This research tests the relative contribution of social distance and spatial distance to the presence of ties between neighborhoods based on youth co-offending. Using official court data from a large U.S. metropolitan area, a set of dyad independence and exponential random graph models are estimated in order to investigate the characteristics of neighborhoods that foster co-offending. Results reveal significant effects of both social and spatial distance. Social distance contributes to network structure net of spatial proximity, though spatial factors better explain the overall network structure. These results have methodological implications for the analysis of spatial effects and criminal behavior.