Poulson (2003) - A third voice - A review of empirical research on the psychological outcomes of restorative justice

*Department of Behavioral Science, Utah Valley State College. This Article was preparedwhile I was in the Department of Psychology, Brigham Young University. E-mail:[email protected].

1MARK S. UMBREIT, VICTIM MEETS OFFENDER 2–5 (1994); HOWARD ZEHR, CHANGINGLENSES: A NEW FOCUS FOR CRIME AND JUSTICE 181 (1990).

2UMBREIT, supra note 1, at 5, 9, 25.

167

A Third Voice: A Review of Empirical Research on thePsychological Outcomes of Restorative Justice

Barton Poulson*

Within the field of restorative justice, at least three voices may be heard: thevoice of theory, the voice of practice, and the voice of research, which is the focusof this Article. Data from seven evaluation studies of restorative justice and courtprograms from around the world are combined on twelve psychologicaloutcomes, such as judgments of fairness, accountability, increased respect, andreductions in fear. Despite substantial differences in the implementation of therestorative justice programs and the persons and cases included, the data wereremarkably consistent: restorative justice outperformed court procedures onalmost every variable for victims and offenders. In addition, connections betweenthe benefits of restorative justice and recent psychiatric research on youth suicideare sketched, with the possible benefits of restorative justice (as compared tocourt) in saving lives. Overall, restorative justice emerged as a compelling andeffective alternative to court in all measured outcomes.

I. INTRODUCTION

A single musical voice can carry a musical tune. With severalcomplementary voices, however, harmony and counterpoint can add to therichness and depth, creating a qualitatively distinct experience. A similar,enriching complementarity can also be found in restorative justice, where thevoices of theory, practice, and research can contrast and support one another,leading to a broader and more useful understanding of restorative justice.

Restorative justice, briefly defined, is an approach to postcrime reparationthat focuses on healing the harm done, promoting accountability and personalresponsibility, and encouraging the active participation of the victim, offender,and other concerned parties. Restorative justice fundamentally views crime as aviolation against a person rather than against an abstract state or law.1 It has beenpracticed for centuries in various forms of indigenous justice and has morerecently become formally incorporated in practices such as victim-offendermediation, family group conferencing, sentencing circles, community reparationboards, and other postcrime forms.2 In its essence, restorative justice involves the

UTAH LAW REVIEW [2003: 167168

3See generally PAUL MCCOLD, RESTORATIVE JUSTICE: AN ANNOTATED BIBLIOGRAPHY (1997)(providing comprehensive overview of literature on restorative justice); RESTORATIVE COMMUNITYJUSTICE: REPAIRING HARM AND TRANSFORMING COMMUNITIES (Gordon Bazemore & Mara Schiffeds., 2001) (providing collection of articles that advocate use of restorative justice principles);RESTORATIVE JUSTICE: INTERNATIONAL PERSPECTIVES (Burt Galaway & Joe Hudson eds., 1996)(providing collection of essays on restorative justice); THE SPIRITUAL ROOTS OF RESTORATIVEJUSTICE (Michael L. Hadley ed., 2001) (providing collection of essays discussing various religiousviews of restorative justice); UMBREIT, supra note 1, at 15 (evaluating four restorative justicestudies); John Braithwaite, Restorative Justice: Assessing Optimistic and Pessimistic Accounts, 25CRIME & JUST. 1 passim (1999) (synthesizing arguments for and against restorative justice); JohnR. Gehm, Victim-Offender Mediation Programs: An Exploration of Practice and TheoreticalFrameworks, 1 W. CRIMINOLOGY REV. 1 passim (1998) (providing survey of both theoretical andpractical applications of restorative justice), available at http://wcr.sonoma.edu/v1n1/gehm.html.

victim of a crime and the offender (in victim-offender mediation) and possiblyother family members or friends (in family group conferencing) in a negotiationthat is facilitated by a third-party neutral (a mediator). The negotiation is designedto explore the subjective and objective impact of the crime and, where possible,to restore the victim to his or her precrime circumstances, although this is anadmittedly idealistic goal. More frequently, the practical goal of restorative justiceis to provide some material benefit to the victim and, perhaps more importantly,to restore some peace of mind and well-being in his life, as well as to provideincreased accountability for and responsiveness to the needs of the offender.

Because of its unique responsiveness to the idiosyncratic situations andhumanity of victims and offenders, restorative justice has found advocates indiverse domains. These are the sources of the harmonizing, contrapuntal voicesin the grand oeuvre of restorative justice. The first voice is that of theory, whichcan be found among ethicists, legal philosophers, criminal justice experts, andpolicy makers. The practice of restorative justice, the second voice, can be foundamong lawyers, mediators, juvenile court judges and administrators, andcommunity activists. These first two voices are excellently represented by thearticles in this volume and elsewhere.3 The third voice in the opus is that ofresearch, which is found among experimentalists, evaluators, and policyresearchers. It is also the voice of this Article. Taken together, the three voicesgive a richer understanding of the promises and challenges of restorative justice.

The purpose of this Article is to briefly review and synthesize the findingsof the research literature on restorative justice and court. Although this field isrelatively new to the research world, the amount of research is expanding. Thisreview will collect empirical data from several previously published studiesregarding outcomes of restorative justice and court based practices. It will thenpresent the combined data in a more accessible format, draw connections betweenrestorative justice and other areas of psychological interest, and emphasize thepossibilities that may be addressed by the three voices of restorativejustice—theory, practice, and research.

No. 1] A REVIEW OF EMPIRICAL RESEARCH 169

4Specifically, such studies address judgments of fairness and fear, but not recidivism. SeeWilliam R. Nugent et al., Participation in Victim-Offender Mediation and the Prevalence andSeverity of Subsequent Delinquent Behavior: A Meta-Analysis, 2003 UTAH L. REV. 137.

5See infra Part II.B.

II. METHOD

This Article reviews and synthesizes data from several published evaluationsof restorative justice and court based programs from around the world. The firstsection of this part presents the criteria for selecting studies to be included in thisreview. The second section describes in detail the studies included. The thirdsection lists the outcomes derived from these studies for inclusion in this review.The final methodological section describes the method for extracting andcombining data from these studies.

A. Selection Criteria

In choosing studies to review for this Article, three major criteria were usedto assess over one hundred books, articles, and chapters for potential inclusion.First, the studies needed to address the psychological outcomes of restorativejustice.4 The list of outcomes was to be determined in part by the data included inthe studies selected. Consequently, a complete list of outcomes could not bedetermined until the appropriate studies were selected for inclusion.5 Second,these studies needed to contain quantitative data, as this permits the statisticalsynthesis of findings. Third and finally, the studies needed to contain data frompersons who participated in restorative justice procedures and from persons whoparticipated in court. This last restriction permitted the direct comparison of theefficacy of restorative justice compared to adjudication.

B. Studies Included

Seven studies out of more than one hundred met the three major criteria. Thestudies were published between 1980 and 1999. They included four evaluationsof victim-offender mediation programs and three family group conferencingprograms. The programs were based in the United States, Canada, England, andAustralia. Across the seven studies, data were gathered from a total of 4602respondents (1297 victims in restorative justice, 1189 victims in court, 1077offenders in restorative justice, and 1039 offenders in court).

