Investigating Life Course Offender Subgroup Heterogeneity: An Exploratory Latent Class Analysis Approach

Abstract

Within developmental criminology, a common classification of offender behavior is life course trajectories broken into discrete groups of early onseters, late onseters, persisters, or desisters. Yet some investigators state that this is an oversimplification of offending behavior and may not be applicable to females and males. The current study utilizes National Youth Survey data and exploratory latent class analysis to determine whether substantive latent classes exist within offender life course trajectories for females and males and to examine differences between the genders across these groups. The analyses reveal the presence of 4 types of latent subgroups: (a) female-majority de-escalators, (b) male-majority persistent de-escalators, (c) male-dominated persisters, and (d) male-majority chronic fluctuators. Post hoc analyses reveal similarities and differences among the latent groups. Research implications of this study suggest further explorations into whether strict 2-pronged developmental models are indeed appropriately tapping and capturing the full essence of criminal careers for both females and males. The results suggest that continued programming that disrupts delinquent peer associations and reduces consumption of drugs and/or alcohol may be promising in promoting desistance—particularly for females.

Keywords:

• offending trajectories
• latent subgroups
• subgroup heterogeneity
• persisters
• desisters
• de-escalation
• fluctuation
• gender

Notes

De-escalation refers to a reduction in offending. The reduction could be in terms of the number of offenses or a shift from serious offending to less serious offending patterns.

Subgroup heterogeneity can be generally defined as unique differences within subgroups.

The sixth and seventh waves of the NYS were each collected over a 3-year period. However, respondents were still asked to provide information on their year-to-year offending histories, which left no gaps in information across the observation period.

Of the original 1,725 respondents, 1,517 had data across all Waves 4–7.

Attrition for the NYS is a concern highlighted by researchers (see Brame & Piquero, 2003). However, despite the attrition rate, MacMillan (2000) stated that there is an absence of systematic selection bias for those remaining in the sample.

With the exception of those items that overlapped with their index offenses scale or were considered serious offenses. Items included bought stolen goods, carried a hidden weapon, stole something worth less than $5, engaged in prostitution, sold marijuana, sold hard drugs, engaged in disorderly conduct, went joyriding, and stole things worth between $5 and $50.

For a summary of latent class analysis, see McCutcheon (1987), Dayton (1998), and von Eye and Clogg (1994).

No significant age differences were found among the clusters. Thus, aging-out interpretations for any of the cluster differences are not likely.