Making a place for space: Using spatial econometrics to model neighborhood effects

Researchers interested in neighborhood effects face many methodological challenges, including, among others, the unobserved selection into neighborhoods that might be associated with the outcome under study, the identification of endogenous effects, or the dependence of results on a neighborhood’s spatial scale. However, another issue has received little attention. Though theoretical explanations commonly stress the interdependence of individual actions in neighborhoods and the spatial diffusion processes between neighbors, the quantitative methods used to evaluate these approaches usually do not directly model such spillover and multiplier effects from one observation to another. Instead, they control for the resulting spatial correlation; for example, by means of multilevel models. Using spatial weights matrices that reflect the extent to which individual neighbors mutually influence each other, this article demonstrates how spatial econometrics are a means for the simultaneous modeling of different mechanisms of neighborhood effects. It further demonstrates how spatial Durbin models reproduce results from multilevel and instrumental variable models without the need to rely on aggregated characteristics of predefined neighborhoods