restorative justice: the evidence

A review of research on restorative justice around the world, across 36 direct comparisons to conventional criminal justice.

  • By Lawrence W Sherman and Heather Strang
  • Jerry Lee Program of Randomized Trials in Restorative Justice
  • In collaboration with Geoffrey Barnes Sarah Bennett Caroline M Angel
  • Dorothy Newbury-Birch Daniel J Woods Charlotte E Gill
  • Please direct all correspondence to

A note on format

This report follows a modified UK Cabinet Office format of unpacking the content like a website: a one-page abstract, a three-page executive summary, and a detailed introductory section that summarises the findings of the full report. This format enables the users of the document to read – or circulate – either the abstract alone, the abstract plus the three-page executive summary, the abstract plus the introductory overview (section 1) of the full report, or the entire document. This format is designed to provide different levels of detail for different kinds of readers, and to support the different ways in which information is digested for different stages and levels of discussion and decision making.

Published by the Smith Institute ISBN 1 905370 16 4

This report, like all Smith Institute monographs, represents the views of the authors and not those of the Smith Institute.

© The Smith Institute 2007


Preface                                                                                                                                                   3

By Wilf Stevenson, Director, Smith Institute

Abstract                                                                                                                                                  4

About the authors                                                                                                                                   5

Foreword                                                                                                                                                6

By Jerry Lee, President of the Jerry Lee Foundation

Acknowledgements                                                                                                                                7

Executive summary                                                                                                                               8

  1. Introduction and overview                                                                                                                 12

Key findings: tables 1 to 5

  1. A tale of three RJ conferences                                                                                                          26
  2. Varieties of restorative justice                                                                                                          32
  3. The process of restorative justice                                                                                                     36
  4. Restorative justice and the rule of law                                                                                              44
  5. How we know what works – and what doesn’t                                                                                  48
  6. Restorative justice in the UK: the four settings                                                                                52
    1. Adult justice
    1. Youth justice
    1. Schools
    1. Neighbourhood policing
  7. Keeping promises versus following orders: how RJ compares                                                          58
  8. Reducing harm to victims                                                                                                                 62
  9. Reducing repeat offending                                                                                                              68
  10. Rare events: when RJ backfires – and why                                                                                      74
  11. Other effects on crime                                                                                                                     78
  12. Getting cases brought to justice                                                                                                     82
  13. Could RJ reduce the financial costs of justice?                                                                              86
  14. More justice, less crime: a way forward                                                                                          88

Bibliography                                                                                                                                         92


Wilf Stevenson, Director, Smith Institute

The Smith Institute is an independent think tank which has been set up to undertake research and education in issues that flow from the changing relationship between social values and economic imperatives. In recent years the institute has centred its work on the policy implications arising from the interactions of equality, enterprise and equity.

In 2004-05 the Smith Institute ran a highly successful series of seminars looking at case studies of the use of restorative justice techniques among criminals and their victims, in schools and within communities and neighbourhoods. Building on the impressive accounts of how powerful restorative justice techniques could be, as a way both of changing behaviour and of mitigating harm, this independent report was commissioned by the Smith Institute in association with the Esmée Fairbairn Foundation in order to examine the evidence on restorative justice (RJ) from Britain and around the world.

The aim of the project was to bring together the results of RJ trials in order to set out a definitive statement of what constitutes good-quality RJ, as well as to draw conclusions both as to its effectiveness with particular reference to reoffending and as to the role that RJ might play in the future of Britain’s youth and criminal justice systems.

The Smith Institute thanks Sir Charles Pollard, Rob Allen and Professor Mike Hough for their hard work as members of the steering committee convened to commission and oversee the academic rigour of this report.

The Smith Institute gratefully acknowledges the support of the Esmée Fairbairn Foundation towards this publication.


A review of research on restorative justice (RJ) in the UK and abroad shows that across 36 direct comparisons to conventional criminal justice (CJ), RJ has, in at least two tests each:

  • substantially reduced repeat offending for some offenders, but not all;
  • doubled (or more) the offences brought to justice as diversion from CJ;
  • reduced crime victims’ post-traumatic stress symptoms and related costs;
  • provided both victims and offenders with more satisfaction with justice than CJ;
  • reduced crime victims’ desire for violent revenge against their offenders;
  • reduced the costs of criminal justice, when used as diversion from CJ;
  • reduced recidivism more than prison (adults) or as well as prison (youths).

These conclusions are based largely on two forms of restorative justice (RJ): face-to-face meetings among all parties connected to a crime, including victims, offenders, their families and friends, and court-ordered financial restitution. Most of the face-to-face evidence is based on consistent use of police officers trained in the same format for leading RJ discussions. These meetings have been tested in comparison with conventional criminal justice (CJ) without benefit of RJ, at several stages of CJ for violence and theft:

  • as diversion from prosecution altogether (Australia and US);
  • as a pre-sentencing, post-conviction add-on to the sentencing process;
  • as a supplement to a community sentence (probation);
  • as a preparation for release from long-term imprisonment to resettlement;
  • as a form of final warning to young offenders.

Violent crimes

Six rigorous field tests found RJ reduced recidivism after adult or youth violence. Three of these were randomised controlled trials

(RCTs), conducted with youth under 30 in Canberra, females under 18 in Northumbria, and (mostly) males under 14 in Indianapolis. Reasonable comparisons also show effects for adult males in West Yorkshire and the West Midlands, as well as for violent families in Canada.

Property crimes

Five tests of RJ have found reductions in recidivism after property crime. Four were RCTs done with youth: in Northumbria, Georgia, Washington and Indianapolis. Diversion of property offenders to RJ, however, increased arrest rates among a small sample of Aboriginals in Canberra.

Victim benefits

Two RCTs in London show that RJ reduces post-traumatic stress; in four RCTs RJ reduces desire for violent revenge; in four RCTs victims prefer RJ over CJ.

RJ versus prison

In Idaho an RCT of RJ as court-ordered restitution did no worse than short jail sentences for youth. In Canada adults diverted from prison to RJ had lower reconviction rates than a matched sample of inmates.

Offences brought to justice

Five RCTs in New York and Canberra show diversion to RJ yields OBTJ (offences brought to justice) rates 100% to 400% higher than CJ, including for robbery and assault, when offenders take responsibility but need not sign full admission to crime.

A way forward

The evidence on RJ is far more extensive, and positive, than it has been for many other policies that have been rolled out nationally. RJ is ready to be put to far broader use, perhaps under a “Restorative Justice Board” that would prime the pump and overcome procedural obstacles limiting victim access to RJ. Such a board could grow RJ rapidly as an evidence-based policy, testing the general deterrent impact of RJ on crime, and developing the potential benefits of “restorative communities” that try RJ first.

About the authors

Lawrence W Sherman is the Wolfson Professor of Criminology at the University of Cambridge,1 Director of the Jerry Lee Center of Criminology at the University of Pennsylvania, Past President of the American Academy of Political and Social Science, the American Society of Criminology and the International Society of Criminology, and senior author of the 1997 report to the US Congress Preventing Crime: What Works, What Doesn’t, What’s Promising.

Dr Heather Strang is the Director of the Centre for Restorative Justice at the Australian National University, Lecturer in

  1. From Easter Term 2007

Criminology at the University of Pennsylvania, Director of the Canberra RISE project (now in its 12th year), senior author of the Campbell Collaboration Systematic Review of Restorative Justice, Fellow of the Academy of Experimental Criminology and author of the 2002 Oxford Press book, Repair or Revenge: Victims & Restorative Justice.

For more information about the work of the authors on restorative justice, see:


Jerry Lee, President of the Jerry Lee Foundation

It is almost 10 years since the Attorney General of the US, Janet Reno, submitted an independent, scholarly report to the US Congress entitled Preventing Crime: What Works, What Doesn’t, What’s Promising (1997).2 When I read that report shortly after it was published, I immediately sensed a sea change in the way democracies would talk about crime prevention. No longer would we focus just on ideology. Evidence would soon take a much larger role in the debate. Within a year, officials of at least 12 different nations would consider the report and its policy implications, from Seoul to Stockholm, from Wellington to London.

As a consumer of social science research in my business life for over four decades, I brought as sceptical an eye to that report as to any other. I came away extremely impressed with the rigour of the framework, its 1-to-5 scientific methods scale, and its comprehensive search for evidence on what works. The next day, I contacted the senior author of both that report and the current review of evidence on restorative justice, Lawrence Sherman.

Since then, the Jerry Lee Foundation has been pleased to support many reviews and evaluations of crime prevention programmes. By far the most comprehensive of these investments has been in the Jerry Lee Program of Randomized Controlled Trials in Restorative Justice, which began in 1995 with a multimillion- dollar investment by the Australian government. With its expansion into the UK with funding from the Home Office in 2001, the programme has now completed 12 rigorous tests of restorative justice in a wide range of settings and populations.

  2. and

The programme’s co-directors, Lawrence Sherman and Heather Strang, are also undertaking a wider review of the evidence on restorative justice for the International Campbell Collaboration, a peer-reviewed network for the systematic review of the effectiveness of public programmes.3 No one is better qualified than they are to prepare a crisp, readable assessment of the evidence on RJ.

The evidence clearly suggests that RJ is a promising strategy for addressing many of the current problems of the criminal justice system. More important, it is a strategy that has been subjected to rigorous testing, with more tests clearly implied by the results so far. The development of RJ in the UK over the past decade is a model in the evidence-based approach to innovations in public policy. Like the old story of the tortoise and the hare, the evidence on RJ cannot be gathered by rushing ahead. The evidence so far suggests that sure and steady wins the race.

The race for all of us is to reach a world of less crime and more justice. An endless increase in the prison population seems unlikely to achieve those goals. This report points out ways to bring more offences to justice, and perhaps reduce the cost of justice, while reducing the personal cost of crime to victims. No other policy I have seen would put the victim so clearly “at the centre” of a larger community in which we are all interdependent. How and when to use RJ most effectively is a matter that evidence can help decide. With this report, that evidence should now be more accessible to all.


Lawrence W Sherman and Heather Strang

The authors owe thanks to many people and institutions. The report was made possible by Smith Institute Director Wilf Stevenson, with his colleagues Ben Shimshon and Konrad Caulkett, and by financial support from the Esmée Fairbairn Foundation and its Director Dawn Austwick. That foundation also supported our completion of the London tests of restorative justice in crown courts. Much of the other material on which the report is based would not have been possible without grants from the Jerry Lee Foundation, the Smith Richardson Foundation, the Home Office, the Australian Criminology Research Council, the Australian Department of Health, the Australian Department of Transport & Communications, the Australian National University, the US National Institute of Justice, and the Albert M Greenfield Endowment of the University of Pennsylvania. The work these institutions supported is entirely the responsibility of the authors, and is not intended to represent the position of any of these sponsors.

Many people leading operating agencies in Australia, the UK and the US made possible the research summarised here. From the first RJ conference we saw in 1993, led by New South Wales Police Sergeant Terry O’Connell, to the early support of Metropolitan Police Commissioners John Stevens and Ian Blair in mid-2000, research and development in RJ has depended on innovative and visionary police professionals. These include Canberra Police Chief Peter Dawson of the Australian Federal

Police, British Chief Constables Crispian Strachan, Peter Neyroud and Michael Craik, Chief Inspector Jane Simmons and Inspector Brian Dowling, and the many dedicated RJ facilitators they recruited and led. The judiciary provided equal support from Lord Chief Justice Woolf, Lord Justice John Kay, and Judges Shaun Lyons and Shirley Anwyl, who helped recruit many of their colleagues to the Crown Court experiments. Eithne Wallis and Ray Fishbourne led the National Probation Service in support of RJ tests across England, Phil Wheatley led HM Prison Service in its support of RJ under even the most demanding conditions, Lord Norman Warner and Professor Rod Morgan provided steadfast Youth Justice Board support, and Dru Sharpling led London’s valiant test of conditional cautioning.

The Smith Institute’s steering group for this report led by Sir Charles Pollard, with members Rob Allen, Professor Mike Hough and Peter Micklewright, provided clear and timely guidance. Professor Paul Wiles and his colleagues at the Research Development & Statistics Directorate provided invaluable advice and support at many stages of this work, as did Professor Joanna Shapland and her team at Sheffield, and Professor David Farrington and Loraine Gelsthorpe at Cambridge. Finally, the global evidence on RJ would not be where it is today without the normative and scientific scholarship of our very special colleague, Professor John Braithwaite.

Executive summary

Purpose and scope

This is a non-governmental assessment of the evidence on restorative justice in the UK and internationally, carried out by the Jerry Lee Center of Criminology at the University of Pennsylvania for the Smith Institute in London, with funding from the Esmée Fairbairn Foundation. The purpose of this review is to examine what constitutes good-quality restorative justice practice, and to reach conclusions on its effectiveness, with particular reference to reoffending.4

Varieties of restorative justice

The review employs a broad definition of restorative justice (RJ), including victim-offender mediation, indirect communication through third parties, and restitution or reparation payments ordered by courts or referral panels. Much of the available and reasonably unbiased evidence of RJ effects on repeat offending comes from tests of face-to-face conferences of victims, offenders and others affected by a crime, most of them organised and led by a police officer; other tests cited involve court-ordered restitution and direct or indirect mediation.

What we found

Repeat offending

The most important conclusion is that RJ works differently on different kinds of people. It can work very well as a general policy, if a growing body of evidence on “what works for whom” can become the basis for specifying when and when not to use it. As tables 1 to 3 show, rigorous tests of RJ in diverse samples have found substantial reductions in repeat offending for both violence and property crime. Other tests have failed to find such effects, but with different populations, interventions or comparisons. In one rare circumstance, a small sample of Aboriginals in Australia, an offer of face-to-face RJ (and its partial completion) appears to have caused higher rates of repeat offending than CJ. This very limited evidence of backfiring can be balanced against the potential RJ may have as a full or partial alternative to incarceration for young adult offenders, who had much lower two-year reconviction rates (11%) in one Canadian study (N =138) than a matched sample (37% reconviction) who served their sentence in prison.

In general, RJ seems to reduce crime more effectively with more, rather than less, serious crimes. The results below (tables 1 to 3) suggest RJ works better with crimes involving personal victims than for crimes without them. They also suggest that it works with violent crimes more consistently than with property crimes, the latter having the only evidence of crime increases. These findings run counter to conventional wisdom, and could become the basis for substantial inroads in demarcating when it is “in the public interest” to seek RJ rather than CJ.

Victim effects

The evidence consistently suggests that victims benefit, on

average, from face-to-face RJ conferences. The evidence is less clear about other forms of RJ, with no unbiased estimates of the effects of indirect forms of RJ on victims. But when victims willingly meet offenders face to face, they obtain short-term benefits for their mental health by reduced post-traumatic stress symptoms (PTSS). This may, in turn, reduce their lifetime risks of coronary disease (which PTSS causes in military veterans), as well as reducing health costs paid by taxpayers.

Offences brought to justice

When RJ has been offered to arrestees before charging in New York and Canberra, RJ has always brought at least twice as many offences to justice – and up to four times as many. Whether such effects could be even greater with widespread take-up of RJ across a community is a major question to be answered.

A way forward

There is far more evidence on RJ, with more positive results, than there has been for most innovations in criminal justice that have ever been rolled out across the country. The evidence now seems more than adequate to support such a roll-out for RJ, especially if that is done on a continue-to-learn-as-you-go basis. Such an approach could be well supported by a “Restorative Justice Board” (RJB), modelled on the Youth Justice Board but on a smaller scale. An RJB could prime the pump for RJ, proposing new statutes and funding new solutions to the obstacles that now limit victim access to RJ. An RJB could monitor RJ practices, design tests of new RJ strategies, and continue to recommend systemic changes needed to make RJ as effective as possible. It could, in effect, take RJ from the drawing board to its widespread construction, while also remaining at the drawing board for on-going improvements in design based on new evidence.

How we found it

Searching for evidence

The search process for this review built on the literature search protocol approved by the International Campbell Collaboration for the authors’ registered and on-going review of the effects of face-to-face restorative justice for personal victim crimes.5 The search has been expanded for this review to encompass other forms of restorative justice and other kinds of crimes.

The following search strategies were used to identify evaluations of the effectiveness of RJ at helping victims and reducing reoffending:

  • searches of online databases;
  • searches of online library catalogues;
  1. All opinions and conclusions in this document are those of the authors and not of any governmental or private agencies that have funded any of the research the document reviews.
  • searches of existing reviews of the literature on the effectiveness of RJ;
  • searches of bibliographies of publications;
  • examination of publications already in our possession;
  • referrals by experts in the field.

Both published and unpublished reports were considered in these searches. The searches were international in scope, but were limited to studies written in English.

Weighing the evidence

For all questions of the causal effect of RJ on such outcomes as victim mental health and repeat offending, we restricted our review to reasonably unbiased estimates of the difference that RJ made in comparison to some form of CJ. We followed the methods used by the National Institute of Health and Clinical Excellence (NICE) to assess evidence on the effectiveness of medical treatments. These methods (NICE, 2006) require us to use the “PICO” principle (population, intervention, comparison and outcome), asking, with every study examined, for exactly what population the RJ intervention, in contrast to what comparison group, produced what outcomes.

In assessing the strength of the evidence in each study that offered a reasonably unbiased PICO analysis, we were able to apply the Home Office (2004) standards for reconviction studies. These standards are based in part on the Maryland scientific methods scale (Sherman et al, 1997), which set a minimum threshold of level 3 for the Maryland report to the US Congress, Preventing Crime. Level 3 requires that the outcomes of at least two relatively similar P and C (population and comparison) groups are compared with (P) and without (C) the intervention. This review adopts that threshold, so that all statements about what works to reduce repeat offending or improve victim outcomes are based on a comparison between reasonably similar cases receiving RJ or not receiving RJ. For questions of implementation and description, the report incorporates both qualitative and before/after quantitative research designs.

Studies selected

The search process and eligibility criteria resulted in the identification of 36 tests eligible for inclusion in our quantitative review of the impact of RJ. These consisted of 25 reasonably unbiased estimates of the impact of RJ on repeat offending, six reasonably unbiased estimates of the effects of RJ on victims, and five estimates of the effects of diversion from prosecution to RJ on offences brought to justice. These studies and point estimates are listed in tables 1 to 5 in the “Summary” section below.

Synthesising the evidence

As the NICE (2006) manual for developing guidelines for practice indicates, it is important to avoid over-mixing of results from substantially heterogeneous populations, interventions, comparisons or outcomes (“PICOs”). Equations that lump together studies into “meta-analyses” with great differences on these dimensions may yield an overall estimate of “effect”, but remain unclear as to the effect of what variety of intervention on which outcome for which population. A more conservative approach is to limit combinations of studies into “average” effects only when they share similar “PICOs”. Given the diverse nature of the studies identified for this review, it is usually necessary to treat each study as the only point estimate of its particular PICO characteristics.

The review makes cautious exceptions to that rule on a limited basis. We report the findings on repeat offending grouped separately by property and violent crime, so that the reader may look for patterns in relation to this basic distinction in the kind of harm (physical or non-violent) that offenders do to victims. What we do not do is “vote count” the studies, declaring a verdict about whether RJ “works” or does not “work”, either in general or in relation to specific characteristics of populations or interventions. The reason for that rule is that the available tests are by no means a fair “vote” from all possible tests. We do total the numbers of findings in different directions within broad domains, but this is merely for the convenience of the reader, who will want to do it anyway. We provide it only to emphasise the caution that is needed in interpreting the numbers.


  1. Introduction and overview


Introduction and overview

Fifteen years ago, Gordon Brown proposed that government should be “tough on crime, tough on the causes of crime”.6 That proposal succinctly captured the complexity of the crime problem, and thus the complexity of effective responses to crime. It implied that any financial investment in crime prevention strategy should be backed by good evidence of effectiveness, something distinctly lacking in a simple call for being “tough”. By even raising the issue of the causes of crime, the proposal opened the door to inventing new ways to deal with those causes.

The search for causes of crime logically begins with criminal justice policy itself. Three of every four new criminal convictions in England and Wales are reconvictions of previously convicted offenders.7 At the least, this fact suggests a missed opportunity for more effective and preventive sentencing practices when offenders are convicted. At the worst, it suggests that the criminal justice system itself is a cause of crime – a cause on which government should be tough.

Many aspects of criminal justice have been blamed for causing crime among convicted criminals. Inadequate or ineffective rehabilitation programmes, lack of drug treatment, insufficient funding for resettlement after prison, and other specific policies have all been nominated as causes. Others have suggested something far more fundamental: the way in which society and government thinks about the actual and potential connections between victims, criminals and society. A “war-on-crime” (or team sports) mentality of “us versus them” recalls a classic American cartoon that appeared on Earth Day 1970: “We have met the enemy and he is us.”8 This view blames our failure to see how interdependent all members of our society are, with many law-abiding people being criminals, victims or both at some point in our lives.

Restorative justice is a way of thinking about what is best for the many connections among crime victims, their offenders and the criminal justice process. Restorative justice advocates suggest that conventional assumptions about these connections may be wrong: that victims should be at the centre rather than excluded from the process, that victims and offenders are not natural enemies, that victims are not primarily retributive in their view of justice, that prison is not necessarily the best way to prevent repeat crime. The erroneous assumptions of conventional justice, the advocates suggest, contribute to rising public dissatisfaction with justice across the common law countries. This report considers whether restorative justice can do better, starting with more realistic and factual premises.

Offenders and victims, for example, are often assumed to be fundamentally different kinds of people. That assumption is largely mistaken, and can have critical and potentially devastating consequences for the administration of justice. It may induce more desire to seek revenge against offenders, when

victims might prefer to evoke a sense of remorse for the wrong. The mistaken premise obscures the reality that most criminals have themselves been victims, some from an early age. Thus criminals and victims often have much in common from which to build closer social bonds.

Another assumption of conventional justice is that the necessarily adversarial character of lawyers in court requires adversarial relations between victims and offenders as well. This adversarial assumption oversimplifies the roles played by citizen participants in the justice system, or the roles victims and their supporters would like to play. Restorative justice, at least in principle, seeks ways for victims and offenders to co-operate in preventing future crime and repairing past harms.

The classic mistaken assumption of conventional justice is to punish criminals as if they will never come back from prison to live among us. But with rare exceptions, they all come back. When they do, we depend on them not to cause more harm in the community. We are all interdependent in a shrinking world: criminals, victims, and the wider society. High rates of reconviction suggest that we are not doing what is needed to support that interdependence.

The doctrine that tougher punishment deters crime by making offenders fearful has been widely falsified for many kinds of offenders (Sherman, 1993). The restorative justice theory is that justice can prevent crime by making offenders feel more sympathy for their victims. This premise may be just as plausible as the deterrence doctrine, if just as unreliable. If restorative justice can work to prevent crime and repair harm, it seems likely to do so by fostering remorse, not fear. The emotions of anger, shame, guilt and regret form a complex cocktail of feelings associated with crime and justice. If we are to make progress in achieving the crime prevention goals of justice, it may happen from better understanding of how we can mobilise those emotions more effectively.

Restorative justice (RJ) is the prime but not only example of the recent trend towards a more “emotionally intelligent” approach to criminal justice (Sherman, 2003). This report reviews what we know, with some confidence, about the effects of different approaches to RJ. It is distinctively contemporary in its emphasis on feelings and bonds among people, both within offenders’ families and in their connections to victims and their families. Whether RJ can serve the people emotionally connected to each

  1. Anderson, Paul and Mann, Nyta Safety First: The Making of New Labour

(Granta Books, 1997), p243, footnote 23).

  1. Home Office, Sentencing Statistics 2004, table 6.2 (available at atistics%202004%22.
  2. Kelly, Walt Pogo (April 22, 1970) The cartoon satirises an 1813 battle report of the US Navy.

criminal incident remains to be seen. What is clear is that in a large number of crimes, RJ facilitators can readily identify people who care about the victims and offenders enough to try to deal with the aftermath.

Two claims

Restorative justice is a strategy that many people have advocated for responding to crime and intentional harm. Like many such strategies, it embraces a variety of forms with a single list of hypothesised outcomes. This report examines the available (and reliable) evidence on those hypotheses that has been produced to date, at least in the English language and within the scope of our search. It begins by stating the hypotheses that the 36 tests we found set out to examine.

These hypotheses can be summed up as two major claims, one about procedures, and one about effectiveness. The procedural claim is that restorative justice (RJ) is seen by victims and offenders as a more humane and respectful way to process crimes than conventional justice (CJ). The effectiveness claim is that RJ is better than CJ in producing important results that we want from justice: less repeat offending, more repair of harm to victims, fewer crimes of vengeance by victims, more reconciliation and social bonding among families and friends affected by crime, and more offences brought to justice.

Promising evidence

A systematic review of tests of these hypotheses offers promising evidence in support of both claims, although with caveats. Victims and offenders who participate in RJ are generally quite pleased with its procedures, more so than with CJ. Some of that evidence may be due to self-selection bias, but other tests eliminated that bias by giving participants little or no choice. This preference is accompanied by strong evidence that RJ is at least as effective in producing the desired results of justice as CJ, often more so, and only rarely (if powerfully) counterproductive. There are also indications of possible cost savings. This evidence is highlighted in the sections that follow below.

Few advocates claim that RJ should ever be used when an offender denies having committed a crime. Rather, the effectiveness claim suggests that RJ will foster more offenders agreeing to accept responsibility for having caused harm criminally. The hypothesis is that RJ would thus help to bring more offenders and offences to justice, since fewer offenders will deny responsibility if offered the prospect of RJ than they do with CJ. In fact, offenders in five controlled tests in New York City and Canberra readily took responsibility for serious crimes and “declined to deny” their guilt – choosing instead the prospect of participation in deciding what should be done about their crimes.

Varieties of restorative justice

Most of the evidence we highlight (tables 1 to 5) is based on just one of the many varieties of restorative justice (see section 3), one that is possibly most consistent with the broad definition

referenced by the Home Office (2003) strategy document, as quoted from Marshall (1999: 5): a process “whereby parties with a stake in a specific offence collectively resolve how to deal with the aftermath of the offence and its implications for the future”. That variety is face-to-face conferences of offenders, victims and their supporters.

This method can substantially reduce repeat offending for some (but not all) kinds of crimes and offenders (tables 1 to 3 and section 10). It can reduce victims’ desire for violent revenge against the offender (table 4 and section 9). Victims also suffer less intense post-traumatic stress symptoms after face-to-face restorative justice (Angel, 2005), returning to work and normal life sooner than they do without it – which should, in turn, reduce the long-term severity and costs of such health problems as coronary heart disease (Kubzansky et al, 2007).

Another “direct” form of restorative justice is victim-offender mediation, at which both victim and offender are present, often without other people affected by the crime, and where a mediator negotiates between them.

“Indirect” restorative justice usually describes any process by which offenders and victims communicate only through third parties, but not face to face. These methods include “shuttle communication”, in which a mediator or facilitator may carry messages by phone or in person between victims (or victims’ representatives) and offenders (or their representatives). They also include one-way communications such as letters of apology from offender to victim, or letters describing a crime’s impact from the victim to the offender. Perhaps the most indirect form of RJ is court-ordered restitution (which has become a substantial source of imprisonment in the US for offender failure to pay). Youth referral panels in the UK may also order restitution. When this form of RJ has been put to controlled tests, it has reduced recidivism in both adult and juvenile samples, but not consistently so (tables 1 to 2 and section 10).

Stages of the criminal process

Each of these forms of RJ has been employed in many different stages of the criminal process, and in different venues, as well as outside the CJ process altogether. RJ in schools, for example, can be used as an alternative to formal processing of young people involved in bullying. When arrests are made, RJ has been used as a diversion from criminal prosecution, as in the Canberra experiments. Prosecution in the UK has, under the Criminal Justice Act 2003, on occasion been suspended pending a “conditional caution” while the offender completes an RJ process (but only after making full admissions). When prosecuted offenders plead guilty, RJ before sentencing can provide evidence of mitigation that should normally reduce the length of imprisonment (R v Collins, 2003; R v Barci, 2003). When offenders are sentenced to community (probation) supervision, RJ can be used to determine conditions of that sentence and possible reparation to the victim. When offenders are sentenced

to long-term imprisonment, they can be invited to undertake indirect or direct RJ in preparation for resettlement.

Offences and offenders

The modern use of RJ began primarily in the context of youth justice, often for such minor offences as shoplifting and vandalism. In recent years it has expanded into serious offences by adults, including robbery, burglary and assault. From the standpoint of the process, the largest distinction among offence types is the “victimed” versus non-victim offences. The latter include drink-driving or shoplifting, in which no individual who was personally harmed can meet with the offender face to face. Theories underlying RJ, such as Braithwaite (1989), logically imply that RJ should work better at reducing reoffending with victimed offences. Any interest in RJ as a policy to help victims prefers victimed crimes by definition.

This review focuses not on a global assessment of whether RJ “works” as a one-size-fits-all strategy, but rather what the evidence suggests about what works for whom. There is only limited evidence on the large number of possible combinations of offence type, RJ method, stage of the criminal justice process, and offender age/gender/race/prior record characteristics. The vast scope of these specific circumstances limits the extent to which any review can draw firm conclusions about “what works best”. It is even possible that the differences in outcomes observed in these different categories could be due to chance, or to idiosyncratic aspects of sample selection or community characteristics. All the more reason, then, for laying out the sometimes conflicting evidence within offence types across different types of offenders, different stages of the criminal process, or different nations or cultures.

Searching for evidence

This review is based on a systematic search of the databases below from 1986 through to 2005, since the terms “restorative justice” and “conferencing” were not in use prior to 1986. Each of these databases is readily accessible online:

  • National Criminal Justice Reference Service (NCJRS)
  • Criminal Justice Abstracts;
  • Sociological Abstracts;
  • Criminal Justice Periodicals Index;
  • Dissertation Abstracts;
  • Social Science Abstracts.

The search terms to be used were “restorative and justice or conferenc/e/ing with reoffending, recidivism and evaluation”.

Weighing and summing up the evidence

The best methods of research synthesis for such a review are subject to substantial scientific and policy debate. This review cuts through some of that debate by relying on the Cabinet Office’s “PICO” model adopted from the methods of the National

Institute for Health & Clinical Excellence (2006: 16): population, intervention, comparison, outcomes. This model is used to weigh and sum up large bodies of research evidence for guidance to medical practitioners. It encourages greater specificity in definitions of both populations and interventions. It also requires comparisons with other treatments to be used as the basis for assessing an outcome, with reasonably unbiased selection of the comparison cases.

In weighing the evidence on RJ, this review follows the Home Office (2004) adaptation of the University of Maryland’s scientific methods scale (Sherman et al, 1997). This scale orders the level of internal validity, or the control of bias in drawing inferences of causation, on a five-point scale. Following the precedent of the Maryland report to the US Congress (Sherman et al, 1997) and the NICE (2006) standards, this review excludes impact evaluations on crime and victim outcomes that fall below level 3. That is, it does not consider research on RJ reasonably unbiased if it lacks direct comparisons between cases given RJ and similar cases given other treatments. Our review does, however, consider direct comparisons made between the actual recidivism of offenders given RJ to those offenders’ predicted recidivism based on the OGRS2 scale, which has been validated by Home Office research on samples other than the RJ cases (see Miers et al, 2001). This method is arguably equivalent to a level 4, as demonstrated in the evaluation methods literature (eg Berk and Rossi, 1998).

