The Society for Social Work and Research

2014 Annual Conference

January 15-19, 2014 I Grand Hyatt San Antonio I San Antonio, TX

Will the Real Nonparticipants Please Stand Up? Exploring the Role of Treatment Contamination Among Nonparticipants Receiving Complementary Prison Programming

Friday, January 17, 2014
HBG Convention Center, Bridge Hall Street Level (San Antonio, TX)
* noted as presenting author
Nora Wikoff, MSW, Doctoral Student, Washington University in Saint Louis, St. Louis, MO
Purpose:This paper examines whether education and employment programs reduce recidivism among young, high-risk male prisoners. Most of these men embarked upon criminal activity during early adolescence, when they were still too young to work. By the time they entered prison, four in ten had less than a 12th grade education and roughly one-third were unemployed. Incarceration further impedes men’s post-release employment options and prospects for financial self-sufficiency.

Prison education and employment programs may help them finish their education, develop job skills, and find good jobs upon release. Unfortunately, most programs show limited improvements in participants' post-release employment and recidivism rates. Unobserved differences between individuals, particularly differences that correlate with both treatment assignment and treatment outcome, can lead to biased treatment effects. Employment-focused programs may not address the educational and skills deficits that had rendered men eligible for employment-focused programs, so pre-existing differences continue to diminish participants’ interest and aptitude for work, relative to nonparticipants. Social norms surrounding work engagement among emerging adults may also contribute to weak program effects, as studies have found that younger prisoners are less responsive to jobs programs than older prisoners.

Prior evaluations have examined the effect of employment programming separate from educational programming, even though both types of programs should increase participants’ employability. This likely underestimates the benefits of each program type on participants’ outcomes, because the comparison group of employment program nonparticipants includes educational program participants as well as true nonparticipants. 

Method:Using longitudinal data on formerly imprisoned adult men from the Severe and Violent Offender Reentry Initiative (SVORI) experiment, this study investigates whether poor treatment outcomes reflect unobserved heterogeneity, selection bias, and treatment noncompliance. This study uses baseline interview data on prison program participation and administrative data on lifetime arrest records to estimate the effects of participation on post-release recidivism. This paper presents findings from a hazard model that analyzed the joint effects of employment programming, prison work, and educational programming on time to first arrest. The analytic model includes three employment-focused programs, to minimize bias resulting from spillover effects, and to estimate the cumulative benefits of programming and work activity. Propensity score analysis is used to differentiate self-selection bias from signaling and human capital effects.

Results:Across participant and nonparticipant groups, men reported needing education and/or employment services at significantly higher rates than other sorts of assistance. Despite high levels of need, less than half of men engaged in education programming before release from prison and only 28% had participated in prison employment programs. Controlling for treatment noncompliance and unobserved heterogeneity among participant and nonparticipant groups improved estimates of treatment effects and provided support for employment-focused programming to reduce recidivism.

Conclusions and Implications: The findings suggest that programs should offer comprehensive education and employment programs to narrower subsets of prisoners who exhibit the strongest interest in working. This paper discusses implications for employment-focused programming design, and it identifies analytic techniques that program evaluations can use to reduce unobserved heterogeneity that contributes to biased parameter estimates.