Design: A quasi-experiment was conducted in which participants in PIECP were compared to individuals participating in traditional industries (TI) or activities other than industrial work (OTW). Regressions included Cox Proportional Hazard, Weibull Model, Lognormal Model, Ordinary Least Squares, probit, and Heckman Selection.

Findings: One regression model found that the treatment group had significantly delayed time until re-arrest, delayed time from post-release employment to job loss, higher probability of obtaining employment, and higher earnings than one or both of the comparison groups; however, these findings did not hold up across other models. No significant differences were observed in any model with respect to the time until re-conviction or the time from release until obtaining employment.

Limitations: Selection bias may result from the voluntary nature of the program and from administrators possibly selecting applicants who were more likely to succeed; the comparison group may have received parts of the intervention; 20-30% of study participants were lost due to missing data.

Sample Size: Total N: 890 (Treatment group: 303; Comparison group: 587) (Note: values estimated using percentages provided by the researcher)

Follow-Up Period: 2 to 7.5 years

Design: A quasi-experimental design was employed. Statistical techniques included logistic regression, Cox proportional hazard models, propensity score matching, and other statistical controls for group differences.

Findings: No significant effects were found for prevalence of re-arrest, time to first re-arrest, prevalence of recommitment to federal prison, or time to recommitment to federal prison.

Limitations: Program participation was voluntary, so unmeasured motivation could introduce selection bias and influence analysis outcomes.

Sample Size: Total N: 13,066 (Treatment group: 1,946; Comparison group: 11,120)

Follow-Up Period: 3 years after release for dichotomous outcomes; 2 years after release for survival models

Design: A quasi-experimental study was conducted with a comparison group of prisoners released during the same time period as the treatment group who had maxed out of their prison sentences and who therefore were not placed in the Day Reporting Center (DRC). Logistic regression and survival analyses were used to assess group differences in recidivism.

Findings: DRC treatment was associated with significantly lower likelihood of re-arrest, re-conviction, and re-incarceration.
DRC participants were also more likely to avoid re-arrest for a longer period of time than the comparison group, but this result did not reach statistical significance (though it did approach significance at p=.059).

Limitations: Regression models did not control for either risk score or in-prison variables, such as behavior in prison programming; there may be inherent differences between groups that contribute to the observed treatment effects.

Sample Size: Total N: 335 (Treatment group: 135; Comparison group: 200)

Follow-Up Period: 2.5 to 3.5 years

Design: The study employed a quasi-experimental design with a comparison group that was matched to the treatment group on age, race, current offense, arrest history, and gang involvement. Cox proportional hazard models were used to examine differences in re-arrest over the follow-up period.

Findings: Relative to comparison group subjects, BRI participants were significantly less likely to be re-arrested over the follow-up period.

Limitations: Treatment and comparison groups may vary on unobserved characteristics; the comparison group consisted of individuals released 2 years earlier than members of the treatment group.

Sample Size: Total N: 417 (Treatment group: 108; Comparison group: 309)

Follow-Up Period: 3 years

Design: A quasi-experimental study was conducted with a comparison group of prisoners who were released during the same year as the treatment group and who had maxed out of their prison sentences and therefore were not placed in the Halfway Back (HWB) Program. Logistic regression was used to assess group differences in recidivism.

Findings: HWB treatment was associated with lower rates of re-arrest, re-conviction, and re-incarceration during the follow-up period, and HWB participants were also significantly more likely to avoid re-arrest longer than the comparison group.

Limitations: Regression models did not control for either risk score or in-prison variables, such as behavior in prison programming; there may be inherent differences between groups that contribute to the observed treatment effects.

Sample Size: Total N: 381 (Treatment group: 181; Comparison group: 200)

Follow-Up Period: 2.5 to 3.5 years

Design: A quasi-experimental design was employed, with propensity score matching used to select comparison group members from administrative data.

