A U.K. Cost-Benefit Analysis of Circles of Support and Accountability Interventions

Abstract

Circles of Support and Accountability (CoSA) aim to augment sex offender risk management at the point of community reentry by facilitating “Circles” of volunteers who provide support, guidance, and advice, while ensuring that the offender remains accountable for their actions. In this study, the authors provide (a) a rapid evidence assessment of the effectiveness of CoSA in reducing reoffending, and (b) a U.K. cost-benefit analysis for CoSA when compared to the criminal justice costs of reoffending. From the study analysis, the average cost of a “Circle” was estimated to be £11,303 per annum and appears to produce a 50% reduction in reoffending (sexual and nonsexual), as the estimated cost of reoffending was estimated to be £147,161 per offender, per annum. Based on a hypothetical cohort of 100 offenders—50 of whom receive CoSA and 50 of whom do not—investment in CoSA appears to provide a cost saving of £23,494 and a benefit-cost ratio of 1.04. Accounting for estimates that the full extent of the cost to society may be 5 to 10 times the tangible costs substantially increases estimated cost savings related to CoSA.

Keywords

sexual offender management treatment effectiveness recidivism Circles of Support cost-benefit analysis

Introduction

The continuing incidence of sexual violence presents a significant challenge to law enforcement agencies, policy makers, medical services, and providers of therapeutic interventions for both victims and perpetrators. Sexual crime directed toward adults and children is rightly considered a gross violation of the person and has been linked to a great number of long-term negative consequences for victims, in terms of mental health, revictimization, and adult social and sexual functioning (e.g., Andersen, Tomada, Vincow, Valente, & Polcari, 2008; Easton, Coohey, O’Leary, Zhang, & Hua, 2011; Hillberg, Hamilton-Giachritsis, & Dixon, 2011; Itzin, Bailey, & Bentovim, 2008). As such, crimes of this nature carry heavy legal penalties, including substantial custodial and community sentences. Sexual offenses incite a great deal of fear among the general public, leading to variety of criminal justice policies designed to protect communities from further sexual victimization (Levenson, Brannon, & Fortney, 2007).

Consequently, releasing and reintegrating an individual convicted of sexual offenses into the community after prison is a sensitive issue that needs to be managed with careful consideration of the needs of the individual being released, the victims, and the community in general (Wilson, Cortoni, & McWhinnie, 2009). On release, the criminal justice system is tasked with balancing the safety of the communities into which the offender is released with the need to support the offender in developing an offense-free life. As Willis, Levenson, and Ward (2010) note, desistence for sex offenders on reentry to the community is “dependent on specific environmental conditions, for example, stable housing, access to employment opportunities, cognitive transformations, and social support” (p. 553).

One approach that seeks to provide these positive environments is that of Circles of Support and Accountability (CoSA). In 1994 a Mennonite pastor agreed to provide support to a recently released sex offender and asked a group of his congregation to volunteer a small amount of their time to assist him in establishing himself in the community. The approach was then subjected to a successful pilot project by Correctional Service of Canada and the Mennonite Central Committee of Ontario (MCCO), who were able to operationalize the approach and its protocols (Correctional Service of Canada, 2003). In light of the success of this pilot, the Circles approach has since been adopted in a number of other countries, including the United States, England and Wales, Scotland, the Netherlands, and Belgium.

According to Wilson, Bates, and Vollm (2010), a U.K. “Circle” consists of a group of four to six trained volunteers from the local community and an experienced Circle coordinator who provide a social network around the offender, referred to as the core member (CM). The volunteers provide support and practical guidance (e.g., in obtaining employment or housing), help the CM to recognize patterns of potentially risky thoughts/behaviors, and ensure that the CM is accountable and takes responsibility for his or her own ongoing risk management. The CM is involved in all planning and decision making and, along with the volunteers, signs a contract committing to the Circle and its aims. The Circle then meets regularly, and volunteers spend individual time with the CM either face to face or via telephone, for the duration of the Circle (typically 1 year or more). The aim of the Circle is to assist the CM in reintegrating responsibly into the community, acting as a support and safety mechanism for both the CM and the community, and enhancing public safety when there is a perceived element of risk.

In order to establish the benefit of CoSA in terms of its use as a supplementary risk management strategy for sex offenders, it appears timely to assess and collate any previous attempts to evaluate the effectiveness of the approach, as there are currently no systematic reviews of specific CoSA approaches to sex offender management. Funding organizations are not only concerned that supported programs achieve their aims and objectives but also that their impact can be justified in terms of money spent (Friendship, Falshaw, & Beech, 2003). To this end, this study was commissioned by Circles U.K. and conducted by the authors at the University of Birmingham.¹ In this study we aimed to provide the following: (a) a rapid evidence assessment outlining the research base for the effectiveness of the CoSA approach in reducing reoffending, and (b) a cost-benefit analysis of the potential cost savings related to any reduction in reoffending in comparison to the costs to the criminal justice system of reoffending by sex offenders.

