Model Evaluation and Choice

The paper begins with a reexamination of the goals of model evaluation and choice. Desirable properties for an empirical model are stated as situational, a priori, observational and specification consistency principles. If these consistency principles are accepted as a basis for model evaluation, then one obvious approach to model choice is to consider the relative goodness of the candidate models in terms of each of the four consistency principles. Specific examples from our own empirical research are used to illustrate the advantages of the suggested multidimensional approach to model evaluation and choice.