Classification and Rating of Firms in the Presence of Financial and Non-financial Information
by Thomas Mählmann of the University of Cologne
Abstract: Under the proposed Basel Capital Accord, internal credit ratings are expected to gain in importance because of their potential role in determining the adequacy of regulatory capital. A standard method of deriving an internal (sub) rating consists of estimating a credit scoring function and assigning rating grades based on credit scores. However, the usual purpose of credit scores is to classify firms into two groups: good and bad. Using a model for the joint distribution of financial and non-financial risk factors, we study the effect of common data sample characteristics on the accuracy of 2-group classification and ratings of firms on the basis of their credit scores as derived from two well known scoring techniques: linear discriminant analysis and logistic regression. Our main focus is on bias in scoring function estimation and it is shown that such a bias has a strong negative impact on the accuracy of ratings, but not on that of classification. Therefore this study suggests that a Basel II conforming rating system should be based on a scoring technique that at least leads to consistent (i.e. asymptotically unbiased) estimates of scoring function coefficients for a wide class of risk factor distributions.
Keywords: credit ratings, credit scores, bias.