A Multinomial-Choice Model of Neighborhood Effects

Abstract

This paper provides a model of individual decisionmaking in the presence of neighborhood effects. By neighborhood effects, we refer to interdependencies between individual decisions and the decisions and characteristics of others within a common neighborhood. As such, neighborhood effects are forms of social interactions; in fact the terms would seem to be interchangeable in many contexts. Within economics, there has developed an increasing recognition that neighborhood effects play a possibly major role in explaining a range of individual behaviors; rich empirical and theoretical literatures have developed. A failing of much of this work is the absence of strong connections between theory and empirics. This paper addresses this limitation by providing a model of multinomial choice with neighborhood effects. The model represents a generalization of the binary choice model of William Brock and Steven Durlauf (2001a,b). The model is econometrically implementable and so has the potential of allowing for structural estimation of neighborhood effects. Section I outlines a basic choice model with neighborhood effects. Section II specializes the structure using a multinomial logit framework. Section III considers econometric implementation. Section IV discusses some limitations of the framework and proposes some research directions. Proofs of all theorems may be found in Brock and Durlauf (2001c). I. Models of neighborhood effects We consider I individuals who are members of a common neighborhood. Our objective is to characterize the individual choices of each i, i ω (made from some set Ω i) and the vector of 1 neighborhood choices ω. From the perspective of theoretical modeling, it is useful to distinguish between three types of behavioral determinants: a vector h i of deterministic (to the modeler) characteristics, a vector ε i of random characteristics, and the subjective beliefs an individual has about behaviors in the neighborhood, represented as a conditional probability e measure () µ i ω. Individual choices i ω are described as maximizing some payoff function V,