What Works With Sex Offenders? Results From an Evaluation of Minnesota Circles of Support and Accountability [Corrections Today]

The use of the circles of support and Accountability (COSA) model with high-risk sex offenders began in a small Mennonite community in Canada in the early 1990s. Grounded in the tenets of the restorative justice philosophy, the COSA model attempts to help sex offenders successfully reenter the community and, thus, increase public safety by providing them with social support by meeting their employment, housing, treatment and social needs. Each COSA model consists of anywhere between four and six community volunteers - one of whom is a primary volunteer - who meet with the offender on a regular basis. The results from several evaluations of a Canadian COSA model suggest it significantly reduces sex offender recidivism.1

In 2008, the Minnesota Department of Corrections (MnDOC) implemented Minnesota Circles of Vil Support and Accountability (MnCOSA), a sex offender reentry program for males based on the Canadian COSA model. The impetus for starting MnCOSA was rooted not only in the promising results reported in the initial evaluation completed by Wilson and colleagues,2 but also in the findings from a 2008 study that examined the impact of broad community notification on sex offender recidivism.3 In determining that broad community notification significantly reduces sexual recidivism for Level 3 sex offenders (i.e., those determined to be high risk), the 2008 study found that sexual recidivism rates were highest among Level 2 sex offenders (those determined to have moderate risk). Therefore, when it began in 2008, MnCOSA targeted Level 2 sex offenders as part of a risk-management strategy to reduce sexual recidivism.

Comparisons Between the Canadian and Minnesota COSA Programs

The design and operation of MnCOSA has been similar, in many ways, to the COSA model. First, MnCOSA consists of a core member (the sex offender) and anywhere between four and six volunteers from the community. Second, the circle uses a covenant, which delineates the responsibilities to which both the core member and circle volunteers agree. Third, volunteers receive training following a selection and screening process. Fourth, the goal for each circle is to provide the core member with support during the first 12 months he is in the community. Fifth, in Minnesota, efforts have been made to establish outer circles that help support inner circles in their work.

The inner circle consists of the four-to-six community volunteers who regularly meet with the core member to provide support and help him remain accountable. The outer circle is comprised of community-based professionals (psychologists, law enforcement officers, supervision agents, social service workers, etc.) who volunteer their time to support the inner circle in its work.

Despite these similarities, however, there have been several notable differences between MnCOSA and the Canadian COSA. First, COSA was very much a grass-roots effort when it originated within a small Mennonite community, and later gained government support and involvement. In contrast, with MnCOSA, it was MnDOC (a government agency) that implemented the program and undertook outreach efforts to attract community interest and participation. Second, due perhaps to the different origins, the Canadian COSA grew organically over time, whereas MnCOSA has been developed more systematically - which is likely a consequence of the emphasis placed on evaluating the program since its inception. Third, in what is likely another consequence of the different origins for each program, COSA volunteers were mainly recruited from faith communities, while MnCOSA has experienced very little success in recruiting volunteers from local churches. Instead, MnCOSA has relied on students from local colleges and universities as a primary source of volunteers. Fourth, whereas Canadian COSA focused on working with offenders released at the expiration of their sentence, MnCOSA has not used the absence of post-release supervision as a selection criterion for core members. Fifth, while the circle process in the Canadian COSA often begins prior to an offender's release from prison, in Minnesota it begins in prison at least four weeks prior to release. Finally, under the Canadian COSA model, circles usually meet in private homes following an offender's release from prison. Under MnCOSA, circles meet in secure public venues.

MnCOSA Research Design

The MnCOSA evaluation used a randomized controlled trial to determine whether it had an impact on recidivism. Prior to randomly assigning eligible offenders to either the experimental MnCOSA or control groups (offenders not involved in MnCOSA). MnCOSA staff recruited volunteers from the community to form a circle around a soon-to-be released Level 2 sex offender who was returning to Dodge, Fillmore, Hennepin, Olmsted or Ramsey Counties. The program involves many counties because in general, the results did not show significant differences between counties, though the research design was developed to account for this possibility. Between 2008 and 2011, 31 sex offenders participated in MnCOSA and were released from prison. Recidivism outcomes for these offenders were compared to those of the 31 sex offenders in the control group. Recidivism data on all 62 offenders were collected through the end of 2011.

The MnCOSA evaluation also assessed whether the program is cost-effective by comparing program operating costs with the costs resulting from recidivism. To determine whether MnCOSA has produced a benefit resulting from reduced recidivism, the study compared the number of offenses committed by offenders in the MnCOSA and control groups. The costs of these offenses were then monetized based on cost of crime estimates developed through prior research.

Results

Recidivism. The data in Figure 1 show that MnCOSA participants had lower recidivism rates than the offenders in the control group. For example, 39 percent of the MnCOSA participants had been rearrested for a new offense by the end of December 2011, compared with 65 percent of the control group offenders. The results also show that 26 percent of the MnCOSA participants were reconvicted for a new offense compared to 45 percent in the control group. In addition, 10 percent of the MnCOSA participants were reincarcerated for a new criminal offense compared to 26 percent of the control group. Further, compared to the offenders in the control group who had a technical violation revocation rate of 68 percent, MnCOSA participants had a rate of 48 percent. Lastly, 48 percent of MnCOSA offenders returned to prison for a new offense and/or a technical violation - versus 61 percent of those in the control group.

The results from Cox regression analyses - which controlled for time at risk for recidivism and other observed differences between the two groups - showed that participating in MnCOSA had a statistically significant effect on three of the five recidivism measures. MnCOSA participation significantly lowered the risk of recidivism by 62 percent for rearrest; 72 percent for technical violation revocation; and 84 percent for any return to prison. Due mainly to the small sample size and short follow-up period, MnCOSA did not have a statistically significant effect at the 0.05 level for reconviction or new offense reincarceration.

