• 1 Introduction

  In recent years, restorative justice practices that embrace student accountability¹x At minimum, restorative justice practices employ a narrow definition of accountability inclusive of an offender admitting blame to a victim, and making amends to the victim as far as possible. For a nuanced discussion of accountability in relation to youth restorative conferencing, see Wood (2020). and reintegration have been introduced in schools across the United States (US) in an effort to replace and move away from the punitive zero-tolerance exclusionary policies popularised as a method for addressing school misbehaviour in the 1990s.
  Beginning in the 1980s, the punitive and exclusionary aspects of the US criminal justice system spilled over into the K-12 educational system and have been normalised over the past 30 years (Armour, 2013). Student misbehaviour like classroom disruptions, disrespect and truancy shifted from being viewed as relatively benign behaviour of children and adolescents to a sign of uncontrollable and dangerous youth. Rooted in the US criminal justice system, the criminalisation of youth behaviour justified a paradigm shift within school discipline practices. Beginning in the late 1970s and early 1980s, US criminal justice systems at the federal and state levels began shifting from rehabilitative policies towards retributive policies rooted in incapacitation (Hagan, 2010; Mauer, 2006). The move towards retributive criminal justice policy coincided with rising crime rates and a public opinion shift towards more punitive approaches (Gerber & Engelhardt-Greer, 1996). As the adult-based criminal justice system became more punitive, so did the US juvenile justice system (Feld, 2017). This trend eventually trickled down to the school setting.
  Most legislation concerning school discipline is the product of state law. While each state’s statutes differ, all have been directly influenced by the federal Gun-Free Schools Act of 1994. The Gun-Free Schools Act was, by its very terms, meant to eliminate the danger presented by firearms in schools, but its impact was much broader. Although the Act is aimed at school policy related to students who bring weapons to school, it signalled the US government’s acceptance and approval of exclusionary discipline based on zero-tolerance enforcement policies. Perhaps more importantly, the Gun-Free Schools Act mandated that every state, as a requirement of eligibility for federal education funding, adopt its own zero-tolerance policy for students possessing guns at school. In the years following the adoption of the Gun-Free Schools Act, disciplinary policies for myriad violations in the school setting became more harsh and exclusionary (Gonzalez, 2012; Nussbaum, 2018).
  Punitive and exclusionary policies offer a short-term solution to problematic student behaviour, but they come with latent and long-term consequences. Despite punitive and exclusionary policies and procedures for school discipline appearing increasingly ineffective, they have been institutionalised in the school setting in the US (Beger, 2002; Dinkes, Forrest, Cataldi & Lin-Kelly, 2007; Jacobsen, 2020; Simon, 2007). In addition, zero-tolerance, punitive-centric policies unjustly further the disenfranchisement of minority student populations through racial disparity in criminal justice contact, punishment and educational regression through exclusion (Bishop, 2005; Fronius, Persson, Guckenburg, Hurley & Petrosino, 2016; Jacobsen, 2020; Jones, 2016; Losen, 2014; Skiba, Shure & Williams, 2011). Zero-tolerance disciplinary policies utilise a risk-management approach that has long since been identified within the criminal justice system. This approach has been instrumental in creating and perpetuating the school-to-prison pipeline (Rocque & Snellings, 2018; Welch, 2018). Because zero-tolerance policies and risk-management approaches disproportionately impact racial and ethnic minorities, they further exacerbate the disproportionate minority contact with the juvenile justice system and contribute to minority students entering the school-to prison pipeline in similarly disproportionate numbers (Epstein, Blake & Gonzalez, 2017; Losen, 2014; Rocque, 2010; Skiba et al. 2011; Welch, 2018). As a result, researchers, policymakers and academics have called for alternatives to punitive and exclusionary school discipline policies (Nussbaum, 2018).
  Beginning in the late 1990s and early 2000s, restorative practices became an increasingly popular alternative to the zero-tolerance disciplinary policies within both the US juvenile justice system and US schools (Bazemore, 1998; Kurki, 2000; McGarrell, Olivares, Crawford & Kroovand, 2000; United States Department of Justice, 1998). By 2016, restorative practices had been implemented within schools in more than half of the US states (González, 2016). Owing to the decentralised nature of US schools, there are countless local or regional restorative practice models that vary in structure, process and programme fidelity. These contextual differences, rooted in decentralised school systems and leadership, may contribute to a dearth of research despite recent restorative justice popularity.
  Overall, restorative practices rely on the victim, the offender and the community to repair the harm caused by an individual’s actions. Unlike the exclusionary practices of retributive, zero-tolerance policies, restorative practices engage all the parties affected by a harmful action to create inclusion-based outcomes that repair harm, hold offenders accountable and provide for reparation (Bazemore, 1998; Zehr, 1990). Restorative practices can take many forms, and many have overlapping techniques for delivering a justice that is restorative in nature. Some of the many examples include restorative circles, family-group conferencing, victim impact panels, reparative boards, restorative community service and victim-offender mediation (Walgrave, 2011). The commonality across this continuum of progressive disciplinary approaches is a focus on healing harms inflicted on relationships, restoring dignity and respecting the voices of victims, offenders and the community (Vaandering, 2014).
  Further evaluation is necessary, but restorative practices have produced promising results in academic outcomes, discipline, attendance and other crucial aspects of the educational system (Fronius et al., 2016; Gonzalez, Sattler & Buth, 2019). And restorative practices may also show promise in reducing the disparate outcomes of exclusionary-discipline policies affecting marginalised student populations that contribute to the school-to-prison pipeline (González, 2015; Hurley, Guckenberg, Persson, Fronius & Petrosino, 2015).
  In this study, we examine what characteristics contribute to participant satisfaction within restorative circles. Extending previous work that measures overall satisfaction in restorative justice processes, we inquire into satisfaction-outcome differences across types of restorative circles and participant demographics within those circles. We estimate the effect of these characteristics across incidents and then disaggregate participants by incident type, estimating participant satisfaction among circles involving specific incident types. We further estimate the potential effect of participants’ race and gender on incident types and individuals’ roles within restorative circles.

• 2 Literature review

  2.1 Zero-tolerance policies and the school-to-prison pipeline

  Zero-tolerance polices had established a foothold within the US educational system by the late 1980s and early 1990s (González, 2012; McKenna & White, 2018; McMorris, Beckman, Shea, Baumgartner & Eggert, 2013, Simon, 2007). By 1997, about 79 per cent of American schools had adopted zero-tolerance policies in response to drugs, weapons and violence (Armour, 2013; González, 2012; Madfis, 2016; Monterastelli, 2017; Nussbaum, 2018; Wilson, 2014). The original intent of zero-tolerance policies was to address serious behavioural incidents in the school environment. But they quickly became a normative reaction to a broad range of deviant behaviour (McMorris et al., 2013; Nussbaum, 2018). Zero-tolerance policies are informed by the classical-school notions of general and specific deterrence, retribution and incapacitation (González, 2012; McMorris et al., 2013; Simon, 2007). The punishments meted out under this philosophy were intended to deter others from misbehaving or causing harm. Zero-tolerance policies are lauded by their advocates for their ability to incapacitate would-be offenders. Removing the perceived source of harm from the classroom or school is believed to ensure a safe school environment that is beneficial to learning (Freiburger & Jordan, 2016; González, 2012; McMorris et al., 2013; Sumner, Silverman & Frampton, 2010). Exclusionary punishments stemming from zero-tolerance policies have short-term benefits; they are inexpensive, swift and simple to enact.
  As a result of these perceived benefits, suspensions from school nearly doubled between 1974 and 2000. They jumped from 1.7 million in 1974 to 3.1 million in 2000 (González, 2012). During the 2015-2016 school year, 2.7 million students were removed from school via out-of-school suspension (Department of Education, 2018). In sum, schools redefined student misbehaviour; they imposed harsher sanctions on tardiness, disrespect, non-compliance and other minor disruptive behaviour. The criminalisation of school produced a net-widening effect because a wider range of student behaviour was labelled as misbehaviour and sanction-worthy (Sumner et al., 2010). According to Wilson (2014), a student who has been suspended is five times more likely to drop out of school and three times more likely to have a juvenile justice contact in the following year. Suspended students are at an increased risk of being referred to the juvenile justice system; they are removed from the learning environment, have increased opportunities to get in trouble owing to lack of supervision, and have less accessibility to resources that can assist in school performance and in reducing deviant behaviour (Fronius et al., 2016; Jacobsen, 2020; McMorris et al., 2013; Monterastelli, 2017; Sumner et al., 2010). Despite these critiques, zero-tolerance policies remain both prevalent and popular.
  The existing literature suggests that zero-tolerance/punitive policies have created a school-to-prison pipeline and contributed to the deepening of racial inequality in the US. González (2012) suggests that zero-tolerance policies in schools have contributed to tripling the prison population from 1987 to 2007, despite the decline of youth crime over the last two decades. Further, the zero-tolerance policies that stem from punitive-centric, normalised ideologies of the criminal justice system share similar flaws in regard to the reinforcement of racial disparities (Rocque & Snellings, 2018; Sykes, Piquero, Gioviano & Pittman, 2015). Minority youth, specifically African American and Hispanic students, experience exclusionary punishment more often than white students (Armour, 2013; Fabelo et al., 2011; Fronius et al., 2016; González, 2012; Losen & Skiba, 2010; Sumner et al., 2010). This is particularly troubling given that research has not found evidence for African Americans being more likely to commit serious offences when compared with students of other races (Armour, 2013; Skiba, Michael, Nardo & Peterson, 2002).
  A contributing factor is the increased surveillance common in urban schools (Simon, 2007). Highly populated schools in urban geographical areas tend to have increased measures of surveillance, such as metal detectors or security guards (Lawrence, 1998). As a result, misbehaviour is more likely discovered and punished, increasing the chances of future legal, academic and disciplinary problems for the student (Fronius et al., 2016; González, 2012; McKenna & White, 2017; Sumner et al., 2010; Welch, 2018).
  It is not just minority students who are affected by zero-tolerance policies. Students with unaddressed mental health issues are also at risk of becoming increasingly disenfranchised through exclusionary sanctions (Krezmien, Leone & Achilles, 2006; Skiba et al., 2002). Monterastelli (2017) reports that students with unaddressed mental health issues attending schools that utilise zero-tolerance policies are at a higher risk of receiving disciplinary action.
  Education is a powerful agent of youth socialisation, and youth undergo numerous critical periods of development while they are school aged (González, 2012). Students who commit non-violent harm and are punished by exclusionary measures lose access to a major contributor to socialisation. This creates a social handicap for the student who is no longer exposed to an influential agent of socialisation (Jacobsen, 2020; McMorris et al., 2013; Sumner et al., 2010). Research has shown that high drop-out rates, difficulty progressing to the next grade, other academic problems and difficulty obtaining employment are latent consequences of zero-tolerance and exclusionary procedures (Armour, 2013; González, 2012; McMorris et al., 2013; Rocque & Snellings, 2018; Sumner et al., 2010). In addition, there is little evidence to show that high rates of suspension or expulsion lead to any improvement in school safety or future student behaviour (González, 2012; Losen, 2014; Nussbaum, 2018; Riestenberg, 2001; 2012).

