Research Paper: Restorative Justice

CRIMINAL JUSTICE AND BEHAVIOR, 201X, Vol. XX, No. X, Month 2021, 1 –18.

DOI: https://doi.org/10.1177/00938548211008488

Article reuse guidelines: sagepub.com/journals-permissions

© 2021 International Association for Correctional and Forensic Psychology

1

Community ServiCe outComeS in JuStiCe-involved youth

Comparing restorative Community Service to Standard Community Service

ABERE SAwAqDEH CHURCH Washington State University Orlando VA Healthcare System

DAVID K. MARCUS Washington State University

ZACHARy K. HAMILTON University of Nebraska Omaha

Traditional mandated community service (CS) typically consists of picking up trash or performing manual labor, distanced from the community. Some juvenile justice programs have begun to implement restorative community service (RCS) pro- grams that enable youth to complete meaningful CS projects in a shame-free manner alongside community members. This study compared RCS with a standard community service (SCS) program in two counties in washington State on psychosocial outcomes, including attitudes, peer relationships, school conduct, academic performance, and substance use. Recidivism was also examined. RCS was associated with reduced substance use and fewer school conduct difficulties compared with SCS, and also positively influenced peer relationships and attitudes. These findings suggest that adding a restorative component to CS may improve psychosocial outcomes for justice-involved youth, but there was no evidence that adding a restorative component to CS led to reduced recidivism. Additional systematic studies are needed to determine whether these findings replicate.

Keywords: community service; restorative justice; juvenile justice; probation; propensity score matching

AuthorS’ note: This article is based on a dissertation submitted by the first author, under supervision of the second author, in partial fulfillment of the requirements for a doctoral degree in clinical psychology. The authors would like to thank Clark County Juvenile Court, Kitsap County Juvenile and Family Court Services, and the Washington State Center for Court Research for their time and effort dedicated to this study. Abere Sawaqdeh Church is now at the Orlando VA Healthcare System. The views expressed in this article are those of the authors and do not necessarily reflect the position or policy of the Department of Veterans Affairs or the U.S. government. Correspondence concerning this article should be addressed to Abere Sawaqdeh Church, Orlando VA Healthcare System, 13800 Veterans Way, Orlando, FL 32827, USA; e-mail: abere.sawaqdeh@ gmail.com.

1008488CJBXXX10.1177/00938548211008488Criminal Justice and BehaviorChurch et al. / Psychosocial outcomes of Community Service research-article2021

2 CRIMINAL JUSTICE AND BEHAVIOR

Community service (CS) is one of the most common sentences for low-level offenses in the United States (wood, 2012), including for justice-involved youth (referred to as youth throughout the article). Initially implemented in the late 1960s, more than 95% of county courts use CS as an alternative to incarceration, either as a standalone or additive sentence to probation and other alternative sanctions (Bazemore & Maloney, 1994; McIvor, 2016). Standard community service (SCS) programs require youth to engage in unpaid menial work (e.g., picking up trash) either in isolation or with other youth (Bouffard & Muftić, 2006). Individuals sentenced to SCS are often identified by orange vests or work crews. Restorative community service (RCS) programs attempt to transform the use of CS in juvenile populations by providing more meaningful work experiences and emphasizing greater community involvement (wood, 2012). Beginning in 2001, the Clark County Juvenile Court (CCJC) in washington State implemented one of the first RCS programs. The court determined that work crews led to minimal benefits and sometimes created prob- lems (e.g., conflicts between youth in work crews), and the court moved to a fully restor- ative model.

RCS is based on the principles of restorative justice (RJ). In addition to addressing the needs of the victims and communities, and providing opportunities for making reparations, RJ interventions attempt to foster growth and reintegration into the community for justice- involved individuals (Ryals, 2004; Strang & Braithwaite, 2017). Because youth are still forming their identities and developing socially and cognitively, they are presumed to be responsive to interventions that encourage the development of prosocial identities, which has been a rationale for implementing restorative programs in juvenile courts (Suzuki & wood, 2018). A 3-year qualitative study that assessed the implementation of RCS in Clark County (wood, 2012) described three primary ways that RCS differed from SCS: engage- ment in “real-work” (wood, 2012, p. 19), defined as service with tangible benefits to the community, frequently involving skill-based work; required interaction with mentors and community members; and preservice orientation on the restorative approach. RCS is viewed as the forum for youth to learn about the RJ approach and take action to contribute and integrate back into their community. Other restorative programs typically included in youth’s case plans (e.g., Victim–offender mediation, Making Things Right Program) com- plement RCS by providing opportunities for those most directly affected by the crime to be involved in the restorative process. For instance, victim–offender mediation enables victims and youth to discuss the crime and its consequences to promote healing for victims and awareness for youth (Hansen & Umbreit, 2018).

One of the primary functions of juvenile sanctions is to prevent future criminality. This aim is often accomplished by influencing intermediate outcomes associated with criminal- ity, including attitudes, peer relationships, school conduct, academic performance, and sub- stance use (Elonheimo et al., 2009; Murray & Farrington, 2010; Tanner-Smith et al., 2013). As Barnoski (2004) noted, “one of the most important and consistently identified factors linked to criminal behavior is antisocial attitudes, values, and beliefs” (p. 103). In addition, almost 50% of justice-involved youth suffer from a substance use disorder (Chassin, 2008), and substance use is associated with higher rates of academic problems (Bugbee et al., 2019). Academic failure often precedes delinquency (Murray & Farrington, 2010), whereas increased school attendance, positive academic values, and school activity involvement are associated with decreased contact with the justice system (Farrington et al., 2012). These psychosocial variables also influence other life domains, such as income and occupational

Church et al. / PSyCHOSOCIAL OUTCOMES OF COMMUNITy SERVICE 3

success (Spengler et al., 2018). Delinquent peer relationships and negative attitudes are associated with mental illness (Sha, 2006), and substance use is associated with sexual assault, injury, and death (Hawkins et al., 1992).

