FORUM on Corrections Research

The effect of prison employment and vocational/apprenticeship training on long-term recidivism

The Post-release Employment Project was designed to evaluate the impact of prison work experience and vocational training on offender behaviour after release into the community. Data were collected on more than 7,000 offenders between 1983 and 1987.

Preliminary findings were reported in 1991, after all the offenders in the study had been released into the community for at least one year. However, this article examines results based on a much longer post-release period-as long as 10 years for many of the offenders. The article reviews the design and methodology of the study, briefly discusses its initial findings and then, most important, provides greater detail on the long-term results. Study design and methodology Unlike most studies of prison vocational training or work experience, the Post-release Employment Project was designed as a prospective longitudinal evaluation.

Inmates were selected for the study if they had worked in prison industry or had received vocational instruction or apprenticeship training while incarcerated. Of those selected, 57% had worked exclusively in prison industry, 24% had received vocational or apprenticeship training (or both) and 19% had a combination of work experience and vocational / apprenticeship training.

Since the requirement for work experience or vocational training made it impossible to randomly assign inmates to a study or control group, a quasi-experimental design was used to choose comparison group inmates from the "reservoir" of all other inmates released in the same calendar quarter as study participants.

Having to control for potential bias in the selection of study participants is a common problem in evaluation studies. Selection bias is based on the belief that selection is never random-there is always a process that determines how participants are selected for programs. It suggests that the selected participants may have unique characteristics that increase their probability of successful outcome-even without program intervention.

Selection bias was addressed in this study by using a two-step statistical matching procedure. A comparison group of inmates was generated that was, in theory, equivalent to the study group in every way but participation in institutional work or vocational training.⁽²⁾ Initial results The study's initial findings can be grouped into three areas- institutional adjustment, halfway house experience and post-release results.

The initial results indicated that inmates who participated in work, vocational training or apprenticeship programs were less likely than the comparison group inmates to receive a misconduct report during their last year of incarceration.

When program participants did receive a misconduct report, it was less likely to be for serious misconduct. Program participants were also rated as more responsible by their unit teams.

Many federal inmates are initially released to a halfway house rather than directly into the community. A halfway house provides a structured setting that allows offenders to work in the community but receive closer supervision than they would on ordinary conditional release.

Interestingly, the comparison group inmates were just as likely as program participants to complete their halfway house stay without committing misconduct resulting in a return to prison. However, the program participants were 24% more likely to obtain a full-time or day-labour job during this time.

Offender post-release outcome data were collected by contacting probation officers. In the U.S. federal justice system, probation officers supervise both offenders who receive a simple probation sentence and offenders who receive a prison sentence and are subsequently released before the end of their sentence.

Information was gathered over a one-year period on whether offenders were re-arrested or had their conditional release revoked, their ability to find employment, and the amount of legal wages they earned.

By the end of the year, 10.1% of the comparison group inmates had been re-arrested or had their conditional release revoked, compared with 6.6% of the program participants. This difference is statistically significant.

Further, 72% of the program participants found and maintained employment during this period, compared with just 63% of the comparison group inmates. This difference is also statistically significant.

Finally, although the difference is not statistically significant, the average working program participant was earning more ($821 per month) than the average working comparison group inmate ($769 per month). Long-term findings Although the initial results of this study were encouraging, we were interested in whether the demonstrated differences would hold up over a longer time period. We were unable to re-assess employment and earnings, but were able to re-analyze recidivism among members of the two groups. Most of the offenders examined in this follow-up had been released at least eight years, some as long as 12 years, previously.

Automated Bureau of Prisons records were used to determine whether the program participants or comparison group inmates had been re-incarcerated federally for a new offence or had had their conditional release revoked.

This use of federal data meant that offenders could have been convicted of, and incarcerated for, a state offence without it being recorded by the study. However, there is no reason to believe that there would be differential rates of prosecution and conviction as a result of their study group assignment, thereby introducing unknown bias into the follow-up data.

This follow-up study examined the amount of time offenders were in the community before being returned to custody. If correctional industries work and vocational training do have a positive effect, program participants would be expected to remain in the community longer.

Male and female offenders were examined separately, as women are less likely to reoffend than men. Confirming these expectations, just 19.3% of the female offenders were returned to custody, compared with 31.6% of the male offenders. However, on average, men were returned to custody after 811 days, while women were returned to custody after 647 days. Therefore, although fewer women were likely to fail on conditional release, those who did failed much earlier than men. This may have interesting implications for the design of male and female offender programs.

The program participants were divided into three subgroups- those who had participated in prison industry (57%), those who had participated in vocational training or apprenticeship programs (24%) and those who had participated in both prison industry and training programs (19%).

Two recidivism measures were used in this study-commission of a new offence and conditional release revocation. However, neither participation in prison industry or training programs or both had an impact on the recidivism of female offenders, regardless of the recidivism measure used.

In contrast, the male prison industry subgroup had 20% longer survival (on conditional release) times than the comparison group when recidivism was defined as a new offence, while the training program subgroup had 28% longer survival times than the comparison group. Both differences were statistically significant.

A similar trend was also present for the "combined" subgroup, although the difference was not statistically significant. This is probably because this was a relatively small subgroup.

However, when recidivism was defined as revocation of conditional release, neither prison industry or training program participation, or both, had an impact on the recidivism of male offenders. What does it mean? It appears that prison employment in an industrial work setting and vocational or apprenticeship training can have both short- and long-term effects that reduce the likelihood of recidivism, particularly for men.

Although this participation does not seem to impact on long-term conditional release revocation, it does seem to reduce the return to custody for new offences. On average, offenders who receive new sentences will spend much longer in prison than those who just have their conditional release revoked. Therefore, correctional industries work and training programs could help to reduce prison populations.



⁽¹⁾U.S. Federal Bureau of Prisons, NALC 202, 3201st Street NW, Washington, D.C. U.S.A. 20534. The opinions expressed in this article are solely those of the authors and do not represent the official position or policies of the U.S. Bureau of Prisons or the U.S. Department of Justice.

⁽²⁾For information on the technicalities of the procedure please see W. G. Cochran and D. B. Rubin, "Controlling Bias in Observational Studies: A Review," Sankhya, 35, 4 (1973): 417-446. See also P. R. Rosenbaum and D. B. Rubin, "The Central Role of the Propensity Score in Observational Studies for Causal Effects," Biometrika, 70, 1 (1983): 41-55.