  2.2 Restorative justice theory and practice

  We cannot fully explore the theoretical underpinnings of restorative justice practice or the philosophy of a justice that is restorative, as either of these endeavours would require much more space than can be afforded in this article. The authors of this article recognise the cursory nature of the review of theory and philosophy of restorative justice that follows. This review is meant to afford a basic introduction to those new to the concept of restorative justice. For those lacking any exposure to restorative philosophy and practices, these concepts and practices may seem strange. This is likely because restorative practices exist outside of, and in stark contrast to, the dominant paradigm of retributive justice that has dominated the criminal justice system for some time.
  The theoretical foundation of restorative practices can be found in a range of traditional criminological theories. First and foremost among these theories are labelling theory and reintegrative shaming. Early labelling theory is often traced to the work of Frank Tannenbaum and Edwin Lemert. In fact, Lemert utilised Tannenbaum’s labelling process in his own general theory of deviance articulated in Social pathology (1951). Lemert was able to clarify and further conceptualise the basic idea that reacting in a certain fashion to deviant behaviour may cause the individual to engage in secondary deviance and may lead to the adoption of a deviant identity. More specifically, he theorised that stigmatising the person strengthens deviant conduct and eventually leads to acceptance of the deviant social status.
  The concept of stigmatising was later included in an integrated theory offered by John Braithwaite. In 1989, Braithwaite published his first edition of Crime, shame, and reintegration. A central thesis of his theory is that reoffending will be higher when shaming is stigmatising and lower when shaming is reintegrative. According to Braithwaite (1999):

  Stigmatisation is disintegrative shaming, in which no effort is made to reconcile the offender with the community. The offender is outcast, her deviance is allowed to become a master status, degradation ceremonies are not followed by ceremonies to decertify deviance (101).

  But, noting that stigmatising shaming is not the only option, Braithwaite states:

  Reintegrative shaming is shaming followed by efforts to reintegrate the offender back into the community of law-abiding or respectable citizens through words, or gestures of forgiveness or ceremonies to decertify the offender as deviant (100-101).

  The causal logic within labelling theory and reintegrative shaming represents the theoretical foundation of restorative practices. Restorative practices are particularly informed by labelling theorists’ belief that by labelling an individual as a criminal, one may actually be reinforcing the delinquent label and pushing one further into a life of crime and harm. These beliefs are manifested in restorative practices through an effort to acknowledge that while the individual’s actions were harmful to another, the individual is not a bad person. Furthermore, the community is there to demonstrate support for the individual’s efforts to live without further harming others. In so doing, the individual is not stigmatised but treated as a person of social worth. This form of reintegrative shaming encourages the individual to self-define as a good person rather than as a criminal. Consequently, individuals are less likely to act in a more harmful or criminal fashion. Reintegrative shaming is most effective before the offenders accept their deviant status and begin to follow deviant subcultures (Mongold & Edwards, 2014).
  A popular definition of restorative justice provided by Tony Marshall defines restorative justice as ‘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’ (1996: 37). The process begins with the realisation that crime signifies injury. Repairing the harm/injury is one of the central goals of restorative processes. Harm is not minimised, but humanised. Accountability for the harm is central to a justice that is restorative. Those who commit harms have a need to be held accountable for their actions (Zehr, 1990). Understanding and acknowledging the harm produced by their actions, along with accepting responsibility for their decisions and harm-causing behaviour, while difficult, can be very poignant. Restorative practices attempt to guide offenders towards understanding the harm they have caused to people and their communities. When offenders recognise that harm has occurred and take ownership of those harm-producing actions, they have been held accountable.
  Repairing the harm experienced is another fundamental goal of restorative practices. Dennis Sullivan and Larry Tifft (2001) liken restorative justice to needs-based justice. A needs-based justice response represents an attempt to achieve justice by addressing the needs of the persons involved. Achieving justice begins with the needs of the parties involved, including victims, offenders and the community. Common needs for victims might include financial needs caused by the harm, but restorative practices are meant to go beyond mere financial compensation to address other, non-financial harms experienced by victims. These non-financial needs might include the need to share one’s story, psychological and emotional needs or perhaps the need for information or an apology.
  Offenders also have needs that cannot be neglected if we are to achieve a true needs-based justice. Offenders are often victims themselves in many ways (Gerkin, 2009). They are victims of violence, aggression or neglect; and many lack emotional support and care networks, creating psychological and emotional needs of their own. Other needs for offenders include the need to be empowered and to have their behaviours challenged. Offenders also share the need to have their emotional and psychological needs addressed (Walgrave, 2011). Many offenders want someone to listen to their truth. They too need emotional support and care. They may desire forgiveness, acceptance and the assurance that they are not bad people but simply made bad decisions. Meeting offender needs is essential to ensure that they desist from harmful activities in future.
  The fundamental goals of restorative justice include accountability, meeting needs, repairing harm and community safety. Restorative practices require stakeholder participation to accomplish these goals. All stakeholders must be present and willing to participate for their needs to be identified, for offenders to be held accountable and for harm to be repaired. Finally, all stakeholders have a role in the reintegrative shaming process. This is essential to the future desistance of harm-producing behaviour. At their core restorative practices are relational.
  The relational component of restorative practices is the underpinning of what makes this approach a paradigm shift as to how harms are addressed (Zehr, 1990), by examining not only the harm-producing behaviour of the conflict but also, through dialogue, the relationship between victims, offenders and community members’ past, present and future relationships (Hollweck, Reimer & Bourchard, 2019; Llewellyn & Parker, 2018; Morrison & Vaandering, 2012; Riestenberg, 2012).
  Restorative practices encourage participation of all members affected by the harm and use communication to explore and understand subjective experiences. Stakeholders are encouraged to participate in the problem-solving process to remediate the harm and strive for the desired outcomes of decreasing future harm by strengthening and reinforcing relationships within the community (Fronius et al., 2016; Sumner et al., 2010). Victims, offenders and other interested parties – like the community – take part in a process of healing the injury experienced by the victim. Community is defined broadly and could refer to a geographical neighbourhood as well as a school community. The point is to recognise that communities experience harm when a community member is harmed. Potential feelings of loss of safety are one example of community harm. Recognising the harm to communities, restorative justice calls on all affected parties to participate in restorative processes. According to Zehr and Mika (1998), restorative justice gives the community and its members active roles in doing justice. They state that ‘the justice process draws from community resources and, in turn, contributes to the building and strengthening of community’ (53). Communities are empowered through participation, and their capacity to recognise and respond to harmful behaviours is increased (Zehr & Mika, 1998). Through participation in the process, the parties are able to hold offenders accountable, repair the harm experienced and encourage the individual responsible for the harm to refrain from future harmful behaviours.

  2.3 Restorative justice in the school setting

  Implementation of restorative philosophy is promising in schools. The school environment offers uniquely supportive opportunities to its members compared with other institutions; it is a crucial source of socialisation and normative relational engagement of youth (González, 2012; Morrison & Vaandering, 2012). González (2012) explains that students and the school community are entwined by a social contract. The educational system’s responsibility to restore harm caused by a student, imposed by the binding societal agreement, is an underlying assumption of restorative practices. Zero-tolerance policies rely on the punitive nature of the criminal justice system, disrupting and interfering with the social contract between the educational system and students.
  School-based restorative justice relies on a whole-school approach in creating a safer environment (González, 2012; González, et al., 2019). Under a restorative philosophy, excessive use of exclusionary sanctions is inappropriate. Reliance on punitive-centric, exclusionary policies mutes stakeholders’ voices in the decision-making process and hinders progression in normative social engagement (Fronius et al., 2016). Inclusion is crucial in implementing restorative practices. Researchers argue that allowing students to participate in the problem-solving process will increase perceived legitimacy and fairness in institutional power and control. The shift in perception could lead to an improvement in self-regulation and compliance (Fronius et al., 2016; Tyler, 2006; Zehr, 2002). Inclusion of stakeholders in the problem-solving process promotes an environment that strives for stronger community, improved relationships, increased empathy and proactive interventions. School-based restorative justice teaches students problem-solving and de-escalation skills (Sumner et al., 2010).
  School-based restorative justice procedures create social sub-contracts among members of the school environment to resolve dilemmas. Stakeholders in a conflict mutually agree to specific obligations they establish in their social subcontract to repair harm within the community (Sumner et al., 2010). The terms of agreement define the harm caused, humanise involved stakeholders, help develop problem-solving skills, establish accountability and offer support for the victim (McMorris et al., 2013). Reliance on agreements between stakeholders reserves exclusionary sanctions for the most serious offences, potentially amending the school-to-prison pipeline (Sumner et al., 2010).