Many of the UK evaluations of restorative justice compare cases in which RJ was completed with cases in which it was offered but refused, by either offenders or victims (see eg Miers et al, 2001). While such comparisons may appear to be level 3 or 4 designs, the treatment-related difference between the two groups makes any estimate of the causal effect of treatment potentially quite biased. It is not possible with such designs, no matter how many cases they may include in their samples, to eliminate the key alternative rival hypothesis: that there is something about those cases that makes consent versus refusal a confounding predictor of recidivism, in uncertain and unspecifiable ways. We excluded studies in which that method is used for estimating either offender or victim outcomes.

Similarly, many studies include only “completers” of RJ in their samples, with no data on those who were offered or initiated RJ but did not complete it (eg Vignaendra and Fitzgerald, 2006). Except in very large samples with a valid model predicting the reasons for non-completion that could be applied to a comparison group (Angrist, Imbens and Rubin, 1996), comparisons of completers with a control group not offered the treatment are inherently biased (Gorman, 2005). This applies equally to contemporary and historical controls (see eg Stone et al, 1998, which is excluded for that reason). Such studies are thus also excluded as evidence of the impact of RJ. For related reasons, correlational studies comparing offenders who do and do not get RJ (level 1) are also excluded, as well as before/after studies

(level 2) that are vulnerable to regression-to-the-mean.9

Our key findings, summarised in tables 1 to 5, rely on 23 reasonably unbiased point estimates of the impact of RJ on repeat offending, six estimates of effects on victims, and five estimates of effects of diversion from prosecution to RJ on offences brought to justice. The studies and their key data are listed in tables 1 to 5 below. While all level 5 studies employed random assignment to determine whether or not a case received RJ, some of the conclusions (or what statisticians call “point estimates”) are derived from subgroups of the full sample that was included in the random assignment sequence (Piantadosi, 1997: 211).

Missing evidence

In assessing the prospects for a major expansion in the use of RJ, the review found no evidence of the kind that might be needed to roll out a national policy. That evidence about the effects of up-scaling RJ for widespread use – whether it would produce “collateral benefits” or harms for entire communities beyond those we can observe in studies limited to comparisons of individual cases – would include the following questions:

  • Would broader use of RJ encourage more witnesses to come forward to help police solve more crimes, bringing more offences and offenders to justice by more public confidence in justice and the law?
  • Would broader use of RJ encourage more offenders to accept responsibility for their offences, increasing offences brought to justice?
  • If an entire community or basic command unit adopted RJ as the initial response to most crimes, would that policy weaken the general deterrent effects of the law – or strengthen compliance with the law by increasing its legitimacy?
  • If RJ became more widely known and understood by the public through far more frequent use, would it attract even more victim and offender consent to participate than the substantial levels it has achieved in some pilot tests?

These and other questions about large-scale RJ could be answered with evidence generated by the scientific method, using communities as the unit of analysis. They cannot, however, be answered with evidence generated solely at the individual or case level in small-scale pilot tests. Thus the implication of this conclusion is that progress in evidence-based restorative justice is likely to depend on whether future testing of RJ is conducted on a neighbourhood-wide or community-wide basis (section 15).

The reasons why

One key question the Smith Institute asked us to examine is why RJ works when it does work. The short answer is that we cannot tell much from the available evidence, but there are some theories that could guide further analysis. The modern revival of RJ has been long on theory, but shorter on tests of those theories. Even when RJ itself is subjected to rigorous testing, the theories

that could explain its effects are often much harder to test. This situation is not uncommon in science, as in the case of antibiotics, which cure infections for reasons that are not fully understood. Yet there is no doubt that understanding the reasons why RJ works – or doesn’t – could help improve predictions and policies about when to use it or not.

A central theory about RJ highlights a massive difference from the theory of conventional justice – and also explains why conventional justice fails to deter repeat convictions far more often than not (see section 11). That theory, based on “defiance theory” (Sherman, 1993), is that:

  • People who commit crimes often believe, or convince themselves, that they are not acting immorally.
  • RJ engages such people in a moral discussion about whether crime is wrong.
  • An RJ discussion can lead offenders to redefine themselves as law-abiders, and to agree that they are not the kind of people who would do immoral things.
  • That discussion would lead to the conclusion that what they did was in fact immoral, and that they should therefore not repeat such behaviour. (Whether they do anyway remains the key empirical question about this theory.)

In contrast, criminal law doctrine presumes that people know they are doing wrong, and that only fear of punishment can stop them from repeating their crime. Thus punishment is required in order to deter them (and others) from doing such wrongs. While such neoclassical theory therefore centres on punishment, RJ centres on persuasion. The aim of punishment is to enhance fear of further punishment; the aim of persuasion is to enhance moral support for voluntary obedience of the law (Braithwaite, 1989; Tyler, 1990; Sherman, 1993, 2003). A theory of obeying the law by persuasion is thus the ultimate commitment to a rule of law.

The rule of law

One frequent objection to RJ is that it may undermine the rule of law, encouraging community abuses of individual rights at the expense of universal principles and rights. This review finds no evidence for this concern. As long as RJ is conducted under the United Nations principles for RJ (see section 5), the major issue for the rule of law will be one that has long plagued conventional justice as well: disparity in severity of penalty. Both RJ and CJ appear equally subject to challenges of disparities and a wide variance of discretion, as former Lord Chief Justice Woolf pointed out in his advice to judges on the testing of RJ in 2002.10

  1. As a report to the Canadian government has noted, “reasonably well-designed studies of the impact of restorative justice programmes on recidivism are few” (Bonta et al, 1998: 4). 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.

Table 1: Reasonably unbiased tests of RJ effects on repeat offending after processing for violent crime

Place,   SMS (internal validity) levelReferencePopulationInterventionComparisonOutcome
Bethlehem, Pennsylvania, USA Level 5McCold and Wachtel, 1998Hispanic (51%) and white youth N = 111Diversion to face-to- face RJ conferences, prior to victim or offender consentConventional juvenile prosecutionNo difference in intention-to-treat analysis; high refusal rate post-random assignment
Canberra, Australia Level 5 (subgroup analysis)Sherman, Strang, Barnes and Woods, 2006Non-Aboriginal defendants under 30 N = 97Diversion to face-to- face RJ conferences, with consent of offender prior to victim consentConventional juvenile prosecutionRJ-assigned had 84 fewer arrests per 100 offenders per year than CJ-assigned (P = .026)
Canberra, Australia Level 5 (subgroup analysis)Sherman, Strang, Barnes and Woods, 2006Aboriginal defendants under 30 N = 14Diversion to face-to- face RJ conferences, with consent of offender prior to victim consentConventional juvenile prosecutionSubgroup sample size too small for adequate power to test effect
Indianapolis, USA Level 5McGarrell et al, 2000Youth N = 251Face-to-face RJ conferencesConventional juvenile diversion to a range of other programmes28% rearrest rate at six months for RJ vs 34% for CJ (P < .05); effect decays by 12 months
Newfoundland and Labrador, Canada Level 4Pennell and Burford, 2000Violent families N = 32Face-to-face RJ conferencesConventional criminal justice and social service response N = 3150% before/after reduction in frequency for RJ families vs 27% increase for CJ families (P = .005)
Northumbria, UK Level 5 (subgroup analysis)Sherman, Strang, Barnes and Newbury- Birch, 2006Female youth N = 44Face-to-face RJ conferences in addition to final warnings by policeConventional final warnings by police only118 before/after fewer arrests per 100 offenders in RJ group vs 47 fewer arrests in CJ group (P =.012)
Northumbria, UK Level 5 (subgroup analysis)Sherman, Strang, Barnes and Newbury- Birch, 2006Male youth N = 64Face-to-face RJ conferences in addition to final warnings by policeConventional final warnings by police onlyNo RJ-CJ difference
West Yorkshire, UK Level 4Miers et al, 2001Young adult violence and property offenders (58% given custodial sentences) N = 153Pre-sentence indirect, some direct mediation, requiring victim participation but rarely face-to-face; not reported to sentencing courtOffenders’ own predicted recidivism rate, based on external model (OGRS2 score)Two-year reconviction rate = 44% vs 58% predicted (P = .01)
West Midlands, UK Level 4Miers et al, 2001Young adult violence and property offenders (52% given custodial sentences) N =147Pre-sentence only; offenders told mediation would be reported to court and could help reduce sentenceOffenders’ own predicted recidivism rate, based on external model (OGRS2 score)Two-year reconviction rate 44% vs 57% convicted (P =.01)
Kings County (Brooklyn), New York, USA Level 5Davis et al, 1981Adult felony defendants (family = 50%; acquaintance = 40%; violent = 60%; property 40%) N = 465Diversion from prosecution to direct mediation; 56% completedProsecution as usual: 27% conviction rate; 72% dismissed or absconded; jail sentences = 2.5%No difference in four- month post-disposition (control absconders excepted!) rate of calling police: RJ =12%; prosecution = 13%. Arrests of victim OR defendant = 4% in both groups
  Total = 10      Reductions = 6 Increases = 0 No effect = 4

Table 2: Reasonably unbiased tests of RJ effects on repeat offending after processing for property crime

Place, SMS (internal validity) levelReferencePopulationInterventionComparisonOutcome
Northumbria, UK Level 5 (subgroup analysis)Sherman, Strang, Barnes and Newbury- Birch, 2006Male youth N = 100Face-to-face RJ conferences in addition to final warnings by policeConventional final warnings by police onlyRJ = 88 fewer arrests per 100 offenders per year before/after vs 32 fewer for CJ (P < .05)
Northumbria, UK Level 5 (subgroup analysis)Sherman, Strang, Barnes and Newbury- Birch, 2006Female youth N = 28Face-to-face RJ conferences in addition to final warnings by policeConventional final warnings by police onlySubgroup sample size too small for adequate power to test effect
Canberra, Australia Level 5 (subgroup analysis)Sherman et al, 2006aWhite youth arrested for crimes with personal victims N = 228Face-to-face RJ conferencesConventional juvenile prosecutionNo difference
Canberra, Australia Level 5 (subgroup analysis)Sherman et al, 2006aAboriginal youth arrested for crimes with personal victims N = 23Face-to-face RJ conferencesConventional juvenile prosecutionComparing two years after to two years before, RJ = 288 more arrests per 100 offenders per year; CJ = 66 fewer arrests per 100 offenders per year (P = .049)
Bethlehem, Pennsylvania, USA Level 5McCold and Wachtel, 1998Hispanic (51%) and white youth N = 181Face-to-face RJ conferencesConventional juvenile prosecutionIntention-to-treat analysis shows marginally significant (P = .11) higher offending in RJ
Indianapolis, USA Level 5McGarrell et al, 2000Youth N = 381Face-to-face RJ conferences, with consent of offenderConventional diversion to a range of non-RJ programmesRJ = 15% repeat offenders at six months, vs CJ = 27% (P <.05); difference fades by 12 months
Clayton County, Georgia, USA Level 5Schneider, 1986Youth N = 128Court-ordered restitutionConventional probationRJ = 26% decline in arrest frequency vs no change in CJ (P value not reported)
Boise, Idaho, USA Level 5Schneider, 1986Youth N = 181Court-ordered restitutionEight days in jail on weekendsNo difference
Oklahoma City, USA Level 5Schneider, 1986Youth N = 182Court-ordered restitutionConventional probationNo difference
Washington, DC, USA Level 5Schneider, 198673% youth property offenders; 27% violent offenders N = 411Court-ordered restitution, with consent of offender (40% dropout)Conventional probationIntention-to treat case RJ = 12% less arrest frequency before/after; CJ = 7% increase (P < .05)
Winnipeg, Canada Level 4Bonta et al, 1998Male adult property (61%) and violence (35%) offenders, accepted at a 33% rate from defence lawyer referrals, all likely to be sent to prison for at least six months N = 142Court-ordered financial restitution and RJ face to face (10%) or by letter of apology N = 75Matched prison inmates with similar profiles N = 67Two-year reconviction rate for 75 RJ = 11%, for 67 ex-prison inmates = 37% (P < .05)
New Zealand Level 4Triggs, 2005Court-referred offenders who completed RJ (25%) N = 192Face-to-face RJ conference with diverse participantsOffenders’ own predicted recidivism rate, based on external modelNo significant difference in two-year reconvictions between RJ (41%) and predicted (45%)
Total = 12    Reduction = 5 Increase = 2 Match or beat custody = 2 of 2

Table 3: Tests of the effects on repeat offending of RJ for samples of all or partially non-victim offences

Place, SMS (internal validity) levelReferencePopulationInterventionComparisonOutcome
Canberra, Australia Level 5Sherman, Strang and Woods, 2000Drink-driving offenders caught in random breath tests N = 900Face-to-face RJ conferences with five family members or supporters; sometimes community representedProsecution in court, six months’ loss of driver’s licence, name published in newspaperNo RJ-CJ difference in before/after difference of frequency in repeat offending
Canberra, Australia Level 5Sherman, Strang and Woods, 2000Youth shoplifters N = 143Face-to-face RJ conferences with five family members or supporters; sometimes store representedConventional prosecution in juvenile courtNo RJ-CJ difference in before/after difference of frequency in repeat offending
Indianapolis, USA Level 5McGarrell et al, 2000Youth public order offenders N = 143Face-to-face RJ conferencesConventional diversion to a range of non-RJ programmesRJ = 28% rearrest after 12 months; CJ = 45%
Total = 3      No difference = 2 RJ decrease = 1 RJ increase = 0

Table 4: Reasonably unbiased estimates of effects of RJ on crime victim outcomes

Place, SMS (internal validity) levelReferencePopulationInterventionComparisonOutcome
Canberra, Australia Level 4.5Strang, 2002Victims of violent crime by offenders under 30, or of property crime by offenders under 18 N = 232 (Two separate RCTs combined)Diversion to face-to- face RJ, with consent of offender prior to victim consentConventional prosecution in juvenile or adult courtAnger at justice process: CJ = 32% RJ = 18% Desire to harm offender: CJ = 20% RJ = 7% Preference for process: RJ = 69% CJ = 48% Satisfaction with outcome: RJ = 60% CJ = 46% (All with P = .05 or less)
London, UK Level 4.5Angel, 2005Victims of robbery or burglary N = 216 (Two separate RCTs combined)Face-to-face RJ in addition to CJ, with consent of offender prior to victim consentConventional prosecution in court without RJPost-traumatic stress symptoms scores for: RJ = 9 CJ = 14 (P < .01)
London, UK Level 4.5Angel, 2005Victims of robbery or burglary N = 207 (Two separate RCTs combined)Face-to-face RJ in addition to CJ, with consent of offender prior to victim consentConventional prosecution in court without RJPost-crime impact on employment scores for: RJ = 16% CJ = 25% (P < .12)
Canberra, Australia, and London, UK Level 4.5Sherman et al, 2005Eight-point estimates of male and female victims of violent and property crimes Total N = 445 (Four RCTs disaggregated by gender)Face-to-face RJ conferences in addition to or instead of CJ, with consent of offender prior to victim consentConventional prosecution in courtMean proportions desiring violent revenge against offender: RJ = 4% CJ = 14% (P < .001)
Indianapolis Level 4.5McGarrell et al, 2000Victims of youth offenders, latter aged seven to 14 N = 92 (Low response rates, data “descriptive”)Diversion to face-to- face RJAny of 23 court- ordered diversion programmesSatisfied with how case was handled: RJ = 90% CJ = 68% Would recommend to other victims: RJ = 98% CJ = 25%
Bethlehem, Pennsylvania, USA Level 1McCold and Wachtel, 1998Victims of violent and property crime by offenders under 18 N = 180Diversion to face-to- face RJ before consent of offender or victimConventional prosecution in juvenile courtSatisfied with the way case was handled: RJ = 96% RJ decline = 73% CJ = 79% (P < .01) Satisfied offender was held accountable: RJ = 93% RJ decline = 77% CJ = 74% (P = .05)
Total six tests summarised with 10- point estimates    In 10 of 10 estimates, victims favour RJ over CJ

Table 5: Reasonably unbiased estimates of effects of RJ on offences with victims brought to justice

Place, SMS (internal validity) levelReferencePopulationInterventionComparisonOutcome
Kings County (Brooklyn), New York, USA Level 5Davis et al, 1981Adult violent and property felony defendants N = 465Diversion to direct face-to-face mediationConventional prosecutionOffences brought to justice: RJ = 56% CJ = 28% Ratio 2:1
Canberra, Australia Level 5Strang et al, 1999Defendants under age 30 charged with violent offences N = 65Diversion to face-to- face RJ, with consent of offender prior to victim consentConventional prosecutionOffences brought to justice: RJ = 89% CJ = 44% Ratio 2:1
Canberra, Australia Level 5Strang et al, 1999Youth under 18 charged with property crimes against personal victims N = 126Diversion to face-to- face RJ, with consent of offender prior to victim consentConventional prosecutionOffences brought to justice: RJ = 92% CJ = 27% Ratio 3:1
Canberra, Australia Level 5Strang et al, 1999Licensed drivers arrested for drink- driving N = 773Diversion to face-to- face RJ, with consent of offender prior to victim consentConventional prosecutionOffences brought to justice: RJ = 99% CJ = 87% Ratio 1.1:1
Canberra, Australia Level 5Strang et al, 1999Youth arrested for shoplifting from large stores N = 87Diversion to face-to- face RJ, with consent of offender prior to victim consentConventional prosecutionOffences brought to justice: RJ = 93% CJ = 18% Ratio 4:1
Summary Five tests    Ratio 5:0 tests of RJ increases of offences brought to justice over CJ

Any “balanced” perspective on criminal law is a subject of intense academic and public debate. Adding victims’ benefits and crime prevention into the mix of criminal law offends a purely retributive view of “just desserts”, leading some scholars to find RJ incompatible with traditional criminal justice (see section 5). Yet those concerns are also present in CJ, with RJ seen by others as a way to help rebalance CJ itself. The evidence shows that RJ leads to more opportunities for offenders to help both victims and themselves, posing a win-win scenario more often than the zero-sum game assumed by RJ opponents (Strang, 2002: 155-191).

The major factor affecting disparity of severity is the problem of victim refusal to participate when an offender is willing. That problem, however, has been resolved in part by R v Barci. This case requires judges to give offenders mitigation of time in prison based on their willingness to participate in RJ, regardless of any victim’s agreement to do so. Victim-absent conferences are also possible under those circumstances, which could allow the theory of RJ to be implemented in a similar fashion to what happens when victims are present (see section 3).

Concerns about inconsistency in the severity of punishment associated with RJ can also be dealt with in a variety of ways, including judicial review (as in New Zealand) and good practice for insuring compliance with outcome agreements, as demonstrated by the Metropolitan and Northumbria police (see section 7). It can also be managed by Crown Prosecution Service oversight, under the Criminal Justice Act 2003. These and many other solutions are possible if evidence continues to grow of the benefits to be gained from making RJ more widely available. That evidence, however, must distinguish between cases in which RJ is likely to work, have no effect, or even cause more crime.

Reliance on evidence to decide when RJ is “appropriate” can yield very different decisions from relying on theory or subjective bias. One example is the usual proposal to limit the use of RJ to “lesser” crimes, including juvenile offences, but not to allow it for “serious” crimes. Yet the evidence suggests RJ may be most effective when the crimes are most serious. For minor crimes, RJ is no better than CJ in reducing repeat offending among shoplifters, drink-drivers, and teenage property offenders in Canberra. For major crimes RJ has succeeded better than CJ in reducing repeat offending among felony defendants in New York City, violent white people under 30 in Canberra, and violent white girls under 18 in Northumbria. Banning RJ for serious crimes would destroy the chance to prevent many thousands more such offences. Nor is it clear that there is any principled basis for selectively allowing, or banning, RJ – other than the principle of harm reduction, which indicates its use with serious crime.

Predicting RJ effects on repeat crime by offenders

The question of repeat offending is often thought to be a small or limited part of the crime problem. UK evidence suggests otherwise. According to Home Office data, 76% of all persons

sentenced for indictable offences in 2003 in England and Wales were people with prior convictions.11 If prior arrests had been used, the percentage of convicted criminals who had indications of prior crime would probably have been even higher. Based on people either caught or actually punished, then, most crime is repeat crime.

How can we tell how much crime RJ might prevent? A reliable prediction of how RJ will affect repeat offending by offenders attending conferences must first be based on unbiased evidence. Much of the evidence about RJ, even with very large sample sizes, is uninformative for this purpose because it is plagued with the bias of self-selection (eg Vignaendra and Fitzgerald, 2006). This means that the kinds of offenders who complete RJ may be substantially different from those who do not, in ways that may predict their risk of repeat offending regardless of RJ. If people with lower risk of repeat offending are more likely to be offered, accept or complete RJ than people with higher risks of repeat offending, then the reason why RJ cases had less repeat offending would not be the result of RJ. The correlation with RJ would be what statisticians call a “spurious” association that reflects some third, underlying cause, rather than the effect of RJ. Much of the positive evidence on RJ suffers from this fatal flaw, especially the direct comparisons between RJ completers and RJ refusers (see section 6).

Much of the evidence claiming that RJ does not reduce crime may suffer a different bias: the bias of measurement. Using police records on thousands of cases, for example, as the sole measure of the delivery of RJ may fail to detect an enormous variability in the content and intensity of the RJ experience. When RJ (or any programme) is rolled out quickly on a wide scale, there is a risk that many conferences will just “go through the motions” to “tick off a box”, rather than treating each case as a kind of surgical procedure requiring careful advance planning, preparation and follow-up. With such heterogeneity of the RJ being delivered, the research is biased against finding any effect of “good practice” RJ, because no measurement was taken of the elements of good practice (see section 6).

Randomised controlled trials (RCTs) provide the best opportunity to control both selection and measurement biases. RCTs generally remove selection bias because they first obtain consent and then assign RJ to some (but not all) of those consenting. RCTs also measure consistency of delivery of RJ, so it is clearer just what is being tested. Finally, the best RCTs analyse their data based on assignment rather than completion of RJ, so that any self- selection bias in completion is eliminated. While this procedure dilutes the effects of RJ, it maintains the capacity of the research to rule out spurious causes of a difference in repeat offending. RCTs are not the only kind of evidence that can help predict the effects of RJ on crime, but they are now available in greater

  1. Home Office, Sentencing Statistics 2004, table 6.2 (available at atistics%202004%22)

abundance for RJ than for any other response to crime ever tested.

These controlled tests show that face-to-face RJ, consistently delivered by facilitators (mostly police officers) trained by the same Australian RJ training firm, has reduced repeat offending on three continents, for highly specific populations, all of which are identified by characteristics that existed before random assignment and are therefore considered by statisticians to be appropriate for subgroup analyses that produce statistically significant within-group differences by treatment (see section 10):

  • violent offenders under age 30 in Canberra (main effects);
  • violent girls under 18 in Northumbria;
  • male property offenders under 18 in Northumbria;
  • property and violence offenders aged seven to 14 in Indianapolis.

Tests of other varieties of RJ, primarily court-ordered restitution or intervention of various descriptions, has also reduced crime among:

  • violent families in Newfoundland and Labrador;
  • adult male property and violence offenders diverted from prison in Winnipeg (and recidivism no worse than from jail for youth in Idaho);
  • youth property offenders in Clayton County, Georgia;
  • youth violence and property offenders in Washington, DC.

The rigorous RCT methods, however, have also found evidence that the same kind of RJ delivered by the same facilitators made little or no difference in repeat offending rates among police-selected samples of:

  • property offenders under 18 in Canberra;
  • violent males under 18 in Northumbria;
  • property offenders under 18 in Bethlehem, Pennsylvania.

Most important, these same RCT methods have found that face- to-face RJ offered as a diversion from court causes a substantial increase in the frequency of arrests among a small sample of Aboriginals under 18 in Canberra arrested for property crimes, when compared with Aboriginals randomly assigned to court for similar offences (see section 11). The Bethlehem findings, with many Hispanics in the sample, also veer close to significance (P = .11) in the same direction.

These findings raise important “reason why” questions about the relationship between RJ and social “marginality”, with both negative and positive implications (see section 11). One question is whether offenders from deeply alienated social groups, such as Australian Aboriginals or American Hispanics, will react very badly to appeals for obedience to the laws of a government they perceive as illegitimate, if they do not believe their crimes to be wrong. The same question could be raised about Islamist radicals (who believe murder in the name of God to be moral) or certain

Afro-Caribbean gangs (who may see violence as part of the moral rules of a business enterprise).

The other question is whether the emotional power of RJ could be customised to engage whatever authority structure may be effective inside such alienated groups. Mobilising older Aboriginal males to attend RJ conferences, for example, could increase the perceived legitimacy of the process among such offenders – changing their minds about the morality of obeying the law more than a white police officer or white crime victim might be able to.

Insight about other reasons RJ results, in general, vary widely among different kinds of offenders can be gained from a quasi- experimental (non-RCT) study of differences in repeat offending among offenders under age 18 in South Australia. This found, controlling for other predictors of recidivism, that the lowest repeat offending rates followed conferences during which offenders showed remorse, and in which agreements were reached by a clearly consensual process among the people in the room (Hayes and Daly, 2003). Similar findings (Morris and Maxwell, 2005) have been reported in long-term follow-up of juvenile cases in New Zealand (see section 10).

The large magnitude of the RJ effects in this evidence – both good and bad – suggests that RJ is like a powerful drug that needs to be carefully tested for specific kinds of cases before it is put into general practice. Just as penicillin can cure infections, but cannot cure cancer or diabetes, RJ can reduce crime for some kinds of offenders but not others. And just as some people are so allergic to penicillin that it can almost kill them, some offenders may find RJ so enraging or humiliating that they are provoked into committing even more crime than they would have done without RJ. This evidence might be taken as a reason to ban RJ altogether, but only one such very strong reaction has been found to date, among a small sample of Australian youth.

All interventions in medicine, or agriculture, or public health (such as vaccination programmes) cause harm under some circum- stances. Yet that is rarely sufficient reason to impose a complete ban on a treatment that offers benefits under some conditions. Instead, the usual response is to predict as reliably as possible when there will be harm, and then to apply limited prohibitions of the intervention under those specific circumstances. That approach has allowed antibiotics to save millions of lives. It could also allow restorative justice to prevent millions of crimes.

Predicting RJ effects on victims

The evidence on victims is far more consistent than it is on offenders. On average, in every test available, victims do better when they participate in RJ than when they do not. Victims may report dissatisfaction in the (very infrequent) cases when offenders refuse to accept responsibility, or if offenders fail to appear at a conference as agreed, or when offenders fail to complete outcome agreements. Yet the very high rate of

offender attendance, remorse and apologies in RJ conferences far outweighs these exceptions, protecting victims from being “re-victimised” during the RJ. Instead, from Canberra to London to Indianapolis, victims who go to RJ conferences report that they are glad they went. The benefits they describe include less fear of the offender, less anger at the offender, and greater ability to get on their lives (see section 9).

The reductions in anger at the offender extend to victims admitting they have less desire for physical revenge against the offender after RJ than before – a result confirmed in some tests by far higher levels of desire for vengeance among victims assigned to control group status than those assigned to an RJ conference.

In London, the effects of RJ on victims of burglary and robbery include large reductions in their post-traumatic stress symptoms. Compared with victims willing to meet with willing offenders but who were not randomly assigned to RJ, victims who were assigned to (and completed) RJ reported greater ability to return to work, to resume normal daily activities, to sleep better at night and to stop their “racing thoughts”. Long-term data on these victims may reveal how much RJ has improved their health, given other evidence that comparable levels of post-traumatic stress elevate risks of coronary heart disease (Kubzansky et al, 2007). That, in turn, could potentially reduce National Health Service costs in an amount sufficient to justify spending on RJ as a disease prevention strategy for crime victims.

Could RJ reduce government spending?

Healthcare for crime victims is only one of several possible avenues by which RJ could reduce government spending (see section 14). Perhaps the largest opportunity is in the reduction of the prison population, especially (as Lord Woolf has suggested) in the area of short sentences. Two cases in the Court of Appeal (Collins and Barci) have already ruled that length of custody should be reduced for offenders who offer to participate in RJ. For somewhat less serious offences in which courts may be inclined to give a short custodial sentence, the addition of RJ could possibly tip the balance to keep them out of custody altogether. Even in robbery cases, London Crown Court judges have said that they had withheld a custodial sentence due to offender participation in RJ (see sections 2 and 4).

The possible substitution of RJ for prison is even more attractive if it would result in less crime. The reduction of reconvictions in Winnipeg (Bonta et al, 1998) with RJ as an alternative to prison provides evidence of that possibility. Even if there is merely no increase in crime, the cost savings could be achieved with zero impact on public safety. Evidence that in Boise, Idaho, youths did no worse with RJ than they did with eight days in jail is especially relevant to the debate over short sentences in the UK. At the pro rata cost of some £35,000 per year for each UK prison sentence, one offender kept out of prison for one year would cover the costs of more than 50 RJ conferences (at £25 per hour of police work for an average of 20 hours per conference, plus supervisory

and overhead costs). That would equal one week of custody for each RJ conference.

Put another way, if only one in 50 RJ conferences prevented a year in custody, that alone could cover the costs of the conferences. The money for one year could thus be saved in one of two ways: by reducing sentence length, or by reducing the costs of repeat offending and reincarceration.

Another way that RJ could save money is in fees paid to lawyers by the government for appearances in court and at police stations. This would require many more defendants than at present to admit guilt in anticipation of a conditional caution involving RJ (see section 14). Each admission of guilt – and diversion to RJ – could save thousands of pounds in legal fees for both defence and prosecution.

How can RJ best be delivered?

RJ has often been delivered badly in the UK by programmes operated on a take-it-or-leave-it philosophy. Courts have been asked – even in person by the Lord Chancellor – to refer cases prior to sentencing, but they did not. Victims have been sent letters inviting them to attend referral panels (on a non- negotiable date), but they did not. Prosecutors have been asked to approve cases for diversion to RJ as conditional cautioning, but they did not. Magistrates have been asked to delay sentence so that RJ conferences might be held, but they did not. RJ was there for the taking, but the current system just left it. The exception may be the youth conferencing programme in Northern Ireland, which – unlike RJ in England and Wales – is statutory. The evaluation indicates that the Northern Ireland programme works well, with cases routinely referred to RJ by criminal justice agencies as the law intends (Beckett et al, 2005; Campbell et al, 2006).

It thus appears likely that RJ programmes can be successfully run only via statutory and regulatory engineering of the criminal justice system (see sections 3, 13 and 15). Statutory requirements could ultimately include an adjournment after each guilty plea to allow police to visit victims and offenders, who could then decide whether to participate in RJ. That decision could be informed by risk assessments conducted by police, who would also draw on the latest research about the kinds of cases for which RJ may be more or less effective (or could potentially backfire). Statutes could also allow or require custody suite sergeants to explain the possibility of diversion to RJ in a conditional caution if offenders admit their guilt upon arrest, which is exactly what police in Canberra were able to do without legislation. Such a practice would be contentious in the UK if seen as an “inducement”, but could be cast as part of a standard information package that would be provided to all arrestees.

