December 10, 2015 | Backgrounder on Crime
To date, we do not know enough about what works in helping former inmates safely and successfully reintegrate into society. In fact, the scientifically rigorous evaluations of prisoner reentry programs that use random assignment—the gold standard of evaluation designs—have found, at best, mixed evidence of success. This conclusion is particularly relevant for employment-based reentry programs. A large-scale multisite experimental evaluation of the federal government’s Reintegration of Ex-Offenders (RExO) program, an employment-based intervention, found the program to be ineffective. This conclusion has significant implications for congressional consideration of versions of the Second Chance Reauthorization Act (S. 1513 and H.R. 3406).
In 2014, the Justice Department’s Bureau of Justice Statistics (BJS) issued a study on the recidivism rates of former prisoners released from 30 states in 2005. The BJS found that 67.8 percent of released prisoners were arrested for a new crime within three years. The recidivism rate for the five-year period was 76.6 percent.
This bad news should come as no surprise. A 2002 BJS report found similar trends. Since the 1990s, the Justice Department has poured significant funding resources into prisoner reentry programs. To date, however, we do not know enough about what works in helping former inmates safely and successfully reintegrate into society.
The quality of the research used to answer this question can greatly influence the answer. For example, quasi-experimental designs that employ scientific methodology, but do not use random assignment frequently fail to account for individual differences that affect program outcomes. This failure leaves open the possibility that the underlying differences between the groups of former inmates receiving and not receiving reentry services, not the program, caused the observed impact. While quasi-experimental designs may use sophisticated techniques, experimental evaluations are still considered better at producing reliable estimates of program effects. Compared with experimental designs, evidence in criminal justice policy indicates that quasi-experimental evaluations are more likely to find favorable intervention effects and less likely to find harmful intervention effects. Simply put, randomized experiments provide more reliable impact assessments than quasi-experiments do.
For example, propensity score analysis, a popular quasi-experimental technique that tries statistically to account for possible differences in the characteristics of program participants, has been applied to assess the effectiveness of prisoner reentry programs. Propensity score quasi-experiments have found three prisoner reentry programs—InnerChange Freedom Initiative, EMPLOY, and the Boston Reentry Initiative—to be effective. The problem with this technique when used for the InnerChange and EMPLOY programs is that it compares offenders volunteering for services with less motivated offenders not volunteering for services. For the Boston Reentry Initiative, program administrators carefully selected inmates for participation, and the evaluation compared these preferred inmates with inmates not selected for participation. The effect of this selection bias is nearly impossible to remove from assessments of effectiveness performed by propensity score analysis.
In sum, propensity score analysis is not a valid substitute for experimental evaluations. Propensity score analysis’s main failing is that the method cannot remove the effects of selection bias caused by unobserved variables, such as an offender’s motivation to change his or her life for the better. No matter the level of statistical sophistication, the method cannot account for unobserved factors. Thus, policymakers should exercise caution when interpreting the results of evaluations using this method.
In addition, policymakers need to ensure that programs are evaluated with regard to their impact on the primary purpose for which they have been established. When assessing the impact of prisoner reentry programs, the most important outcome measure is recidivism. Some have questioned the emphasis on recidivism as a measure of effectiveness compared with other measures that assess adjustment or reintegration of former prisoners into society. While intermediate measures, such as finding employment and housing, may be important, these outcomes are not the ultimate goal of reentry programs. If former prisoners continue to commit crimes after going through reentry programs, then the programs cannot be judged effective. For this reason, policymakers should place primary importance on measures of recidivism when judging the effectiveness of prisoner reentry programs.
In fact, the scientifically rigorous evaluations of prisoner reentry programs that use random assignment—the gold standard of evaluation designs—tend to find mixed results. A large-scale multisite experimental evaluation of the federal government’s Reintegration of Ex-Offenders (RExO) program found the program to be ineffective. This conclusion has significant implications for advocates of federal funding of prisoner reentry programs and for congressional attempts to pass versions of the Second Chance Reauthorization Act (S. 1513 and H.R. 3406).
There is considerable debate over the effectiveness of corrections and reentry programs. Some have concluded that several types of programs are effective, while others have cast doubt on the ability of these programs to reduce recidivism. Prisoner reentry programs operated by secular or faith-based organizations offer a wide range of services that potentially could address recidivism. However, there are not enough scientifically rigorous evaluations of secular or faith-based prisoner reentry programs to conclude that these programs are effective. This review of the prisoner reentry literature is thus limited to evaluations that use random assignment.
A search of the evaluation literature identified nine experimental evaluations and one evaluation that used experimental and quasi-experimental methods. Their results are summarized below.
Reintegration of Ex-Offenders (RExO) Program. Begun as a combined initiative of the U.S. Departments of Labor and Justice and other federal agencies in 2005, the RExO program provides grants to local organizations to administer employment-focused prisoner reentry programs.
The Department of Labor just released the two-year results of the RExO Random Assignment Evaluation. The large-scale multisite experimental evaluation assessed the impact of federal grants on reducing recidivism of former prisoners. Specifically, the evaluation assessed the effectiveness of federal grants to 24 local employment-based reentry programs, many of which were operated by faith-based organizations.
Former prisoners who were at least 18 years of age, were previously incarcerated for a minimum of 120 days, were not convicted of sex-related offenses, and did not have a violent crime as their most recent offense were eligible for participation. Of the 4,655 eligible individuals, 60 percent were randomly assigned to the program group, while the remaining 40 percent were randomly allocated to the control group.
