Abstract
We conducted a systematic review of studies reporting on the effectiveness of Circles of Support and Accountability (Circles). Circles use volunteers to provide support for sex offenders living in the community. We searched 10 databases up to the end of 2013 and identified 3 relevant outcome studies. An additional 12 papers or reports were identified by searching reference lists, Google, and contacting key authors and Circles providers to obtain unpublished data. These 15 studies comprised one randomized controlled trial, three retrospective cohorts with matched controls, and 11 case series. The majority reported measures of recidivism, particularly reconviction. The 4 studies with controls generally reported that participation in Circles was associated with lower recidivism although there were few statistically significant differences. Few studies examined changes in risk or psychosocial outcomes. A number of methodological issues are discussed. Longer term, prospective follow-up studies with control groups are required to address these issues.
Keywords
Introduction
This article reports the findings of a systematic review of research relating to the effectiveness of a social intervention—known as a Circle of Support and Accountability (CoSA, or “Circle”)—that aims to reduce the likelihood of reoffending among high-risk sex offenders who have been released from prison by providing support from a small number of community volunteers.
Interventions for Sex Offenders
The effectiveness of interventions for sex offenders in reducing recidivism is mixed; whereas some research points to positive outcomes (from certain types of interventions), other research suggests little effect. A meta-analysis by Hanson et al. (2002), comprising 43 studies reporting on institution-based treatments, community-based treatments, and treatment in both settings, found a lower rate for sexual offense recidivism for those offenders who received treatment, over a mean follow-up of 46 months, though benefits were modest with 12.3% recidivism in those who received interventions compared with 16.8% in those who did not. People who dropped out of treatment had higher sexual recidivism rates than those who completed, regardless of the type of treatment and even when the treatment was found to be ineffective. Post 1980, cognitive-behavioral-based treatments appeared more effective than earlier treatments. Conversely, Dennis et al. (2012) found no evidence in favor of psychological interventions for sex offenders in reducing sexual recidivism. Their systematic review of randomized controlled trials (RCTs) identified the need for more good quality studies.
Consequently, what is clear is that despite the provision of sex offender treatment programs (SOTPs), sexual reoffending remains a possibility, and with potentially very serious consequences. There is therefore a need to further improve the effectiveness of SOTPs or to develop interventions to further reduce the risk of sexual reoffending following completion of these programs both during and after incarceration. In addition, there is a group of offenders who, for various reasons, do not participate in sex offender programs but are nevertheless released back into the community when their sentence expires; these individuals are of particular concern as their risk remains unmodified (Criminal Justice Joint Inspection, 2012).
Irrespective of whether or not sex offenders have participated in treatment interventions, the reintegration of sex offenders is an emotive subject and successful reintegration into the community can be difficult. However, successful reintegration is integral to reducing the likelihood of reoffending. For example, isolation and loneliness are potential risk factors in sexual offending (Edwards & Hensley, 2001; Marshall, 2010). Thus, communities that fail to reintegrate offenders may, counterintuitively, also fail to quell important risk factors. Circles are an important development in terms of such community reintegration.
The Circles Model
Circles are an innovative community approach to the reintegration of sex offenders following release from prison. Circles began in Canada in 1994 (Hannem & Petrunik, 2007) as a one-off response to a particularly high-risk sex offender who was about to be released from prison with no supervision or support in the community. The history of this one case—and the start of the movement more generally—is well-documented elsewhere (e.g., Hannem & Petrunik, 2007; Hanvey, Philpot, & Wilson, 2011; R. Wilson, McWhinnie, & Wilson, 2008; R. J. Wilson & Prinzo, 2001a). In summary, the first Circle was formed by members of a Mennonite congregation who agreed to form a support group around the offender. The group had the dual role of providing him with support through the very hostile response from the community and monitoring his behavior to ensure that the community was kept safe. The group was effective at helping the man to remain offense-free and its model was swiftly replicated. What began as a one-off example of community action has spread into an international movement (Hanvey & Höing, 2012); Circles are now provided in Canada, the United States, the United Kingdom, the Netherlands, Belgium, Catalonia, Bulgaria, and Latvia, with various other countries (e.g., France, Hungary, Ireland) registering keen interest and some taking steps toward implementation.
How Circles Work
The way that such Circles operate has been described elsewhere (e.g., Hannem, 2013; Höing, Bogaerts, & Vogelvang, 2013; C. Wilson, Bates, & Völlm, 2010), but the essence is that they provide both support and accountability. The sex offender, or core member (CM) as he or she is known, is given support by a group of four to six volunteers (who are in turn supported by a Circle coordinator). Volunteers are from within the community, with a range of ages and backgrounds, and meet with the CM on a weekly and sometimes daily basis. Help is provided to carry out basic aspects of community life such as looking for work or accommodation and adjusting to life outside of prison. In addition, the volunteers hold the CMs to account for their own reintegration (and in terms of managing their own risk, for example, sticking to their desistance plan and avoiding triggers to offend). The relationship between the CMs and volunteers is one of openness and honesty, which is essential to the success of the Circle; being able to ask (and respond to) potentially challenging questions—including relating to offense-supportive beliefs—is pivotal. The volunteers also provide a monitoring function, and Circle coordinators work closely with statutory agencies (in the context of the United Kingdom, this means working with the Multi-Agency Public Protection Arrangements (MAPPA), alerting the relevant agencies of any potential risk posed by the CM. Having volunteers with such close insight into the life of the CM is invaluable in helping statutory agencies to act swiftly and prevent further reoffending.
