Modeling Ratings Using Discrete Choice Approach

This paper is devoted to simulations of rating migrations with discrete choice model approach. Data on changes in ratings is often characterized by abundant status quo outcomes. It was hypothesised that different macro-economic states might have an impact on the way rating agencies generate their estimations of company’s financial strength. To verify this idea a cross-nested ordered probit (CNOP) model which can account for this kind of heterogeneity in data generating process was applied to a sample of ratings migrations of 72 financial institutions. Estimation results obtained via CNOP model were tested with results of other discrete-choice model to reveal if they are significantly better with Vuong test and information criteria. Both test favoured CNOP model which proved that the proposition discussed in this paper is correct.