Restorative Justice:
An Evaluation of the
Restorative Resolutions Project

Solicitor General Canada

October 1998

Report No. 1998-05
Cat. No. JS42-84/1998E
ISBN: 0-662-27346-X

Acknowledgements

There have been two interim evaluations of the Restorative Resolutions program in Winnipeg. The first evaluation was completed by Gord Richardson and Burt Galaway from the University of Manitoba. They worked together with program staff led by Michelle Joubert and established the framework for collecting the data. Their work allowed us to continue with the subsequent evaluations, culminating in the present report. We are indebted for their efforts.

We would like to thank Yvonne Lesage, Michael Gray, and Debbie Carriere for their assistance in collecting some of the data used in constructing the present databases, Kevin McAnoy for gathering and coding the studies in the meta-analysis and Janice Martens for information on the amounts of restitution paid by the offenders. Our appreciation is also extended to Vic Bergen and Ron Parkinson for facilitating the data collection at Headingly and Milner Institutions and for providing the recidivism information. Of course, our appreciation is extended to the staff of Restorative Resolutions for their co-operation and dedication to maintaining the databases used in this evaluation.

The opinions expressed are those of the authors and do not necessarily represent the official position of the Ministry of the Solicitor General of Canada.

Table of Contents

Acknowledgements *

Table of Contents *

Introduction *

Results and Discussion *

Summary and Conclusions *

References *

Appendix A *

Restorative Justice: An Evaluation of the Restorative Resolutions Project

The program evaluated in this report is based upon principles of restorative justice and a concern for maintaining offenders safely in the community. There are many such programs operating in Canada (Church Council on Justice and Corrections, 1996). However, the Restorative Resolutions project operating in Winnipeg is one of the few restorative justice programs that had from its inception an evaluative component. Consequently, the present evaluation, based upon the program period of October 1, 1993 to May 9, 1997, represents an important contribution to our knowledge of community-based restorative justice. The evaluation is intended to inform not only the continued development of the Restorative Resolutions program but other programs that attempt to divert offenders from prisons within a restorative justice context.

Restorative Justice

Restorative justice is an approach for dealing with offenders that differs from traditional criminal justice processing. In North America, criminal behaviour is primarily viewed as an act against the state that requires retribution. Violation of law is met with punishment and the legal consequences of violating laws expresses society's condemnation of the act and demonstrates the application of justice. In addition, the administration of penalties is intended to deter the offender and others who may be tempted to transgress society's norms.

During the last decade however, empirical evidence has accumulated indicating that criminal justice sanctions have little impact on the recidivism of offenders (Andrews & Bonta, 1998; Andrews, Zinger, Hoge, Bonta, Gendreau & Cullen, 1990; Gendreau & Goggin, 1996). Moreover, the emphasis on the offender has been criticized by some victim groups who have felt abandoned and betrayed by the criminal justice system. These two developments have fostered the growth in restorative justice approaches (Messmer & Otto, 1992).

Restorative justice is a new way of "settling disputes"(Hudson & Galaway, 1996). The victims, their families and friends, and the broader community are seen as the recipients of the harm caused by the offender's actions. Restorative justice approaches seek to repair this harm by direct contact between victim and offender rather than solely by the state. The process of reparation involves bringing together the offender, victim and community to seek solutions that, to the greatest extent possible, satisfy all parties. Through this process of mediation, reparations are negotiated and the process of forgiving and healing is initiated. At an operational level, restorative justice practices are most evident in victim-offender reconciliation meetings (Hudson & Galaway, 1996; Umbreit, 1994). However, restitution and community service programs may also be included (Hudson, 1992; Walgrave, 1992). More recently, approaches originating in Aboriginal communities such as sentencing circles and family case conferencing have found their way into the mainstream North American criminal justice system. These programs share many features of restorative justice (e.g., mediation, community involvement) although there has been some debate over their benefits (LaPrairie, 1998; Umbreit & Zehr, 1996).

The roots of restorative justice may be traced back to the victim-offender reconciliation programs (VORP) developed in the early 1970s. The first program was established in Kitchener, Ontario, in 1974. Under the sponsorship of the local Mennonite Church, the program used structured mediation techniques in face-to-face meetings between offender and victim. The purpose of these meetings was to deal with the need by both parties to gain information about the criminal justice process and to resolve the victim's emotional upset. VORP differs from the victim-offender meetings in restitution programs in that there is a greater emphasis on reconciliation than on monetary repayment.

Subsequent to the Kitchener program, a VORP was established in Elkhart, Indiana in 1978 (once again, by the local Mennonite Church). Since then, VORPs have proliferated. Umbreit's (1994) survey of programs in the United States found 50 Victim Offender Reconciliation Programs in 1986, growing to 123 by 1994. The growth of VORP was not limited to the United States. In Canada, where VORP was first established, there are an estimated 26 programs and in Europe the number of VORPs exceeds 500 (Church Council on Justice and Corrections, 1996; Umbreit, 1994).

There are two important features of restorative justice: 1) victim and community participate in the administration of justice, and 2) offenders are managed in the community. The emphasis on the victim is particularly important. Advocates of restorative justice have been highly critical of the criminal justice system for neglecting the victims of crime and focusing on the punishment and rehabilitation of offenders. Meeting the offender addresses certain victim needs (e.g., emotional satisfaction and personal healing) and brings the victim's perspective into the administration of justice. As a result, restorative justice provides a sense of victim empowerment.

Although the victim plays a central role in a restorative justice philosophy, there is also the goal of managing the offender within the community as opposed to relying on imprisonment to deal with the offender. Most restorative justice programs present themselves as an alternative to incarceration and an option to traditional prosecutory processes (Nuffield, 1997). The programs target offenders either prior to conviction (most often in the case of juveniles) or prior to sentencing (usually with adults). The goal is to avoid a custodial sentence and maintain the offender in the community.

It is also worth noting that the principles of restorative justice are applicable throughout the criminal justice process. For example, victims and offenders could meet while the offender is incarcerated and preparing for conditional release. Applications of this kind are relatively uncommon at this point in time.