Hanvey (2011) argued that the Circles model is most closely associated with the therapeutic approach advocated in the strengths-based good lives model (GLM; Ward & Stewart, 2003); thus, the work that is carried out by Circles may in some respects parallel and build upon treatments that offenders may receive while incarcerated. The GLM approach focuses on improving how the offender functions as a person by enhancing his or her capabilities to attain goals, or primary human goods, through socially acceptable means (Ward & Gannon, 2006). Through the provision of practical help to offenders (e.g., through finding work, appropriate friendships, or housing), Circles might assist in achieving the goods required for the offenders to lead a “good life.” In addition, Circles may reduce risk factors associated with antisocial and offending behavior, namely, social exclusion and antisocial attitudes. Social exclusion may reduce prosocial behavior in general (Baumeister, De Wall, Ciarocco, & Twenge, 2005; Twenge, Baumeister, De Wall, Ciarocco, & Bartels, 2007) while antisocial orientation and deviant sexual interests are key factors in sexual reoffending (Hanson & Morton-Bourgon, 2004, 2005). Circles target some of these risk factors by aiming to improve insight, problem solving, social skills, coping and self-regulation skills, social integration, and participation in society (Höing et al., 2013).
The majority of Circles research is related to the effectiveness of Circles, particularly whether Circles are effective at helping CMs to remain offense-free. However, Circles are provided on a relatively small scale due to the relatively low number of sex offenders being released at any one time, the need for willing volunteers, and the novelty of the Circles approach. Therefore, large-scale effectiveness studies are rare.
The main aims of this article are twofold: to provide a systematic review of studies relating to Circle effectiveness to date and to use the findings of that review, including reflection upon the strengths and limitations of previous studies, to inform the design of future effectiveness studies. It is by understanding the limitations of previous research that the most effective study designs can be created and the most fruitful future avenues identified.
Method
We identified studies and reports investigating the effectiveness of Circles on relevant outcomes in sex offenders. The review question was defined in terms of population, interventions, comparators, outcomes, and study design (PICOS) as recommended by the Centre for Reviews and Dissemination (2008).
Participants
Adult (18 years+) sex offenders. We expected the majority would be males although we did not exclude females.
Interventions
A Circle as described above. The Circle will be operational while the CM is in the community although the Circle may have been initiated prior to the CM’s release from prison or other custody.
Comparators
Sex offenders not receiving a Circle.
Outcomes
Our primary focus was on studies reporting recidivism outcomes: reconviction for any offense and any sexual offense, reoffending, arrest, recall, and breach of license. Given that Circles provide support for other needs, and the potential impact of those needs on recidivism, other outcomes such as psychosocial adaptation, housing, relationships, and employment were also recorded. Accordingly, the search strategy did not define specific outcomes.
Study Design
RCTs are the preferred way of assessing the effectiveness of interventions. However, given the ethical considerations of RCTs in forensic populations, matched control studies, risk norm, and cohort studies have been included in similar reviews (Elliott & Beech, 2011, 2013) and so are included here. Our inclusion criteria for study design therefore were RCTs, matched control studies, cross-sectional studies, prospective studies, retrospective studies, risk norm studies, case series, and cohort studies. We excluded reviews, expert opinions, editorials, single case reports, and any other nonempirical papers.
Databases were searched up to the end of 2013. We searched Medline, Embase, PsycInfo, Applied Social Sciences Index and Abstracts (ASSIA), International Bibliography of the Social Sciences (IBSS), Sociological Abstracts, Allied and Complementary Medicine Database (AMED), Cumulative Index to Nursing and Allied Health Literature (CINAHL), Web of Science, and LexisLibrary. Reference lists of identified papers were checked. In addition, a Google search was conducted for any gray literature. Key authors and Circles providers in the United Kingdom and Netherlands were contacted to obtain unpublished data.
Search terms were (“circle$ of support and accountability” OR “circle$ AND support AND accountability” OR “circle$ of support” OR “CoSA”) AND (sex$ offend$). This syntax was adapted to match the capabilities of the databases.
Data analysis
Risk ratios were calculated for studies reporting reconviction rates for CMs and control groups.