Findings: Rates of conviction for a new felony offense were 19% lower for the treatment group than the comparison group, and rates of conviction for any new offense were 22% lower for the treatment group; both findings were statistically significant.

Limitations: Participation in the program was voluntary, potentially creating issues of selection bias; the comparison group consisted of individuals released from custody approximately 3 years before members of the treatment group.

Sample Size: Total N: 128 (Treatment group: 64; Comparison group: 64)

Follow-Up Period: 2 years

Design: This study used a quasi-experimental design. Statistical analysis included OLS regression, difference-in- difference specification, propensity score methods, and linear probability models.

Findings: Recidivism was significantly lower among the treatment group 2 and 3 years after release; mean quarterly earnings were significantly higher among the treatment group in all 3 years after release; and, in two of three statistical models, employment probability was significantly higher among the treatment group in all 3 years.

Limitations: Self-selection bias may be present, given that the treatment group volunteered to participate in the program, while the comparison group did not.

Sample Size: Total N: 9,221 (Treatment group: 2,186; Comparison group: 7,035)

Follow-Up Period: 12-36 months after release

Design: A quasi-experimental design was employed with a matched comparison group composed of individuals released during the same time as the treatment group. A proportional hazards Weibull model was used to determine treatment effects on the likelihood of recidivism.

Findings: ComALERT treatment was associated with significant reductions in re-arrest and re-conviction over the 2-year follow-up period, but differences in re-incarceration failed to reach significance. Treatment was also associated with increases in employment and earnings, though these findings are likely due in part to the fact that transitional jobs were provided as part of the program.

Limitations: Some parolees who were assigned to ComALERT did not attend; because the study did not employ random assignment, it is possible that treatment and comparison groups vary on unobserved characteristics.

Sample Size: Total N: 896 (Treatment group: 448; Comparison group: 448)

Follow-Up Period: 2 years

Design: This study utilized a quasi-experimental design with logistic regression to control for group differences. The comparison group included individuals who were released during the same period as the treatment group and would have been eligible for work release but did not participate in the program.

Findings: A significant reduction in the rate of re-conviction was observed for the treatment group over the 3-year period. However, no significant differences were observed with respect to the rate of violent felony convictions.

Limitations: Findings may not be generalizable to a broad prison population; participants were found to be systematically different from non-participants; the study is based on observational data and relies upon limited statistical controls.

Sample Size: Total N: 15,326 (Treatment group: 11,413; Comparison group: 3,913)

Follow-Up Period: 36 months

Design: Researchers conducted a quasi-experiment featuring all participants from from program inception in 1999 to 2004. The comparison group was composed of non-participants who were also in custody between 1999 and 2004 and who were matched to the treatment group on the basis of risk to recidivate, need level, ethnicity, number of violent offenses, and age. Cox proportional hazards regression was used to evaluate the program’s effect on recidivism.

Findings: A survival analysis found no significant differences between the treatment and comparison groups in terms of either new convictions or new violent convictions.

Limitations: A significant difference between the treatment and comparison groups was found with respect to motivation for treatment, but this difference was not controlled for in regression analysis.

Sample Size: Total N: 966 (Treatment group: 500; Comparison group: 466)

Follow-Up Period: Average of approximately one year

Design: A quasi-experimental design was employed, and multivariate logistic regression was used to control for group differences. The comparison group was drawn from parolees who were eligible for the program but whom administrators did not select to receive services.

Findings: Those who received a pre-release assessment were more likely to attend post-release mental health treatment, and those attending post-release treatment were in turn less likely to recidivate.

Limitations: Significant differences were found in demographic, parole region, offense category, and mental health status variables between treatment and comparison groups; attendance in the program is at least partially at the discretion of the parolee, which may result in selection bias between the treatment and comparison groups.