Establishing the Effectiveness of CoSA

Design: Rapid Evidence Assessment

Policy makers often require the synthesis of knowledge in an area within a shorter time frame than that required to produce a systematic review, which can typically take up to 12 months and utilizes a team of researchers. Rapid reviews, such as the Rapid Evidence Assessment, “streamline traditional systematic review methods” (Ganann, Ciliska, & Thomas, 2010, p. 58) by limiting searches (by year, databases, language, etc.) or employing fewer researchers for data extraction. Consequently, though the process is expedited, it may lead to missing data and introduce biases (Ganann et al., 2010); this is, however, balanced against the time and resources available. This project was completed within a time frame of 12 weeks.

Selection of Studies

Computer searches of Digital Dissertations, National Criminal Justice Reference System (NCJRS), PsychINFO, PubMed, Scholar Portal, and Web of Knowledge were conducted using the following key terms: sex* offen* and circle*. Additional searches were made of reference lists of collected articles, review articles, and contacts with researchers with links to CoSA organizations in Europe and North America. The search was limited to studies published after 1995 (the year of the first CoSA). One researcher was employed to review abstracts, assess methodological quality, and extract data.

PICO (Participants, Interventions, Comparators, Outcomes) Elements

Reviews in health care interventions typically report their method in terms of their PICO elements (participants, interventions, comparators, outcomes; Centre for Reviews and Dissemination, 2008), as follows:

Participants

To be included in the REA the studies needed to include an identifiable sample of adult (18+) male CoSA CMs.

Interventions

CoSA CMs were defined as individuals who received a community-based intervention by a CoSA provider, where trained volunteers provide a supportive community for the individual offenders, holding that person accountable for their actions.

Comparators

Randomized control studies are typically preferred in systematic review; however, there are ethical challenges associated with random assignment for criminal offenders. RCTs were, therefore, included in the criteria, but relevant matched control studies using non-CoSA sex offenders, risk norm, and cohort studies were also considered eligible.

Outcomes

Selected studies reported reoffending/reconviction outcomes for the samples. Reconviction is defined as the conviction for another offense during a specific follow-up period derived from official records of criminal convictions (Friendship et al., 2003). Other outcome data (e.g., employment, educational, health, housing, etc.) were also collected and reported.

Although this is a U.K.-based study, based on U.K. criminal justice costs, it was decided that this study would consider any quality studies into the effectiveness of CoSA regardless of the location of their sample. This decision was made due to a lack of quality studies into the effectiveness of CoSA in the United Kingdom specifically. Consequently, although it meant that we may not be able to generalize the findings to a U.K.-based population, any well-conducted cross-cultural comparisons (e.g., other European or North American samples) are included in this study in order to make estimations of effectiveness.

Initial Results

Four studies were identified (see Table 1). Two were classified as published papers (Bates, Saunders, & Wilson, 2007; Wilson et al., 2009), one was classified as an unpublished document (government report; Wilson, Picheca, & Prinzo, 2005), and one was a report in preparation (Bates, Macrae, Webb, & Williams, in press). Two were from a single research team in Canada and two were from a single research team at Hampshire Thames Valley Police, United Kingdom. The studies were all recent, ranging from 2005 to 2011. Table 1 provides details of the included studies.

Table 1.

Studies Examining the Effectiveness of CoSA (Circles of Support and Accountability).

Study	Sample description	CoSA core members (CM), n	Comparison group	Characteristics analyzed
Wilson, Picheca, & Prinzo (2005)	Retrospective sample of CoSA CMs taken from a Canadian pilot project.	60	60	Matched on age, Static-99, RRASOR, deviance, victim characteristics, and time at risk. “Recidivism” compared with matched sample (though not defined)—appears to be reoffending.
Bates, Saunders, & Wilson (2007)	U.K. follow-up cohort study of CoSA CMs from the HTV Circles organization.	14	0	Sexual reconvictions and questionnaire/qualitative data.
Wilson, Cortoni, & McWhinnie (2009)	Canadian national replication of Wilson et al (2005).	44	44	Matched on age, Static-99, RRASOR, deviance, victim characteristics, and time at risk. Recidivism (new charge or conviction) compared to matched controls and actuarial norms.
Bates, Macrae, Webb, & Williams (in press)	U.K. descriptive cohort study of CoSA CMs from the HTV Circles organization.	60	0	Sexual reconvictions and self-report, OASys (criminogenic factor database), and qualitative data.

To provide the most robust estimate of the effectiveness of CoSA in reducing reoffending, we sought to calculate an aggregated percentage reduction. It should be noted that in cost-benefit analyses the aggregation of the size of the effect in identified studies alone (including outcomes from meta-analysis) is not necessarily indicative of a positive result. Effect size alone has been found to be a poor indicator for overall net benefit, and the use of effect size alone can often lead to different policy recommendations (i.e., producing a highly significant reduction in offending can still produce a negative cost benefit; Marsh, Chalfin, & Roman, 2008). However, figures for percentage reduction were only available for the Canadian samples. Consequently, in order to include the U.K. data to the overall percentage reduction, and hence provide a more valid and culturally balanced comparison to the financial data, the raw Bates et al. (in press) data were acquired and a risk-norm analysis was conducted.