Results
Studies Identified
The search of the databases identified 68 papers of which 29 were unique. Of these 29, the majority described the concept or the development of Circles while only 3 were relevant outcome papers (Bates, Macrae, Williams, & Webb, 2012; Duwe, 2012; R. J. Wilson, Cortoni, & McWhinnie, 2009). A further 5 papers (Bates, Saunders, & Wilson, 2007; Bates & Wager, 2012; Bates, Williams, Wilson, & Wilson, 2014; R. J. Wilson, Picheca, & Prinzo, 2007; R. J. Wilson & Prinzo, 2001a) and 4 reports (Fox, 2013; Haslewood-Pócsik, Smith, & Spencer, 2008; McCartan et al., 2014; Quaker Peace and Social Witness, 2005) that met the inclusion criteria were identified from Google searches. One unpublished report (Earnshaw, 2014) and two unpublished studies (Clarke, Warwick, & Völlm, n.d.; Höing, Vogelvang, & Bogaerts, n.d.) were also identified. These 15 studies are summarized in Table 1.
Empirical Studies of Circles of Support and Accountability.
Note. CoSA = Circles of Support and Accountability; CI = confidence interval; RRASOR = Rapid Risk Assessment for Sex Offense Recidivism; IMPACT = Innovation Means Prisons and Communities Together.
Identified in literature database search.
This information is also available in R. J. Wilson, Picheca, and Prinzo (2005) and R. J. Wilson and Picheca (2005).
This information is also available in R. J. Wilson, Picheca, McWhinnie, and Cortoni (2011).
This information is also available in R. J. Wilson and Prinzo (2001b).
This information is also available in Circles South East (2012).
The Samples
Location and study type
The studies were from the United Kingdom (predominantly England; n = 9), Canada (n = 3), the United States (n = 2), and the Netherlands (n = 1). In the only RCT, Duwe (2012) followed up 31 moderate-risk sex offenders receiving a Circle in Minnesota, United States and 31 controls. There were also three retrospective cohort studies with matched controls and 11 case series.
Inclusion of CMs from previous studies and updated outcomes
Several of the papers updated previous studies by reporting increased sample sizes as more Circles developed within the project or the project area expanded. R. J. Wilson and Prinzo (2001a) briefly described the outcome of 30 CMs in South Central Ontario after which R. J. Wilson et al. (2007) expanded the study by adding a further 30 CMs and then compared these 60 CMs with 60 controls.
In the United Kingdom, Bates et al. (2007) described the outcome of 16 CMs in Hampshire and Thames Valley (HTV) Circles before Bates et al. (2012) increased this sample and described the outcome of 60 CMs. Bates et al. (2014) then described 71 of the first 100 Circles South East (CSE) Circles—renamed after HTV Circles expanded to include the county of Kent in 2010. Bates et al. (2014) excluded 19 CMs whose Circle had been operational for less than 6 months and a further 10 CMs under their “90-day rule” whereby CMs who spent less than 90 days in a Circle were deemed not to have had the opportunity to benefit from being in a Circle. This was the only U.K. study with a control group—a sample of sex offenders not receiving a Circle. The controls were sex offenders who had been referred to a Circle, considered to be suitable, but did not receive a Circle, and were broadly matched on risk status and period of community follow-up (Bates et al., 2014). The majority of these U.K. data were subsumed when Clarke et al. (n.d.) described the characteristics of the first 275 men who participated in a Circle under the auspices of Circles UK. Where available, Clarke et al. also reported follow-up data for the Circles that had ended; follow-up data were available for 185 CMs although there were missing data for many of the outcomes
In total, these 15 studies comprised approximately 446 Circles. This figure excludes the smaller samples where CMs were counted in larger samples.
Recidivism
Defining recidivism
Most of the studies reported recidivism, which comprised definitions that either grouped outcomes or reported specific types of outcomes (e.g., reconviction). Bates et al. (2007) and Quaker Peace and Social Witness (2005) reported instances of recidivist behaviors that included conduct indicative of previous offense patterns as well as reoffending and reconviction. Others defined recidivism as being charged for or convicted of a new offense (R. J. Wilson et al., 2009; R. J. Wilson & Prinzo, 2001a) and sexual recidivism as either being charged with a new sexual offense or breaching a court-imposed condition (R. J. Wilson et al., 2007). However, the majority of studies reported specific outcomes: reconviction (7), recall (5), breach of an order (4), rearrest (3), new charges (3), revocation (2), new Sexual Offenses Prevention Order (SOPO; 2), failure to comply with a Sex Offender’s Register (SOR; 1), resentence to prison (1) and any reincarceration (1).
Impact upon recidivism
Overall, there was some evidence that CMs had lower rates of general recidivism than controls. The few cases of sexual recidivism were reported as being less severe than the index offense. However, there were few significant differences between CMs and controls. These findings will now be discussed in more detail.
Duwe (2012) found that CMs had lower rates of recidivism on all five measures compared with the randomized controls but only rearrest was significantly lower using chi-square. However, this was for any offense rather than specifically sexual offenses. A reduction in sexual offenses was not demonstrated as only one of the 31 controls and none of the 31 CMs were rearrested for a sexual offense. Duwe (2012) also found that, despite the small sample size, Cox regression analyses showed significant reductions in time to recidivism for three of the five measures of recidivism used in this study (any arrests, technical violation revocations, and any reincarceration) for those who received a Circle compared with the controls.