Policymakers should take special note of the scientific rigor of this multisite experimental evaluation. First, the experimental design of the RExO evaluation is important because random assignment is the only way to determine with a high degree of certainty the effectiveness of the grant-funded reentry programs. Second, basing the evaluation on multiple sites allows policymakers to assess the national effectiveness of the reentry grants.
The RExO grantees mainly provided three types of services: mentoring, employment services, and case management and supportive services. Program group members were more likely to obtain employment-focused services, including vocational and job-readiness training, than their counterparts in the control group. Additionally, they were more likely to receive mentoring services. The authors of the evaluation conclude that “the RExO program had a substantial impact on the number and types of services study participants received during the two-year follow-up period.”
How Did RExO Affect Employment and Earnings? The services provided by the RExO grantees had a small effect on the employment and earnings of participants. One year after random assignment, 71.3 percent of the program group worked at all, compared with 67.9 percent of the control group—a small but statistically significant beneficial impact. However, during the following year, the rates of participation in work failed to be statistically different between the program and control groups, indicating that the initial small benefit disappeared. For total days employed over the two-year period, the program group averaged 286.7 days, while the control group averaged 274.3 days—a statistically indistinguishable difference of 12.4 days. On average the program group earned a statistically significant $883 more in income than the control group over the two-year period.
Did RExO Reduce Recidivism? Based on administrative data, the services provided by the RExO grantees failed to improve the recidivism, convictions, and reincarceration rates of participants. Over the two-year follow-up period, 42.0 percent of the program group and 43.2 percent of control groups were arrested—a statistically indistinguishable difference of 1.2 percent. Further, the services provided by the RExO grantees failed to affect convictions for new crimes, including violent, property, and drug crimes. Members of the program group were no more or less likely to be admitted to prison for new crimes or parole/probation violations. Members of both groups were incarcerated for an average of 76 days during the two-year period. The evaluation authors conclude that the criminal justice results based on administrative data provide “no evidence whatsoever of any impacts of RExO.”
CEO Prisoner Reentry Program. The Center for Employment Opportunities (CEO) Prisoner Reentry Program is an employment-based program that immediately places recently released prisoners in transitional jobs, usually with New York City or New York government agencies. The transitional jobs approach uses employer subsidies to encourage the hiring of former prisoners. The work performed consisted mainly of maintenance, repair, and janitorial activities. While working their transitional jobs, participants receive assistance in finding permanent, unsubsidized employment.
A large-scale experimental evaluation with a sample size of 977 former prisoners found that CEO Prisoner Reentry Program participants had a mixed impact on employment and recidivism. According to the evaluation, “CEO substantially increased employment early in the follow-up period, but the impact faded over time as program group members left the transitional jobs. CEO had no positive impact on unsubsidized employment for the full sample.” The initial impact on employment was driven by the transitional jobs subsidies. During the course of the two-year evaluation, 59.6 percent of intervention participants found unsubsidized employment, compared with 62.8 percent of the control group—a statistically indistinguishable difference of 2.7 percent.
Over the three-year follow-up period, the rate of ever being employed in a subsidized or unsubsidized job was 83.8 percent for the intervention group and 70.4 percent for the control groups—a highly statistically significant difference of 13.4 percent. However, 63.7 percent of the intervention group reported ever working in an unsubsidized job, compared with 69.0 percent for the control group—a difference of –5.3 percent that was not statistically significant at the traditionally accepted level. Thus, the program appears to be ineffective at moving participants into unsubsidized employment.
Compared with the control group, the intervention group did not have statistically different arrest rates two years and three years after release from prison. After two years, the intervention group had an arrest rate of 37.7 percent, compared with the 41.8 percent arrest rate for the control group—a statistically indistinguishable difference of 4.1 percent. Over three years, the arrest rate for the intervention group was 48.1 percent and for the control group was 52.8 percent—a statistically insignificant difference of 4.7 percent. However, CEO had more success at lowering conviction rates. After two years, the intervention group had a conviction rate of 30.5 percent, compared with the 38.3 percent conviction rate for the control group—a statistically significant difference of 7.7 percent. This difference in convictions is explained by the fact that the intervention group was less likely to be convicted of misdemeanors. For felony convictions, there was no difference between the groups. For the three-year conviction rate, the difference between the intervention and control groups was not statistically different at the traditionally accepted level.
For the two-year and three-year follow-ups, the intervention group was less likely to be incarcerated in jail or prison. For the two-year follow-up, the intervention group had a reincarceration (returning to jail or prison) rate of 49.5 percent, compared with the 55.4 percent reincarceration rate for the control group—a statistically significant difference of 5.9 percent. For the three-year follow-up, the intervention group had a reincarceration rate of 58.1 percent, and the control group had a reincarceration rate of 65.0 percent. The difference of 6.9 percent was statistically significant. This result appears to be driven by the intervention group being statistically less likely to land in jail. The difference in the prison reincarceration rates between the intervention and control groups was not statistically different. At the end of the three-year follow-up, 25.4 percent of the intervention and 30.0 percent of the control groups were incarcerated in jail or prison—a statistically insignificant difference of 4.6 percent.
The goal of the CEO reentry programs was to provide immediate employment and income to former inmates after their release. However, only 41 percent of study participants were enrolled in the evaluation within three months of their release from prison. Limiting the data to this subsample of former prisoners with less than three months between release and random assignment, the analysis found recidivism results similar to the full sample. There were no statistically significant differences at the traditionally accepted level on ever being arrested during the two-year and three-year follow-up periods. However, members of the intervention group were less likely to be convicted of a crime during both follow-up periods. The beneficial effect appears to be driven by lower rates of misdemeanor convictions, rather than felony convictions. Further, intervention group members of this subsample were less likely to be incarcerated during the follow-up periods, mainly due to lower jail admissions, not to differences in prison admissions.