Other requirements could include the “best practice” methods of contacting victims face to face by knocking on doors when victims cannot be reached by phone or by post. Similar standards

could be set for contacting offenders on bail, and for mobilising all participants the day before, and the day of, an RJ conference.

Most of the rigorous evidence on RJ to date is based on police administration of the RJ processes. Yet the expense of trained officers providing these services makes some police leaders reluctant to commit that expense to RJ. While there are no direct comparisons between RJ provided by police or members of other professions (with similar cases), there is a (still untested) way to reduce the cost of restorative policing. Testing police community safety officers assigned as “RJ officers” to work with a small number of sworn constables could offer a way of providing the benefits of association with police protection, as well as incorporating more community-based knowledge of the cultural dimensions of bringing together offenders and victims.

A competing viewpoint is that RJ facilitators should have no affiliation with criminal justice agencies (Roche, 2003). There is no evidence to suggest, however, that this practice can even attract substantial numbers of crime victims, let alone produce powerful reductions in offending or victim harm. The New Zealand system of social workers leading RJ conferences, for example, has a much lower victim attendance rate than the police-led programmes evaluated elsewhere (Maxwell and Morris, 1993). The best evidence for large-scale delivery of RJ, at present, is associated with police-organised and -led conferences.

More justice, less crime: a way forward

Restorative justice offers a strategy for holding more offenders accountable, with many more victims helped, with more crimes prevented, and with the costs of government reduced. The evidence so far suggests that many elements of this strategy can work with some kinds of offenders and offences. That evidence is far more extensive, and positive, than the evidence base for most national roll-outs of new criminal justice policies in any govern- ment. This conclusion would support a decision to roll out RJ as well.

Building on the promise of the evidence requires two conditions. In the short run, it requires an institutional focus for the development of RJ as a major shift in justice policy, as distinct from a minor programme on the margins. In the long run, it requires a continuing growth in the evidence of its effectiveness in order to withstand theoretically based attacks that it is not “tough” enough or may cause more crime. Both of these may be achieved by creating a stand-alone RJ agency comparable to the Youth Justice Board.

A “Restorative Justice Board” (RJB) could provide the focus and leadership for overcoming the obstacles to delivering RJ on a widespread basis in England and Wales. It could consist of five or seven members drawn from diverse constituencies concerned with justice policy, including victims, courts, police, probation, prisons, treatment professionals and the public. It could be empowered to set standards and make recommendations for statutory and policy changes. It could be given a budget to

deliver two products. One would be the implementation of high- quality RJ on a much wider scale, fostered by investing the most in communities achieving the most RJ and by recommendations for statutory or other systemic changes needed to broaden access to RJ. The other focus of the budget would be research and development, investing in testing new ways to deliver RJ even in communities most challenged by crime.

This second focus of the RJB would thus be able to fill the major evidence gap found in this review: community-level impact assessments. Unlike the abundant evidence of the effect of RJ on individual victims and offenders, there is no evidence on how a widespread use of RJ would affect community rates of crime and respect for the law. In order to estimate the effects of RJ as a general (national) policy, it is necessary to conduct evaluations using neighbourhood crime rates, rather than individual offending patterns, as the unit of analysis. This review found no such tests to date. Such evidence could, however, be far more relevant to policy making than further research using individual cases as the unit of testing. Rolling out RJ right across a neighbourhood – from every stage of the criminal justice system to every civic organisation and governmental institution dealing with harms – could yield the best evidence on the effects of building a restorative society. It could also help to inform an RJB of the best practices to encourage in making funding decisions and setting standards.

Best practice for RJ

The ultimate purpose of this review is to identify “best practice” for RJ. To do that in an evidence-based way requires a focus on the two best-studied forms of RJ, face-to-face conferences and court-ordered restitution. The evidence is far from definitive, but it is at least suggestive. That suggestion is the following “best practice” in at least setting priorities for investments in RJ:

  • RJ seems to work best when it is focused on the kinds of offences that have a personal victim, who can – at least in principle – be invited to meet with the offender. The major criteria for “working” in this claim include helping victims and reducing reoffending.
  • RJ seems to work best when it is focused on violent crime, rather than property crime, with major exceptions: burglary victims gain reduced post-traumatic stress symptoms, and property offenders may commit less crime in future (or at least no more) if they get RJ than if they get prison.
  • RJ may be best able to reduce court and imprisonment costs, as well as crime and its medical and financial impact on victims, if it is used as a form of diversion from CJ – including prosecution, or on a post-conviction basis, as a diversion from likely incarceration.

With the benefit of this evidence, and increasing dissatisfaction with the rising costs of prisons, restorative justice offers the entire UK a 21st-century alternative response to the challenge of crime in a free society.


  1. A tale of three RJ conferences


A tale of three RJ conferences

This section presents three examples of the most tested form of RJ, face-to-face conferences. The examples all show how victims are helped, even when – as in the first example – the offender’s pattern of violent crime continues unabated. The cases are not intended to be representative, except of cases with the most serious injuries. These cases illustrate both the possibility and the benefits of using RJ in such cases, against conventional wisdom. The cases also show, by contrast to conventional justice procedures, just how much does not happen – that could happen – in the UK to deal with the emotional impact of crime on the social network of caring people connected to offenders and victims.

  1. A robbery in London

Two crime victims sat facing each other in HMP Holloway one October night in 2002. The younger one – call her Natalie – had been raped at age 19 and 21, and indecently assaulted when she was eight years old. The older one – call her Carol – had been violently robbed of her handbag, with a glass bottle smashed over her head; her injury required over 70 stitches. Natalie (the rape victim) was 21 years old. Carol (the robbery victim) was 56. Both were Afro-Caribbean women. The reason the two victims sat across from each other in prison was that Carol’s robber was Natalie.

They met that night to discuss the harm caused by the robbery, to which Natalie had just pled guilty in Crown Court. Carol never knew that Natalie had been a crime victim, let alone been raped. She did not know whether the meeting would affect the length of prison sentence Natalie would receive. Carol did know that for seven months, since Natalie robbed her, Carol had been unable to return to her job as a caregiver in a nursing home. Not since the attack had Carol carried a handbag, nor even left her home. Carol was suffering from severe post-traumatic stress symptoms, not unusual for victims injured in crimes. Such stress can foster disease and reduce the quality of life. Thus her daughter and her daughter’s partner went along as her “supporters” for the conference, at the invitation of Nick Cole, the experienced Met police officer who organised and led the event.

PC Cole opened the conference by introducing everyone present, and said that the group was there to discuss three questions:

  • What happened?
    • Who was affected by it and why?
    • What should be done to try to repair the harm?

At the end of the meeting, he said, the group would prepare a written statement of what Natalie had agreed to do. He and Natalie would sign the agreement, and submit it to the Crown Court judge in the case. The judge was free to ignore the statement or take it into consideration as a possible “mitigation” of the harm done by the crime.

PC Cole turned to Natalie to describe what she was doing that night in May, and how the robbery had happened. Natalie said she was using crack cocaine at the time, and needed money to buy her next hit. She was looking for someone to rob. Carol was the first woman with a handbag that Natalie saw. She had tried to grab the handbag and run, but Carol had held on to it tightly. So Natalie had hit Carol over the head with the bottle to get her to release the bag. Carol had let go, and Natalie had run off.

PC Cole asked Carol if she wanted to add anything to that description. Carol said not a word. She sat there with her arms folded and her head facing down, showing no emotion, as if in a catatonic state. But Carol’s daughter quickly chimed in, describing the injuries in great detail. She also described the sudden change in Carol’s life in the aftermath of the crime.

When PC Cole asked “who was affected and how”, Carol’s daughter described the ripple effects of the crime on the many family members supported by Carol as head of the family. More effects were described by Natalie’s grandmother, who was horrified to hear the details of the crime. “Suppose somebody had hurt me the way you hurt Carol,” said Natalie’s grandmother, looking directly at Natalie. “Would you want somebody to do that to me?”

Now in tears, Natalie said she would not want anyone to do that to her grandmother, or to Carol. She said how sorry she was that she had done it. She stood up and crossed the circle to kneel directly in front of Carol, who she asked to forgive her. But Carol remained silent, eyes down.

Natalie pleaded again for forgiveness, saying it was just the drugs; she had never meant to hurt Carol. That led the discussion to PC Cole’s third question, “What should be done to repair the harm?” Carol’s daughter stressed the need for drug treatment, which Natalie’s family supported as well. Someone also raised the idea of an anger management course, which Natalie said she would be happy to attend.

Everyone assumed that Natalie would be sent to prison for the robbery. They all hoped that prison would provide a chance for drug treatment and rehabilitation. Natalie promised she would work hard at her treatment. But no one discussed just how much Natalie had to do to turn her life around.

New Scotland Yard’s first crime report on Natalie named her as the victim of an indecent assault at the age of eight. The next report named her as an offender arrested for assault at age 16. Between March 1997 and May 2002 – about five years – Natalie was arrested 28 times in London, averaging over five arrests per year. Four of those arrests were for robberies, including the May 2002 attack on Carol. Natalie was also named as the victim in these crimes and dates: a rape in 2000; an assault in 2000; another rape in 2002.

Natalie had received a series of community sentences (probation) for her previous crimes, culminating in an 84-day prison sentence from which she had been released just 75 days before robbing Carol on 25 May 2002. She was arrested for that crime on 1 June, and had been in prison awaiting trial and sentence for exactly five months on the night of the restorative justice conference. The question for her, and everyone else present that night, was whether Natalie’s remorse for having hurt Carol so badly would be enough to help her turn her life around.

The question, in other words, was what would happen by placing the present victim – Carol – at the centre of a face-to-face process of justice, one that is far more up close and personal than the formal rituals she had already been through in a Crown Court. Would it prevent Natalie from committing more crimes? Would it help Carol to heal the trauma that had kept her from going back to work?

The answer was that the process helped Carol immensely, but had no apparent effect on Natalie. Carol had not spoken throughout the RJ conference. When it was over and the participants stood up to take tea and biscuits, Carol stayed in her seat. After general conversation was under way, Carol suddenly said loudly, “Young lady, come over here!” Natalie stood before Carol, who took her hand and started to pray that Natalie would turn her life around. Natalie, once again moved to tears, promised she would.

The items in the outcome agreement were as follows:

  • Natalie to receive help for drug problem.
    • Natalie to attend anger management course.
    • Natalie to contact victim when ready (contact details through PC Cole).
    • Natalie promises to behave herself and become a better person.

Carol went out the next day to buy a handbag. Then she went back to work. The RJ conference had restored and transformed Carol’s life.

It had not changed Natalie’s.

Natalie was sentenced to five years in prison. With time off for good behaviour and time served before sentence, she was released 24 months later, in late 2004. Six weeks after her release, she was arrested for yet another robbery. As of autumn 2006, she was still in prison.

  1. A serious assault in Canberra

Ten years ago in the Australian capital city of Canberra, two heroin addicts sat in a circle with a uniformed Australian Federal Police officer and a Protestant pastor. They were discussing an assault by one addict that had almost killed the other, knocking out his teeth and causing massive loss of blood.

The crime had occurred shortly after one heroin addict, Bob, had

been released from prison. He was filled with ideas of revenge. He had heard that while he was in prison his girlfriend had been raped by the other addict, Sam. Bob knew Sam slightly because they both patronised the same heroin dealer. Bob came looking for Sam on the day of his release. When he found Sam, months of rage fuelled the beating Bob gave him. He left Sam with broken bones, his front teeth all knocked out, bleeding profusely. Sam lost three litres of blood and would have died if he had not been found in time. Bob was soon arrested. He most certainly would have gone back to prison if he had been charged and found guilty in court of this assault. Instead, police decided that an RJ conference might work better. After submitting the case to the randomised controlled trial of RJ at the Australian National University’s RISE (Reintegrative Shaming Experiments) Project, the case was randomly assigned to a conference.

Usually RJ conferences are attended by the family and friends of both the victim and the offender, but Bob and Sam had no one who wanted to come with them. Only the church pastor, Joe, who knew both of them because of his work with the drug-using community, offered to come and to support them both.

In the conference Bob freely acknowledged responsibility for the injuries he caused to Sam. He had no remorse at all. The RJ police facilitator asked Bob what he had been thinking when he attacked Sam. Bob explained about his girlfriend (with whom he had broken up since the time of the assault). Sam was sheepish about this accusation – the first explanation he had heard for what happened – and responded by saying, “I didn’t go out of my way to rape her.”

An hour’s discussion ensued in which Bob insisted on the moral rightness of his actions, while Sam spoke only of the injuries he had suffered. In Sam’s own estimation, the attack had been completely unjustified, despite his admitted rape of Bob’s former girlfriend. Sam was particularly concerned about his teeth – he had found out they would cost A$3,000 to repair. When he said he wanted A$3,000 from Bob, Bob laughed.

Bob knew, however, that if this RJ conference failed, he would have to go to court – and almost certainly go back to prison. In the time between the incident and the conference his anger had cooled and he was now feeling some regret, if not remorse. Sam, for his part, knew that if the conference failed he would be living in fear of the next time he encountered Bob – a very likely event given their way of life, and use of the same heroin dealer. Sam trusted neither his own feelings of anger and vengefulness, nor Bob’s, in that eventuality.

Finally Joe, the pastor, proposed that Sam would forgo his claim against Bob for monetary restitution of damages. In return, Bob would agree always to stay 500 metres away from Sam, even when they were both visiting their drug dealer. Six weeks after the conference Sam said he was still unhappy about his teeth, but enormously relieved that there had been no further trouble

with Bob. Nor did he expect there to be. Five years later, neither Bob nor Sam had been arrested again in Canberra – for any offence against anyone.

  1. Another robbery in London

One night in December 2004, a young man walked into a room in a Highgate (London) police station with a young woman and their baby. Sitting in the room awaiting their arrival were a middle-aged couple, Anne and Terry, as well as a criminologist (Heather Strang) and two male cameramen. Two cameras were set up (by advance consent of all parties) to videotape the conference.12 A Scotland Yard police constable in plain clothes, PC Mark Davies, introduced everyone and began an RJ conference.13

Anthony and his partner, Christy, had fallen on hard times. Both were young newcomers to London and had faced many difficulties since the birth of their son. Anthony had worked in the building industry until an accident resulted in his being off work for several weeks prior to his offence. Both he and Christy had also been the victim of robbery, in separate incidents. Their social isolation meant that there was no one they could turn to for assistance when their precarious situation led to a financial crisis. Anthony’s attempted robbery of Anne’s purse in the street near where both of them lived was amateurish. He was quickly apprehended.

When Anne, a journalist, and her husband, Terry, a school headmaster, arrived at the RJ conference they were extremely sceptical about Anthony’s motives for meeting them. They were unimpressed by his repeated apologies and said they believed that Anthony had committed many such offences before this one. Anthony, with his many tattoos and body piercings, may have given the impression of prison experience. This robbery was, however, his first recorded offence. In expressing his remorse, he said he could well understand that Anne and her husband would feel quite unforgiving. That is why he wanted to tell them face to face how very sorry he was for causing them trouble and pain.

Anne responded that she had not been traumatised by the incident, but she felt angry about it because it diminished feelings of trust and confidence in the local community. At this, Anthony’s partner, Christy, said that she had felt the same way when she had been the victim of an attempted street robbery. Christy’s robber had hurled their baby in his pushchair into the road. This was why she had been so angry at Anthony for his actions.

Christy went on to describe her own feelings of outrage and disgust when she found out what Anthony had done. As Anthony cradled their sleeping child in his lap, she described eloquently the consequences of the incident for their relationship: she said that she had no patience at all with Anthony’s excuses for his actions, even though she acknowledged their poverty and anxiety about the welfare of their son. As far as she was concerned there was no possible justification for the robbery

attempt. She had come close to leaving Anthony, despite her own desperate circumstances, because she found it so hard to forgive him. She then said that his extreme remorse gave her some hope about their remaining together because she loved him for that, though she did not expect Anne and her husband to set any store by it.

Christy’s description of the state of her relationship with Anthony proved to be a powerful turning point for the conference. After a moment’s silence Terry turned to Anthony and said: “You may go to prison on Monday, but you are a very lucky man.” Anne and Terry agreed that the best outcome of the conference as far as they were concerned would be for Anthony to get himself some job training and to get back into employment as soon as he was able. They hoped very much that he would not receive a custodial sentence and that he and Christy would be able to survive the stresses and strains of their present circumstances. As one viewer of the videotape, Malcolm Gladwell, quoted their exact words in the New Yorker:

“If there is anything I can do, please say it,” Anthony says.

“I think most of what you can do is between the two of you, actually,” Anne says to Anthony and Christy. “I think if you can put your lives back together again, then that’s what needs to be done.” 14

After the conference, the Crown Court judge in Anthony’s case was informed that there had been an RJ conference, and that it had been filmed. He asked to see the three hours of videotape from both cameras: only rushes were available, so the judge watched six hours of conference proceedings. In view of Anthony’s remorse and previous good character, the judge said in court that he had decided for the first time in his career not to impose a custodial sentence for a robbery. Twenty-one months later there was no record of Anthony having reoffended.

  1. What can we conclude from these cases?

Little can be concluded from three examples about the average effects of RJ on key outcomes. But much can be concluded about the nature of the process itself, and the central importance of narrative in the emotional processing of the aftermath of a crime. As Malcolm Gladwell wrote of the third case (Anthony the robber) in his review of a book about how people crave, offer and understand explanations for events (Tilly, 2006):

Watching the conference is a strange experience, because it is utterly foreign to the criminal process of which it is ostensibly a

  1. This case was excluded from the sample of the randomised trial in order to minimise any heterogeneity in the way in which RJ was delivered, since it was the only case to be videotaped.
  2. The resulting videotape of this conference was viewed by New Yorker magazine writer Malcolm Gladwell (author of The Tipping Point) and analysed in that publication in 2006, the text of which (called “Here’s Why”) is posted online at

part. There is none of the oppressive legalese of the courtroom. Nothing is “alleged”; there are no “perpetrators”. The formal back-and-forth between questioner and answerer, the emotionally protective structure of courtroom procedure, is absent. Anne and Terry sit on comfortable chairs facing Christy and Anthony. They have a conversation, not a confrontation. They are telling stories, in Tilly’s sense of that word: repairing their relationship by crafting a cause-and-effect account of what happened on the street.

Why is such storytelling, in the wake of a crime, so important? Because, Tilly would argue, some social situations don’t lend themselves to the easy reconciliation of reason and role …

Tilly argues that [conflicts between narratives and codes] are endemic to the legal system. Laws are established in opposition to stories. In a criminal trial, we take a complicated narrative of cause and effect and match it to a simple, impersonal code: first-degree murder, or second-degree murder, or manslaughter. The impersonality of codes is what makes the law fair. But it is also what can make the legal system so painful for victims, who find no room for their voices and their anger and their experiences. Codes punish, but they cannot heal.

So what do you do? You put Anne and her husband in a room with Anthony and Christy and their baby boy and you let them talk. 15

Gladwell’s analysis does not apply fully to all forms of restorative justice; only the face-to-face conference allows the chance to “let them talk”. But the craving for a “story”, an explanatory narrative for why and how a crime happened, may be present in almost all victims. Similarly, the craving to offer an explanation may be present in many or even most offenders, once they come into direct contact with their victims. To the extent that explanations do help to repair relationships, this may explain why victims derive so much benefit from (at least face-to-face) RJ.

To the extent that giving such explanations to victims may encourage remorse by offenders, it may be one of the most emotionally powerful tools in the armament of criminal justice. Whether these effects are differentially produced, if at all, from different kinds of RJ remains an unanswered question. The question we can answer is what varieties RJ can take in the modern world.

  1. Gladwell at


  1. Varieties of restorative justice


Varieties of restorative justice

Restorative justice means different things to different people. The evidence on repeat offending in this report focuses heavily on the use of face-to-face meetings of victims, offenders, and their respective families and “supporters”, in response to what is recognised by all those parties as a criminal act. The purpose of this section is to clarify that focus in light of other approaches to RJ, regardless of the evidence or theoretical power that may be associated with them.

This section briefly describes some varieties of restorative justice found in the world today, as well as throughout human history. It does so by focusing on the major dimensions on which RJ varies – the fundamental choices people make about who is to be brought to RJ, for what kinds of actions, in what kind of relationship to conventional justice – rather than by attempting to describe the thousands of RJ programmes in use today in India, China, South Africa and many other parts of the world.

But underlying all of these dimensions is a definition of RJ that has been widely accepted across the modern world, and adopted in publications and research grants by the Home Office. That definition builds on the ancient customs of human societies from Europe to New Zealand, from the Arctic to Arabia, from Kenya to Bangladesh. Just as those ancient customs varied on such process dimensions as the involvement or exclusion of women, or such outcome dimensions as the cost of reparations to be paid in compensation for the harm caused by different kinds of crimes, the modern revival of RJ varies widely on similar dimensions of process, scope and outcome.

What all definitions of restorative justice share is a common moral vision: that justice requires more than the infliction of a “just dessert” of pain on an offender. Whether RJ should include such pain is a point of controversy and difference across the varieties of RJ. But the evidence suggests that RJ starts with the common premise that offenders should somehow try to make the world a better place than they left it after they committed a crime. Once that cornerstone is laid, RJ can then be built in two major ways: by authorities imposing on the offender an act of restitution, or, less controversially, by a process involving the people most affected by a crime – including the offender, whose consent to a collective agreement is considered essential to the definition of RJ. However RJ is achieved, it is consistent with the moral vision of redemption through acknowledgement of responsibility for having caused harm.

  1. One definition of restorative justice

The most generally used definition of “consensual” RJ is now almost a decade old (Marshall, 1999):

… a process whereby all the parties with a stake in a particular offence come together to resolve collectively how to deal with the aftermath of the offence and its implications for the future.

This definition is far from universal, but it serves as a starting point. There are many varieties of RJ that achieve what Marshall describes. These include:

  • face-to-face conferences of victims, offenders and stakeholders;
    • face-to-face mediation (without supporters present);
    • indirect, “shuttle diplomacy” mediation;
    • victim-absent discussions with offender and supporters about crime;
    • offender-absent discussions with victim and supporters about crime;
    • sentencing circles led by a judge (First Nations people in Canada).

Notably, Marshall’s definition of restorative justice does not include a penalty of court-ordered restitution to a victim, which is central to the meaning of the term in many locations, and in many of the empirical tests of RJ. Thus the extent to which RJ represents a voluntary agreement by an offender is the first of many dimensions found in the actual use of the concept of RJ by people processing crimes around the world. Some, like the authors of the UN standards, would say that it is not RJ if it is not voluntary. Empirically, however, it is necessary only to classify each test correctly on this dimension, and to call the reader’s attention to this distinction in reading and interpreting the evidence in tables 1 to 5.

  1. Major dimensions of restorative justice
  2. Scope: criminal versus non-criminal

RJ is a conflict resolution technique not confined to incidents defined by law as criminal. It is increasingly used in school and workplace settings. It is also the basis for attempts to resolve the consequences of transgressions by the state through truth and reconciliation commissions, most notably in South Africa.

  1. Standing: diversion versus supplementation

RJ in criminal justice has been most commonly used to divert offenders from the formal justice system. These are usually, but not always, juveniles, often first-time offenders, who have committed minor offences. In some cases (see section 2) diversion has also been used with extremely serious adult cases. RJ is also used in addition to normal prosecution through the criminal courts for offences too serious, given our current state of knowledge, to be dealt with exclusively by RJ.

  1. Whose harm: personal versus collective victims

While the theory of RJ predicts it will work most effectively when personal victims are involved, RJ is also feasible when the victims are a collectivity. Collective victims include both corporations (such as with shoplifting) and communities (such as with disorder offences or vandalism of such public property as schools or parks). We note that stakeholders to crimes are defined by


their emotions, regardless of a crime’s physical impact on them or on property that they own. Some might say that certain stakeholders in a collectivity are really “personal victims” of a crime, if the crime caused them enough emotional harm. One example is the zookeepers at a Midlands zoo where youths abused and killed a wallaby. The deep pain felt by the zookeepers led them to participate in a police-led RJ conference with the offenders.

  1. Process: face-to-face versus all other methods

Face-to-face RJ entails direct deliberation among those affected by a crime. The emotional power generated in this way may be critical to the process. When one or all of the parties are unwilling to meet, however, indirect mediation or discussion with only the victim or the offender is often used to process the crime.

  1. Convening: negotiator or facilitator

Victim-offender mediation programmes typically involve a mediator whose responsibility is to negotiate between the parties to achieve a satisfactory outcome. This may take place with both parties present or by the mediator shuttling between them. Conferencing programmes typically involve a facilitator whose responsibility is to provide a forum in which the parties can negotiate directly with each other. The facilitator’s role is to ensure that the parties stay focused on the reasons for the meeting and remain civil in their communication. The facilitator is not expected to participate or lead the substance of the discussion.

  1. Outcome: restitution-centred versus other restoration

RJ in the form of court-based restitution programmes and some victim-offender mediation programmes usually focus on financial restitution to the victim. These typically pay less attention to victims’ emotional harm.

3.3 Why focus on face-to-face RJ?

The evidence and the vision considered in this report are primarily, but not exclusively, about face-to-face RJ. Given the limits of length in a document addressed to busy policy makers, the report largely takes this focus for three reasons. First, the principles of the moral vision of RJ to strengthen interdependent communities resonate more with face-to-face RJ than with any other approach. Second, a preponderance of unbiased evidence about RJ comes from the face-to-face variety. Third, the criminological theory derived from the evidence on patterns of criminal offending and desistance over the life-course predicts that only face-to-face RJ could provide an experience with enough emotional power to substantially reduce repeat offending.

3.4. Stages in the criminal process: inventory of face-to- face RJ models in relation to conventional justice

Within the domain of face-to-face RJ, a further variety of settings for RJ have emerged in relation to the criminal process. These include, but are probably not limited to, the following models:

  • police diversion from prosecution to a caution, even for very serious offences (such as the Canberra RISE model);
  • in youth justice, as a way of delivering standard cautions, reprimands or final warnings, or referral panels;
  • diversion from prosecution by prosecutors (for example, conditional cautioning under the Criminal Justice Act 2003);
  • diversion from prosecution by court order or statute (for example, various Australian mainstream models, the New Zealand mainstream model for juveniles and to a more limited extent for adults);
  • “sentencing circles” combining elements of traditional First Nations conflict resolution techniques with formal court- based justice (in Canada);
  • victim-offender mediation and victim-offender reconciliation programmes (scattered throughout Europe and the US, often community-based);
  • conditional caution preserving possibility of prosecution (UK test model);
  • post-conviction, pre-sentence (UK test model);
  • post-sentence, case management (UK test model);
  • post-sentence, pre-release (UK test model);
  • consensual processes not brought to the attention of the formal justice system (UK restorative policing).

3.5 Summary

The label of “restorative justice” has been applied to a wide range of programmes, with one common element: offenders doing something constructive to make the world a better place, rather than simply having pain inflicted on them. This broad definition is an opportunity for flexibility as well as for confusion.

Confusion results when, for example, the term “restorative” is used to describe prison inmates cleaning up a rubbish-filled lot, or when juvenile delinquents are required by a youth offending team referral panel to remove graffiti from park benches. While it is true that in such activities offenders are in some sense “restoring” the world to the status quo before the rubbish or graffiti was there, such efforts do nothing to put the victim at the centre. That is, the restoration to the community has no connection to the restoration to the actual victims of the offenders performing such activities.

From the perspective of the personal victims of these offenders’ crimes, such community service per se offers little restoration of the emotional or psychological state the victims were in before the crime occurred. Even from the perspective of the offenders themselves, the experience of such “constructive” work is unlikely to cause an emotional revelation of the moral truth that harming other people is wrong. And from the perspective of the loved ones and friends – of both victim and offender – who were affected by the crime, the mere act of community service has not been tested to see if it can mitigate the psychological harm of crime.


The kind of RJ that could have wide-ranging benefits for everyone directly affected by a crime requires more than community service. Restoring victims, reducing repeat offending, and building community commitment to the rule of law are all part of the moral vision of RJ. The evidence about whether RJ can achieve those results is largely limited to the variety of RJ designed to achieve them: face-to-face meetings. That design is

also consistent with criminological theories based on thousands of criminal careers about “epiphanies” and “narratives” about who ex-offenders are and what they do (Maruna, 2001; Giordano et al, 2002). For these reasons of goals, evidence and theory, most of the report focuses on what we know about face-to-face restorative justice: its processes, compliance levels, and long- term effects on offenders and victims.


  1. The process of restorative justice


The process of restorative justice

RJ depends entirely on a process of bringing offenders and victims together, either in person or by communication through third parties. The process consists of a sequence of three key stages: finding cases, getting consent and reaching agreements. When RJ is used as a supplement to CJ processes, a fourth stage may arise: the transmission of an agreement to criminal justice decision makers.

In some tests, this process has been completed in substantial volume, despite the lack of public familiarity with the idea. In other tests, the process has been mired in the first or second stages. Either no cases can be found (or referred by officials), or offenders and victims refuse to consent.

These problems might be due to sheer start-up inertia. They may fade away if RJ is pump-primed with enough cases to become commonplace to the public (and perhaps appear in television dramas). Whatever the evidence shows to date on the extent to which victims and offenders are willing to communicate, it is reasonable to predict that take-up rates for RJ should increase steadily as public awareness of the process increases. As in any other “tipping point” of social change in technologies, from fax machines to mobile phones, public acceptance of the technology rises with greater experience and a critical mass of participation (Gladwell, 2000). The problem, to which we return in section 13, is how to get to that tipping point.

  1. Finding cases

Many evaluation reports on RJ in the UK (Meirs et al, 2006; Shapland et al, 2004, 2006) stress the difficulty of recruiting victims and offenders into RJ processes. Where this problem is found, it is usually associated with a decision to test RJ as an “add-on” to an agency or organisation that already has plenty of other work. In contrast, when a team of people has been funded to do RJ – and nothing but RJ – all the time, the case-flow of case recruitment has been far more successful.

At its peak in Canberra in 1995-97, for example, the RJ unit in the Australian Federal Police/Australian Capital Territory was organising and conducting up to five conferences a week in a community of under 300,000 people. This was in an agency with a total strength of some 500 officers, 100 of whom had been trained to lead RJ conferences on request of a central RJ co-ordinating unit. Scaled up in terms of London police numbers, that would equal some 300 RJ conferences a week, or almost 16,000 per year. The production of RJ was simplified by two conditions: 1) many conferences addressed victimless crimes, and 2) AFP procedures required no formal consent of victims – who were simply asked to name a time and date they could attend such a conference. More important, perhaps, was the in-person or telephone conversations police officers had with the victims, offenders and sometimes supporters before a conference occurred, which produced a very high show-up rate for all parties (Strang, 2002).

The most important reason the Canberra RISE project found so many cases, however, was that it was able to work with some 500 arresting police officers as independent decision makers. Each one of the officers was empowered to refer a criminal case to RJ as a diversion from prosecution. Under ACT policies, police constables were given broad discretion, in consultation with the station sergeant in the custody suite booking area, to decide whether an arrest should result in a police caution, a prosecution, or a diversion to RJ. In the UK, by contrast, these powers have now been relocated with the Crown Prosecution Service, which by written policy (and statute) has excluded many of the kinds of cases that were referred to RJ in Canberra.