The other study from the United States (Fox, 2013) was unusual in that Circles were provided for nonsex offenders (n = 9) as well as sex offenders (n = 12). There were no instances of sexual recidivism in either of the groups. Three of the sex offenders (25.0%) and three of the nonsex offenders (33.3%) were revoked for violating the conditions of their release, such as lack of residence or not complying with the conditions of their release (K. Fox, personal communication, February 20, 2014).
In Canada, R. J. Wilson et al. (2007) examined 60 CMs and 60 controls and found that CMs had a significantly lower rate of sexual recidivism (5.0%) compared with controls (16.7%). Incidentally, the three instances of sexual recidivism by CMs were the same three instances recorded in R. J. Wilson and Prinzo’s (2001a) smaller study of 30 CMs where 10.0% had recidivated. R. J. Wilson et al. (2007) described the three new instances of sexual recidivism by CMs as “qualitatively less severe or invasive than the offense for which they had most recently served sentence” (p. 332). Such a reduction in harm was not evident in the 10 new instances of sexual recidivism by the controls. However, there were no statistical differences between CMs and controls in the time to first failure or in the overall recidivism rates (i.e., sexual and nonsexual) although the latter approached significance and the authors argued it should be seen as socially significant.
R. J. Wilson et al. (2009) examined a different cohort of sex offenders (44 CMs and 44 controls) to replicate the R. J. Wilson et al. (2007) study albeit with a smaller sample and shorter follow-up. They found that CMs had a significantly lower rate of any recidivism (11.4%) compared with controls (38.6%). They also reported that CMs had a significantly lower rate of sexual recidivism (2.3%) compared with controls (13.7%). However, Elliott and Beech (2013) noted that the assumptions of the chi-square test had not been met and reanalyzed these data using Fisher’s Exact Test, which showed that the difference was not significant (albeit marginally). R. J. Wilson et al. (2009) then examined a subsample of 19 CMs and 18 controls with a 3-year follow-up period and found significant differences between CMs and controls for sexual recidivism (0%; 27.8%), violent recidivism (including sexual recidivism: 10.5%; 61.1%), and any recidivism (including sexual and violent recidivism: 10.5%; 66.7%), respectively. While there were no significant differences in the STATIC-99 (Hanson & Thornton, 1999) risk scores between the subsample of CMs and controls, there were significant differences in the amount of new charges/convictions: The two CMs and 12 controls who had recidivated had incurred a total of 5 and 45 charges/convictions, respectively.
In England, while Bates et al. (2014) identified that 54 of the 71 CMs (76.1%) had not faced any legal sanctions during the follow-up period, comparable figures were not available for the 71 controls because data regarding recall to prison were not available. Three CMs (4.2%) and two controls (2.8%) were reconvicted of a noncontact sex offense. Three controls (4.2%) and one CM (1.4%) were convicted of a contact sexual offense; however, the CM’s conviction was for a historical offense and therefore not classed as a reconviction. While there were fewer sexual reconvictions than predicted using the Risk Matrix 2000 for CMs and for Controls after a 5-year follow-up, these differences were not significant. Other comparisons also showed no significant differences between CMs and controls, including breaching a SOPO or failing to comply with the SOR (these two outcomes do not create more victims and are not classed as sex offenses). Given that Bates et al. (2014) found that only three CMs were reconvicted of a sexual offense, it is not surprising that the smaller samples from the same Circles project, reported in earlier publications, found that no CMs were reconvicted of sexual offenses (Bates et al., 2007; Quaker Peace and Social Witness, 2005).
Risk
Perhaps surprisingly, little attention has been given to changes in risk as the Circle progressed. Only three studies examined changes in risk (Bates & Wager, 2012; Earnshaw, 2014; Höing et al., n.d.). Each used the Dynamic Risk Review (DRR) which was based on the Structured Assessment of Risk and Need (SARN; HM Prison Service, 2005) and was developed by Circles UK to provide a structured means of evaluating a CM’s progress by assessing dynamic risk factors. The DRR is administered approximately every 3 months by the Circle coordinator in collaboration with the Circle volunteers. Bates and Wager (2012) identified five factors on the original DRR (17-items scored 0-6 on a 7-point scale). After excluding two factors, one for poor internal reliability and one for comprising only one item, Bates and Wager reported changes in DRR scores for the 13 CMs who had completed three DRRs. There were significant changes in two of the three factors: Inappropriate Sexual Attitudes (Factor 2) and Overconfident Hostile Sexualization (Factor 3). Overall, positive reductions in scores for Factor 2 and Factor 3 were noted for 61.5% and 77% of CMs, respectively. In contrast, scores for two CMs (15%) indicated their risk had increased. However, the sample size was small, it is not known what constitutes a meaningful reduction, and changes cannot be attributed to the Circle.