Psychosocial outcomes are assessed for all youth sentenced to probation in washington State to facilitate outcome studies and predict recidivism. The Positive Achievement Change Tool (PACT; Barnoski, 2004) is a risk assessment tool that includes measures of attitudes, peer relationships, school conduct, academic performance, and substance use. These vari- ables are considered dynamic risk or protective factors of recidivism and were treated as psychosocial outcomes in this study. The primary goal of this study was to determine whether RCS (Clark County, wA) led to greater improvement in psychosocial outcomes compared with SCS (Kitsap County, wA). Evaluating RCS extends the RJ literature by assessing whether this action phase of RJ contributes to youth outcomes.

rCS And PSyChoSoCiAl outComeS

youth attitudes toward themselves and others are likely to change as a result of demon- strating competence, value, and reliability during meaningful service projects (Bazemore & Stinchcomb, 2004). Interactionist researchers suggest that identity development can be facilitated by “amend-making activity” (Bazemore & Karp, 2004, p. 19) that enables youth to empathize with others and understand how their actions can benefit their community. Unlike SCS, RCS requires youth to interact directly with community members and may provide more opportunities to enhance attitudes. Bazemore and Schiff (2001/2015) indi- cated that the reintegration of youth into the community in a shame-free manner can lead youth to experience cognitive changes and view themselves as productive citizens. Therefore, our first hypothesis is that RCS will lead to the development of more positive attitudes (e.g., empathy, responsibility) than SCS.

whereas RCS programs require youth to work alongside community members and men- tors, SCS programs typically have youth complete CS solitarily or with other justice- involved youth (Morris & Tonry, 1991). According to Gifford-Smith and colleagues (2005), a common element across ineffective juvenile interventions “is the aggregation of deviant youth” (p. 262). Group interventions for antisocial youth may have iatrogenic effects because youth reinforce each other’s problematic behaviors and beliefs during the interven- tion (i.e., peer deviancy training; Dishion et al., 1999). Such harmful effects may, however, only develop under certain conditions (e.g., no parental involvement, minimal supervision; Dishion & Tipsord, 2011), which may be more common in SCS programs. On the contrary, youth engaging in RCS may be more likely to form positive peer relationships by interact- ing with community members and mentors who are present during RCS projects (Flanagan et al., 2015). Therefore, Hypothesis 2 is that compared with SCS, RCS will result in more positive peer relationships.

CS can enable youth to develop improved self-control (Bouffard & Muftić, 2007; Gelsthrope & Rex, 2004), which has a direct influence on academic performance and con- duct (Job et al., 2015). Once youth have demonstrated the ability to complete CS in a con- strained environment with high external motivations (e.g., avoiding detention), these experiences may provide youth with evidence that they can control their maladaptive behaviors (i.e., improve self-efficacy) in other environments, such as at school (Caldwell & Smith, 2006). Enhancing self-efficacy also increases resistance to problem behaviors,

4 CRIMINAL JUSTICE AND BEHAVIOR

including school conduct issues such as suspensions/expulsions (Valdebenito et al., 2019). RCS may provide more opportunities than SCS to develop self-control and self-efficacy during more meaningful service work, which frequently requires more focus and skill- building than the menial work commonly completed during SCS. School conduct outcomes may also be positively influenced by interactions with mentors and other community vol- unteers. Positive social support has been viewed as a form of informal social control of youth, which can prevent delinquency (Bazemore & McLeod, 2012; Chouhy et al., 2020).

Completing meaningful service work and increasing positive social support likely also enhances academic performance by allowing youth to gain the confidence and support needed to succeed academically (Mackinnon, 2012). Although there is no research on the academic benefits of CS in justice-involved youth, there is evidence that CS performed by nonoffending youth is associated with increased academic achievement, particularly if the CS experience promotes autonomy and social relatedness (e.g., Astin & Sax, 1998), which is emphasized in RCS programs. Therefore, Hypotheses 3 and 4 are that, compared with SCS, RCS will result in improved school conduct and academic performance.

Kristjansson et al. (2010) found that peer use and peer approval of substance abuse were the second and third most important predictors of substance use in adolescents. RCS pro- vides more opportunities than SCS to develop positive social support, which may influence substance use outcomes. CS that facilitates positive interpersonal or mentoring experiences is associated with decreased substance use (Geller, 2011; Taylor et al., 1999). A randomized study of at-risk middle school students found that adding a mentoring component to CS was associated with lower levels of substance use when compared with standalone CS (Taylor et al., 1999). RCS may also enhance self-efficacy more than SCS through meaningful ser- vice projects, which is also associated with lower levels of substance use (Kadden & Litt, 2011). Therefore, Hypothesis 5 is that, compared with SCS, RCS will result in a greater decrease in substance use.

Although not specific to CS, RJ approaches, compared with traditional justice pro- grams, are associated with reduced recidivism, increased restitution compliance, and improved satisfaction with the legal process (see Latimer et al., 2005 for a meta-analysis). Early studies also suggested that nonmenial service work, coupled with positive commu- nity interactions, may also be associated with lower recidivism rates (McIvor, 2016). Furthermore, changes in attitudes, peer relationships, school conduct, academics perfor- mance, and substance use are all associated with reduced offending (Barnoski, 2004; Chassin, 2008; Elonheimo et al., 2009; Murray & Farrington, 2010; Tanner-Smith et al., 2013). Because evidence suggests that RJ interventions more positively influence these intermediate outcomes, Hypothesis 6 is that RCS would also result in lower recidivism than SCS (Hypothesis 6).

thiS Study

Although there is promising research and theoretical support for the proposed benefits of RCS, methodological limitations (e.g., lack of appropriate comparison groups) and limited research on CS outcomes in justice-involved youth limit confidence in these findings. Thus, it is still unknown whether reintegrative aspects of RCS (e.g., restorative training, meaning- ful service work, mentorship, and nonstigmatizing community interactions) are associated with more positive outcomes compared with SCS. The primary goal of the study was to determine whether different approaches to CS are associated with different psychosocial

Church et al. / PSyCHOSOCIAL OUTCOMES OF COMMUNITy SERVICE 5

outcomes. Specifically, the study compared the Clark County RCS program with the Kitsap County SCS program. Propensity score matching (PSM) procedures were used to account for observed covariates that influence outcomes or predict selection into a program. A com- parison group from a neighboring county (Kitsap County) was selected to control for unob- served factors that could influence a youth’s participation in RCS. Using a comparison group from a court that offered SCS as a sanction, enabled group assignment based on location and implementation strategies rather than unsuitability. Although Clark County is larger than Kitsap County (474,643 vs. 266,414; Data USA, 2019), the two counties are comparable in terms of demographics and economics. The counties have similar median household incomes (US$73,026 in Kitsap, US$74,747 in Clark), poverty rates (9.88% in Kitsap, 10.3% in Clark), median ages (39.1 years in Kitsap, 38.6 years in Clark), median property values (US$326,200 in Kitsap, US$329,200 in Clark), and racial demographics (76.6% white in Kitsap, 78.3% white in Clark).