Evaluations of Restorative Justice

The evaluation research on restorative justice programs varies from general descriptions of program processes and anecdotal accounts of their value to more carefully conducted studies with comparison groups. From a methodological perspective, anecdotal accounts provide the weakest form of evidence. The majority of the better evaluations focus on the success of the programs in achieving restorative justice goals. That is, their success in bringing together the victim and offender, arranging restitution and community service agreements, alleviating victim emotional upset, etc. Assessing the value of the program vis-à-vis these goals is entirely consistent with a restorative justice perspective.

Evidence on the success of programs in achieving restorative justice goals has been quite favourable. Both victims and offenders report satisfaction with the results of their reconciliation meetings, subsequent restitution (if any), and community service agreements. Clearly, such findings are important for the maintenance of these programs. After all, if staff or clients see no value in the service then the future of the program is in doubt. However, the general importance of the "consumer satisfaction" studies is attenuated by the difficulties many programs experience in trying to bring the victim and the offender together. For example, Gehm (1990) found that 53% of the victims from six VORPs refused to meet the offender. Therefore, evidence of program satisfaction is often based upon highly selected samples.

Further problems with the interpretation of the high satisfaction ratings is the influence of various types of offenders and victims. The majority of VORPs involve juvenile offenders who have committed relatively minor offenses. Typically, adult victims are more lenient toward younger offenders and those who have committed less serious crimes (Gehm, 1990) thereby influencing the high satisfaction ratings. Regarding victims, Umbreit (1990) has described three types of victims that may respond differentially to offenders. There is the "healer", the "fixer" and the "avenger". The "healer" is interested in the rehabilitation of the offender and the "fixer" wants to be compensated for the damages resulting from crime. Obviously, the "avenger" type of victim may not be responsive to meeting the offender nor to give positive evaluations of restorative justice practices.

Arranging victim-offender mediations and achieving victim satisfaction are not the only goals of restorative justice approaches. Community safety is also a goal (Bazemore, 1996). Reasonably well-designed studies of the impact of restorative justice programs on recidivism are few. For example, McCold's (1997) bibliography of 552 reports on restorative justice identified only two reports that had a comparison group and provided recidivism outcome data.

To further explore the impact of restorative justice programs on recidivism, we undertook a brief meta-analytic review of the literature. Meta-analysis is a quantitative approach to summarizing the literature and it has largely replaced the more traditional, narrative literature review. Beginning with McCold's (1997) bibliography and conducting our own search of the literature, we identified 14 evaluations of restorative justice programs yielding 20 effect size estimates (phi coefficients).

The criteria for including a study were that there must be a comparison group and that the recidivism outcome must be reported in a way to permit the calculation of the phi statistic. The phi statistic is a measure of association used to evaluate the relationship between two dichotomous variables. In our case, we assessed the association between the presence or absence of restorative justice and recidivism (yes/no). We followed the coding procedures used by Andrews, Zinger et al. (1990) in their meta-analysis of the offender rehabilitation literature. In addition, multiple effect sizes within a single study were averaged following the procedures described by Bonta, Law and Hanson (1998) in order to produce one effect size per study. The detailed results of the meta-analysis are shown in Appendix A.

The average phi coefficient, after adjustments for sample size and base rates, was .08. This value corresponds to approximately an 8% decrease in recidivism associated with programs that had restorative justice features. Although these results are promising, there was considerable variation among the studies. Some studies reported very large decreases in recidivism (e.g., Heinz et al., 1976) and some found increases in recidivism (e.g., Bonta et al., 1983). In addition, all the studies reviewed in the meta-analysis had methodological shortcomings. None used random assignment and few used matched comparison groups.

To illustrate some of the difficulties encountered in evaluations of restorative justice programs, we take one of the more sophisticated outcome studies, Umbreit's (1994) quasi-experimental evaluation of four VORPs. The programs targeted juveniles (average age of 15 years) who were mostly first offenders (73%). One of the comparison groups consisted of victims and offenders matched on age, sex, race and offence type but who were not referred to mediation.

As with most evaluations of VORPs, Umbreit found high levels of satisfaction (over 90%) with the victim-offender reconciliation meetings. However, there was a 64% attrition rate (p. 62) and moreover, 95% of the mediations resulted in restitution agreements. Thus, program satisfaction was based upon a select sample of victims who received some monetary payment. Nevertheless, an important finding was that victims participating in VORP were less upset by the crime and reported reductions in their fear of being victimized.

One year post-program follow-up for new offences found that the matched comparison group had a recidivism rate of 27% while the juveniles who participated in VORP showed a lower rate of 18%. The differences however, were not statistically significant. Umbreit (1994; p. 117) explained the lack of statistically meaningful differences in recidivism by writing "it is naïve to think that a time-limited intervention such as mediation by itself (perhaps four to eight hours per case) would be likely to have a dramatic effect on altering criminal and delinquent behaviour". Returning to restorative justice principles, Umbreit reminds the reader that the high levels of victim satisfaction are something that traditional criminal justice processing has failed to demonstrate.

In summary, studies of restorative justice clearly show the complexity of implementing and evaluating an approach that is relatively new in North America. The introduction of the victim and the community in criminal justice processing requires consideration of factors normally ignored in mainstream criminal justice. The research thus far has shown that restorative justice approaches can have a significant impact on the views of victims toward offenders and the criminal justice system. Regarding recidivism, the effectiveness of restorative justice programs is small, but positive. Most studies however, are based upon juvenile samples and all have serious methodological limitations.

The Restorative Resolutions Program

Many restorative justice programs assume a victim advocacy perspective. The Restorative Resolutions (RR) program however, is somewhat unique in this regard. RR operates through the John Howard Society of Manitoba, an offender oriented voluntary sector agency. Nevertheless, RR does attempt to follow the restorative justice principles of redressing the harm to victims, encouraging community involvement in the criminal justice process and managing the offender in the community. Providing an alternative to incarceration within a restorative justice context is one of the most important features of RR and it was a major reason for undertaking an evaluation of the program.

The RR program began in October of 1993 as a demonstration project. Referrals to the program were invited from Community and Youth Corrections, Crown and defence attorneys, judges, community agencies, family members, and the accused themselves. The major purpose of RR was to provide a community-based alternative to incarceration. Therefore, procedures were adopted to ensure that those accepted into RR were likely to receive a prison sentence were it not for the program.