Further information on changes in DRR scores was available in a paper prepared for a Circles UK Trustees Meeting (S. Hanvey, personal communication, June 27, 2014) at which Earnshaw (2014) reported a larger sample of 52 CMs with at least three completed DRRs. There was a mean reduction in total score of 11 points from their first DRR to their last. However, Earnshaw noted considerable variation; one CM’s DRR score had reduced by 41 points, whereas for some other CMs, DRR scores had increased by 20 points (although the actual scores were not reported). Furthermore, Earnshaw noted that the scores for the “sexual interest” and “offence related attitude” items were often very low at all time points, which may indicate that the CMs or other Circle members felt uncomfortable broaching these important areas. Earnshaw also reported that more than 60% of scores indicated that engagement with appropriate hobbies and interests was unproblematic but that more than 70% of DRRs indicated that the CMs were not in employment. The DRR has also been used in the Netherlands where Höing et al. (n.d.) found a nonsignificant decrease in risk in 13 CMs between 6 months and 12 months after the beginning of the Circle.
Psychosocial Outcomes
While good quality reconviction data are important, measures of recidivism do not capture the extent of the impact of participating in Circles. Psychosocial adaptation, housing, relationships, and employment are also important in assessing the efficacy of Circles, not least as they have been linked with recidivism. However, while examples of these outcomes are discussed in qualitative studies (not reviewed here), few quantitative data were available.
Volunteers supported and assisted CMs in a variety of ways such as accessing employment (Bates et al., 2012; Bates et al., 2007) and accommodation (Bates et al., 2007). Furthermore, there were improvements in prosocial attitudes (Bates et al., 2012; McCartan et al., 2014), participation in prosocial activities (McCartan et al., 2014), emotional well-being (Bates et al., 2012), and self-esteem (Höing et al., n.d.). While tangible improvements were documented, none of these studies used control groups.
Positive changes in several psychosocial factors were reported when Bates et al. (2012) examined CMs’ files to identify the dynamic risk factors categorized in the Offender Assessment System (OASys) risk assessment tool (Home Office, 2002). Bates et al. (2012) asserted that, in some cases, their Circle was instrumental in assisting with these improvements: 42 CMs (70%) showed improvement in emotional well-being, 37 CMs (62%) had displayed attitudes and behaviors that were prosocial, 30 CMs (50%) had increased their engagement in age-appropriate relationships, 29 CMs (48%) had improved links with their families and increased their support networks, 29 CMs (48%) were encouraged and supported by the volunteers to access employment and education, 17 CMs (28%) experienced financial difficulties with which the volunteers were able to advise and assist them, 17 CMs (28%) were given assistance in accessing housing support, and 13 CMs (22%) were encouraged to seek medical help and, in some cases, reduction of stress and depression could be linked to the actions of the volunteers. Specific details of the support given were provided in three case studies. However, the effectiveness of the support provided, for example, in improving financial circumstances, was not reported.
Similarly, Bates et al. (2007) reported the role that Circles played in the resettlement progress of 16 CMs, such as assisting with a move to independent accommodation and in the process of disclosure of offenses to prospective employers. Again, longer term outcomes were not reported.
Some significant differences for CMs between the beginning and the end of a Circle have been found (Clarke et al.,n.d.). For example, more CMs were in a relationship at the end of the Circle. However, data on the type or quality of the relationship were not available as these were dichotomous classifications of “yes” or “no.” Also, fewer people were unemployed at the end of the Circle—two thirds compared with three quarters.
McCartan et al. (2014) reviewed case files of 32 CMs from two pilot sites and found that two thirds received support to engage in prosocial activities, including safe leisure activities, use of a gym (with a volunteer in attendance), social events, volunteering, support groups, education courses, and employment. McCartan et al. (2014) also found that two thirds were recorded as self-reporting some positive changes in attitudes and motivations.
Höing et al. (n.d.) found self-esteem had significantly increased at 12 months into the Circle. Three of the six factors identified on a measure of volitional skills also showed significant improvements: self-soothing, emotion regulation, and internal locus of control. Other measures such as participation in society (which included dichotomous measures of housing, stability of housing, employment, having leisure time activities, and being a member of an association) and size of social network had not significantly improved at 12 months.
Cost-Benefit Analysis
Further to the specific outcomes of CMs identified in this review, others have modeled the cost-benefit of providing Circles. Overall, these studies indicate that costs are outweighed by benefits, though the extent to which this is the case varies according to the type of calculation undertaken.
In addition to the RCT of Circles in Minnesota (MnCoSA), Duwe (2012) modeled the cost-benefit of Circles within the first 4 years by comparing the cost of operating Circles with the costs associated with rearrest (chosen instead of reconviction because of the short follow-up period) and reincarceration. The costings mainly relate to nonsexual offenses because only one of the 85 rearrests was for a sexual offense. Nevertheless, Duwe (2012) concluded “The cost-benefit ratio indicates that for every dollar spent on MnCoSA, the State of Minnesota has seen an estimated benefit of US$1.82, which results in an 82% return on investment” (p. 160).