Methodologies for assigning people to restorative justice or court variedbetween the studies. In most research situations, the ideal method is randomassignment, in which a random process (e.g., flipping a coin or consulting a tableof random numbers) is used to assign cases to either restorative justice or court.Random assignment is generally considered the best way to create groups that are


6Again, those who opted out of restorative justice were grouped with the court/non-restorativejustice group for purposes of this review.

7See, e.g., HEATHER STRANG ET AL., EXPERIMENTS IN RESTORATIVE POLICING: A PROGRESSREPORT ON THE CANBERRA REINTEGRATIVE SHAMING EXPERIMENTS (RISE) 7–28 (1999) (providingrich data relating to RISE), available at http://www.aic.gov.au/rjustice/rise/progress/1999.html.

8R.C. DAVIS ET AL., MEDIATION AND ARBITRATION AS ALTERNATIVES TO PROSECUTION INFELONY ARREST CASES: AN EVALUATION OF THE BROOKLYN DISPUTE RESOLUTION CENTER (FIRSTYEAR) (1980).

9Unless otherwise noted, all facts in Part II.B.1 are taken from DAVIS ET AL., supra note 8, at14–34.

reasonably equivalent on all known variables (e.g., age of offender, nature ofcrime, offender’s criminal history, nature of relationship between victim andoffender) as well as unknown or unmeasured variables (e.g., psychologicalfunctioning of offender and victim, history of abuse, level of moral reasoning).Two of the seven studies (including the largest study) used random assignmentto create equivalent groups for restorative justice and court, while another studyused randomization with an additional condition for those who opted out ofrestorative justice (these participants were grouped with the court/nonrestorativejustice group for purposes of this review).

In situations where randomization is not feasible because of small numbersof participants, a decision to conduct the study was made after people had alreadybeen assigned to conditions, or because randomization was not permitted by theagencies involved, matching is a viable alternative. In matching, a comparisongroup is deliberately compiled that resembles the experimental group on allknown, relevant variables. There is no known probability, however, that the twogroups will be equivalent on other dimensions. One study utilized matchedassignment with the possibility of opting out of restorative justice.6

Finally, for the remaining three studies (including the two smallest) theresearchers did not attempt to create equivalent groups. In these studies, all of theparticipants had been originally assigned to restorative justice and comparisonswere made between those who accepted or rejected this assignment.

As a final methodological note, many of these studies7 had unusually richdata sets. This review does not reflect the entirety of the available data from anyone of the studies. Instead, this review focuses on selected questions that weretypically addressed in more than one study and that addressed psychologicaloutcomes of interest. Readers are encouraged to see each original research paperfor more detail and more outcomes.

1. The Brooklyn Study8

The earliest study selected for this review is an evaluation of a victim-offender mediation center in Brooklyn, New York.9 Like many mediation centers,this one was staffed by community volunteers who received approximately fifty


hours of training in victim-offender mediation. What was unusual was that themajority of cases referred were felony arrests, excluding cases involvingprostitutes or pimps, minors living with parents who were the other party,incompetent parties, cross-complaints (i.e., all parties were defendants), firstdegree assault or higher, hospitalized minors, critically injured adults, recoveredguns, gun wounds, forcible rape, arson or attempted arson, misdemeanors, priormediation, or assault on a police officer. Consequently, felony arrests were theexclusive focus of the evaluation.

The study covered a period of four months in the first year of the mediationprogram, from September 1 to December 23, 1977. The Brooklyn Criminal Courtreferred approximately 550 cases to the mediation center each week. Only 8 to10% of referred cases were approved for mediation by all parties (the two mostrestrictive criteria were a prior relationship between the parties and a civilcomplaint). The total of 465 cases that were approved for mediation during thefour-month evaluation period were then randomly assigned to either mediation (N= 259) or prosecution (N = 206). A large percentage of cases assigned to theseconditions, however, were not actually mediated or prosecuted. In 12% of themediation cases the victim declined mediation, and in another 32% of the casesthe victim and/or the offender was absent. In 41% of the cases assigned toprosecution, the defendant had not been asked to come to court. In another 17%the defendant did not appear when requested (consequently, the cases went todisposition without them). Nevertheless, these cases were kept in the study inorder to retain the random assignment of cases to mediation or prosecution, aswell as to accurately reflect the common experiences of victims in bothconditions.

The majority of victims (63%) were female and relatively young (median ageof twenty-nine years, with 7% eighteen years or younger). Most victims were alsopoor and slightly less than half (47%) had finished high school. Demographicdetails were not provided for offenders.

Qualitative and quantitative data were gathered from in-person and telephoneinterviews conducted at three separate times for victims (at time ofcourt/mediation, after mediation/disposition, and four months later) and at onetime (four months after mediation/disposition) for offenders. The interviewsfocused on descriptions of the incident, expectations for mediation/adjudication,and the desired role for the participant in the proceedings.


10PAUL MCCOLD & BENJAMIN WACHTEL, U.S. DEP’T OF JUSTICE, RESTORATIVE POLICINGEXPERIMENT: THE BETHLEHEM PENNSYLVANIA FAMILY GROUP CONFERENCING PROJECT (1998).

11Unless otherwise noted, all facts in Part II.B.2 are taken from MCCOLD & WACHTEL, supranote 10, at 7.

12David B. Moore & John M. McDonald, Achieving the “Good Community”: A Local PoliceInitiative and Its Wider Ramifications, in PERCEPTIONS OF JUSTICE: ISSUES IN INDIGENOUS ANDCOMMUNITY EMPOWERMENT 143, 145 (Kayleen M. Hazlehurst ed., 1995).

13DAVIS ET AL., supra note 8.14UMBREIT, supra note 1, at 209–44.

2. The Bethlehem Study10

The second study evaluated a family group conferencing program inBethlehem, Pennsylvania.11 This program was based on programs implementedin Australia, and used a substantially revised version of the “Wagga Wagga”model,12 which is named for the city where it was developed. Like the WaggaWagga model, the program in Bethlehem was staffed by police officers (eighteenin this case) who had received a three-day training conducted by the developersof the family group conferencing program in Australia. Unlike the Brooklynproject evaluated in Part II.B.1,13 the Bethlehem project only included juvenilefirst-time offenders (i.e., those who had not been through the juvenile probationsystem), and it excluded felony offenders. It also excluded drug and alcoholcrimes, sex offenses, and certain simple assaults.

The majority of offenders were male (between 53% and 71% for the threegroups). The largest ethnic group for offenders was Latino (49 to 51%), followedby Caucasian (35 to 44%), and African-American (1 to 14%). The most commonage of offenders was between fourteen and fifteen (31 to 34%), with about one-quarter (23 to 25%) under the age of thirteen. Demographic data were notreported for victims. The evaluation covered a period from November 1, 1995 toMay 1, 1997. A total of 140 property crime cases and seventy-five violent crimecases were included. Two-thirds of each kind of crime were randomly assignedto family group conference, and one-third of the cases of each kind were assignedto prosecution. However, participants in 50% of the property crime cases and 68%of the violent crime cases declined family group conferencing. Consequently,these cases were prosecuted and were included in the study as a third group ofthose who declined conferencing (although for analytical purposes in this study,they were included with the prosecution cases as the two groups did notexperience restorative justice principles).

Data were gathered using a survey developed by Mark Umbreit.14 Of the 233offenders who were mailed a survey, 156 (67%) responded. Of the 180 victimswho were mailed a survey, approximately 121 (67%) responded. Data were alsogathered from parents of the participants and from police officers, although thesedata were not included in this analysis due to lack of suitable comparisons fromother studies.


