Study 1
To investigate the effect of the Allegheny County Jail-Based Reentry Case Management Program (Reentry 2), Willison, Bieler, and Kim (2014) used a quasi-experimental design, assessing the program’s impact on recidivism. Data (such as demographics, criminal offending risk scores, and criminal history) was drawn from Allegheny County administrative database. To conduct their analysis, the researchers drew a sample of more than 100,000 individuals who had been sentenced to the Allegheny County Jail between 2008 and 2012. From this sample, 250 were identified as program participants.
Propensity score matching was conducted to create a balance between the treatment group (participants in the reentry program) and the comparison group. The matching variables included race, gender, citizenship status, marital status, origin of driver’s license, age, the number of prior arrests, and the incarcerated person’s Proxy Triage Risk Screener score. If an individual’s matching score was too high or too low to correspond to another’s incarcerated person's score, the incarcerated person was removed from the sample. Following propensity score matching, there were 220 program participants and 220 matched comparison incarcerated persons.
The treatment group was 85 percent male, had a mean age of 30, and was 58 percent Black and 42 percent white. In addition, 80 percent of the participants were single. Approximately 5 percent of the treatment group was considered low risk, 46 percent considered medium risk, and 49 percent considered high risk. The comparison group was similar on baseline characteristics. The comparison group was 83 percent male; had an average age of 29; and was 57 percent Black, 41 percent white, and 2 percent other. Additionally, 79 percent of the comparison group was single. Of this group, 6 percent was considered low risk, 47 percent considered medium risk, and 47 percent considered high risk.
Logistic regression and Kaplan-Meier survival curves were used to analyze the impact of the program on probation violations and the probability of rearrest during the first 1,000 days after release from jail. No subgroup analyses were conducted.