Guglielmo Maria Caporale , Roman Matousek and Chris Stewart Rating Assignments : Lessons from International Banks March 2009

This paper estimates ordered logit and probit regression models for bank ratings which also include a country index to capture country-specific variation. The empirical findings provide support to the hypothesis that the individual international bank ratings assigned by Fitch Ratings are underpinned by fundamental quantitative financial analyses. Also, there is strong evidence of a country effect. Our model is shown to provide accurate predictions of bank ratings for the period prior to the 2007 – 2008 banking crisis based upon publicly available information. However, our results also suggest that quantitative models are not likely to be able to predict ratings with complete accuracy. Furthermore, we find that both quantitative models and rating agencies are likely to produce highly inaccurate predictions of ratings during periods of financial instability.