15STRANG ET AL., supra note 7.16Unless otherwise noted, all facts in Part II.B.3 are taken from STRANG ET AL., supra note 7,

at 1–2, and the earlier version of the study, LAWRENCE W. SHERMAN ET AL., EXPERIMENTS INRESTORATIVE POLICING 4–5 (1997), available athttp://www.aic.gov.au/rjustice/rise/progress/1997.html.

17UMBREIT, supra note 1; see also Mark S. Umbreit & Robert B. Coates, Cross-site Analysisof Victim-Offender Mediation in Four States, 39 CRIME & DELINQ. 565, 570–79 (1993) (analyzingadditional data from this evaluation).

18Unless otherwise noted, all facts in Part II.B.4 are taken from UMBREIT, supra note 1, at29–30, 35–40, 56 tbls.6 & 7.

3. The Canberra Study15

The most ambitious of the seven studies included here is this large-scale (N= 2353), multiyear (with working papers published from 1997 to 2000)randomized experiment of a family group conferencing program run by police inCanberra, Australia.16

Four classes of crimes were included in this study: (a) drunk driving (N =1683), which could involve an offender of any age with blood alcohol contentover .08 but no accidents or damage associated with the arrest; (b) juvenile (underthe age of eighteen) property offending with personal victims (N = 295); (c)juvenile shoplifting detected by shop security (N = 212); and (d) youth (under theage of thirty) violent offenses (N = 163). Data from the four offense categorieswere combined for the current analysis.

Cases were referred to family group conferencing by the police, who hadreceived special training in screening cases. Cases were then randomly assignedto conferencing or court. Assignment also remained more consistent than in mostother studies, with 83 to 99% of the cases processed as assigned, varyingaccording to the category of crime. Offenders were primarily male (52 to 98%),with average ages of fifteen to seventeen for juvenile crimes and thirty to thirty-two for drunk driving, and 76 to 96% were born in Australia (0 to 14% wereAboriginal or Torres Strait Islanders). Demographic data were not reported forvictims.

A particularly rich set of data regarding offender outcomes was gatheredfrom questionnaires and observations of conferences. A smaller amount of datawas collected from victims and from police facilitators via questionnaires. Thesequalitative data were not included in the present quantitative review.

4. The United States Study17

In a summary and comparison of several victim-offender mediation programsfor juvenile offenders in the United States, Umbreit presented data from 1153interviews conducted in Albuquerque, New Mexico; Minneapolis, Minnesota;Oakland, California; and Austin, Texas.18 The programs in Albuquerque,


19MCCOLD & WACHTEL, supra note 10, at 51.20See infra Parts II.B.5 through II.B.7.21MARK S. UMBREIT, CTR. FOR RESTORATIVE JUSTICE & MEDIATION, MEDIATION OF

CRIMINAL CONFLICT: AN ASSESSMENT OF PROGRAMS IN FOUR CANADIAN PROVINCES (1995),available at http://ssw.che.umn.edu/rjp/Resources/Resource.htm.

22Unless otherwise noted, all facts in Part II.B.5 are taken from UMBREIT, supra note 21, at1.

Minneapolis, and Oakland were private nonprofit community groups that operatedin conjunction with the juvenile courts. The program in Austin was administeredby the juvenile court.

From 1990 to 2001, a total of 2799 victims and 2659 offenders were referredfor restorative justice by local juvenile courts and probation staffs, along with asmaller number by prosecuting attorneys and police. The majority of referrals(87%) involved property crimes (typically burglary or vandalism), whereas theremaining 13% involved violent crimes (typically minor assaults). The averageage for offenders was fifteen (ranging from seven to eighteen). The majority ofoffenders (86%) were male and Caucasian (54%), although ethnicity varied fromone city to another. The great majority of offenders had no prior criminalconvictions.

Cases that were referred to victim-offender mediation were compared withmatched cases drawn from adjudicated cases. A third group consisted of cases thatwere referred to mediation but chose instead to go to court. For purposes of thisanalysis, these cases were combined with matched group of court participants.

Data were gathered using an interview format that included forced-choicequestions as well as open-ended statements. The same interview schedule wasused in the Bethlehem study19 and the three studies below.20 Interviews wereconducted over the telephone approximately one week and again two months afterthe mediation.

5. The Canada Study21

In an evaluation that strongly paralleled the above analysis of programs inthe United States, Umbreit also studied Canadian victim-offender mediationprograms in Langley, British Columbia; Calgary, Alberta; Winnipeg, Manitoba;and Ottawa, Ontario.22 All four programs were administered by private, nonprofitgroups. The programs were oriented primarily towards juvenile offenders,although two of the four in this study also accepted adult offenders. Themediations were conducted by volunteers who had received thirty to forty hoursof training in addition to supervised apprenticeships or observations.

During the two-year study period (1991 to 1993), 4445 cases were referredto mediation, of which 1736 were mediated. Referrals came primarily fromprobation officers, judges, and the Crown Attorney’s Office. The most common


23UMBREIT, supra note 1, at 211.24MARK S. UMBREIT & CLAUDIA FERCELLO, CTR. FOR RESTORATIVE JUSTICE & MEDIATION,

INTERIM REPORT: CLIENT EVALUATION OF THE VICTIM/OFFENDER CONFERENCING PROGRAM INWASHINGTON COUNTY (MN) (1997), available athttp://ssw.che.umn.edu/rjp/Resources/Resource.htm.

25Unless otherwise noted, all facts in Part II.B.6. are taken from UMBREIT & FERCELLO, supranote 24, at 2.

26UMBREIT, supra note 21, at 4–6.27UMBREIT, supra note 1, at 211 (explaining format of interviews).

offenses included assault, mischief, and breaking and entering. A total of 610(including victims and offenders) interviews were conducted. Of the 323 victimsinterviewed, 59% were male, 86% were white, and the average age was thirty-three. Of the 287 offenders interviewed, 80% were male, 80% were white, and theaverage age was twenty-four. All cases were originally assigned to restorativejustice. Participants in victim-offender mediation were compared with those whoopted to go to court.

Telephone interviews, using the form described above,23 were conductedwith participants two months after the mediation session or the time that the casewas disposed of in court. Additional qualitative data came from interviews withcourt officials, program staff and volunteer mediators, records data, and casestudy qualitative research techniques. These data, however, are not included in thecurrent review due to lack of parallel data from other studies.

6. The Minnesota Study24

Mark Umbreit and Claudia Fercello conducted a focused evaluation of a newfamily group conferencing program in Washington County, Minnesota, betweenMarch and May 1996.25 Cases were referred for conferencing by judges, countyattorneys, social services, law enforcement, and other government agencies. Themost common offenses were damage to property, assault, disorderly conduct, andtheft.

A total of fifty-eight conference participants were interviewed. The thirty-nine offenders (twenty-six in family group conferencing and thirteen in court) hadan average age of 15.8, although any age could participate. The majority of of-fenders (69%) were white, and slightly over half (54%) were male. Nineteenvictims were also interviewed (twelve in conferences and seven in court), but nodemographic information was provided for this group.