Similarly, the five youth offending teams that co-operated with the Northumbria police in testing RJ for youth justice agreed as a matter of policy to refer all cases resulting in a reprimand or final warning to the police RJ unit. This provided a very high volume of cases for the specially trained RJ officers to work on, yielding both victim and offender consent in over a third of all property and assault crimes brought to justice without prosecution (Shapland et al, 2006).

Northumbria’s adult cases for similar crimes in magistrates’ courts, in contrast, were far harder to identify. Neither CPS nor court clerks actually implemented agreements to notify the police of eligible cases as convictions were reached. Only after months of trying alternative strategies was the RJ unit able to find an effective link for case referral: the pre-sentence report (PSR) staff of the regional branch of the National Probation Service. Each time a PSR was requested for an offender convicted of an eligible offence type, probation staff faxed a copy of the request to the police RJ unit, which in turn was able to contact the offender and victim to organise RJ cases prior to the offender’s sentencing date.

The Northumbrian adult cases, however, also illustrate the precarious position of RJ programmes in seeking cases without statutory requirements, even when the Lord Chancellor meets with the people concerned. For several years, in 2001-04, police and Home Office officials asked the magistrates’ courts not to sentence cases “there and then” when offenders offered guilty pleas, and where no PSR would be requested – the majority of the adult convictions for eligible cases. Instead, police asked the courts to adjourn the cases for sentencing four weeks later, to allow time to organise RJ before sentence. Court officials said they would do so, but in practice the courts continued to sentence “there and then”. Even a meeting of some 200 magistrates with the Lord Chancellor failed to change the local custom, which made RJ-before-sentence impossible to organise in the majority of eligible cases.

The question of “finding” a case is closely linked to the step in the criminal process at which RJ is being attempted. A time window

deadline is associated with most cases, determined by the date at which the next step in the process must occur by law. An arrest must generally be diverted before the case is prosecuted; a conviction must not lead to immediate sentencing if RJ is to occur pre-sentence; a probation plan cannot be based on RJ unless an RJ process occurs at the very beginning of the sentence; and inserting RJ into a prison pre-release process must be done before the inmate is actually released. The advantage of tying RJ to these deadlines is that they provide a focus, and some incentive, for offenders to co-operate with the process and accept responsibility for their crimes. The disadvantage of tying RJ to these deadlines is that the CJ officials with primary control of the case can use the deadlines to foreclose the possibility of RJ.

While some tests of efforts to refer high volumes of cases to RJ have succeeded in the UK, other tests provide evidence of a need for statutory requirements for referral (see section 13).

  1. Consent to RJ

Once cases can be identified in good time, the UK record on obtaining consent to RJ from victims and offenders is excellent. Compared with the scepticism often heard about crime victims meeting or communicating with their offenders, the rates of victim take-up have been substantial – at least in pilot tests focused on RJ.16 The low rates of victim involvement in routine youth justice, however, indicate the challenge of making RJ succeed without a unit with a special brief for RJ. If RJ is one of dozens of performance indicator targets to be met, it seems unlikely to get top priority. If RJ is the only performance target a unit must meet, however, the track record is far better.

The best track records in the UK so far have been found in adult probation (Meirs et al, 2001) and in the Metropolitan Police-led London Crown Court cases, as well as with youth reprimands and final warnings led by a specialist RJ team in Northumbria (Shapland et al, 2006). In these settings, consent and completion of RJ processes all happen with enough frequency to maintain morale and competence in facilitating the actual content of RJ. This includes the critical work of preparing for a conference or other RJ process, delivering the process to a mutually satisfactory agreement, and then ensuring that the agreements are met by the offender.

The typical process in these units is to offer RJ to offenders first, without saying whether the victim is willing. This practice is based on a desire not to bother crime victims with the emotional work of deciding whether or not to communicate with the offender – unless the offender is willing to communicate and appropriately takes responsibility. Thus in the initial meetings with offenders by RJ facilitators, the primary concern is whether an offender is suitable for RJ, regardless of stated intent. If offenders deny their guilt, express intense anger or give other indications of posing a risk to victims or supporters in an RJ process, facilitators have readily excluded them.

At the same time, there has been no attempt to require evidence of “remorse” as a precondition for RJ. Many offenders agree to RJ in a matter-of-fact way, in a one-on-one meeting with a facilitator. It is when they get into an RJ process that they may become quite emotional and make apparently sincere expressions of remorse and apology. As facilitators often say, “RJ doesn’t screen for remorse; it aims to achieve remorse.” This basis of the facilitator’s consent to have an offender participate has resulted in no documented cases (to our knowledge) of offenders verbally abusing victims, let alone behaving violently, in any RJ process in the UK.

Once an appropriately willing offender has given consent, RJ facilitators contact victims to seek their consent. This is a key issue for “best practice”, since there are at least three ways to obtain victim consent: by letter, by phone call, or in person. Our evidence suggests that best results are likely to be obtained by facilitators meeting in person with victims prior to any RJ process, especially face-to-face RJ conferences. By contrast, it is common youth justice practice to send crime victims invitations- by-letter before a youth offender panel meets with their offenders. This practice yields a low victim take-up rate. It may also yield problems in victim or offender reaction to a meeting with three or more people who had no direct stake in the crime – in stark contrast to Marshall’s (1999) definition of RJ.

Whether victims choose to participate in RJ depends on many considerations: who asks them, in what fashion, and with what kind of priority given to their convenience and emotional state. All these factors appear to influence the likelihood of victims taking part. The resulting take-up rate can vary enormously.

For example, Gehm (1990) found that 47% of the victims from six US victim-offender reparation programmes (VORPs) agreed to meet their offender, after strenuous efforts to reach agreement. Where efforts have not been so strenuous, as with the New Zealand family group conferencing programme (Morris and Maxwell, 1993) or the referral panels operated by the Youth Justice Board (Miers et al, 2001), victim involvement has been much lower.

Within England and Wales, some of the highest victim take-up rates are for the most serious offences. About half of all victims of serious burglary and robbery who were approached (after offender consent) by Metropolitan Police officers in the Justice Research Consortium’s London Crown Court studies agreed to meet their offenders.17 Comparable if somewhat lower rates were found for victims of violent offenders sentenced to prison or probation in the Thames Valley, who were approached by the

  1. In the JRC experiments as of mid-2003, victim take-up ranged from 38% for the Thames Valley prison study to 78% for the Northumbria youth property study.

Offender take-up ranged from 57% for the Northumbria magistrates’ court adult property study to 82% for the Northumbria magistrates’ court adult assault study.

  1. JRC, or the UK studies of the Jerry Lee Program of Randomized Controlled Trials in Restorative Justice.

National Probation Service-Thames Valley/HM Prison Service RJ team for the JRC tests.

The highest overall take-up rates we have found were in the Northumbrian youth justice experiments, in which 39% of eligible youth assault cases and 46% of eligible youth property crime cases resulted in a joint victim-offender agreement to complete RJ. Without a special RJ unit in youth offending teams, however, only 5% to 15% of invited victims attend youth referral panel meetings (Crawford and Burden, 2005; Newburn et al, 2002). This range of response reflects the importance, and structural conditions, of the substantial efforts taken to encourage victim participation in RJ.

For example, Crawford and Burden (2005: 38) report that with youth justice referral panels: “[A]dministrative pressures of time militate against active encouragement of victim attendance, in particular the 20-day timetable for the initial panel meeting as set by national standards [and the fact that] attending panels can place an inconvenient burden upon victims.” Lack of awareness by the general public about RJ may also contribute to the low take-up in youth justice, given the difficulties of suggesting the value of attending an event about which there is little public knowledge.

By contrast, the facilitators involved in organising RJ conferences for the JRC experiments in RJ spent on average 18 hours in organising each successful conference. This included time spent talking to offenders, but in general more time was needed to explain the process to victims. They also arranged conferences as far as possible to suit victims’ convenience. Notwithstanding the need to conduct most of the London and Thames Valley conferences within prison, with all the attendant difficulties of bringing people together in that setting, these efforts paid off in 90% or higher rates of victim attendance as promised. (RJ in prison also has the advantage of allowing RJ facilitators to find non-appearing offenders, who are occasionally too ashamed or fearful of meeting their victim to come out of their cells.)

  1. Reaching victim-offender agreements

Even when RJ facilitators succeed in getting both offenders and victims to agree in principle to undertake RJ, there are a number of reasons why RJ conferences may not go ahead to reaching a victim-offender agreement. While we lack good evidence on indirect RJ, we have ample data on face-to-face processes, which define full success as an agreement completed at a conference with both victim and offender present. For many reasons, this may be harder than simply sentencing an offender in court – a process that is far from perfect due to offender failure to appear. But it is possible to make it easier, as the evidence shows.

It is a common experience of victims in the UK that the RJ event cannot be held at a time and place convenient to them, so they decline to attend. There are also many reasons why offenders prevent face-to-face RJ from being delivered as promised. The

offender may withdraw his/her admissions or guilty plea; something may be discovered to render the case ineligible for RJ; at the last minute the offender perhaps cannot be found; the offender may fail to attend the conference. Most rarely, offender and victim meet but a satisfactory outcome agreement cannot be reached.

These offender compliance (and sometimes facilitator delivery) obstacles in the process pose a major issue for RJ. When, for whatever reason, victims have been promised an RJ conference that is not held, they are often very dissatisfied. In RISE, for example, about a third of victims assigned to CJ described themselves as “angry” about their experience – three times as many as those who completed RJ. But proportionately even more victims were angry if they had been promised RJ that never actually happened.

The evidence suggests that these problems can be solved, at least in the UK, when special RJ units are employed to do face-to-face RJ and nothing else. For indirect RJ, we have no precise evidence about rates of completing agreements given consent. But for the face-to-face RJ tests, we can be very specific and even hopeful. Across the eight Justice Research Consortium tests in 2001-04, a total of 883 cases were randomly assigned to RJ or CJ status. Of the 444 cases assigned to RJ, 372 (84%) were completed with both victims and offenders present for a face-to-face discussion. Of the remaining 16%, some 8% (35) of the cases had no conference at all, 8% (34) had a conference with the offender that the victim declined to attend, and two had a conference with the victim present but no offender. The 92% offender completion rate compares favourably with the 74% offender appearance rates in a London court, as reported in recent years.18

The evidence is clear that systems can be designed to bring the RJ process to completion. The key problem is not whether such systems can be created, but whether sufficient funding and statutory support will be provided to create them (see section 13 below). When the processes are completed, they produce an agreement based on a discussion of the kinds of issues that victims – at the centre of the process – wish to discuss.

Issues for discussion

Here again, we rely on the evidence about face-to-face RJ. In personal discussions among victims, offenders and other stakeholders to the crime, the following issues usually arise. It is less clear whether these issues arise or are discussed to the same degree during indirect RJ processes. But we can say that victims of crime widely share these concerns. They want to have a discussion of:

  • What happened? (This includes “Why me?” or how the offender picked the victim for the crime; the answer that it
  1. Lord Chancellor’s Department Extended Court Sitting Hours: Pilot Final Evaluation Report v 1.0 (21 November 2002), p.ii

was “chance” or opportunistic is usually assuring to the victim.)

  • Who was affected by the crime and how?
  • What should be done to try to repair the harm?

Robust discussions of these questions can last from one to three hours. The first hour can be an emotionally intense period of venting anger. The second hour can be a time for expressions of remorse, apologies and forgiveness. Whatever happens in the beginning and middle of an RJ conference, the third and final phase almost always reaches an agreement on what the offender should do. This agreement has at least two dimensions. One is its emotional meaning for the participants in the room. The other is its symbolic meaning for people outside the room, especially those charged with assessing what it means for community standards of justice.

In a face-to-face conference, the discussion of what should be done can become a dance of mutual altruism. Offenders say they want to help the victim. Victims often say they just want the offender to help himself. Offenders often focus on promises of material reparation. Victims often refuse those promises, asking instead for promises that offenders will seek and complete drug treatment, rehabilitation or education. If no one has any ideas, police facilitators may sometimes suggest some. Canberra facilitators, for example, often suggested community service work for the Salvation Army, a donation of money to the victim’s favourite charity, or a donation of blood to the Red Cross. The facilitator’s iconic power is the one-page “outcome agreement” form on which something must be written. Unless the group agrees on what to write on the form, the facilitator will (usually) not adjourn the conference. Only rarely is resistance by the offender an obstacle to adjournment. More often it may be a sheer lack of ideas or knowledge of available programmes.

The tea break: building social bonds

When an agreement is finally (or speedily) reached, the group’s reward for its work – if its facilitator has been trained by the Australian school of police-led RJ in many of the RCTs (tables 1 to 4) – is a break for tea or coffee with biscuits. While the facilitator writes up and makes photocopies of the agreement, an informal discussion occurs among the participants. This is a frequent opportunity for building social bonds across the divide of offenders’ supporters and victims’ supporters. These bonds may then help to create informal controls that support the offenders’ compliance with the “outcome agreement”.

After the tea break, the facilitator asks the group to sit down again. The facilitator reads the agreement, and asks again if everyone is happy with it. At that point the offender and the facilitator sign

the agreement. In some cases, a victim may sign as well. The agreement then becomes an icon for a promise to keep, a symbol of the offender’s commitment to repair the harm. It also becomes a tool for future monitoring of the offender’s compliance with the agreement. That part of the process, like the completion of a criminal sentence, may take years after the case is “resolved”. Due to its connection with crucial issues of crime prevention, it merits a further discussion later in the report (see section 8).

  1. Reporting RJ to officials

Once an RJ agreement is reduced to writing, it can be used for many serious purposes. From length of sentence to the elements of a probation plan, the RJ agreement can represent a major moral and material part of an offender’s life. But because it is designed to reflect a group consensus in an emotionally intense discussion, it is not intended to look like a court document, a medical report, or a pre-sentence report. It is, rather, as short as a single sentence may be, for example, as written in the minutes of a business meeting. To those who were not present at an RJ conference, it may even appear trivial or insufficient. Crown Court judges in London who have read them for the Justice Research Consortium experiments have been far less impressed with the signed paper promising to, for instance, ”seek drug treatment” than they have by actually observing a videotape of a conference, or even attending a conference in person (but not for a case in their own court).

A less labour-intensive solution (than watching an entire RJ conference) is the kind of report that the Justice Research Consortium used in London, with support from the Esmée Fairbairn Foundation, in the second part of its test of RJ for burglary cases. After consulting with several Crown Court judges about the format for such a report, the JRC consistently followed the protocol used below. What follows is a verbatim copy of what was submitted to the sentencing judge in this case, with only the names of the parties involved being altered to protect their anonymity.

As the case shows, the participants chose to include their views on whether the offender should be sentenced to prison. This issue often arose in pre-sentence conferences, but with a variety of responses. Some victims said they had no desire to influence the sentence, and even said it would be wrong for them to do so. More often, victims were inclined to recommend against custody. After sitting with the offender for up to three hours, sceptics may suggest that this was another example of the “Stockholm syndrome” of victim identification with an enemy. Yet the ways in which victims have talked about this issue suggest that they see it as in their own interest, and society’s, to try to prevent another crime by the offender. They express little confidence that prison will prevent that crime.

Protocol for an RJ conference record, as submitted to sentencing judge

This is a chronological summary of the conference that led to the attached outcome agreement. The report includes some direct quotes and “summarised quotes” from participants. It also includes specific information identified by the judges as important to the report. These are as follows.

  • Remorse expressed by defendant.
    • Degree to which defendant takes responsibility for crime.
    • Defendant’s commitment to change.
    • Victim statements about defendant, offence or conference including:
      • offender remorse;
      • relief at learning the crime occurred by chance with little planning;
      • increased feelings of safety and security;
      • empathy for the offender;
      • views about sentencing.

Conference details

Date of conference: 12-11-2003 Time: 7:45PM-10:45PM

Place of conference: North London Restorative Justice Centre Offence: Burglary

Offence date:

Date of next appearance:


  1. PC Mark Davies – conference facilitator
  2. Joe Bloggs – defendant
  3. John Doe – victim
  4. Frank Sinatra – victim’s friend (1)
  5. Matthew Perry – victim’s friend (2)
  6. Ginger Rogers – defendant’s girlfriend

Produced by conference observer

Sarah Bennett Restorative Justice Centre 407-409 Archway Road London N6 4NW

Tel: 020 8345 2225

Mobile: 07980 834 072

Part 1: What happened

The facilitator introduced all the participants and asked the defendant to describe what happened. The defendant described how he had recently been released from prison and, following an appointment with his probation officer, was walking home with no money to purchase heroin or a present for his son’s birthday. He came to the victim’s house, which appeared to be unoccupied. He knocked on the door and when no one answered, explored the exterior of the house until he located a side window where the frame was in poor repair and was able to gain entry. Once in the house, the defendant used gloves found in the

victim’s house to search through his belongings and ultimately stole a stereo and jewellery. When he left the house, he removed the gloves and dropped them in the victim’s garden. DNA evidence from these gloves identified the defendant four months later. The defendant sold the items for £40 – £20 went towards heroin and he gave £20 to his son.

The victim asked how the defendant could know that no one was at home at the time. The defendant replied that he picked the house for no other reason than that it was empty and he had a “gut feeling” but he knocked on the door to be sure. The defendant said that whilst he was not proud of it, he had a certain skill for identifying empty houses. The victim asked several questions including whether the defendant had committed the crime on his own, how long the defendant had been a prolific burglar and if he had committed any crime since. The defendant replied that he did not have a co-offender and whilst he had committed a lot of crime in the past and had “been in every prison except Holloway” he had not offended since this burglary.

The defendant said he “felt disgusted when I found out it [the house] belonged to an older person”. “If I had known it was a pensioner’s house I would have come right out.” To this the victim’s friend said “it was not the point” whose house he stole from but that he chose to steal at all. When he was caught, the defendant said he felt a huge relief. He explained that he wanted to clean up his life and whilst he had always given “no comment” interviews, he had pleaded guilty right away, against his solicitor’s advice. In the time between the offence and the arrest, he had taken active steps to deal with his heroin addiction by referring himself to a drug treatment programme and was now self-employed as a painter and decorator with guaranteed work with a building company.

The victim and defendant debated throughout the conference what jewellery had been stolen. The defendant agreed that he had taken some of the items listed but “with hand on heart” vehemently denied taking any of the watches, locket or bracelet. The victim said: “I know what is gone … the things that went were of great sentimental value.”

Part 2: Who was affected and how

The facilitator asked all present how they were affected by the burglary at the time and how things had been since. The victim replied that he returned home following an appointment with a hip replacement specialist. He entered the house and saw the box in which he kept his parents’ jewellery on the floor. The victim explained that he was very sentimental and when his mother died three years ago (and his father some years before) he had made a choice to keep their jewellery and possessions in drawers that he used daily so he would always be reminded of them. The victim said he did not sleep the night of the burglary as he was afraid someone might return and therefore spent most of the night attempting to secure the place by screwing down windows, which was very painful given his bad hip. The victim

commented how surprised he was that the defendant had not knocked over, destroyed or vandalised anything in the house. The defendant stated: “If I am not going to take it, I’m not going to break it.”

Since the burglary, the victim said he had paid thousands of pounds to install a burglar alarm and new windows. However, he is “terrified” when he leaves the house and constantly wonders “what am I going to come back to … I don’t see that I will ever relax again.” Despite this expense however, the victim said what he wanted most out of the meeting was to get back the jewellery that had great sentimental value.

The victim’s friend said he was worried about the victim and joked about how the phone bills have been “racking up”. He and his son (victim supporter 2) had driven all the way from Preston to be at the conference and would be returning right after the conference (arriving back home at 4am). The friend went on to say how impressed he was that the defendant had taken such an active step in his own drug rehabilitation and employment.

The defendant’s girlfriend highlighted how she had found out about the offence when the defendant gave her son £20 which she knew he did not have earlier that day. The girlfriend said: “I was hurt at the time; I have a dad your age and I could not stand for this to happen to him.” She had thrown him out of the house but in time he had made real efforts to sort his life out. She highlighted that the defendant “has changed” and she was so proud of the steps he had taken since committing the offence in terms of his drug addiction and employment.

Part 3: What can be done to repair the harm caused

The victim said that he had not been looking forward to the conference as he thought it would “rake everything up again and I would wring your neck”. However, he felt much better for attending and said: “I consider myself to be a good judge of character and you have come across favourably … I am upset that you categorically deny taking the items but if I take what you say as genuine then we have a mystery.” The defendant said: “I assure you, if I had seen a gold watch – I would have seen it and taken it.”

The defendant stated that he had never seen what people go through as a result of his offending and had never met someone like this before. “This has been a real eye opener; I know my apology may be an insult but I genuinely mean it.”

The defendant offered to decorate the victim’s house from “top to bottom”. The victim declined, however, as he said his house is a shrine to his mother and being very sentimental he did not want to change a thing. The defendant offered to paint the exterior of the house, which the victim declined as well as it had been painted this year. The defendant offered to buy a dog for security purposes, but the victim said he likes to travel and might not be able to take it with him.

All participants agreed that the defendant had made active steps to deal with his heroin addiction, a catalyst for his offending behaviour. The defendant stated that he thought it was likely that he would get a custodial sentence. The victim’s friend said “to send you back to prison now would be a waste – you are off drugs, you have a job; I personally think you should be given 12 months’ probation. Going to prison for you right now

… I don’t see the point of it.” The victim said he did “not consider prison to be a deterrent at all … It is to your credit that you have got off heroin.” The defendant replied that the sentencing judge would be looking at all his prior offending and “has to be seen as giving justice”.

The conference ended with all participants shaking hands.

The outcome agreement

  1. Joe has made a commitment to move out of his estate and will start researching areas and letting agents in an effort to do so, after Christmas.
  2. Joe wants to work with diverting youth from drugs. He will consult probation and social services on which schemes are available.
  3. Joe has agreed to write to all participants to apprise them of his progress. This was agreed by all present.
  4. All participants (victim and supporters) agreed that prison is not an appropriate sentence. They felt community sentence would achieve more.

While the narrative report is far more informative than the RJ outcome agreement printed – in its entirety – in the final section on page 41, it is not clear how much of that detail would be absorbed by a Crown Court judge. A survey of Crown Court judges for the Esmée Fairbairn Foundation found that almost half of the judges to whom these reports had been sent could not recall having seen one. While judges may be flooded with paperwork in making sentencing decisions, it is also true that the delivery of paperwork to a judge’s office does not guarantee that it will find its way, in time, into the case file. In the absence of a statutory requirement for such reports to be produced and reviewed prior to sentence, it may be that the use of such reports will fail to have much impact on sentencing practices.

Nor did the outcome agreements themselves have any apparent influence on sentencing, despite the Collins and Barci cases. Several JRC comparisons between the pre-sentence cases randomly assigned RJ and those not found no difference in the

length of sentence. While reduction in custody length was manifestly not a purpose of the project, it was a question raised by numerous participants. Some may even welcome the news that RJ did not affect sentencing, in hopes that it will nonetheless affect future criminal behaviour. At the least, the result controls for sentencing effects, so that any differences in subsequent arrest rates can be attributed to the RJ process as an add-on to otherwise consistent criminal justice practices.

What the role of RJ reports does raise is questions about the rule of law. Since RJ has been accepted by the High Court in relation to the law of mitigating harm as a sentencing criterion, it would seem to be essential for counsel to make any RJ known to the court before sentencing. The fact that this does not always happen does not mean that RJ, per se, violates the rule of law. It does, however, raise one more item for a long list of matters that could be tidied up by a sustained institutional commitment to developing RJ as an enhancement to the rule of law.


  1. Restorative justice and the rule of law


Restorative justice and the rule of law

Restorative justice is an ancient practice that has been codified in Roman law, Sharia law, and many other documents. Restorative justice is arguably the most common approach to law in most societies throughout most of human history (Braithwaite, 1998). Whether that makes it compatible with the principles of the “rule of law” that emerged during the 18th- century European Enlightenment, however, is a different question from its ample precedents.

Lord Justice Bingham has recently addressed the definition of a rule of law.19 He notes that there is a need for greater clarity in both legislative and judicial uses of the concept, and proposes eight “sub-rules” to define the rule of law. Each of these principles, we conclude, is potentially consistent with the use of restorative justice in all of its varieties noted above. As in any other system of administering justice, there is always a potential for mal-administration that breaks the rules of the system itself. But that is a different question from the one put to us by the Smith Institute: whether “RJ is philosophically compatible with traditional approaches to criminal justice, or whether it represents a conflicting paradigm” (emphasis added).20

There is no doubt that RJ represents an alternative paradigm to the conventional approaches to criminal justice of the past century or so. In that respect, it is a possible response to Lord Woolf’s (1999) call for the discipline of criminology to develop a more effective alternative to conventional approaches, which have, in his view, been failing to achieve their goals. We conclude that this paradigm is both a promising alternative to those conventions and a compatible extension of them, under the principles articulated by Lord Bingham:

  1. The law must be accessible and so far as possible intelligible, clear and predictable.

Comment: The statutes prohibiting assault and robbery, burglary, car theft and other crimes are just as clear under RJ as under conventional justice (CJ).

  1. Questions of legal right and liability should ordinarily be resolved by application of the law and not the exercise of discretion.

Comment: RJ is clearly not a system for resolving issues of legal right and liability. It is a system for dealing with the question imposed by decisions of legal right and liability. Whether an offender pleads guilty in court, or tells a police officer he would be willing to meet with the crime victim and take responsibility for the harm he has caused (or “decline to deny the offence”, as it is put in New Zealand), the question of liability for the crime is always resolvable under the rule of law, and not by the discretion of the stakeholders in an RJ deliberation. The discretion entailed in RJ is no different in principle from the discretion found in CJ:

in the decisions to arrest, to charge and to sentence, but not to decide whether someone is liable for a crime despite their denial of liability. RJ simply transfers discretion, after an accused person has voluntarily accepted responsibility for a crime, about “what should be done to repair the harm” to the people vitally concerned with the question. All of the ground rules for that transfer and the decisions to be reached can be established and regulated by case law or statute.

  1. The laws of the land should apply equally to all, save to the extent that objective differences justify differentiation.

Comment: RJ is clearly based upon principles of equality, and may reinforce them in powerful ways through the intensely personal exchanges among parties affected by crimes.

  1. The law must afford adequate protection of fundamental human rights.

Comment: The voluntary and consensual nature of RJ processes has, to date in the UK, consistently allowed for rights to withdraw from the process and return to a CJ approach.

  1. Means must be provided for resolving, without prohibitive cost or inordinate delay, bona fide civil disputes which the parties themselves are unable to resolve.

Comment: While RJ has been evaluated primarily in relation to criminal processes, it could well be extended into alternative dispute resolution for civil matters. Its costs are potentially far lower than litigation, and may provide greater access to dispute resolution.

  1. Ministers and public officers at all levels must exercise the powers conferred on them reasonably, in good faith, for the purpose for which the powers were conferred and without exceeding the limits of such powers.

Comment: This principle has been applied in many RJ projects by police officers, prosecutors and other officials who convene or facilitate RJ resolutions.

  1. Adjudicative procedures provided by the state should be fair.

Comment: To the extent that adjudication is defined as a determination of legal liability (or guilt), RJ does not alter adjudicative procedures at all. To the extent that it may supplant determinations as to sanctions for crime, it is only by mutual consent of all concerned. Whether the community is adequately

  1. Lord Justice Bingham “The Rule of Law”, sixth Sir David Williams Lecture, Faculty of Law, University of Cambridge, 16 November 2006. (
  2. Letter from Wilf Stevenson (30 June 2006).


represented in a stakeholder conference is an important issue that future RJ projects must continue to address. Many systems for insuring community interests, such as the possibility for judicial review of RJ decisions in New Zealand, are available for addressing this concern. Thus RJ is not, in principle, at all incompatible with Lord Bingham’s requirement of fairness, given the existing discretion in sentencing and the current variation in sentences for apparently similar offences committed by offenders of similar backgrounds.21

  1. The rule of law requires compliance by the state with its obligations in international law, the law which whether deriving from treaty or international custom and practice governs the conduct of nations.

Comment: RJ neither entails nor requires a breach of this principle.

The underlying tension between RJ and CJ is not about the rule of law. It addresses a question that Lord Bingham does not even include in his definition of the rule of law. That question is: “Who owns a crime?” As Norwegian criminologist Nils Christie (1977) has suggested, the rise of European monarchy resulted in offences against people becoming transformed into offences against the Crown – and, by modern extension, against the state. In Christie’s argument, the Crown “stole” the offences against the people, and used them to seize assets of offenders for the monarch’s own use rather than to provide compensation to victims.

A 21st-century view of this question need not view crimes as “property” owned by victims or states, but rather as problems to be addressed jointly by nation-states and their citizens most directly affected by specific crimes. In this view, RJ can become a “rebalancing” of the interests of crime victims and the state that may improve public respect for law. RJ may afford the opportunity for maintaining the capacity of justice to protect the community in future, and maintain fairness about past actions. At the same time, it may promote a greater sense of achieving what is fair to victims. The evidence reported here suggests that there is a substantial improvement in victims’ own sense of fairness when RJ is added to the tools of justice than when RJ is not available.

United Nations principles

One reassuring safeguard in best practice for RJ is adherence to the UN principles. In mid-2006, for example, the Northern Ireland Office adopted these principles in its Draft Protocol for Community-based Restorative Justice Schemes, at point 6:

Subject to the other provisions of this Protocol, schemes will adhere to the relevant sections of the UN “Basic Principles on the use of Restorative Justice Programmes in Criminal Matters”, in particular the following:

  • restorative processes should be used only with the free and voluntary consent of the parties (which may be withdrawn at any time);
  • agreements should be arrived at voluntarily and should be reasonable and proportionate;
  • disparities leading to power imbalances, and the safety of the parties, should be taken into consideration in referring a case to, and during, a restorative process;
  • parties should have the right to legal advice about the process;
  • before agreeing to participate, parties should be fully informed of their rights, the nature of the process, and the possible consequences of their decision;
  • neither victim nor offender should be coerced, or induced by unfair means, to participate in the process or to accept the outcome.

Most of the RJ programmes reviewed for this report, and all of the 12 controlled tests of RJ that the authors have conducted in the UK and Australia, have followed these UN principles. They are strongly supported by British (as well as Irish) culture, and would seem likely to achieve high levels of compliance. Even conventional criminal justice (CJ), of course, has occasional lapses, as when a judge falls asleep during a trial. The question of such lapses is not whether to abolish the system of justice, but rather how to hold it accountable to scrutiny and correction. In terms of visibility to anyone besides the offender, CJ rarely has more people in the room than RJ (Roche, 2003: 194-198). RJ would therefore seem to have just as much (or more) potential for detection of violations of relevant rules as CJ.

In sum, we conclude that, on the evidence so far, RJ does not conflict with the rule of law. Nor does it necessarily conflict with the basic framework of common law. What it does offer is an alternative to conventional interpretations of that framework as they have developed in the industrial world. As a public safety strategy for the post-industrial era, RJ may offer better results within the same basic principles. By providing more opportunities for questions and answers, face-to-face or otherwise, it may actually make law far more accessible to the people. The evidence of satisfaction with RJ suggests that it may reinforce the rule of law. There is no evidence that wider use of RJ would undermine the rule of law.

