Multinomial Choice with Social Interactions

Abstract

This paper develops a model of individual decisionmaking in the presence of social interactions when the number of available choices is finite. We show how a multinomial logit model framework may be used to model such decisions in a way that permits a tight integration of theory and econometrics. Conditions are given under which aggregate choice behavior in a population exhibits multiple self-consistent equilibria. An econometric version of the model is shown to be identified under relatively weka conditions. That analysis is extended to allow for general error distributions and some preliminary ways to account for the endogeneity of group memberships are developed.