Unraveling Victim-Offender Overlap: Exploring Profiles and Constellations of Risk

Abstract

Victim-offenders are generally considered a distinct group with one or more shared characteristics; however, some have suggested possible victim-offender subgroups with varied victimization-offending patterns. The potential for victim-offender subgroups was assessed within a nationally representative sample of 1,000 youth using latent class analysis. Regression of relevant covariates on analytically derived subgroups allowed for further identification of the nature of these groups. Three victim-offender types emerged: the general victim-offender class linked extensive victimization and offending by acutely angry and anxious youth from extremely strained families; the bullied-combative class linked being bullied with peer assault by younger, emotionally vulnerable youth; and the abused-substance use class linked child maltreatment and sexual violence with substantial substance use by poorly supervised, angry, and anxious youth.

Keywords:

• victimization
• offending
• taxonomy

Acknowledgments

The data used in this journal article were made available by the National Data Archive on Child Abuse and Neglect, Cornell University, Ithaca, New York, and have been used by permission. Data from the Developmental Victimization Survey 2002–3 were originally collected by Heather A. Turner and David Finkelhor at the University of New Hampshire. Neither the collector of the original data, the funder, the National Data Archive, Cornell University, or its agents or employees bear any responsibility for the analyses or interpretations presented here.

Notes

1. The Ousey et al. (2011) article addressed limitations and the likelihood of diverse findings related to victim-offender overlap due to the utilization of samples from various subsets of the population. They used a representative sample from a state with a predominately rural population and noted that dissimilar findings may develop and be supported by studies investigating victim-offender overlap utilizing a different type of sample—such as a sample of older, high-risk youth from an urban environment (i.e., Pittsburgh Youth Study).

2. A latent class model was also specified for the aggregated sample with no subgroups to assess whether such a model might provide a better fit to the data. This captures the generalist argument that there are no qualitatively distinct subgroups and only one class of individuals is identifiable in the data. Given some recent controversy surrounding the specification of latent class models as pertains to inherent assumptions about the presence of groups and their subsequent identification (e.g., Sampson & Laub, 2005) we specified a one class LCA model in an attempt to assess whether a single distribution might better capture the relationship studied here. The results indicated that the BIC for an aggregated or one-class model specification (18472.55) was not superior to the multiclass models reported in Table 2. This provides some empirical support for the attempt to fit this type of model to the data.

3. The descriptions of the three classes of victim-offenders provide fuller elaboration of typical members of each subgroup; however, this is a probabilistic process where not all youth fit within the given descriptions. For example, some members of the abused-substance use class were male and had not experienced sexual victimization, and certain members of the bullied-combative class were girls whose parents were not necessarily critical, abusive, or stressed.