  2.4 Research findings on restorative justice in schools

  ‘Despite the popularity of [restorative justice] in the US, most programmes are still at the infancy stage’ (Fronius et al., 2016: 18; 2019; see also, Guckenberg, Hurley, Persson, Fronius & Petrosino, 2015). Therefore, the research on such programmes is limited. Most studies, including the present one, represent descriptive research utilising small, local samples without a comparison group, as opposed to rigorous longitudinal evaluation studies (Fronius et al., 2016; 2019).
  Still, the extant literature evidences many promising outcomes and recommends continued and expanded use of restorative justice practices in schools as an overwhelming policy suggestion, particularly as a replacement for harsh zero-tolerance and retributive disciplinary practices (Fronius et al., 2016; 2019; Gonzalez, 2012). The positive outcomes associated with restorative justice practices evidenced in the research literature thus far include improved academic performance (Gonzalez et al., 2012; Jain, Bassey, Brown & Karla, 2014; Lewis, 2009; McMorris et al., 2013; Norris, 2009), increased attendance and graduation rates (Baker, 2009; Jain et al., 2014; McMorris et al., 2013; Riestenberg, 2001), and positive school climate and community relationships, including outcomes like improved school environment and increased parent, staff, and community engagement (Gonzalez et al., 2012; Health and Human Development, 2012; Jain et al., 2014; McMorris et al., 2013; Mirsky, 2007; Mirsky & Watchel, 2007; Tolefree, 2017; Voight, Austin & Hanson, 2013).
  And many negative outcomes associated with zero-tolerance disciplinary practices are nullified by restorative practices. These outcomes include declining frequency of student altercations like fighting and bullying (Armour, 2013; Baker, 2008; Davis, 2014; González, 2012; Sumner et al., 2010; Suvall, 2009) and improved school safety (Knight & Wadhwa, 2014; Morrison, Blood & Thorsborne, 2005; Riestenberg, 2012; Schumacher, 2014). Restorative justice practices are also associated with declining rates of suspension and expulsion (Armour, 2013; Baker, 2008; Davis, 2014; Sumner et al., 2010) since these practices aim to resolve issues amicably and keep students in schools.
  Additionally, restorative justice practices result in a more balanced and equitable response to student behaviour. While zero-tolerance disciplinary practices have been shown to lead to disparities in suspensions and expulsions by race and gender, producing significantly higher expulsion rates for male students of colour and disparate criminal justice consequences along the school-to-prison pipeline, restorative justice practices have been shown to produce more equitable outcomes in response to student behaviour, thereby reducing these disparities (González, 2015; Gregory, Bell & Pollock, 2014; Karp & Breslin, 2001; Payne & Welch, 2010, 2015; Stinchcomb, Bazemore & Riestenberg, 2006). Furthermore, a study by Anyon and colleagues (2014) found that restorative justice participation not only created more equitable behavioural outcomes for students by race but also acted as a protective factor that reduced the number of out-of-school suspensions for students of colour. And a study by Gregory and colleagues (2014) attributed the decline in racial disparities arising from restorative practices to improved teacher-student relationships stemming from satisfaction with the restorative justice experience. Alternatively, in a study of racial composition and the use of restorative practices, Payne and Welch (2015) found that schools with a larger proportion of black students were less likely to implement restorative practices instead of relying on punitive disciplinary policies. Furthermore, in an ethnographic study of three New York City public schools that had implemented restorative practices, Lustick (2017) concluded restorative practices were used to reinforce traditional normative social control in the school setting rather than implementing and developing a relational pedagogy.
  Many of the associated outcomes of restorative practices may be attributed to high rates of participant satisfaction and agreement (Fronius et al., 2016; 2019; Sumner et al., 2010). Participant satisfaction is integral to restorative justice process and may be linked to many of the process’s positive outcomes. As Bradshaw and Umbreit (2003) note in regard to victim satisfaction,

  victim satisfaction is important because it provides a means by which victims can have input into the restorative justice process: satisfaction data can provide feedback about program services; identify problems and needs…; and provide data that increase knowledge about the process of mediated dialogue that results in improved victim services. (71)

  Although Bradshaw and Umbreit (2003) focused on the role of victims and improvement of victim services within the criminal justice system, the importance of satisfaction likely extends to all participants in the restorative justice process, particularly in a school setting where the goal of educators and participants is to resolve behavioural conflicts among youth.
  Many studies show that restorative circles result in high rates of participant satisfaction and agreement (Cameron & Thorsborne, 2001; Fronius et al., 2016; 2019; Sumner et al., 2010). An early study by Cameron and Thorsborne (2001), conducted on one of the first schools to adopt restorative justice practices in Queensland, Australia, found high rates of satisfaction among participating students. Students reported feeling engaged in the process, feeling the process was fair, and general satisfaction with the experience (Cameron & Thorsborne, 2001). Similarly, a study by Riestenberg (2001) found that satisfaction was highest in the South St. Paul district, where restorative justice practices were more fully implemented compared with two other Minnesota school districts with varying degrees of programme implementation. The majority of student, staff, parent, applicants and victim participants reported satisfaction with the process, indicating high levels of feeling confident, grateful, supported and hopeful (Riestenberg, 2001).
  More recent studies on participant satisfaction reveal similar findings. Sumner and colleagues (2010) interviewed a sample of middle-school students and teachers in a school in Oakland, California, and found that most participants felt the restorative process was beneficial and allowed an opportunity for students to resolve conflicts and be heard. A study by Watts and Robertson (2019) surveyed student participants and found that most were highly satisfied with the restorative process. More specifically, over 90 per cent of participants agreed that they ‘had a chance to say what I needed to say’ (98 per cent), they were ‘treated fairly’ (95 per cent), that ‘everyone helped make the agreement’ (95 per cent), and that they ‘know an adult at school to go to’ (95 per cent) for help.
  Despite the importance and focus on satisfaction within the research literature, studies to date are highly descriptive and have yet to explore restorative circle conditions or participant characteristics related to satisfaction. This represents a gap in the literature that the authors seek to address. A deeper understanding of the characteristics associated with improved participant satisfaction can inform future restorative justice efforts to produce additional benefits and outcomes in school settings.

• 3 Current study

  The restorative practices implemented across five West Michigan school sites are referred to as restorative circles. Each restorative circle is facilitated by a trained restorative justice facilitator.²x Facilitators successfully completed the International Institute for Restorative Practices (IIRP) two-day training programme. For further information regarding IIRP facilitator training please visit iirp.edu/professional-development/restorative-justice-conferencing. The role of the restorative justice facilitator is to assist individuals in making their own decisions about how to resolve each conflict. It is the role of the facilitator to make sure everyone is heard. The facilitator is a neutral party in the restorative circle, helping everyone to communicate their feelings and explore options. The facilitator does not judge the parties or determine outcomes.
  Restorative circles begin with invited individual meetings with the restorative justice facilitator. During the individual meeting each student is provided with an opportunity to communicate their own version of the conflict/event, and the facilitator describes the restorative process for the student. After the individual meetings, and upon student agreement, the parties are brought together for their restorative circle. At this meeting the facilitator uses restorative questioning to spur dialogue, encourage participation, reflection and recognition. Typical questions include, but are not limited to:

  • What happened?

  • What do you need to do to make things right?

  • What impact has this incident had on you and others?

  • What can you do if a similar thing happens again?

  Throughout the process, the facilitator takes notes and asks supplemental questions as needed. As the students come to an agreement, the facilitator captures the agreement in writing. The facilitator enables the parties to decide for themselves what the agreement should include. The facilitator then reads the agreement to the students, making any necessary changes that have been agreed on. The agreement is then printed and signed by all participants. At the conclusion of the restorative circle each participant is asked to complete a survey about their perceptions of the restorative process and outcomes.
  The present study aims to identify characteristics related to successful outcomes in restorative justice circles. More specifically, it seeks to identify what factors contribute to participant satisfaction. Towards this end, 2017-2018 survey data (n = 1,313) is utilised across five West Michigan schools to determine how participant and restorative circle characteristics contribute to participant satisfaction within OLS regression models among the full sample, as well as within restorative circles involving specific incident types. In addition, a second stage in the analysis further examines participant characteristics, including race and gender, and whether these characteristics might moderate these relation­ships.

  3.1 Data and methods

  3.1.1 Data

  Data collection took place across five West Michigan schools that began implementing restorative circles in 2016 and 2017. Restorative circles were implemented by third-party facilitators trained in restorative justice practices and were provided by the Dispute Resolution Center of West Michigan, a non-profit community mediation centre that has been servicing West Michigan for over 30 years. Data collection took place from March 2017 to July 2018 and includes 576 restorative circles, involving participants engaging in 1,742 unique roles.
  Data for the present study is derived from three sources. First, restorative circle facilitators collected data on incident type, participants’ demographic characteristics and role in the dispute, and other characteristics of the restorative circle. The 1,742 cases represent at least 965 unique individuals; 645 of these participants were involved in a single circle, while 320 were engaged in two or more circles (for 116 respondents the number of times participating in a restorative circle could not be determined). In addition, corresponding school-level data on student detentions, classroom detentions, classroom disputes, school suspensions, expulsions, unexcused absences and student population statistics were obtained from each site.
  Participants who took part in a restorative circle were also surveyed about their experience. Survey data includes responses from each participant on items like level of satisfaction within the restorative circle process, whether participants listened to the respondent within the circle, whether participants were treated fairly, whether apologies were given or received on either side of the circle, and other items related to the restorative justice experience. Those included in the survey data (75.4 per cent response rate, n = 1,313)³x Administrative/facilitator data was collected from restorative circle participants who chose not to participate in the restorative circle survey (n = 429). Therefore, comparison of these individuals with those who are included in the survey data (n = 1313) is possible. A significantly greater proportion of restorative justice participants who did not complete the survey took on the role of witness (20.7 per cent), and a significantly smaller proportion of those not included in the survey data took on the role of victim or aggressor (34.8 per cent) compared with those included in the survey (7.2 per cent witnesses and 52.6 per cent victims/aggressors). Additionally, a greater proportion of those not included in the survey data fulfilled the role of ‘other’ compared with those who complete the survey (10.7 per cent compared with 4.3 per cent, respectively). Those included in the survey participated in a significantly greater number of restorative circles (mean = 2.04) compared with those who did not complete the survey (mean = 1.70), and those included in the survey include a significantly higher proportion of repeat participants (61.9 per cent) compared with those who did not participate in the survey (55.6 per cent). Survey participants include a greater proportion of Hispanic individuals (32.8 per cent) compared with those who did not participate (24.3 per cent). Additionally, those who did not participate are significantly older by an average of 0.6 years and are enrolled in 1.5 grade levels higher on average. A significantly greater number of incidents involving staff/student conflicts and classroom disruptions took place among those who chose not to participate in the survey. However, a significantly greater number of physical altercations, incidents involving threats and/or intimidation, cyber conflicts and incidents involving gossip and/or rumour spreading occurred among those who participated in the survey compared with those who did not. Those included in the survey are significantly more likely to have reached an agreement, spent significantly more time in the restorative circle and saved significantly more total suspension days compared with those who did not participate. participated in 486 restorative circles. These circles included 452 participants who participated in one circle and 256 participants who engaged in two or more circles during the data collection period (for which at least one survey was completed), as well as 63 individuals who could not be assigned owing to missing identifiers.

