Is reduced contact with the formal labor market over time associated with heightened recidivism risk?

Abstract

This research addresses whether reductions in formal labor market contact over time lead to heightened recidivism risk among the formerly incarcerated. To address this question, this research draws on a sample of 2,050 Ohio ex-prisoners. First, using group-based trajectory modeling, subjects are categorized into four distinct categories of employment stability, reflecting longitudinal trajectories of formal labor market contact. Then, event-history analysis is used to empirically assess the implications of declining contact with the formal labor market on recidivism risk. Results demonstrate that (1) the recidivism risk of subjects with declining employment stability diverges substantially from the recidivism risk of subjects with high stability over time, (2) the recidivism risk of subjects with declining stability converges with the recidivism risk of subjects with low stability or no employment over time, and (3) heightened recidivism risk among subjects with declining stability occurs contemporaneously to reductions in formal labor market contact. Thus, this research provides novel evidence that dislocation from the formal labor market over time heightens recidivism risk among the formerly incarcerated and suggests that employment-based reentry programming may need to increasingly focus on helping the formerly maintain employment over longer periods of time.

Keywords:

• recidivism
• life course
• employment stability
• reentry

Acknowledgements

The author gratefully acknowledges the support of award #1459214 from the National Science Foundation and the Ohio State University Criminal Justice Research Center.

Notes

1 As a sensitivity analysis, I replicated my findings using a combined recidivism measure comprising both technical and new-crime reincarceration events. The pattern of findings does not change when this is done.

2 As a sensitivity analysis I replicated the multivariate analysis using logistic regressions and parametric survival analysis (selecting a lognormal distribution). Apart from minor differences in coefficients, the pattern of findings was effectively identical across all model specifications.