Transitional Jobs Reentry Demonstration. The Transitional Jobs Reentry Demonstration (TJRD), a large-scale experimental evaluation, tested the effectiveness of providing temporary paid jobs, support services, and job placement assistance to prisoners returning to the community. The Joyce Foundation used a competitive process to select and fund employment-focused prisoner reentry programs in four Midwest cities: Chicago, Detroit, Milwaukee, and St. Paul. The experimental evaluation randomly assigned 1,813 former prisoners to the TJRD intervention group and the control group. Members of the control group received basic job search and referral services without subsidized employment.
The TJRD targeted men who were at least 18 years of age and who had been released from state prison within 90 days before participating in the demonstration. The release within 90 days requirement was based on the findings of the CEO Prisoner Reentry Program. On average, members of the intervention and control groups were enrolled in the evaluation 44 days after their release from state prison. While there were some limitations on participation at certain sites, men with all types of criminal histories were generally allowed to participate.
The TJRD jobs programs in the four sites offered temporary, minimum-wage employment that provided 30 hours to 40 hours of paid work each week. These temporary jobs were subsidized by TJRD—meaning that the employers were given subsidies to hire former inmates. Rather than focusing on developing occupation-related skills, TJRD programs aimed to detect and deal with behavior and performance issues that arose at work sites. In addition, participants were provided assistance with finding permanent employment. At the Milwaukee and St. Paul sites, TJRD participants were awarded retention bonus payments totaling up to $1,500 for staying in their subsidized, transitional jobs. However, the effect of the retention bonuses was not tested experimentally, so these results are not reliable. Members of the intervention group who were placed in transitional jobs spent an average of 11 weeks in the subsidized jobs.
Did TJRD Increase Employment and Earnings of Participants, Compared with the Control Group? In the short term, the answer is yes. For the first year after random assignment, measured quarterly, the intervention group had higher rates of overall employment that were statistically significant, compared with the overall employment rates of the control group. In the second year, the quarterly differences in overall employment were not statistically significant at traditionally accepted levels of statistical significance. Thus, the initial impact appears to be driven by offering subsidized employment to the intervention group. During the first three quarters after random assignment, the intervention group had statistically higher subsidized employment rates than the control group, but the differences became statistically indistinguishable in quarters four through eight.
Over the course of two years, 94.3 percent of the intervention group reported being employed at least once, compared with 65.3 percent of the control group—a statistically significant difference of 29.0 percent. However, only 41.0 percent of the intervention group and 40.4 percent of the control group members ever obtained unsubsidized employment two years after random assignment. This minuscule difference of 0.6 percent was statistically insignificant. In fact, the participation rate of the control group in unsubsidized jobs over two years was 65.0 percent, compared with 60.4 percent for the intervention group—a statistically significant difference of 4.6 percent. As for earnings obtained from unsubsidized jobs, the difference between the intervention and control groups was statistically insignificant.
The author concluded: “[T]here is little evidence that the transitional jobs programs, as implemented, led to better unsubsidized employment outcomes over a two-year period. These results are consistent with the findings of other evaluations of transitional jobs programs.”
Despite These Underwhelming Employment and Earnings Findings, Did TJRD Affect Recidivism? Overall, the authors concluded, “For the full sample, the TJRD transitional jobs programs did not significantly affect key measures of recidivism in Year 2 or over the two-year follow-up period as a whole.” Over the course of two years, the arrest rates for the intervention and control group were 55.2 percent and 51.8 percent, respectively—a statistically insignificant difference of 3.4 percent. TJRD also failed to affect convictions, prison admissions for new crimes, parole/probation violations, and total days spent in prison after random assignment.
Interestingly, when the recidivism rates are analyzed by site, intervention group members in Detroit were more likely to be admitted to prison, especially for technical violations, than their counterparts. The admission-to-prison rate for the Detroit intervention group was 17.8 percent, compared with 9.9 percent for the control group—a statistically significant harmful difference of 7.9 percent. When prison admissions were limited to new crimes, the difference between the groups was statistically indistinguishable. However, the prison admission rates for parole/probation violations were 10.3 percent for the intervention group and 4.2 percent for the control group—a statistically significant harmful difference of 6.1 percent. Further, members of the intervention group averaged 24.9 more days in prison after random assignment, compared with their counterparts. The findings for the other sites failed to reach traditional levels of statistical significance.
Minnesota Comprehensive Offender Reentry Plan. The Minnesota Comprehensive Offender Reentry Plan (MCORP), a reentry program for inmates released to Hennepin, Ramsey, Dodge, Fillmore, and Olmstead Counties, began in 2008. For the small-scale experimental evaluation, inmates were randomly assigned to intervention and control groups at least 60 days prior to their scheduled release date. To be eligible to participate, in addition to the county requirement, inmates needed to have at least six months of community supervision remaining on their sentences and not be required to register as a sex offender. From January to September 2008, 269 eligible inmates were randomly assigned to the intervention and control groups.
MCORP attempted to increase the level of services to inmates while they were institutionalized and released to the community. With about half of the caseload of agents supervising members of the control group, “MCORP agents focused on helping offenders access services related to employment, vocational training, education, housing, chemical health, mentoring, faith-based programming, and income support.” The evaluation found that the intervention group was more likely to obtain employment, housing, social support, mentoring, educational services, and income support. They were no more likely to receive faith-based services and vocational training than members of the control group.