As with Umbreit’s assessment of Canadian programs,26 all fifty-eightparticipants were originally assigned to restorative justice. Participants in victim-offender mediation were compared with those who opted to go to court. Finally,data were gathered via telephone interviews using a modified version ofUmbreit’s questionnaire.27


28MARK S. UMBREIT & ANN WARNER ROBERTS, CTR. FOR RESTORATIVE JUSTICE &MEDIATION, MEDIATION OF CRIMINAL CONFLICT IN ENGLAND: AN ASSESSMENT OF SERVICES INCOVENTRY AND LEEDS (1996), available at http://ssw.che.umn.edu/rjp/Resources/Resource.htm.

29Unless otherwise noted, all facts in Part II.B.7. are taken from UMBREIT & ROBERTS, supranote 28, at 5–12.

30UMBREIT, supra note 1, at 211.

7. The England Study28

The last study included in this review is an evaluation of a victim-offendermediation program in Coventry and Leeds, England.29 These programs differedin one significant way from the other programs: although some cases wereconducted directly (i.e., face-to-face), the majority of mediations (84%) wereindirect, similar to what is sometimes called “conciliation” or “shuttlediplomacy.” Cases were referred by juvenile courts, magistrates courts, and theCrown Court. Cases were mediated by paid mediators (sessionally paid mediatorsin Leeds and full time in Coventry).

In the study period of 1991 through 1993, a total of 123 cases were mediated.Most cases involved property offenses (e.g., burglary and theft) and minor assault,although some more serious cases were also referred. The majority of victims(69%) were male, with an average age of forty. Offenders were almost exclusively(94%) male, with an average age of twenty-four. Cases were assigned into threegroups: direct mediation (nineteen victims and sixteen offenders), indirectmediation (twenty-five victims and fourteen offenders), and those who werereferred to mediation but chose to go to court instead (twenty-six victims andtwenty-three offenders). The first two groups were combined into a single“mediation” group for purposes of this analysis. No attempt was made at random-assignment or matching due to time and complexity involved, as well as smallsample sizes at some of the sites.

Telephone and in-person interviews were conducted after mediation ordisposal of the case using Umbreit’s questionnaire,30 which was adapted for usein Great Britain.

C. Outcomes Included

After selecting the seven studies for inclusion, the studies were thenreviewed for potential outcome variables to be included in this review. The goalwas not to use all possible outcome variables but instead to use outcomes that: (a)plainly reflected psychological processes (e.g., affect, behavior, and cognition);(b) would be of interest to restorative justice stakeholders; and (c) appeared,ideally, in multiple studies. The selected outcome variables were:


31The text accompanying each of the attached tables provides examples of this “coding.”

The perception that the criminal justice system was fair; Satisfaction with thehandling of case; Whether the participant felt that he or she had an opportunity totell his or her story; The perception that his or her opinion was adequatelyconsidered; The perception that the judge or mediator was fair; The belief that theoffender was held accountable; Whether an apology was offered by the offenderfor forgiveness was expressed by the victim; The participant’s perception that theoutcome was fair; The participant’s satisfaction with the outcome; Whether theparticipant’s of other party’s behavior improved; Whether the victim was stillupset about crime; Whether the victim was afraid of revictimization.

D. Data Extraction

All seven of the studies reported outcomes as the percentage of participantsresponding positively to each question. This greatly facilitated the synthesis ofdata. The number of victims in restorative justice from each study who agreedwith an item, for example, was simply added for an overall frequency, which wascompared against the total number of victims in restorative justice to yield apercentage for agreement. At times it became necessary to exercise judgment indetermining whether questions were sufficiently similar and whether responseswere coded in comparable fashions.31 Unfortunately, none of the studies presentedthe data in a way that permitted subanalyses by participant characteristics (e.g.,gender, age, and race/ethnicity) or case characteristics (e.g., category of crime).Consequently, all analyses are limited to the participant’s status as a victim oroffender and whether restorative justice or standard adjudication was used.

E. Results

Overall, restorative justice practices substantially outperformed court onalmost every item for both victims and offenders. The twelve figures present thedata for each of the variables in aggregated and disaggregated form. In the charts,the vertical bars indicate the percentage of victims and offenders in restorativejustice and court who agreed with each item. The narrow black lines at the top ofthe gray bars are 95% confidence intervals. These indicate the likely range ofvariation for the percentage in the larger population. As a simple heuristic, thepercentages for two groups may generally be considered reliably different fromeach other if their confidence interval lines do not overlap. The tables below thecharts provide the percentage of agreement, broken down by the studies thatprovided the data and by status as a victim or offender. The percentages of


32The odds-ratio is a measure of effect size (i.e., the difference between two groups, regardlessof sample size). It is calculated in the following manner: Take, for example, the number of victimsin restorative justice who agreed with the statement and divide it by the number of victims inrestorative justice who disagreed with the statement (i.e., the odds of agreement for victims inrestorative justice). Then divide this number by the similarly calculated odds of agreement forvictims in court. The ratio of these two numbers (i.e., the ratio of the odds, or the odds-ratio) is thenused to index the extent to which the two odds differ. An odds-ratio of 1 indicates no difference(i.e., identical odds), whereas an odds-ratio greater than 1 indicates greater odds of agreement forthe restorative justice group. An odds-ratio of less than 1 (but odds-ratios are always greater than0) indicates greater agreement for the court group. It should be noted that the reciprocal of an odds-ratio (1/odds-ratio) indicates a relationship of the same strength but in the opposite direction (i.e.,the other group has greater agreement).

33Chi-square (P2) is an inferential test that indicates whether a given effect is likely due tochance. A probability level less than .05 indicates that there is less than a 5% chance that thedifferences in agreement between the restorative justice and court groups emerged through randomchance. Such an unlikely difference is conventionally referred to as “statistically significant.”

34The effect did not reach conventional levels of statistical significance for the smallest of thestudies, The England Study, see UMBREIT & ROBERTS, supra note 28, but the direction of the effectwas consistent with the other studies (i.e., participants in restorative justice showed higher levelsof agreement than participants in court), implying that this non-significant result is more likely dueto a small sample size (fifty-four victims) than to a small effect.

participants in restorative justice and court that agreed with the statement arecompared using odds-ratios32 and the chi-square statistic.33

Table 1 and Figure 1 address the participants’ belief that they had beentreated fairly by the criminal justice system. It shows that both victims andoffenders in restorative justice were significantly and substantially more likely(3.4 times more likely for victims and two times more likely for offenders) tobelieve that the criminal justice system was fair than were victims or offenders incourt. This difference was consistent across all five studies that included thisoutcome.34


Table 1. Percentage of Respondents Who Believed That the Criminal JusticeSystem Was Fair

RestorativeJustice Court Odds

Source Agree N Agree N Ratio P2 p

McCold &Wachtel (1998)a

96% 54 80% 67 6.02 6.86 .009

Umbreit (1994)b 83% 204 58% 193 3.60 30.92 <.000

Umbreit (1995)b 80% 174 43% 124 5.30 43.35 <.000

Umbreit &Roberts (1996)b

59% 34 50% 20 1.42 0.38 .537

VictimsCombined

82% 466 56% 404 3.43 65.20 <.000

Offenders

McCold &Wachtel (1998)c

97% 67 87% 88 4.72 4.62 .032

Strang et al.(1999)d

94% 508 83% 455 3.01 27.27 <.000

Umbreit (1994)b 89% 178 80% 202 2.05 5.97 .015

Umbreit (1995)b 80% 156 56% 121 3.14 18.52 <.000

Umbreit &Roberts (1996)b

89% 27 56% 23 6.29 6.94 .008

OffendersCombined

91% 936 78% 889 2.66 52.97 <.000

a “Did you experience fairness with the way your case was handled?” b “Did youexperience fairness within the justice systems in your case?” c Did you experiencefairness within the justice system in your case?” d “Treatment was fair overall?”


