SUMMARY
The paper proposes a methodology to assess the ratings attributed to the customers of a financial institution. The process is based on the use of a posteriori likelihood distributions developed from the concepts of Bayesian statistics. The banking practice has lacked a more effective ex-post assessment of the credit risk grades (which are often called score or classification) ascribed by independent risk assessment companies. The lack of this information has certainly had a negative impact on the analysis of the consistency between the grades given to potential borrowers and their actual performance and may have caused biased results. The method proposed here is designed to fill this gap. It intends to check whether the a priori ratings are really adequate to the specific customer group of a bank. Thus, the article aims at contributing to credit risk assessment by establishing a systematic and statistically consistent procedure that makes it possible to assess the ratings proposed by external risk assessment companies. Key words: credit rating, credit risk, credit risk assessment.