The offenders referred to RR were to meet the following primary conditions:

In addition to the above, there were further precautions taken against net-widening. Efforts were made to accept offenders with a history of probation breaches and prior incarceration into the program. However, offenders who committed sexual assaults, gang related or drug related offences or who were involved in domestic violence were excluded. Notwithstanding these secondary considerations, the Crown's recommendation for a custodial sentence remained the major criteria for acceptance into RR. Thus, even first time offenders were eligible provided the Crown was recommending a custodial sentence of at least six months.

Once RR staff accepted the offender into the program, work began on developing an individualized restorative plan. Included were attempts to contact the victim and members of the community in order to enlist their help in the development of the plan. Where appropriate the plan also addressed the treatment needs of the offender. If accepted by the judge, RR staff implemented the restorative justice plan and provided, or obtained, the necessary services outlined in the plan.

The Interim Evaluations

There have been two earlier evaluations of RR. The first evaluation was conducted by Richardson and Galaway (1995). At that time, the program was approximately half way through its three year funding arrangement. In general, the results suggested that RR was targeting offenders who were likely to receive a sentence of incarceration. Referrals however, were considerably below projections. As a consequence, after referrals were screened according to the RR criteria, plans developed and finally accepted by the Court, only 32 offenders were in the program. The low number of clients was attributed to the developmental phase of the program. During the early phase of the program, considerable time was spent publicizing RR and encouraging referrals.

Richardson and Galaway (1995) also reported findings gathered from victims and a public opinion survey. Many victims were reluctant to participate with either the program or the evaluation. Victims reported that they wanted to put the experience behind them and that they were too busy to become involved. Only two victims agreed to be interviewed by project staff. Interestingly, half of the 16 offenders interviewed by the evaluators felt that most offenders would not want to meet their victims.

As part of the Winnipeg Area Survey, 814 people were asked a number of questions about their views of restorative justice principles. Considerable public support for these principles was expressed. Nearly three-quarters of the respondents (72%) said that they would be willing to participate in victim-offender mediation, a far cry from the actual experience of RR staff. Furthermore, two-thirds preferred to receive restitution rather than see the offender receive a custodial sentence.

The second evaluation was conducted to take advantage of a longer time frame and therefore, a larger sample. Bonta and Gray (1996) extended the first evaluation by an additional 14 months (up to April, 1996). The rate at which offenders were accepted into the RR program remained unchanged between the two evaluations despite the fact that the criteria of the Crown's recommendation was reduced to a nine month sentence. During the first 31 months, 54 clients were accepted into the program (from 190 referrals).

The program still appeared to target prison-bound offenders. All offenders had a recommended sentence of nine months or more. In comparing the results from the two evaluations, there was also an increase in the use of restorative justice practices. The first evaluation found that one-third of the plans included restitution and 37% included community service work. In the second evaluation, restitution was included in slightly over one-half of the plans and community service was a feature in nearly all of the plans (96.6%). RR staff contacted 122 victims but the majority of the victims (79.5%) did not wish to meet the offender. There were only 11 face-to-face meetings between victim and offender.

Bonta and Gray (1996) conducted a preliminary outcome evaluation. A post-program outcome evaluation was not possible as many of the offenders in RR were given lengthy probation dispositions. Therefore, a one year, in-program evaluation was conducted. For 35 offenders who were sentenced to RR and were on the program for at least one year, 80% were successful. Two of the seven failures were due to a new offence. This success rate compared favourably to a group of probationers with similar risk-needs profiles.

The Present Evaluation

The present evaluation adds to the earlier reports by extending the time frame to May 9, 1997. Moreover, a concerted effort was made to verify all available data elements against multiple sources and to minimize missing data. The earlier reports used databases where information was missing for a large number of cases. Official files were reviewed and missing information entered. In the process, inconsistencies in the data were cross-referenced with correctional files and errors corrected. As a result, some data presented in this report may not be entirely consistent with that reported earlier. However, even after these efforts, some information remained missing.

Results and Discussion

Restorative justice programs have multiple goals. Success can be measured in terms of the appropriate targeting of clients, achieving restorative justice goals such as arranging victim-offender meetings, and reducing recidivism. As an alternative to imprisonment, RR also needs to demonstrate that it does not widen the net and bring additional individuals under correctional control. All of these goals have merit and the overall value of a program rests upon how many, and how well, the goals are met. In this evaluation, we report upon the success of RR in reaching these goals. Specifically, we present the results regarding: 1) client target selection, 2) doing restorative justice, 3) providing an alternative to imprisonment, and 4) reducing recidivism.

Before proceeding, we have a few comments on the use of statistical significance testing. Throughout the report, numerous comparisons and predictions are made concerning the offenders who were under RR supervision and offenders who did not participate in the RR program. When differences are found or relationships identified, the question is asked whether the findings are due to chance. The commonly accepted practice in the social sciences is that findings with probabilities less than 5% are "statistically significant". Statistically significant findings are usually expressed using the probability value, p < .05. Sometimes, we may report values at p < .01 (findings due to chance are less than one in one hundred) and p < .001 (one in one thousand). Finally, we would note that when sample size is small, it becomes more difficult to identify statistically significant relationships. When this situation arises, we try to mark it for the reader.

1. Client Target Selection

There are a number of steps before an offender becomes a client of RR. Offenders must be referred, meet program criteria, have plans developed and have the plans accepted by the Court. Along this path, there are many points where offenders are lost to the program. Figure 1 shows the significant attrition that occurs from referral to RR to the point where the offender is sentenced to RR. There were 297 referrals to RR during the period covered in this evaluation (October 1, 1993 to May 9, 1997). However, only 99 offenders actually had their plans accepted by the Court.

Referrals to RR. Slightly more than three-quarters of the referrals to RR were males and a little more than half were single. The average age of the referred cases was 27.8 years (SD = 9.7). One referral was a young offender not accepted by the program. Given that RR was intended for offenders likely to receive a prison sentence, it was not surprising that 65% of the referrals had prior contact with the criminal justice system. Table 1 presents information on the source of the referrals and the characteristics of the offenders referred to the program.