Sample Size: Total N: 60,912 (Treatment group: 32,322; Comparison group: 28,590)

Follow-Up Period: 12 months

Design: A quasi-experimental design was used in which two comparison groups were matched to the treatment group—one composed of individuals participating in traditional industries (TI) and the other composed of those who participated in activities other than work (OTW) while incarcerated. The treatment group was drawn from PIECP programs in five states. Findings were evaluated using survival analyses.

Findings: The treatment group took more time to be re-arrested, re-convicted, and re-incarcerated; spent more time employed; and earned higher wages relative to one or both of the comparison groups. All of these findings were significant.

Limitations: The process by which individuals were enrolled in treatment versus comparison activities is unclear, leaving open the possibility of selection bias; the comparison groups were not matched to the treatment group on criminal history.

Sample Size: Total N: 6,464 (Treatment group: 2,333; Comparison group (TI): 1,863; Comparison group (OTW): 2,268)

Follow-Up Period: 2-7.5 years

Design: The researchers used a partially randomized design in which the treatment group consisted of those assigned to Project Greenlight; one comparison group consisted of those enrolled in the regular Transitional Services Program, operating at the same facility as Project Greenlight; and the second comparison group consisted of those released with no pre-release programming. Cox regression was used to determine differences between groups in recidivism.

Findings: Project Greenlight programming was associated with an increased likelihood of re-arrest and parole revocation and a reduced time until re-arrest relative to both of the comparison groups.

Limitations: Assignment to the treatment and comparison groups occurred in varied and inconsistent ways, potentially introducing systematic bias into the study (however, no pre-existing differences between groups were observed).

Sample Size: Total N: 735, (Treatment group: 344; Comparison group (TSP): 278; Comparison group (Upstate): 113)

Follow-Up Period: 1-2 years

Design: A quasi-experimental design was used in which the comparison group was drawn from a group of mentally ill prisoners released about 4 years before the treatment group. Logistic and Cox proportional hazards regression were used to evaluate recidivism outcomes and to control for pre-existing differences between groups.

Findings: The treatment group experienced significantly better recidivism outcomes (in terms of both the likelihood of re-conviction and the time until re-conviction) than the comparison group.

Limitations: The comparison group was released 4 years before the treatment group, so the groups may have faced different economic and other circumstances upon release; there may be selection effects resulting from the fact that treatment group participants were hand-selected by corrections staff and a Statewide Review Committee, while the comparison group was not.

Sample Size: Total N: 387 (Treatment group: 100; Comparison group: 287)

Follow-Up Period: 18 months

Design: A two-site quasi-experimental design was employed in which the treatment group was released with Medicaid benefits on the day of discharge, while the comparison group was released without benefits. Generalized estimating equations and Cox proportional hazards regression were used to evaluate recidivism outcomes.

Findings: In both study sites, Medicaid benefits were found to be associated with reduced recidivism, but this effect generally was not significant when controlling for the rate of accessing mental health services in the community.

Limitations: The effects of Medicaid may differ from state to state, depending on what services are covered, thereby limiting the generalizability of the findings; study results are only applicable to individuals eligible for Medicaid and who retain their benefits while incarcerated or whose benefits are reinstated upon release.

Sample Size: Pinellas County: Total N: 2,419 (Treatment group: 1,877, Comparison group: 542), King County: Total N: 5,189 (Treatment group: 3,346; Comparison group: 1,843)

Follow-Up Period: 12 months

Design: A quasi-experimental design was conducted in which the treatment group consisted of individuals participating in the treatment program while incarcerated and the comparison group consisted of inmates who were eligible for the program but did not have enough time left on their sentence to complete it and therefore were not enrolled. Logistic regression was used to determine the program’s impact on post-release outcomes.

Findings: The likelihood of remaining abstinent from alcohol and substance use was higher for program participants at three-months post-release, but in regression models, no significant differences were found with respect to rearrest, reincarceration, or mental stability at either three or twelve months post-release.