  1. See Lord Woolf’s letter to Crown Court judge David Pitman at



  1. How we know what works – and what doesn’t


How we know what works – and what doesn’t

  1. The logic of cause and effect

One of Britain’s greatest gifts to the world is the scientific and statistical understanding of cause and effect (Salsburg, 2001). The stunning achievements of British thinkers in this domain, including immigrants and citizens from other nations who trained or worked here, is often forgotten in modern discussions of evidence-based policy. While Isaac Newton is well remembered for discovering the uniformity and invariant nature of gravity, less famous British scientists discovered how to reveal causation of the widely variable phenomena that are characteristic of criminal justice.

Foremost among these contributions is the work of Sir Ronald Aylmer Fisher, whose research on agriculture in the 1920s laid the foundation for studying all phenomena that are diverse and variable (including human responses to medicine, music and criminal sanctions). Fisher’s (1935) invention of randomised controlled trials (RCTs) was first employed in medicine in 1948 to demonstrate that a new antibiotic could cure tuberculosis. Since then an estimated 1 million RCTs worldwide have been undertaken to test medical procedures.

The basic logic of causation within diverse populations, Fisher suggested, lies in a comparison of outcomes between members of a population with a certain characteristic and those without it. If all other characteristics are identical (or “held constant”), then any difference in outcomes can be attributed to the one characteristic that differs. Thus if 500 tuberculosis patients are divided into two groups at random (with an equal chance of placement in either group), the groups will usually not differ on the average age, weight or other characteristics that could affect their health. So if one group is given an antibiotic, while the other is not, any difference in death rates between the two groups can be said to be caused by the antibiotic.

While other methods are used to test causation in diverse populations, they are often described as “approximations” of Fisher’s experimental design. The term “quasi-experiment” is sometimes used to describe groups that are selectively matched, rather than assigned at random. Such quasi-experiments vary in their ability to rule out selection bias, or pre-existing differences in characteristics of populations that are the true cause of differences in outcomes (like death or crime), rather than the treatment being evaluated. Such quasi-experiments have been ranked from 1 to 5 on a scale of internal validity (Sherman et al, 1997) called the “scientific methods scale” (SMS) or “Maryland scale” that has been adapted by the Home Office in its reviews of evidence on what works to prevent crime.

The evidence presented on RJ is a mix of RCTs, quasi-experiments and matched comparisons. This review uses the Maryland scale to identify a threshold of results that are reasonably unbiased by self-selection into or out of RJ. As in the Maryland report to the

US Congress on what works to prevent crime (Sherman et al, 1997), the present report uses level 3 on the SMS as the minimum criterion for inclusion in the list of findings included in tables 1 to 5 above. This means that at least one population receiving RJ is compared with a similar population not receiving RJ, with the definition of “similar” excluding the idea that one group accepted RJ but the other group declined. While such comparisons have been reported in the evaluation literature, the decision to reject or accept RJ is so fundamentally related to its effects that it creates a strong suspicion of influencing the outcomes of crime rates or victim satisfaction.

Within the Home Office scale of approximations of Fisher’s randomised experiments, tables 1 to 5 employ two variations on the theme. One is the use of subgroups of randomly assigned cases, identified on the basis of fixed characteristics that preceded random assignment. These include gender and race. While the cases were not randomly assigned with separate randomisation formulae for each subgroup (called “blocking”), they are virtually the equivalent of that procedure from the standpoint of removing self-selection bias.22 The other variation on the Home Office scale is the designation of “4.5” for RCT findings about victims (table 4 above). Because the victims were not themselves randomly assigned to RJ, but were closely tied to the cases that were randomly assigned, we treat the effects of RJ on victims as unlikely to be influenced by self-selection bias, but not as unlikely as if they had been randomised themselves.

Sample sizes are an important part of our assessment of the evidence, although we follow no rigid formula for inclusion or exclusion. Because both the diversity of people within the sample (which statisticians call standard deviation or variance) and the magnitude of effects can affect statistical power, we avoid any minimum sample size as a cut-off criterion; this approach differs from Home Office practice.

Tables 1 to 5 display sample sizes in two ways, in order to emphasise the difference between experimental and non- experimental findings. Where allocation has been based on random assignment of cases, a single sample size (designated by the letter “N”) is shown to indicate the total population allocated – usually with about half the cases placed in the RJ group. Where allocation has not been based on random assignment, the sample size of the treated population is indicated under the “population” column, with the “N” size of the matched comparison sample (if any) indicated under the “comparison” column.

In sum, how we know what we know about the effects of RJ is based on comparisons of similar groups according to the same

  1. As one statistician notes, “… simple randomization in each of several strata is equivalent to randomization overall. Because a simple random process has no memory of previous assignments, stratification serves no purpose …” Piantadosi, 1997: 211.


outcomes, such as future crime. These comparisons are made by making the groups as similar as possible, so that the main difference between them is that one group went through (or at least began) RJ processes and the other did not, after both groups said they were willing to do so.

  1. Searching for eligible tests of RJ

This review is based in part on a systematic search of the following databases from 1986 through to 2005 (the terms “restorative justice” and “conferencing” were not in use prior to 1986):

  • C2-SPECTR;
    • [US] National Criminal Justice Reference Service (NCJRS);
    • Criminal Justice Abstracts;
    • Sociological Abstracts;
    • Criminal Justice Periodicals Index;
    • Dissertation Abstracts;
    • Social Science Abstracts.

The search terms used were “restorative” and “justice” or “conferenc/e/ing” with “reoffending”, or “recidivism”, or “evaluation”. In addition, reviews of the effects of restorative justice on repeat offending and victim reactions, including satisfaction with the handling of their cases, were examined for references. Experts in the field were also contacted. Unpublished evidence is included where available, including up-to-the- minute analyses of the 12 main tests and additional subgroup analyses of the Jerry Lee Program of Randomized Trials in Restorative Justice. The review is also based on qualitative evidence on implementation and theoretical issues. The reviewers searched reference lists, contacted other authors, conducted electronic searches, and examined all reports related to restorative justice in the programme of the American Society of Criminology from 1997 to 2005.

While some databases were restricted to particular periods of time, electronic searches were not otherwise limited by date. Indexes were searched in which non-English publications were expected to appear, but only reports written in English were considered for the review. In January 2005, one reviewer electronically searched 24 databases related to criminal justice, law, and related areas of social science. The most common search was applied to databases indexed by Cambridge Scientific

Abstracts; these databases were searched using the following terms: ((restorative AND (justice OR sentenc*)) OR (mediate OR mediation OR restitution OR conferencing AND ((criminal OR offender OR perpetrator) AND victim))) AND (reoffend* OR recidiv* OR victim) AND (ab=random* OR ab=controll*). All databases searched and the particular terms used to search each database are listed in figure 1.

  1. Selecting the eligible studies

One reviewer checked titles and abstracts to identify studies that could be excluded based on information provided in the title or abstract. When a study could not be excluded based on that information, more information was obtained by retrieving the article or by contacting the authors.

The search identified articles in languages other than English. The reviewers are not aware of any completed or on-going RCTs or level 3 or 4 tests that have not been reported in English, but the reviewers are unable to conclude that none would be identified by combing these articles or by conducting a broader search.

Two reviewers extracted information from the full text of articles when published reports were available. Other information was obtained directly from investigators, including our own on-going work.

In order to update the 2005 search for the present (non- systematic) review, we examined the abstracts or full reports for all of the tests of RJ currently listed by the Restorative Justice Consortium in London. Three reviewers participated in this process, again applying the level 3 threshold to anything that claims a “result” or “consequence” of RJ, or that RJ “caused” an outcome. The surviving studies of all these searches are listed and summarised in tables 1 to 5 above.

In summing up the evidence, this review does not employ meta- analyses, forest graphs or other summary tools for averaging the results of effectiveness research. By NICE standards, there is too much diversity in populations, interventions and comparisons to yield a meaningful average of outcomes. This review takes the more conservative approach of listing results that specify populations and comparisons (summarised above in tables 1 to 5), and discussing any discernible tendencies towards patterns.


Figure 1: Electronic searches

  Academic Index1) (restorative OR mediat* OR restitution OR conferencing) AND victim AND random*; 2) (restorative OR mediat* OR restitution OR conferencing) AND offender AND random*; 3) (restorative OR mediat* OR restitution OR conferencing) AND crim* AND random*  16
Bibliography of Nordic Criminology(“restorative justice” or mediation or conference or restitution) AND (criminal OR offender OR perpetrator) AND (random or randomly or randomized)44
Criminal Justice Abstracts((restorative AND (justice OR sentenc*)) OR (mediate OR mediation OR restitution OR conferencing AND ((criminal OR offender OR perpetrator) AND victim))) AND (reoffend* OR recidiv* OR victim) AND (ab=random* OR ab=controll*)  19
Criminal Justice in Denmark(“restorative justice” or mediation or conference or restitution) AND (criminal OR offender OR perpetrator) AND (random or randomly or randomized)7
  Criminology((restorative AND (justice OR sentenc*)) OR (mediate OR mediation OR restitution OR conferencing AND ((criminal OR offender OR perpetrator) AND victim))) AND (reoffend* OR recidiv* OR victim) AND (ab=random* OR ab=controll*)  47
  Dissertation Abstracts((restorative AND (justice OR sentenc?)) OR (mediate OR mediation OR restitution OR conferencing AND ((criminal OR offender OR perpetrator) AND victim))) AND (reoffend? OR recidiv? OR victim) AND ab(random? OR controll?)  18
International Bibliography of the Social Sciences((restorative AND (justice OR sentenc*)) OR (mediate OR mediation OR restitution OR conferencing AND ((criminal OR offender OR perpetrator) AND victim))) AND (reoffend* OR recidiv* OR victim) AND (ab=random* OR ab=controll*)  0
Index to Current Urban Documents1) restorative justice; 2) conferencing; 3) mediation1
Index to Foreign Legal Periodicals1) kw criminal OR kw offender OR kw perpetrator; 2) kw restorative justice or kw mediation or kw conferencing or kw restitution; 3) #1 & #2  27
  Infotrac Onefile1) (restorative OR mediat* OR restitution OR conferencing) AND victim AND random*; 2) (restorative OR mediat* OR restitution OR conferencing) AND offender AND random*; 3) (restorative OR mediat* OR restitution OR conferencing) AND crim* AND random*  16
NCJRS Abstracts Database (NCJRS Virtual Library)((restorative AND (justice OR sentenc*)) OR (mediate OR mediation OR restitution OR conferencing AND ((criminal OR offender OR perpetrator) AND victim))) AND (reoffend* OR recidiv* OR victim) AND (random* OR controll*)  146
  PAIS International((restorative AND (justice OR sentenc*)) OR (mediate OR mediation OR restitution OR conferencing AND ((criminal OR offender OR perpetrator) AND victim))) AND (reoffend* OR recidiv* OR victim) AND (ab=random* OR ab=controll*)  0
  PILOTS Database((restorative AND (justice OR sentenc*)) OR (mediate OR mediation OR restitution OR conferencing AND ((criminal OR offender OR perpetrator) AND victim))) AND (reoffend* OR recidiv* OR victim) AND (random* OR controll*)  3
PolicyFile1) Keyword: restorative justice; 2) Keyword: conferencing; 3) Keyword: victim mediation8
Political Research On-line  Subject: restorative justice  10
  PsychInfo((restorative AND (justice OR sentenc*)) OR (mediate OR mediation OR restitution OR conferencing AND ((criminal OR offender OR perpetrator) AND victim))) AND (reoffend* OR recidiv* OR victim) AND (ab=random* OR ab=controll*)  37
Social Services Abstracts((restorative AND (justice OR sentenc*)) OR (mediate OR mediation OR restitution OR conferencing AND ((criminal OR offender OR perpetrator) AND victim))) AND (reoffend* OR recidiv* OR victim) AND (ab=random* OR ab=controll*)  2
  Social Work Abstracts((restorative AND (justice OR sentenc*)) OR (mediate OR mediation OR restitution OR conferencing AND ((criminal OR offender OR perpetrator) AND victim))) AND (reoffend* OR recidiv* OR victim) AND (ab=random* OR ab=controll*)  0
  Sociological Abstracts((restorative AND (justice OR sentenc*)) OR (mediate OR mediation OR restitution OR conferencing AND ((criminal OR offender OR perpetrator) AND victim))) AND (reoffend* OR recidiv* OR victim) AND (ab=random* OR ab=controll*)  13
Sociology: A SAGE Full-Text Collection((restorative AND (justice OR sentenc*)) OR (mediate OR mediation OR restitution OR conferencing AND ((criminal OR offender OR perpetrator) AND victim))) AND (reoffend* OR recidiv* OR victim) AND (ab=random* OR ab=controll*)  6
  TRANwebKeyword Anywhere: (“restorative justice” mediation conference restitution) AND Keyword Anywhere: (criminal offender perpetrator) Keyword Anywhere: (criminal offender perpetrator)  4
Worldwide Political Science Abstracts((restorative AND (justice OR sentenc*)) OR (mediate OR mediation OR restitution OR conferencing AND ((criminal OR offender OR perpetrator) AND victim))) AND (reoffend* OR recidiv* OR victim) AND (ab=random* OR ab=controll*)  0
 Total Hits424


  1. Restorative justice in the UK: the four settings


Restorative justice in the UK: the four settings

Recent RJ practices in England and Wales have been developed in at least four major settings: adult and youth justice, schools, and neighbourhood policing.

  1. Adult justice

As recent Home Office reports show (Miers et al, 2001; Shapland et al, 2004, 2006), RJ has been used in multiple settings for adult justice. These programmes have almost all used RJ as a supplement to CJ, rather than as an alternative to it. One of the earliest and most common forms of RJ is in the post-sentencing stage of the criminal process. Probation agencies have been early prime sponsors and advocates of such efforts to bring victims and offenders together. Until the major Home Office initiatives funded in 2001 (Shapland et al, 2004), probation programmes created the greatest volume of adult RJ (Miers et al, 2001). In these programmes, communicating with victims, either by letter or face to face, became part of the process of serving a community sentence.

More recently, RJ has been conducted after a guilty plea but before a sentencing decision (see cases in section 2). In these settings, the RJ process becomes information that a judge can use to decide the appropriate sentence. The fact that RJ occurred or that the offender offered to undertake it can then be treated as mitigation of the harm of the offence.

The most striking aspect of RJ in adult justice is the wide range of offenders and offences for which RJ has been employed. Contrary to the conventional wisdom heard in discussions with government officials as recently as November 2006, RJ has not been limited to minor offences, and certainly not to juveniles. Since the 1980s, RJ has been used in UK cases of relatively serious violence and property crime. From 2001 through to 2005, RJ was used in the UK with robbery and burglary offences, many of them committed by offenders with long criminal histories. Offenders with up to 50 or more prior convictions have met with victims and apologised for their actions, with demonstrable benefits for the victims – including reductions in post-traumatic stress symptoms (see section 9). This programme, however, has been discontinued due to lack of funding.

Even the most widespread form of violence – domestic violence, long considered off limits to RJ by central government policy – has been the subject of RJ efforts for adults in Hampshire. The “Dove Project” has conducted RJ-style family group conferences on the Pennell and Burford (2000) model under the auspices of Hampshire County Council, which has supported up to 600 conferences a year. While it is not clear from published sources how cases are referred, one Hampshire Constabulary report describes the Hampshire model as the one used previously with effective results in Canada:23

  • Stage 1 – the family (includes wider family and friends) meet

with professionals and information is shared regarding the incidents and problems surrounding the violence.

  • Stage 2 – the family have a time together in private without professionals present, to plan ways of making the home a safer place.Stage 3 – is a chance to confirm the plan with the agencies to ensure it is supported by all present.

These conferences appear to be led by non-profit community organisations under grants from Hampshire County Council.24 And as noted in table 1, a level 4 Canadian evaluation demonstrated substantial reductions in family violence relative to similar families not receiving RJ, the results of which were presented in the constabulary report in 2002.

Other RJ efforts may be under way across the UK, well under the research radar. There may well be, for example, traditional forms of RJ being used in communities of South Asian Britons, even with criminal offences that may or may not have been reported to police. The robust tradition of RJ in such communities is not always consistent with the rule of law, especially if it results in agreements for imposing physical punishments on offenders. But in a related example, there was a public outcry across Australia over a youth agreeing, in an RJ conference, to wear a T-shirt saying, “I am a thief.” The challenge of law is not necessarily to foreclose such community justice, but rather to monitor and regulate it under nationally agreed principles.

What is not clear in the UK is the future of diversion from prosecution in adult RJ. While the Home Office (2003) consultation document and the Criminal Justice Act 2003 adopted it in principle, diversion on a wide scale has been difficult to achieve. This might have been easier to accomplish when charging decisions were still made by police rather than by co-located Crown Prosecution Service lawyers. But the advent of conditional cautioning, based on offenders’ agreement to undertake RJ, coincided in history with the transfer of charging decisions to the CPS (in early 2004). There may be some RJ under way on this basis around England and Wales, but if so it has not been visible. With half the offences in the country, London’s CPS was barred by national policy from pursuing a high volume of cases as diversion to RJ. These issues are considered further in section 13.

7.2 Youth justice

The Crime & Disorder Act 1998 was intended (among other things) to make RJ a major part of youth justice in the UK. The act also provided for new institutions, including the creation of

  1. Chief Inspector Paul Bright, Hampshire Constabulary, Report of a 2001-2002 Fulbright Scholarship, found at 1159FF386AFA/0/Fulbright.pdf

multi-agency youth offending teams (YOTs) to manage young offenders and to convene community leaders in “youth offender referral panels” to help guide community sentences to delinquents. But the act also required the Youth Justice Board and YOTs to manage custodial sentences of young offenders in HM prisons, with many other tasks related to court convictions. In general, the evidence suggests that these other missions have created constraints on the capacity of YOTs to deliver RJ to victims, especially in first offender cases receiving reprimands and final warnings. The evidence on these questions is reviewed below (see section 9) in the discussion of victim effects. It is fair to say that youth justice in the UK is not predominantly delivered in ways that are based on RJ principles, although there is ample statutory room to introduce them on a case-by-case or more general basis.

7.3 Restorative justice in schools

Perhaps the least-known arena for RJ lies at the borders of criminal and youth justice, in the management of antisocial behaviour in schools. Because it lies outside the criminal process, we go beyond mere description in this subsection in order to offer a more complete picture of what is known about the varieties of school-based RJ, as well as what little is known about its effects on students’ behavioural problems.

Persistent problems in the school environment have long been addressed outside the criminal process by using restorative techniques. In recent years, those approaches have become more formalised as comprehensive and replicable programmes for addressing a wide range of school problems. These include bullying, truancy, disciplinary issues arising from defiant and disrespectful behaviour, everyday conflicts, disputes and petty crimes.

Following a programme introduced into Australian schools in the mid-1990s (Cameron and Thornsborne, 2001), similar RJ interventions have been tried in the US and the UK. These range from proactive, informal practices in everyday interaction between students and staff – usually known broadly as restorative practices – through to full-blown RJ conferencing to address specific and serious incidents of harm in the school environment (Wachtel and McCold, 2001; Hopkins, 2004). A number of programmes have been devised (mainly in the US) to teach RJ principles for managing relationships within schools. If a public health framework is employed, these can be described as primary-level interventions, trying to prevent harm in as-yet- unharmed populations. The Collaboration for Academic Social & Emotional Learning, based at the University of Illinois at Chicago, was established in 1994 by Daniel Goleman (author of Emotional Intelligence). It has described (but not evaluated) many of these programmes on its website (see

Although there have been high hopes for RJ in schools, problems have frequently been encountered in implementation. Traditional disciplinary approaches tend to predominate, even in schools

experimenting with RJ on a limited basis. Anything short of a whole-school RJ approach appears to be difficult to manage and difficult to assess in terms of impact. Evaluation has been patchy and usually not rigorous, but the following programmes have been subjected to some level of evaluation.

RJ as conflict resolution

Some programmes have used restorative principles for responding to conflicts in a non-violent manner, but without the intervention of an RJ facilitator. Instead, students are taught in advance how to deal with any conflicts they may have in future. The Help Increase the Peace Project (HIPP) in Baltimore, for example, takes students through a series of workshops designed to develop mutual respect and understanding. A pre-post evaluation (Woehrle, 2000) indicated that students who completed the workshops dealt more constructively with conflict.

Another programme focused on student conflict resolution skills is the Resolving Conflict Creatively (RCC) programme in New York City. This ambitious K-12 intervention involved the development of social and emotional skills through 25 or more workshops over the course of a school year. The subject of a major pre-post evaluation effort involving 5,000 students, 300 teachers and 15 elementary schools (Aber, Brown and Henrich, 1999, as cited in Morrison, 2005), the programme reduced crime and antisocial behaviour among participants compared with non-participants. There were, however, fewer positive outcomes for boys, younger students and high-risk students.

The Responsible Citizenship Programme piloted in an Australian elementary school also aimed to build school community and resolve conflict peacefully (Morrison, 2001, 2006b). It involved 30 students aged 10-11 years. A pre-post evaluation found that students’ feelings of safety at school increased along with levels of respect, empathy and participation in the school community.

Responding to incidents

More intensive programmes that are focused on incidents of harmful behaviour affecting the whole school community include peer mediation and problem-solving circles, both of which entail a third person initiating dialogue among those directly affected by an incident. This dialogue does not, however, rise to the level of organisation of a professionally led RJ conference.

Peer mediation attempts to employ fellow students trained in mediation techniques as the neutral party to resolve conflict. Despite their wide popularity, with thousands of programmes in existence in many countries, a systematic review shows non-significant or weak effects (Gottfredson, 1997).

Problem-solving circles aim to improve students’ capacity for collective dispute resolution by introducing them to the structures of RJ conferencing and encouraging them to solve

classroom problems for themselves. The evaluation of a programme in an Australian elementary school (Morrison and Martinez, 2001, as cited in Morrison, 2005) compared one classroom (N = 12 students) with two other classrooms acting as controls. Teachers reported a number of benefits to the experimental classroom, including the provision of a safe place to express problems, the open expression of feelings, and better communication and support between students. Students said they thought teachers held bullies more accountable and that bullying had decreased.


The face-to-face RJ conference is the most intense and personal kind of RJ, focused on specific incidents and individuals. In the school setting it has been structured along similar lines to conferences in criminal justice. These conferences have been attended by the person responsible for the harm, the person experiencing the harm, and their respective communities of support. This model has been used in Australia, Canada and the UK. Some limited evidence is available about its effectiveness in resolving school conflicts and disputes.

The first evaluation of school-based RJ conferencing was reported by Cameron and Thorsborne (2001). Describing outcomes from 89 conferences in Queensland, the report details RJ for serious assaults (about half of all cases), bullying, property damage, theft, truanting and other more minor criminal incidents. The report shows positive findings from all participants on measures of engagement with the process, procedural fairness and satisfaction with the experience. It also shows high levels of compliance by offenders with agreed conference outcomes. Morrison (2006a) reports that these results have been largely replicated in studies she cites by: Calhoun (2000) in Calgary, Canada; Hudson and Pring (2000) in Thames Valley, UK; Ierley and Ivker (2002) in Colorado; and Shaw and Wierenga (2002) in Victoria, Australia. Findings reported are from post- conference interviews, with no comparison group.

Another study focused on conferencing for disciplinary problems was carried out by the University of Waikato in 34 New Zealand schools. It found that RJ was successful in reducing the number of suspensions (Dravery et al, 2006), although that measure can be interpreted either as a reduction in conflict or as a reduction in punitive response to conflict. Despite the high level of interest generated by the programme, there has been no systematic introduction of restorative conferencing into New Zealand schools. The authors conclude that conferencing for incidents of crime and discipline in the school environment is not enough. They say that for RJ to be accepted as an alternative to traditional punitive responses, it is necessary for restorative practices of all kinds to be used in all kinds of interactions.

An evaluation of Hampshire County Council’s application of family group conferencing in schools, as part of its extensive RJ programme in social welfare and justice settings, has been

conducted by the University of Sheffield. The programme was designed to address the needs of students involved in truancy, bullying or other behavioural difficulties. It worked through referral to the county council’s Project in Education, rather than being located in the schools themselves. Morrison (2005) reports that the evaluation (cited as Marsh, 2004) consisted of interviews with students, their families and professionals. Most participants were positive about the process and links between home and school appeared to be improved, though only if schools were closely involved with the programme.

Whole-school programmes

Whole-school approaches of the kind recommended by the Waikato team include a variety of RJ practices as primary, secondary and tertiary interventions. The largest evaluation of RJ in schools in the US using RJ practices in staff-student interactions – as well as conferencing for specific incidents – was conducted by the Minnesota Department of Children, Family & Learning (MDCFL, 2002). Its primary outcome measure concerned the effectiveness of RJ at reducing the use of suspensions and expulsions. It concluded that restorative measures of many different kinds could be an effective alternative in addressing disciplinary issues. However, baseline data are incomplete, outcomes varied from school to school and the evaluation reports are so varied in their content that the results are unclear (Morrison, 2006a).

The whole-school approach was also evaluated by a Scottish study carried out by the University of Edinburgh and the University of Glasgow (Kane et al, 2006). The evaluation of RJ trialled in 18 schools of different levels and settings consisted of interviews with staff and students, observation of RJ in operation and surveys of participants. It found evidence of a change in culture towards disciplinary practices in elementary schools but less in high schools, where considerable staff resistance was encountered towards restorative alternatives to punishment.

A national evaluation of a whole-school approach in the UK via the Restorative Justice in Schools Programme was carried out by the Youth Justice Board for England and Wales (YJB, 2004). This involved nine local youth offending teams working with 26 schools across the country. The programme included informal restorative practices as well as formal RJ conferences of varying kinds, though the actual restorative practices varied widely from place to place. Almost 600 conferences were held for incidents of bullying, name calling, gossiping, family feuds, conflict with teachers, and minor property and violent crime. The report concluded that if a conference was held, it was usually successful in resolving the dispute.

The YJB (2004) evaluation used a comparison school in each of the nine project areas for evaluation purposes. It found through a student survey that the three schools using RJ for the full three years of the evaluation had less bullying behaviour. It found no significant differences between project and comparison schools,

however, in the remaining 23 schools, where the programme had lasted only 18 months. A teacher survey also indicated encouraging differences between the programme and non- programme schools in improved student behaviour, though staff remained firm in their view that exclusion was an effective way of dealing with behaviour problems. Inconsistent reporting practices meant that it was not possible to conclude whether or not RJ approaches affected the level of exclusion in the schools. In addition, almost half of all staff in the RJ schools reported at the end of the evaluation they knew little or nothing about RJ; those who felt they did know about RJ often had significant misconceptions.

The YJB (2004) report concluded that if RJ is implemented “correctly” (without specifying exactly what this meant), “it may be a useful resource that improves the school environment and enhances the learning and development of young people” (YJB, 2004: 65). Moreover, while there were no statistically significant effects on student attitudes, the RJ programme schools reported some important within-school improvements in student attitude compared with non-programme schools. The report observed that since the programme came to an end in 2004, there was little value in making specific recommendations, though it did suggest that any future introduction of RJ in schools should seriously involve the Department for Education & Skills, should require a whole-school approach, and should clearly articulate what RJ means in the school context.

RJ in schools remains an attractive possibility. Insufficient evidence is so far available about its effectiveness. Evidence that is available suggests that:

  • Tensions frequently exist between restorative principles and prevailing practices for school discipline, typically involving punishment. Such tension militates against successful implementation. Staff resistance to testing such a radically different approach (as RJ frequently appears to be) requires careful explanation and a great deal of support.
  • Successful implementation of RJ, even on a pilot basis for fair evaluation, requires broad institutional commitment within the relevant educational authority, the involvement and strong support of the school leadership, and adequate training in restorative techniques for staff.
  • Evaluations require more clarity about outcome measures. Culture change is a difficult concept to measure, while simpler indicators such as numbers of suspensions and expulsions may not be sufficiently clear as measures of harmful conduct.
  • Using restorative conferencing for specific incidents, isolated from other restorative practices, appears to have limited value for school conduct generally.
  • A whole-school approach, with RJ being used throughout the institution, rather than limited to specific classes or specific circumstances, appears to be essential for a satisfactory

test of RJ.

The implication of these conclusions for the future of RJ in criminal justice may be similar. All tests of RJ to date in criminal justice have been limited to case-specific tests. Perhaps whole communities or court areas need to adopt RJ in order to give it a fair test as a crime prevention strategy (see section 15).

7.4. Neighbourhood policing

The use of restorative policing in neighbourhood policing is an evolving idea. Police community safety officers or regular constables may find opportunities to divert cases from the criminal process if they can be resolved more effectively with RJ in collaboration with parents, neighbours or community leaders. The potential for this approach may depend on both the extent of social capital in a neighbourhood and on the extent to which local police have mapped the social networks that can be mobilised for RJ. But where both requirements are met, there appears to be great potential for returning to what is often celebrated as the “golden era” of neighbourhoods.

In this era, it is said, village bobbies on foot who knew everyone could operate a system of less formal justice. Their use of local knowledge to handle crime often prevented the escalation of conflicts into serious violence, but it entailed the risks attendant to all unaccountable processes. Whether such neighbourhoods or villages ever existed in quite this way is open to debate (see eg Banton, 1964; Cain, 1973). Nonetheless, the value of such conditions seems theoretically powerful in contemporary society.

It is even possible that RJ itself may help to foster the social cement that holds neighbourhoods together. The current discussions of neighbourhood-based RJ properly address the question of how such “street” practices can be made transparent and accountable to the wider society. They could also, in principle, become a way to reintroduce police diversion decisions on the Australian model, at least on a pilot basis, to see what crime reduction effects it might achieve on a “whole- community” basis.

The four institutional settings for RJ each provide opportunities for growth and development. Of these four, the greatest potential lies in adult justice. That is where large volumes of crimes go unsolved, probably for lack of victim and witness confidence in the system. Adult justice is where some 65,000 offenders are incarcerated, and many more are under community supervision. That is where most of the serious violence is treated, for which recidivism rates remain high. That is also where the lion’s share of the costs of crime are borne by the entire nation. The rest of this report therefore focuses on the evidence about adult – and, to a lesser extent, youth – justice as a setting for restorative strategy.


  1. Keeping promises versus following orders: how RJ compares


Keeping promises versus following orders: how RJ compares

Motivating offenders to keep promises is a key step in reducing repeat offending. RJ shares with CJ the tough question of how to ensure that offenders keep their promises in the aftermath of a crime. Problems of compliance in conventional criminal justice include ensuring attendance at court, at probation and at drug treatment or other rehabilitation programmes, payment of fines, payment of restitution to victims, and other promises.

Supporters of CJ assume that custody is an effective “stick” and that RJ may not benefit from that threat. Yet the data on compliance with CJ are not encouraging as evidence of deterrent effects of the custody “stick”. Offenders often return to prison for breaking promises, not for committing new crimes. Rather than being deterred by returning to prison, they often seem surprised.