  3.1.2 Measures

  3.1.2.1 Dependent variable

  Participant satisfaction is a multidimensional 13-item scale (α = 0.91) that includes participants’ responses to items intended to gauge their level of satisfaction with participation in a restorative justice circle.⁴x All respondents were provided with the same survey instrument (Appendix A). Students were advised to skip questions that did not apply to their role within the dispute (i.e. victims would not be expected to take responsibility for their actions). Surveys were subsequently cross-checked during data input. Each included item asked respondents to answer whether they ‘strongly agree’, ‘agree’, are ‘unsure’, ‘disagree’ or ‘strongly disagree’ with statements such as the ‘staff explained how a restorative circle works’, ‘I feel that I was treated fairly in the restorative circle’ and ‘the restorative circle helped me understand how my actions affected others’. These and other included items, listed in their entirety in Appendix A, loaded on the first factor included in the multidimensional scale and all factor scores on this factor were over 0.61. A second factor included in the participant-satisfaction scale included respondents’ ranking of agreement with items like ‘I feel the other person took responsibility for their behaviour’, ‘I feel the other person listened to me during the restorative circle’ and ‘I have apologised for my actions’. These and additional included items listed in Appendix A all had factor loadings that were over 0.53. Each of the thirteen items in the multidimensional scale were coded from ‘1’ to ‘5’, with ‘5’ indicating ‘strongly agree’ and ‘1’ indicating ‘strongly disagree’. The average of each of the thirteen items was calculated for each respondent.

  3.1.2.2 Independent variables

   

  Participant characteristics

  Gender, grade and race. Demographic characteristics, including participants’ gender as reported by administrators (coded as ‘male’ = ‘1’ and ‘female’ = ‘0’) and participants’ grade level in school⁵x Grade was used in place of participants’ age since less missing data exists for grade compared with age and since they are highly correlated (r = 0.90, p < 0.01). and their race, are included as participant characteristics. Participants’ race was dummy coded as ‘0’ or ‘1’ for categories of ‘white’, ‘black’, ‘Hispanic’ and ‘other’. ‘Other’ races included Asian, Native Indian or Alaska Native, Bosnian and Dominican. When respondents reported multiple races, race was coded according to the minority category given societal interpretations and racial constructions.
  Role.⁶x Categories for role were developed on the basis of a pre-test of the data collection method. The role that each respondent took on in each incident was classified by administrators as ‘victim’, ‘aggressor’, ‘victim/aggressor’, ‘witness’, ‘participant’, ‘teacher’, ‘other’ and ‘subject of rumour’. Categories of ‘witness’ and ‘participant’ were combined in view of substantive similarities. Additionally, two cases classified as ‘subject of rumour’ were recoded into the ‘victim’ category, and three ‘teachers’ were recoded into the ‘other’ category. The resulting categories (‘victim’, ‘aggressor’, ‘victim/aggressor’, ‘witness/participant’ and ‘other’) were dummy coded ‘0’ or ‘1’ for each respondent.

   

  Restorative circle characteristics

  Time in circle, number of participants and agreement reached. The time each participant spent within the restorative justice circle was coded in minutes for each respondent. The number of participants within the circle was also totalled, ranging from 1 to 30 (mean = 6.06). In addition, whether the restorative justice circle resulted in an agreement being reached was included. Nearly all (98.16 per cent) circles ended in an agreement being reached, consistent with prior literature (Fronius et al., 2016; Sumner et al., 2010).
  Times participated. The number of times each respondent was involved in a restorative circle was counted for each participant. Each restorative circle was then coded ranging from ‘1’ to ‘15’, indicating ‘first’, ‘second’, ‘third’, and subsequent restorative circle participation. The ordering of participation was determined by the month and year that each circle took place. However, when a participant was involved in more than one circle in the same month, the sequential case number was used to determine which occurrence took place first.⁷x For one participant, the order of participation in restorative circles could not be determined owing to a missing case number. Therefore, this participant’s circles were coded using the sequence, ‘1’, ‘4’, ‘4’, ‘4’, ‘5’, ‘6’.
  Incident type. The type of incident each respondent was engaged in that led to their participation in a restorative circle was dummy coded as ‘1’ or ‘0’ for each incident category. Facilitators coded as many incident types as applied to each circle. For example, an individual may have participated in a restorative circle that was involved in a ‘verbal conflict’ as well as ‘friendship issues’ and a ‘physical altercation’. Other incident types include ‘threats/intimidation’, ‘cyber conflict’, ‘taunts/teasing’, ‘gossip/rumours’ and ‘classroom disruption’.
  Facilitator and school. The trained facilitator who led each circle and the school that each participant attended where the circle took place were also included as controls. Facilitators and schools have been deidentified and relabelled (i.e. ‘School 1,’ ‘Facilitator 5’).

  3.1.3 Missing data

  Since the focus of this study is on incidents involving students, participants who were involved in restorative circles that sought to resolve ‘student/staff conflicts’ (n = 146) were excluded from the analysis, resulting in 1,167 remaining restorative circle participants involved in student conflicts. An additional 34 cases (2.9 per cent) could not be categorised by incident type because of missing data and were therefore excluded from the analysis. Thirteen missing responses on gender, fifteen missing responses on race and eleven missing responses on grade were imputed using valid data associated with respondents’ participation in another restorative circle. Missingness on the variable ‘grade’ was adjusted to reflect the month and date of participation (it was assumed that participants moved on to the next grade level each year unless otherwise indicated by existing data). Fifty cases (4.2 per cent) were dropped owing to missing data on the respondent’s role in the incident that led to participation in a restorative circle. Additionally, 19 (1.6 per cent) cases were dropped owing to missing data on race, 6 (0.5 per cent) were dropped owing to missing information of gender, and 22 were dropped owing to missing data on ‘agreement’ (1.9 per cent). All missing data (less than 5 per cent on all variables) associated with continuous variables were mean imputed to retain these cases within the analysis, resulting in a final sample of n = 1,035 (88.7 per cent of the total sample after excluding cases of student/staff conflicts).

  3.1.4 Analytic strategy

  OLS regression models are used to estimate the relationship between participant and restorative justice circle characteristics and participant satisfaction. We begin by estimating the effect of these characteristics on participant satisfaction for all incidents. We then disaggregate participants by incident type to estimate the relationship between participant and restorative circle characteristics on participant satisfaction among restorative circles involving specific incident types, including verbal conflicts (n = 350), friendship issues (n = 243), physical altercations (n = 205) and threats/intimidation (n = 164).⁸x Incident types with inadequate sample size to estimate these effects are excluded from this stage of the analysis (i.e. cyber conflict (n = 53), taunts/teasing (n = 65), gossip/rumours (n = 65), classroom disruption (n = 72)).
  The second stage of the analysis proceeds to estimate potential moderating effects of participant characteristics on satisfaction. Given that previous punitive polices have been shown to produce disproportionate negative consequences based on individuals’ racial and gender characteristics, we further estimate the potential effects of participants’ race and gender by examining whether these characteristics interact with individuals’ roles within the restorative justice circle. To do so, we add interaction terms for race and role of participant (race*role of participant) and gender and role of participant (gender*role of participant) to each model to estimate the potential moderating effect of participant demographic characteristics (race and gender) and role in the circle on participant satisfaction.

  3.2 Results

  3.2.1 Descriptives

  Table 1 shows the means or percentages and standard deviations for each study variable. The sample is 37.3 per cent male and the mean grade level of participants is 7.92. A nearly even proportion of participants are black (35.0 per cent) and Hispanic (35.3 per cent), followed by 26.6 per cent white and 3.2 per cent ‘other’ races. The majority of participants (58.6 per cent) were classified as participating in the combined role of ‘victim/aggressor’.

  Table 1 Descriptive statistics

  	Mean/%	SD	Min.	Max.
  Dependent Variable
  Participant Satisfaction –‍ ‍13-‍item scale (α = 0.91)	4.22	0.59	1.38	5.00
  Independent Variables
  Participant Characteristics
  Gender (Male = 1)	37.29%	0.48	0.00	1.00
  Grade	7.92	1.60	6.00	12.00
  Race
  White	26.57%	0.44	0.00	1.00
  Black	34.98%	0.48	0.00	1.00
  Hispanic	35.27%	0.48	0.00	1.00
  Other	3.19%	0.18	0.00	1.00
  Role in Circle
  Victim	10.92%	0.31	0.00	1.00
  Aggressor	12.08%	0.33	0.00	1.00
  Victim/Aggressor	58.65%	0.49	0.00	1.00
  Witness/Participant	13.53%	0.34	0.00	1.00
  Other	4.83%	0.21	0.00	1.00
  Restorative Circle Characteristics
  Time in Circle (Minutes)	80.51	44.16	2.00	240.00
  Number of Participants	3.76	4.02	1.00	29.00
  Times Participated	1.90	1.74	1.00	15.00
  Agreement Reached	98.16%	0.13	0.00	1.00
  Incident Type (can be more than one)
  Verbal Conflict	33.82%	0.47	0.00	1.00
  Friendship Issues	23.48%	0.42	0.00	1.00
  Physical Altercation	19.81%	0.40	0.00	1.00
  Threats/Intimidation	15.85%	0.37	0.00	1.00
  Cyber Conflict	5.12%	0.22	0.00	1.00
  Taunts/Teasing	6.28%	0.24	0.00	1.00
  Gossip/Rumours	6.28%	0.24	0.00	1.00
  Classroom Disruptions	6.96%	0.25	0.00	1.00
  Facilitator
  Facilitator 1	23.29%	0.42	0.00	1.00
  Facilitator 2	27.15%	0.44	0.00	1.00
  Facilitator 3	29.76%	0.46	0.00	1.00
  Facilitator 4	7.44%	0.26	0.00	1.00
  Facilitator 5	12.37%	0.33	0.00	1.00
  School
  School 1	23.29%	0.42	0.00	1.00
  School 2	27.15%	0.44	0.00	1.00
  School 3	2.42%	0.15	0.00	1.00
  School 4	27.34%	0.45	0.00	1.00
  School 5	19.81%	0.40	0.00	1.00

  The average amount of time spent in a restorative justice circle is 80.5 minutes, and circles include 3.8 participants on average, although the number of participants ranged from 1 to 29. Participants were involved in 1.9 restorative circles, on average, throughout the period of data collection, and agreement was reached in nearly all circles (98.2 per cent). The most common incidents brought to restorative justice circles are characterised by issues including verbal conflicts (33.8 per cent), followed by problems arising from friendship issues (23.5 per cent), physical altercations (19.8 per cent) and/or threats and intimidation (15.8 per cent).
  The average ranking on the scale of satisfaction with restorative circles is 4.22 (sd = 0.59). As shown in Appendix A, respondents reported the highest average agreement with the statement ‘staff explained how a restorative circle works’ (mean = 4.49, sd = 0.64) and the lowest average agreement with the statement ‘the other person apologised for their actions’ (mean = 3.95, sd = 1.13).