C r im in a l J u s t ic e S y s te m W a s F a ir

R e s to ra t iv eJ u s t ic e(N = 9 3 6 )

R e s to ra t iv eJ u s tic e ( N = 4 6 6 )

C o u r t (N = 8 8 9 )

C o u rt(N = 4 0 4 )

0 %

2 5 %

5 0 %

7 5 %

1 0 0 %

V ic t im s O ffe n d e rs

Figure 1. Percentage of Respondents Who Believed That the Criminal JusticeSystem Was Fair (With 95% Confidence Intervals).

Participants in restorative justice were also significantly more likely to besatisfied with the way that their case was handled than were participants in court(2.8 times more likely for victims, and 1.9 times more likely for offenders; seeTable 2 and Figure 2). This pattern was consistent for all six studies, although theeffects were not statistically significant for the smaller studies. This, too, wasmore likely due to small sample sizes and the attendant low statistical power thanto negligible effects.


Table 2. Percentage of Respondents Who Were Satisfied With the Handling ofTheir Case




96% 54 76% 67 7.52 9.26 .002

Strang et al.(1999)b

68% 77 48% 75 2.29 6.18 .013

Umbreit (1994)c 79% 204 57% 199 2.84 22.46 <.000

Umbreit (1995)c 78% 178 48% 134 3.84 30.29 <.000

Umbreit &Fercello (1997)c

100% 12 88% 7 - 1.51 .220

Umbreit &Roberts (1996)c

62% 42 58% 26 1.18 0.11 .745

VictimsCombined

78% 567 56% 508 2.75 57.64 <.000

Offenders


97% 67 92% 88 2.85 1.78 .182

Umbreit (1994)c 87% 181 79% 205 1.79 4.38 .036

Umbreit (1995)c 74% 157 53% 126 2.52 13.49 <.000

Umbreit &Fercello (1997)c

88% 26 62% 13 4.49 3.57 .059

Umbreit &Roberts (1996)c

79% 29 55% 22 3.13 3.43 .064

OffendersCombined

84% 460 73% 454 1.92 16.14 <.000


S a tis f ie d w ith H a n d lin g o f C a s e



C o u r t(N = 5 0 8 )

C o u r t (N = 4 5 4 )

0 %

2 5 %

5 0 %

7 5 %

1 0 0 %


a “How satisfied were you with the way the justice system handled your case?” b“Satisfied with the way case was dealt with?” c “How satisfied were you with theway the justice system handled your case?”

Figure 2. Percentage of Respondents Who Were Satisfied With the Handling ofTheir Case (With 95% Confidence Intervals)

Participants in restorative justice were more likely to believe that they hadbeen able to tell their story during the proceedings than were court participants(8.8 times more likely for victims, and 4.1 times more likely for offenders; seeTable 3 and Figure 3). The generalizability of this finding, however, is limitedbecause it is derived from only one study for victims and two studies foroffenders, although the effect is consistent in both studies.


Table 3. Percentage of Respondents Who Believed That They Had an Opportunityto Tell Their Story



Victims

Davis et al.(1980)a

94% 112 64% 55 8.81 24.85 <.000

Offenders

Davis et al.(1980)b

90% 62 44% 43 11.45 26.18 <.000

Strang et al.(1999)c

88% 508 66% 455 3.70 65.82 <.000

OffendersCombined

88% 570 64% 498 4.11 86.14 <.000

a “Had an opportunity to tell story? b “Story was heard by judge or mediator?” c“Had an opportunity to express my views?”


35STRANG ET AL., supra note 7, at 5.

H a d O p p o r tu n ity to T e ll S to ry



C o u r t(N = 5 5 )

C o u r t (N = 4 9 8 )

0 %

2 5 %

5 0 %

7 5 %

1 0 0 %


Figure 3. Percentage of Respondents Who Believed That They had anOpportunity to Tell Their Story (With 95% Confidence Intervals).

Victims in restorative justice were not significantly more likely to believethat their opinions were adequately considered (see Table 4 and Figure 4),although this non-effect is undoubtedly due to the high levels of agreement thatwere evident for victims in both court (92%) and restorative justice (94%). Itsimply was not possible for restorative justice to do much better than court,although the odds-ratio was still favorable at 1.28. For offenders, however, therewas no ceiling effect, and restorative justice participants were significantly morelikely to believe that their opinions were adequately considered than wereoffenders in court, with an odds-ratio of 2.1. As with the last outcome, however,generalization is greatly limited by the fact that data came from only one studyeach for victims and offenders. As a compensatory note, the data for the offenderscame from a particularly strong and varied experiment,35 which moderates thislimitation.


O p in io n W a s A d e q u a te ly C o n s id e re d

R e s to ra t iv e J u s tic e ( N = 5 4 )


C o u r t(N = 6 7 )

C o u r t(N = 4 7 0 )

0 %

2 5 %

5 0 %

7 5 %

1 0 0 %


Table 4. Percentage of Respondents Who Believed That Their Opinion WasAdequately Considered



Victims


94 54 92 67 1.28 0.11 .735

Offenders


72 508 55 470 2.12 31.12 <.000

a “Do you feel your opinion regarding the offense and circumstances wereadequately considered in this case?” b “Treatment took account of what I said.”

Figure 4. Percentage of Respondents Who Believed That Their Opinion WasAdequately Considered (With 95% Confidence Intervals).

Restorative justice participants were also more likely to say that the mediatorhad been fair than participants in court were to say that about the judge (2.3 and6.0 times more likely for victims and offender, respectively; see Table 5 andFigure 5). Data from only two studies went into this result, and the effect was


small for victims and nonsignificant for offenders in the smaller of the twostudies, although the effect was always in the same direction.

Table 5. Percentage of Respondents Who Believed That the Judge or MediatorWas Fair



Victims

Davis et al.(1980)a

88% 112 76% 55 2.32 3.96 .047

Offenders

Davis et al.(1980)a

89% 62 86% 43 1.32 0.21 .645


91% 508 60% 446 6.87 127.03 <.000

OffendersCombined

91% 570 63% 489 6.05 123.01 <.000

a “Judge or mediator fair?” b “Police were fair at treatment?”


J u d g e /M e d ia to r W a s F a ir



C o u r t(N = 5 5 )

C o u r t (N = 4 8 9 )

0 %

2 5 %

5 0 %

7 5 %

1 0 0 %


Figure 5. Percentage of Respondents Who Believed That the Judge or MediatorWas Fair (With 95% Confidence Intervals).

Results were similar for the belief that the offender had been heldaccountable (see Table 6 and Figure 6), with participants in restorative justicebeing significantly more likely to respond in the affirmative (4.9 times more likelyfor victims and 4.8 times more likely for offenders). It is particularly importantto note that only 49% of the offenders in court felt that they had been heldaccountable for their actions. Thus, an offender’s chances of having to answer forhis or her actions were the same as a coin flip. Fortunately, the level of agreementwent up by thirty-three percentage points in restorative justice, which was adramatic improvement.


