At the time of Strang’s 2001 review, restorative justice was a relatively new alternative to traditional criminal justice responses and as a result, few evaluations of restorative justice programs had been conducted. In the intervening years, the number of evaluations (both process and outcome) has grown markedly. Beven et al. 2011 note that a key criticism of restorative justice has been that it fails to deliver the much talked about restoration of victims and offenders. However, proponents suggest that this criticism is based on a misinterpretation of the goals of restorative justice; that it is unrealistic to think in terms of ‘undoing’ the crime, but that the impact of restorative justice can be measured through more specific goals (Beven et al. 2011).
The question of whether restorative justice ’works’ has often been answered with reference to the traditional criminal justice for which it is an alternative and as with any intervention in the criminal justice field, the effectiveness of restorative justice has often been based on its impact on reoffending. In recognition of the aim of restorative justice to be more inclusive of victims, to encourage the offender to take responsibility for their crime(s) and to repair the harm caused by the crime, evaluations have also assessed participant (particularly that of victims) satisfaction. This section summarises the results of evaluations concerning these two areas of interest—reoffending and victim satisfaction—and looks briefly at other aspects of restorative justice programs that have been evaluated.
A note on methodological issues
An assessment of restorative justice outcomes needs to be mindful of the variation in eligibility and processes implemented across the jurisdictions (as set out in section two of this report), and also to take into consideration the different methodological approaches to analysis and the substantial variation in the quality of studies. This is particularly important for studies comparing restorative justice outcomes with those of traditional criminal justice processes. Where the methodologies used differ significantly, it is not possible to compare the results of one study with another. Following a review of research comparing court with restorative justice Weatherburn, McGrath and Bartels (2012: 799) noted that the findings of many studies were compromised by ‘small sample size, limited controls for selection bias, selective attrition, ambiguous comparison groups, and conclusions unwarranted by the evidence presented’.
One key methodological issue that is often not adequately addressed is that of selection bias. Fewer restorative justice options are available for offenders who commit serious offences and who have a history of offending, with the exception of VOMs taking place at the post-sentence stage. This can lead to a selection bias, as low rates of reoffending are more likely to be found among offenders eligible for restorative justice than the more serious offenders for whom it is less likely to be an option. Randomised controlled trials (RCTs) are considered the most rigorous method of addressing selection bias. They involve the random allocation of eligible offenders and cases to two conditions—a restorative justice group or a control group. Weatherburn and Macadam (2013: 3) note that such an approach is ‘rare’ and that a widely used alternative is control through statistical techniques. Many studies control for relevant factors through regression analysis, or more recently, propensity score matching, which involves matching those who receive a treatment (in this context, those who participate in a restorative justice process) with ‘untreated individuals’ (Poynton 2013: 13), that is offenders who are processed through the traditional criminal justice system. While it is acknowledged that ‘propensity score matching is not a substitute for a randomised controlled trial…[it can have] a distinct advantage’ in producing matched control and treatment groups in some contexts (Smith & Weatherburn 2012: 6).
A further issue that is not adequately addressed in many evaluations is that of self-selection bias, arising when those deemed eligible and offered restorative justice, or ‘treatment’, choose not to proceed (Bergseth & Bouffard 2007). This is often dealt with by employing an ‘intention to treat’ design, where results are analysed based on treatment assigned rather than treatment actually delivered and experienced (Bergseth & Bouffard 2007). This is important from a policy perspective; if the program being tested becomes a policy, it is essential to take into account the reality that it will not always be delivered as intended and to calculate its benefit based on the actual take-up rate not an assumed one.
Impact on reoffending
While reducing reoffending is not the only goal of restorative justice, it is critical to be confident that restorative justice does not lead to increases in reoffending. Reducing recidivism is anticipated as an outcome due to the engagement of informal social controls through the inclusion of family, supporters and community representatives (Daly 2001) and the impact of meeting one’s victim face to face (Strang 2002). Informal social control is widely believed to influence offending. Hirschi’s (1969) social control theory was grounded in the belief that bonds with prosocial values, people and institutions prevent people from engaging in criminal behaviour. The social bonds required to prevent rule-breaking are achieved through four elements—commitments, attachments, involvements and beliefs. In essence, the poor opinion that friends and families may have of an individual’s deviant behaviour has the effect of inhibiting rule-breaking. Once offending has taken place, Braithwaite (1989) posits that reintegrative shaming inhibits further offending. Restorative justice processes encourage reconnection with prosocial bonds and one indication of the success of this through reduced reoffending.