Attitudes, peer relationships, school conduct, academic performance, and substance use (the dependent variables) were analyzed in two ways. First, the study examined whether psychosocial outcomes improved after completing either CS sanction. Second, to deter- mine whether RCS led to more positive outcomes compared with SCS, the outcomes of each court were compared. Recidivism was assessed at the end of probation from court records and compared by CS program type. Throughout these analyses, length of proba- tion was included as a covariate because the longer the probation, the more likely out- comes are subject to change (watts, 2016). Probation periods are also commonly influenced by outcome variables in this study, such as probation violations for school conduct issues or recidivism (Phelps, 2018). Also, in the final matched sample, the SCS group had signifi- cantly longer probation periods. Propensity scores were also included as a covariate in all the outcome analyses to account for residual covariate imbalance between RCS and SCS (Harder et al., 2010).

method

PArtiCiPAntS And ProCedure

Data were available on youth assigned to RCS in Clark County, or SCS in Kitsap County, as a requirement of probation for an offense that the youth had admitted guilt for between January 1, 2004, and December 31, 2017. For both courts, all youth were eligible to be assigned CS if they were sentenced to probation. In Clark County, approximately 80% to 90% of youth were assigned RCS as part of their court order. If RCS was not assigned as part of a requirement of probation, it was likely because the youth was assigned other com- parable restorative sanctions offered by the court. In Kitsap County, almost 98% of youth completed CS as a requirement of probation and, according to the court, if this option was not granted, it was likely due to the youth being very close to the age of 18 years at the time of disposition. youth on probation in Clark County were typically assigned 8 to 24 hr of RCS for all offenses, whereas youth in Kitsap County could be assigned up to 150 hr of CS for each offense. Clark County only provided data on youth who completed at least one RCS project, but neither court identified the exact number of CS hours completed by each youth or which youth in the sample participated in each type of CS work. Therefore, these variables were not controlled for in this study.

The original sample consisted of 1,719 youth from Clark County and 684 from Kitsap County. Prior to the matching procedures, 149 youth from Clark County and 25 youth from

6 CRIMINAL JUSTICE AND BEHAVIOR

Kitsap County were excluded because they only completed a baseline PACT assessment. There were also missing data from 337 (15%) youth on the academic outcome variable. It is likely that these data were missing because the PACT administrators were required to confirm academic progress with outside sources (e.g., parents, teachers), which was not always possible (Barnoski, 2004). These youth were excluded from academic outcome analyses but retained for the other analyses. The sample prior to the matching procedures included 2,229 youth. Following matching (see the following), the final sample was 1,669 (RCS n = 1,043, SCS n = 626).

PACT and recidivism data from each county court were provided by the Administrative Office of the Courts. All study procedures were approved by the washington State University institutional review board. youth who were assigned to CS, completed their probation,1 and received a final PACT assessment were included in the analyses, regardless of how many CS hours they completed. This method was used to avoid biasing the results toward com- pleters of CS. For example, if youth disproportionately dropped out of RCS and those who stayed in RCS improved the most (perhaps because of greater perseverance), a completer analysis would misleadingly suggest that RCS was the more effective approach. Thus, an intention-to-treat (ITT) analysis, which avoids confounding participant motivation with intervention efficacy, is better suited for informing policy decisions about which form of CS to adopt.

rCS verSuS SCS

RCS and SCS differed in many ways. RCS included a preservice orientation on the restorative approach for youth and the community member in charge of each service proj- ect. This training involved an introduction to the RJ approach, as well as a discussion regarding the harm caused by the offense, the importance of accountability, and the value of the service work. During the RCS projects, other community members were encouraged to have conversations with youth about making things right but to refrain from having detailed conversations about the youth’s specific offenses to facilitate shame-free reintegration. RCS was viewed as an opportunity for the youth to learn the value behind contributing and to feel more connected to their community. SCS did not require preservice training.

In contrast to SCS, RCS projects required youth to engage in meaningful service projects. Examples of RCS projects included Habitat for Humanity building projects, working in home- less shelters, growing food with 4-H for local food banks, and building play structures in public housing projects. youth assigned to RCS also interacted with at least one mentor at each RCS project site, who provided them with support throughout their service experiences. Mentors were typically college student volunteers who were trained in the restorative approach and did not evaluate or supervise youth, which promoted the development of nonhierarchical relationships based on shared interests or experiences. SCS did not require meaningful service work or mentorship (see Table 1 for a further comparison of RCS and SCS).

outCome meASureS

PACt

The PACT provided the primary source of outcome data. The PACT is an evidence-based risk assessment designed for youth aged from 12 to 18 years who are sentenced to proba- tion. Approximately, 10,000 to 15,000 youths are assessed annually. The PACT includes a

Church et al. / PSyCHOSOCIAL OUTCOMES OF COMMUNITy SERVICE 7

prescreen and a full assessment (Asscher et al., 2015). The prescreen is a shortened version of the full assessment that provides a risk assessment status of low, moderate, or high based on the youth’s criminal and social history, which are two of the most important predictors of recidivism. Risk assessment status and criminal history score were taken from the PACT prescreen assessment and used as pretreatment covariates for the PSM procedures. The criminal history score was composed of items assessing the age of onset of criminal behav- ior and the type of current offense, as well as frequency and type of prior offenses.

The full PACT assessment required 1 to 3 hr to complete (Van der Put et al., 2014). The initial full assessment (i.e., baseline assessment) was completed as a semi-structured inter- view. Follow-up assessments were administered using computer software. Moderate-high and high-risk youth were reassessed every 90 days, whereas low or moderate risk youth were reassessed every 180 days.2 This study used the baseline and final assessment. The baseline assessment was based on the previous 6 months for each youth, and the final assessment was based on the previous 4 weeks. The following outcome measures from the PACT were included as dependent measures in the study: Attitudes (nine items on opti- mism, impulsivity, self-control, empathy, respect for the property of others, respect for authority figures, attitude toward antisocial behavior, attitude toward prosocial rules in

Table 1: RCS in Clark County Versus SCS in Kitsap County

CS component RCS in Clark county SCS in Kitsap county

Framework Preservice training on restorative approach of CS required.

No training on restorative approach of CS

CS is framed as a personal obligation and a way to reintegrate the youth back into the community.

CS is framed as a punishment.

Service work RCS youth participate in nonmenial “real- work.”

It is not required that CS involve in “real- work” and the work is often menial.

Participating organizations must demonstrate that the CS provides tangible benefits to the community.

Participating organizations do not have to demonstrate that the CS provides tangible benefits to the community.

RCS staff guide youth in choosing CS assignments based on the needs of the victim and community, as well as the youth’s skills and preferences.

SCS youth are given a list of CS assignments to choose from without guidance on what CS would address the needs of the victim or community.

RCS requires that the CS hours are proportional to the offense and set the upper limit of CS hours to 24 to shift focus on the quality (vs. the quantity) of the CS.

A youth can receive up to 150 hr of CS for each offense.