  3.2.2 Multivariate results

  The multivariate models estimating the effect of participant and restorative circle characteristics on participant satisfaction are presented in Table 2. The first column of results presented in Table 2 includes ‘All Incidents.’ Findings presented for Model 1 within this column show that participants’ role as a ‘witness/participant’ (b = 0.14, p < 0.05) and ‘other’ participants’ role (b = 0.25, p < 0.01) have a significant and positive effect on participant satisfaction compared with the reference-category role of ‘victim/aggressor’. In addition, several characteristics of restorative circles, including the amount of time spent in the restorative circle (b = -0.00, p < 0.01) and the number of participants within the circle (b = -0.03, p < 0.01), have a significant and negative effect on participant satisfaction, while other characteristics of the circle, including the number of times a respondent participated in a circle (b = 0.40, p < 0.01) and whether or not an agreement was reached (b = 0.64, p < 0.01), are associated with increased participant satisfaction. Together, these findings indicate that as more individuals join and as time spent in a particular circle increases, the likelihood that participants will be ‘highly satisfied’ with their experiences in the restorative circle decreases; while reaching an agreement, previous experience in a restorative justice circle and being in a role that is less central to the conflict (‘witness/participant’ and ‘other’) are related to higher satisfaction.

  Table 2 OLS regression of participant and restorative circle characteristics on satisfaction

  	All Incidents (n = 1,035)		Verbal Conflicts (n = 350)		Friendship Issues (n = 243)		Physical Altercation (n = 205)		Threats/Intimidation (n = 164)
  	Model 1	Model 2	Model 1	Model 2	Model 1	Model 2	Model 1	Model 2	Model 1	Model 2
  Participant Characteristics
  Gender (male = 1)	-0.064	-0.023	-0.032	0.013	-0.057	0.040	0.057	0.048	0.032	0.116
  	(0.038)	(0.043)	(0.063)	(0.070)	(0.110)	(0.124)	(0.073)	(0.087)	(0.117)	(0.142)
  Grade	0.005	0.008	-0.051	-0.056	0.034	0.039	0.011	0.017	-0.074	-0.053
  	(0.021)	(0.022)	(0.041)	(0.041)	(0.052)	(0.052)	(0.040)	(0.042)	(0.063)	(0.065)
  Race (white = reference category)
  Black	0.054	0.034	0.041	0.007	-0.062	-0.164	0.077	0.123	0.158	0.160
  	(0.045)	(0.051)	(0.078)	(0.087)	(0.105)	(0.114)	(0.083)	(0.097)	(0.126)	(0.145)
  Hispanic	-0.009	0.007	-0.050	0.005	-0.105	-0.176	0.164	0.188	0.030	0.066
  	(0.048)	(.053)	(0.086)	(0.096)	(0.105)	(0.107)	(0.094)	(0.111)	(0.139)	(0.162)
  Other	0.014	0.024	-0.011	0.000	-0.006	-0.010	0.300	0.282	-0.340	-0.258
  	(0.107)	(0.108)	(0.191)	(0.191)	(0.164)	(0.162)	(0.256)	(0.260)	(0.368)	(0.376)
  Role in Circle (victim/aggressor= reference category)
  Victim	-0.029	0.044	0.007	0.153	-0.134	-0.348	0.058	-0.011	0.016	0.051
  	(0.057)	(0.120)	(0.101)	(0.223)	(0.131)	(0.293)	(0.107)	(0.960)	(0.159)	(0.390)
  Aggressor	-0.039	0.005	-0.078	-0.051	0.027	-0.578*	0.054	0.226	-0.095	0.034
  	(0.056)	(0.118)	(0.090)	(0.248)	(0.154)	(0.290)	(0.095)	(0.207)	(0.160)	(0.377)
  Witness/Participant	0.136*	0.134*	0.311**	0.316**	-0.013	-0.027	-0.020	-0.025	0.392**	0.394**
  	(0.061)	(0.061)	(0.107)	(0.107)	(0.127)	(0.126)	(0.130)	(0.131)	(0.155)	(0.155)
  Other	0.246**	0.255**	0.178	0.155	0.066	0.065	0.423	0.423	0.347	0.295
  	(0.097)	(0.097)	(0.545)	(0.545)	(0.200)	(0.199)	(0.343)	(0.347)	(0.360)	(0.367)
  Interaction Effects
  Black*Victim		-0.61		-0.007		0.442		-0.148		-0.257
  		(0.143)		(0.254)		(0.380)		(0.261)		(0.392)
  Black*Aggressor		0.190		0.236		0.933**		-0.211		0.400
  		(0.134)		(0.264)		(0.337)		(0.220)		(0.419)
  Hispanic*Victim		-0.055		-0.248		0.357		0.151		0.082
  		(0.140)		(0.257)		(0.354)		(0.250)		(0.440)
  Hispanic*Aggressor		-0.058		-0.160		1.093**		-0.277		-0.028
  		(0.141)		(0.269)		(0.415)		(0.243)		(0.399)
  Male*Victim		-0.083		-0.141		-0.454		0.103		0.002
  		(0.114)		(0.202)		(0.339)		(0.215)		(0.351)
  Male*Aggressor		-0.218*		-0.179		-0.109		-0.009		-0.562
  		(0.107)		(0.175)		(0.346)		(0.204)		(0.317)
  Restorative Circle Characteristics
  Time in Circle	-0.002**	-0.002**	-0.001	-0.001	-0.003**	-0.033**	-0.003	-0.003	-0.002	-0.002
  	(0.001)	(0.001)	(0.001)	(0.001)	(0.001)	(0.001)	(0.002)	(0.002)	(0.002)	(0.002)
  Number of Participants	-0.033**	-0.033**	-0.014	-0.015	-0.048	-0.049	-0.042	-0.032	0.058	0.057
  	(0.007)	(0.007)	(0.029)	(0.029)	(0.037)	(0.037)	(0.052)	(0.054)	(0.033)	(0.033)
  Times Participated	0.040**	0.039**	0.055*	0.056*	0.021	0.024	0.080**	0.081**	0.032	0.026
  	(0.011)	(0.011)	(0.026)	(0.026)	(0.021)	(0.021)	(0.024)	(0.224)	(0.030)	(0.030)
  Agreement Reached	0.638**	0.634**	1.198	1.17	–	–	0.527*	0.530*	–	–
  	(0.172)	(0.172)	(0.771)	(0.780)	–	–	(0.220)	(0.224)	–	–
  School/Facilitator (School 1/Facilitator 1 = reference category)
  School 2/Facilitator 2	-0.188**	-0.181**	-0.267*	-0.284*	-0.274	-0.274	-0.310*	-0.299*	-0.256	-0.193
  	(0.068)	(0.069)	(0.129)	(0.129)	(0.148)	(0.147)	(0.146)	(0.152)	(0.217)	(0.223)
  School 3/Facilitator 3	-0.085	-0.082	-0.165	-0.161	0.298	0.163	-0.050	-0.083	-0.016	-0.041
  	(0.123)	(0.124)	(0.186)	(0.187)	(0.324)	(0.329)	(0.359)	(0.370)	(0.289)	(0.304)
  School 4/Facilitator 3	0.082	0.089	0.106	0.089	0.215	0.188	-0.208	-0.214	-0.102	-0.038
  	(0.063)	(0.063)	(0.113)	(0.114)	(0.145)	(0.145)	(0.147)	(0.151)	(0.187)	(0.191)
  School 5/Facilitator 4	0.029	0.015	0.237	0.268	0.186	0.108	-0.342	-0.420	0.410	0.669
  	(0.095)	(0.096)	(0.142)	(0.145)	(0.208)	(0.207)	(0.362)	(0.380)	(0.445)	(0.471)
  School 6/Facilitator 5	0.131	0.123	0.291*	0.300*	0.501**	0.514**	-0.070	-0.077	0.261	0.314
  	(0.086)	(0.086)	(0.138)	(0.140)	(0.203)	(0.201)	(0.181)	(0.185)	(0.226)	(0.235)
  Constant	3.7041**	3.701**	3.360**	3.419**	4.448**	4.472**	3.821**	3.730**	4.715**	4.462**
  	(0.271)	(0.272)	(0.897)	(0.907)	(0.463)	(0.458)	(0.433)	(0.453)	(0.579)	(0.610)
  R2	0.142	0.149	0.257	0.273	0.192	0.239	0.191	0.207	0.136	0.175

  ^*p = 0.05, ^**p = 0.01. Unstandardised regression coefficients are reported; standard errors in parentheses (n = 1,035).