Did the Beneficial Impacts Have an Effect on Recidivism? The follow-up period for study participants averaged 16 months after release. A survival analysis that assessed the impact of MCORP on length of time until study participants recidivated found that MCORPS reduced recidivism. Specifically,
Controlling for the effects of the pre-release predictors, MCORP lowered the hazard ratio by 37% for rearrest, 43% for reconviction, and 57% for new offense reincarceration. That is, MCORP offenders reoffended less often and more slowly than the offenders in the control group; as a result, they survived longer in the community without a new offense.
However, MCORP failed to have any effect on technical violation revocation rates. By standards of recidivism research, the 16-month follow-up period is too short to make judgements about the long-term effectiveness of the program.
Minnesota High-Risk Revocation Reduction (HRRR) Program. The goal of the HRRR program was to reduce the recidivism of high-risk offenders in Minnesota. For the small-scale experiment, 239 adult male release violators who spent two to six months in two state prisons were randomly assigned to intervention and control groups. The prisoners were previously released from a state prison, but were reincarcerated for violating the conditions of their supervised release. The control group received normal case management services, while the intervention group was given access to “supplemental case planning, housing, employment, mentoring, cognitive-behavioral programming, and transportation assistance services.” Specifically, the intervention group was offered 75 days of transitional housing, 16 weeks of subsidized employment, weekly life skills counseling, weekly mentoring sessions, and three public transportation passes.
HRRR was funded by a federal demonstration grant authorized by the Second Chance Act. Recidivism rates were assessed one year to two years after release. Based on one-to-two years of post-release data, a survival analysis found statistically significant impacts on revocations and reconvictions, but HRRR had no effect on rearrest and reincarceration. Specifically, “HRRR participation significantly reduced the risk of a new supervised release revocation by 28% and the risk of a reconviction by nearly 42%.” Even though HRRR participants were less likely to be returned to prison for technical violations (revocations) and convicted of new crimes, they were still no more or less likely to be arrested for new crimes or returned to prison after being convicted of a new crime.
Southern California Employment-Focused Reentry Program. A small-scale random assignment evaluation of a nationally recognized employment-focused prisoner reentry program operating in Southern California was found to have no statistically measurable effect on recidivism. The program uses the STRIVE employment model, which is provided by more than 25 community-based organizations throughout the nation. The reentry program, which provided vocational and employment readiness training as well as placement services, failed to increase employment and reduce recidivism. A total of 217 former inmates were randomly assigned to intervention and control groups. Participants were eligible for participation as long as they had no convictions for arson or sexual offenses and could pass a drug test.
Twelve months after random assignment, 29.8 percent of the intervention group and 27.1 percent of the control group were employed full time. The part-time employment rates of the intervention and control groups were 12.5 percent and 9.4 percent, respectively. The differences between the full-time and part-time employment rates were not statistically different, suggesting that the vocational services provided were ineffective.
The self-reported arrest rates at the 12-month follow-up were 45.2 percent for the intervention group and 49.4 percent for the intervention and control group—a statistically insignificant difference. Similarly, the difference between the incarceration rate at the one-year follow-up period—47.1 percent for the intervention group and 48.2 percent for the control group—was statistically insignificant. Based on official records, 66.3 percent and 69.2 percent of the intervention and control groups were arrested during the two-year follow-up period. The difference was statistically insignificant.
The authors infer that there “is little support among experimental studies that offering employment programs to formerly incarcerated offenders reduces their risk of recidivism.”
Milwaukee Safe Street Prisoner Release Initiative (PRI). Operated by the Wisconsin Department of Corrections, PRI provided “reach in” services, including drug abuse counseling, anti-gang counseling, and employment preparation, to inmates six months prior to their release, along with after-release employment-focused services. Through the Comprehensive Anti-Gang Initiative created in 2006, the U.S. Department of Justice provided grant funding to Milwaukee to operate PRI. After release, PRI offered subsidized employment for six months and assistance in finding unsubsidized employment. Improving the employment opportunities of the released inmates was the principal instrument for reducing recidivism.
To assess effectiveness, a small-scale experimental evaluation of 236 high-risk offenders with histories of violence and gang involvement who were scheduled to be released in Milwaukee was performed. During 2009, 106 inmates and 130 inmates were randomly assigned to the intervention and control groups, respectively. While in prison, members of the intervention group were offered vocational skills assessments, vocational training, soft-skills training, restorative justice counseling, Community Corrections Employment Program (CCEP), alcohol and drug treatment, remedial education, and Breaking Barriers, a life skills and behavioral/cognitive change program. After release, the intervention group was eligible for substance-abuse treatment and continued CCEP services, which included subsidized employment and job search assistance.
When assessing the PRI’s impact on quarterly employment outcomes, the evaluators limited the analysis to nonrecidivists (those individuals who were not sent back to prison). In each of the four quarters of the one-year follow-up, members of the treatment group had higher rates of employment, measured by working at least one hour per quarter, than the employment rates of the control group. These findings are statistically significant. According to the authors, the intervention group was about 20 percentage points more likely than the control group to be employed for at least an hour during each quarter. As for earnings per quarter, members of the intervention group earned $334 more in the first quarter and $470 more in the second quarter than the earnings of the control group. These differences were statistically significant; however, this initial beneficial effect disappeared during the third and fourth quarters. The fading of the beneficial effect on earnings is likely due to the employment subsidies lasting only six months after release.
