Study 1
Schochet, Burghardt, and Glazerman (2001) used an experimental design with random assignment to assess the effects of the Job Corps program on criminal justice involvement, employment, earnings, and time employed. This was the first nationally representative experimental evaluation of a federal employment program. Between Nov. 17, 1994, and Dec. 16, 1995, nearly 81,000 eligible applicants nationwide were randomly assigned to either enroll in a Job Corps program group or to a control group which did not receive Job Corps services (although they could participate in other training or education programs). This large sample was followed for 4 years after random assignment to the treatment group (Job Corps program services) and the control group (not receiving Job Corps program services).
Outcome measures for the study were obtained from two sources: survey data and administrative earnings records. The surveys were conducted at baseline (shortly after random assignment) and at 12-, 30-, and 48-month follow-up periods. Surveys were conducted primarily by telephone and, if necessary, in person. Final analysis and results are based on participants who completed a 48-month follow-up interview. This resulted in 6,828 youths in the Job Corps program group (an 81 percent response rate) and 4,485 youths in the control group (a 78 percent response rate). Educational attainment, employment and job characteristics, receipt of public assistance, and criminal offending were all measured in the participant surveys. Educational attainment was measured by items such as number of programs that youths participated in; attainment of degrees, diplomas, or certificates; and number of hours in training. Employment was measured with similar items such as number of hours worked in a week, number of weeks employed, and job characteristics. Receipt of public assistance was measured with items asking about sources of income (unemployment insurance), receiving assistance in the form of Temporary Assistance to Needy Families or food stamps, and living in public housing. Criminal offending was measured by items asking about number of arrests, number of convictions, most serious charge for which one has been convicted, substance use, time spent in drug or alcohol treatment, and time spent in jail or on probation/parole.
The administrative earnings records (tax data) were collected for 9 years after random assignment. By year 9, participants in the original sample were between the ages of 25 and 33. This data came in two forms: 1) annual summary earnings reported by employers to the Internal Revenue Service and 2) wage records reported to state unemployment insurance agencies. These two sources of tax data cover most workers in formal jobs. Agricultural labor, railroad workers, and some domestic service workers are not covered in either data source. This tax data was used as another source of information to corroborate employment and earnings data collected in the participant surveys.
The study population was mostly male (60 percent) and younger than 20 (73.2 percent were 19 or younger). Most of the sample were racial or ethnic minorities (47.4 percent African American, and 17.7 percent Hispanic) and did not have high school diplomas or GEDs (77 percent). Almost one third of the female participants (28.7 percent) reported having children of their own. The majority of the sample reported never having been arrested (76.6 percent); only a small number stated they had been arrested for a serious crime (4.7 percent). Owing to the random assignment and large sample sizes, there were no significant statistical differences between the treatment and control groups at baseline.
The analysis used average impact estimates per eligible applicant. These were obtained by computing differences in mean outcomes between all treatment group members and all control group members. Weights were used in all calculations and analyses to adjust for the sample and survey designs. Notably, these impact estimates represent the effects of Job Corps relative to other employment and training programs in the community, but not relative to the absence of training. These impact estimates represent the incremental effects of the Job Corps program relative to other programs in which control group members participated. Subgroup analysis was not conducted.