  Table 2 also presents findings for models estimating the effect of participant and restorative circle characteristics on participant satisfaction disaggregated by incident type. The Model 1 columns continued in Table 2 present the results for restorative justice circles that attempted to address an issue that included a verbal conflict (n = 350), a friendship issue (n = 243), a physical altercation (n = 205) and threats and/or intimidation (n = 164), respectively. Similar results to those found within the analysis inclusive of the full sample are also present within the disaggregated models. For example, the role of ‘witness/participant’ is statistically significant compared with the role of ‘victim/aggressor’ for incidents involving verbal conflicts (b = 0.31, p < 0.01) and for incidents involving threats and/or intimidation (b = 0.39, p < 0.01); but there is no significant effect of participant’s role for incidents involving friendship issues or physical altercations. The number of times a participant was involved in a restorative circle has a significant and positive effect on their satisfaction in circles involving both verbal conflicts (b = 0.06, p < 0.05) and incidents involving physical altercations (b = 0.08, p < 0.01) but not friendship issues or threats/intimidation. Lastly, time spent in the restorative circle has a small but significant negative effect on participants’ satisfaction only in circles involving friendship issues (b = -0.00, p < 0.01).
  It is also worth noting that none of the demographic characteristics of participants, including gender, grade level or race, are directly related to participant satisfaction in the model including all incidents or in the models disaggregated by incident type. Prior research shows that punitive policies in schools have disproportionately involved, and have had negative effects on, students, based on race and gender. Stage two of the analysis examines whether restorative justice circles have similar effects as the punitive policies they are attempting to replace by analysing how race and gender, and their potential interaction, affect student satisfaction with participation in restorative justice circles. To examine this, we added interaction terms for race and role of participant (race*role of participant) and gender and role of participant (gender*role of participant) to each model.
  These results are presented for all incidents and by incident type in the Model 2 columns of Table 2. Findings show that gender and participant role interact to have a significant effect on satisfaction, as indicated by the significant interaction term for male*aggressor (b = -0.22, p < 0.05). This relationship is illustrated in Figure 1. As the bar graph displays, girls have higher satisfaction in the role of aggressor compared with boys in the role of aggressor.

  Participant satisfaction in all circles for girls and boys separated by role.
  

  The interaction effect among these variables can be more clearly seen in Figure 2. The role of aggressor has a positive effect on satisfaction for girls (the grey line in Figure 2), while the effect of being an aggressor is negative for boys (the black line in Figure 2).

  Effect of aggressor role on participant satisfaction in all circles for girls and boys.
  

  Similarly, the effect of being an aggressor in incidents involving friendship issues interacts with race to have a significant effect on participant satisfaction. As shown in Figure 3, both black (b = 0.93, p < 0.01) and Hispanic (b = 1.09, p < 0.01) students’ participation in a restorative justice circle involving friendship issues in the role of an aggressor is related to higher participant satisfaction compared with white participants and participants of ‘other’ acting in the same role.⁹x Within the restorative practices circles categorised as ‘friendship issues’ no individuals were categorised as victims.
  More specifically, Figure 4 shows that the effect of being an aggressor on satisfaction is positive for black and Hispanic participants but negative for white and ‘other’ races, indicating that race significantly moderates the relationship between role and satisfaction in circles involving friendship issues.

  Participant satisfaction in circles involving friendship issues for participants by race and separated by role.
  
  Effect of aggressor role on participant satisfaction separated by race.
  

• 4 Discussion

  Previous restorative practice research findings that highlight improved school environment, engagement and academic success are promising (González et al., 2016; Health and Human Development, 2012; Jain et al., 2014, McMorris et al., 2013; Mirsky, 2007; Mirsky & Wachtel 2007; Tolefree, 2017; Voight et al., 2013). This study continues within the restorative practice research trend focusing on descriptive analysis and fills a gap in the literature identifying characteristics related to successful outcomes in restorative justice circles. Acknowledging differences in characteristics of participants and these characteristics’ association with improved participant satisfaction not only informs future restorative justice efforts but also acknowledges the complex differentiation among disputes and disputants.
  In particular, the study’s main findings show that characteristics of restorative circles (i.e. the number of participants, amount of time spent in the restorative circle, the number of times respondents have participated in a circle and whether or not an agreement was reached) significantly impact participant satisfaction, indicating that specific conditions within the restorative process can be cultivated to achieve more favourable outcomes for stakeholders. Significant effects related to gender and participant role in restorative practice satisfaction were also found; specifically, girls were found to have a higher participant satisfaction in the role of dispute aggressor compared with boys in the role of aggressor. In addition, the effect of being an aggressor in friendship-issue disputes was found to have a higher participant satisfaction for black and Hispanic participants than for white students. These findings reinforce and expand on previous middle-school student restorative practice satisfaction outcomes found by Watts and Robertson (2019) and add to the plethora of positive satisfaction outcomes across the existing restorative practice literature (Cameron & Thorsborne, 2001; Fronius et al., 2016, 2019; Riestenberg, 2001; Sumner et al., 2010; Watts & Robertson, 2019).
  Along with the findings reported within this study, the authors would be remiss not to acknowledge particular weaknesses in the study design and data. In particular, this study is descriptive in nature. While the design is additive, with the inclusion of further depth related to disputant and dispute characteristic variables, the research design is not experimental or longitudinal, and it is unlikely that restorative practice programme implementation created school culture change within such a short time frame. Furthermore, a common criticism and disadvantage of survey research, in general, is that respondents may not be comfortable or may not be encouraged to provide accurate, honest answers (Babbie, 1990; Nardi, 2006). While there is no indication that this is the case, and while facilitators were provided with clear instruction on fidelity in data collection, given the exceptionally high rate of satisfaction among respondents, this general methodological criticism must be acknowledged. In addition, a clearer understanding of participant satisfaction may be possible with the inclusion of race and gender variables of restorative practice facilitators within the statistical model. Finally, missing data as a result of third-party facilitator data collection omissions and school administrator differences in data collection across sites created analysis challenges whereby just over eleven per cent of the eligible cases for analysis were excluded. In addition, a small percentage (<5 per cent) of continuous variables’ missing data were mean imputed for retention purposes within the analysis.
  At the heart of a whole-school intervention platform reliant on restorative practices is replacing overt zero-tolerance disciplinary practices with a focus on repairing social sub-contracts between disputants to a normative state (González et al., 2019; McMorris et al., 2013). Teaching middle-school students self-reliance in solving disputes through restorative practices that emphasise inclusion and on-going student relationships repairs harm without exclusion. Accountability for harm caused can be achieved without methods explicitly reliant on exclusion; and restorative practices attempt to achieve this justice with the inclusion of all parties involved in a dispute (Marshall, 1996; Sullivan & Tifft, 2001; Zehr, 1990). Developing the skills of self-regulation, problem-solving, ownership of responsibility and humanisation of others are an inherent component of K-12 education as well as restorative justice theory.
  Replacing zero-tolerance disciplinary practices, ingrained in decades of school administration policy and practice, will not be easy. Policymakers are slowly realising that the pendulum has swung too far toward an over-reliance on severity and incapacitation as a disciplinary outcome. Beyond egregious acts suitable for exclusionary approaches, there is palpable damage to individuals and communities where zero-tolerance policies rooted in banishment from normative educational opportunities exist. These damages are specifically prevalent in communities of colour and among minority students, who continue to bear the brunt of these policies as represented within the school-to-prison pipeline.

• 5 Conclusion

  Pragmatic and incremental changes to school discipline policies are needed. Whole-school approaches rooted in restorative practices that address school discipline are promising. Yet without additional systemic work on the part of school administrators, communities and policymakers, punitive approaches will be perpetuated alongside burgeoning restorative practices. It is important to note not only when restorative practices should be implemented within schools, but also when they should not. Restorative practices should not be implemented when they are used as a net-widening tool, when they perpetuate bias, or when they do not meet the specific needs of victims, offenders and the community (Lustick, 2017). Restorative practices should be used when addressing vulnerable populations and problems such as the adultification of black girls (Epstein et al., 2017). School disciplinary changes applying restorative practices towards specific types of incidents, with students who are most likely to engage in restorative practices, and who have shown significant levels of participant satisfaction, is not wholesale change but positive nonetheless.
  It is not clear from the current research that restorative practices are a one-size-fits-all solution to school discipline any more than current policies weighted towards zero-tolerance approaches. The authors suggest that expanded restorative training, application, data collection and analysis are needed across school settings and within different incident types and disputants. Future work focusing on informed data analysis rooted in experimental and longitudinal designs, coupled with expanded training of restorative theory and practices applied to situationally specific incidents and student dyads, will serve to incrementally change the culture of school discipline processes and pull the overcorrections of exclusion approaches implemented in the past back towards a balanced justice approach. This approach continues the praxis for prescriptive change and pushes the research on restorative practices from the tethers of theoretical postulation and descriptive analysis towards pragmatic application and causal inference. Prior research indicates that suspension and exclusion are being used inordinately for minor classroom misbehaviour and attendance problems (Fronius et al., 2016; Skiba et al., 2014). Whole-school initiatives applying restorative practices (González et al., 2019) enhance inclusion and ameliorate the social inequality and stigmatisation of current exclusionary practices prevalent within school discipline policy and practice.

References

• Anyon, Y., Jenson, J., Altschul, I., Farrar, J., McQueen, J., Greer, E., Downing, B. & Simmons, J. (2014). The persistent effect of race and the promise of alternatives to suspension in school discipline outcomes. Children and Youth Services Review, 44, 379-386. doi: 10.1016/j.childyouth.2014.06.025.

• Armour, M. (2013). Ed White middle school restorative discipline evaluation: implementation and impact, 2012/2013 sixth grade. Austin: University of Texas.

• Babbie, E. (1990). Survey research methods (2nd ed.). Belmont: Wadsworth.

• Baker, M. (2008). DPS restorative justice project: executive summary. Retrieved from http://restorativesolutions.us/wp-content/uploads/2013/11/RestorativeSolutions-DPSRJ-ExecSum07-08.pdf (last accessed 7 May 2021).

• Bazemore, G. (1998). Restorative justice and earned redemption: communities, victims, and offender reintegration. American Behavioral Scientist, 41(6), 768-813. doi: 10.1177/0002764298041006003.

• Beger, R. (2002). Expansion of police power in public schools and the vanishing rights of students. Social Justice, 29, 119-130.

• Bishop, D. (2005). The role of race and ethnicity in juvenile justice processing. In D. Hawkins & K. Kempf-Leonard (eds.), Our children, their children: confronting racial and ethnic differences in American juvenile justice (pp. 23-82). Chicago: University of Chicago Press.

• Bradshaw, W. & Umbreit, M. (2003). Assessing satisfaction with victim services: the development and use of the victim satisfaction with offender dialogue scale (vsods). International Review of Victimology, 19, 1, 71-83.

• Braithwaite, J. (1999). Crime shame and reintegration. Cambridge: Cambridge University Press.

• Cameron, L. & Thorsborne, M. (2001). Restorative justice and school discipline: mutually exclusive? In H. Strang & J. Braithwaite (eds.), Restorative justice and civil society (pp. 180-194). Cambridge: Cambridge University Press.

• Davis, F. (2014). Discipline with dignity: Oakland classrooms try healing instead of punishment. Reclaiming Children and Youth, 23, 38-41.