Restorative justice processes are also underpinned by reintegrative shaming theory. Conferencing, one of the more common forms of restorative justice, is thought to be more effective than court processes in reducing reoffending due to the different stigmatising effects of each. That is, stigmatising of offenders in traditional processes serves only to reinforce their deviant behaviour, whereas conferences stigmatise the behaviour and not the individual (Smith & Weatherburn 2012), the importance of which is set out in Braithwaite’s theory of reintegrative shaming (1989). As evident in the brief summaries of evaluations of the various restorative justice programs reviewed in section two, evidence of the effectiveness of restorative justice is mixed. A meta-analysis of 22 studies examining the effectiveness of 35 individual restorative justice programs found that restorative justice was more effective than traditional criminal justice approaches, leading to reduced reoffending (Latimer, Dowden & Muise 2005). Similar findings were reported by Sherman and Strang (2007) in a review of research comparing restorative justice outcomes with those from conventional processes. They reported ‘substantial reductions in repeat offending for both violence and property crime’ that restorative justice was found to be more effective with more serious offences and for crimes involving personal victims (Sherman & Strang 2007: 8).
To date, most studies have focused on the impact of restorative justice on reoffending among juvenile offenders due largely to the widespread use of restorative justice with this demographic. Smith and Weatherburn (2012) reviewed a range of studies that compared the impact of reoffending between those who undertook youth conferencing and those who only appeared in Children’s Court, as well as studies that compared conferencing and other measures such as cautions, mediation and orders to pay restitution. Authors of the review concluded that there was ‘little basis for the confidence that conferencing reduces re-offending at all’ (Smith & Weatherburn 2012: 6) and cited several methodological problems in studies on conferencing, such as a failure to adjust for differences between control and treatment groups, small sample sizes and restricted definitions of reoffending, among others. Consequently, Smith and Weatherburn (2012) revisited the question of whether conferencing was more effective in reducing reoffending than court using propensity score matching (see summary above). Their comparison of reoffending between young people participating in Youth Justice conferences in New South Wales and those eligible for a conference but processed through the Children’s Court in 2007 found no significant differences between the two groups in the proportion who reoffended, the length of time to first (proven) reoffence, the level of seriousness of reoffending, or the number of proven offences. This led to the conclusion that conferencing for young offenders in New South Wales ‘is no more effective than the NSW Children’s Court in reducing juvenile offending among young person’s eligible for a conference’ (Smith & Weatherburn 2012: 1).
Similar findings have been reported in samples of young adults. A recent BOCSAR study examined the likelihood of reoffending among two groups—young adults who participated in Forum Sentencing in New South Wales and young adults who were eligible but sentenced in conventional courts (Jones 2009). The study found no evidence that Forum Sentencing had an impact on the key areas of concern; that is, reducing likelihood of reoffending, increasing the time taken to reoffend, reducing the frequency of subsequent offending and reducing the seriousness of subsequent offending. One possible explanation cited was the large number of offenders having committed ‘victimless’ crimes (such as motor vehicle regulatory offences). However, even after removing these offences from the analysis, there was no effect of Forum Sentencing. Further, the study did not apply appropriate restrictions on the control group to match this with the intervention group (eg eligibility for the control group was not restricted to defendants facing possible imprisonment) and did not reduce selection bias by conducting analyses on the basis of ‘intention-to-treat’. These design issues were addressed in a subsequent evaluation of Forum Sentencing, which also used propensity score matching to minimise selection bias (Poynton 2013). That evaluation also found no evidence that participation in Forum Sentencing reduces reoffending, although the findings should be interpreted cautiously given the short (12 month) follow-up period. Further, as the author notes (and as is discussed in section two of this report), reducing reoffending is one of six objectives of the program and it may be that Forum Sentencing results in other benefits while not performing worse than court on reoffending.
While RCTs in this field are indeed rare, there are several notable exceptions that add significantly to the evidence on the impact of restorative justice on offending. In Australia, the RISE project, an independent evaluation of the effectiveness of diversionary conferencing for victims and offenders, was undertaken by the Australian National University in partnership with ACT Policing between 1995 and 2000. Four experiments sought to test the impact of restorative justice in the Australian Capital Territory through the random assignment of offenders to either be dealt with through the courts or a diversionary conference for:
- violent offenders under 30 years;
- juvenile offenders charged with shoplifting from department stores;
- juvenile offenders who committed property crime against personal victims; and
- drink driving offences committed by both young and adult offenders.
