This study presents an alternative way of estimating credit transition matrices using a hazard function model. The model is useful both for testing the validity of the Markovian assumption, frequently made in credit rating applications, and also for estimating transition matrices conditioning on firm-specific and macroeconomic covariates that influence the migration process. The model presented in the paper is likely to be useful in other applications, though we would hesitate to extrapolate numerical values of coefficients outside of our application. Transition matrices estimated this way may be an important tool for a credit risk administration system, in the sense that with them a practitioner can easily forecast the behavior of the clients ‘ratings in the future and their possible changes of state.
