Optimal method of credit rating division based on maximum credit difference degree

Credit rating has an extremely important impact on modern society. It will mislead the creditors and social public if the credit rating division is unreasonable. In 2011, Standard & Poor's lowered the sovereign credit rating of United States to AA+ from AAA which caused the turmoil in the global financial markets. The essence of credit rating is to classify the customers according to their credit level which means customers with different credit risk level should be included in different credit rating. The internationally popular credit rating agencies like Moody, often has the unreasonable phenomenon that customers with higher loss given default (LGD) while the credit level not low for China's loan customers. Our research constructs nonlinear programming model to divide the credit rating according to the LGD pyramid and maximum credit difference degree as the standard, and then we make an empirical research with the loan data of a bank in China. The special and contributions of this paper lie in three aspects: Firstly, we build up a nonlinear programming model to divide the credit rating with the objective function that the sum of credit score difference ∑(Pmkk - P1k+1) is maximum, which ensure customers with different credit status are more likely to be divided into different credit level, we can avoid the unreasonable phenomenon that customers with big credit status difference are divided into the same level. Secondly, we construct a nonlinear programming model to divide the credit rating with the constraint that the LGD is strictly increasing with credit rating from high to low, which can meet the pyramid standard that customers with lower LGD should be divided in higher level, we can avoid the unreasonable phenomenon that customers with higher LGD while the credit level not low. Thirdly, we make an empirical study with 1814 small business loan data of a Chinese commercial bank in recent 20 years and its research result indicates that the method of credit rating division in this paper not only meet the pyramid standard that customers with lower LGD should be divided in higher level, but also own the advantage that it can ensure the customers with different credit status are divided into the different level. © 2017, Editorial Board of Journal of Systems Engineering Society of China. All right reserved.