Risk-Norm Designs Using Bates et al. (2007) and Bates et al. (In Press) Data

The data in both Bates et al. (2007) and Bates et al. (in press) studies did not include a comparison with a control sample. We did, however, obtain the raw data for these studies for the purpose of this analysis. These data included category allocations for the Risk Matrix 2000 (RM2000; Thornton et al., 2003) actuarial risk assessment tool. The RM2000 categories are ordinal groupings along the risk continuum with higher categories representing relatively higher levels of risk of reconviction—low, medium, high, and very high risk (Thornton, 2007). Access to these data allowed for a basic risk-norm design analysis (e.g., Cortoni & Nunes, 2007) to be conducted. The Bates et al. (2007) data, however, reported only an average follow-up time (time at risk) of only 18 months and as such a comparison with the 3- and 5-year follow-up data provided for the RM2000 was not considered appropriate. Consequently, a risk-norm analysis was only provided for the Bates et al. (in press) data.

A risk-norm design compares the actual reoffending rate of a sample of offenders with a projected reoffending rate based on norms for a risk assessment tool (Wilson et al., 2009). Although the individual RM2000 scores (i.e., 0-10) were not available, we did obtain individual category membership (i.e., low, medium, high, very high). Therefore, we could not, as Wilson et al. (2009) were able, calculate the exact probability of reconviction for each CM. Instead, we took the number of CMs in each group and calculated the number of offenders in a group whose size would be used to predict reoffending based on the RM2000 norms. See Table 2 for a breakdown of those offenders predicted to reoffend by RM2000/S based on two reported normative samples,² as follows: (1) the latest norms supplied for RM2000/S (Thornton, 2007), and (2) a more recent large-scale U.K. analysis reported by Barnett, Wakeling, and Howard (2010). It should also be noted that the length of follow-up for the norms are longer than those of the CoSA sample (3 years follow-up for CoSA and 4 years follow-up for RM2000); therefore, the Bates et al. (in press) data may underestimate the sexual reconviction rates in the Bates et al. sample.³

Table 2.

Comparison Between Number of Actual Reoffenders in Bates, Macrae, Webb, & Williams (in press) and Projected Numbers Based on RM2000s Norms Reported by Thornton^a (2007) and Barnett et al. (Barnett, Wakeling, & Howard, 2010; rounded totals in parenthesis).

	Bates et al. (in press)		Thornton (2007) norms		Barnett et al. (2010) norms
		Actual reoffenders	% expected to reoffend		% expected to reoffend
Follow-up length	N	3 years	5 years	Predicted^a	4 years	Predicted^a
Low risk	16	1	3	0.48	0.7	0.11
Medium risk	17	2	13	2.21	2.8	0.48
High risk	13	1	26	3.38	9	1.17
Very high risk	8	1	50	4	27.3	2.18
Total	54	5	—	10.1 (10)	—	3.9 (4)

The number of offenders predicted to reoffend (e.g., Bates et al., in press; n × percentage expected to reoffend).

Based on the two Canadian studies and the risk-norm analysis, estimates for the effect of CoSA on both sexual and nonsexual reoffending rates were calculated.

Sexual Reoffending

The four papers include nine comparisons between a CoSA group and either a matched-sample group or normative risk data for sexual reoffending. Table 3 provides the calculated percentage change in each of these analyses and their related statistical significance level.

Table 3.

The Various Percentage Changes in Sexual Reconviction Reported in Analyses Between CoSA (Circles of Support and Accountability) and Control (or Risk-Norm) Samples.

			Sample size		Number of offenders reconvicted
S. No.	Study	Description	CoSA	Control	CoSA	Control	% change^b	Statistical significance
1	Wilson, Picheca, & Prinzo (2005)	Main comparison	60	60	3	10	70	χ²(1) = 4.23, p < .05
2	Wilson, Cortoni, & McWhinnie (2009)	Main comparison	44	44	1	6	83	ns (FET: p = .11)¹
3	Wilson et al. (2009)	Fixed 3-yearr follow-up	19	18	0	5	100	FET: p < .05¹
4	Wilson et al. (2009)	Cohort versus Static99 “high risk”	44	44	1	7	86	ns
5	Wilson et al. (2009)	Cohort versus Static-99 “routine”	44	44	1	3	67	ns
6	Bates, Macrae, Webb, & Williams (in press)	Cohort versus RM2000.s^a	54	54	5	10	50	ns
7	Bates et al. (in press)	Cohort versus Barnett^a	54	54	5	4	−25	ns

Conducted by the authors of this report using data provided by the authors of the original report.

The percentage change between CoSA and comparison samples was calculated thus: Change = (RC Ctrl – RC CoSA) / RC Ctrl.

The original data presented compromises the assumptions of chi-square (50% of 2 × 2 table cells < 5) so we present Fisher’s Exact Test (FET).

Table 3 shows that sexual reoffending is reduced by 70% and 83% in the two largest significant matched control studies, 100% in a significant fixed follow-up comparison, and 62% on average across the four nonsignificant analyses (the last four rows of Table 3). Taking the average of these three figures, weighted to account for sample size, we estimate an approximate reduction in sexual reoffending as a result of CoSA of 61% over an average 3.5-year follow-up period. Meta-analysis, using a fixed-effects model, shows that, combined, the results indicate a highly significant reduction in sexual recidivism (OR = 0.31, 95% CI = 0.12-.070, p <.01).

