The spatial configuration of land use can have a significant impact on both market and non-market outcomes. One type of land use configuration that has received considerable attention in recent decades - because of its impact on the efficiency of public service provisioning, on environmental quality and on the productivity of agricultural land - is fragmented development. Many theories have been presented to explain these fragmented development patterns including variable densities, competing land uses and heterogeneity in agents' expectations. More recently, a number of authors have hypothesized that it is the reduction in congestion and the increase in open space amenities in exurban areas that have attracted people thus creating an increase in low-density, scattered development. While many of these hypotheses have been addressed empirically, most of the work has looked at these factors separately. In addition, most of the previous work has been conducted using global parametric models, which make it difficult to capture the spatial heterogeneity of the impact of the different factors and the resulting response. To address these shortcomings it is necessary to have micro-level data that can capture the spatio-temporal nature of the land conversion process and can separately identify the most important factors, on both sides of the market, as well as have a modeling technique that can capture the potential for spatial heterogeneity and non-stationarity in the effects of the variables across space. In this paper, we fill this gap in the existing land use literature by combining several unique panel datasets of historical land use change for Carroll County, Maryland with an original nonparametric modeling technique to separately identify the heterogeneous effect of amenities, local congestion and real options prices effects on the exurban land conversion decision. The results from our model show that factors on both sides of the market matter and that these factors are not spatially stationary. We find that much of the fragmented growth pattern observed in our study region can be explained by the differential effect of local land use interactions across space. This result is particularly important for policy makers interested in protecting farmland, reducing public costs and protecting valuable ecosystem services and biodiversity.