The studies found that both victims and offenders reported that conferences were fairer than court proceedings and that there were greater benefits for victims who attended conferences (including feeling less fearful and having their sense of security restored (Strang et al. 2011). Reoffending was statistically significantly lower among offenders who were conferenced following a violent offence than among offenders who were processed through the courts (Sherman, Strang & Woods 2000). Offending rates among this sample dropped by 38 crimes per 100 offenders per year. No differences were found between the court and diversionary conference groups in the experiments involving juvenile shoplifters and juvenile property offenders, and the fourth experiment involving drink-driving offences resulted in a small increase (6 crimes per 100 offenders per year) in offending (Sherman, Strang & Woods 2000).
A further seven RCTs were conducted by the Justice Research Consortium (JRC) in the United Kingdom to assess the effectiveness of a scheme funded by the Home Office. These included the following:
- Two RCTs in London with adult offenders, one involved offences of burglary of a dwelling and the other involved street crime offences.
- Three RCTs in Northumbria. One involved cases with an identifiable victim for adult offenders at the pre-sentence stage, a second involved an identifiable victim of a property offence committed by young offenders and the third involved cases with an identifiable victim of a violent offence committed by young offenders.
- Two RCTs in the Thames Valley which both involved adult offenders who had committed violent offences.
Across the seven RCTs, offenders who participated in restorative justice were found to commit significantly fewer offences in the two year period following the conference than were offenders who were assigned to a control group (Shapland et al. 2008). Neither the likelihood of reconviction, nor the severity of reconviction, differed significantly between the two groups. With the exception of adult property offenders in Northumbria dealt with in the magistrates’ courts, no single RCT resulted in a statistically significant impact on reoffending (Shapland et al. 2008). More recently, Strang et al. (2013) completed a systematic review of the effect of face-to-face restorative justice conferencing on reoffending and victim outcomes, such as victim satisfaction, material restoration and emotional restoration. This review encompassed two of the three Canberra RISE RCTs, seven UK experiments and one RCT in the United States, and shows that across all 10 RCTs, restorative justice performed better than court in reducing reoffending (Strang et al. 2013).
A randomised controlled trial of a family conferencing program in Indianapolis, known as the Indianapolis Experiment, was conducted in 1997 (McGarrell et al. 2000) and research in 2007 published findings relating to reoffending among this group following a 24 month period (McGarrell & Hipple 2007). The original study involved the random allocation of first-time young offenders to either a control group (ie juvenile court) or a treatment group (ie family group conferencing). Of the total 782 youths who participated, 400 were allocated to the family conference group and 382 to the control group. The outcome measures were the rate and timing of reoffending, and the incidence of reoffending among the two groups. Results indicated that participation in Family Group Conferencing was related to a decreased incidence of reoffending and a longer period to the first reoffence. A slightly greater proportion of young offenders in the treatment group ‘survived’; that is, they were not arrested for a new offence during the 24 month period, (52%, n=207), than did those in the control group (46%, n=176). There was significant difference in the survival rates of the two groups, with the control group failing at a faster rate in weeks 14 to 32 (McGarrell & Hipple 2007). Young offenders in the treatment group had an average of 1.29 rearrests in the 24 months following treatment, compared with an average of 1.67 rearrests among the control group.
Further, several studies have examined factors associated with reoffending, such as offence types and offending histories, among participants in restorative justice programs. In assessing data from RISE, which utilised RCTs, Sherman, Strang and Woods (2000) found that rates of reoffending were lower for violent offenders following conferencing rather than court processing, but not for property offenders.
Assessments of the impact of restorative justice conferencing on reoffending among youths in southeast Queensland found that reoffending was less likely among offenders for whom conferencing was the first intervention compared with young offenders whose first intervention involved a caution or court (Hayes & Daly 2004). In another study examining conference-specific factors, offender characteristics and reoffending, Hayes and Daly (2003) reported that young offenders who experienced remorse and whose outcomes were reached by consensus were also less likely to reoffend.
While the evidence is not overwhelming at present, there is a growing body of evidence that supports the assertion that restorative justice can reduce reoffending, however, more attention needs to be paid to the results of rigorous studies in order to state conclusively that it is ‘a less expensive and more efficient way’ (Weatherburn & Macadam 2013: 1) of addressing offending.