The cost-benefit analyses by Elliott and Beech (2011) provide further evidence of the cost savings of Circles in the United Kingdom. The authors modeled costs based on a hypothetical 50 sex offenders having a Circle and 50 not having a Circle. They calculated a small cost-benefit of participating in Circles (£0.18 in savings for every £1 invested in Circles) although the cost savings substantially increased when they modeled intangible costs to the society (Elliott & Beech, 2011). Elliott and Beech (2013) used slightly different coefficients to model the cost-benefits of participating in Circles and also concluded there was a small cost-benefit of participating in Circles ( £0.04 in savings for every £1 invested in Circles) and larger cost savings when intangible costs to the society were considered. However, Elliott and Beech (2013) acknowledged that the different ways in which Circles are provided make it difficult to estimate costs.
Discussion
This article reports on a systematic review of quantitative studies that have evaluated the effectiveness of Circles. First, we discuss the findings from our review of the literature. Second, we discuss the methodological issues and limitations of the research included in this review. Third, we discuss the programmatic issues associated with Circles. Fourth, we describe the limitations of our review. Finally, we outline the directions for future research.
The Effectiveness of Circles
There were few statistically significant differences between CMs and controls in the outcomes reported. However, where there were significant differences, CMs fared better than controls (e.g., R. J. Wilson et al., 2009).
Reconviction
The studies were not able to demonstrate a reduction in reconvictions for sexual offenses. This is not surprising given the methodological limitations outlined below. Some studies provided evidence that fewer CMs than controls were reconvicted for any type of offense.
Recidivism
The broader outcome of recidivism was the main outcome reported with less attention given to other outcomes. This may be because of the ethos of “no more victims” and therefore needing to quantify whether or not CMs have created more victims. It may be that funding and support from respective Governments also require a focus on recidivism. If the effectiveness of Circles is measured solely by a reduction in sexual recidivism for CMs, again one would have to conclude that data are not yet available to support their effectiveness. This does not mean that Circles are not effective—rather, we need studies with a sufficiently long follow-up. Given that reductions in recidivism mean fewer victims, it has been argued that differences that approach statistical significance could be considered socially significant (R. J. Wilson et al., 2007).
Importantly, there is no evidence that participating in a Circle has an adverse effect on outcomes. Likewise, the cost-benefits of participating in Circles, financially and to the wider community, add support for Circles (Duwe, 2012; Elliott & Beech, 2011, 2013). This is particularly important given the reduction in funding for statutory and voluntary organizations (Circles South East, 2012).
Psychosocial outcomes
Clearly, providing support and opportunities to be active members of the community is important for the rehabilitation and reintegration of sex offenders (Brown, Spencer, & Deakin, 2007). Therefore, more attention should be given to recording other outcomes to assess how CMs are functioning, integrating, and coping in the community. There were specific examples of where Circles have provided support and had an impact on the CM (e.g., Bates et al., 2012; McCartan et al., 2014). Clarke et al. (n.d.) noted that some changes between the beginning and end of a Circle could not necessarily be attributed to the CM participating in the Circle. Indeed some changes would likely have happened over time regardless of participating in a Circle. However, due to the methodology of most of the studies, comparable outcomes for non-CMs were not available.
Being able to resume normal lives after punishment for an offense is an important part of desistance theories (Willis, Levenson, & Ward, 2010). By helping to reduce social isolation, Circles support desistance from sexual offending. This facilitation of community inclusion, together with nonjudgmental social support, is consistent with the GLM (Fox, 2017). However, the theoretical framework on how Circles work requires further development. This is particularly important given that desistance is a process that may include relapses (Willis et al., 2010). Nonetheless, what sets this aside from many other interventions is that the support is given by people from the local community who are volunteering to help the individual reintegrate and desist from sexual offending.
Methodological Issues—Limitations of the Research Included
We will now discuss the limitations of the study design, small sample size, short follow-up, data sources, and Circles dosage.
Study design
There was only one RCT. Duwe (2012) found that the demand to participate in a Circle exceeded the availability of a Circle. Thus, overcoming ethical concerns surrounding withholding an intervention in an RCT—albeit an intervention that at the time had even less of an evidence base. The majority of outcome studies were case series. These are easier to conduct and considerably less expensive than RCTs but they do not have control groups with which to compare the outcomes of people participating in Circles.
Two of the three matched control studies matched the two groups a priori based on risk (all detained until their sentence was completed), general criminality, release date, and previous sexual offender treatment (R. J. Wilson et al., 2009; R. J. Wilson et al., 2007). However, both studies noted that the matching was less than exact given that that there were differences on one of the two actuarial risk assessments for sexual reoffending. One where the Circles group were higher risk (R. J. Wilson et al., 2007) and one where the controls were higher risk (R. J. Wilson et al., 2009). The third matched control study reported their groups were broadly matched on risk status and time in the community (Bates et al., 2014).
