Modelling Credit Rating Transition Matrix

Construction and analysis of credit rating migration probability matrices is an integral part of business planning and credit risk assessment of a credit institution's portfolio of borrowers. A study of S&P Agency ratings migration matrices for 2000-2019 showed that the matrices are of a mixed type (TTC-PIT) and are unstable over time. In this paper, we specify and evaluate a model for predicting changes in the probability matrices of credit rating migration with changes in the business cycle phase. The model specification utilizes Markov chain property in continuous time and applies an exogenous GDP growth variable as a predictor. The model is estimated through maximum likelihood method. Constructed model is significantly better than the model without including exogenous variable at the 5% significance level. Obtained coefficient estimates for the exogenous variable are significant. Higher annual GDP growth leads to an increase in the likelihood of rating upgrades and a decrease in the likelihood of downgrades or defaults. However, the model does not provide accurate predictions of changes in transition probabilities – the model's response to changes in GDP may be either insufficient or excessive for individual years.