Did PRI Reduce Recidivism? A survival analysis found that the probability of individuals in the intervention group being arrested, compared with the control group, was about 39 percent lower on a month-to-month basis. This decrease in the probability of being arrested was statistically significant. However, the difference in the average number of arrests during the one-year follow-up was not statistically different between the intervention and control groups. In fact, 28 percent of both groups were arrested for serious (violence and weapons-related) crimes.
When new arrests are analyzed during the six-month, nine-month, and 12-month follow-up periods, members of the intervention group had statistically significant lower arrest rates. As for imprisonment one year after release, the difference in reincarceration rates was not statistically significant. Given that the follow-up period for this evaluation is only one year, policymakers should exercise caution when interpreting these results. Short-term impacts, especially with prisoner reentry programs, tend to disappear with longer follow-up periods. Hopefully, the authors of this evaluation will update their research as more data become available.
Multimodal Community-Based Reentry Program. A small-scale experimental evaluation with a sample size of 511 assessed the effectiveness of a multimodal, community-based prisoner reentry program that focused on substance-abuse treatment for offenders with substance dependencies. Based upon the “best practices literature,” the substance-abuse treatment employed a cognitive-behavioral therapeutic model that assumes substance abuse is a learned behavior that can be unlearned and/or replaced. In addition, drug testing was also implemented to hold program participants accountable. A total of 263 former prisoners were assigned to the intervention group, while 248 former prisoners were assigned to the control group. Control group members received regular community supervision.
Prisoners with high assaultive risk classifications, sex-offense histories, arson histories, pending new felony charges, or physical or mental conditions were barred from eligibility. The first phase of the program (30 days to 45 days after release) entailed a residency in a secure transitional facility that focused on finding housing and employment, enrolling in job-readiness and life-skills training, and initiating outpatient substance-abuse treatment. The second phase started when former prisoners were allowed to move into approved home residences. Continued participation in substance-abuse treatment and testing were required for one year post-release.
Participants in the experimental evaluation were followed for two years. The following outcomes were assessed: relapse (testing positive for drug use); rearrest for new felony; and reincarceration for technical violation or conviction for a new crime. For the relapse outcome of testing positive for drug use at least one time, the program had no effect on the percentage of subjects testing positive. However, the program did affect the percentage of drug tests in which study participants were found to be using illegal substances. Members of the intervention and control groups tested positive for drug use in 26 percent and 21 percent of their respective tests—a difference that was statistically significant. A survival analysis found that the mean survival time until relapse was 286 days for the intervention group and 338 days for the control group—a statistically significant difference. Thus, the intervention group had higher rates of testing positive for drug use and survived shorter periods of time until testing positive for drug use, meaning that the program had a harmful impact. Otherwise, the program has no effect on rearrest and reincarceration rates.
Washington State Work Release. During the early 1990s, a small-scale evaluation in Seattle randomly assigned 218 eligible prisoners to serve out their sentences or enter work-release facilities. Participants in the intervention group were required to be involved in gainful employment or job training while participating in the program. These work-release participants were obligated to remain in their work-release facilities unless they were engaged in approved work and other activities.
One year after random assignment, work-release participants had a rearrest rate of 22 percent compared with the recidivism rate of 30 percent of the non–work release participants. However, this difference of 8 percent was statistically insignificant, meaning that the difference cannot be attributed to participation in the work-release program. The program had no impact on any incarcerations (jail or prison) for new crimes, but the intervention group was statistically less likely to be jailed. Further, a cost-effectiveness analysis demonstrated “basically no differences in costs between work releases and inmates completing their full terms in prison.”
Project Greenlight. Project Greenlight, a short-term, prison-based reentry program in New York City, applied cognitive-behavioral skills training to prisoners eight weeks before their release. The program mainly endeavored to increase “post-release outcomes by (1) incorporating an intensive multimodal treatment regimen during incarceration and (2) providing links to families, community-based service providers, and parole officers after release (although there was no actual community follow-up).” The cognitive-behavioral skills training approach used by Project Greenlight is labeled as a “What Works” or “evidence-based” model based on the results of previous research.
A small-scale evaluation found that Project Greenlight “did not reduce recidivism and may actually have increased it.” The evaluation used a mixed design that combined a quasi-experiment design for the first five months of assigning inmates to the program with random assignment design during the last six months. Project Greenlight participants were compared with a group of inmates that did not receive any pre-release transition services and with a group that received alternative transition services.
Compared with the inmate group that received the alternative transition services, Project Greenlight participants had a 41 percent higher chance of arrest after one year. Project Greenlight participants did not have statistically different arrest rates compared with inmates receiving no services. While the impact of Project Greenlight appears to be harmful, policymakers should exercise caution when interpreting the findings due to the mixed methodology that degrades the validity of the findings.
While the evaluations of CEO, MCORP, and PRI found varying degrees of success, these impacts either did not last longer than short-term assessments or were not assessed over longer follow-up periods. Further, inconsistent effects of the employment-focused CEO could not be replicated in the TJRD evaluation.
Many advocates of social programs have adopted the language of the evidence-based policy movement. Under the evidence-based policy movement, programs found to be effective using rigorous scientific methods are deemed “effective” or “evidence-based” and held up as “model” programs to be replicated. The assumption is that the same successful impacts found at a particular setting can be replicated in other settings or on the national scale.