Any Reoffending (Sexual and Violent Offenses)

Sexual reoffending, as we have calculated earlier, is not the only form of reoffending that will create costs to the criminal justice system. The cost-benefit analysis that we aim to provide is based on the cost of rearrest, reconviction, and return to prison. Therefore, we calculated the reduction in any reoffending (so a combination of sexual and violent offenses) as a result of CoSA. The four studies include three comparisons between a CoSA group and either a matched-sample group data on a combination of sexual and violent reoffending. Table 4 provides the calculated percentage change in each of these analyses and their related statistical significance level. In order to conduct a risk-norm analysis for the two U.K. studies, the RM2000/V (risk of violent reoffending) and RM2000/C (risk of sexual or violent reconviction combined) were used. Statistics from the two Hampshire Thames Valley data sets (Bates et al., in press, 2007) did not include either the CM’s RM2000/V⁴ or RM2000/C⁵ scores and, hence, comparisons to the normative data for nonsexual offenses was not possible.

Table 4.

The Various Percentage Changes in Any Reconviction Reported in Analyses Between CoSA (Circles of Support and Accountability) and Control (or Risk-Norm) Samples.

			Sample size		Reconvictions (RC)
S. No.	Study	Description	CoSA	Control	CoSA	Control	% Change^a	Significance
1	Wilson, Picheca, & Prinzo (2005)	Main comparison	60	60	17	26	35	χ²(1) = 2.94, p < .09^b
2	Wilson, Cortoni, & McWhinnie (2009)	Main comparison	44	44	5	17	71	χ²(1) = 8.73, p < .01
3	Wilson et al. (2009)	Fixed 3-year follow-up	19	18	2	12	83	χ²(1) = 12.39, p < .01

The percent change between CoSA and comparison samples was calculated thusly: Change = (RC Control – RC CoSA) / RC Control.

Originally reported as p < .07.

As shown in Table 4, any reoffending is reduced by 35%, 71%, and 83%, respectively, in the three Wilson et al. (2005, 2009) analyses. Taking an average of the two⁶ main outcomes in Table 4, weighted to account for sample size, we estimate an approximate reduction in any reoffending as a result of CoSA of 50% over an average 3.75-year follow-up. Meta-analysis, using a fixed-effects model, shows that, combined, the results indicate a highly significant reduction in sexual recidivism (OR = 0.29; 95% CI = 0.12-0.67, p < .01).

Baseline U.K. Reoffending Rates

In order to establish the cost of reoffending for non-CoSA offenders it was also necessary to establish the base rate of reoffending by sex offenders in the United Kingdom. The latest reoffending rates for sex offenders in general were sourced from the U.K Home Office figures. These figures, however, were based on a 1-year follow-up. The likelihood that criminal behavior will reappear decreases the longer the person has abstained from that behavior; for example, the rate of reoffending between the time of release from prison and Year 2 is not likely to be the same as the rate of reoffending between Years 10 and 12 (Harris & Hanson, 2004). Hence, it is important that any estimate of reoffending be “time defined” (i.e., matched for follow-up time), to ensure that each group has the same opportunity or “time at risk” to commit subsequent offenses.

The average follow-up time for the CoSA samples were 3.75 years and so this seems inappropriate as a comparison. As such, we decided to use the latest figures from Barnett et al.’s (2010) assessment of the applicability of the RM2000. They report 4-year reoffending rates in the United Kingdom at 15.1% for sexual and/or nonsexual violent offenses.

Cost-Benefit Analysis

Cost-benefit analysis (CBA) is a well-known analysis technique and is used to assess the effectiveness of interventions in a way that allows choices to be made regarding the allocation of resources among competing approaches (see Swaray, Bowles, & Pradiptyo, 2005, for a review). Cost-benefit analysis is used for multiple purposes in addition to assessing intervention effectiveness. As Swaray et al. note, CBA identifies the scope of the problem (i.e., that effective interventions need also to be as cost-beneficial as possible), measures the effects of an intervention, monetizes the stream of intervention costs and outcomes, and subjects those estimates to a sensitivity analysis. Having established the effectiveness, we can then analyze the efficiency in terms of whether the benefits of the intervention outweigh the costs (Marsh et al., 2008).

Method

In order to calculate the potential monetary benefit of CoSA we first sought to calculate three figures: (1) the cost of “a Circle,” (2) the cost of reoffending, and (3) the base rate of reoffending by sex offenders. Direct and indirect operation in a Circle by established CoSA providers⁷ were calculated along with an estimate of the number of Circles per year. This enabled us to calculate an average “per Circle” running cost. This average cost was calculated to be £11,303 per Circle, per annum. These costs are shown below in Table 5.

Table 5.

The Cost of “a Circle”: Average Number of Circles per Year = 20.

	Cost	Average cost (£)
Direct costs	Salary costs	123,168^a
	Recruitment—volunteer	793^a
	Training—staff	400
	Training—volunteer	5,633
	Travel—staff	12,579
	Travel—volunteer	15,817
	Meeting costs	13,838
	Telephone—staff	267
	Telephone—volunteer	6,579
	Other costs	46
	Subtotal	179,120
Indirect costs	Media	800^a
	Premises costs	3,667^a
	Office running costs	29.736^a
	Allocation of central costs	12,735^a
	Subtotal	46,938
	Total	226,079
	Total “per Circle”	11,303

Includes estimates for CoSA organizations who receive some resources (e.g., office space, staff costs, etc.) “in kind” from the National Probation Service.