Table 6. Percentage of Respondents Who Believed That the Offender Was HeldAccountable



Victims

McCold &Wachtel(1998)a

93% 54 75% 67 4.33 6.62 .010

Umbreit &Fercello(1997)b

90% 12 33% 7 18.27 6.72 .010

VictimsCombined

92% 66 71% 74 4.90 10.17 .001

Offenders

McCold &Wachtel(1998)c

91% 67 87% 88 1.52 0.63 .426

Strang et al.(1999)d

81% 508 42% 454 6.00 158.43 <.000

OffendersCombined

82% 575 49% 542 4.82 137.22 <.000

a “Was the offender adequately held accountable for the offense committed?” b

“Was offender adequately held accountable?” c “Were you adequately heldaccountable for the offense you committed?” d “Was offender adequately heldaccountable?”


36As a caveat, in STRANG ET AL., supra note 7, at 67, drunk driving cases were excluded fromthe data (a total of 378 cases) because these cases did not involve any personal victims to whom theoffender could apologize (i.e., none of the drunk driving arrests involved accidents or injuries). Ifthese cases are included, offenders in restorative justice are still more likely to apologize (40% of716 offenders) than are offenders in court (20% of 681), although the effect is not as pronounced(odds-ratio = 2.65, P2(1, N = 1397) = 65.11, p < .001).

37As with apologies, cases of victimless drunk driving were excluded from the Canberra studyassessment of forgiveness. See STRANG ET AL., supra note 7, at 67. Were these 901 additional casesto be included, as before, the effect would still be significant. In fact, it would be more pronounced,with 43% of 647 victims in restorative justice offering forgiveness, as opposed to 9% of 636 victimsin court (odds-ratio = 7.30 P2(1, N = 1283) = 186.91, p < .001).

O ffe n d e r W a s H e ld A c c o u n ta b le

R e s to ra t iv eJ u s tic e ( N = 6 6 )


C o u r t(N = 7 4 )

C o u r t (N = 5 4 2 )

0 %

2 5 %

5 0 %

7 5 %

1 0 0 %


Figure 6. Percentage of Respondents Who Believed That the Offender Was HeldAccountable (With 95% Confidence Intervals).

One of the major advantages of restorative justice practices is that offendersmay spontaneously apologize to the victims for the impact of their actions. Thisis an outcome that can have particularly important psychological benefits to thevictims. It is also one of the strongest effects in this review: offenders were 6.9times more likely to apologize to the victim in restorative justice settings than incourt. To put this difference into perspective, very nearly three out of four (74%)offenders in restorative justice apologized, whereas almost three out of four (71%)offenders in court did not apologize. This effect was consistently large across thefour studies that tracked apologies.36 Just as significant, Table 7 and Figure 7 alsoshow that victims in restorative justice were 2.6 times more likely to forgive theoffender than were victims in court.37


Table 7. Percentage of Cases Where an Apology Was Offered or Forgiveness WasExpressed



Victims

Strang et al.(1999)a

43% 296 22% 186 7.30 18.25 <.000

Offenders


67% 142 39% 132 4.07 30.37 <.000

Umbreit(1995)c

84% 140 30% 121 12.25 78.22 <.000

Umbreit &Fercello(1997)d

92% 26 54% 13 9.80 7.63 .006

Umbreit &Roberts(1996)e

80% 30 27% 22 10.81 14.60 <.000

OffendersCombined

74% 338 29% 228 6.88 125.13 <.000

Note. Drunk driving was excluded from the Strang et al. (1999) data becausethese cases did not include a personal victim (i.e., no accidents were included).

a “Percent of offenders who receive forgiveness.” b “Percent of offenders whoapologize.” c “An apology by the accused/offender was found.” d “Did youapologize?” e “Offender apologized to victim.”


38DAVIS ET AL., supra note 8, at 49–56.

F o rg iv e n e s s / A p o lo g y

0 %

2 5 %

5 0 %

7 5 %

1 0 0 %

V ic tim s O ffe n d e rs

R es to ra tiv e J u s tice C ou rt

Figure 7. Percent of Cases Where an Apology Was Offered or Forgiveness WasExpressed (With 95% Confidence Intervals).

Consistent with other outcomes, participants in restorative justice were morelikely to rate the outcome of their proceedings as fair than were participants incourt (2.6 times more likely for both victims and offenders; see Table 8 andFigure 8). However, data for this question come from only one study,38 with arelatively small sample size, so caution should be exercised in generalizing thisparticular result. The same study also showed that victims in restorative justicewere 2.3 times more likely to be satisfied with the outcome of the proceedingsthan were victims in court (see Table 9 and Figure 9). Although the effect was inthe same direction for offenders, with an odds-ratio of 1.6, it did not reachstatistical significance.


O u tc o m e w a s F a ir

R e s to ra t iv e J u s tic e ( N = 1 6 0 )

R e s to ra t iv eJ u s t ic e(N = 8 4 )

C o u r t(N = 1 1 9 )

C o u r t(N = 4 5 )

0 %

2 5 %

5 0 %

7 5 %

1 0 0 %


Table 8. Percentage of Respondents Who Believed That the Outcome Was Fair



Victims

Davis et al.(1980)a

77% 160 56% 119 2.63 13.84 <.000

Offenders

Davis et al.(1980)a

79% 84 59% 45 2.61 5.82 .016

a “Believe outcome fair?”

Figure 8. Percentage of Respondents Who Believed That the Outcome Was Fair(With 95% Confidence Intervals).


S a t is f ie d w ith O u tc o m e


R e s to ra t iv eJ u s t ic e(N = 8 4 )

C o u r t(N = 1 1 9 )

C o u r t (N = 4 5 )

0 %

2 5 %

5 0 %

7 5 %

1 0 0 %


Table 9. Percentage of Respondents Who Were Satisfied With the Outcome



Victims

Davis et al.(1980)a

73% 160 54% 119 2.30 10.81 .001

Offenders

Davis et al.(1980)a

77% 84 67% 45 1.65 1.50 .220

a “Satisfied with outcome?”

Figure 9. Percentage of Respondents Who Were Satisfied With the Outcome(With 95% Confidence Intervals).

Participants in restorative justice also ended up with significantly betterperceptions of the other parties’ behavior than did participants in court (2.4 timesmore likely for victims and 1.9 times more likely for offenders; see Table 10 andFigure 10). Again, this result depends on only two studies, one of which is notsignificant for offenders. The effect is relatively robust in this context, as always,can be generalized only through more extensive research.


Table 10. Percentage of Respondents Whose Perception of the Other Party’sBehavior Improved



Victims

Davis et al. (1980)a 62% 85 40% 55 2.45 6.49 .011

Offenders

Davis et al. (1980)a 63% 65 61% 33 1.09 0.04 .847

McCold & Wachtel(1998)b

80% 67 56% 88 3.16 9.92 .002

OffendersCombined

72% 132 57% 121 1.88 5.71 .017

a “Complainant’s [Defendant’s] perception of defendant’s [complainant’s]behavior improved.” b “What is your attitude toward the victim? (percentpositive).”


39To make the odds-ratio more compatible with other analyses, the reciprocal of the odds-ratiocan be used (1/0.46 = 2.17) to show that lessened fear is twice as likely in restorative justice.

B e tte r P e rc e p t io n o f O th e r 's B e h a v io r

R e s to ra t iv eJ u s tic e ( N = 8 5 )


C o u r t(N = 5 5 )

C o u r t (N = 1 2 1 )

0 %

2 5 %

5 0 %

7 5 %

1 0 0 %


Figure 10. Percentage of Respondents Whose Perception of the Other Party’sBehavior Improved (With 95% Confidence Intervals).

