The empirical clustering of crimes

Despite wide use among researchers and law enforcement, crime classification structures have received little empirical study to validate their theoretical models. Specifically, few studies have examined the empirical structure of crime categories across a large range of offenses to examine adherence to traditional classification systems (e.g., violent vs. non-violent; crimes against persons, property, and society). This study first identified how crimes clustered together empirically in a sample of Florida jail inmates (Aim 1; N = 832, 68% men, 72% White, 83% non-Hispanic) and then tested the replicability of that grouping in a second sample of justice-involved individuals in Illinois (Aim 2; N = 794, 62% men, 51% Black, 96% non-Hispanic). The psychological and behavioral correlates of the factors were then explored (Aim 3). Results from exploratory factor analysis in Sample 1 and exploratory structural equation modeling in Sample 2 indicated a 4-factor structure best fit the data, supporting and expanding the Uniform Crime Report/National Incident-Based Reporting System 3-factor model (i.e., crimes against persons, property, and society). Property and societal crime factors were highly replicable across samples; two factors related to violent crimes were present in both samples but differed in their composition, suggesting the need for more research to understand violent crime heterogeneity. Correlations between the crime factors and external criteria indicated mixed support for the construct coherence of the factors. Most consistently, all crime factors were related to callous unemotional, impulsive, and aggressive traits across samples, suggesting that these characteristics offer little discriminatory ability in understanding criminal behavior. • Crime classification systems need empirical validation for theoretical models. • Factors tested through EFA, CFA, and ESEM, and correlations for construct validity. • Two justice-involved samples used to explore validity and reliability. • A 4-factor structure best fit, supporting and expanding the UCR/NIBRS system. • Property and societal factors were highly replicable, but violent crimes were not.