An estimate for the cost of the criminal justice response to reconviction, per offender, was calculated. We estimated tangible costs (direct costs of the criminal justice process) and intangible costs (the indirect cost of crimes to health, education, and extra costs to policing) for reoffending. Inflation calculators were used to provide estimates of these costs in 2009 (the latest inflation figures available). The estimated cost per reoffense was calculated to be £147,161 per offender. These costs and estimates are shown in Table 6.

Table 6.

Calculating the Cost of Reconviction.

	Description	Cost (£)	Source (original source costs in parenthesis).
Tangible costs	Criminal justice cost	15,766^a	Social Exclusion Report (2002) report regarding the cost of each offense leading to reconviction (£13,000).
	Crown Court cost	39,405^a	Harries’ (1999) study regarding Crown Court costs for each offense (£30,500).
	Prison service cost	45,000	Hansard’s (2010) parliamentary answer suggests regarding provision of a new prison place (£45,000; Hansard, 2010, c1251W).
	SOTP cost	8,090^a	Home Office (2006) Freedom of Information Request regarding cost of SOTP (£7,500).
	Subtotal	108,261
Intangible costs	Non–criminal justice costs	38,900^a	The cost of crimes committed (e.g., hospital treatment, victim services, property damage, out-of-pocket victim expenses, etc.). Government Spending Review (2000) cited In Social Exclusion Report (2002; £37,500).
	Total	147,161

Updated costs using Bank of England Inflation Calculator (http://www.bankofengland.co.uk/education/inflation/calculator/index1.htm)

The intangible costs outlined in Table 6 are those indirect costs identified as being borne by other related agencies, out-of-pocket costs to victims, and potential commercial costs. It has, however, been noted that there are many hidden intangible costs produced as a consequence of the deleterious effects on the victims of sexual abuse that may increase the true cost of reoffending, potentially by up to 10 times the intangible costs (McGurk & Hazel, 1998; Miller, Cohen, & Wiersma, 1996). This main analysis is based on the potential cost benefits of CoSA as an intervention and, therefore, it may not seem appropriate to include these hidden intangible costs in the main analysis, as they are unlikely to be borne directly by those providers. It is important though to acknowledge that reoffending has personal and societal costs and as such we have provided an analysis that includes these costs as an additional analysis in this article.

Modeling the Results of CoSA Interventions

To conduct the CBA we caulculated costs for a cohort of 100 hypothetical offenders released from prison per annum into the community, 50 in CoSA Circles, and 50 who were not. A decision tree models the two populations and their reoffending patterns . See Figure 1 for this decision tree.

Figure 1.

Cost-benefit analysis decision tree.

We can input our estimated reoffending and CoSA effectiveness data into the CBA decision tree. First, we have estimated that 15.1% of the non-CoSA cohort is predicted to reoffend (base reoffense rate). Second, we have estimated that there will be an approximate 50% reduction in any new offenses in the CoSA group. According to the figures presented above, we would predict the number of the cohort reoffending in Table 7.

Table 7.

The Breakdown of Reoffending by a Cohort of 100 Offenders Released From Prison Based on Predicted Base Rates of Reoffending and Reductions in Any New Offenses Due to CoSA (Circles of Support and Accountability).

Cohort	n	Any reoffense	No reoffense
No CoSA	50	8^a	42
CoSA	50	4^b	46
Total	100	11	89

15.1% of 50 non-CoSA offenders: (50 × 0.151).

50% fewer than the eight non-CoSA offenders: (8 – [8 × 0.5]).

Applying the costs calculated in Tables 5 and 6 produces the costs per outcome seen in Table 8. These costs are based on tangible criminal justice costs.

Table 8.

Cost Matrix for CBA (Cost-Benefit Analysis; Tangible Costs Only).

	Any reoffense			No reoffense
	Recidivism	Cost per person (£)	Total cost (£)	Recidivism	Cost per person (£)	Total cost (£)	Total cost (£)
Non-CoSA (Circles of Support and Accountability)	8	147,161	1,177,288	42	0	0	1,177,288
CoSA	4	158,464	633,856	46	11,303	519,938	1,153,794
						Cost saving (£)	23,494

We can see from the figures that the net benefit for the 50 offenders provided with CoSA is £23,494. Therefore, based on a total cost of £565,150 (50 offenders receiving CoSA at £11,303) and a total benefit of £588,644 (4 fewer offenders at £147,161 each), this results in an estimated benefit-cost ratio of 1.04.

Cost-Benefit Analysis (Including Intangible Criminal Justice Costs)

This analysis outlines the potential savings based on the costs that may be borne by criminal justice agencies, other related agencies, and the out-of-pocket costs to victims. Other researchers, however, have noted that the full extent of the cost to society through the deleterious effects on the victims of sexual abuse may be more in the region of 5 to 10 times the intangible costs (McGurk & Hazel, 1998; Miller et al., 1996) and these upper limits have been previously adopted to adjust cost-benefit analyses of sex offender interventions (e.g., Donato & Shanahan, 1999). These estimates are based on the amounts of money that nonaffected individuals (e.g., juries in compensation cases, research participants, etc.) place on the long-term physical and emotional damage caused by child sexual abuse on its victims.

In what follows, we present the potential “full societal costing” of the benefit of the reduction in reoffending due to CoSA (using the figures in Table 7). If we multiply the non–criminal justice costs in Table 6 by 5 times (taking the total cost of reconviction to £302,761—see Table 9) and 10 times (£497,261—see Table 10) we obtain different costs per outcome.