The final two questions were asked only of victims. The first was whetherthe victim was still upset about the crime after having participated in restorativejustice or court. In both studies that asked this question, victims who participatedin restorative justice were half as likely (odds-ratio=0.46)39 to feel upset about thecrime afterwards than were victims who went to court (see Table 11 and Figure11).


V ic t im S t il l U p s e t A b o u t C r im e

V ic t im s( N = 3 1 6 )

V ic t im s(N = 2 4 4 )

0 %

2 5 %

5 0 %

7 5 %

1 0 0 %

R e s to ra t iv e J u s t ic e C o u r t

Table 11. Percentage of Victims Who Were Still Upset About the Crime



Victims

Davis et al.(1980)a

23% 159 48% 120 0.32 19.12 <.000

Umbreit (1995)b 53% 157 66% 124 0.58 4.83 .028

VictimsCombined

38% 316 57% 244 0.46 20.51 <.000

a “Angry at defendant?” b “Remaining upset about the crime?”

Figure 11. Percentage of Victims Who Were Still Upset About the Crime (With95% Confidence Intervals).

Finally, Table 12 and Figure 12 show that victims in restorative justice wereone-third as likely (odds-ratio=0.34 or, for comparison, 1/0.34 = 2.94) to beafraid of revictimization than were victims in court. This effect that was consistent


40UMBREIT & ROBERTS, supra note 28, at 24.

across four studies, although it did not reach statistical significance in the EnglandStudy40 because of a small sample size.

Table 12. Percentage of Victims Who Were Afraid of Revictimization



Victims

Davis et al.(1980)a

21% 159 40% 120 0.40 11.95 .001


9% 66 26% 51 0.30 5.63 .018

Umbreit (1995)c 11% 178 31% 132 0.28 19.30 <.000

Umbreit &Roberts (1996)d

16% 43 33% 24 0.39 2.59 .108

VictimsCombined

15% 446 34% 327 0.34 38.04 <.000

a “Fear reverenge?” b “Anticipate offender will revictimize me.” c “Fear of beingrevictimized by the same accused/offender?” d “Fear of revictimization?”


V ic t im A fra id o f R e v ic t im iz a t io n

V ic t im s( N = 4 4 6 )

V ic t im s(N = 3 2 7 )

0 %

2 5 %

5 0 %

7 5 %

1 0 0 %

R e s to ra t iv e J u s t ic e C o u r t

Figure 12. Percentage of Victims Who Were Afraid of Revictimization (With95% Confidence Intervals).

III. DISCUSSION

This review summarized data on twelve psychological outcomes from aninternational collection of seven studies comparing restorative justice programsand court procedures. The results, to be frank, were extraordinary. For victims,restorative justice outperformed court on every outcome except for considerationof opinion. For offenders, restorative justice outperformed court on everyoutcome except for satisfaction with outcome. In no case did court perform betterthan restorative justice. Rather, restorative justice always held an advantage, butin certain cases the advantage (or, more frequently, the sample size) was not largeenough to reach statistical significance. In fifty separate statistical comparisons,restorative justice came out ahead of court on all fifty, significantly so in thirty-eight (76%) of them. With larger samples, every single comparison in this studycould show statistically significant advantages for restorative justice over court.

The seven studies included in this review varied in their form of restorativejustice (victim-offender mediation or family group conferencing), the country inwhich they were carried out (United States, Canada, England, and Australia), theirsample sizes (from 58 to 2353), and their methods for assigning victims andoffenders to either restorative justice or court (random assignment, matching, orself-selection, with and without the possibility of opting out of an assignedtreatment). What role did these differences play? Apparently, not much. Not oncedid court show an advantage over restorative justice—not even a shadow of anonsignificant advantage. This level of consistency is exceptionally unusual,particularly in light of the fact that studies and variables were not chosenaccording to their results (i.e., selected only if they provided data favorable to


41See, e.g., STRANG ET AL., supra note 7, at 7–28 (exhibiting perhaps most extensive data set).42See generally TONY F. MARSHALL & SUSAN MERRY, CRIME AND ACCOUNTABILITY:

VICTIM/OFFENDER MEDIATION IN PRACTICE 107–72 (1990) (providing various qualitative factorsfor research of mediation and reparation in England); GABRIELLE M. MAXWELL & ALLISONMORRIS, FAMILIES, VICTIMS, AND CULTURE: YOUTH JUSTICE IN NEW ZEALAND 136–64, 172–91(1993) (looking at impact of youth justice law in New Zealand); SUE WARNER, MAKING AMENDS:JUSTICE FOR VICTIMS AND OFFENDERS 60–78 (1992) (discussing views of both victims and accusedin reparation and mediation program in Scotland); Mark S. Umbreit & Betty Vos, HomicideSurvivors Meet the Offender Prior to Execution: Restorative Justice Through Dialogue, 4 HOMICIDESTUD. 63, 68–75 (2000) (describing qualitative effects of first death row mediation/dialoguesessions in Texas).

43For an example of such a longitudinal analysis of the Salt Lake City, Utah, victim-offendermediation program, see Jaysen R. Oldroyd, An Analysis of Restorative Justice in the UtahMediation Program 6 (2000) (unpublished senior honors thesis, University of Utah) (on file withUniversity of Utah Marriott Library).

44Nugent, supra note 4, at 147–59.

restorative justice) but simply by the existence of data that met the selectioncriteria (e.g., quantitative, comparative, psychological outcomes).

Despite these successes, there are a few important limitations andunanswered questions that remain. Several of these limitations were inherent inthe design of this review. First, this review relied exclusively on a limited numberof quantitative indicators of success and failure in administering justice. Many ofthese studies had much more extensive data that would permit richer analyses inanother context.41 Second, this review was restricted to quantitative measures, asthey facilitate the compilation and comparison of data from multiple sources. Thequalitative distinctiveness of the participants’ experiences in justice settingswould be of at least equal importance in understanding the implications of thesepractices.42

Third, the design of this review is focused primarily on data that weregathered at the time of the mediation or court disposition (or shortly thereafter).Although several of the studies included longitudinal data, the time spans andnature of the data collected subsequent to the proceedings were not consistentfrom one study to another. There is reason to believe that satisfaction with justiceproceedings wanes over time, depending primarily on the offender’s ability tomake good on the agreement or court order.43 At the same time, the metanalysisof recidivism rates44 shows that restorative justice can have important long-termeffects, at least for offenders (specifically, a 32% reduction in recidivism). Itwould, of course, be a theoretically straightforward exercise to collectlongitudinal data on the outcomes included in this study. Such data wouldcontribute substantially to an ameliorated understanding of the positivecontributions of both court procedures and restorative justice.

Three other limitations pose potentially more significant restrictions on thegeneralizability of this review. The first is that the crimes included in the sevenstudies in this review, while they covered a wide range of behaviors, did not


45See, e.g., Umbreit & Vos, supra note 42, at 67 (exemplifying study including severelyviolent crime).

46Id. at 74–82.47For example, in the Canada study, UMBREIT, supra note 21, at 4–6, and the Minnesota study,

UMBREIT & FERCELLO, supra note 24, at 5, all of the cases were originally assigned to restorativejustice. The court group was then composed of cases where the assignment to restorative justice wasrejected by the participants.

