Study 1
Cohen and Piquero (2008) used a quasi-experimental study design to compare YouthBuild participants with a similar youth cohort, the Second Philadelphia Birth Cohort sample. This cohort included all youths who had prior police contact involving an alleged criminal offense, who had been born in Philadelphia in 1958, and who had lived there until age 18. Thirty YouthBuild grantee sites from 29 cities were chosen to participate in the study, based on geographic dispersion, in both urban and rural areas. Eligible individuals included those referred to YouthBuild by the courts as a diversion program to avoid incarceration, those who had served time in prison or jail and were referred by the criminal justice system for reentry, or those who had been convicted of a crime and served time in prison or jail previously who had self-referred to YouthBuild for education or job training.
A total of 388 YouthBuild participants were included in the study. The average age was 19.6 years. The majority of the participants were male (85 percent) and nonwhite (76 percent). Almost all participants (97 percent) had received a prior arrest, with 70 percent having a prior misdemeanor and 46 percent having a prior felony. Sixty percent had served time in juvenile detention, while 40 percent had served time in an adult correctional facility. Only 11 percent had a high school degree (6 percent) or GED (5 percent) at the time of entry.
The primary outcome of interest was recidivism. Recidivism in the YouthBuild Offender Project included any conviction, incarceration, or parole revocation. For participants in the Philadelphia Cohort comparison group, recidivism referred to any police contact involving a criminal offense (for juveniles) or any arrest (for adults). YouthBuild participants were followed, on average, for 10.3 quarters, or approximately 31 months, while 2-year, follow-up recidivism data was available for the Philadelphia Cohort comparison group (through age 25). There were no subgroup analyses.
There were a number of limitations to the study. First, the birth cohort sample used as a control group is not the most ideal comparison, especially given that no information on the number of participants in this group or their demographic characteristics were provided in the report. There were also no baseline comparisons made between the two study groups, and researchers were not able to use the same measure of the outcome variable (recidivism) to compare the two groups after the program. For the birth cohort control group, recidivism referred to any police contact involving a criminal offense (in the case of juveniles) or arrest (for adults). However, recidivism for the YouthBuild group referred to any conviction, incarceration, or parole revocation. Additionally, no information to support the fidelity of the program implementation was presented.
Study 2
Miller and colleagues (2016) used a randomized controlled design in which eligible participants were randomly assigned through a lottery process to receive the YouthBuild program or to a control group (youths assigned to the control group received information about other services in the community). A total of 75 YouthBuild programs from 29 states, the U.S. Virgin Islands, and Washington, D.C., were selected to participate in the study. The programs took place in areas that ranged from densely populated urban centers to rural areas. Eligible individuals included out-of-school youths, ages 16 to 24, who had dropped out of school before completing high school and were either from low-income or migrant families, were in foster care or aging out of it, were formerly convicted people, had disabilities, or were children of incarcerated parents.
A total of 3,929 participants were included in the study (2,700 assigned to the YouthBuild program, and 1,229 assigned to the control group). The average age for the full sample was almost 20 years, with 33 percent aged 16 to 18 years, 46.3 percent aged 19 to 21 years, and 20.7 percent aged 22 years or older. The majority were male (64.1 percent) and Black (62.9 percent). Of the remaining sample, 15.3 percent were white, 14.6 percent were Latino, 6 percent were other, and 1.1 percent did not specify. Only 9.2 percent of participants had a high school diploma or equivalent at the time of entry. There were no statistically significant differences between the groups on baseline characteristics.
The outcomes of interest included educational attainment, work, and youth development. Participant surveys were administered to a random subset of study participants 12 and 30 months after study entry to obtain information on participation in education and training, educational attainment, and work. They were also asked to report whether they were currently happy and whether their overall health was good. Depression was measured using the Patient Health Questionnaire (PHQ-9), a 9-item scale used to diagnose depression in clinical settings. Participants provided responses to indicate the frequency of occurrence of depression symptoms. Participants who scored a 10 or higher were considered to exhibit signs of major depression. Social support was measured using a 6-item scale, with higher scores indicating stronger social support. Self-esteem was measured using the 10-item Rosenberg Self-Esteem Scale, with higher scores indicating higher levels of self-esteem. Regression models were estimated to determine the effects of the program. Subgroup analyses were performed based on participants’ age, sex, and education level.