Table 9.

Cost Matrix for 5 Times Non–Criminal Justice Costs.

	Any reoffense			No reoffense
	Recidivism	Cost per person (£)	Total cost (£)	Recidivism	Cost per person (£)	Total cost (£)	Total cost (£)
Non-CoSA (Circles of Support and Accountability)	8	302,761	2,422,088	42	0	0	2,422,088
CoSA	4	314,046	1,256,256	46	11,303	519,938	1,776,194
						Cost saving (£)	645,894

Table 10.

Cost Matrix for 10 Times Non–Criminal Justice Costs.

	Any reoffense			No reoffense
	Recidivism	Cost per person (£)	Total cost (£)	Recidivism	Cost per person (£)	Total cost (£)	Total cost (£)
Non-CoSA (Circles of Support and Accountability)	8	497,261	3,978,088	42	0	0	3,978,088
CoSA	4	508,564	2,034,256	46	11,303	519,938	2,554,194
						Cost saving (£)	1,423,894

We can see from the figures that the cost saving for the 50 offenders provided with CoSA, when intangible costs are estimated to be 5 times the non–criminal justice costs, is £645,894. Based on a total cost of £565,150 (50 offenders receiving CoSA at £11,303) and a total benefit of £1,211,044 (4 fewer offenders at £302,761 each), this results in an estimated benefit-cost ratio of 2.14.

We can see from the figures that the cost saving for the 50 offenders provided with CoSA, when intangible costs are estimated to be 10 times the non–criminal justice costs, is £1,423,894. This produces a benefit-cost ratio of 0.56—every £1 invested in CoSA returns £0.56 in savings. Based on a total cost of £565,150 (50 offenders receiving CoSA at £11,303) and a total benefit of £1,989,044 (4 fewer offenders at £497,261 each), this results in an estimated benefit-cost ratio of 3.52.

Discussion

In this investigation a cost-benefit analysis of CoSA interventions was presented based on estimations of reductions in offending due to CoSA and U.K. estimated costs and recidivism rates. Although limited data are available on CoSA outcomes, and the quality of that data varies, it was estimated that there is a reduction of 50% in all types of reoffending for CoSA CMs compared with no intervention. The delivery of CoSA in the U.K. costs, on average, £11,303 per “Circle.” The U.K. costs of reconviction are estimated to be £147,161. According to these estimations, CoSA appears to be cost beneficial—although the related cost-benefit ratio is not substantially large. The net benefit of CoSA over reconviction was calculated to be £23,494 per year, per 100 offenders, equating to a cost-benefit ratio of 0.02. To put these figures into context, a recent review of cost-benefit analyses for 106 criminal justice programs (including sex offender treatments, drug treatments, basic education, electronic monitoring, etc.) between the years of 1988 and 2005 found the average cost benefit to be $13,711 (around £8,500; Marsh et al., 2008).

Hence, it would appear that few criminal justice interventions produce vast cost savings. If these estimations do prove to be a realistic and accurate reflection of cost-effectiveness, then the CoSA program would appear to fall within a reasonable range for interventions within these populations. This is important within a rehabilitative political climate. If a rehabilitative option can, at the very least, break even in terms of reducing offending to an extent that it doesn’t produce additional cost to criminal justice agencies then it may be viewed as a preferable option to a no-treatment alternative. That a number of those studies identified in this investigation relating to CoSA (e.g., Bates et al., in press, 2007; Wilson et al., 2009) report evidence that the offenders themselves consider CoSA to be a positive influence on their behavior and their ability to reintegrate into the community lends further credence to the provision of CoSA where appropriate being preferable to not providing any postrelease intervention at all.

It is also worth noting that any reconviction of an offender depends on those behaviors being detected. As Zedlewski (2009) states, a notable weakness in criminal justice evaluations is the reliance on recidivism, the outcome measure, as this is simply the recorded outcome of a single crime (or series of offenses) and it is not, therefore, a true measure of whether the offender is more or less active as a result of the intervention. So is it likely that recidivism rates are being underestimated in both CoSA and non-CoSA groups. It is, however, not necessarily a failing of CoSA if reconviction rates for CMs are not lower than those who do not have a Circle. An effective Circle will provide support in the absence of offending behaviors and accountability where offending does occur. That is not to say that all new offences by a CM would be detected, but the additional offender contact and accountability provided by a Circle may inflate the detection of new offenses in the CM group. This may be especially so for those that are more difficult to detect, such as breaches of parole conditions.

It is important too to note that the full extent of the cost to society through the deleterious effects on the victims of sexual abuse may be much greater than those that represent direct tangible costs. These costs might include costs to the victim, such as additional health care or loss of income; related to this, costs to the victim’s family or dependents; costs to the offender’s families in terms of loss of income, or potential loss of social benefits; and also indirect costs to commercial enterprises and the economy in general. As such, a further analysis found that CoSA may potentially produce more encouraging cost benefits when the estimated nontangible effects of sexual abuse on the victims are taken into consideration. The real cost saving to society may be in the region of about £650,000-£1.4m.