48See DAVIS ET AL., supra note 8, at 14–18; STRANG ET AL., supra note 7, at 5.

include severely violent crimes (e.g., homicide or sexual assault) that have beensuccessfully processed with restorative justice techniques.45 Rather, the greatmajority of cases included in this review addressed relatively minor juvenilecrimes, such as theft or vandalism. It would be instructive to examine the relativebenefits of restorative justice for severe crimes, as soon as enough data becomeavailable.

The second major restriction is that in all of these studies, and in restorativejustice generally, culpability is never a question. Cases are only processed inrestorative justice after guilt has been established, frequently through confessionby a contrite offender. As such, restorative justice operates primarily at thesentencing level. Although this is often seen as an advantage of restorative justice,as court trials generally maintain the rights of the defendant (particularly for dueprocess) more completely than the informal, personal procedure of restorativejustice, it is a restriction nonetheless. The experience of restorative justice withcases of severe violence shows that initial belligerence on the part of offender orresistance on the part of the victim does not permanently rule out restorativejustice.46 Instead, as long as the offender eventually accepts responsibility for hisor her behavior and shows a desire to make things better for the victim, thenrestorative justice can be effectively applied at a later date— sometimes as late astwenty years after the crime. For now, it would be important to explore more fullythe empirical consequences of the requirement that offenders accept fullresponsibility for their crimes as a prerequisite to participation in restorativejustice.

A third potentially significant restriction of this study has to do withconsistency of assignment. That is, in all seven studies cases were assigned tocourt or restorative justice through random assignment, matching, or self-selection. However, not all cases were ultimately processed as assigned. Some ofthe studies created additional groups for such cases47 and others acknowledgedthat cases deviated from original assignment,48 but in both studies those caseswere analyzed as though they were still in their originally assigned conditions.This choice represents a reasonable methodological choice, as it maintains theoriginal random assignment that is necessary for accurate statistical inference. Onthe other hand, it may prove useful in the future to break down results by bothoriginal assignment (court or restorative justice) and final processing. This would


49See supra Part I.50Doug Gray et al., Utah Youth Suicide Study, Phase I: Government Agency Contact Before

Death, 41 J. AM. ACAD. CHILD ADOLESCENT PSYCHIATRY 427 (2002).51Unless otherwise noted, all facts for this study are taken from Gray et al., supra note 50, at

428–32.

permit the research consumer to make decisions about the best way to parse thedata for his or her purposes.

IV. SUMMARY

The data in this review were consistently favorable to restorative justicewhen compared to adjudication. If outcomes such as fairness, accountability,satisfaction, contrition and forgiveness, emotional well-being, and feelings ofsafety are important, then restorative justice is the clear choice.

V. LIFE AND DEATH?

In addition to summarizing existing evaluation research on restorativejustice, another stated purpose of this Article was to “draw connections betweenrestorative justice and other areas of psychological interest.”49 It is at this pointthat this review takes temporary leave of the purely descriptive and enters thespeculative (but reasonable and vital).

One outcome that is of paramount significance to all people is whether aperson lives or dies. Doug Gray, a psychiatrist at the University of Utah, has takenspecial interest in this issue and has extensively researched adolescent suicide. Inone major project,50 Gray and his colleagues examined 151 youth suicides in Utahover a three-year period with the goals of identifying predictors of suicide risk andplanning appropriate interventions.51

This study looked at a wide range of readily available predictors of suicide(i.e., information that is routinely gathered and available from public records). Ofthe 151 Utah youths to commit suicide, 19% had been referred to child protectiveservices for abuse, 24% tested positive by medical examiners office for substanceuse, 27% had been referred to a mental health agency, 27% were behind schedulefor graduation from high school, 34% had been subject to some form of schooldiscipline, such as suspension or dismissal, 39% were receiving educationalsupport, and 40% had been participating in extracurricular activities.

However, the largest percentage (63%) had been referred to the juvenilejustice system. A single encounter with the juvenile justice system doubled theodds of suicide for a youth (compared to nonreferred youths). Eight or morereferrals led to a fivefold increase in the odds of suicide. The connection betweenjuvenile offenses and suicide risk, while correlational, is shocking and significant.


52Doug Gray et al., Juvenile Offenders: Suicide Risk and Psychiatric Symptoms, SCI. PROC.AM. ACAD. CHILD & ADOLESCENT PSYCHIATRY 16, 105.

53Unless otherwise noted, all facts for this study are taken from Gray et al., supra note 52.54Gray et al., supra note 50, at 432–33.55William R. Nugent et al., Participation in Victim-Offender Mediation and Reoffense:

Successful Replications?, 11 RES. SOC. WORK PRAC. 5, 8–9 & 15 fig.1 (2002).56William R. Nugent & Jeff Paddock, The Effect of Victim-Offender Mediation on Severity

of Reoffense, 12 MEDIATION Q. 353, 360 (1995).

In another study,52 Gray and his colleagues also provided a more in-depthlook at the mental health of youths who were referred to juvenile court.53 For onemonth they assessed the psychopathology for every person to be referred to thejuvenile courts in Salt Lake City, Utah. They found that, overall, mental healthwas poor. Forty-nine percent of all adolescents (ages twelve to seventeen) referredto court had levels of psychopathology that would merit outpatient treatment; 17%had inpatient levels of psychopathology. For juveniles who had been referredeight or more times (the group at greatest risk of suicide), the levels ofpsychopathology were 66% and 27%, respectively.

Given the seriousness of youth suicide and the great risk faced by certainyouth (particularly repeat offenders), Gray and his colleagues are designing anintervention for youth offenders with the goals of directly improving mentalhealth and reducing recidivism.54 By bearing directly on these two significantcorrelates of youth suicide, Gray and his colleagues are hoping to reduce the riskof suicide in the youth population overall.

Where does restorative justice come into all of this? Restorative justicedecreases recidivism. Research by Nugent, Umbreit, Winnimaki, and Paddockshowed that after one year, 19% of 619 participants in restorative justicereoffended, compared to 28% of 679 participants in non-restorative justice.55 Thisdifference represents a 32% reduction in recidivism (odds-ratio = 0.60, P2(1, N= 1298) = 14.50, p < .001). In another study, Nugent and Paddock also found thatreoffenders who participated in restorative justice tended to commit less severeoffenses than other offenders.56

Inasmuch as participation in restorative justice is correlated with reducedrecidivism and reduced recidivism, in turn, is correlated with decreased suiciderisk, it seems reasonable to believe that participation in restorative justice mayreduce suicide risk. This is, of course, conjecture. No studies have directlydemonstrated the link between restorative justice and offender suicide. Still, thepossibility is too important to overlook. This possibility may be pursued in at leasttwo ways: (1) addressing severe mental health problems (including suicide) aspotential outcomes to be assessed in future evaluations of restorative justiceprograms; and (2) involving restorative justice procedures in existing suicide andpsychopathology prevention programs. Either alternative would help elucidate thepotential connection between restorative justice and the very life of at-riskparticipants.


VI. CONCLUSION

Seven studies from around the world, using different methodologies toevaluate different versions of restorative justice programs and compare them withadjudication, found unanimous support for the positive outcomes of restorativejustice. It is possible, as well, that the benefits of restorative justice may even savelives. The three voices of restorative justice—theory, practice, and research—canuse this consistency as an obbligato in sustaining and enriching the collective artof justice and compassion.

Poulson (2003) - A third voice - A review of empirical research on the psychological outcomes of restorative justice

Documents