• Department of Education. (2018). 2015-16 civil rights data collection: school climate and safety. Retrieved from https://www2.ed.gov/about/offices/list/ocr/docs/school-climate-and-safety.pdf (last accessed 21 July 2021).

• Dinkes, E., Forrest-Cataldi, E. & Lin-Kelly, W. (2007). Indicators of school crime and safety: 2007. (NCES 2007-NCES 2008) U.S. Department of Education and Justice. Washington, DC: Government Printing Office.

• Epstein, R., Blake, J. & Gonzalez, T. (2017). Girl interrupted: the erasure of black girls’ childhood. Center for Poverty and Inequality: Georgetown Law. Retrieved from www.law.georgetown.edu/poverty-inequality-center/wp-content/uploads/sites/14/2017/08/girlhood-interrupted.pdf (last accessed 7 May 2021).

• Fabelo, T., Thompson, M.D., Plotkin, M., Carmichael, D., Marchbanks, M.P. & Booth, E.A. (2011). Breaking school rules: a statewide study of how school discipline relates to students’ success and juvenile justice involvement. New York: Council of State Governments Justice Center.

• Feld, B. (2017). The evolution of the juvenile court: race politics and the criminalization of juvenile justice. New York: NYU Press.

• Freiburger, T.L. & Jordan K.L. (2016). Race and ethnicity in the juvenile justice system. Durham: Carolina Academic Press.

• Fronius, T., Persson, H., Guckenburg, S., Hurley, N. & Petrosino, A. (2016). Restorative justice in U.S. schools: a research review. San Francisco: WestEd.

• Fronius, T., Persson, H., Guckenburg, S., Hurley, N. & Petrosino, A. (2019). Restorative justice in U.S. schools: an updated research review. San Francisco: WestEd.

• Gerber, J. & Englehardt-Greer, S. (1996). Just and painful: attitudes toward sentencing criminals. In T. Flanagan& D. Longmire (eds.), Americans view crime and justice: a national public opinion survey (pp. 62-74). Thousand Oaks: Sage.

• Gerkin, P. (2009). Participation in victim-offender mediations: lessons learned from observations. Criminal Justice Review, 34, 226-247.

• González, T. (2012). Keeping kids in schools: restorative justice, punitive discipline, and the school to prison pipeline. Journal of Law and Education, 41(2), 281-335.

• González, T. (2015). Socializing schools: addressing racial disparities in discipline through restorative justice. In D. Losen (ed.), Closing the school discipline gap: equitable remedies for excessive exclusion (pp. 151-165). New York: Teachers College Press.

• González, T. (2016). Restorative justice from the margins to the center: the emergence of the new norm in school discipline. Howard Law Journal, 60(1), 267-308.

• González, T., Sattler, H. & Buth, A. (2019). New directions in whole-school restorative justice implementation. Conflict Resolution Quarterly, 36, 207-220.

• Gregory, A., Bell, J. & Pollock, M. (2014). How educators can eradicate disparities in school discipline: a briefing paper on school based interventions. Bloomington: Discipline Disparities Series, The Equity Project at Indiana University. Retrieved from www.issuelab.org/resources/17517/17517.pdf (last accessed 7 May 2021).

• Guckenberg, S., Hurley, N., Persson, H., Fronius, T. & Petrosino, A. (2015). Restorative justice in U.S. schools: summary findings from interviews with experts. San Francisco: West Ed. Retrieved from https://jprc.wested.org/wp-content/uploads/2015/11/​1447101213resourcerestorativejusticeinusschoolssummaryfindingsfrominterviews​withexperts.pdf (last accessed 21 July 2021).

• Gun Free Schools Act, 20 U.S.C. § 7961 (1994).

• Hagan, J. (2010). Who are the criminals? The politics of crime policy from the age of Roosevelt to the age of Reagan. Princeton: Princeton University Press.

• Health and Human Development Program. (2012). Workbook for improving school climate & closing the achievement gap, 2nd edition: using your California healthy kids and California school climate surveys. Sacramento: California Department of Education.

• Hollweck, T., Reimer, K. & Bouchard, K. (2019). A missing piece: embedding restorative justice and relational redagogy into the teacher education classroom. The New Educator, 15(3), 246-267. doi: 10.1080/1547688X.2019.1626678.

• Hurley, N., Guckenberg, S., Persson, H., Fronius, T. & Petrosino, A. (2015). What further research is needed on restorative justice in schools? Retrieved from https://jprc.wested.org/wp-content/uploads/2015/07/Restorative_Justice_Future_​Research.pdf (last accessed 7 May 2021).

• Jacobsen, W. (2020). School punishment and interpersonal exclusion: rejection, withdrawal, and separation from friends. Criminology, 58(1), 35-69. doi: 10.1111/1745-9125.12227.

• Jain, S., Bassey, H., Brown, M. & Karla, P. (2014). Restorative justice in Oakland schools implementation and impact: an effective strategy to reduce racially disproportionate discipline, suspensions and improve academic outcomes. Oakland: Data in Action. Retrieved from https://www.ousd.org/cms/lib07/CA01001176/Centricity/Domain/134/OUSD-RJ%20Report%20revised%20Final.pdf (last accessed 21 July 2021).

• Jones, C. (2016). Is disproportionate minority contact improving? an exploratory analysis of the arrest, confinement, and transfer stages in rural and urban settings. Journal of Ethnicity in Criminal Justice, 14(1), 40-57. doi: 10.1080/15377938.2015.1052172.

• Karp, D. & Breslin, B. (2001). Restorative justice in school communities. Youth and Society, 33(249), 254-255. doi: 10.1177/0044118X01033002006.

• Knight, D. & Wadhwa, A. (2014). Expanding opportunity through critical restorative justice: portraits of resilience at the individual and school level. Schools: Studies in Education, 11(1), 14-16. doi: 10.1086/675745.

• Krezmien, M., Leone, P. & Achilles, G. (2006). Suspensions, race, and disability: analysis of statewide practices and reporting. Journal of Emotional & Behavioral Disorders, 14(4), 217-226. doi: 10.1177/10634266060140040501.

• Kurki, L. (2000). Restorative and community justice in the United States. In M. Tonry (ed.), Crime and justice (pp. 235-303, Vol. 27). Chicago: University of Chicago.

• Lawrence, R. (1998). School crime and juvenile justice. New York: Oxford University Press.

• Lemert, E. (1951). Social pathology: a systematic approach to the theory of sociopathic behavior. New York: McGraw-Hill.

• Lewis, S. (2009). Improving school climate: findings from schools implementing restorative practices. Bethlehem, PA: International Institute for Restorative Practices.

• Llewellyn, K. & Parker, C. (2018). Asking the ‘who’: a restorative purpose for education based on relational pedagogy and conflict dialogue. The International Journal of Restorative Justice, 1(3), 399-412. doi: 10.5553/IJRJ/258908912018001003005.

• Losen, D. (2014). Closing the school discipline gap: equitable remedies for excessive exclusion (disability, equity, and culture). New York: Teachers College Press.

• Losen, D. & Skiba, R. (2010). Suspended education: urban middle schools in crises. Retrieved from https://www.ousd.org/cms/lib07/CA01001176/Centricity/Domain/134/OUSD-RJ%20Report%20revised%20Final.pdf (last accessed 21 July 2021).

• Lustick, H. (2017). “Restorative justice” or restorative order? Discipline practices in urban public schools. Urban Education, 1-28. doi: 10.1177/0042085917741725.

• Madfis, E. (2016). ‘It’s better to overreact’: school officials’ dear and perceived risk of rampage attacks and the criminalization of American public schools. Critical Criminology, 24, 39-55. doi: 10.1007/s10612-015-9297-0.

• Marshall, T.F. (1996). The evolution of restorative justice in Britain. European Journal on Criminal Policy and Research, 4(4), 31-43.

• Mauer, M. (2006). Race to incarcerate. New York: The New Press.

• McGarrell, E., Olivares, K., Crawford, K. & Kroovand, N. (2000). Returning justice to the community: the Indianapolis juvenile restorative justice experiment. Retrieved from https://www.hudson.org/content/researchattachments/attachment/253/indianapolis_juvenile_restorative_justice_experiment.pdf (last accessed 21 July 2021).

• McKenna, J.M. & White, S.R. (2018). Examining the use of police in schools: how roles may impact responses to student misconduct. American Journal of Criminal Justice, 43, 448-470.

• McMorris, B.J., Beckman, K.J., Shea, G., Baumgartner, J. & Eggert, R.C. (2013). Applying restorative practices to Minneapolis public schools students recommended for possible expulsion: a pilot program evaluation of the family and youth restorative conference program. Minneapolis: University of Minnesota. Retrieved from www.legalrightscenter.org/uploads/2/5/7/3/25735760/lrc_exec_summ-final.pdf (last accessed 7 May 21).

• Mirsky, L. (2007). Safer saner schools: transforming school cultures with restorative practices. Reclaiming Children and Youth: The Journal of Strength Based Interventions, 16(2), 5-12.

• Mirsky, L. & Wachtel, T. (2007). “The worst school I’ve ever been to”: empirical evaluations of a restorative school and treatment milieu. Reclaiming Children and Youth: The Journal of Strength Based Interventions, 16(2), 13-16.

• Mongold, J. & Edwards, B. (2014). Reintegrative shaming: theory into practice. Journal of Theoretical and Philosophical Criminology, 6, 205-212.

• Monterastelli, S. (2017). Every hand’s a loser: the intersection of zero-tolerance policies, mental illness in children and adolescents, and the juvenile justice system. Law & Psychology Review, 41, 209-228.

• Morrison, B., Blood, P. & Thorsborne, M. (2005). Practicing restorative justice in school communities: the challenge of culture change. Public Organization Review, 5, 335-357. doi: 10.1007/s11115-005-5095-6.

• Morrison, B. & Vaandering, D. (2012). Restorative justice: pedagogy, praxis, and discipline. Journal of School Violence, 11(2), 138-155. doi: 10.1080/15388220.2011.653322.

• Nardi, P. (2006). Doing survey research: a guide to quantitative methods (2nd ed.). Boston: Pearson.

• Nussbaum, L. (2018). Realizing restorative justice: legal rules and standards for school discipline reform. Hastings Law Journal, 69, 583-646. doi: 10.2139/ssrn.3039752.