This faulty reasoning is based upon the “single-instance fallacy.” This fallacy occurs when a person believes that a small-scale social program that appears to work in one instance will yield the same results when replicated elsewhere. Compounding the effects of this fallacy, we often do not truly know why an apparently effective program worked in the first place. So how can we replicate it?
The social programs, including prisoner reentry programs, that have been deemed “effective” and serve as “model” programs have often been implemented under optimal conditions by highly trained professionals operating under ideal conditions. In addition, the conditions under which these programs operate are carefully monitored to ensure that the participants receive the intended level of treatment. In the real world, program conditions are almost always less than optimal. Merely replicating an “evidence-based” program does not necessarily mean that it will produce the same results. In addition, the federal government has a poor track record of replicating successful programs on a national scale.
TJRD could not replicate the positive results from CEO. Further, the approaches used by MCORP and PRI need to be replicated in different settings to adequately determine the merits of these approaches.
The experimental evaluations discussed in this review offer mixed, at best, evidence of success. This conclusion is particularly relevant for employment-based reentry programs. A previous scholarly review of random assignment evaluations of employment-focused prisoner reentry programs concluded that these programs do not reduce recidivism. Developing a better understanding of how former prisoners desist from crime has important ramifications for advocates of prisoner reentry programs and victims of crime.
The results of scientifically rigorous evaluations, especially the RExO evaluation, raise serious doubts about the overall effectiveness of federal involvement in subsidizing state and local prisoner reentry programs and congressional attempts to pass versions of the Second Chance Reauthorization Act (S. 1513 and H.R. 3406). One assumption made by supporters of prisoner reentry programs is that helping released prisoners find employment reduces recidivism rates. In fact, both reauthorization efforts devote federal resources to funding transitional jobs programs that subsidize the employment of former prisoners. The causal assumption underlying this approach is that an employed ex-con will be less likely to commit crime. In essence, workplace attachment is thought to encourage “desistance” from crime. (Desistance is what criminologists call the process of criminals giving up on criminal behavior.)
As criminologists Torbjørn Skardhamar of Statistics Norway and Jukka Savolainen of the University of Nebraska at Omaha point out, the critical question concerns the timing of employment transitions in relation to desistance from crime. There are two hypotheses on the relationship between employment and desistance. The turning point hypothesis postulates that desistance occurs after released prisoners find employment, while the maturation perspective presumes that desistance from crime occurs before the acquisition of legitimate employment. If the maturation perspective is a more accurate explanation, then helping released prisoners find employment before they are ready to give up criminal behavior may be unproductive.
Criminologist Ray Paternoster and his colleagues are positing a similar theory that the process of changing an offender’s identity from a criminal to a law-abiding citizen is a complex process that needs to precede finding legitimate employment. For instance, former prisoners need to realize that criminal offending is more costly than beneficial. Once this realization occurs, the individual can adopt a more pro-social identity that eschews “quick and easy money,” such as theft and drug dealing, for more conventional employment.
Some evidence supports this theory. Based on a sample of 783 recidivist males from Norway, Skardhamar and Savolainen found that most of the offenders turned away from criminal behavior before finding legitimate employment and that becoming employed was not linked to reduced criminal behavior—adding support for the maturation hypothesis. Giving up on the criminal lifestyle precedes finding and maintaining employment.
If the perspective of Paternoster and his colleagues is a more accurate explanation of the process of giving up on crime, helping released prisoners find employment before they are psychologically ready to give up criminal behavior may be unproductive. This means that prisoner reentry efforts that rely mainly on job training and subsidized jobs are not likely to succeed beyond a few small-scale examples. The results of the RExO evaluation certainly undercut the assumption that federally funded employment-based reentry programs can be effective.
Prisoner reentry programs that adopt a multifaceted approach, such as MCORP, may have more realistic chances of success. Nevertheless, more experimental evaluations, especially large-scale multisite evaluations, are needed to shed light on what works and what does not.—David B. Muhlhausen, PhD, is a Research Fellow for Empirical Policy Analysis in the Center for Data Analysis, of the Institute for Economic Freedom and Opportunity, at The Heritage Foundation.
 Matthew R. Durose, Alexia D. Cooper, and Howard N. Snyder, “Recidivism of Prisoners Released in 30 States in 2005: Patterns from 2005 to 2010,” U.S. Department of Justice, Office of Justice Programs, Bureau of Justice Statistics, April 2014, http://www.bjs.gov/index.cfm?ty=pbdetail&iid=4986 (accessed September 29, 2015).
 Patrick A. Langan and David J. Levin, “Recidivism of Prisoners Released in 1994,” U.S. Department of Justice, Office of Justice Programs, Bureau of Justice Statistics, June 2, 2002, http://www.bjs.gov/index.cfm?ty=pbdetail&iid=1134 (accessed September 29, 2015).
 For the purposes of this review of the scientific literature, prisoner reentry programs are defined as programs that provide services to recently released inmates. By this definition, correctional programs that provide services only to currently incarcerated individuals are excluded from this review. However, programs that provided services inside prison and outside correctional institutions are included as prisoner reentry programs.
 After conducting a meta-analysis of 308 criminal justice program evaluations, Professor David Weisburd of George Mason University and his colleagues found that weaker evaluation designs, such as quasi-experiments, are more likely to find favorable intervention effects and less likely to find harmful intervention effects. They caution that quasi-experimental designs, no matter how well designed, may be incapable of controlling for the unobserved factors that make individuals more likely to respond favorably to the intervention. David Weisburd, Cynthia M. Lum, and Anthony Petrosino, “Does Research Design Affect Study Outcomes in Criminal Justice?” The ANNUALS of the American Academy of Political and Social Sciences, Vol. 578, No. 1 (November 2001), pp. 50–70.
