This introduction places in context the papers on credit risk modelling contained in the special issue. We explain why credit risk modelling has become such a focus of interest for practitioners and financial supervisors. Even though, as we explain, the current modelling technologies have significant weaknesses, they offer the possibility of major changes in the ways banks are managed and regulated. The main impediment to greater use of these models, especially by regulators, is the difficulty involved in back-testing the risk measures they produce. We suggest some thoughts on how back-testing and other types of model assessment might be performed.

The distribution of ratings changes plays a crucial role in many credit risk models. As is well-known, these distributions vary across time and different issuer types. Ignoring such dependencies may lead to inaccurate assessments of credit risk. In this paper, we quantify the dependence of rating transition probabilities on the industry and domicile of the obligor, and on the stage of the business cycle. Employing ordered probit models, we identify the incremental impact of these factors. Our approach gives a clearer picture of which conditioning factors are important than comparing transition matrices estimated from different sub-samples.