Abstract
This paper uses discriminant and logit analyses to develop prediction models to identify bank acquisition targets. We consider several methodological issues, such as whether the choice of the estimation technique, the selection of variables, the use of raw versus industry relative data, the train-and-test sampling scheme, and the criteria for model evaluation affect the predictive accuracy of the developed models. Both estimation methods generate remarkably similar model performance rankings, while differences are revealed in the relative importance of variables when using raw versus industry relative data. We find that in most cases there is a fair amount of misclassification, consistent with previous studies in non-financial sectors, which is hard to avoid given the nature of the problem. © 2009 Elsevier B.V. All rights reserved.