 Grant Duwe and Michelle King, “Can Faith-Based Correctional Programs Work? An Outcome Evaluation of the InnerChange Freedom Initiative in Minnesota,” International Journal of Offender Therapy and Comparative Criminology, Vol. 57, No. 7 (July 2013), pp. 813–841.
 Grant Duwe, “The Benefits of Keeping Idle Hands Busy: An Outcome Evaluation of a Prisoner Reentry Employment Program,” Crime and Delinquency, Vol. 61, No. 4 (May 2015), pp. 559–586. EMPLOY is the actual name of the program, not an acronym.
 Anthony A. Braga, Anne M. Piehl, and David Hureau, “Controlling Violent Offenders Released to the Community: An Evaluation of the Boston Reentry Initiative,” Journal of Research in Crime and Delinquency, Vol. 46, No. 4 (November 2009), pp. 411–436.
 For detailed discussion of the methodological weaknesses of propensity score analysis, see David B. Muhlhausen, Do Federal Social Programs Work? (Santa Barbara, CA: Praeger, 2013).
 Christy A. Visher and Jeremy Travis, “Transitions from Prison to Community: Understanding Individual Pathways,” Annual Review of Sociology, Vol. 29 (2003), pp. 89–113.
 Andrew Wiegand et al., Evaluation of the Re-Integration of Ex-Offenders (RExO) Program: Two-Year Impact Report, Social Policy Research Associates, May 2015, http://wdr.doleta.gov/research/FullText_Documents/ETAOP_2015-04.pdf (accessed August 19, 2015).
 Steve Aos, Marna Miller, and Elizabeth Drake, “Evidence-Based Public Policy Options to Reduce Future Prison Construction, Criminal Justice Costs, and Crime Rates,” Washington State Institute for Public Policy, October 2006, http://www.wsipp.wa.gov/rptfiles/06-10-1201.pdf (accessed November 20, 2015), and Doris Layton Mackenzie, What Works in Corrections: Reducing the Criminal Activities of Offenders and Delinquents (Cambridge: Cambridge University Press, 2006).
 David J. Farabee, Rethinking Rehabilitation: Why Can’t We Reform Our Criminals? (Washington, DC: AEI Press, 2005).
 Further, two evaluations that compromised their original random assignment designs were excluded from this review. Linda G. Smith and Denise R. Suttle, An Outcome Evaluation of the Pennsylvania Department of Corrections’ Community Orientation and Reintegration Program, International Association of Reentry, November 20, 2008, and Minnesota Department of Corrections, Final Report on Serious Offender Accountability Restoration (SOAR) Project, October 2006, http://www.doc.state.mn.us/pages/files/large-files/Publications/ProjectSOAR_FinalReport10-06.pdf (accessed October 19, 2015).
 Wiegand et al., Evaluation of the Re-Integration of Ex-Offenders (RExO) Program: Two-Year Impact Report.
 Ibid., pp. I-7 and I-11.
 Ibid., p. II-2.
 Ibid., pp. II-6–II-7, Table II-1.
 Ibid., p. II-8.
 Ibid., pp. III-3, Table III-1.
 Ibid., pp. IV-3, Table IV-1.
 Ibid., p. IV-15.
 Cindy Redcross et al., Transitional Jobs for Ex-Prisoners: Implementation, Two-Year Impacts, and Costs of the Center for Employment Opportunities (CEO) Prisoner Reentry Program, MDRC, August 2009, http://www.mdrc.org/publication/transitional-jobs-ex-prisoners (accessed October 19, 2015).
 Ibid.; and Cindy Redcross et al., More Than a Job: Final Results from the Evaluation of the Center for Employment Opportunities (CEO) Transitional Jobs Program, U.S. Department of Health and Human Services, Administration for Children and Families, Office of Planning, Research and Evaluation OPRE Report No. 2001-08, January 2012, http://www.acf.hhs.gov/programs/opre/resource/more-than-a-job-final-results-from-the-evaluation-of-the-center-for (accessed October 19, 2015).
 Redcross et al., More Than a Job, p. 27.
 Redcross et al., Transitional Jobs for Ex-Prisoners, p. 48, Table 4.1.
 Redcross et al., More Than a Job, p. 30, Table 3.1.
 Redcross et al., Transitional Jobs for Ex-Prisoners, p. ES-8, Table ES.1, and Redcross et al., More Than a Job, p. 32, Table 3.2.
 Redcross et al., Transitional Jobs for Ex-Prisoners, p. ES-8, Table ES.1.
 Redcross et al., More Than a Job, p. 32, Table 3.2.
 Redcross et al., Transitional Jobs for Ex-Prisoners, p. ES-8, Table ES.1.
 Redcross et al., More Than a Job, p. 32, Table 3.2.
 Redcross et al., Transitional Jobs for Ex-Prisoners, p. ES-8, Table ES.1.
 Redcross et al., More Than a Job, p. 32, Table 3.2.
 Redcross et al., Transitional Jobs for Ex-Prisoners, p. 10.
 Ibid., pp. 71–72, Table 5.3, and Redcross et al., More Than a Job, pp. 41–42, Table 3.4.
 Erin Jacobs, “Returning to Work After Prison: Final Results from the Transitional Jobs Reentry Demonstration,” MDRC, May 2012, http://www.mdrc.org/publication/returning-work-after-prison/file-full (accessed September 8, 2015).