Rather than using more prison to enforce promises offenders make, RJ could provide an alternative. It may be that RJ can do as well by providing a better “carrot” or motivation to keep promises. When offenders agree to terms that are dictated by a sentence, they have little emotional connection to the court officials offering those terms. But when offenders sit face to face with a crime victim for two to three hours, they may – or may not – develop a strong emotional commitment to that victim. Compliance with outcome agreements is important for the integrity and reputation of the RJ programme as well as for the sake of victims. Measured simply by average rates of promise keeping, the evidence to date is indeed encouraging that RJ may do better.

Compliance with outcome agreements

The key test of RJ on this point is whether offenders fulfil the outcome agreements they have reached in a consensual group discussion. Such agreements are reached with offender consent in a very high proportion of cases: 98% of JRC conferences held succeeded in reaching consensual outcome agreements across eight experiments in London, Northumbria and the Thames Valley, according to the University of Sheffield’s evaluation (Shapland et al, 2006). A similar percentage was reported in the four Canberra RISE tests (Strang, 2002). The number of items in each agreement varied, but almost always there were several components. The timescale for completion was usually within six months. Both offenders and victims often emphasise altruistic reasons for taking part (Shapland et al, 2006: 65).

The data are less clear, but still promising, for offender fulfilment of those consensual agreements. We reach this conclusion in part because of, and not just despite, patchy compliance checking in the Australian RISE project – where police claimed a compliance rate of around 70% of all conferenced offenders. Learning from this earlier experience, police and probation facilitators in the UK’s JRC studies developed an intense monitoring regime to determine whether offenders had in fact fulfilled their

agreements. When offenders were found not to have complied, facilitators encouraged them to do so if the outcomes were feasible. In general, offenders could not walk away from their agreements without at least being prodded to fulfil them – repeatedly if necessary.

Sometimes the offenders’ undertakings proved not to be achievable. For example, offenders who had agreed to pay compensation were often unable to do so because of receiving a custodial sentence. Those who had promised to comply with drug treatment were not always able to be placed on a drug programme (Shapland et al, 2006: 68). Across all eight Justice Research Consortium UK experiments, Shapland et al report that 36% of all agreements were fully completed and a further 52% were partially completed, leaving only 11% definitely not completed. They conclude (p68): “Given the seriousness of the offences and the entrenched nature of many of these problems, these are in fact very low failure rates.”

Despite the paucity of evidence on compliance with RJ outcome agreements in other projects, what evidence we have shows that RJ exceeds the rate of compliance with court-ordered outcomes. In the Canadian meta-analysis of 22 studies that examined the impact of 35 RJ programmes, Latimer et al (2001) found that participation in RJ significantly improved the likelihood of offenders complying with restitution agreements over court- ordered payments. Cormier (2002) reports that a (non- randomised) evaluation of a Canadian victim-offender mediation programme conducted by Nuffield (1997) estimated that victims seeking restitution for material harm were more than four times as likely to receive it as victims whose cases went to court.

There also appears to be substantial evidence that compliance with court-ordered fines is problematic. An Australian study in the state of Victoria found that 79% of fines were unpaid for more than one year, and that the overall recovery rate was 44% (Victoria Auditor-General, 1998). This comports with a study in South Australia reported by Chapman et al (2003) that found a collection rate of 51%. Similarly, the Canberra (ACT) magistrates’ court clerk estimated that during the RISE experiments only about 40% of court-imposed fines were ever collected (personal communication to H Strang, 2000). The situation in England and Wales is little better: Chapman et al (2003) report that the estimate of the UK select committee of public accounts was less than two-thirds of fines collected, at vast public cost: in 2000/01 alone, £74 million in financial penalties were written off.

Once again, the evidence from three highly developed, long-term RJ programmes in the UK shows a documented success. Across the eight separate tests of the RJ units in the Metropolitan Police, the Northumbria Police, and the Thames Valley probation and prisons services, 89% of promises made in RJ were kept, at least


in part. By simple comparison to 66% of UK fines collected, RJ does better.

Whether the comparison is reasonable depends on how difficult it may be to pay a fine versus meeting the RJ agreement. Even if there was less money to be paid in the RJ outcomes, that does not mean that it was easier than paying a fine. For a chronic offender to attend a drug treatment programme is a big step

forward. If such results are more likely with RJ than with CJ, we may have found the “missing link” between sentencing and crime prevention. If RJ can foster more commitment to completing rehabilitation programmes, then it could be focused on specifically indicated programmes for each offender. While this approach has not been attempted or tested, it is a logical next step in building on the promising evidence to date.



  1. Reducing harm to victims


Reducing harm to victims

One observation about the cost-effectiveness of RJ focuses entirely on victims. Even if RJ has no effect on crime, it may still be a useful strategy if it helps victims. And if by helping victims it reduces the costs of other services to victims by more than the cost of RJ, then RJ would be cost-effective. On the grounds solely of reducing cost to taxpayers, RJ may be a very beneficial strategy. On the grounds of putting victims at the centre, RJ may be a very appealing strategy. On the grounds of helping victims for the intrinsic merit of that goal, the evidence for RJ is compelling.

To be sure, this evidence has limits. Most victims will never learn who committed a crime against them, since most crimes go unsolved. RJ cannot help those victims directly. Even when an offender is identified, some offenders will refuse to accept responsibility, or to engage in RJ on any basis. RJ cannot help their victims. And even when offenders are willing to engage in RJ, some victims (or their families) will prefer not to. RJ cannot help those who will not help themselves.

All of these categories of excluded victims could, in theory, shrink steadily if RJ became more widely available. RJ could be more appealing to offenders’ families and friends, who might be moved more often to turn in their loved ones and offer evidence against them – if only as a way to get them into RJ. More offenders could agree to it because they would become more familiar with RJ. So could more victims, who might have attended other people’s RJ conferences as supporters.

Thus while the evidence to date is limited to the “pioneers” of RJ, there is no reason to assume that the proportions of criminal cases they represent must always be that small. Nor does it mean that RJ would have the same effects on a broader sample of all victims. The only way to find out is by testing RJ on a widespread basis. The evidence to date supports, at least, the cost- effectiveness of such a test.

  1. Effects of RJ on victims

While the effect of RJ on offenders appears variable, the effect of RJ on victims who agree to participate are far more consistent. Evaluation results almost always indicate a high level of satisfaction with the process, despite a wide range of rigour in those tests. The quality of many evaluation studies in relation to victims is as problematic as it is for offenders; often they are the same studies. Yet the results stand up to the toughest test: comparing all victims randomly assigned to RJ (whether or not it was delivered or completed) to similar victims whose cases received normal CJ processing.

Less compelling is the evidence that victims who completed RJ were more satisfied than victims who refused RJ. Such findings may reflect pre-existing differences in personality, or even opinions about RJ, rather than the effects of RJ itself. The problem of “completers’ evidence” runs through all the victim

studies that do not compare victims whose cases are randomly assigned (or not) to RJ. Indeed, even in these studies, it is not possible to have the views of the victims who did not consent to be considered for RJ. In any event, it is likely that some of the reluctance is due to the unfamiliarity of the general public with RJ, plus some misconceptions about what it entails. Greater availability of RJ, together with information about the very positive views of the victims who have chosen to participate, is likely to increase the proportion of victims willing to participate in the future.

Some reluctance by victims to risk encountering their offenders is undoubtedly due to anger, fear and other emotions. The level of reluctance appears to be mediated by the setting and perhaps by the seriousness of the offence. We do not know whether victims who choose to participate have the capacity to self?select in such a way that they are more likely to benefit than victims who decline.

There is also good reason to think that RJ is not a remedy that is right for every victim. For a small minority within well-conducted studies, it is clear that RJ was a negative experience that did not improve their situation and may have made it worse. Conferences are inherently more risky ventures than normal criminal justice processing: the latter may do little to help victims, but little to harm them either when they have no role to play. Confronting the (fortunately rare) unremorseful offender in an RJ conference, however, may appear to be a significant risk – even if the payoff is substantial when the encounter results in a sincere expression of apology.

At present we do not have the tools to predict for which victims RJ is likely to be beneficial and for the (probably rare) cases in which it will be counterproductive. Nevertheless, such strong and consistent positive findings about victim benefits in the great majority of cases lead us to conclude that victims will generally benefit from participation whenever they have the opportunity to do so.

9.2 Evidence from randomised controlled trials

  1. RISE

(Sherman et al, 2005; Strang, 2002)

Even though RISE employed a randomised design, the victim analysis constitutes a quasi-experimental analysis only (level 4.5) because it was the case rather than the victim that was randomly assigned. However, the high participation rate of eligible victims, as well as their high interview response rate (approximately 90%), gives us confidence that there was little selection bias inherent in the research design (Strang, 2002). If anything, the study underestimates the effects of RJ because 23% of victims who were promised a conference did not actually attend one,

due to various difficulties of implementation (Strang, 2002: 81).

When victims were asked whether they were satisfied with the way their case was dealt with by the justice system, there was a statistically significant difference between the court-assigned and the RJ-assigned victims (46% vs 60%). Significantly more of those who actually experienced an RJ conference were satisfied, compared with those whose cases were dealt with in court (70% vs 42%). There was no difference here between property and violence victims.

As a further indicator of overall treatment satisfaction, all RISE victims were asked whether they were pleased with the way their case had been dealt with (whether by court or by conference), or whether they would have preferred the alternative treatment. Significantly more conference victims than court victims agreed they were pleased to have been treated the way they were (69% vs 48%). Property victims were slightly more pleased than violence victims.

Victims thought offender apologies were important in bringing about emotional restoration. Ninety percent of all victims, whether assigned to court or to conference, felt they should receive an apology. One of the most important differences was thus found in the prevalence of apologies for each group. For those victims assigned to a conference, 72% said their offender had apologised (and 86% of those who actually experienced a conference) compared with 19% of those assigned to court. Furthermore, more conference-assigned victims than court- assigned victims said they felt the apologies they received were sincere (77% vs 41%).

It seems likely that these findings are related to another striking difference between the conference-assigned and court–assigned victims: desire for revenge. Overall, 20% of court victims said they would harm their offenders if they had the chance, compared with 7% of conference victims. This difference was especially stark for victims of violent crime whose cases went to court, nearly half of whom (45%) wanted to harm their offenders, compared with 9% of those who went to a conference (Strang, 2002). Table 4 shows that when we extended this analysis to London, computing separate point estimates for men and women victims in each of four separate experiments, the pattern across the eight tests of the revenge-reduction hypothesis was supported as statistically significant and substantial. On average, there was two-thirds less desire for revenge across the eight groups receiving RJ than their comparison victims receiving CJ (Sherman et al, 2005).

  1. Bethlehem Restorative Policing Experiment

(McCold and Wachtel, 1998)

Results from these experiments with first-time juvenile property and violence misdemeanour offenders are complicated by the fact that consent of participants was sought after random

assignment rather than before. As a result, a substantial number of victims or offenders assigned to the conference group declined to participate in RJ and their cases went to court. Conferences were conducted in only 42% of cases randomly assigned to conference. Because victim impact results were not reported according to the randomised assignment (as “intention- to-treat”), but rather were reported on the basis of treatment as delivered, we actually code this study not as a randomised trial (level 5) but as a correlation study (SMS level 1). We include it in table 4 to provide clarity about its status among readers familiar with the study. Victim results for the two experiments (violence and property) were combined in the original report.

Among the RJ completers, McCold and Wachtel found 96% victim satisfaction with cases randomly assigned to conferences (and where the case was dealt with in a conference), compared with 79% when cases were assigned to court and 73% among victims whose case went to court after the option of a conference was declined. Victims who attended a conference also more often felt that the process was fair (96%), that the offender was held accountable (93%), that their opinion had been adequately considered (94%) and that the offender had apologised (96%). It is important to note that the satisfaction level was lowest among those victims who were promised, but did not get, RJ. This provides evidence in support of the practice of asking offenders first, before even raising the prospect of RJ with victims who may later have it denied. Similar results were found among the victims in RISE whose cases were randomly assigned to the RJ group but who were not ultimately given RJ.

What we cannot tell from this design, as implemented and reported, is what the effects of RJ would be on victims if consent had been sought prior to random assignment. A comparison between victims of criminals who were willing to do RJ, but not offered it, would be necessary to draw an inference about the causation of victim outcomes by RJ rather than selection bias.

  1. Indianapolis Juvenile Restorative Justice Experiment

(McGarrell et al, 2000)

This study of young (seven to 14 years) first-time property and minor assault offenders and their victims also found markedly higher levels of satisfaction among victims whose cases were randomly assigned to a conference rather than an array of other court diversions. In this experiment, however, random assignment was almost always followed, so that analysis suffers less offender self-selection bias than in Bethlehem. Furthermore, 97% of victims said they felt involved with the way their case was dealt with, compared with 38% of victims in the control group, and 95% felt they had been able to express their views, compared with 56% of the control group.

  1. Justice Research Consortium’s eight trials of RJ in London,

Northumbria and Thames Valley

(Shapland et al, 2006)

The University of Sheffield’s evaluation of the JRC’s series of randomised trials in the UK does not directly compare the views of victims whose cases were assigned either to normal CJ or to normal CJ plus RJ. However, it contains a number of measures of victims’ experience of RJ, derived from interviews from over 200 of the approximately 450 victims involved in RJ in the eight trials. Overall they found that about 85% of victims (and 80% of offenders) were satisfied with their experience. In particular, only 12% of victims (and 10% of offenders) expressed any doubt about the outcome agreement reached at the end of the conference, and almost all thought it was fair. Looking at various dimensions of satisfaction, the evaluation found that more than 70% of victims in all eight experiments said they found the conference useful and fair and that it had given them a sense of closure about the offence. Interestingly, given the highly variable length of time that had elapsed between the incident and the conference (from a very few weeks to several years), most said that the RJ had been held about the right length of time after the incident.25

Besides the findings relating to macro-level victim satisfaction concerning both process and outcome in the London trials, there are strong indications of psychological benefits to traumatised victims at the individual level. A study embedded in the larger RCT was conducted to determine the effect of RJ on post- traumatic stress symptoms (PTSS) experienced by burglary and robbery victims (Angel, 2005). While few of these crime victims had PTSS symptoms severe enough to be classified as having a clinically diagnosable disorder (PTSD), their symptoms were nonetheless substantial.

A total of 137 individuals whose cases had been randomly assigned either to court only or to court plus RJ were assessed using a standardised psychological instrument (the “impact of event scale revised”) to measure PTSS. Victims who had experienced RJ consistently scored lower than control participants, both immediately after their cases were finalised and also six months later. The item concerning return to employment and normal activities is reported separately in table 4, even though the significance level is marginal, for two reasons. One is that a substantial effect of RJ on employment could produce major cost savings for government and economic productivity. The other reason is that the magnitude of the effect of RJ on return to employment is quite substantial, as distinct from its 12% probability of error. It suggests that robbery and burglary victims take 50% longer to return to work without RJ than if they have received it.

Given the low propensity of victims to seek assistance for trauma associated with their experience and the sometimes devastating consequences in terms of health, employment and quality of life, RJ offers encouraging prospects for victim recovery that are not otherwise available. The long-term effects of lowering even sub-clinical PTSS symptoms is suggested by a prospective study of PTSS among almost 3,000 military veterans, almost all of

whom fell below the threshold of clinical PTSD. Nonetheless, their PTSS levels predicted substantial increases in coronary heart disease (CHD) and myocardial infarctions (Kubzansky et al, 2007). These biomedical consequences of PTSS suggest that a broader view is needed of the true costs of crime, and of the cost-saving benefits for helping crime victims with their PTSS.

Another striking finding from the London trials, as noted above, concerned victims’ unresolved anger towards their offenders and their desire for revenge, which mirrored the results from RISE referred to above. Indeed, remarkably consistent results emerge across violent and property crime and for male and female victims in both sites. While victims of violent crime in both Canberra and London showed higher levels of desire for revenge than victims of property crime, nonetheless, for both offence types and both sexes, RJ reduced vengeance in the predicted direction (Sherman and Strang, forthcoming).

9.3. Evidence from non-RCT studies

Victim benefits have been remarkably similar in findings from both RCTs and other evaluations. Broadly speaking, victims who agree to participate report extremely high levels of satisfaction with the process on the outcomes they say matter most – the opportunity to confront their offenders and express themselves directly to them. Consistent with findings from conferencing programmes (see eg Strang, 2002), victims tend to show satisfaction based on greater participation in the justice process rather than according to the extent of financial or other material restitution ensuing from the encounter with the offender.

Like the RCT-design studies, these evaluations consistently show that RJ appears to benefit victims along several dimensions. It can reduce victim anger, anxiety and fear of revictimisation and increase their perceptions of fair process. The meta-analysis of 35 RJ programmes carried out by Latimer, Dowden and Muise (2001) found that they increased victim (and offender) satisfaction generally. The following studies relate to various forms of RJ, including victim-offender mediation, victim-offender reconciliation programmes and family group conferencing and its effects on victims:

  1. Victim-offender mediation in North America

(Umbreit and Coates, 1992, 1993; Umbreit, 1994, 1995, 2001; Umbreit, Coates and Kalanj, 1994; Umbreit, Coates and Roberts, 2000; Umbreit, Coates and Vos, 2001; Nugent et al, 2003).

  1. As part of their evaluation, the University of Sheffield team also reported on two non-RCT schemes that the Home Office funded under the same arrangements as the JRC, Connect and REMEDI. Both these schemes offered both direct and indirect mediation, giving the evaluators an opportunity to make a preliminary assessment of the relative merits of the two techniques. They observe that, while victims may initially be more attracted to indirect contact with their offenders, indirect mediation makes it difficult to have an agreed outcome or to refer to future-oriented matters and that it is associated with lower levels of satisfaction. They further observe that almost all victims who experienced a face-to-face meeting with their offenders did not regret it and conclude that direct mediation is more helpful than indirect.

In a series of studies, mostly relating to juvenile offenders, stretching over more than a decade, Umbreit and his colleagues consistently report high levels of victim satisfaction arising from the opportunity to tell their offenders directly about the effects of the crime and to negotiate restitution. They also found that restitution per se was the least important issue for the victims who completed the process.

  1. 1990s RJ evaluations in England and Wales

(Miers et al, 2001)

This major evaluation of seven RJ schemes operating across the country in the late 1990s was mixed in its findings, though many of the problems reported concerned issues of implementation rather than dissatisfaction among participants. Interviews with 23 victims found that while most felt the intervention had made an impact on the offender, they were not always convinced that the offender had made sufficient amends. Nevertheless, they indicated that they appreciated the opportunity to be involved and to get the information they wanted and over two-thirds said that they were satisfied with their involvement.

  1. Before/after RJ effects in RCTs

(Strang et al, 2006)

In RISE and the JRC tests in the UK, victims who had participated in RJ conferences were asked a series of questions after the conference about how they felt both before and after the conference. Although there are shortcomings in asking the respondent to remember feelings from before the event, the consistency of responses lends confidence to the findings; in any event, it was not practical to obtain baseline information on emotional dimensions beforehand. Nor was it practical to ask

control victims similar questions on a before-and-after basis. The findings therefore have the status of quasi-experimental, level 2 results that are more qualitative than quantitative. While they do not meet the threshold for inclusion in the tables, they are presented here because they address substantial issues of victim impact.

In these before/after, retrospective comparisons, consistent and statistically significant differences were found. When victims reported on their before-conference and after-conference feelings, there were large differences in Canberra, London and Northumbria on all the following dimensions: fear of the offender (especially for violence victims); perceived likelihood of revictimisation; sense of security; anger towards the offender; sympathy for the offender and the offender’s supporters; feelings of trust in others; feelings of self-confidence; anxiety.

9.4. Victims’ opinions of RJ across the process

The quality of studies relating to victims’ opinions or views of RJ are highly variable: they rarely include an explicit comparison group and almost never do they consider the views of victims who consented to RJ but for whom the RJ was not delivered for any reason. Nonetheless, across all these studies including many kinds of offence type the conclusions are clear: when victims consent to meet their offender in an RJ conference they are usually satisfied with their experience provided that 1) the RJ meeting happens as promised and 2) the offender complies with the undertakings they made during the conference. Furthermore, available evidence shows that these victims are far more satisfied than their counterparts whose cases are dealt with in the formal justice system (Strang, 2002; McCold and Wachtel, 1998; McGarrell et al, 2000).


  1. Reducing repeat offending


Reducing repeat offending

The variable effects of RJ on repeat offending are listed in tables 1 to 3 above. The patterns we have been able to discern from this evidence so far are reported in the summary. The key finding is that RJ may work better with more serious crimes rather than with less serious crimes, contrary to the conventional wisdom. The research is therefore discussed in this section in the same structure as in tables 1 to 3: organised by results for violent, property and non-victim crime.

  1. Recidivism after violent crime

The success of RJ in reducing, or at least not increasing, repeat offending is most consistent in tests on violent crime (table 1). Whether we consider just randomised experiments, or include quasi–experiments as well, we find no evidence of increased repeat offending with RJ after violent crime. We also find, in some tests, substantial reductions in recidivism after violent crime.

Randomised experiments

The largest effect of RJ on recidivism after violent crime found in a randomised experiment so far has been in the Canberra RISE project. In a two-year-before, two-year-after comparison, the frequency of arrest among white people under 30 years of age who were assigned to RJ dropped by 84 per 100 offenders more than in the control group. This effect was found despite a less-than-perfect implementation of RJ in the group assigned to that treatment. The arrest measure included all kinds of crimes, from offences with victims to non-compliance with court appearances. These effects were not found among Aboriginal offenders, but the sample size for that stratum was too small for an adequately powered test (Sherman et al, 2006a).

Almost as impressive was the effect found among girls in Northumbria (Sherman et al, 2006b). Females under 18, whose parents joined in their consent to an RJ conference as part of a final warning for assault, were randomly assigned to either the conference or a standard “talking-to” by a police officer with a parent present. Those who were assigned the conference almost always attended one. Compared with those not assigned a conference, they had twice as great a reduction in arrests per 100 offenders (118 fewer in the year-before/year-after comparison) than the CJ group (only 47 fewer). Thus we can estimate that RJ among predominantly poor, violent white girls in Northumbria prevented 71 crimes leading to arrest per 100 offenders, relative to CJ.

Similar effects were found among children aged seven to 14 in Indianapolis (McGarrell et al, 2000). Youth accused of violent crime were randomly assigned to RJ conferences or a wide range of other programmes of diversion from prosecution. None of the comparison treatments involved a meeting with the victim. The RJ group was reported to have had a 28% prevalence of rearrest at six months (after only) compared with 34% for the control group.

The first reported randomised experiments in diversion from prosecution to RJ – those conducted in Bethlehem, Pennsylvania – are included in table 1 for clarity, even though little can be concluded from them (McCold and Wachtel, 1998). As noted above, the random assignment in these experiments was done prior to seeking offender consent, which guaranteed a high refusal rate. (Most experiments seek consent first and then randomise cases to RJ or controls, answering the question of “what effect RJ has on those who say they are willing to undertake it”.) Since it is possible to combine the refusers and completers into a comparison of the groups randomly assigned to RJ and control, that analysis stands as a level 5 RCT. The result is no difference in repeat offending between the two “intention- to-treat” groups. With some half of the RJ offenders refusing to undertake RJ, however, it is difficult to draw any conclusions from this experiment about the effect of RJ on recidivism after violent crime.

Much the same can be said for the Brooklyn, New York, experiment in diversion to RJ for (predominantly) violent felony cases (Davis et al, 1981). While the evidence is reported as a level 5 intention-to-treat analysis, 44% of the cases assigned to RJ never undertook it. Similarly, 72% of offenders prosecuted in court had their cases dismissed or absconded. The comparison of 27% of prosecutions resulting in criminal convictions to 56% of RJ cases with completed mediations across 100% of both groups yielded no difference in recidivism.

A better test of RJ also failed to show an effect (Sherman et al, 2006b). The white male youths in Northumbria, in the same experiment that found a strong effect of RJ on violent white girls’ recidivism, had similar recidivism rates regardless of how their final warning was delivered. Meeting with the victim in an RJ conference yielded no less recidivism than a “talking-to” by a police officer. Equally important, it yielded no increase in recidivism. This finding stands in clear contrast to the reductions in recidivism after RJ found in RCTs among white male youth in Canberra and Indianapolis, but it is unwarranted to analyse this question as a vote count. Rather, the question remains why RJ works in some samples and not others.

Quasi-experimental tests

As is often the case in evaluation research generally, the less- than-RCT evidence is far more consistent in showing benefits of treatment (see Weisburd et al, 2001). Three reasonably strong quasi-experimental tests of RJ on violent (or groups of predominantly violent) offenders all found substantial reductions in repeat offending after RJ compared with CJ.

Perhaps the least known of these tests is the Canadian programme of RJ for violent families (Pennell and Burford, 2000). This comparison of families offered a family group conference (as described in section 7) with families not offered such a

conference (in the same years) addresses a very volatile issue in RJ. While the sample size is small, the apparent effects of RJ are large. In the RJ group, for example, emergency visits to the home dropped from the before to the after period by over half, compared with a 50% increase in the comparison families. Nearly identical differences were found in reports of child abuse or neglect. While the sample included many aboriginal Inuit people, the sample size was too small to compute separate effects by ethnicity.

Two UK evaluations of (mostly) indirect RJ in the 1990s found statistically significant, although smaller, differences between predicted and actual offending rates after RJ treatment (Meirs et al, 2001). In West Yorkshire, a mixed sample of violent and property offenders were given a mix of indirect and direct RJ treatments. The predicted rate of reconviction based on their record prior to the treatment was then found to be about a third higher than the observed rate of reconviction after RJ, for a relative reduction in recidivism of 25% after RJ. A West Midlands pre-sentence mediation programme prior to sentencing produced almost identical results with a similar size and mix of sample. In both tests, slightly over half of the offenders also received custodial sentences.

  1. Recidivism after property crime

The use of RJ for property crime produces less consistency and magnitude of effects on recidivism than is found in RJ for violent crime (table 2). The evidence here includes two findings of increased offending, as well as five findings of crime reductions. More impressive, however, is the fact that RJ did as well as, or better than, prison time as measured by recidivism in two separate tests.

Randomised experiments

The best result of RJ for property crime in the UK so far is found among white male youth in poverty areas of Northumbria (Sherman et al, 2006). Randomly assigned RJ conferences among youths whose parents joined in their consent to RJ were delivered to half of the cases in which victims were also willing to undertake RJ. The other half received the standard final warning “talking-to” by a police officer. The one-year before/after comparison of arrest rates in the RJ group showed 88 fewer arrests per 100 offenders. In the control group, the reduction was only 32 fewer arrests per 100 offenders. Thus RJ had more than twice the size of the reduction in offending measured by arrest as did CJ. The number of females enrolled in the sample was too small to yield a valid estimate of RJ effects for that stratum, but large enough to drive the effects on the overall sample into a difference that failed to achieve statistical significance.

The Northumbria results stand in sharp contrast to the Canberra results (Sherman et al, 2006a), where non-Aboriginal youths under 18 had no different rates of arrest frequency after RJ than after CJ (in a two-year before/after comparison). The good

effects of RJ on Northumbrian youth property crime stand in even sharper contrast to the Canberra results for Aboriginal youth, a small sample of whom (N = 23) showed such a large negative effect of RJ that the result was statistically significant (P = .049). RJ caused an increase of 228 arrests per 100 offenders per year among Aboriginal property crime arrestees, while conventional juvenile prosecution caused a reduction of 66 arrests per year per 100 offenders. This means that RJ roughly tripled the detected offending rates of these people under 18. Why that happened, and its broader implications, are discussed below in section 11.

The Indianapolis experiment with youths aged seven to 14 did much better than its counterpart in Canberra did with older juveniles of both races (McGarrell et al, 2000). Unfortunately, no breakdown by race in recidivism results has been published for the Indianapolis test. But the main effects show that RJ conferences cut in half the percentage of juveniles who had any new arrest after random assignment, relative to the control group that was not given RJ (15% vs 27%).

Most of the randomised trials of RJ for property crime, however, did not employ the face-to-face RJ conference strategy. Rather, they were primarily experiments in indirect RJ conducted by Anne Schneider (1986) in a nationwide evaluation of restitution programmes in the US in the 1970s. These programmes varied in their details of treatment and consent, but were consistently implemented and reported as level 5 RCTs. Two of the four RCTs showed that court-ordered restitution, randomly assigned, yielded lower rearrest rates than conventional probation. This was despite one of the tests (in Washington, DC) suffering a 40% dropout from its consent-based programme. The intention-to- treat analysis preserved the level 5 test, showing 12% lower arrest frequency for consenting cases assigned to RJ and a 7% increase for consenting cases assigned to non-RJ probation. This stands in contrast to a similar comparison in Oklahoma, where no effect on recidivism was found.

Perhaps the most important finding of the four Schneider (1986) experiments is the Boise, Idaho, result. In this test, court-ordered restitution in the community was compared with a sentence of eight days in jail, served primarily on weekends. With a sample of 181 randomly assigned cases, this test may be one of the best tests of the “short, sharp shock” theory long popular with British courts. What it found was substantively significant because the outcome lacked statistical significance. That is, there was no benefit from putting offenders in jail, compared with restitution, despite the far higher costs of incarceration.

Quasi-experimental tests

An even more impressive comparison of RJ to prison was found in a quasi-experimental test in Canada (Bonta et al, 1998). In this test, 142 predominantly (61%) property offenders coupled indirect or (in 10% of cases) face-to-face RJ with court-ordered restitution. All of them were rated after conviction but before

enrolment as “likely” to be sent to prison for six months, but most of them were not. They were matched with a comparison group of 67 inmates with similar profiles who had not been offered this programme. The two-year prevalence of reconviction for the comparison group after they were released from prison was 37% – three times the 11% rate of the RJ group. This difference was statistically significant evidence that RJ after serious crimes can reduce recidivism better than six months or more in prison. If that comparison were to hold up in the UK, it might be possible to use RJ to cut the prison population in half with no change, or a reduction, in the crime rate.

RJ performed less well in the New Zealand pilot of court-referred RJ for adults (Triggs, 2005: 15). In this study, only 25% of referred offenders actually completed the RJ as requested. The actual two-year prevalence of reconviction (41%) among the 193 completers was then compared with their predicted reconviction rate (45%), with the difference not enough to be statistically significant. The broad mix of property, violent and non-victim offences (including traffic) in this sample, however, limits what we can conclude from it.

  1. Recidivism after non-victim crime

This subsection examines the effects of RJ on recidivism after offences such as shoplifting, drink-driving, and public disorder (table 3). While RJ is often most widely acceptable when used for such low-seriousness offences, it generally does worse in these tests than in tests on more serious crimes.

Randomised experiments

Two of the four RCTs in the Canberra RISE project drew samples from these offences (Sherman et al, 2000). One was for drink- driving, detected by random breath tests at police roadblocks. Anyone with a measured blood alcohol content of .05 or higher was then randomly assigned to RJ or prosecution. The difference meant no criminal record or publication of the offender’s name in the newspaper if they received RJ. It also meant that they did not lose their driver’s licence for six months or pay a mandatory fine, although they often agreed to pay larger amounts to a charity voluntarily in the RJ agreement. In terms of drink-driving behaviour, there was no difference in repeat offending. In the initial analysis, in fact, there was a small increase in total offending associated with RJ.