Consistent with the earlier evaluations of RR, defense attorneys provided the majority of referrals to the program. There was also significant variation in the type of offender referred by different sources. Referrals from the Crown were the most conservative (see Table 2). On average, the referrals from the Crown had fewer prior probation failures and prior custodial sentences (0.4 and 1.4 respectively). In addition, only 16.2% were charged with a violent offence. Self-referrals, not surprisingly had the highest average number of probation failures (2.0) and prior incarcerations (6.6).

Program Acceptance. One hundred and seventy-four of the 297 referrals met the selection criteria for RR and were subsequently accepted into the program by project staff. The most important criteria for acceptance into the program was a Crown recommendation of a custodial sentence exceeding six months. In 91.4% of the cases accepted by RR staff, the Crown's recommendation was at least six months. Specific information on the sentence recommended by Crowns for both the accepted and rejected referrals was available for only 61 cases. Offenders accepted by RR had longer sentence recommendations (18.0 months, SD= 9.5, n = 45) than the 16 offenders rejected by RR staff (9.7 months, SD = 6.0; t = 3.3, p < .01). These findings strongly support the conclusion that RR staff were targeting offenders likely to receive a custodial sentence thereby providing an alternative to incarceration.

Beyond the differences in the Crown's recommendation, there were few other differences between the offenders judged to meet the criteria for RR and those who were not accepted (Table 3). First time offenders were over-represented with 43.7% of first offenders accepted by RR compared to only 14.8% who were rejected (c ² = 21.77, df = 1, p < .001). In addition, offenders charged with crimes against the person and Aboriginal/Metis offenders were less likely to be accepted by the program (c ² = 5.21, p < .05 and c ² = 7.36, p < .01 respectively).

Note: p = probability level; ns = nonsignificant
Numbers vary due to missing information. Five pending cases were in the Not Accepted group.

Further analyses of the available data on the Aboriginal referrals found no differences between those who were rejected by RR and those accepted. There were no differences in age and gender between the two groups and they showed similar criminal histories (e.g., number of charges, prior incarcerations). The only difference found was that a significantly higher proportion of Aboriginal offenders than non-Aboriginal offenders accepted for the program by RR staff were charged with a crime against the person (42.1% vs 23.9%; c ² = 4.89, p < .01).

One important, non-offender factor that may have influenced acceptance into the RR program was the referral source. Even though the Crown made relatively few referrals, when referrals were made to RR, most of the cases (83.8%) were accepted. Next to the Crown, 69.3% of the referrals from defense attorneys were accepted by RR. Referrals by probation officers and by the offender himself/herself were less likely to be successful (34.9% and 20.5% respectively).

To summarize, both the Crown's recommendation for a custodial sentence and referral to RR were associated with acceptance into the program. The first finding is consistent with the stated referral criteria of a custodial sentence recommendation for RR. A custodial recommendation ensures that the program provides a true alternative to incarceration. The high likelihood of being accepted by RR when the referral comes from the Crown suggests a number of mediating factors. First, the concerted efforts of RR staff to educate Crown attorneys about the program may have resulted in Crowns being more familiar with the referral criteria. As a result, the Crown may be in a better position to make more appropriate referrals. Second, referrals from the Crown also signal to RR that investing in the case will likely lead to the Court's endorsement of the plan, a hypothesis that is confirmed from findings reported later in the report.

Another critical factor in being accepted by RR is the offender's motivation to participate in the program. If the offender was judged by staff as unwilling to assume responsibility for his/her behaviour or unwilling to meet the victim, then they were screened from the program. Unfortunately, data was unavailable on how many referrals were actually refused by the program because of poor motivation.

Once accepted by RR, individualized plans were developed. In general, 85.6% of the 174 offenders accepted by RR had a plan developed or a plan was pending. Developing a community management plan involved a considerable amount of the staff's time. On average, 25.5 hours were required per case. However, only 120 plans were formally presented to the Court with an average of 112.0 days (SD = 68.0) elapsing from acceptance by RR staff to the Court's endorsement of the plan. Plans were not developed for 25 offenders and another 20 were pending. Nine plans were developed but never submitted. Of the 120 plans submitted to the Court, the judge rejected 18 plans and three others were pending. In the final analysis, the Court accepted the plans for 99 offenders or 82.5% of the cases submitted to the Court. Available data on 97 offenders showed an average sentence of probation of 28.5 months (SD = 6.8). Most of the plans were accepted without modification (73.2%). For 22.7% of the cases, the Court added conditions to the plan and made deletions in only 4.1% of the cases.

Offenders who had their plans accepted by the Court did not differ from those rejected with respect to race, employment, marital status, and gender. Offenders who had committed a crime against the person were less likely to have the plan accepted. Of the 18 rejected plans, 11 offenders (61.1%; c ² = 7.04, p < .01) committed a crime against the person. Once again, the influence of the Crown in the process was evident. All but one of the 57 offenders who were supported by the Crown had their plan accepted by the Court. Only 21.1% of offenders who committed a crime against the person received the support of the Crown.

In light of the high attrition from acceptance by RR staff to the Court's disposition of the case (from 174 to 99), an analysis of the drop-outs was conducted. Assessments on the Manitoba Risk-Needs classification instrument showed no differences in risk-needs scores. The average score for program placements was 10.3 and for drop-outs the average score was 10.2 (t = -.09). In addition, there were no differences in terms of prior criminal history, age, gender, race and source of income.

As noted earlier, the selection criteria with respect to the Crown's recommendation of a custodial sentence changed over time. This led us to question whether there were also changes in the type of offender accepted by RR over the three and one-half years of the project. Table 4 shows the characteristics of offenders during the three phases of the project. Although there appears to be a slight trend to accepting a lower risk offender, most of the changes were statistically nonsignifcant. For example, the average risk-needs scores as measured by the Manitoba classification instrument decreased since the beginning of the program. The only statistically significant changes were with respect to the type of offence and race. Compared to the beginning of the program, RR accepted fewer offenders who committed crimes against the person (c ² = 3.91, p < .05) and Aboriginal/Metis offenders (c ² = 5.69, p < .05).