Community interactions

Youth in the RCS program always work alongside community volunteers and mentors.

Youth in the SCS program often work in isolation or with other youth.

Community members perform the same work as youth unless the job requires extensive training or increases liability (e.g., driving a vehicle or handling hazardous material).

There is often a difference between what CS work youth perform and what CS work community volunteers perform, if they are present.

Community members are encouraged to interact with the youth on a personal level to foster connections among the youth, the service work, and the community.

Community members often do not interact with youth on a meaningful level.

Note. RCS = restorative community service; SCS = standard community service; CS = community service.

8 CRIMINAL JUSTICE AND BEHAVIOR

society, and accepts responsibility for antisocial behavior), Peer Relationships (three items on current peer association, current admiration/emulation of antisocial peers, and current resistance to antisocial peer influence), School Conduct (two items on school conduct and expulsions/suspensions), Academic Performance (three items on involvement in school activities, school attendance, and academic performance); and Substance Use (three items on current substance use, impairment from drug use, and impairment from alcohol use).

within the PACT manual, items are assigned an absolute value (0–3) relative to how much the item response is associated with recidivism according to evidence-based assess- ments, with higher scores indicating either more risk or more protective weight (Barnoski, 2004; see Supplemental Material, available in the online version of this article, for a list of items and values). To maintain the evidence-based weights within each item, but differenti- ate between risk and protective responses within the current analyses, all protective item responses were coded as positive (e.g., 2 = strong protective, 1 = weak protective), and all risk item responses were coded as negative (e.g., −3 = highest risk, −2 = strong risk, and −1 = weak risk). Neutral items were coded as 0. The items within each outcome measure were summed for the outcome analyses. For instance, the nine items within the attitude measure were summed to create a total Attitudes score.

demographics, recidivism, and other evidence-Based interventions

Demographics variables, whether other interventions were completed by the youth (i.e., Functional Family Therapy or Aggression Replacement Training), and recidivism data were provided by the Administrative Office of the Courts (Table 2). Recidivism was defined as any adjudicated misdemeanor and felony offense (excluding traffic and parking infractions and sentencing violations) committed within 18 months after the youth was assigned to CS (Barnoski, 2004). Because recidivism was based on adjudicated offenses, a 12-month adju- dication period was included in the calculation of recidivism to account for the period between the date of the offense and the resulting adjudication.3 Recidivism was coded as a dichotomous variable.

AnAlytiCAl PlAn

Prior to the outcome analyses, PSM procedures were used to approximate group compa- rability on observed baseline covariates. Propensity scores represent the probability of being in a group (bound between 0 and 1), given a set of pretreatment covariates (Miller et al., 2015). To estimate propensity scores, binary logistic regression models were con- ducted to predict the probability of a categorical outcome (i.e., assignment into the SCS group) at various levels of independent variables (i.e., pretreatment covariates; Ho et al., 2007). Pretreatment covariates included age, race, gender, month/year youth began CS,4 risk assessment status (low, moderate, or high), and criminal history score. These covariates were selected because they may have been associated with treatment assignment or out- come measures (Rubin & Thomas, 1996).

Because the other evidence-based programs (i.e., Functional Family Therapy or Aggression Replacement Training) provided by both courts may have had a disproportionately large positive effect on the outcome variables (Khan et al., 2015), exact matching for this variable was implemented within matching procedures to emphasize the importance of a large-effect variable (Ho et al., 2007). Specifically, youth in the SCS group who completed zero, one, or

Church et al. / PSyCHOSOCIAL OUTCOMES OF COMMUNITy SERVICE 9

two of these evidence-based interventions were matched to youth in the RCS group who completed the same number of evidence-based interventions. Following this exact matching procedure, youth who completed SCS were randomly sorted and then matched with the near- est youth from the RCS group with the most similar propensity score based on the other covariates. A 2:1 nearest neighbor matching, without replacement, within a caliper of 0.15 SDs was implemented. This matching procedure allowed two RCS youth to be matched to one SCS youth and ensured that RCS and SCS youth did not differ on their estimated pro- pensity score by more than 0.15 SD units. youth (n = 560) who did not have a closely matched pair were excluded. Standardized difference values were calculated before and after matching procedures to determine how much the procedure improved the balance between the treatment and comparison groups (Miller et al., 2015). Standardized difference values greater than .10 indicate moderate imbalance, whereas values greater than .20 indicate severe imbalance (Austin, 2009).

Once the PSM procedure was completed, analyses of covariance (ANCOVAs) were conducted to compare the differences in psychosocial outcomes at pre/post assessment for all youth (within group), as well as the difference in outcomes between youth who received RCS and youth who received SCS (between group), while controlling for

Table 2: Matched Study Groups Characteristics

Covariate

Means and proportions before matching Means and proportions after matching

RCS (n = 1,570)

SCS (n = 659)

Absolute standardized difference (d)

RCS (n = 1,043)

SCS (n = 626)

Absolute standardized difference (d)

Age 15.10** 14.74 .23 14.87 14.80 .04 Gender (% male) 76% 74% .03 75% 75% .01 Race * .12 .05 White 85% 79% 83% 81% Black 11% 14% 13% 14% Asian 3% 4% 3% 3% American 1% 3% 2% 1% Indian/Alaskan Other 0.2% 0.3% 0.2% .2% EBP ** .51 .08 None 66% 57% 63% 58% One 26% 32% 28% 31% Two 8% 11% 9% 10% Month/year youth begana 08/2011** 05/2012 .24 09/2011 12/2011 .09 Risk status * .12 .06 High risk 26% 29% 31% 29% Med risk 40% 41% 42% 41% Low risk 29% 26% 28% 30% Criminal Hx 6.32** 5.57 .31 5.79 5.60 .11

Note. Bold text indicates severe imbalance, |d| > .25; italicized text indicates moderate imbalance, |d| > .1. RCS = restorative community service; SCS = standard community service; EBP = evidence-based program (i.e., Functional Family Therapy or Aggression Replacement Training); CS = community service; Hx = history. aThis variable was a numerical data point that represented the number of days from the onset of the study (i.e., January 1, 2004) to when the youth were assigned to CS. The data were transformed after propensity score matching to a month/year to present the data more clearly. *p < .01. **p < .001.

