Estimating the neighborhood influence on decision makers : theory and an application on the analysis of innovation decisions

When making decisions, agents tend to make use of decisions others have made in similar situations. Ignoring this behavior in empirical models can be interpreted as a problem of omitted variables and may seriously bias parameter estimates and harm inference. We suggest a possibility of integrating such outside influences into models of discrete choice decisions by defining an abstract space in which agents with similar characteristics are neighbors who possibly influence each other. In order to correct for correlations between the characteristics, the design of this space allows for nonorthogonality of its dimensions. Several Monte Carlo simulations show the small sample properties of spatial models with binary choice. When applying the estimator to innovation decisions data of German firms, we find evidence for the existence of neighborhood effects.