Notes: T1 = October 1, 1993 to February 28, 1995

T2 = March 1, 1995 to April 30, 1996

T3 = May 1, 1996 to May 9, 1997

In summary, RR was quite successful in targeting offenders who were likely to receive a custodial sentence. Almost all of the offenders (91.4%) had Crown recommendations for a custodial sentence of at least six months. Furthermore, the majority of the offenders accepted by RR had prior experience with the criminal justice system averaging three prior incarcerations. Once accepted by RR, nearly four months passed before the plan was submitted to the Court. During this time considerable effort was invested by RR into developing a restorative justice plan. Only one offender committed a new offence during this time period (see Figure 1). Finally, we can conclude that the offenders who had their plans accepted by the Court represented a group of offenders diverted from incarceration.

2. Doing Restorative Justice

Restorative justice encourages the involvement of the victim and the community in the criminal justice process. In addition, the offender is held responsible for making reparation to the victim and community for the harm caused by his/her actions. The success of RR in meeting these goals was evaluated from information provided in RR plans and from information stored in the case and victim data bases.

There were 249 victim files available for review. RR staff contacted 209 (83.9%) victims. Employees of businesses represented the largest group of victims (41.5%) followed by private individuals (29.8%) and owners operating their own business (16.5%). The losses reported to the police varied from the trivial ($20 of candy) to goods in excess of $20,000 (e.g., stolen automobiles). Physical injuries to the victims were quite rare (4.9% of the cases), but 22.2% of the victims reported some form of psychological suffering.

RR staff attempted to contact victims and request their participation in the development of a restorative justice plan. Thirty-four victims could not be contacted due to difficulties in locating them. Only 10.3% (25) of the 243 victims for whom we have information actually met the offender. This low rate is not uncommon in many restorative justice programs and may be explained by a number of factors. Some victims may be reluctant to meet the offender because they wished to forget the experience. Others may not have been sufficiently affected by the victimization to seek some form of emotional closure to the experience. Recall that only a minority of the victims reported physical or psychological harm. However, for the 25 face-to-face meetings that did occur, mutually satisfying agreement was reached in 24 cases and apologies were made in person to 22 victims. In addition to the personal apologies, another 58 victims were sent written apologies from the offenders.

Although relatively few victims met the offender, there were other indications of victims participating in the criminal justice process and benefiting from the offenders' efforts at reparation. A majority of victims (78.6%) provided victim impact statements and, as already noted, 58 victims (23.9%) received written apologies from the offenders. Reparation was also evident from restitution and community service agreements.

Restitution was ordered by the Court for 53 of the 94 offenders placed on the program. The amounts ordered to be paid ranged from $200 to $42,000 with an average restitution order of $5,622. Information regarding the amounts actually paid by the offenders was available for 51 cases. Twenty-one offenders paid in full, with another nine cases still active. Thirteen offenders did not pay any restitution. The average amount of restitution paid by the RR clients was $2,563. In total, $130,741.37 was paid to crime victims.

Since the Richardson and Galaway (1995) report, community service has become an important component in the plans of RR clients. In the first interim evaluation, Richardson and Galaway (1995) found evidence of community service in only 37% of the plans. In response to this finding, RR increased their efforts to correct this shortcoming. By the second evaluation (Bonta & Gray, 1996), community service was found in 96.6% of the plans. This high rate of community service has continued. However, based upon the complete time span in this evaluation, 69% of the plans specified community service. The number of hours of service ranged from 50 hours to 800 hours (the average number of hours was 175.9). Unfortunately, records regarding the Court's agreement with RR's recommendation for community service and the completion of community service agreements were unavailable.

Restorative justice was not the sole feature of the plans developed by staff. Almost all of the 174 plans (96.7%) included recommendations for counselling or treatment services. Attention to the individual needs of offenders are often overlooked in diversion programs (Nuffield, 1997) even when offenders clearly present such needs. Analysis of the risk-needs profiles of the offenders showed alcohol/drug problems among more than half of the offenders (58.3%). Employment problems were identified in 64.8% of the cases and unstable family/marital relationships were very common among these offenders (70.9%). Also noteworthy was that 44.9% were assessed as having emotional difficulties and 60.6% were identified as associating with companions who had a negative influence.

When evaluating RR's success in following restorative justice principles, comparisons to traditional probation are instructive. Both restitution and community service are options available to probation. Therefore, one may ask if probation, as it normally operates, can be as effective in following restorative justice principles as a more formal program such as RR. In this case, the answer to the question is no. Examining our database of over 1,000 probationers from Manitoba, the rates of restitution and community service are much lower. Compared to probation cases, RR plans were more likely to include restitution (56.4% vs 24.9%) and community service (96% vs 13.8%). The average amount of restitution paid by RR clients ($2,563) was also greater than the amounts ordered for probationers ($1,517). Finally, we cannot overlook the success of RR in obtaining apologies from offenders to their victims - something very rare in traditional probation.

In summary, RR has been relatively successful in adhering to restorative justice principles. Victims were contacted and invited to participate, community service had become almost universal feature of the plans and restitution agreements were much higher than that found in probation. In total, over $130,000 was paid to victims of crime. Lastly, and importantly, RR paid significant attention to meeting the needs of offenders. Recent theorizing in the areas of restorative justice and offender rehabilitation has suggested that both approaches have much in common and together they can contribute to enhanced confidence in the criminal justice system (Crowe, 1998). RR has made offender treatment planning an integral component of the restorative justice plan and an additional valuable feature of the program.

3. An Alternative to Incarceration?

Working with offenders in the community is a valued principle in restorative justice programs, but it need not be the only goal. Some restorative justice programs, including RR, also aim to provide an alternative to incarceration. This is certainly one of the stated goals of RR. The problem with many alternative to incarceration programs is that they deal with relatively low risk offenders who would have received a less intrusive intervention were it not for the availability of the "diversion" program (Nuffield, 1997).

Earlier we reported that over 90% of the RR offenders had sentence recommendations of six months or more. Although a Crown may recommend a custodial sentence, the Courts may choose a noncustodial disposition. To increase our confidence that the 99 offenders placed into RR would have likely received a prison sentence, their risk level was assessed by the Manitoba Risk-Needs classification instrument. The average risk-needs score of the RR offenders was 10.4, slightly higher than the average score of 8.8 found among probationers in Manitoba. Following Manitoba's classification guidelines for probation, only 9.9% were classified as low risk (scores 0 to 5). 50.5% were medium risk (6 to 11) and 39.6% were classified as high risk (scores of 12 or more). That is, approximately 90% of the RR clients were classified as medium risk-needs or higher suggesting that RR was indeed targeting an appropriate group for diversion from prison.

