A Bayesian inference model for the credit rating scale

Purpose This paper aims to provide a theoretical explanation of why credit rating agencies typically disclose credit risk of issuers in classes rather than publishing the qualitative ranking those classes are based upon. Thus, its goal is to develop a better understanding of what determines the number and size of rating classes. Design/methodology/approach Investors expect ratings to be sufficiently accurate in estimating credit risk. In a theoretical model framework, it is therefore assumed that credit rating agencies, which observe credit risk with limited accuracy, are careful in not misclassifying an issuer with a lower credit quality to a higher rating class. This situation is analyzed as a Bayesian inference setting for the credit rating agencies. Findings A disclosure in intervals, typically used by credit rating agencies results from their objective of keeping misclassification errors sufficiently low in conjunction with the limited accuracy with which they observe credit risk. The number and size of the rating intervals depend in the model on how much accuracy the credit rating agencies can supply. Originality/value The paper uses Bayesian hypothesis testing to illustrate the link between limited accuracy of a credit rating agency and its disclosure of issuers’ credit risk in intervals. The findings that accuracy and the objective of avoiding misclassification determine the rating scale in this theoretical setting can lead to a better understanding of what influences the interval disclosure of major rating agencies observed in practice.