Credit rating model development: An ordered analysis based on accounting data

In this paper we propose and test a methodology for constructing a credit rating model. We follow a polytomous ordered probit analysis leading to the specification of statistically significant credit rating intervals. We test our model with accounting data of Greek listed firms over the years 2004-2013, a period which includes both the pre-crisis growth and the crisis phase of the Greek economy and the stock market. Using the empirically-based rating categories that the model generates endogenously, we observe not only a clear and timely response of ratings to the changing economic environment, but we also obtain significant predictive ability over a period of one, two and three years.