Impact on victim satisfaction
Historically, restorative justice was once widely used as a response to wrongdoing. As crimes committed against an individual became crimes against the ‘King’s peace’ and later, against the state, the role of the victim was eroded (Strang & Sherman 2003). The reintroduction of restorative justice practices alongside the adversarial system is seen as redressing the balance. Although victims are considered to play a central role in restorative justice processes, this has been disputed (Richards 2009). Commentators have noted a range of indicators (low number of victims attending conferences, use of restorative justice for ‘victimless’ crimes such as graffiti, eligibility criteria being based on offenders) that restorative justice is primarily about reforming offenders than repairing the harm caused to victims (Richards 2009).
In developing the NSW Community Conferencing for Young Adults Pilot, the necessity of involving victims in the program was a contentious point during the development phase (Hart & Pirc 2012). Although it was considered that forums should only proceed if a victim(s) chose to participate, it was disregarded as it seemed to be ‘usurping the role of the court’ (Hart & Pirc 2012: 75) and to be unfair to the offender as willing offenders would then be excluded. The decision was taken that a victim would be consulted in the early stages and their wishes considered but that forums would go ahead without their involvement if need be.
The benefits of restorative justice are not restricted to offenders alone. In an analysis of what victims need from restorative justice, Strang (2002) found that victims in the traditional criminal justice system commonly experience:
- a lack of attention to ‘non-material dimensions of victimisation’, for example, anger, fear and mistrust;
- no focus on repairing the injury caused by crime;
- failure of the criminal justice system to clearly communicate with victims regarding the status of the case;
- failure to provide victims with a legitimate and active role when dealing with offences committed against them; and
- perceptions of a lack of procedural fairness and dissatisfaction with outcomes due largely to having been excluded from the decision-making process.
It follows that given the widespread dissatisfaction with traditional responses, an important indication of the effectiveness of restorative justice practices when it comes to repairing the harm caused by crime are the responses of victims themselves. This is primarily examined through measuring victim satisfaction.
While the impact of restorative justice programs on reoffending is mixed, evidence from a range of studies indicates positive outcomes for victims who participate in face-to-face restorative justice programs in Australia and overseas (Hayes 2005; Latimer, Dowden & Muise 2005; Strang et al. 2006). Studies also consistently reported that victims felt they had been treated fairly and with respect (Hayes 2005; Shapland et al. 2007; Strang 2002).
In light of Hart and Pirc’s (2012) comments regarding the role of victims in forums in New South Wales and the incorporation of restorative justice into existing criminal justice processes, the question must be asked—is placing the victim at the centre of the process reasonable and feasible? An entirely different conceptualisation of ‘justice’ in the Western world is required to achieve this. As stated in section one of this report, restorative justice departs from the conventional criminal justice system in conceptualising crime as conflict between individuals, which leads to harm to victims and communities rather than a violation of law to be punished by the state. As a result, the role of the victim in restorative justice is a crucial one. In addition to being integral in the restoration of harm caused, being the party who has been harmed, ‘victims turn out to be exceptionally important in requiring offenders to understand and take responsibility for the harm they have caused’ (Strang 2004: 77). Positive outcomes reported from the ACT Scheme, which is victim-centric and designed to meet the needs of victims by empowering them to have a voice in how an offence has affected them, suggest that it is entirely feasible to place the victim at the centre of the process. Shapland et al. (2007) found that a larger proportion of victims who participated in restorative justice processes across the seven RCTs in the JRC scheme in the United Kingdom reported being satisfied than did victims who were in the control groups (72% cf 60%). Similar proportions of restorative justice victims and control group victims reported feeling the process was fair and significantly more restorative justice victims reported feeling more secure than did victims in the control groups (Shapland et al. 2007). The Indianapolis Experiment also reported increased satisfaction among victims who participated in conferences and both victims and offenders were more likely to recommend conferencing than were those who were allocated to the control group (McGarrell et al. 2000).
Importantly, Weatherburn and Macadam (2013) identify the need for research to examine whether victim’s who report satisfaction following participation in restorative justice processes remain satisfied if an offender fails to complete the agreed undertakings.
While research has commonly examined victim satisfaction with restorative justice, it has generally overlooked their other experiences, although some have measured the impact on feelings of anger or fear (McGarrell & Hipple 2007; Shapland et al. 2007; Strang 2002) and perceptions of safety (Strang 2002). Although reparation is considered ‘a defining characteristic of restorative justice and a key benefit for victims of crime’ it is not necessarily relevant for all victims (Stubbs 2007: 171). Although the findings should be interpreted with caution due to the small sample sizes involved, Beven et al. (2011) found that victims who participated in conferences were more likely than those whose cases were dealt with in the traditional court process to have higher perceptions of safety and in fact, on average, these perceptions of safety were higher than pre-offence levels. Beven et al. (2011) also found that victim satisfaction was not related to satisfaction with the outcome of the process; that is, a victim’s overall satisfaction level was not affected by whether an outcome was lenient or harsh. The studies reviewed by Sherman and Strang (2007) consistently reported benefits to victims in face-to-face conferences, including reduced post-traumatic stress which led to improved mental health (Angel 2005; Angel et al. forthcoming).
