We develop a model of land use conversion that incorporates local spillover effects among spatially distributed agents. The model is used to test the hypothesis that fragmented patterns of development in rural-urban fringe areas could be due to negative externalities that create a ‘repelling’ effect among residential land parcels. Identification of the hypothesized interaction effect is complicated by unobserved, spatially correlated heterogeneity. Using an identification strategy that bounds the interaction effect from above, we find empirical evidence that is consistent with a theory of negative interactions among recently developed residential subdivisions in exurban Maryland. The result offers an alternative explanation for low density sprawl to that which is frequently posited in the economics literature and one with potentially quite different efficiency implications.