Spatial Determinants of Urban Crime

DC’s open data portal provides information on places of worship in 2019, most notably, the longitude–latitude coordinates of each POW. We identified 742 POW in the dataset after eliminating 32 cases with coordinates outside of the study area or with duplicate coordinates. Although some prior studies have theorized that the effects of POW differ by the religion or denomination of POW [8 (link), 9 (link), 44 ], DC only classifies its POW by seven religions, of which 97% of them are determined to be Christian. Denomination information was not provided for places of worship. We also determined that the name of POW was not sufficient for accurately classifying POW into denominations. Thus, we have created an index of the number of all places of worship, aggregated to block groups.
One of the most enduring correlates of spatial crime patterns is the presence of facilities that provide an opportunity structure for offenders, targets, and weak guardianship to converge in space and time [18 , 19 (link)]. Accordingly, we constructed several variables to capture such facilities. These facilities include the number of onsite alcohol outlets (i.e., bars, night clubs, and taverns), offsite alcohol outlets (i.e., liquor stores and convenience stores), check-cashing stores, and retail districts/centers (e.g., shopping malls and plazas). We also include a dichotomous variable for the presence of a DC metro station (1 = Yes and 0 = No).
It is also necessary to control for sociodemographic characteristics that have been linked to the spatial distribution of crime [25 (link), 45 (link)–47 ]. Drawing on data from the U.S. Census Bureau we create measures of various sociodemographic characteristics. In particular, we utilize the American Community Survey (ACS) five-year estimates from 2015 to 2019, aggregated to block groups. To capture differences in economic hardship, we account for poverty (%) in block groups. We computed a Herfindahl index of five ethnic groups (white, Black, Latino, Asian, and other races) to account for the ethnic heterogeneity of block groups. The concentration of both Black (%) and Latino (%) residents are also included to account for populations that have been historically marginalized by the political economy of place [48 ]. Furthermore, we employ a variable of homeowners (%) as a proxy for residential stability and we control for two types of housing characteristics: the number of housing units (/100) and occupied units (%). Finally, we created a variable of the population (/100) along with a variable that specifically captures the age group with the highest rate of offending and victimization, that is, persons aged 15 to 29 (%). Descriptive statistics for all measures are shown in Table 1.