conduction model between non-performing loan rate and macro factors is constructed. The conduction model is estimated by using the method of quasi-independent regression and partial least square, respectively. The comparison results show that when there is correlation between variables, the coefficient test can not pass by using the method of seemingly independent regression, the sign is abnormal, and so on. The partial least square method is a good solution to this problem. After the construction of the conduction model, Monte Carlo simulation method is used to simulate the variation of residual terms in the autoregressive equation of macroeconomic variables. Through the residual terms, the pressure impact scenarios are obtained, and the credit risk pressure test scenario model of commercial banks is constructed. The results show that the ratio of non-performing loans (NPLs) of banks increases significantly under the impact of the decline of real gross national product (GNP). In order to carry out credit risk stress testing by using the stress test model based on partial least square method, commercial banks in China need to strengthen the monitoring of relevant macroeconomic variables and establish a timely and effective feedback reporting mechanism.
【学位授予单位】：湖南大学
【学位级别】：硕士
【学位授予年份】：2012
【分类号】：F832.33;O241.5

【参考文献】