These estimates, though, are based on costs from the United States and Australia; hence, it would be of benefit to analyze the potential worth the U.K. population places on the non–criminal justice costs created by child sexual abuse. Indeed, having more well-defined and reliably collected data on outcomes related to societal integration by offenders may also contribute positively to future analyses of CoSA. Anecdotally, the CoSA model appears to have a positive impact on employment and securing housing for CMs. If it were possible to establish the extent to which this may be the case and a financial value then any additional benefits of additional tax revenue and reduced benefits of employment could be integrated into a cost-benefit model. There is an inherent difficulty in establishing intangible cost of sexual assault, as multiple victimization (i.e., one offender having a number of victims or one victim being linked to multiple perpetrators) and willingness to access services and the varying costs of those services, all combine to create very different costs for victimization.

There are a number of limitations in this analysis that should be taken into account when evaluating the outcomes. First, both the figures for the effectiveness and the financial figures presented in the CBA represent estimations produced by the authors for the purpose of this article. All efforts have been made to make these figures as evidence based as possible; however, they may not truly reflect the costs and benefits of CoSA. For example, it was difficult to estimate the cost of a Circle, due to operational and primary funding differences between different providers. The effectiveness figures were taken from a small number of mainly Canadian studies, with small numbers of CMs and, therefore, may not directly translate to the effectiveness of CoSA in the United Kingdom. There is certainly a pressing need for a large-scale RCT-style study of the effectiveness of CoSA in the United Kingdom on which we can base future policy and financial examinations.

Second, there were also some methodological compromises, for example, the use of the REA over a full systematic review, and the ability to explore nondatabase sources was limited (i.e., to contact other researchers in the field). Those U.K. data that were available were not randomized-controlled studies, and hence, we sought to estimate effectiveness using the risk-norm design (Cortoni & Nunes, 2007). In the risk-norm design it was difficult to match variables such as time at risk between studies and any estimation of effectiveness will be dependent on the quality and reliability of suitable risk assessment instruments and reconviction data.

There is continuing interest in developing and providing novel management solutions for sex offenders in the community. CoSA is a fairly new model, designed to augment management for sex offenders who may benefit from additional social support at the point of reentry to the community and is still being developed in terms of providing an increasingly effective service. As with the vast majority of criminal justice interventions, policy makers are keen to balance effectiveness and the costs to the state. As criminal justice agencies seek to provide both evidence-based and cost-effective programs, future analyses of emerging interventions should seek to include information about the related cost benefits (Zedlewski, 2009). The flexible models in which CoSA are delivered will continue to make it a difficult program on which to estimate costs and as such further efforts should be made to assess effectiveness in order to present the most accurate case. However, should these estimates be found realistic, then where the alternative is not providing a service for reentry into the community and where practitioners find that their clients are suitable, CoSA can be an effective tool.

Footnotes

Acknowledgements

The authors would like to thank Prof. Nick Freemantle at the University College London Medical School for his assistance in the statistical analysis of the effectiveness data. We also thank Jack Bergman at Avans University of Applied Sciences for his comments on a previous analysis of these data.

Declaration of Conflicting Interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This paper was commissioned and funded by Circles U.K.

Notes

References

Andersen

S. L.

Tomada

Vincow

E. S.

Valente

Polcari

(2008). Preliminary evidence for the sensitive periods in the effect of childhood sexual abuse on regional brain development. Journal of Neuropsychiatry and Clinical Neuroscience, 20, 292-301. doi:10.1176/appi.neuropsych.20.3.292

Barnett

G. D.

Wakeling

H. C.

Howard

P. D.

(2010). An examination of the predictive validity of the Risk Matrix 2000 in England and Wales. Sexual Abuse: A Journal of Research and Treatment, 22, 443-470. doi:10.1177/1079063210384274

Bates

Macrae

Webb

Williams

(in press). Ever-increasing Circles: A descriptive study of Hampshire and Thames Valley Circles of Support and Accountability 2002-9.

Bates

Saunders

Wilson

(2007). Doing something about it: A follow-up study of sex offenders participating in the Thames Valley Circles of Support and Accountability. British Journal of Community Justice, 5, 19-42.

Centre for Reviews and Dissemination. (2008). Systematic reviews: CRD’s guidance for undertaking reviews in health care. York, UK: York Publishing Services.

Correctional Service of Canada. (2003). Circles of Support and Accountability—guide to project development. Ottawa, Ontario, Canada: Author. Retrieved from http://www.csc-scc.gc.ca/text/prgrm/chap/circ/proj-guid/index-eng.shtml

Cortoni

Nunes

K. L.

(2008). Assessing the effectiveness of the National Sexual Offender Program (Research Report R-183). Ottawa, Ontario, Canada: Correctional Service of Canada. Retrieved from http://www.csc-scc.gc.ca/text/rsrch/reports/r183/r183-eng.shtml#_Risk

Donato

Shanahan

(1999). The economics of implementing intensive in-prison sex-offender treatment programs (Trends and Issues in Crime and Criminal Justice No. 143). Retrieved from http://www.aic.gov.au/publications/current%20series/tandi/121-140/tandi134.aspx

Easton

S. D.

Coohey

O’leary

Zhang

Hua

(2011). The effect of child sexual abuse on psychosexual functioning during adulthood. Journal of Family Violence, 26, 41-50. doi:10.1007/s10896-010-9340-6

10.