10 CRIMINAL JUSTICE AND BEHAVIOR

differences in length of probation and propensity scores. The interaction between Type of CS (RCS vs. SCS) and Time (before and after CS) was examined to determine whether one form of CS resulted in greater change. Bonferroni post hoc tests, which adjust for multiple comparisons, were used to evaluate the group comparisons. Assumptions for each ANCOVA were checked prior to computing the analyses (Rutherford, 2001). Levene’s tests were used to assess for homoscedasticity, which were not significant for any of the dependent variables. To test for linearity of the continuous independent vari- ables, each covariate was squared and added to the model to test for significance. The squared term for probation length was significant in each analysis. These results indicated significant curvature between probation length and each outcome. Polynomial terms for probation length were added to each outcome analyses to account for the curvilinear rela- tionship (Rutherford, 2001), but this addition did not change any of the main findings. Because recidivism was a dichotomous outcome, a binary logistical regression was used to compare recidivism rates between RCS and SCS while controlling for a youth’s length of probation and propensity score.

reSultS

mAtChing ProCedureS And PreliminAry AnAlySeS

Table 2 summarizes the proportions of covariates before and after the matching proce- dures. Standardized difference values were used to quantify how much PSM improved the balance between the treatment and comparison groups. Table 2 also reports the standardized bias for each covariate in the propensity score model before and after matching procedures. After matching, all covariates were sufficiently well-balanced (i.e., no covariate value exceeded .10). Some moderate bias remained for propensity score (d = .16), which was included as a covariate in all the outcome analyses. This final step accounted for residual covariate imbalance between the groups (Harder et al., 2010).

There was a significant difference in length of probation between the RCS and SCS groups in the final matched sample t(1,668) = 6.05 p = .012, with SCS youth having significantly longer probation periods in 30-day/1-month increments (M = 28.47 months, SD = 18.67) compared with RCS youth (M = 23.12 months, SD = 16.75). This differ- ence further supported the need to include length of probation as a covariate in all analy- ses. Composite reliability coefficients for the summated outcome measures were as follows: Attitudes (.87), School Conduct (.75), Academic performance (.61), Peer rela- tionships (.77), and Substance use (.77). Skew and kurtosis values were acceptable for the Attitudes, School Conduct,5 Academic Performance, and Peer Relationships outcome measures. The distributions for the Substance Use outcome measure, however, were sub- stantially skewed. Based on the distributions and the items comprising the measures, Substance Use outcomes were fit within ordinal categories. There were three distinct groups: youth who did not engage in substance use, youth who engaged in substance use but had no or minimal impairment (problems in one or no domains), and youth who engaged in substance use and had high impairment (problems in more than one domain). Other studies using the PACT data (e.g., Van der Put et al., 2014) have created similar categories for substance use. Consequently, multinomial logistic regression was used in place of an ANCOVA to determine the effect of Type of CS and Time on the substance use outcome while controlling for the covariates.

Church et al. / PSyCHOSOCIAL OUTCOMES OF COMMUNITy SERVICE 11

PSyChoSoCiAl outComeS

Table 3 provides the means and standard errors for the mixed ANCOVAs. The results of the two-way mixed ANCOVA showed that there was a small main effect of Time, F(1, 1,666) = 8.14, p = .004, ηp

2 = .005) on Attitudes,6 with outcomes significantly improved between baseline (M = 1.43, SD = 10.10) and follow-up (M = 2.23, SD = 10.90). However, the effects were better explained by the significant interaction between Time and Type of CS on Attitudes, F(1, 1,666) = 6.63 p = .011, ηp

2 = .004. Supporting Hypothesis 1, post hoc pairwise comparisons indicated that Attitudes significantly improved for the RCS group over time, whereas Attitudes did not significantly change over time in the SCS group (Supplemental Figure S1, available in the online version of this article).

There was a small main effect of Time, F(1, 1,666) = 16.63, p < .001 ηp 2 = .01 on Peer

Relationships, with outcomes significantly improved between baseline (M = −1.27, SD = 3.22) and follow-up (M = −1.28, SD = 3.52). Again, the effects were better explained by the significant interaction between Time and Type of CS on Peer Relationships, F(1, 1,666) = 9.83, p = .006, ηp

2 = .006. Consistent with Hypothesis 5, post hoc pairwise comparisons

indicated that Peer Relationships significantly improved for RCS group over time, whereas Peer Relationships significantly worsened over time in SCS group (Supplemental Figure S3, available in the online version of this article).

There was a small main effect of Time, F(1, 1,386) = 19.16, p < .001, ηp 2 = .014 on

School Conduct, with outcomes significantly improved between baseline (M = −1.22, SD = 2.65) and follow-up (M = 0.10, SD = 2.49). The effects were better explained by the

Table 3: Type of Community Service by Time Interaction on PaCT Domains Controlling for Covariates

RCS SCS

Psychosocial Outcome M SE M SE

Attitudes pre 2.83 .31 −1.09 .41 Attitudes post 4.13*** .31 −1.11 .41 Time × Type of CS F 6.63* ηp

2 .004 Peer relationships pre −0.94 .10 −1.83 .14 Peer relationships post −0.07* .10 −2.25** .14 Time × Type of CS F 9.83** ηp

2 .006 School conduct prea −1.19 .10 −1.27 .11 School conduct post 0.35 .08 −0.30 .11 Time × Type of CS F 11.78** ηp

2 .008 Academic performance prea −1.89 .11 −1.84 .15 Academic performance post −1.61 .12 −1.77 .15 Time × Type of CS F 1.05 ηp

2 .001

Note. Covariates appearing in the model are evaluated at the following values: Probation length = 753.91, propensity score = .32323. PACT = Positive Achievement Change Tool; RCS = restorative community service; SCS = standard community service. aThe sample size for School Conduct and Academic Performance was 866 for RCS and 520 for SCS. *p < .05. **p < .01. ***p < .001.

12 CRIMINAL JUSTICE AND BEHAVIOR

interaction between Time and Type of CS on School Conduct, F(1, 1,386) = 11.78, p = .001, ηp

2 = .008. Post hoc pairwise comparisons indicated that School Conduct signifi- cantly improved for both groups over time. Supporting Hypothesis 3, School Conduct improved significantly more in the RCS compared with the SCS group (Supplemental Figure S2, available in the online version of this article). There was a small significant main effect of Time, F(1, 1,386) = 8.51, p = .004, ηp

2 = .006 on Academic Performance, with outcomes improved between baseline (M = −1.69, SD = 3.41) and follow-up (M = −1.54, SD = 3.66). Contrary to Hypothesis 4, there was not a significant interaction between Time and Type of CS, F(1, 1,386) = 1.05 p = .35, ηp