There are likely to be many factors associated with the effectiveness of interventions such as the offending history, the needs of the offender, risk level, statutory supervision, and participation in other sexual offender treatment programs. For example, Circles complement sexual offender treatment programs (Bates et al., 2014) and CMs may have completed such programs or may have participated in one alongside their Circle (Clarke et al.,). Likewise, members of control groups may have participated in sexual offender treatment programs. The aim of such treatment is, of course, to reduce the risk of reoffending. A meta-analysis of other types of sexual offender treatment studies comprising control groups, albeit mainly nonrandomized, found that the sexual recidivism rate was 37% lower overall for the treated sex offenders compared with the controls (Schmucker & Lösel, 2008). This again supports the need for randomization (Duwe, 2012) or including previous treatment as one of the matching criteria (R. J. Wilson et al., 2009; R. J. Wilson et al., 2007). Furthermore, outcomes were not reported in a way that distinguished between whether or not CMs had completed, or were completing, other SOTPs in addition to participating in a Circle. More data are required to establish whether a combination of Circles and other interventions are more effective at reducing reconviction or whether Circles are sufficient on the own.
Non-RCTs have to consider how to assess noncompleters. There will be many reasons why CMs do not complete their Circle but the effectiveness of Circles can only really be measured if noncompleters are included. Given Circles were set up to offer support to those sex offenders who are expected to fail in the community very quickly, it is antithetical to the ethos of Circles to exclude CMs who drop out quickly when assessing their effectiveness. Elliott (2014) criticized Bates et al. (2014) for including people who withdrew from the process after being assessed as suitable for a Circle in the control group, rather than in the treatment group. Elliott rightly argued that withdrawal may indicate a lack of motivation to stop offending and it could therefore potentially overinflate differences between the two groups if noncompleters were included as part of the control group. The one RCT (Duwe, 2012) randomized offenders who were willing to participate. Nonetheless, an RCT would address the issue of CMs’ motivation outlined by Elliott (2014) if data were analyzed by “intention to treat” whereby noncompleters would be included in the treatment group. Hanson et al. (2002) asserted that it is important to promote equivalence between treatment and comparison groups and that this matters more than sample size.
Limitation of small sample size
An additional methodological limitation in the literature reviewed was the small sample sizes in most studies resulting in a failure to detect statistically significant differences. Despite detecting some significant differences, Duwe (2012) acknowledged that the reconviction outcome measure was not sufficiently powered to detect a meaningful difference (or effect size) between CMs and controls. This can lead to a Type II error whereby the null hypotheses—that no difference exists between the groups—is not rejected when it is false. Power, the probability of avoiding a Type II error, is conventionally set at 80%. Achieving this level of power depends on the sample size, the effect size, and the probability level at which the effect is considered statistically significant (conventionally > 95%). Studies need to have a sufficient number of participants to detect a meaningful effect but not too many as to detect a trivial effect. Furthermore, the inclusion of too many participants could raise ethical concerns if participants were denied an effective intervention or given an intervention which, with fewer participants, could have been shown to be ineffective.
Short follow-up
Short follow-up periods are problematic given that a key indicator in assessing the efficacy of Circles is being able to show a reduction in sexual offending. However, low rates of recidivism, particularly for sex offenders, present a challenge and therefore long follow-up periods, for example, 5 to 10 years, are required for studies of sex offenders (Hanson & Bussière, 1998). Duwe (2015) pointed out that low rates of sexual recidivism leave little room for Circles and other interventions to substantially reduce sexual recidivism. Further to the low rates, Cann, Falshaw, and Friendship (2004) demonstrated a slow rate of reconviction when they followed up sex offenders released from prisons in England and Wales in 1979. While only 10% had been reconvicted of a sexual offense within the first 2 years following discharge, this increased to one quarter after 21 years (Cann et al., 2004). This serves to underline the need for longitudinal studies, in particular those that include control or match-control participants.
Limitation of data sources
Reconviction rates for sexual offenses can vary depending on the data source (Falshaw, Bates, Patel, Corbett, & Friendship, 2003). However, the majority of studies did not explicitly report which data sources were used to acquire outcomes such as reconviction. In most studies, it is likely that CM case files were the only source used but such files were not compiled for the purpose of research. While CM case files may contain information from official sources (e.g., the Police National Computer in England and Wales), provided by other agencies working with the Circle, these data may not be up-to-date. For future research, careful consideration should be given to which data sources are used to acquire outcome data and these sources—and the period covered—should be stated. A judiciously designed prospective study should allow comparable data to be available for CMs and controls. This is often not the case with retrospective studies as Bates et al. (2014) found when data regarding recall to prison were not available for the controls.
Circles dosage
While some studies reported data for the length of time in a Circle, specific information relating to the dosage of Circles was not reported. Dosage would also include the frequency of meetings, whether they were group or individual meetings, how many volunteers were in the Circle, and the duration of meeting. Again, this information may not have been recorded, particularly in the earliest Circles, or readily available to the researchers. However, this seems particularly important to collect as part of future research as Circles are not a manualized treatment with a prescribed schedule—rather they are unique to the individual CM and his or her circumstances. Clearly, it is difficult to assess the efficacy of an intervention if no one “completes” a “sufficient” dosage. However, what this is remains unclear.