Limitations: Small sample size and collinearity threaten the strength of regression models; analyses do not control for two statistically significant differences between the treatment and comparison groups; cases with insufficient follow-up data are excluded; the rate of program non-completion was high (75%).

Sample Size: 3-month follow-up: Total N: 166 (Treatment group: 109; Comparison group: 57); 12-month follow-up: Total N: 125 (Treatment group: 70; Comparison group: 55)

Follow-Up Period: 1 year

Design: A group of court-defined first-time DWI (driving while intoxicated) offenders participating in a DWI programming were randomly assigned to participate in Victim Impact Panels. Treatment effects were assessed using logistic regression and chi square tests.

Findings: No significant differences were found with respect to re-arrests for DWI or self-reported DWI.

Limitations: Sample size was relatively small; self-report data used at 2-month follow-up; no main treatment effect determined.

Sample Size: Total N: 99 (Treatment group: 56; Control group: 43)

Follow-Up Period: 2 years

Design: A randomized design was employed in which the control group received all services in the facility except for the psychoeducational component of the EQUIP program. No significant differences were found between groups with respect to several demographic and mental health characteristics.

Findings: Female treatment group members were significantly less likely than female control group members to recidivate, but no significant difference was found for males.

Limitations: The control group participated in many of the same programming and services as the treatment group; non-completers across both the treatment and control groups were excluded from analysis.

Sample Size: Total N: 250 (Treatment group: 128; Control group: 122)

Follow-Up Period: 6 months

Design: This study used a randomized controlled trial in which participants had been diagnosed with an Axis I Psychiatric Diagnosis and were placed on parole. No statistical controls for pre-existing group differences were included.

Findings: The rate of returning to jail and the number of days spent in jail during the 12-month program were significantly lower among treated individuals than controls. However, neither the rate of re-conviction during the 12-month treatment phase nor the rate of return to jail during the 6 months after program termination was significantly lower among the treatment group.

Limitations: The treatment group consisted of older individuals and more females than the control group, and no statistical methods were employed to control for these or other potential between-group differences.

Sample Size: Total N: 448 (Treatment group: 225; Control group: 223)

Follow-Up Period: 18 months

Design: This study utilized a quasi-experimental design. The comparison group was incarcerated during the same time period as the treatment group, and the groups were matched by sex, offense type, age, race, and years incarcerated. Statistical analyses involved logistic regression models.

Findings: 17% of the treatment group was re-incarcerated, compared to 24% of the comparison group, and a logistic regression model indicated that this difference was significant. Furthermore, 69% of the treatment group obtained employment, compared to 58% of the comparison group—also a significant difference.

Limitations: The treatment group may differ from the comparison group in areas other than the treatment; the author was unable to control for members of the comparison group who may have previously received parts of the treatment or similar services; matching was conducted on a limited number of variables.

Sample Size: Total N: 827 (Treatment group: 424; Comparison group: 403)

Follow-Up Period: Recidivism: 36 months after release

Employment: 12 months after release

Design: Participants were randomly assigned to either the treatment group, which received therapeutic community treatment and optional aftercare, or the control group, which received other mental health services. Researchers examined both the full treatment group as well as those who participated in aftercare and those who did not. Statistical analyses involved logistic regression models and Kaplan-Meier survival methods.

Findings: Beneficial treatment effects were found for a host of substance abuse outcomes, as well as reincarceration; however, no effects were found for several other recidivism measures. When the treatment group was split into those who received aftercare and those who did not, only the group that received aftercare displayed lower recidivism than the control group.

Limitations: Although the study used random assignment for the in-prison component of the treatment, participation in the aftercare component was voluntary, introducing the possibility of self-selection bias into some of the sub-group analyses; follow-up data was unavailable for 75% of the original sample, and attrition rates differed among groups; deviations from the random assignment process occurred in 21% of cases.