2 = .001 on Academic Performance. A multinomial logistic regression examined the effect of Type of CS, Time, the interac-

tion between Time and Type of CS, Probation Length, and Propensity Score on Substance Use. Based on the model, there was a significant effect for Time and Propensity Score, as well as the interaction between Time and Type of CS. The coefficient for Propensity Score was −2.124, 95% confidence interval (CI) = [−3.04, −1.21]. Thus, the odds of being in the high impaired Substance Use group decreased by 2.1% for each 1% increase in the propen- sity score. Based on the main effect of Time, youth in both RCS and SCS were less likely to be in one of the more impaired Substance Use groups at follow-up. The main effect of Time was best explained by the interaction between Time and Type of CS. The effects and odds ratios for Substance Use between the groups and across time are shown in Table 4. Odds ratios were calculated based on the average probation time of 754 days to interpret the interaction of Time by Type of CS. For RCS, the odds of being in one of the more impaired Substance Use groups decreased by 51.5% during the program. For SCS, the odds of being in one of the more impaired Substance Use groups decreased by 31.5% during the program. Before the program, the odds of being in one of the more impaired Substance Use groups in the RCS program were 4.8% higher than they were in the SCS program. After the program, the odds of being in one of the more impaired Substance Use groups were 35.9% lower for the youth in the RCS program than in the SCS program. Supporting Hypothesis 2, the change from before to after CS was greater in the RCS program than it was in the SCS pro- gram for youth, with an average probation period. In addition, following CS, the odds of being in a more impaired Substance Use group were higher for those in the SCS program than in the RCS program.

reCidiviSm

A binary logistical regression comparing recidivism rates between RCS and SCS account- ing for Probation Length and Propensity Score was significant, χ2(4) = 39.31, p < .001 (Table 5). The model explained 3.2% (Nagelkerke R2) of the variance in recidivism and

Table 4: Program Comparisons on Substance Use across Time

Program Time Program Time [Relative] effect

Difference (logit scale) Odds ratio

Restorative After Restorative Before −0.7237 0.485 Traditional After Traditional Before −0.3777 0.685 Restorative Before Traditional Before 0.0466 1.048 Restorative After Traditional After −0.2994 0.741

Note. Odds ratios were calculated based on the average probation time of 754 days.

Church et al. / PSyCHOSOCIAL OUTCOMES OF COMMUNITy SERVICE 13

correctly classified 64% of cases. Contrary to Hypothesis 6, there was not a statistically significant effect for Type of CS. Specifically, 35.3% of SCS youth and 34.2% of RCS youth committed at least one offense during the follow-up period. For each additional 30 days of probation, the odds of recidivism increased by 1% for youth in the RCS group and by 2.4% for youth in the SCS group. Thus, Probation Length had a stronger effect on the probability of recidivism in the SCS group.

diSCuSSion

This study examined whether RCS was associated with more positive psychosocial out- comes when compared with SCS in justice-involved youth. The hypotheses were that RCS would lead to improved outcomes compared with SCS in the domains of attitudes, peer relationships, school conduct, academic performance, and substance use because RCS aims to improve attitudes of youth and also includes more meaningful service work and positive social support experiences. An additional hypothesis was that these intermediate psychoso- cial outcomes would contribute to reduced rates of recidivism for RCS youth compared with SCS youth. RCS was associated with more positive attitudes (Hypothesis 1) and peer relationships (Hypothesis 5) at the end of probation, whereas there were no changes in atti- tudes following SCS, and SCS was associated with poorer peer relationships. School con- duct and substance use outcomes improved for both groups. However, the odds of being in one of the more impaired substance use groups were lower in the RCS program than in the SCS program (Hypothesis 2), and RCS was associated with fewer negative behaviors at school compared with SCS (Hypothesis 3). Although academic performance improved in both groups, contrary to Hypothesis 4, the RCS group did not improve more than the SCS group. Contrary to Hypothesis 6, recidivism rates did not differ between RCS and SCS.

RCS uses restorative principles to engage youth, while addressing the needs of the vic- tims and the community (Ryals, 2004; Strang & Braithwaite, 2017). For the Clark County RCS program, participation in RCS included mentorship, attending restorative training, and working alongside community members while completing meaningful service work. These RCS enhancements, as compared with SCS, likely contributed to the greater improvements in outcomes. Specifically, the opportunities to evolve identities, build positive relationships, and increase self-efficacy may explain why RCS youth had significantly better attitudes and behavioral outcomes compared with SCS youth (Caldwell & Smith, 2006; Tolan et al., 2014). It is likely that task orientation in CS diverted attention from urges and maladaptive behaviors for both programs (Caldwell & Smith, 2006), which may have contributed to

Table 5: logistic Regression Predicting likelihood of Recidivism based on Type of Community Ser- vice, Probation length, and Propensity Score

Variable B SE B Wald p value Odds ratio

TYPE_CS(1) −0.375 .193 3.773 .052 0.687 Probation length 0.009 .005 4.088 .043 1.009 Propensity score −1.066 .530 4.042 .044 0.344 TYPE_CS(1) by probation length 0.015 .006 6.334 .012 1.015 Constant −0.520 .211 6.083 .014 0.594

Note. This is a dichotomous variable that differentiates RCS and SCS. TYPE_CS = type of community service; RCS = restorative community service; SCS = standard community service.

14 CRIMINAL JUSTICE AND BEHAVIOR

improved substance use and school conduct outcomes in both counties. RCS’s inclusion of more meaningful service work likely allowed youth to develop additional skills and a sense of accomplishment leading to more improved outcomes. Furthermore, the mentoring oppor- tunities incorporated into RCS likely fostered improvements in relationships, contributing to more improved peer relationships and reduced substance use (Geller, 2011; Kristjansson et al., 2010; Taylor et al., 1999). In contrast, SCS programs in the juvenile justice system may have provided opportunities for peer deviancy training when the CS was performed in a group format with minimal supervision (Dishion et al., 1999). These different implemen- tation practices may explain why SCS was associated with worse peer relationships at the end of probation. Therefore, even if courts do not choose to move toward a fully restorative model, it may be beneficial to consider increasing the supervision of youth completing mandated CS in groups and providing opportunities for youth to interact directly with com- munity members.

Although this study found significant and meaningful differences between RCS and SCS on a number of the psychosocial outcomes, there were no significant differences in rates of recidivism between RCS and SCS. Reducing re-offending remains a primary goal of juve- nile sanctions. whether the apparent failure of RCS to result in a greater reduction in recidi- vism is the result of measurement or methodological limitations (e.g., lack of longer term follow-up) or whether RCS is not more effective than SCS at reducing recidivism remains an important subject for future study. Nevertheless, it is important to consider that RCS achieved comparable recidivism rates to SCS with significantly fewer total hours of CS. A large number of CS hours may be more difficult for youth to complete, leading to more involvement in the juvenile justice system for noncompliance. Indeed, youth in the SCS group had significantly longer probation periods compared with RCS youth. Courts may consider focusing on the quality of the CS assignments instead of the quantity, and then assess whether this shift leads to a conservation of resources and decreased involvement of youth in the justice system.

