Spatial and temporal correlations of crime in Detroit: Evidence from spatial dynamic panel data models

Abstract

Since the infamous riot of 1967, high crime rates and negative media reports have labeled the city of Detroit as one of the most dangerous cities in the United States. Using a Spatial Dynamic Panel Data model with both individual and time fixed effects to capture the unobserved heterogeneity as well as the time varying common factors, we investigate the spatial and temporal interactions of criminal activities among the block groups in Detroit. The results indicate that the crime incidents in a block is correlated with the average crime incidents in neighboring block groups contemporaneously, with an estimated coefficient of 0.4758, and the block crime incidents is also correlated with the average crime incidents in neighboring blocks from the previous year, with an estimated coefficient of 0.1572. And crime incidents in a block are positively correlated with its own crime incidents in the previous year. The findings are robust against different model specifications based on alternative spatial weights matrices. The results for both violent crimes and property crimes also suggest strong spatial and temporal correlations among neighboring blocks, providing suggestive and preliminary evidence for policy implementation.