Sample Size: Total N: 139 (Treatment group: 75; Control group: 64)

Follow-Up Period: 1 year

Design: This study used a quasi-experimental design with statistical controls for differences between groups. The comparison group was selected from a population of parolees in 1996. Statistical analysis involved the use of a logistic regression model.

Findings: A logistic regression model found no statistically significant evidence that the treatment had an impact on recommitment to prison.

Limitations: Statistical controls were limited to age, race, and original offense; information with which to assess whether there is a potential for selection bias that could be responsible for any treatment effects was insufficient.

Sample Size: Total N: 530 (Treatment group: 107; Comparison group: 423)

Follow-Up Period: 2 years

Design: Quasi-experimental design with matching on age, race, sex offender status, criminal history, and other characteristics. Comparison group consisted of individuals not placed in halfway houses. Logistic regression was used to control for differences between groups.

Findings: Regardless of inmate’s risk level, programs overall showed a 6% reduction in recidivism. 47% of the programs were found to have a beneficial treatment effect. Effects were strongest for high-risk offenders, while a harmful effect on low-risk offenders was observed.

Higher program quality was also associated with greater impacts on recidivism.

Limitations: No information was provided on the treatments/ programs comparison group received (if any).

Sample Size: Total N: 6,795 (Treatment group: 3,737; Comparison group: 3,058)

Follow-Up Period: 24 months

Design: This study utilized a quasi-experimental design with logistic regression to control for group differences. The comparison group consisted of inmates in the Beaver County Jail who did not participate in the work release program.

Findings: A correlation matrix indicated that program participation was significantly correlated with reduced recidivism. However, significant group differences disappeared when controlling for demographic and criminal history variables in the logistic regression model. ssion model.

Limitations: The program appears to be voluntary, so results may reflect selection bias; relevant variables that may affect recidivism, such as drug abuse and length of incarceration, were not included in the logistic regression models; data on post-release incarceration appear to have been limited to local correctional facilities.

Sample Size: Total N: 448 (Treatment group: 142; Comparison group: 306)

Follow-Up Period: 36 months after release

Design: A randomized study was conducted in which participants from economically depressed communities who had official arrest histories were assigned to treatment or control groups and required to submit self-reported work, crime, and arrest information at 9-month intervals for up to 3 years.

Findings: Program participants over 26 years of age had lower rates of recidivism and longer times to arrest than participants under 26 and the control group. However, participants under 26 had no significant program impacts, with the exception of those who were gainfully employed, who had lower recidivism rates than their under-26 control group counterparts.

Limitations: Drug users and youth dropouts with arrest histories were included in the sample along with recently release prisoners, drawing into question the validity of findings specifically for the returning prisoner population; dependent variables were based on self-reported arrests, potentially overstating a deterrence effect.

Sample Size: Total N: Varied across analyses (Treatment group: 1,821 to 2,052; Control group: 1,937 to 2,210)

Follow-Up Period: 36 months

Design: A randomized controlled trial was employed in which treatment group members were incarcerated at one of two social-therapeutic institutes in Germany, while control group members were assigned to normal prisons.

Findings: A smaller proportion of the treatment group received a new sentence than the control group over the follow-up period, but this difference was not significant.

Limitations: No analysis of pre-existing differences between groups was conducted.

Sample Size: Total N: 223 (Treatment group: 111; Control group: 112)

Follow-Up Period: 5 years

Design: A quasi-experimental design with statistical controls for group differences was conducted. The treatment group consisted of those enrolled into any Preventing Parolee Crime Program (PCPP), whereas the comparison group consisted of those opting not to enroll in any PPCP and paroled during the study period.

Findings: Participation in PPCP was associated with a significant reduction in re-incarceration during the follow-up period relative to the comparison group.

Limitations: Because treatment group participants opted to enroll in the program whereas the comparison group did not, it is possible that the effects attributed to the program may actually be due to motivational or other differences between groups.