Although most agencies emphasize the need to implement evidence-based practices, recent findings have indicated a drift and reduction of effectiveness for brand-name youth programs (Knoth et al., 2019; Peterson, 2017). For lower risk youth, there is a lack of program development and availability. Although frequently viewed as a low-impact sanc- tion, a common assumption is that CS serves as a learning experience that may change youth perceptions toward future justice system involvement. Given the promising, but far from definitive, results from this study, jurisdictions that use SCS should consider adopt- ing RCS, using a clinical trial model. In other words, they could begin to implement RCS by randomly assigning youth to either an RCS or SCS condition. Such random assign- ment would eliminate the need for propensity analysis and the risk of potential uncon- trolled confounds. Because it is not yet certain that RCS yields superior outcomes, such an approach would be ethical and would ensure that future approaches to CS would be empirically informed.

This study had many strengths (e.g., use of a well-validated risk assessment tool; repeated measures that incorporated self-report, as well as reports from parents, teachers, and social and health professionals). There were also a number of limitations. The PSM required the elimination of several hundred youth from Clark County. Although this pro- cedure was required to achieve balance on the covariates, it resulted in some loss of power. Also, PSM cannot remove all selection bias and unobserved characteristics could

Church et al. / PSyCHOSOCIAL OUTCOMES OF COMMUNITy SERVICE 15

have influenced outcomes. Ideally, randomized control trials (RCTs) should evaluate the effect of the program but conducting RCTs in juvenile justice settings with comparable sample sizes may not be feasible in many jurisdictions due to logistical barriers (e.g., funding, staff, and time; Miller et al., 2015). A recent study demonstrated that similar treatment effects can be found between effectiveness studies and RCT conditions if the individuals in both settings are comparable in terms of inclusion/exclusion criteria and matched on baseline covariates with PSM (Lutz et al., 2016). Thus, although randomiza- tion was not possible in this study, PSM may be able to estimate the treatment outcomes that could be found in an RCT. Another limitation of the study was that socioeconomic status, current criminal charge, and other programs included in the youth’s case plan (except Functional Family Therapy or Aggression Replacement Training) were not avail- able to be included as covariates. The current adjudicated offense, however, was included in the calculation of the criminal history score.

The sample was also limited to youth who were assigned to and completed probation, and therefore cannot generalize to youth who are assigned CS but do not complete proba- tion. Furthermore, the majority of youth in both programs were white males, and these results may not generalize to female youth and youth of color. In the general populations of Clark and Kitsap counties, between 76% and 78% of individuals identify as white, whereas more than 80% of youth assigned to CS were white, although ethnic minorities are usually overrepresented in the juvenile justice system (Bishop et al., 2010). Ethnic minority youth are likely to receive more severe sentences (e.g., detention, transfer to adult court, and boot camps) than their white counterparts, even when controlling for sentence severity or dif- ferential offending rates (Leiber & Peck, 2012). CS may be viewed as a low-impact sanc- tion, but it is unclear whether racial biases influenced sentencing.

Measurement constraints were also a concern. Specifically, official records of recidivism (i.e., adjudicated offenses) likely represent a conservative estimate of re-offending. It was also not possible to determine how many CS hours a youth completed or when the youth completed their CS hours during their probation period. The observation period for recidi- vism began when CS was assigned at the beginning of the probation period. Therefore, it is possible that some youth recidivated prior to completing CS, but “completers” were less likely to re-offend. A more detailed analysis of hours and completion may reveal an impor- tant dosage effect.

The type of CS work completed is an important area for future study. Given the variety of tasks that are classified as CS nationally, isolating the effective components of both RCS and SCS would provide an important contribution. For instance, Astin and Sax (1998) found that education-related CS was associated with more positive academic outcomes than other types of CS (e.g., human needs, public safety, or environmental work). Future studies should also directly attempt to evaluate potential mechanisms of change of RCS (e.g., mentoring experiences, meaningful work, decreased shame, increased self-efficacy, and community connectedness) to determine which components of RCS lead to more positive outcomes.

ConCluSion

Overall, our current findings indicate that RCS may have a more positive impact than SCS on attitudes, peer relationships, school conduct, and substance use, despite requiring consider- ably fewer CS hours. In an era where the risk, need, and responsivity model (Andrews &

16 CRIMINAL JUSTICE AND BEHAVIOR

Bonta, 2010) has focused juvenile correction’s attention on high-risk and high-need youth, RCS provides a noteworthy alternative. RCS may be a cost-effective intervention that improves intermediate targets (i.e., psychosocial outcomes). Although refinements and calibrations to RCS content and eligibility may be needed to further improve outcomes, the current findings are promising. Agencies with reduced or a limited array of programs should consider RCS as a way of improving youth outcomes and extending the continuum of services available.

orCid id

Abere Sawaqdeh Church https://orcid.org/0000-0001-8714-5736

SuPPlementAl mAteriAl

Supplemental Figures S1–S3 are available in the online version of this article at http://journals.sagepub.com/ home/cjb

noteS

1. The requirements for completing probation differed for each youth but completion status indicated that the youth com- plied with all requirements of their probation.

2. It should be noted that although the Positive Achievement Change Tool (PACT) prescreen was designed to “screen-out” or divert low-risk youth from supervision, it is a guideline, not a requirement. Therefore, a portion of sample youth in each county received a full PACT, despite being identified as low risk through the prescreen tool.

3. where applicable, follow-up periods extended past the age of 18 years into the adult criminal justice system. 4. This variable was a numerical data point that represented the number of days from the onset of the study (i.e., January

1, 2004) to when the youth was assigned to CS. These values were transformed after propensity score matching (PSM) to a month/year format to present the data more clearly in Table 2.

5. Skewness and kurtosis values for School Conduct were also slightly elevated. School Conduct was also transformed into an ordinal variable with similar categories as Substance Use and analyzed with a multinomial logistic regression. Results were almost identical to the results from the two-way analysis of covariance (ANCOVA). Therefore, the original analyses were retained.