Studies have also examined other aspects of restorative justice, such as offender satisfaction and compliance with restitution. For example, Latimer, Dowden and Muise (2005) found that offenders who participated in restorative justice practices were more satisfied than offenders who were processed through courts. A recent study that re-examined data from the ACT RISE experiments reported that restorative justice conferences did not have a significant impact on the perception of future offending among juveniles, although participation in a conference influenced positive perceptions on repaying the victim and society and on feeling repentant (Kim & Gerber 2011).
A small number of studies have examined the impact of participation in restorative justice practices on compliance with outcomes, primarily on restitution compliance (Latimer, Dowden & Muise 2005). A meta-analysis of these studies found substantially higher rates of compliance among offenders following restorative justice processes than those processed through traditional responses. Similar findings were reported by Sherman and Strang (2007) in their assessment of the effectiveness of restorative justice, which found that offenders in restorative justice were more likely to comply with outcomes than offenders with court-ordered outcomes. The same review found that the desire among victims for revenge against offenders was lower among restorative justice participants compared with those experiencing formal responses and could lead to less violence in communities (Sherman & Strang 2007).
Others have considered the cost-effectiveness of restorative justice processes. A recent BOCSAR report that found that overall, Youth Justice Conferences were more cost-effective, costing 18 percent less per person, than the Children’s Court in New South Wales (Webber 2012). More specifically, court-referred youth conferences cost four percent per person less and police-referred youth conferences cost 45 percent per person less than processing young offenders through the NSW Children’s Court. Similarly, KPMG’s (2010) evaluation of youth conferencing in Victoria found that the cost of referring a youth to conferencing was lower than the cost of administering a Community Based Order. Cost-effectiveness is an aspect requiring further research. In a tighter fiscal environment, the introduction or continuation of restorative justice programs (and similarly, other alternative approaches) will be significantly determined by their ability to deliver positive outcomes in a cost-effective manner.
Sherman and Strang (2011: 14) note the value for policymakers in assessments of effectiveness that consider the reduced costs of formal criminal justice options as a result of restorative justice and ‘relative to the cost of providing [restorative justice conferences]’. That is, if restorative justice leads to reduced costs for policing, court and sentencing administration, prisons and community corrections but is relatively expensive to run, then it is not a cost-effective option. In an assessment of the seven UK RCTs, Shapland et al. (2008) compared the costs of crime prevented with the costs incurred in delivering restorative justice conferences, excluding costs associated with first establishing a restorative justice scheme. The seven experiments produced a benefit associated with decreased reconviction when compared with the running costs. At an aggregate level, the cost of running restorative justice conferences in London was £598,848 compared with £8,261,028 in the costs of crime prevented. Across the Northumbria RCTs, the costs of crime prevented (£320,125) were again greater than the running costs (£275,411). This was also the case in the Thames Valley RCTs where costs of crime prevented (£461,455) were greater than the running costs (£222,463).
Finally, gaps have also been identified in knowledge of the impact of remorse in restorative justice. In particular, Weatherburn and Macadam (2013) note the need for further research to examine remorse between offenders who have participated in restorative justice and those who have been processed through the traditional criminal justice system.
The evidence on the impact of restorative justice on reoffending is mixed but a growing body of research suggests positive impacts for both victims and offenders. Research has demonstrated that contrary to popular opinion, restorative justice may be more effective for more prolific offenders; that it has the ability to prevent some offenders from further criminal activity, slows the offending of others (although it leaves some others unaffected); that it is more effective for violent rather than property offences, more effective post- rather than pre-sentence and while there are differing expectations of what the process offers, there is clear evidence that those willing to engage in the process do benefit (Strang 2010).
While the ability of restorative justice to reduce reoffending is still contested, a focus on reoffending outcomes alone fails to capture the extent of other benefits, such as, victim satisfaction, offender responsibility for actions and increased compliance with a range of orders, among others. Studies examining the impact of restorative justice have shown different results for different offences and offenders, as is the case for many other forms of intervention, including processing through the courts, among others. What is critical in terms of reoffending, is to be confident that restorative justice does not lead to increases in reoffending.