Interpolating Socioeconomic Data for Crime Regression Models at the Neighborhood Level

Abstract

Prior studies of spatial crime patterns often turn to the U.S. Census administrative units as proxies for neighborhoods. We know less about the feasibility of using areal interpolation methods to help estimate neighborhood-level regression models. We used ArcGIS Pro spatial interpolation methods to create neighborhood estimates of socioeconomic characteristics based on the Census data. Then, we examined the relationship between neighborhood characteristics, including racial heterogeneity, residential mobility, immigration, and concentrated disadvantage, and crime at several levels of analysis. We demonstrated that the spatial interpolation methods can offer reliable and accurate estimates of neighborhood measures when socioeconomic data are only available at the Census block group and tract levels. Moreover, we found a significant association between concentrated disadvantage and crime at all levels of analysis. This study has important implications for social and systems science research with focus on area characteristics to predict spatially distributed outcomes, including crime.