• Payne, A. & Welch, K. (2010). Modeling the effects of racial threat on punitive and restorative school discipline practices. Criminology, 48(4), 1019-1062. doi: 10.1111/j.1745-9125.2010.00211.x.

• Payne, A. & Welch, K. (2015). Restorative justice in schools: the influence of race on restorative discipline. Youth & Society, 47(4), 539-564. doi: 10.1177/0044118X12473125.

• Riestenberg, N. (2001). In-school behavior intervention grants: a three-year evaluation of alternative approaches to suspensions and expulsions. Roseville: Minnesota Dept. of Children, Families & Learning.

• Riestenberg, N. (2012). Circle in the square: building community and repairing harm in schools. St. Paul: Living Justice Press.

• Rocque, M. (2010). Office discipline and student behavior: does race matter? American Journal of Education, 116, 557-581. doi: 10.1086/653629.

• Rocque, M. & Snellings, Q. (2018). The new disciplinology: research, theory, and remaining puzzles on the school-to-prison pipeline. Journal of Criminal Justice, 59, 3-11. doi: 10.1016/j.jcrimjus.2017.05.002.

• Schumacher, A. (2014). Talking circles for adolescent girls in an urban high school: A restorative practices program for building friendships and developing emotional literacy skills. Sage Open, 1-13. Doi: 10.1177/2158244014554204.

• Simon, J. (2007). Governing through crime: how the war on crime transformed American democracy and created a culture of fear. New York: Oxford University Press.

• Skiba, R., Chung, C., Trachock, M., Baker, T., Sheya, A. & Hughes, R. (2014). Parsing disciplinary disproportionality contributions of infraction, student, and school characteristics to out-of-school suspension and expulsion. American Educational Research Journal, 51, 640-670. doi: 10.3102/0002831214541670.

• Skiba, R., Michael, R., Nardo, A. & Peterson, R. (2002). The color of discipline: sources of racial and gender disproportionality in school punishment. The Urban Review, 34(4), 317-342. doi: 10.1023/A:1021320817372.

• Skiba, R., Shure, L. & Williams, N. (2011). What do we know about racial and ethnic disproportionality in school suspension and expulsion. Briefing paper developed for the Atlantic Philanthropies’ race and gender research-to-practice collaborative, 1-34. Bloomington: The Equity Project at Indiana University.

• Stinchcomb, J., Bazemore, G. & Riestenberg, N. (2006). Beyond zero tolerance: restoring justice in secondary schools. Youth Violence and Juvenile Justice, 4(2), 123-147. doi: 10.1177/1541204006286287.

• Sullivan, D. & Tifft, L. (2001). Restorative justice: healing the foundations of our everyday lives. Monsey: Willow Tree Press

• Sumner, D., Silverman, C. & Frampton, M. (2010). School-based restorative justice as an alternative to zero-tolerance policies: lessons from west Oakland. Berkley: University of California, Berkeley, School of Law.

• Suvall, C. (2009). Restorative justice in schools: learning from Jena high school. Harvard Civil Rights-Civil Liberties Law Review, 44, 547-569.

• Sykes, B., Piquero, A., Gioviano, J. & Pittman, N. (2015). The school-to-prison pipeline in America, 1972-2012. Oxford Handbooks Online. Retrieved from www.oxfordhandbooks.com/view/10.1093/oxfordhb/9780199935383.001.0001/oxfordhb-9780199935383-e-110 (last accessed 7 May 2021).

• Tolefree, T. (2017). A comprehensive approach for implementing school-based restorative practices, restorative practices pilot site: upper Midwest elementary school pre-k-5. Minneapolis: Koinonia Leadership Academy. Retrieved from https://genderpolicyreport.umn.edu/wp-content/uploads/2018/01/KLA-RP-Report-1.15.18-2.pdf (last accessed 7 May 2021).

• Tyler, T. (2006). Why people obey the law. Princeton: Princeton University Press.

• United States Department of Justice. (1998). Guide for implementing the balanced and restorative justice model. Retrieved from www.ncjrs.gov/pdffiles/167887.pdf (last accessed 7 May 2021).

• Vaandering, D. (2014). Implementing restorative justice in schools: what pedagogy reveals. Journal of Peace Education, 11(1), 64-80. doi: 10.1080/17400201.2013.794335.

• Voight, A., Austin, G. & Hanson, T., (2013). A climate for academic success: how school climate distinguishes schools that are beating the achievement odds. San Francisco: West Ed.

• Walgrave, L. (2011). Investigating the potential of restorative justice practice. Washington University Journal of Law & Policy, 36, 91-139.

• Watts, B.R. & Robertson, K. (2019). Evaluation of the use of restorative practices to reduce school truancy and suspensions. Kalamazoo: Western Michigan University Evaluation Center. Retrieved from https://courts.michigan.gov/Administration/SCAO/OfficesPrograms/ODR/Documents/MI-SCAO%20Report%20Final%2002212019.pdf (last accessed 21 July 2021).

• Welch, K. (2018). The effect of minority threat on risk management and the ‘new disciplinology’ in schools. Journal of Criminal Justice, 59, 12-17. doi: 10.1016/j.jcrimjus.2017.05.006.

• Wilson, H. (2014). Turning off the school-to-prison pipeline. Reclaiming Children and Youth, 23(1), 49-53.

• Wood, W. (2020). The indecent demands of accountability: trauma, marginalization, and moral agency in youth restorative conferencing. The International Journal of Restorative Justice, 32(2), 168-193. doi: 10.5553/IJRJ.000038.

• Zehr, H. (1990). Changing lenses: a new focus for criminal justice. Scottsdale: Herald.

• Zehr, H. & Mika, H. (1998). Fundamental concepts of restorative justice. Contemporary Justice Review, 1(1), 47-56.

• Zehr, H. (2002). The little book of restorative justice. Intercourse: Good Books.

Appendix
Appendix A Participant Satisfaction Items and Factor Loadings

	Mean/%	SD	Min.	Max.
Dependent Variable
Participant Satisfaction - 13-item scale (α= .91)	4.22	0.59	1.38	5.00
Factor 1:
Staff explained how a restorative circle works (.71)	4.49	0.64	1.00	5.00
I feel that I was treated fairly in the restorative circle (.65)	4.34	0.76	1.00	5.00
I feel this program helped me feel better (.65)	4.13	0.92	1.00	5.00
The restorative circle helped me understand how my actions affected others (.64)	4.28	0.78	1.00	5.00
Overall, I am happy with the outcome of my restorative circle (.66)	4.21	0.83	1.00	5.00
I was given the opportunity to ask for what I wanted in the restorative circle (.74)	4.35	0.71	1.00	5.00
If I had another conflict, I would choose to participate in the restorative circle again (.61)	4.22	0.91	1.00	5.00
I have learned new skills that will help me avoid harmful situations in the future (.68)	4.21	0.85	1.00	5.00
I feel that I was able to explain my side of the story (.73)	4.45	0.71	1.00	5.00
Factor 2:
I feel the other person took responsibility for their behaviour (.64)	4.04	0.95	1.00	5.00
I feel the other person listened to me during the restorative circle (.53)	4.16	0.90	1.00	5.00
The other person apologized for their actions (.83)	3.95	1.13	1.00	5.00
I have apologized for my actions (.76)	4.10	1.07	1.00	5.00

Factor loadings noted in parentheses.

Noten

• 1 At minimum, restorative justice practices employ a narrow definition of accountability inclusive of an offender admitting blame to a victim, and making amends to the victim as far as possible. For a nuanced discussion of accountability in relation to youth restorative conferencing, see Wood (2020).

• 2 Facilitators successfully completed the International Institute for Restorative Practices (IIRP) two-day training programme. For further information regarding IIRP facilitator training please visit iirp.edu/professional-development/restorative-justice-conferencing.

• 3 Administrative/facilitator data was collected from restorative circle participants who chose not to participate in the restorative circle survey (n = 429). Therefore, comparison of these individuals with those who are included in the survey data (n = 1313) is possible. A significantly greater proportion of restorative justice participants who did not complete the survey took on the role of witness (20.7 per cent), and a significantly smaller proportion of those not included in the survey data took on the role of victim or aggressor (34.8 per cent) compared with those included in the survey (7.2 per cent witnesses and 52.6 per cent victims/aggressors). Additionally, a greater proportion of those not included in the survey data fulfilled the role of ‘other’ compared with those who complete the survey (10.7 per cent compared with 4.3 per cent, respectively). Those included in the survey participated in a significantly greater number of restorative circles (mean = 2.04) compared with those who did not complete the survey (mean = 1.70), and those included in the survey include a significantly higher proportion of repeat participants (61.9 per cent) compared with those who did not participate in the survey (55.6 per cent). Survey participants include a greater proportion of Hispanic individuals (32.8 per cent) compared with those who did not participate (24.3 per cent). Additionally, those who did not participate are significantly older by an average of 0.6 years and are enrolled in 1.5 grade levels higher on average. A significantly greater number of incidents involving staff/student conflicts and classroom disruptions took place among those who chose not to participate in the survey. However, a significantly greater number of physical altercations, incidents involving threats and/or intimidation, cyber conflicts and incidents involving gossip and/or rumour spreading occurred among those who participated in the survey compared with those who did not. Those included in the survey are significantly more likely to have reached an agreement, spent significantly more time in the restorative circle and saved significantly more total suspension days compared with those who did not participate.

• 4 All respondents were provided with the same survey instrument (Appendix A). Students were advised to skip questions that did not apply to their role within the dispute (i.e. victims would not be expected to take responsibility for their actions). Surveys were subsequently cross-checked during data input.

• 5 Grade was used in place of participants’ age since less missing data exists for grade compared with age and since they are highly correlated (r = 0.90, p < 0.01).

• 6 Categories for role were developed on the basis of a pre-test of the data collection method.

• 7 For one participant, the order of participation in restorative circles could not be determined owing to a missing case number. Therefore, this participant’s circles were coded using the sequence, ‘1’, ‘4’, ‘4’, ‘4’, ‘5’, ‘6’.

• 8 Incident types with inadequate sample size to estimate these effects are excluded from this stage of the analysis (i.e. cyber conflict (n = 53), taunts/teasing (n = 65), gossip/rumours (n = 65), classroom disruption (n = 72)).

• 9 Within the restorative practices circles categorised as ‘friendship issues’ no individuals were categorised as victims.