Sample Size: Total N: 176,941 (Treatment group: 28,598; Comparison group: 148,343)

Follow-Up Period: 1 year

Design: This study included both a randomization component and a quasi-experimental component. A total of 127 prisoners were randomly assigned to either the control or treatment group. To increase the sample size, 48 prisoners were added to the treatment group and matched to 45 prisoners who did not participate in work release. Chi-square tests and logistic regression were used to determine significance.

Findings: At 1 year after assignment to the treatment or comparison group, 22.2% of the treatment group had been re-arrested and 4.7% had been re-incarcerated, compared to 30.2% and 4.7% of the control group, respectively. Analyses found that these differences were not significant.

Limitations: Some members of the sample were randomly assigned while others were matched; on average, the sample had been in the community following release for only 3 months by the end of the 12-month follow-up period, which is a limited time frame with which to measure recidivism outcomes.

Sample Size: Total N: 218 (Treatment group: 112; Control group: 106)

Follow-Up Period: 12 months from day of assignment to the program

Design: This study involved both random assignment and matching components. Program volunteers were randomly assigned to the treatment or control group. Because of the small size of the control group, treatment group participants were matched to an additional population of prisoners on several characteristics. Bivariate tests were used to determine significance of outcomes.

Findings: No significant differences were found on any of the three recidivism outcomes or the nine employment outcomes tested, with the exception that the comparison group was significantly more likely to be arrested—but had a longer time until re-arrest—than both the treatment and control groups.

Limitations: Within the treatment group, final analyses included only participants who graduated from the STEP program; follow-up time for return to prison was tracked for 6 months longer for the comparison group than for the other two groups.

Sample Size: Total N: 299 (final recidivism analysis), (Treatment group: 63; Control group: 36; Comparison group: 200)

Follow-Up Period: 3-12 months after release

Design: Quasi-experimental design with probit and ordinary least squares regressions. Comparison and treatment groups were drawn from administrative data. Treatment group received electronic monitoring, and comparison group was placed in a halfway house.

Findings: No significant difference was found between treatment and comparison groups with respect to recidivism. However, individuals in the halfway house group held employment for a shorter period of time than those on electronic monitoring.

Limitations: Potential for selection bias in that those assigned to halfway houses may have greater needs and lower employment prospects than those assigned to electronic monitoring; use of independent variables in the regression was limited by the data.

Sample Size: Total N: 192 (Treatment group: 98; Comparison group: 94)

Follow-Up Period: 12 months

Design: This study utilized a quasi-experimental design in which the treatment and comparison groups were drawn from seven maximum-security prisons. Statistical analyses involved the use of logit regression models and survival analyses.

Findings: In the logit models, no significant differences were found with respect to recidivism outcomes, including post-release felony arrest and re-admission to prison, over the 24-month follow-up period. Survival analyses also found no significant differences in time until felony re-arrest and re-admission to prison.

Limitations: The program was voluntary, and thus results may reflect pre-existing differences between groups; initial sample size was reduced by about 20% due to missing data.

Sample Size: Total N: 896 (Treatment group: 399; Comparison group: 497)

Follow-Up Period: 24 months after release

Design: A quasi-experiment was conducted using a three-site pooled treatment group matched to a comparison group of individuals who did not learn Transcendental Meditation while incarcerated. Subsequent 15-year follow-up analyses were conducted in one of the three sites (Folsom Prison). Stepwise multiple regression was used to determine treatment effects, and for the long-term outcomes for Folsom Prison participants, survival analysis and analyses of offense severity were also conducted.

Findings: Across sites and at long-term follow-up for one site, findings suggested that the program significantly decreases recidivism.

Limitations: The voluntary nature of the program may produce selection effects (i.e., there may be some differences between individuals who chose to practice meditation and those who do not that researchers were unable to control for).

