How good is "good"? - Managing Subjective Information in Credit Ratings

How good is “good”? Based on the assumption that asymmetric information exists between decentralized relationship managers who create soft information in credit rating systems and a bank’s management, this study deals theoretically and empirically with the existence of rater effects in bank internal credit ratings. If relationship managers have an incentive to manipulate credit ratings, the ratings cannot be easily compared, rating fairness is reduced. Psychometric methods such as Latent Trait and Latent Class Models allow the user to identify patterns in the data, representing different rating strategies. By making use of this information, i.e. by considering the entire data matrix instead of total sums, the classification power of soft data can be considerably increased. To achieve this improvement, no additional information is needed. Since the credit rating literature has remained largely silent on this topic so far and a real-life data set is used, we consider our findings truly unique. key words: asymmetric information, bank internal credit ratings, soft facts, qualitative information, rater effects, fairness of credit ratings, psychometric methods * Bina Lehmann, Center of Finance and Econometrics, University of Konstanz, Wiwi, D147, 78457 Konstanz (Germany), Tel.: +49 – 7531 – 88 3620, e-mail: [email protected]