Cost-beneRt. Because MnCOSA relies heavily on volunteers, the costs to operate the program are confined mainly to project staff salaries and volunteer training and recruitment efforts. As shown in Table 1, it cost nearly $450,000 to operate MnCOSA from 2008-2011. The results also show, however, that the benefits resulting from reduced recidivism amounted to more than $800,000. More specifically, MnCOSA participants were rearrested for 33 fewer offenses than the control group and spent about 100 fewer days in prison following their release.

After subtracting the program operating costs from the recidivism costs avoided, the results in Table 1 show that MnCOSA has, within its first four years of operation, produced an estimated benefit of $363,211, which amounts to $11,716 per participant. The cost-benefit ratio indicates that for every dollar spent on MnCOSA, the state of Minnesota has seen an estimated benefit of $1.82, which results in an 82 percent return-on-investment. To place the MnCOSA benefit-per-participant results in a broader context, it is worth considering a 2006 study by Steve Aos, Elizabeth Drake and Marna Miller on the cost-effectiveness of correctional programs. In their study, 10 programs were identified for adult offenders that produced a monetary benefit, which ranged from $870 to $13,738 per participant.4 Thus, with an estimated benefit of $11,716 per participant, MnCOSA ranks near the top for cost-effectiveness among adult correctional programs in the U.S.

Conclusion

Existing research has shown that providing offenders with a continuum of social support from prison to the community produces better recidivism outcomes.5 Consistent with this evidence, MnCOSA has been effective in helping sex offenders successfully reenter society. It is too early to tell, however, whether MnCOSA is effective in reducing sexual recidivism. The virtual absence of sexual recidivism among the offenders examined in the MnCOSA evaluation reflects the fact that recent sexual recidivism rates among Minnesota sex offenders have been relatively low.6

Reducing sexual victimization is clearly important, and it should remain a top priority for sex offender programming. However, the findings from this evaluation, especially those from the cost-benefit analyses, suggest nonsexual recidivism outcomes are important too. Indeed, of the 85 rearrests among the 62 offenders in this study, only one was for a sex offense. Moreover, the results from the cost-benefit analyses indicate that sex offender programming can still produce substantial benefits by reducing nonsexual recidivism.

MnCOSA is a cost-effective program that produces public safety benefits, but it is effective only for a relatively small number of offenders. MnCOSA has averaged nearly eight circles per year (31 in the first four years), whereas both Canada and the U.K. have averaged closer to 10 circles per year.7 Given the need for at least four volunteers per circle, combined with prevailing public perceptions regarding convicted sex offenders, recruiting enough suitable volunteers from the community to operate the COSA model on a widespread basis is a significant challenge. Because MnCOSA is a high-impact, low-volume program, it should be reserved only for high-risk offenders, which will likely help maximize its cost-effectiveness.

In response to high-profile sex crimes that have occurred during the last several decades, lawmakers have attempted to control sexual recidivism by implementing a variety of policies, many of which have been costly and punitive. In contrast, the COSA model represents an alternative approach to the prevention of sexual violence. Because it relies heavily on volunteers, program operational costs are relatively low. Moreover, the positive recidivism outcomes observed in this study indicate that the COSA model can be implemented effectively within the U.S. Further, due to the differences between programs in Canada and Minnesota, this study implies that COSA is a relatively flexible model that can withstand adaptation in another milieu. Still, the success the Canadian and Minnesota models have had in reducing recidivism is likely due to the fact that both have focused on delivering programming that increases offender accountability and social support.

ENDNOTES

1 Wilson, RJ, J.E. Picheca and M. Prinzo. 2005. Circles of support and accountability: An evaluation of the pilot project in southOntario. Ottawa, Canada: Correctional Service of Canada.

Wilson, RJ., F. Cortoni and AW. McWhinnie. 2009. Circles of support and accountability: A Canadian national replication of outcome findings. Sexual Abuse: A Journal of Research and Treatment, 21(4):412-430.

2 Ibid.

3 Duwe, G. and W. Donnay. 2008. The impact of Megan's Law on sex offender recidivism: The Minnesota experience. Criminology, 46(2):411-446.

4 Aos, S., M. Miller and E. Drake. 2006. Evidence-based public policy options to reduce future prison construction, criminal justice costs, and crime rates. Olympia, Wash.: Washington State Institute for Public Policy.

5 Duwe, G. and M. King. 2012. Can faith-based correctional programs work? An outcome evaluation of the InnerChange Freedom Initiative in Minnesota. International Journal of Offender Therapy and Comparative Criminology, 57(7):813-841.

6 Duwe, G. and P. Freske. 2012. Using logistic regression modeling to predict sex offense recidivism: The Minnesota Sex Offender Screening Tool-3 (MnSOST-3). Sexual Abuse: A Journal of Research and Treatment, 24(4):350-377.

7 Bates, A, R. Macrae, D. Williams and C. Webb. 2011. Ever-increasing circles: A descriptive study of Hampshire and Thames Valley circles of support and accountability 2002-2009. Journal of Sexual Aggression, 25(2):143-165.

Wilson, RJ., F. Cortoni and AW. McWhinnie. 2009. Circles of support and accountability: A Canadian national replication of outcome findings. Sexual Abuse: A Journal of Research and Treatment, 21(4):412-430.

Grant Duwe, Ph.D., is director of research and evaluation for the Minnesota Department of Corrections.