Finkelhor

(2008). Child victimization: Violence, crime, and abuse in the lives of young people. New York, NY: Oxford University Press.

11.

Finkelhor

Turner

Ormrod

Hamby

S. L.

(2010). Trends in childhood violence and abuse exposure: Evidence from 2 national surveys. Archives of Pediatric and Adolescent Medicine, 164, 238-242. doi:10.1001/archpediatrics.2009.283

12.

Friendship

Falshaw

Beech

A. R.

(2003). Measuring the real impact of accredited offending behavior programmes. Legal and Criminological Psychology, 8, 115-127. doi:10.1348/135532503762871282

13.

Ganann

Ciliska

Thomas

(2010). Expediting systematic reviews: Methods and implications of rapid reviews. Implementation Science, 5, 56-66. doi:10.1186/1748-5908-5-56

14.

Hansard

H. C.

(2010, March 3). c1251W. Retrieved from http://www.publications.parliament.uk/pa/cm200910/cmhansrd/cm100303/text/100303w0013.htm#10030367000136

15.

Harries

(1999). The cost of criminal justice (Home Office Research, Development and Statistics Directorate Research Findings No. 103). London: Home Office. Retrieved from http://rds.homeoffice.gov.uk/rds/pdfs/r103.pdf

16.

Harris

A. J. R.

Hanson

R. K.

(2004). Sex offender recidivism: A simple question. Retrieved from http://www.publicsafety.gc.ca/res/cor/rep/_fl/2004-03-se-off-eng.pdf

17.

Hillberg

Hamilton-Giachritsis

Dixon

(2011). Review of meta-analyses on the association between child sexual abuse and adult mental health difficulties: A systematic approach. Trauma, Violence, & Abuse, 12, 38-49. doi:10.1177/1524838010386812

18.

Home Office. (2006). Freedom of information request 3400. London: Author. Retrieved from http://webarchive.nationalarchives.gov.uk/20070701083828/http://www.homeoffice.gov.uk/about-us/freedom-of-information/released-information/foi-archive-offender-management/3400-sex-treatment-programme?view=Standard&pubID=407304

19.

Itzin

Bailey

Bentovim

(2008). The effects of domestic violence and sexual abuse on mental health. The Psychiatrist, 32, 448-450. doi:10.1192/pb.bp.107.018804

20.

Marsh

Chalfin

Roman

J. K.

(2008). What does cost-benefit analysis add to decision making? Evidence from the criminal justice literature. Journal of Experimental Criminology, 4, 117-135. doi:10.1007/s11292-008-9049-1

21.

McGurk

Hazel

(1998). The economic cost of child abuse and neglect in South Australia. Adelaide, Australia: Office for Families and Children & the Australian Institute of Family Studies.

22.

Miller

Cohen

Wiersma

(1996). Victim costs and consequences: A new look (National Institute of Justice Research Report). Washington, DC: U.S. Department of Justice. Retrieved from http://www.ncjrs.gov/pdffiles/victcost.pdf

23.

Social Exclusion Unit. (2002). Reducing re-offending by ex-prisoners. London: Office of the Deputy Prime Minister. Retrieved from http://www.gos.gov.uk/497296/docs/219643/431872/468960/SEU_Report.pdf

24.

Swaray

R. B.

Bowles

Pradiptyo

(2005). The application of economic analysis to criminal justice interventions: A review of the literature. Criminal Justice Policy Review, 16, 141-163. doi:10.1177/0887403404265631

25.

Thornton

(2007). Scoring guide for Risk Matrix 2000.9SVC (unpublished document). Retrieved from http://www.cfcp.bham.ac.uk/Extras/SCORING%20GUIDE%20FOR%20RISK%20MATRIX%202000.9-%20SVC%20-%20(ver.%20Feb%202007).pdf

26.

Thornton

Mann

Webster

Blud

Travers

Friendship

Erkison

(2003). Distinguishing and combining risks for sexual and violent recidivism. Annals of New York Academy of Sciences, 989, 225-235. doi:10.1111/j.1749-6632.2003.tb07308.x

27.

Willis

G. M.

Levenson

J. S.

Ward

(2010). Desistance and attitudes towards sex offenders: Facilitation or hindrance? Journal of Family Violence, 25, 545-556. doi:10.1007/s10896-010-9314-8

28.

Wilson

Bates

Vollm

(2010). Circles of Support and Accountability: An innovative approach to manage high-risk sex offenders in the community. The Open Criminology Journal, 3, 48-57.

29.

Wilson

R. J.

Cortoni

McWhinnie

A. J.

(2009). Circles of Support and Accountability: A Canadian national replication of outcome findings. Sexual Abuse: A Journal of Research and Treatment, 21, 412-429. doi:10.1177/1079063209347724

30.

Wilson

R. J.

Picheca

Prinzo

(2005). Circles of Support and Accountability: An evaluation of the pilot project in South-Central Ontario. Retrieved from http://www.csc-scc.gc.ca/text/rsrch/reports/r168/r168_e.pdf

31.

Zamble

Quinsey

V. L.

(1997). The criminal recidivism process. New York, NY: Cambridge University Press.

32.

Zedlewski

E. W.

(2009). Conducting cost benefit analyses in criminal justice evaluations: Do we dare? European Journal on Criminal Policy and Research, 15, 355-364. doi:10.1007/s10610-009-9108-9