Research note: sensitivity of SES-crime relationships to the inclusion or exclusion of extreme neighborhoods

Abstract

Objectives: This research examines whether SES-crime relationships in Detroit, Michigan, are influenced by including or excluding the highest and lowest one per cent of neighborhoods in SES from correlation and regression analyses. Methods: We studied 879 census block groups in Detroit, Michigan from 2011–2015. SES variables are ratios of (a) per cent in poverty; (b) per cent renters; and (c) the ratio of vacant or demolished units to rented or owned units. Dependent variables include residential burglaries, domestic and non-domestic assaults. Relationships were calculated with and without extreme block groups. Results: SES-crime correlations over space are surprisingly small and can be affected by a few extreme neighborhoods. Some SES-crime correlations are not robust. Regression analyses show that omission of extreme neighborhoods can reverse signs or remove statistical significance. Conclusions: Future researchers should examine the impact of extreme values explicitly before embarking on further analysis.