Despite the efforts of RR staff to target prison bound offenders, the actions of the Court modified the impact of RR in actually diverting offenders from prison. Analysis of Court dispositions found that 18 of the offenders (nearly 19%) received a custodial sentence in addition to RR placement. Their average sentence of imprisonment prior to RR was 4.9 months and ranged up to 24 months (one offender was still in prison at the time of the present review). One third of those with a custodial disposition had an intermittent sentence (average sentence of 2.8 months) and the remainder had an average sentence of 5.9 months.

Offenders who were given a custody sentence in addition to RR were compared with offenders who were placed directly into RR. Table 5 summarizes the comparisons between the two groups of offenders. Very few differences were found. Offenders placed into custody before RR were more likely to have been convicted of a crime against the person. However, the seriousness of the violent crimes, as measured by injury to the victim, was no different than for the offenders placed directly into RR. Overall, risk-needs scores and criminal history were similar for both groups except offenders placed directly under community supervision had more prior breaches of community supervision.

When the nonsignificant differences between the two groups in risk-needs scores and victim injury are considered, the question is raised whether the Courts created a net-widening effect. Moreover, an analysis of the recidivism rates of the two groups using survival analysis found no differences. Unfortunately, we did not survey the Courts and Crowns as to why certain offenders were subjected to custody in addition to RR. Thus, the answer to the question remains unresolved.

Notes: For RR group, n = 78; for RR + Custody, n = 18.

4. Reducing Recidivism

Evaluation Strategy. Typically, the impact of a program on recidivism is measured after participants have completed the program. However, an average probation sentence of 28.5 months for the RR offenders meant that over two years would lapse before significant numbers of offenders would have completed RR supervision. The first offender graduated from RR in 1995 and from the beginning of the program until May 9, 1997 only 15 offenders had graduated from the program. We therefore chose in-program rather than post-program recidivism as our outcome measure.

Recidivism data were drawn from custody records and a probation database provided by the Manitoba Department of Justice (Corrections Division). These computerized files had a number of limitations. For example, the custody records failed to capture the offender who was convicted of a new crime resulting in a fine. The probation database did not record the dates for in-program probation failures. Of course, being provincial records, offenders committing new offences outside of Manitoba would be missed. Therefore, it was important to examine the consistency of the results across our various measures of recidivism.

Two measures of recidivism were developed. The first measure, file named CONVICT, was taken directly from the custody records and it was defined as any new conviction resulting in a custodial disposition. The second measure was drawn from both the custody records and the probation database. This measure was called VIOLATION. It included new arrests and/or convictions resulting in custody or a violation of the conditions of supervision. The analysis of the recidivism rates of the various groups used a number of different follow-up periods. Most of the results reported here will focus on follow-up periods of one year and 18 months. For the RR offenders, the time at risk begins from the date of being placed on the program by the Courts. 94 RR offenders were at risk for at least one year (see Figure 1) and the numbers drop off as the length of follow-up increases. For the 18 RR offenders who received a custody disposition prior to program placement, the time spent in prison was removed from the calculation of the time at risk. For the six offenders serving intermittent sentences, their sentence was removed as a block of time at the beginning of the follow-up.

Simply reporting the recidivism of a group of offenders is insufficient to evaluate the value of a program. It is important to compare the recidivism of program participants to similar offenders who did not have the benefit of the program. Random assignment of subjects to groups, the preferred method for equating groups, was not possible. Therefore, we attempted to match the comparison groups to the RR offenders on certain key variables.

The first comparison group consisted of 70 male inmates detained in Headingly and Milner institutions in March, 1996. In selecting the inmate comparison sample an attempt was made to match the inmates according to the RR sentence recommendation criteria, which was nine months at the time. Efforts were also made to choose inmates with prior probation breaches and incarcerations (the secondary selection criteria for RR). The difficulty with the prison comparison group was that it only included male offenders. Therefore, only the 75 male RR offenders could be used when comparing them to the inmates. In addition, because half of the RR clients were first offenders we could not identify enough inmates to match offenders according to prior custody.

The second and third comparison groups consisted of probationers (male and females). These groups were drawn from a large database of 1,062 Manitoba probationers (Bonta et al., 1994). Comparing RR clients to probationers allowed an analysis of the value of adding restorative justice to a traditional community sanction (i.e., probation). The probation comparison groups also controlled for the effects of being supervised and the increased likelihood of detecting illegal activity. For the inmates, many were released without any further correctional monitoring and control.

The second comparison group of 94 probationers was matched with the RR clients on the following six risk factors: 1) gender, 2) race (Aboriginal/Metis), 3) age (+/- 4 years), 4) risk-needs classification (minimum, medium and high), 5) violent offence, and 6) first offence. Complete matching was successful for gender, violent offence and risk-needs classification, 97.5% successful for race, 94% for age and 90% for first offence.

A third comparison group consisted of 83 probationers who all had either restitution or community service as a condition of probation. These probationers were matched to 83 RR clients on two risk factors: gender and risk-needs classification level. The size of the RR group was reduced to 83 because 11 offenders had no restitution, community service or victim contact evident in their plans. All offenders were successfully matched on gender and 94% were matched on risk-needs level. A comparison of RR clients to this third comparison group assessed whether the services of RR contributed to reductions of recidivism beyond what a traditional community sanction with some restorative justice features can offer.

Recidivism: RR vs Inmates. Over the course of the evaluation period, the sentence recommendation criteria for RR changed. However, the inmate comparison group was chosen using a minimum sentence length of nine months. For this reason, it was important to examine the equivalence of the RR clients and the prison inmates. Differences between the two groups could possibly account for any observed differences in recidivism rates. A comparison of the two groups is shown in Table 6.

The prison sample and RR clients were very similar. In terms of risk-needs classification, the two groups did not differ and they were similar on many personal-demographic and criminal history variables. There were however, significant differences in grade level and history of incarceration with the inmates showing higher risk. A history of incarceration is one factor that has been particularly linked to recidivism (Gendreau, Little & Goggin, 1996).