A second RISE experiment employed a sample of shoplifters under 18 caught by security staff in large department stores. In these cases, the store sometimes sent a security employee as a representative, but they could not claim to have suffered from the offender’s conduct (indeed, their jobs depended on the sustainability of crime, as Karl Marx pointed out about law professors). We therefore treat this test as a non-victim crime. No difference in repeat offending was found between the RJ and CJ groups.

The best result for RJ in RCTs of non-victim offences has been reported by McGarrell et al (2000). This sample focused on such “public order” offences as making noise, drinking under age, and youth curfew violations. In this Indianapolis test of RJ with youths aged seven to 14, the group randomly assigned to non- victim RJ conferences had half the prevalence of repeat arrest of those assigned to other diversion programmes: 28% for RJ versus 45% for the comparison over the next six months after disposition. Whether this effect can only be achieved among very young offenders is a question the evidence cannot yet answer – one that can be raised about all the Indianapolis results.


We found no quasi-experiments that met our minimum eligibility criteria for testing RJ with non-victim crime cases.

  1. Missing evidence on recidivism

The results discussed above generally support the conclusion that RJ works most consistently to reduce repeat offending – and not increase it – with violent crime. The effects of RJ are less predictable, and sometimes counterproductive, for property crime. The evidence is least compelling for the use of RJ with non-victim crime. While it would be inappropriate to quantify this ranking any more precisely than saying it in words connotes, it is still a reasonable summation of the evidence.

The missing evidence, however, could change the content of that summation. More tests of RJ on more specific kinds of offences, organised by such criteria as stranger versus acquaintance crime, injury versus non-injury crime, and other salient criteria, could yield different conclusions about seriousness. Until and unless such tests are conducted, however, the finding on this evidence can stand: RJ works better with more serious offences. The reason why may be consistent with the apparent emotional basis for RJ: that offender remorse for having harmed a victim – perhaps especially victims “like them” rather than socially distant by class, race or income – is what drives any reduction in repeat offending that follows RJ.

Beyond the conclusion that RJ may reduce crime more effectively with more serious crimes, the evidence does not readily allow accurate predictions of the kinds of offences or offenders for whom RJ is more likely to be effective. Such obvious patterns as consistency by race and gender are elusive; the opposite effects of RJ for property crime by young white males in Northumbria versus Canberra is a prime example.

What we can predict, based on the evidence, is that except in rare circumstances RJ is unlikely to cause more repeat offending than CJ. That, combined with clear and consistent average benefits for victims, means that RJ could proceed on a much larger scale, as long as:

  1. criminogenic effects of RJ are not more widespread than is apparent; and
  1. we can tackle the remaining question about the effects of RJ on crime: how does RJ affect general deterrence?

As several criminologists have pointed out, the second question is central to the foundation of criminal law. What effect would living in an “RJ community” have on potential offenders, knowing that their first offence would almost always be dealt with by a restorative process? Would that knowledge encourage

more primary offending, even if RJ reduced repeat offending? Only a randomised comparison of communities with widespread RJ to communities without it can answer that question.

Because of the great importance of the issues of “how rare” the backfiring effects of RJ might be, and what effect RJ would have on general deterrence, we highlight our discussion of them in the next two sections, respectively.


  1. Rare events: when RJ backfires – and why


Rare events: when RJ backfires and why

For many people, the fundamental rule of justice should be the same as in the Hippocratic oath: first, do no harm. The extensive evidence on RJ effects provides substantial reassurance that RJ has a very low risk of causing harm to past crime victims, or to communities and victims suffering from higher rates of repeat offending that might have been caused by RJ. The exceptions to that evidence so far, as reported in section 10, are rare. So far negative evidence is scattered and weak. But any evidence at all of criminogenic effect merits close examination. This section first revisits the evidence. Then it examines the theory that could explain past findings and predict future results.

11.1 Three findings to consider

There are three findings worth discussing on this issue. One is the large effect of RJ on a small sample of Australian Aboriginal youth arrested for property crimes against white victims. A second is the initial evidence of a small increase in total arrests for Canberra offenders assigned to RJ after arrest for drink- driving. The third is a marginally significant increase in offending among Bethlehem (Pennsylvania) youth who were offered RJ after arrest for property crimes.

Effects of RJ on Aboriginal property crime offenders

The clearest evidence of criminogenic effects of RJ is limited to 23 Australian Aboriginal youth charged with property offences. This RCT did not set aside the Aboriginal defendants for a separate, block-randomised sample. The finding, rather, is from what statisticians call a subgroup analysis, or the analysis of interaction effects within a stratum of the larger experiment. The very large magnitude of the effect, however, meant that the effect was clearly not due to chance, even though the sample size was small. While certain governmental rules and systematic reviews exclude findings with samples that are small, there is no statistical necessity to distrust significant findings from small samples when the effect sizes are large and appropriate tests have been used.

Whatever the internal validity of the finding, its external validity – the capacity to generalise from these results to other populations – remains unclear. We do not, and cannot ethically, know whether this finding would apply even in another sample of Aboriginal youth in Canberra. We do not know whether it would apply to Aboriginal offenders in other cities in Australia, in other age groups, or other offence types. We certainly do not know what the limits of the finding are for other minority groups in other countries – especially minority groups of colour with disproportionately high rates of incarceration. We do not know whether there is a potential for criminogenic reactions whenever minority offenders of colour meet with white victims, but not when such offenders meet with victims of their own race. We do not know whether the finding could be limited to the specific context of property offending in the wealthy and highly educated community of Australia’s capital – or if it could be

predictive of RJ effects in all cities, rich and poor, across the developed world. There is no way to answer this question without further evidence.

It is therefore appropriate to say that the question of minority group effects remains a major gap in the evidence on RJ in the UK, as elsewhere. Further analysis of existing studies may shed some light on it, especially as recidivism data accumulate on the Crown Court experiments in London. Ample numbers of minority group offenders in those experiments provide a robust opportunity to explore issues of race and RJ.

Even more useful may be experiments specifically designed for the purpose of the comparison of RJ effects with different combinations of races or ethnicities of offenders. One example would be a suite of four RCTs testing RJ on samples of crimes with, respectively, white offenders and Afro-Caribbean victims, white offenders and white victims, Afro-Caribbean offenders with white victims, and Afro-Caribbean offenders with Afro- Caribbean victims.

RJ for drink-driving

In the first year after random assignment to either RJ or prosecution, Canberra drink-driving offenders in the RJ group had 83% more arrests for all offences in the year after than in the year before (Sherman et al, 2000). The prosecution group, however, had only 30% more arrests. The probability that this difference was due to chance was 13%, or slightly outside the conventional margins of statistical significance. The difference also disappeared at two years before/after, so the effect was transitory. The question is whether it has any significance.

Answering that question requires further evidence on the conferences themselves. In some of the conferences, police invited a “community representative” who lived in the neighbourhood where the arrest occurred. This volunteer citizen was often highly critical of the offenders, all of whom had been required by police to invite five family members or friends. These statements may have been emotionally toxic to the offenders, as an analysis of subgroup results indicated. RJ increased total arrests among cases at which community representatives had been present, but it reduced total arrests among offenders who had no community representative present.

Interpreting these findings has been complicated by the finding in repeated transfers of arrest data from the police that the criminal history of the offenders – for the same years – was different in different data sets transferred. This probably represents the effects of changes in both hardware and other aspects of record keeping. Nonetheless, the finding did appear, in one data set, for the first year of the experiment. It also fits into a larger possible explanation for any criminogenic effects.


Juvenile property crime defendants in Bethlehem, Pennsylvania A majority of the property crime offenders assigned to diversion for RJ in the Bethlehem experiment refused to undertake RJ. This group had the highest prevalence of repeat arrests among the three groups compared by the authors: accepters, refusers and controls. When the repeat offending is combined for both refusers and accepters assigned to RJ, however, the result is a small increase in the RJ group compared with the controls. The difference is again just over the boundary of conventional significance levels, at 11%. Yet given the random assignment of cases, there should not have been a difference in repeat offending between the groups if the refused offer to do RJ had no effect on recidivism. Rather, the data suggest that merely offering RJ to the refusers somehow made them more criminal. In that sense it appears that this majority Hispanic population experienced a criminogenic effect from the way in which their local police may have presented the invitation to RJ.

11.2. Theoretical and policy implications

It is important to note that none of these three findings would be reportable under current Home Office standards of publication. The Aboriginal sample is too small to meet the Home Office minimum, even with random assignment. The other two findings are not within the boundaries of statistical significance. The drink-driving differences also disappeared over time. These are reasons to disregard the findings altogether.

There are two good reasons to give the findings further consideration. One is the implications they may have for the criminological theory of RJ. The other is the implications they may have for policy decisions on scaling up RJ.

All three findings are consistent with predictions made by defiance theory (Sherman, 1993): that some people who are told they have behaved immorally by other people they neither trust nor like will become more criminal in future, rather than less. Defiance theory specifies some characteristics of such people who are “allergic” to criticism from sanctioning authorities they

perceive to be illegitimate. They are disproportionately likely to be lacking in conventional “stakes” in conformity, such as jobs, marriage or respect of mainstream people. They may also be constitutionally sensitive to criticism due to trauma or abuse at an early age, or perhaps nutritionally deficient in foods that promote equanimity.26

While it may be impossible to screen for such reactivity to RJ at the individual level, it may be possible to predict at the neighbourhood or community level of analysis. It may be that RJ simply does not work well where there is a preponderance of such offenders. That may be what happened in both the Aboriginal and Bethlehem samples. And it may also be that the dynamic only applies to property crimes. It is arguable, for example, that a victim of property crime can be seen as part of an unfair system. A victim of violent crime, however, could be seen by the offender as someone who hurts – physically – just like them, and who is worthy of the offender’s sympathy or remorse.

Predominantly white and employed drink-driving offenders also may fit defiance theory, when exposed to criticism from a community representative they have not elected or ever met before. The stinging words of a stranger may evoke anger on a bus or in a playground, let alone in front of a police officer and close friends. Activating the limbic system with a fight-or-flight reaction, such criticism could be carried around in the offender’s head for some months or years. That, in turn, may cause an increased propensity to “get back” at an illegitimate accuser through defiance of such people as representatives of the wider society. It may just make them “mad as hell and not going to take it any more”.

These connections make the three findings more significant in theory than they are in practice. They suggest that policy development may need to be guided by more theory, as well as by more data, just to be sure at every stage of scaling up RJ we “first, do no harm”.

  1. See, for example, Gesch et al, 2002.



  1. Other effects on crime


Other effects on crime

Just as RJ may work best in schools when it is used as a “whole school” strategy rather than for isolated incidents, RJ may work best when it is used across an entire community. Small pilot studies cannot generate such conditions. Public knowledge of RJ does not become wide enough with such pilots to influence community-level effects. But such effects could be the result of a synergy of many cases being handled by RJ at the same time. A high proportion of community residents, for example, might attend the RJ conferences, and begin to understand the process. Evidence to date also suggests they would see it favourably. This raise the question of what benefits – or harms – there may be from RJ on which we have yet to gain evidence.

If RJ became a widespread first response to crime by the criminal justice system, there could be two major effects on crime and justice that cannot be seen in the evidence to date. One is that many more crimes could be detected and brought to justice. The other is that many more crimes could be committed. The evidence to address both these possibilities is missing at present, but could be gathered before very long.

  1. Bringing more crimes to justice

At the individual level of analysis, there is good evidence that RJ can bring more crimes to justice (table 5). If RJ is scaled up across a large portion of all offences in a community, the benefits could be even greater. The major barrier to bringing offences to justice is victim and witness reluctance to risk retaliation or – more important – their time, from involvement with legal formalities. They may also distrust or fear the system itself, in terms of imposing excessive or inappropriate punishments on their loved ones. Hence the vast majority of crimes go unsolved, despite many citizens having the knowledge that could lead to a criminal conviction.

RJ could change all that. If it worked widely and well, it would logically encourage more people to come forward to participate in a process that would be more predictable and convenient than going to court. If offenders themselves accept responsibility at a higher rate because they become more familiar with the RJ process, that would also help solve more crimes. The evidence that RJ can improve trust in justice is substantial (see Strang et al, 1999). That trust could be translated into putting more crimes in the arms of justice, because those arms would be seen as helpful rather than harmful.

  1. Undermining general deterrence?

There is good evidence for discarding the worry that RJ would undermine general deterrence. That worry posits an opposite, but possibly simultaneous, effect of widespread RJ in a community causing more crime as well as solving more crimes. Removing the threat of punishment, it could be argued, would encourage more people to commit their first offence. In principle, it could be seen as giving a lifetime “juvenile offender” status to all adults in the

community. The implication of “one free crime” might be too attractive to resist, unless other reasons besides the prospect of jail time make people restrain from offending.

It is important to note that this question has not been addressed to date in any study, with any research design. The question, in fact, can only arise when RJ is used on such a broad scale that it could theoretically alter the calculations of offenders or potential offenders in deciding whether to commit a crime. Suppose, for example, that a city were to divert over half of all its arrests for violent offences into RJ, as opposed to CJ. Some might argue that a level of diversion that high would undermine the effectiveness of the threat of punishment that results from the expectation that offenders who are arrested will be prosecuted and punished. Many readers may nod their heads in agreement with this hypothesis.

The evidence shows, however, that in England and Wales, diversion of half of all arrests for violent crime already occurs.27 That is, one report claims that around half of all arrests for violent crimes result in a police caution. These cautions are unlikely to involve much restorative work by offenders in relation to victims, based on the evidence of current RJ operations. What effect they have at the individual or community level is unknown. But because it is current practice, rather than the myth of 100% prosecution, which forms the baseline for discussing changes in policy, we may actually be justified in ruling out the “loss of deterrence” scenario.

One policy response to that conclusion might be the opposite: that there should be no cautioning of any violent crime defendants, and that all should be brought into court. Yet these cautions may reflect problems of evidence, limited resources, or other factors associated with the complex requirements of proving a case in court. Many of those requirements vanish when offenders accept responsibility, and proceed directly to a discussion of how they could repair the harm.

It is not clear whether an RJ discussion would decrease – or perhaps increase – the burdensome consequences of committing a crime. From a deterrence standpoint, an RJ agreement could possibly even increase the general deterrent effects of law enforcement on crime. This could be even more likely in contrast to an unconditional caution.

There is no way to resolve this question in theory. What is required to undertake a full assessment of the effects of RJ on crime is a demonstration of RJ as a community-wide phenomenon, a default option by which all cases are initially processed unless offenders or victims refuse. We have predicted

  1. Ford, Richard “Violent Criminals are Avoiding Jail as Police Issue More Cautions”,

The Times (24 November 2006).


that if RJ is used on a much larger scale, from schools to policing to probation, take-up rates may well increase. If and when they do, it would be possible to assess the effects of RJ on a wide range of outcomes, at least if there is a systematic comparison between communities using this approach and similar communities that do not.

Other effects of RJ on society and government could be obtained from community-level tests that have never been done. For example, would an RJ community produce a level of social capital and community building such that those factors alone would help to reduce crime levels? Irrespective of general deterrent effects, would less crime result from an increase that RJ might cause in such characteristics of communities as “intergenerational closure” (where parents know the parents of the young people their own children play with, and vice versa)? Would face-to-face recognition of community residents increase to a point where more informal surveillance would provide a greater level of general deterrence? Would RJ increase community respect for the law, and legitimacy of the police, in ways that would also increase compliance with the law? Or would none of these effects occur?

Related questions could be raised about community support for the rehabilitation of offenders. RJ is often hypothesised as something that could help motivate offenders to seek and complete rehabilitation. It is just as plausible that RJ could motivate law-abiding community members to support offender resettlement in far more powerful ways than at present. RJ agreements between offenders and victims to stay in touch are not uncommon. If a sufficient density of such agreements in a community were to be put in place, it could create a new culture around the idea of helping offenders go straight. The growing use of “circles of support and accountability” for sex offenders serving community sentences, such as the Quaker partnership in Thames Valley, is but one example of such long-term support.28

Finally, the evidence that RJ consistently reduces victim desire for violent revenge against their offenders could result in a reduced level of such violence in RJ communities. Many crimes of violence may be committed out of motives of revenge. Nipping that motivation in the bud with an RJ process could become a further mechanism for RJ to reduce crime. All of these mechanisms, if properly tested, could then be entered into the cost-effectiveness calculations of RJ to be applied by Treasury analysts in estimating the value of investing in its development.




  1. Getting cases brought to justice


Getting cases brought to justice

Before RJ can be rolled out in high volume, substantial barriers to giving more victims RJ must be removed. The evidence shows that when diversion to RJ is available, many more offences can be brought to justice than when it is not (table 5). Current policies and statutes in England and Wales, however, bar the widespread use of diversion to RJ. This section reviews that evidence. It then describes the barriers to diversion, and concludes with an alternative strategy of intensified supplementation of conventional justice with RJ.

  1. RJ effects on offences brought to justice

Table 5 summarises five RCTs testing diversion from court to RJ. One of these tests was conducted in Brooklyn, New York, where prosecutors approved a high volume of serious felonies to be diverted to victim-offender mediation. These cases were all felonies, or crimes punishable in New York State by more than one year in prison. The other four RCTs were in Canberra, encompassing over 1,000 arrests for violent, property, shoplifting and drink-driving crimes.

Table 5 compares the rate at which the randomly assigned cases were “brought to justice” in both the RJ and control groups. The concept of diversion from prosecution is often misunderstood to imply that prosecution would otherwise occur in 100% of the cases. What is actually meant, however, is that prosecution will proceed in the usual way, with offenders absconding and charges dismissed for lack of witnesses. Both these and other contingencies affect the rate at which offences are brought to justice, as the British government defines that phrase, both by prosecution and by other means.

In the Brooklyn experiment (Davis et al, 1981), for example, almost three out of four cases randomly assigned to prosecution as usual were never brought to justice. Dismissals and absconding accounted for most of the case attrition. In the RJ (mediation) group, by contrast, 56% of the offences completed the process. From the perspective of both victims and offenders, it seems fair to say that these mediation cases were “brought to justice”. Certainly, victims were more satisfied in these cases than in the prosecution cases (p51). When the percentage of randomly assigned cases that were resolved in each group is compared as a ratio, the odds of bringing the crime to justice were over 2 to 1 in favour of RJ.

Some might object that the defendants may have been factually innocent in the prosecution group, and that there was no offence to bring to justice in those cases. That hypothesis, however, can largely be eliminated by random assignment. The proportion of truly innocent defendants should have been roughly the same in each group. The fact that offenders in the RJ group were twice as likely to accept an ultimate decree of justice as offenders in the CJ group says much more about the process than it says about true guilt or innocence.

A similar procedure in the four RISE experiments shows surprisingly favourable odds for diversion to RJ over prosecution as usual. A case of violent crime against a youth under 30 is twice as likely to be brought to justice with RJ as with CJ, just as in Brooklyn. A case of property crime with a personal victim against a Canberra juvenile is three times more likely to be brought to justice with a completed RJ conference than a case assigned to prosecution as usual. A case of shoplifting from a department store showed the biggest effect of RJ, with four times higher odds of an offence being brought to justice for the RJ group than for the CJ prosecution group. The only offence type in which the odds of justice were not different was drink-driving, for which the result was largely determined the moment offenders failed the breathalyser test.

Understanding these results does not require the reader to accept a special magic from RJ. Rather, it requires understanding just how difficult it is to bring offences to justice under the strict requirements of the formal legal process. This issue has received public attention in recent years in the UK, but perhaps insufficient nuance. The high attrition rates in prosecution are not necessarily due to incompetence or even intimidation. They are more likely to indicate a disconnection between the unpredictable, multi-day court procedures for trial and the predictable, one-off structure of an RJ conference. Consider how many people must appear at exactly the right time on how many different days in order to achieve a criminal conviction at trial. Every one of those deadlines holds the potential of losing the case to justice. RJ lowers that risk level by lowering the number of times and dates, and even more by making it predictable – and convenient – to the key parties concerned.

The fact that many more offenders take responsibility for the crimes with RJ does not necessarily indicate any kind of unfair inducement. It may simply indicate their preference for a path of least resistance. If RJ makes it quicker and easier for them to “give it up” and get it over with, they may prefer to game the system that way rather than by playing cat and mouse with the court process. Evidence of such attitudes may be indicated by the substantial differences in rates of guilty pleas across demographic groups and regions within Britain. Whether or not we want defendants to think in such terms will not change the truth about how they in fact make such calculations.

Whether getting more offenders to calculate their way into an OBTJ will reduce repeat offending remains unknown. The only way we can find out, of course, is to make RJ much more widely available in the UK, on the same lines as in the US and in Australia. That course of action, however, faces obstacles of policy and statute.

  1. Getting cases and opening doors to RJ

The basic barrier to more widespread use of RJ is the reluctance


to make decisions that could divert offenders from a custodial sentence. While individual judges in London have changed their views on custodial sentences based on RJ, it may be more difficult for policy makers and prosecutors to make such decisions. The fear of something terrible being done by an offender who was diverted from custody may lead to a decision that RJ is “not in the public interest”. As long as such rules are discretionary and based on unspecified criteria for what is in the public interest, RJ leaders will find it difficult to shift such decisions from their current course.

Ironically, there is actually little chance of most potential diversion cases ever resulting in a sentence, let alone a custodial sentence, as shown by the evidence cited above. Yet the one case in a million that becomes a cause célèbre of “soft justice” is enough to prevent the rest from being diverted to RJ.

The failed effort to introduce conditional cautioning with RJ in London in 2004-05 provides ample evidence of this problem. Yet the source of the problem was arguably not with the CPS, nationally or locally. Rather, the statutory restrictions in the Criminal Justice Act 2003 made diversion almost impossible, despite its stated intent to implement Lord Justice Auld’s recommendation for such RJ-based cautions.

This 2003 act required full admissions of a crime in order for an offender to become eligible for RJ diversion. It did not acknowledge the evidence that such admissions are almost never made in the early stages of criminal processing. This provision of the act was inserted after questions were raised in the House of Lords, focused on the propriety of any “inducement” of the offender to make admissions. While the UN standards forbid any unfair inducement, they do not forbid inducement altogether. As a matter of practice, the option of a caution is “explained” to offenders all over the UK every day. Whether it constitutes an inducement is a matter of debate.

What is not a matter of debate is that diversion to RJ via conditional caution simply did not happen in London. Despite substantial investments of public and private funding to train police and prosecutors, to search for eligible cases and bring them forward, and to provide a team of experienced police officers to lead RJ, prosecutors approved fewer than 10 cases in one year. The full admissions requirement for cautioning via RJ effectively prohibited them from doing more, under the requirements of the 2003 act.

In order for RJ to be used more widely, and to bring far more offences to justice, these obstacles must be altered. An evidence- based review of the full admissions requirements, based on the evidence of what is possible, is needed in order to allow more victims the possibility of electing to meet with their offenders. That is not possible as long as offenders are required to meet a bar that is far higher than the standard for diversion eligibility that is widely used in Australia, New Zealand, Canada and the US.

  1. Getting victims RJ with CJ

While diversion may offer the greatest potential for increasing the proportion of offences brought to justice, supplementation may offer more victims RJ. It is also possible to offer more supplementation in the short run, since the process is tested and feasible under current UK practices.

The highest priority for supplementing CJ with RJ would be serious cases under Crown Court review for sentencing decisions. It is these cases in London for which victim benefits have been so clearly demonstrated, with potential cost savings to government as well. Cases are readily identified by pre-sentence report requests from judges to the National Probation Service – as long as pre-sentence reports themselves continue to be conducted. The Metropolitan Police RJ unit that provided conferences in such cases could be reactivated with central government funding, at least in the short run. It could also test the concept of civilian “restorative justice officers”, drawn from more diverse communities than the constabulary. Other regions could replicate the London approach, with broad national benefits for victims.

Another priority would be to resume police-led RJ for youth reprimand and final warning cases of violence and property crimes. The Northumbrian tests could be replicated on more ethnically diverse samples, in order to reduce concern about criminogenic effects in minority populations. The test, and RJ as a policy, could also be extended into RJ conferences after court convictions, via youth offender referral panels.

We are still awaiting results from the post-sentence adult experiments, which may also benefit victims. If there is no indication of criminogenic effects with these populations (inmates and probationers on community sentence), the benefits of RJ for victims could be extended even further.

The supplementation strategy could make RJ far more widespread in the short run. That could lay a foundation of public support for increasing use of diversion. As a first step, it provides the clearest path to getting more cases brought to (restorative) justice.



  1. Could RJ reduce the financial costs of justice?


Could RJ reduce the financial costs of justice?

Whatever the evidence on the benefits of RJ, it cannot be expanded without a consideration of costs. That consideration entails two options. One is adding new money to the criminal justice system: a budget increase. The other is reallocating existing funding for criminal justice: no budget increase. Of the two, the second is clearly more feasible.

There are at least three ways in which RJ could reduce costs to government. One is to reduce the use of courts for processes that fail to bring offences to justice. A second is to reduce the use of prisons for offenders whose incarceration does not prevent total crime. The third is to reduce the health costs incurred by failure to treat crime-related post-traumatic stress symptoms (PTSS).

RJ versus courts

The evidence shows that, when it is allowed, diversion from court to RJ conferences can increase offences brought to justice. What is does not show is just how much money could be saved by such a policy. Diverting more cases from court that would ultimately be dismissed might bring down the costs of prosecution with no impact on offences brought to justice. It could even raise the count of total offences brought to justice, while bringing down the cost per offence brought to justice. These costs would include legal aid solicitors and barristers, court security and clerical personnel, and time spent by police and prison officials in going to and from court. The costs of organising a single RJ conference appear likely to be far less, even with substantial monitoring of post-conference compliance with agreements.

Better and more realistic estimates of these savings, however, require cost calculations that are not bogged down by start-up

costs associated with a pilot. A “scaling up” of RJ to a level of volume in which economies of scale could bring down the unit cost per case would provide more realistic estimates. That is exactly the kind of evidence that could result from the testing of the concept of “RJ communities” (see section 15).

RJ versus custody

At £35,000 per year for each offender in custody, even a 10% reduction in custodial population in favour of an RJ alternative could yield substantial cost savings. The evidence that RJ does as well as custody is admittedly modest, but consistent. The debate over short sentences could be addressed by pitting RJ against short sentences, to see which approach costs less and yields less crime over a multi-year follow-up period. Short sentences are especially vulnerable to the result of more total crime than no custody, on the theory that a short sentence increases offender anger (and desire for revenge) more than enough to make up for any crimes prevented by mere incapacitation behind bars for a short time.

Reduced health costs from RJ

The effects of RJ on PTSS may have substantial impact on long- term health costs. These savings have not been estimated, so they pose another evidence gap in thinking about RJ. But they could be measured in collaboration with the National Health Service or the Medical Research Council. Offering the possibility of discoveries in basic science as well as in treatment outcomes, a long-term test of the effects of RJ could show another way by which RJ could pay for itself.


  1. More justice, less crime: a way forward


More justice, less crime: a way forward

This section begins with a restatement of various conclusions and recommendations offered throughout the report. It then offers a description of the kind of restorative community that could be tested prior to a wider national roll-out of such a concept. It concludes by suggesting an institutional mechanism that could provide energy and guidance for increasing the pace of development for RJ in the UK.

  1. Conclusions and recommendations

Conclusions from the evidence

  1. Crime victims who receive restorative justice do better, on average, than victims who do not, across a wide range of outcomes, including post-traumatic stress.
  1. In many tests, offenders who receive restorative justice commit fewer repeat crimes than offenders who do not.
  1. In no large-sample test has RJ increased repeat offending compared with CJ.
  1. RJ reduces repeat offending more consistently with violent crimes than with less serious crimes.
  1. Diversion from prosecution to RJ substantially increases the odds of an offender being brought to justice.
  1. Diversion from prosecution to RJ is almost impossible in the UK because of an evidentiary requirement that far exceeds the bar found in other common law nations.
  1. The best evidence on success in implementing RJ from tests to date is associated with specially trained police officers providing RJ from a base in a police RJ unit.
  1. RJ does not conflict with the rule of law, nor does it depart from the basic paradigm of the common law of crime.
  1. RJ can do as well as, or better than, short prison sentences, as measured by repeat offending.
  1. RJ reduces stated victim desire for violent revenge against offenders.

Recommendations for evidence-based policy in the UK

  1. RJ can be rolled out across the country with a high probability of substantial benefits to victims and crime reductions for many kinds of offenders.
  1. RJ can be resumed and expanded immediately as a supplement with substantial victim benefits in cases awaiting Crown Court sentencing.
  1. RJ can be resumed and expanded immediately as a supplement

with substantial crime reduction effects to young offenders in final warnings: girls for assault, boys for property crime.

  1. RJ as a supplement can be resumed and expanded for victims with inmates about to return from prison, as soon as repeat offending results in Thames Valley are completed to show no risk of increase from RJ.
  1. RJ as a supplement can be continued for offenders sentenced to community sentences, pending repeat offending results in Thames Valley.
  1. RJ as diversion could provide the basis for far more general use of RJ, with possibly substantial crime reductions, less victim post-traumatic stress, and more offences brought to justice.
  1. A “Restorative Justice Board” (RJB) or similar institution is needed to provide guidance, set standards, monitor programmes, and provide research and development for continuously improving best practice.
  1. An RJB could develop RJ in an evidence-based way, on a continue-to-learn-as-you-go basis.
  1. Testing restorative communities

The major evidence gap in restorative justice is the evidence on its scaling-up. What would happen if RJ were delivered on a widespread basis, rather than in small pilot tests that affect a tiny fraction of cases in any local justice system? Answering that question requires some precision about what such a community would look like. Most important, we must be clear on how its criminal justice system would work.

A short answer to these complex questions is that a restorative community would employ graduated sanctioning of its offenders, using restorative solutions first and then increasingly coercive solutions for those who persist in harming others. This model is derived from Braithwaite’s (2002) sanctioning pyramid, with the highest volume of cases at the bottom, at the level of the most restorative responses to crime. As the community goes up the pyramid, to fewer crimes committed respectively by second, third, fourth and fifth repeat offenders, the criminal penalties become correspondingly less restorative for offenders – but no less so for victims.

Offenders would be subjected to increasing constraints on their freedom if they persisted in crime, but they would always be encouraged to meet with their victims to explain the crime and offer to make amends. The challenge at these higher levels will always be to balance past victim needs against the potential for creating future victims. Electronic tagging and home detention would come into play along with making amends at a relatively low to middle level of the pyramid. When offenders’ crimes become

serious enough, the community would simply have to remove the offender altogether, as in a life sentence for murder. Even then, however, it may benefit the survivors to be able to meet with the offender, as they have done in both the US and the UK.

Most important in this pyramid is what happens at the bottom. That is where a “default” response to first crime at any age would be to seek a restorative process, possibly with diversion from prosecution (up to, but not beyond, offences as serious as those tested in Brooklyn by Davis et al, 1981). This default response may even continue well beyond the first offence, if the offenders seem sincere and victims are willing.

