Accounting for Feedback Effects in Neighbourhoods and Crime Research: How Much Does It Matter?

Abstract Whereas much ecology of crime research employs cross-sectional study designs, these studies assume that crime does not cause various neighbourhood measures to change over time. However, longitudinal research evidence suggests this is not a plausible assumption. The present study explores this question using neighbourhood data in US cities in Southern California covering over 6.7 million residents. Using data from two time points—2000 and 2010—instrumental variable models account for this endogeneity. The results demonstrate sharp differences between models that do or do not account for endogeneity: several measures yield completely reversed results from positive to negative or vice versa, some measures exhibit much stronger results, whereas others have much weaker, or nonsignificant, results. The findings highlight the importance of accounting for endogeneity in the US context, and that the differences are potentially not modest, but large enough that failing to account for endogeneity may limit our understanding of these social processes that criminologists have theorized. Whether such results generalize to non-US contexts will need to be a focus of future research.