Note: p = probability level; ns = nonsignificant

The recidivism outcome for the RR offenders and the inmates are shown in Table 7A. Although the RR male offenders appeared to have lower recidivism rates on the variable CONVICT than the inmates (6.7% vs 14.9%), the differences were not statistically significant. However, at two years, statistically significant differences emerged (11.5% for RR and 33.3% for the inmates; c ² = 3.84, P < .05). Using the measure VIOLATION, which is perhaps more sensitive to problematic behaviour, the RR grouped had significantly lower recidivism rates (c ² = 4.56, P < .05).

Notes: Rest = Restitution; CS = Community Service.

VIOLATION = arrest/conviction with custody/technical breach at 18 mos.

CONVICT = conviction resulting in custody at 12 mos.

The lower recidivism rate for the RR group compared to the prison sample is not only very encouraging but it also suggests future program developments. The similarities of the inmates to the RR clients in terms of personal-demographics and criminal history show that there does exist a potential incarcerated client pool for RR supervision. In light of the difficulties that RR experienced in receiving offenders to the program prior to sentencing, targeting an incarcerated group with the goal of accelerated release may supplement this shortfall.

Recidivism: RR vs Probationers. There were two probation comparison groups. One group was matched to RR clients along six factors. The second group consisted of probationers who had either a condition to pay restitution or provide a community service. As already reported, the matching was quite successful. Table 8 compares the RR offenders to the probationers along some additional characteristics. The three groups were highly similar and only a few differences were found. RR offenders, on average, completed a higher grade in school than the probationers in group A (t = 7.81, p < .001) but they were more likely to have more prior incarcerations (t = 2.49, p < .05) and probation breaches (t = 2.51, p < .05). More importantly however, the RR group scored higher than the two probation groups on the Manitoba risk-needs classification instrument (p < .05).

The recidivism of the RR clients and the two samples of probationers are shown in Table 7A and 7B. For the variable CONVICT, the recidivism rates for the probationers was based upon one year post-program recidivism. The variable VIOLATION was based upon in-program failures while on probation. As both groups were under community supervision, the variable VIOLATION may be the most appropriate measure to use in comparing the recidivism rates of RR offenders and probationers. An 18 month follow-up is used because it represented the average time at risk for failure in the probation database.

Regardless of the measure used, the recidivism rate was significantly lower for the RR clients. When comparing the RR offenders to the probationers who had either a restitution or community service order (see Table 7B), RR clients still showed significantly lower recidivism rates (c ² = 24.98, p < .001 for VIOLATION and c ² = 7.93, p < .01 for CONVICT). In general, it appears that the services provided by RR are associated with a reduction in recidivism when compared to traditional probation services.

The results from the recidivism analysis clearly showed offenders supervised by RR with lower recidivism rates compared to offenders exposed to traditional correctional supervision. We recognize the fact that each comparison group was not drawn from random sampling procedures and each had methodological shortcomings. Nevertheless, the consistency of the findings supports the effectiveness of RR in managing offenders in the community. When we situate the present findings within the general literature on the effectiveness of restorative justice programs, the RR program compares favourably. In our meta-analysis of the literature (Appendix A), the average phi coefficient was .08. Calculation of phi coefficients for recidivism in the present evaluation ranged from .13 (with the inmate comparison) to .22 (with the probation group matched on restitution and community service).

Note: A. Matched on gender, age, race, risk level, first offence, and most serious offence (n = 94).

B. Ordered to pay restitution or participate in community service order. Matched on gender, and risk level (n = 83).

Summary and Conclusions

At the beginning of the "Results and Discussion" section of this report, we remarked that restorative justice programs have multiple goals. The goals include providing restorative justice, functioning as a true alternative to incarceration and enhancing public safety by reducing offender recidivism. We measured RR in achieving these goals and conclude that RR was generally successful.

Compared to regular probationers, RR clients were more likely to make restitution to their victims and to have some form of community service in their supervision plans. Unlike traditional probation, RR did arrange victim-offender meetings although they were not as frequently conducted as hoped. In addition, letters of apologies from the offenders were sent to victims. Considering these findings, RR was generally successful in adhering to the principles of restorative justice.

A notable finding was that RR successfully targeted offenders who were likely to be sent to prison if it were not for the program. Setting aside those who received a custody disposition in addition to RR, RR was a true alternative to incarceration for 81 offenders. The absolute number of offenders diverted may seem low considering the three and one-half year time period covered in the evaluation. However, the low numbers could not be traced to a failure of RR to provide a service. Much depended upon the cooperation and support of the Crowns and the Courts. They appeared to be the key to controlling the flow of offenders to the program. The results of the present evaluation may serve to further educate criminal justice practitioners on the value of RR as an effective alternative to incarceration. The identification of an inmate population for diversion to RR may also serve as an additional source of referral.

Finally, in all but one analysis, RR offenders demonstrated significantly lower recidivism rates than the comparison groups. The calculation of phi coefficients, which ranged from .13 to .22, provide a simple estimate of the reduction of recidivism associated with RR. Namely, a 13% to 22% reduction of recidivism can be expected. This reduction is higher than typically found in restorative justice programs. We suspect that the offender treatment services offered through RR may account for the more positive findings.

We summarize the major findings as follows:

References
(asterisk studies used in meta-analysis)

Andrews, D. A., & Bonta, J. (1998). The psychology of criminal conduct (2^nd ed.). Cincinnati, OH: Anderson.

Andrews, D. A., Zinger, I., Hoge, R. D., Bonta, J., Gendreau, P., & Cullen, F. T. (1990). Does correctional treatment work? A psychologically informed meta-analysis. Criminology, 28, 369-404.

Bazemore, G. (1996). Three paradigms for juvenile justice. In Galaway & J. Hudson (Eds.), Restorative justice: International perspectives (pp. 38-67). Monsey, NY: Criminal Justice Press.

Bonta, J., Andrews, D. A., & Motiuk, L. L. Dynamic risk assessment and effective treatment. Paper presented at the annual meeting of the American Society of Criminology, Phoenix, October 28, 1993.

* Bonta, J. L., Boyle, J., Motiuk, L. L., & Sonnichsen, P. (1983). Restitution in correctional half-way houses: Victim satisfaction, attitudes, and recidivism. Canadian Journal of Corrections, 20, 140-152.