Sample Size: All 3 sites combined: Total N: 511 (Treatment group: 259; Comparison group: 259), Folsom Prison site: Total N: 248 (Treatment group: 120; Comparison group: 128)

Follow-Up Period: All 3 sites: Up to 6 years

Folsom Prison site: Up to 15 years

Design: This study utilized a quasi-experimental design with groups matched by propensity score methods. The comparison group was drawn from a population of prisoners released during the same quarter as the treatment group but who did not participate in UNICOR for more than 6 months and did not complete vocational/apprentice training. Statistical analyses involved the use of chi square tests and a Cox proportional hazard model.

Findings: A survival analysis indicated that males in the treatment group were significantly less likely to recidivate over a follow-up period of 8-12 years. Chi square tests found that the treatment group (including both males and females) was significantly less likely to experience parole revocation and more likely to be employed than the comparison group during the 1st year after release; however, no significant differences were found in wages earned.

Limitations: No significant limitations were identified.

Sample Size: Total N: 4,868 (Treatment group: 2,013; Comparison group: 2,855) (Note: Actual sample included in analyses was considerably smaller and varied by outcome and follow-up period.)

Follow-Up Period: 12 months after release; additional recidivism analysis involved a follow-up of 8-12 years

Design: This study utilized a quasi-experimental design. Members of the comparison group did not participate in the intervention and were released during the same time period as the treatment group. Statistical analyses involved the use of chi square tests and multiple regression models.

Findings: No significant differences were found with respect to either re-arrest or
employment success during the follow-up periods.

Limitations: Because of the voluntary nature of the program, groups may differ in ways other than the treatment; members of the treatment and comparison group may have participated in other reentry programs.

Sample Size: Total N: 1,210 (Treatment group: 690; Comparison group: 520)

Follow-Up Period: Recidivism: 24 months after release

Employment: 5 quarters after release

Design: This study utilized a quasi-experimental design. Members of the comparison group did not participate in the intervention and were released during the same time period as the treatment group. Statistical analyses involved the use of chi square tests and multiple regression models.

Findings: No significant differences were found with respect to either re-arrest prevalence or employment during the follow-up periods.

Limitations: Due to the voluntary nature of the program, groups may differ in ways other than the treatment; limited information is provided on sample characteristics or pre-existing group differences; members of the treatment and comparison group may have participated in other reentry programs.

Sample Size: Total N: 1,210 (Treatment group: 157; Comparison group: 1,053)

Follow-Up Period: Recidivism: 24 months after release

Employment: 5 quarters after release

Design: This study utilized a quasi-experimental design. Members of the comparison group did not participate in the intervention and were released during the same time period as the treatment group. Statistical analyses involved the use of chi square tests and multiple regression models.

Findings: No significant differences were found with respect to either re-arrest or employment during the follow-up periods.

Limitations: Due to the voluntary nature of the program, groups may differ in ways other than the treatment; evaluation of group differences was limited.

Sample Size: Total N: 1,210 (Treatment group: 987; Comparison group: 223)

Follow-Up Period: Recidivism: 24 months after release

Employment: 5 quarters after release

Design: A quasi-experimental design with two comparison groups was utilized. The first group consisted of parolees released 1-5 years earlier than the treatment group, while the second group included parolees released during the same time period as the treatment group. Statistical analyses involved the use of chi-square, Fisher’s LSD, and analysis of variance (ANOVA).

Findings: 75% of the treatment group successfully completed parole, compared to 50% of the first comparison group and 62% of the second comparison group. These differences were statistically significant. The treatment group also experienced significantly fewer arrests at 6 months after release than the first comparison group, but not the second group. However, no significant differences were found with respect to the other seven recidivism outcomes tested.

Limitations: Because the program was voluntary, groups may differ on unobserved characteristics; with respect to the first comparison group, which included individuals released earlier than the treatment group, findings may be due to historical changes rather than to the treatment; groups differed significantly on several characteristics.

Sample Size: Total N: 370 (Treatment group: 140; Comparison group 1: 94; Comparison group 2: 136)

Follow-Up Period: 32 months after release to parole