Among neighbourhood police, RJ could become the tool of first resort for responding to matters when they are called in by citizens. Their operating framework could shift from the recent emphasis on such questions as “Was any law was broken here?” or “Is the evidence sufficient to charge anyone with a crime?” to Sir Robert Peel’s historic emphasis on preventive policing (Miller, 1977) with such questions as these:

  • What happened?
  • Who was affected by it and how?
  • What can be done to repair the harm?
  • Does anyone accept responsibility for having caused the harm?
  • If so, would responsible parties be willing to negotiate how to repair the harm?
  • Even if there is sufficient evidence that a crime has occurred, and that a certain person is responsible, is arrest and criminal prosecution the most effective way to deal with the harm – or is a negotiated agreement more likely to repair past harm and prevent future crime?
  • If the effort to reach such an agreement in the first instance fails, is there a police or RJ unit with special skills that could undertake a second or third attempt to bring offender and victim together in a way that can restore the status quo ante, rather than potentially inflaming the case with legal actions that may harden emotions of anger and distrust?
  • What sanction, if any, is required in addition to the repair of material or emotional harm?

There will continue to be many cases in which police decide to proceed by arrest and prosecution. It is for repeat offenders in such cases that the Canadian idea of sentencing circles, with a judge or magistrate in the room, along with victims and offenders, may need to replace purely community-based justice. Reaching an agreement that can become the basis of a criminal sentence by restorative processes could add extra force to its crime prevention effects. But there must also be safeguards for compliance with sentencing guidelines and other national policies.

The underlying idea is to maximise restorative practices, even while staying within the range of current sentencing outcomes. The same sentence, reached by RJ, may be far more effective in curbing future crime than if it is imposed without such deliber-

ations. This hypothesis can be tested, along with many others like it, in a “nested” design of experiments within experiments. Communities with these and other strategies could provide a variety of models of restorative communities, based upon local preference. That, in turn, could provide the basis for a much larger test comparing restorative communities – possibly centred on magistrates’ court areas – to conventional justice communities.

Among the general public, an RJ community would feature familiarity with RJ processes and their use in resolving conflicts. RJ in schools, for example, could proceed by offering classes to parents or the general public in the RJ processes schools use to deal with bullying, thefts and minor violence among students. Parental involvement in some of these processes would further educate adults in the restorative framework for win-win responses to harms, made possible both by victims’ willingness to talk to those who have harmed the victims, and a harmer’s incentive to take responsibility for their actions as the first step in repairing the harm. This might prepare neighbours to deal with more of their own problems, from barking dogs to parking space disputes, rather than consuming expensive police time on such matters.

In sum, the key features of restorative communities would be, subject in all cases to victim and offender consent:

  • diversion to RJ if there is a willing personal victim on a first offence, for all offences included in the Brooklyn test (Davis et al, 1981);
  • consideration of diversion for subsequent offences;
  • post-guilty plea, pre-sentence RJ for subsequent offences;
  • experimentation with sentencing circles;
  • post-sentence restorative sentence planning for community sentences;
  • combinations of home detention and tagging with RJ;
  • RJ in prison for those about to be released;
  • RJ upon resettlement to form support circles and reintegration;
  • RJ in other community conflicts, including schools and neighbourhoods.

The key hypotheses to be examined for such communities, compared with conventional justice communities, would be that RJ communities, other things equal, would have:

  • less crime;
  • better reduction in harm to crime victims;
  • more offences brought to justice;
  • less cost of justice;
  • greater public confidence in justice.

There is an enormous difference in the capacity to test these hypotheses between developing major pilot sites and conducting small randomised trials. All tests of RJ done to date, as both a diversion and a supplement, have been mere blips on the justice radar screen. None of them have confronted the multiple dimens- ions of turning RJ into CJ, in a mainstream way rather than as an


exotic but rare option. The evidence required must tackle a systemic change in how everyone thinks about justice in a community, from the general public to officials right across every agency of justice. Only then can research determine both the problems of imple- mentation, and the evidence on results, of a large-scale use of RJ.

The crucial decision in testing RJ communities is the definition of a “community”. For this purpose, the scope of a community must be larger than a police beat but smaller than an entire police authority. A valid test would require a large volume of cases ranging across the spectrum of crime seriousness, without too large an area to measure reliably. Perhaps the ideal solution is to link a magistrates’ court to one or several police basic command units, or the evolving equivalent. Several hundred police officers would then become the core of the test, in their application of the questions listed above to the full array of events they are asked to resolve.

The link to a magistrates’ court would provide a coherent focus for using RJ within the criminal justice process. Training of court clerks and magistrates could be focused not only on principles but also on evidence. Progress reports on a monthly or quarterly basis could show how well crime is being prevented (in comparison with OGRS2 scores or predicted local trends) and how many victims are being served by RJ compared with a baseline period. This process of encouraging and supporting an RJ community could be led by existing services, such as probation, or by an evolving institution similar to youth offending teams. Interagency, multidisciplinary, adult-focused RJ teams could provide both the expertise and local evidence for a community-wide drive to increase justice and prevent crime.

There is no doubt that such tests would reveal many problems with the administration and implementation of RJ. By analysing those problems in relation to the operation of CJ, however, the tests we propose would focus on the difference between the systems rather than on some ideal of perfection. Even targets could be defined in relation to such differences, rather than in absolute terms. The evidence needed, like the evidence in hand, is not to show that RJ is a perfect system. All the evidence that we have to date, and all that new tests would need to confirm, is that conventional justice can do better with RJ in its toolbox.

  1. How can development and testing of RJ be accomplished?

In the past five years, several parts of government have been engaged in developing RJ. None, however, has been able to take primary responsibility. The Home Office has invested the most funding, but the cases have remained in the province of the courts, the CPS, the Department for Constitutional Affairs, and even the Department for Education & Skills. No one minister or ministry can be said to have a portfolio for restorative justice that cuts right across government.

The time may be right to pinpoint responsibility and

accountability for leading on restorative justice. An institutional framework for doing that has a precedent in the form of the Youth Justice Board. Such a board brings in representatives of diverse constituencies affected by the policies under development. It may also speak as a body that is quasi- independent of government, and hence able to push the envelope of innovation somewhat further and faster. In contrast to the YJB, however, a “Restorative Justice Board” need not take over any massive on-going operations of a justice system. It need not create RJ teams around the country, although that could be an option to explore. It need not manage a large prison estate, or see its budget consumed by sentencing decisions beyond its control – all of which have restricted the YJB.

Similarly, an RJB could avoid the pitfalls of a national roll-out with no institutional oversight. The conversion of the New Zealand youth justice system to RJ almost two decades ago, for example, lacked such a board. New Zealand also failed to engage some half of the crime victims whose offenders were brought to justice (Maxwell and Morris, 1993). Had there been the accountability and oversight provided by an RJB in New Zealand, there might have been more successful efforts to increase victim participation in the process, thereby greatly increasing the potential return on a national investment in RJ.

Similarly, we note that efforts to advance RJ to date have required extensive consultation across agencies of government. The leadership in such consultations is unclear because each agency is in theory co-equal with all others. Designating a board to take the lead would cut through the unavoidable issues of separate departments sharing pieces of the criminal justice process.

Rather than serving as an operating agency, an RJB could become more of a national improvement agency for restorative justice. Much like the National Police Improvement Agency, an RJB could use information systems, research and standards to foster innovation and improvement. With a reasonable budget, it could provide grants of local assistance to criminal justice agencies willing to invest in RJ, test RJ, or even just explore RJ. Its budget could support projects as modest as reinstituting the Crown Court RJ services, or as broad as piloting a restorative community or two. It could also become a repository of research, training and consultation, providing technical assistance to community groups, schools and human services agencies, as well as justice agencies.

As a board with multiple members, an RJB could have memberships designated for expertise in policing, prosecution, defence, adjudication, probation, prisons, rehabilitation and community organisation. While there are other ways to provide such expertise, such diversity would offer legitimacy to its voice that could help gain public understanding of how RJ works. Whether as a board or some other structure, an agency with pinpointed responsibility to take RJ forward may be the best option for making the most of the evidence so far.


Bibliography of research cited or examined


Bibliography of research cited or examined

Aber, L, Brown, J and Henrich, C Teaching Conflict Resolution: An Effective School-Based Approach to Violence Prevention (New York: National Center for Children in Poverty, 1999)

Angel, C Crime Victims Meet Their Offenders: Testing the Impact of Restorative Justice Conferences on Victims’ Post-Traumatic Stress Symptoms, PhD dissertation (University of Pennsylvania, 2005)

Angrist, J, Imbens, G and Rubin, D “Identification of Causal Effects Using Instrumental Variables” in Journal of the American Statistical Association, 91 (1996), pp444-455

Banton, M The Policeman in the Community (London: Tavistock, 1964)

Beckett, H, Campbell, C, O’Mahony, D, Jackson, J and Doak, J “Interim Evaluation of the Northern Ireland Youth Conferencing Scheme” in Research & Statistical Bulletin, 1/2005 (Belfast: Northern Ireland Office, 2005)

Berk, R and Rossi, P Thinking About Programme Evaluation (Newbury Park, California: Sage, 1998)

Beven, J, Hall, G, Froyland, I, Steels, B and Goulding, D “Restitution or Renovation? Evaluating Restorative Justice Outcomes” in Psychiatry, Psychology & Law 12 (2005), pp194-206

Bonta, J, Wallace-Capretta, S and Rooney, J Restorative Justice:

An Evaluation of the Restorative Resolutions Project (Ottawa: Ministry of the Solicitor-General, 1998)

Bradbury, B Audit Report: Deschutes County Delinquent Youth Demonstration Project, Report No 2002-29 (Eugene, Oregon, 2002)

Braithwaite, J Crime, Shame & Reintegration (Cambridge: Cambridge University Press, 1989)

Braithwaite, J “Restorative Justice” in Tonry, M (ed) Handbook of Crime & Punishment (Oxford: Oxford University Press, 1998)

Braithwaite, J Restorative Justice & Responsive Regulation (Oxford: Oxford University Press, 2002)

Bright, P Research into Domestic Violence & the Use of Family Group Conferences, Fulbright Commission report (2002), posted by Hampshire Constabulary at 5864-478B-965A-1159FF386AFA/0/Fulbright.pdf

Burford, G and Pennell, J Family Group Decision Making: After the Conference – Progress in Resolving Violence & Promoting Well-being: Outcome Report, Volumes 1 & 2 (St John’s, Newfoundland, Canada: Memorial University of Newfoundland School of Social Work, 1997 and 1998)

Cain, M Society & the Policeman’s Role (London: Routledge, 1973)

Calhoun, A Calgary Community Conferencing School Component 1999- 2000: A Year in Review (2000) (http://www.calgarycommunity

Cameron, L and Thorsborne, M “Restorative Justice and School Discipline: Mutually Exclusive?” in Strang, H and Braithwaite, J (eds) Restorative Justice & Civil Society (Cambridge: Cambridge University Press, 2001)

Campbell, C, Devlin, R, O’Mahony, D, Doak, J, Jackson, J, Corrigan, T and McEvoy, K Evaluation of the Northern Ireland Youth Conference Service, NIO Research & Statistical Series: report no 12 (Belfast: Northern Ireland Office, 2005) ( youth_conference_service.pdf)

Carr, C Victim-Offender Reparation Programme Evaluation Report

(Inglewood, California: Centenela Valley Juvenile Diversion Project, 1998)

Chapman, B, Freiberg, A, Quiggan, J and Tait, D Rejuvenating Financial Penalties: Using the Tax System to Collect Fines, discussion paper no 461 (Canberra: Centre for Economic Policy Research, Australian National University, 2003)

Christie, N “Conflicts as Property” in British Journal of Criminology vol 17 (1977), pp1-15

Cormier, R Restorative Justice: Directions & Principles – Developments in Canada (Ottawa: Department of the Solicitor General, Canada, 2002)

Crawford, A and Burden, T Integrating Victims in Restorative Youth Justice, Researching Criminal Justice series (Bristol: Policy Press, 2005)

Crawford, A and Newburn, T Youth Offending & Restorative Justice: Implementing Reform in Youth Justice (Cullompton, Devon: Willan Publishing, 2003)

Daly, K “Making Variation a Virtue: Evaluating the Potential and Limits of Restorative Justice” in Weitekamp, EGM and Kerner, H-J (eds) Restorative Justice in Context: International Practices & Directions (Cullompton, Devon: Willan Publishing, 2003)

Davis, R, Tichane, M and Grayson, D Mediation & Arbitration as Alternatives to Criminal Prosecution in Felony Arrest Cases: An Evaluation of the Brooklyn Dispute Resolution Center (First Year) (New York: Vera Institute of Justice, 1980)

Dignan, J “Repairing the Damage: Can Reparation Be Made to Work in the Service of Diversion?” in British Journal of Criminology vol 32, no 4 (1992), pp453-472

Dignan, J Understanding Victims & Restorative Justice (Cullompton, Devon: Willan Publishing, 2005)

Dravery, W and Winslade, J Developing Restorative Practices in Schools: Flavour of the Month or Saviour of the System? (University of Waikato, School of Education, 2006) (

Evje, A A Summary of the Evaluations of Six California Victim Offender Reparation Programmes, report to the California legislature submitted by the Judiciary Council of California (2000)

Fisher, R The Design of Experiments (Edinburgh: Oliver & Boyd, 1935)

Gehm, J “Mediated Victim-Offender Restitution Agreements: An Exploratory Analysis of Factors Relating to Victim Participation” in Galaway, B and Hudson, J (eds) Criminal Justice, Restitution & Reconciliation (New York: Criminal Justice Press, 1990)

Gesch, B, Bernard, S, Hammond, M, Hampson, S, Eves, A and Crowder, M “Influence of Supplementary Vitamins, Minerals and Essential Fatty Acids on the Antisocial Behaviour of Young Adult Prisoners: A Randomised, Placebo- controlled Trial” in British Journal of Psychiatry 181 (2002), pp22-28

Giordano, P, Cernkovich, S and Rudolph, J “Gender, Crime and Desistance: Toward a Theory of Cognitive Transformation” in American Journal of Sociology 107 (2002), pp990-1064

Gladwell, M The Tipping Point (Boston: Little, Brown, 2000)

Gottfredson, D “School-Based Crime Prevention” in Sherman, L, Gottfredson, D, MacKenzie, D, Eck, J, Reuter, P and Bushway, S (eds) Preventing Crime: What Works, What Doesn’t, What’s Promising – A Report to the United States Congress (Washington DC: National Institute of Justice, 1997)

Griffith, M The Implementation of Group Conferencing in Juvenile Justice in Victoria, paper presented at the Restoration for Victims of Crime conference convened by the Australian Institute of Criminology and the Victims Referral & Assistance Service, Melbourne (September 1999)

Hayes, H and Daly, K “Youth Justice Conferencing and Reoffending” in

Justice Quarterly vol 20, no 4 (December 2003), pp725-764

Hayes, H, Prenzler, T and Wortley, R Making Amends: Final Evaluation of the Queensland Community Conferencing Pilot (Brisbane: Griffith University, 1998)

Holdaway, S, Davidson, N, Dignan, J, Hammersley, R, Hine, J and Marsh, P New Strategies to Address Youth Offending: The National Evaluation of the Pilot Youth Offending Teams (London: Home Office, 2001)

Home Office Restorative Justice: The Government’s Strategy, consultation document on the government’s strategy on restorative justice (2003) ( consult.pdf?view=Binary)

Home Office Home Office RDS & YJB Standards for Impact Studies in Correctional Settings (11 February 2004) (

Hopkins, B Just Schools: A Whole-School Approach to Restorative Justice

(London and New York: Jessica Kingsley, 2004)

Hoyle, C, Young, R and Hill, R Proceed With Caution: An Evaluation of the Thames Valley Police Initiative in Restorative Cautioning (York: Joseph Rowntree Foundation, 2002)

Hudson, C and Pring R Banbury Police Schools Project: Report of the Evaluation, unpublished manuscript held by the Thames Valley Police (2000)

Ierley, A and Ivker, C Restoring School Communities. Restorative Justice in Schools Programme: Spring 2002 Report Card, unpublished manuscript held by the School Mediation Center, Boulder, Colorado (2002)

Judge, N, Mutter, R, Gillett, T, Hennessey, J and Mauger, J Executive Summary: From Evaluation of 30 Restorative Justice FGCs (Essex Family Group Conference – Young People Who Offend Project, Essex County Council, 2002)

Kane, J, Lloyd, G, McCluskey, G, Riddell, S, Stead, J and Weedon, E Restorative Practices in Three Scottish Councils, final report of an evaluation funded by the Scottish Executive Education Department (Edinburgh: Scottish Executive Education Department, 2006)

Kubzansky, L, Koenen, K, Spiro, A, Vokonos, P and Sparrow, D “Prospective Study of Posttraumatic Stress Disorder Symptoms and Coronary Heart Disease in the Normative Aging Study” in Archives of General Psychiatry 64 (2007), pp109-116

Latimer, J, Dowden, C and Muise, D The Effectiveness of Restorative Justice Practices: A Meta-Analysis (Ottawa: Canadian Department of Justice, 2001)

Luke, G and Lind, B “Reducing Juvenile Crime: Conferencing vs Court” in Crime & Justice Bulletin no 69 (Sydney: New South Wales Bureau of Crime Statistics & Research, 2002)

Marsh, P Supporting Pupils, Schools & Families: An Evaluation of the Hampshire Family Group Conferences in Education Project, unpublished manuscript held by the University of Sheffield (2004)

Marshall, T Restorative Justice: An Overview (London: Home Office, Research, Development & Statistics Directorate, 1999)

Marshall, T and Merry, S Crime & Accountability: Victim Offender Mediation in Practice (London: Home Office, 1990)

Maruna, S Making Good: How Ex-Offenders Reform & Reclaim Their Lives

(Washington, DC: American Psychological Association, 2001)

Maxwell, G and Morris, A Families, Victims & Culture: Youth Justice in New Zealand, (Wellington: Social Policy Agency and Institute of Criminology, Victoria University of Wellington, 1993)

Maxwell, G and Morris, A “Family Group Conferences and Reoffending” in Morris, A and Maxwell, G (eds) Restorative Justice for Juveniles: Conferencing, Mediation & Circles (Oxford: Hart, 2001), pp243-263

McCold, P and Wachtel, B Restorative Policing Experiment: The Bethlehem Pennsylvania Police Family Group Conferencing Project (Pipersville, Pennsylvania: Pipers Press, 1998)

McCulloch, H Shop Theft: Improving the Police Response, Crime Detection & Prevention Series, paper 76 (London: Home Office, Police Research Group, 1996)

McGarrell, E, Olivares, K, Crawford, K and Kroovand, N Returning Justice to the Community: The Indianapolis Restorative Justice Experiment (Indianapolis, Hudson Institute, 2000)

Miers, D An International Review of Restorative Justice, Crime Reduction Series, paper 10 (London: Home Office, Policing & Reducing Crime Unit, 2001)

Miers, D, Maguire, M, Goldie, S, Sharpe, K, Hale, C, Netten, A, Uglow, S, Doolin, K, Hallam, A, Enterkin, J and Newburn, T An Exploratory Evaluation of Restorative Justice Schemes, Crime Reduction Research Series paper 9 (London: Home Office, 2001)

Miller, W Cops & Bobbies: Police Authority in New York & London, 1830–1870 (Chicago: University of Chicago Press, 1977)

Minnesota Department of Children, Family & Learning “In-School Behavior Intervention Grants” in A Three-Year Evaluation of Alternative Approaches to Suspensions & Expulsions, report to the Minnesota legislature (2002)

Moore, D and Forsythe, L A New Approach to Juvenile Justice: An Evaluation of Family Conferencing in Wagga Wagga (Canberra: Criminology Research Council, 1995)

Morris, A and Maxwell, G “Restorative Justice in New Zealand” in von Hirsch, A, Roberts, J and Bottoms, A (eds) Restorative Justice & Criminal Justice: Competing or Reconcilable Paradigms? (Oxford and Portland, Oregon: Hart Publishing, 2005)

Morris, A, Maxwell, G and Robertson, JP “Giving Victims a Voice: a New Zealand Experiment” in Howard Journal 32 (1993), pp304-21

Morrison, B “Developing the School’s Capacity in the Regulation of Civil Society”, in Strang, H and J Braithwaite, J (eds), Restorative Justice & Civil Society (Cambridge: Cambridge University Press, 2001)

Morrison, B “Restorative Justice in Schools” in Eliott, E and Gordon, R (eds) New Directions in Restorative Justice (Cullompton, Devon: Willan Publishing, 2005)

Morrison, B “Schools and Restorative Justice” in Johnstone, G and Van Ness, D (eds) Restorative Justice Handbook (Cullompton, Devon: Willan Publishing, 2006a)

Morrison, B Restoring Safe School Communities: A Whole School Response to Bullying, Violence & Alienation (Sydney: Federation Press, 2006b)

Morrison, BE and Martinez, M “Restorative Justice through Social and Emotional Skills Training: An Evaluation of Primary School Students”, unpublished honours thesis, Australian National University, Canberra (2001)

National Institute for Health & Clinical Excellence The Guidelines Manual: Overview & Contents (2006) (

Newburn, T, Crawford, A, Earle, R, Goldie, S, Hale, C, Masters, G, Netten, A, Saunders, R, Sharpe, K, Uglow, S and Campbell, A The Introduction of Referral Orders into the Youth Justice System, HORS 242 (London: Home Office, 2001)

Niemeyer, M and Stichor, D “A Preliminary Study of a Large Victim/Offender Reparation Programme” in Federal Probation 57 (1996), pp48-53

Northern Ireland Office Draft Protocol for Community-Based Restorative Justice Schemes (2006) ( community_based_restorative_justice_schemes.pdf)

Nuffield, J Evaluation of the Adult Victim-Offender Mediation Programme, Saskatoon Community Mediation Services (Regina, Saskatchewan: Department of Justice, Saskatchewan, 1997)

Nugent, W and Paddock, J “The Effect of Victim-Offender Mediation on Severity of Reoffense” in Mediation Quarterly vol 12, no 4 (1995), pp353-367

Nugent, W, Umbreit, M, Wiinamaki, L and Paddock, J “Participation in Victim-Offender Mediation Reduces Recidivism” in Connections Summer 1999, no 3 (1999)

Nugent, W, Williams, M and Umbreit, M “Participation in Victim-Offender Mediation and the Prevalence and Severity of Subsequent Delinquent Behavior: A Meta-Analysis” in Utah Law Review vol 2003, no 1 (2003), pp137-166

Pennell, J and Burford, G “Family Group Decision-Making: Protecting Women and Children” in Child Welfare vol 79, no 2 (March/April 2000)

Piantadosi, S Clinical Trials: A Methodologic Perspective

(New York: Wiley, 1997)

Poulson, B “A Third Voice: A Review of Empirical Research on the Psychological Outcomes of Restorative Justice” in Utah Law Review (2003), pp167-203

Roche, D Accountability in Restorative Justice (Oxford: Oxford University Press, 2003)

Roy, S “Two Types of Juvenile Restorative Programmes in Two Midwestern Counties: A Comparative Study” in Federal Probation 57 (1993), pp48-53

Salsburg, D The Lady Tasting Tea: How Statistics Revolutionized Science in the Twentieth Century (New York: Henry Holt & Co, 2001)

Schneider, A “Restitution and Recidivism Rates: Results from Four Experimental Studies” in Criminology vol 24 (1986), pp533-552

Shapland, J, Atkinson, A, Colledge, E, Dignan, J, Howes, M, Johnstone, J, Pennant, R, Robinson, G and Sorsby, A Implementing Restorative Justice Schemes (Crime Reduction Programme): A Report on the First Year, Home Office online report 32/04 (London: Home Office, 2004) (

Shapland, J, Atkinson, A, Atkinson, H, Chapman, B, Colledge, E, Dignan, J, Howes, M, Johnstone, J, Robinson, G and Sorsby, A Restorative Justice in Practice: The Second Report from the Evaluation of Three Schemes (Sheffield: Centre for Criminological Research, University of Sheffield, 2006)

Shaw, G and Wierenga, A Restorative Practices: Community Conferencing Pilot, unpublished manuscript held at the Faculty of Education, University of Melbourne (2002)

Sherman, L “Defiance, Deterrence and Irrelevance: A Theory of the Criminal Sanction” in Journal of Research in Crime & Delinquency vol 30 (1993), pp445-73

Sherman, L “Reason For Emotion: Reinventing Justice with Theories, Innovations and Research”, American Society of Criminology 2002 presidential address, in Criminology vol 41 (2003), pp1-38

Sherman, L, Gottfredson, D, MacKenzie, D, Eck, J, Reuter P and Bushway, S Preventing Crime: What Works, What Doesn’t, What’s Promising (Washington, DC: US Department of Justice, 1997)

Sherman, L, Strang, H and Woods, D, 2000, Recidivism Patterns in the Canberra Reintegrative Shaming Experiments (RISE) (Canberra: Centre for Restorative Justice, Research School of Social Sciences, Australian National University, 2000) (

Sherman, L and Strang, H “Verdicts or Inventions? Interpreting Results from Randomized Controlled Experiments in Criminology” in American Behavioral Scientist (special issue devoted to experimental methods in the political sciences edited by Donald P Green and Alan S Gerber,

Yale University), vol 47, no 5 (2003), pp575?607

Sherman, L, Strang, H, Angel, C, Woods, DJ, Barnes, GC, Bennett, S, Inkpen, N and Rossner, M “Effects of Face-to-Face Restorative Justice on Victims of Crime in Four Randomized Controlled Trials” in Journal of Experimental Criminology vol 1, no 3 (2005), pp367-395

Sherman, L, Strang, H, Barnes, G and Woods, D Preliminary Analysis of Race, Recidivism & Restorative Justice for Victimed Crimes in Canberra, unpublished manuscript, Lee Center of Criminology, Philadelphia, Pennsylvania (2006a)

Sherman, L, Strang, H, Barnes, GC and Newbury-Birch, D Preliminary Analysis of the Northumbria Restorative Justice Experiments, unpublished manuscript, Lee Center of Criminology, Philadelphia, Pennsylvania (2006b)

Sherman, L and Strang, H “Curing Revenge: Transforming Emotions with Restorative Justice” in Karstedt, S, Loader, I and Strang, H (eds) Emotions, Crime & Justice, proceedings of a conference held in September 2004 at the International Institute for the Sociology of Law, Onati, Spain (forthcoming)

Stone, S, Helms, W and Edgeworth, P Cobb County Juvenile Court Mediation Programme Evaluation (Children & Youth Coordinating Council and the State University of West Georgia, 1998)

Strang, H, Barnes, GC, Braithwaite, J and Sherman, L Experiments in Restorative Policing: A Progress Report on the Canberra Reintegrative Shaming Experiments (RISE) (Canberra: Australian National University, 1999) (

Strang, H Repair or Revenge: Victims & Restorative Justice (Oxford: Oxford University Press, 2002)

Strang, H and Sherman, L “Repairing the Harm: Victims and Restorative Justice” in Utah Law Review (2003), pp15-42

Strang, H, Sherman, L, Angel, C, Woods, D, Bennett, S, Newbury-Birch, D and Inkpen, J “Victim Evaluations of Face-to-Face Restorative Justice Experiences: A Quasi-Experimental Analysis” in Journal of Social Issues vol 62, no 2 (2006), pp281-306

Tilly, C Why? (Princeton, New Jersey: Princeton University Press, 2006)

Triggs, S New Zealand Court-Referred Restorative Justice Pilot: Two-Year Follow-Up of Reoffending (Wellington: Ministry of Justice, 2005)

Trimboli, L An Evaluation of the NSW Youth Justice Conferencing Scheme

(Sydney: New South Wales Bureau of Crime Statistics & Research, 2000)

Tyler, T Why People Obey the Law (New Haven, Connecticut: Yale University Press, 1990)

Umbreit, M “Crime Victims Confront Their Offenders: The Impact of a Minneapolis Mediation Programme” in Research on Social Work Practice vol 4, no 4 (1994), pp436-447

Umbreit, M The Handbook of Victim Offender Mediation (San Francisco: Jossey-Bass, 2001)

Umbreit, M and Coates, R “The Impact of Mediating Victim Offender Conflict: An Analysis of Programmes in Three States” in Juvenile & Family Court Journal (1992), pp21-28

Umbreit, M and Coates, R “Cross-Site Analysis of Victim-Offender Mediation in Four States” in Crime & Delinquency vol 39 (1993), pp565-585

Umbreit, M, Coates, R and Kalanj, B Victim Meets Offender: The Impact of Restorative Justice & Mediation (Monsey, New York: Criminal Justice Press, 1994)

Umbreit, M, Coates, R and Roberts, A “The Impact of Victim-Offender Mediation: A Cross-National Perspective” in Mediation Quarterly 17 (2000), pp215-229

Umbreit, M, Coates, R and Vos, B Juvenile Victim Offender Mediation in Six Oregon Counties: Final Report (Oregon Dispute Resolutions Commission, 2001)

Umbreit, M, Warner, S, Roberts, A, Kalanj, B and Lipkin, R Mediation of Criminal Conflict in England: An Assessment of Services in Coventry & Leeds – Executive Summary (Centre for Restorative Justice & Mediation, School of Social Work, University of Minnesota, 1996)

UK Select Committee of Public Accounts Collection of Fines & Other Financial Penalties in the Criminal Justice System (London: HMSO, 2002)

Victoria Auditor-General Report on Ministerial Portfolios (Melbourne, 1998)

Vignaendra, S and Fitzgerald, J “Reoffending among Young People Cautioned by Police or Who Participated in a Youth Justice Conference” in Contemporary Issues in Crime & Justice no 103 (Sydney: New South Wales Bureau of Crime Statistics & Research, 2006)

Wachtel, T and McCold, P “Restorative Justice in Everyday Life: Beyond the Formal Ritual” in Strang, H and Braithwaite, J (eds) Restorative Justice & Civil Society (Cambridge: Cambridge University Press, 2001)

Walker, L “Conferencing – A New Approach for Juvenile Justice in Honolulu” in Federal Probation 66 (2002), pp38-43

Willcock, R Retail Theft Initiative: Does It Really Work? (K2 Management Development, 1999)

Wilcox, A, Young, R and Hoyle, C Two-Year Resanctioning Study: A Comparison of Restorative & Traditional Cautions, Home Office online report 57/04 (London: Home Office, 2004)

Woehrle, LM Summary Evaluation Report: A Study of the Impact of the Help Increase the Peace Project in the Chambersburg Area School District (Baltimore, Maryland: American Friends Service Committee, 2000)

Wynne, J “Leeds Mediation and Reparation Service: Ten Years’ Experience with Victim-Offender Mediation” in Galaway, B and Hudson, J (eds) Restorative Justice: International Perspectives (Monsey, New York: Criminal Justice Press, 1996)

Youth Justice Board for England & Wales National Evaluation of the Restorative Justice in Schools Programme (2004) (www.youth-justice-

Weisburd, D, Lum, C and Petrosino, A “Does Research Design Affect Study Outcomes in Criminal Justice?” in The Annals of the American Academy of Social & Political Sciences 578 (2001), pp50-70

Woolf, H “Foreword” in Radzinowicz, L Adventures in Criminology (London: Routledge, 1999)