¹University of Florence, Florence, Italy

Correspondence: A Neri, Department of Statistics, University of Florence, Viale Morgagni, 59 50134, Italy. E-mail: aneri@ds.unify.it

Received March 2000; Accepted January 2001.

Abstract

This paper explores the potential of graphical chain models in defining marketing strategies in the consumer credit field. When ordinal variables are involved, the potential can be increased by exploiting the ordinal scale of variables. In the presented research, the effectiveness of ordinal methods is analysed in three important activities for building a score-card: (1) in coarse classifying of variables, (2) in model selection and (3) in measuring the prediction accuracy of the model.