Programmatic Issues
There are programmatic factors that need to be acknowledged as these issues will impact on outcome studies.
Increased detection of recidivism
While many CMs will be subject to statutory supervision—unlike the original Circles—being in a Circle provides an extra layer of surveillance; this increases the opportunity to detect recidivist behavior. This additional layer of surveillance therefore may lead to what appears to be an increase rather than a decrease of recidivism (when compared with controls). This was demonstrated when Quaker Peace and Social Witness (2005) found seven of the eight cases of recidivist behavior were identified by the Circle. Holding CMs to account could be recognized as a positive outcome for Circles. Furthermore, outcomes such as being recalled to prison should be acknowledged as positive rather than negative if the Circle contributed to the outcome, thereby giving the opportunity to address problematic antecedents that otherwise could potentially lead to more serious offending.
Variation between circles
Circles originated in the community and evolved as a community response to sex offenders. However, while all Circles developed with the aim of making their local community safer, caution is required when comparing projects and generalizing outcomes because of the variability across Circles projects. For example, Duwe (2012) pointed out several notable differences between the original Canadian Circles and the Minnesota Circles he investigated: the impetus for the Circles (initiated by the community or the Department of Corrections), the development of Circles (organic or systematic), the recruitment of volunteers (from faith groups or the wider community), postrelease supervision (none or statutory), the commencement of Circles (community or in prison), and the location of meetings (individual’s homes or secure public venues). In the United Kingdom, Circles usually commence on release from prison although in some cases Circles meet with the sex offender in prison prior to being released. However, in one study, the four Circles were only put in place 14 to 30 months after release from prison (Haslewood-Pócsik et al., 2008). An examination of data from Circles UK showed variations in the demographics and background characteristics of CMs within and across the five projects that provided the most Circles (Clarke et al.,). Indeed, the structure of Circles Projects develops based on specific, local constraints (Fox, 2017). There is likely to be more variation in the future in the United Kingdom as projects look to provide support to groups such as young people or people with learning disabilities and develop additional support such as mentoring. Further variation in Circles provision is likely too as they are implemented in new countries (with differing social contexts and criminal justice systems) and this variability is important to consider for future research.
Limitations of the Review
We excluded qualitative studies that “only” reported the experience of Circles by either the CM or other stakeholders. However, consideration of experience would have been helpful given the difficulties in assessing the quality of the outcome and attributing any changes in the CMs’ circumstances to being in a Circle. For example, more information is required about the role of Circles in outcomes, such as detecting offending behavior or assisting to find suitable employment or accommodation. This will require the Circle having the knowledge that their actions affected the outcome and that these outcomes are systematically recorded.
Only one RCT and three retrospective non-randomized-controlled studies were identified and so it was not feasible to conduct a meta-analysis on the recidivism data. There were also insufficient data reported to assess other types of outcome.
Directions for Future Research
Well-designed quasi-experimental designs have their place in adding to the evidence base when evaluating offending behavior programs (Hollin, 2008). Nevertheless, RCTs are perceived as the gold standard methodology for assessing the efficacy of interventions and a larger scale RCT would increase the evidence base. Elliott, Zajac, and Meyer (2013) offered advice on conducting an RCT of Circles. Pursuing an RCT would require overcoming ethical concerns, convincing stakeholders, deciding on the primary outcome, and designing a methodology with a sufficient sample size and follow-up period alongside robust data collection. Longer term follow-up studies would demonstrate whether positive changes, such as gaining employment, had been sustained but would also address some of the practical concerns of researching sexual offending, namely, the slow and low and rates of reconviction.
Subgroups of CMs should also be examined as it is likely that different groups, such as younger or older CMs, will have different needs and outcomes. Furthermore, few women have participated in Circles and more data are required. It is possible that Circles could be adapted for mentally disordered sex offenders (MDSOs), for example, those leaving prison or forensic secure hospitals or being treated by community forensic teams. This would require exploring how Circles could best be adapted, such as providing additional training and support for the volunteers.
Different countries will have different regulations for participating in academic research but CMs should be encouraged to participate in research, either face-to-face or allowing access to records, to evaluate the effectiveness. Such evidence could help allay public fears and increase the opportunities to acquire funding for Circles.
Conclusion
Circles have helped sex offenders in adjusting to their release to the community. The dual components of support and accountability mean that Circles have a role over and above statutory supervision in that they address known risk factors for reoffending and provide an additional layer of surveillance. Data from Circles in the United Kingdom, Canada, and the United States broadly suggest positive outcomes although methodological issues have meant this has not yet been adequately demonstrated, particularly for sexual offenses. Longer term, prospective research would allow further evaluation and more attention should be given to evaluating the role Circles have in the outcomes.
Footnotes
Declaration of Conflicting Interests
The author(s) declared the following potential conflicts of interest with respect to the research, authorship, and/or publication of this article: Birgit Völlm is a former Trustee of Circles UK.
Funding
The author(s) received no financial support for the research, authorship, and/or publication of this article.