The co-offender as counterfactual: a quasi-experimental within-partnership approach to the examination of the relationship between race and arrest

Abstract

Objectives: Estimate the relationship between race and arrest within co-offending partnerships using a quasi-experimental framework. More specifically, this study argues that when two offenders commit an offense together (i.e., co-offend), the characteristics of the offense and victim are the same and can be removed as possible confounding variables. In this way, co-offenders can serve as counterfactual observations to one another, allowing for quasi-experimental analysis of the effects of race on arrest likelihood. Methods: The current study restructures data from the National Incident-Based Reporting System (NIBRS) into a multi-level format wherein level-1 information on offender demographics and arrest are nested within a level-2 file containing information on co-offending partnerships, offense characteristics, and victim characteristics. By restricting the data to co-offending partnerships and examining within-partnership differences in arrest, the analysis examines racial differences in arrest given that two offenders commit the same offense together against the exact same victim. Results: While a traditional logistic regression approach suggests that black offenders are less likely than white offenders to be arrested (OR = 0.749), the quasi-experimental analysis examining within-partnership differences suggests the opposite: black offenders are more likely than their white co-offending partners to be arrested for an offense (OR = 1.031). Conclusions: These results have two implications. First, traditional regression analyses of the relationship between race and arrest may be subject to significant selection and omitted variable bias. Second, there is potential racial disparity in co-offender arrest: black co-offenders are more likely than their white partners to be arrested for the same violent offense.