Bonta, J., & Gray, M. (1996). The Restorative Resolutions project: Phase 2 evaluation. Report available from Solicitor General Canada, Ottawa.

Bonta, J., Law, M., & Hanson, K. (1998). The prediction of criminal and violent recidivism among mentally disordered offenders: A meta-analysis. Psychological Bulletin, 123, 123-142.

Bonta, J., Parkinson, R., & Barkwell, L. (1994). Revising the Wisconsin classification system. Paper presented at the Annual Meeting of the American Society of Criminology, Miami, Florida, November 9-12, 1994.

Braithwaite, J. (1989). Crime, shame and reintegration. New York, NY: Cambridge University Press.

* Butts, J. A. & Snyder, H. N. (1992). Restitution and Juvenile Recidivism (NCJRS No. 137774). Washington, DC: U. S. Department of Justice.

* Cannon, A. & Stanford, R. M. (1981). Evaluation of the juvenile alternative services project (NCJRS No. 080633). Tallahassee, FL: Children, Youth and Families Program Office.

Crowe, A. H. (1998). Restorative justice and offender rehabilitation: A meeting of the minds. Perspectives: Journal of the American Probation and Parole Association, 22, 28-41.

Church Council on Justice and Corrections (1996). Satisfying justice. Safe community options that attempt to repair harm from crime and reduce the use or length of imprisonment. Ottawa, Ontario: Church Council on Justice and Corrections.

Gehm, J. R. (1990) Mediated victim-offender restitution agreements: An exploratory analysis of the factors related to victim participation. In B. Galaway & J. Hudson (Eds.), Criminal Justice, Restitution and Reconciliation (pp. 177-182). Monsey, NY: Criminal Justice Press.

Gendreau, P., & Goggin, C. (1996). Principles of effective correctional programming. Forum on Corrections Research, 8, 38-41.

Gendreau, P., Little, T., & Goggin, C. (1996). A meta-analysis of the predictors of adult offender recidivism: What works! Criminology, 34, 575-607.

* Griffith, W. R. (1983). The effect of Washington, D. C.'s restitution program on the recidivism rates of the disadvantaged serious offender (NCJRS No. 098581). Eugene, OR: Institute of Policy Analysis.

* Heinz, J., Galaway, B. & Hudson, J. (1976). Restitution or parole: A follow-up study of adult offenders. Social Service Review, 50, 148-156.

Hudson, J. (1992). A review of research dealing with the views on financial restitution. In H. Messmer & H.-U. Otto (Eds.), Restorative justice on trial (pp. 239-276). Dordrecht, Netherlands: Kluwer Academic Publishers.

Hudson, J., & Galaway, B. (1996). Introduction. In B. Galaway & J. Hudson & (Eds.), Restorative justice: International perspectives (pp. 1-14). Monsey, NY: Criminal Justice Press.

* Kruissink, M. (1990). The halt program: diversion of juvenile vandals. Dutch Penal Law and Policy: Notes on Criminological Research, 8p.

LaPrairie, C. (1998). The "new" justice: Some implications for Aboriginal communities. Canadian Journal of Criminology, 40, 61-79.

* Levi, K. (1982). Relative redemption: Labeling in juvenile restitution. Juvenile & Family Court Journal, 33, 3-13.

McCold, P. (1997). Restorative justice: An annotated bibliography 1997. Monsey, NY: Criminal Justice Press.

Messmer, H., & Otto, H.-U. (1992). Restorative justice: Steps on the way to a good idea. In H. Messmer & H.-U. Otto (Eds.), Restorative justice on trial (pp. 1-12). Dordrecht, Netherlands: Kluwer Academic Publishers.

Nuffield, J. (1997). Diversion programs for adults. User Report #1997-05. Ottawa: Solicitor General Canada.

* Pearson, F. P. (1988). Evaluation of New Jersey's intensive supervision program. Crime and Delinquency, 34, 437-448.

Richardson, G., & Galaway, B. (1995). Evaluation of the Restorative Resolutions project of the John Howard Society of Manitoba. Report available from Ministry of the Solicitor General of Canada, Ottawa.

* Roy, S. (1993). Two types of juvenile restitution program in two midwestern counties: A comparative study. Federal Probation, 57, 48-53.

* Shichor, D. & Binder, A. (1982). Community restitution for juveniles: An approach and preliminary evaluations. Criminal Justice Review, 7, 46-50.

* Schneider, A. L. (1986). Restitution and recidivism rates of juvenile offenders: Results from four experimental studies. Criminology, 24, 533-552.

Umbreit, M. (1986). Victim/offender mediation: A national survey Federal Probation, 50, 53-56.

Umbreit, M. (1990). The meaning of fairness to burglary victims. In B. Galaway & J. Hudson (Eds.), Criminal justice, restitution, and reconciliation. Monsey, NY: Criminal Justice Press.

* Umbreit, M. S. (1994). Victim meets offender: The impact of restorative justice and mediation. Monsey, NY: Willow Tree Press.

* Umbreit, M. S. & Coates, R. B. (1992). Victim offender mediation: An analysis of programs in four states of the U.S. Minneapolis, MN: Minnesota Citizens Council on Crime and Justice.

Umbreit, M. & Zehr, H. (1996). Family group conferences: A challenge to victim offender mediation? Victim Offenders Mediation Association Quarterly, 7, 4-8.

Walgrave, L. (1992). Mediation and community service as models of a restorative approach: Why would it be better? Explicating the objectives as criteria for evaluation. In H. Messmer & H.-U. Otto (Eds.), Restorative justice on trial (pp. 343-353). Dordrecht, Netherlands: Kluwer Academic Publishers.

* Wiebush, R. G. (1985). Recidivism in the juvenile diversion project of the young volunteers in action program, (NCJRS No. 102325). Columbus, OH: U. S. Department of Justice.

* Wiebush, R. G. (1993). Juvenile intensive supervision: The impact on felony offenders diverted from institutional placement. Crime and Delinquency, 39, 68-89.

Appendix A

Restorative Justice (RJ) and Recidivism: Meta-Analytic Findings

Notes: Rest = Restitution; CS = Community Service

Negative sign indicates increases in recidivism.