 Ibid., p. 4.
 Ibid., p. 5.
 Ibid., p. 7.
 Ibid., p. 35.
 Ibid., p. 7.
 Ibid., p. 11, Figure 1.
 Ibid., p. 13, Table 3.
 Ibid., p. 35.
 Ibid., p. 27.
 Ibid., p. 28, Table 7.
 Ibid., pp. 30–31, Table 8.
 Grant Duwe, “Evaluating the Minnesota Comprehensive Offender Reentry Plan (MCORP): Results from a Randomized Experiment,” Justice Quarterly, Vol. 29, No. 3 (2012), pp. 347–383.
 Ibid., p. 353.
 Ibid., p. 352.
 Ibid., p. 353.
 Ibid., p. 365, Table 2.
 Ibid., p. 368, Table 5.
 Ibid., p. 366.
 Valerie A. Clark, “Making the Most of Second Chances: An Evaluation of Minnesota’s High-Risk Revocation Reduction Reentry Program,” Journal of Experimental Criminology, Vol. 11, No. 2 (June 2015), pp. 193–215.
 Ibid., p. 193.
 Ibid., pp. 195–196.
 Public Law 110–199.
 Clark, “Making the Most of Second Chances,” p. 208, Table 4.
 Ibid., p. 208.
 David Farabee, Sheldon X. Zhang, and Benjamin Wright, “An Experimental Evaluation of a Nationally Recognized Employment-Focused Offender Reentry Program,” Journal of Experimental Criminology, Vol. 10, No. 3 (September 2014), pp. 309–322.
 Ibid., p. 315.
 Ibid., p. 316.
 Ibid., p. 309.
 Philip J. Cook et al., “An Experimental Evaluation of a Comprehensive Employment-Oriented Prisoner Re-Entry Program,” Journal of Quantitative Criminology, Vol. 31, No. 3 (September 2015), pp. 355–382.
 Ibid., p. 363.
 Ibid., p. 370.
 Ibid., p. 371, Table 4.
 Ibid., p. 370.
 Ibid., p. 371, Table 4.
 Ibid., p. 373.
 Ibid., pp. 374–375, Table 6.
 Eric Grommon, William S. Davidson II, and Timothy S. Bynum, “A Randomized Trial of a Multimodal Community-Based Prisoner Reentry Program Emphasizing Substance Abuse Treatment,” Journal of Offender Rehabilitation, Vol. 54, No. 4 (2013), pp. 287–309.
 Ibid., p. 288.
 Ibid., p. 290.
 Ibid., p. 292.
 Ibid., p. 295.
 Ibid., p. 297, Table 3.
 Ibid., p. 298, Table 4.
 Ibid., p. 297, Table 3, and p. 298, Table 4.
 Susan Turner and Joan Petersilia, “Work Release in Washington: Effects on Recidivism and Corrections Costs,” The Prison Journal, Vol. 76, No. 2 (June 1996), pp. 138–164.
 Ibid., p. 157.
 Ibid., p. 160.
 James A. Wilson and Robert C. Davis, “Good Intentions Meet Hard Realities: An Evaluation of the Project Greenlight Reentry Program,” Criminology & Public Policy, Vol. 5, No. 2 (May 2006), pp. 303–338.
 Ibid., p. 307.
 Edward E. Rhine, Tina L. Mawhorr, and Evalyn C. Parks, “Implementation: The Bane of Effective Correctional Programs,” Criminology & Public Policy, Vol. 5, No. 2 (May 2006), pp. 347–357.
 Christie A. Visher, “Effective Reentry Programs,” Criminology & Public Policy, Vol. 5, No. 2 (May 2006), p. 300.
 Wilson and Davis, “Good Intentions Meet Hard Realities,” p. 311.
 Ibid., p. 323.
 For more information on evidence-based policymaking, see David B. Muhlhausen, “Evidence-Based Policymaking: A Primer,” Heritage Foundation Backgrounder No. 3036, October 15, 2015, at http://www.heritage.org/research/reports/2015/10/evidence-based-policymaking-a-primer.
 Stuart M. Butler and David B. Muhlhausen, “Can Government Replicate Success?” National Affairs, No. 19 (Spring 2014), pp. 25–39, http://www.nationalaffairs.com/publications/detail/can-government-replicate-success (accessed April 13, 2015).
 Muhlhausen, Do Federal Social Programs Work? and ibid.
 Muhlhausen, Do Federal Social Programs Work?
 Christy A. Visher , Laura Winterfield, Mark B. Coggeshall, “Ex-Offender Employment Programs and Recidivism: A Meta-Analysis,” Journal of Experimental Criminology, September 2005, Vol. 1, No. 3 (September 2005), pp. 295–316.
 Torbjørn Skardhamar and Jukka Savolainen, “Changes in Criminal Offending Around the Time of Job Entry: A Study of Employment and Desistance,” Criminology, Vol. 52, No. 2 (May 2014), pp. 263–291.
 Ray Paternoster et al., “Human Agency and Explanations of Criminal Desistance: Arguments for a Rational Choice Theory,” Journal of Developmental and Life-Course Criminology, Vol. 1, No. 3 (September 2015), pp. 209–235, and Ray Paternoster and Shawn Bushway, “Desistance and the Feared Self: Toward an Identity Theory of Criminal Desistance,” Journal of Criminal Law and Criminology, Vol. 99, No. 4 (Summer 2009), pp. 1103–1156.
 Paternoster et al., “Human Agency and Explanations of Criminal Desistance.”