6. The Levene’s Statistic was significant for Time Point 1 for Peer Relationships, Time Point 2 for School Conduct, and for both time points for Attitudes outcomes. A significant Levine’s test indicated that the variance between groups was not equal. A weighted least squares approach was used to account for this heteroscedasticity (O’Neill & Mathews, 2000). Specifically, the variances of the residuals for these measures were weighted to allow the groups with smaller variances, which are viewed as more precise, to weigh more than the groups with larger variances. The results were unchanged following these analyses. Therefore, the original model was retained.

referenCeS

Andrews, D. A., & Bonta, J. (2010). The psychology of criminal conduct. Routledge. Asscher, J. J., Van der Put, C. E., & Stams, G. J. J. M. (2015). Gender differences in the impact of abuse and neglect vic-

timization on adolescent offending behavior. Journal of Family Violence, 30, 215–225. https://doi.org/10.1007/s10896- 014-9668-4

Astin, A. w., & Sax, L. J. (1998). How undergraduates are affected by service participation. Service Participation, 39(3), 251–263. Austin, P. C. (2009). Balance diagnostics for comparing the distribution of baseline covariates between treatment groups in

propensity-score matched samples. Statistics in Medicine, 28, 3083–3107. https://doi.org/10.1002/sim.3697 Barnoski, R. (2004). Washington State juvenile court assessment manual, version 2.1. washington State Institute for Public

Policy. Bazemore, G., & Karp, D. (2004). A civic justice corps: Community service as a means of reintegration. Justice Policy

Journal, 1, 1–35. Bazemore, G., & Maloney, D. (1994). Rehabilitating community service: Toward restorative service sanctions in a balanced

justice system. Federal Probation, 58(1), 24–35. Bazemore, G., & McLeod, C. (2012). Restorative justice and the future of diversion and informal social control. In E.

weitekamp & H.-J. Kerner (Eds.), Restorative justice: Theoretical foundations (pp. 143–176). willan. Bazemore, G., & Schiff, M. (2015). Restorative community justice: Repairing harm and transforming communities. Routledge.

(Original work published 2001) Bazemore, G., & Stinchcomb, J. (2004). A civic engagement model of reentry: Involving community through service and

restorative justice. Federal Probation, 68(2), 14–24.

Church et al. / PSyCHOSOCIAL OUTCOMES OF COMMUNITy SERVICE 17

Bishop, D. M., Leiber, M., & Johnson, J. (2010). Contexts of decision making in the juvenile justice system: An organi- zational approach to understanding minority overrepresentation. Youth Violence and Juvenile Justice, 8(3), 213–233. https://doi.org/10.1177/1541204009361177

Bouffard, J. A., & Muftić, L. R. (2006). Program completion and recidivism outcomes among adult offenders ordered to complete a community service sentence. Journal of Offender Rehabilitation, 43(2), 1–33. https://doi.org/10.1300/ J076v43n02_01

Bouffard, J. A., & Muftić, L. R. (2007). The effectiveness of community service sentences compared to traditional fines for low-level offenders. The Prison Journal, 87(2), 171–194. https://doi.org/10.1177/0032885507303741

Bugbee, B. A., Beck, K. H., Fryer, C. S., & Arria, A. M. (2019). Substance use, academic performance, and academic engage- ment among high school seniors. Journal of School Health, 89(2), 145–156. https://doi.org/10.1111/josh.12723

Caldwell, L. L., & Smith, E. A. (2006). Leisure as a context for youth development and delinquency prevention. Australian & New Zealand Journal of Criminology, 39(3), 398–418.

Chassin, L. (2008). Juvenile justice and substance use. The Future of Children, 18(2), 165–183. Chouhy, C., Cullen, F. T., & Lee, H. (2020). A social support theory of desistance. Journal of Developmental and Life-Course

Criminology, 6(2), 204–223. https://doi.org/10.1007/s40865-020-00146-4 Data USA. (2019). Basic demographics comparison charts for Clark County and Kitsap County, WA [Chart]. https://datausa.

io/profile/geo/kitsap-county-wa/?compare=clark-county-wa Dishion, T. J., McCord, J., & Poulin, F. (1999). when interventions harm: Peer groups and problem behavior. American

Psychologist, 54(9), 755–764. Dishion, T. J., & Tipsord, J. M. (2011). Peer contagion in child and adolescent social and emotional development. Annual

Review of Psychology, 62, 189–214. Elonheimo, H., Sourander, A., Niemelä, S., Nuutila, A. M., Helenius, H., Sillanmäki, L., Ristkari, T., & Parkkola, K.

(2009). Psychosocial correlates of police-registered youth crime. A Finnish population-based study. Nordic Journal of Psychiatry, 63(4), 292–300. https://doi.org/10.1146/annurev.psych.093008.100412

Farrington, D. P., Loeber, R., & Ttofi, M. M. (2012). Risk and protective factors for offending. In B. welsh & D. P. Farrington (Eds.), Oxford handbook of crime prevention (pp. 46–69). Oxford University Press.

Flanagan, C. A., Kim, T., Collura, J., & Kopish, M. A. (2015). Community service and adolescents’ social capital. Journal of Research on Adolescence, 25(2), 295–309. https://doi.org/10.1111/jora.12137

Geller, J. D. (2011). Probing the relationship between volunteering and substance abuse in adolescence: A test of two expla- nations [Unpublished doctoral dissertation]. Vanderbilt University.

Gelsthrope, L., & Rex, S. (2004). Community service as reintegration: Exploring the potential. In G. Mair (Ed.), What matters in probation (pp. 229–254). willan.

Gifford-Smith, M., Dodge, K. A., Dishion, T. J., & McCord, J. (2005). Peer influence in children and adolescents: Crossing the bridge from developmental to intervention science. Journal of Abnormal Child Psychology, 33(3), 255–265. https:// doi.org/10.1007/s10802-005-3563-7

Hansen, T., & Umbreit, M. S. (2018). State of knowledge: Four decades of victim-offender mediation research and practice: The evidence. Conflict Resolution Quarterly, 36(2), 99–113. https://doi.org/10.1002/crq.21234

Harder, V. S., Stuart, E. A., & Anthony, J. C. (2010). Propensity score techniques and the assessment of measured covari- ate balance to test causal associations in psychological research. Psychological Methods, 15(3), 234–249. https://doi. org/10.1037/a0019623

Hawkins, J. D., Catalano, R. F., & Miller, J. y. (1992). Risk and protective factors for alcohol and other drug problems in adolescence and early adulthood: Implications for substance abuse prevention. Psychological Bulletin, 112(1), 64–105. https://doi.org/10.1037/0033-2909.112.1.64

Ho, D. E., Imai, K., King, G., & Stuart, E. A. (2007). Matching as nonparametric preprocessing for reducing model depen- dence in parametric causal inference. Political Analysis, 15, 199–236. https://doi.org/10.1093/pan/mpl013

Job, V., Friese, M., & Bernecker, K. (2015). Effects of practicing self-control on academic performance. Motivation Science, 1(4), 219–232. https://doi.org/10.1037/mot0000024

Kadden, R. M., & Litt, M. D. (2011). The role of self-efficacy in the treatment of substance use disorders. Addictive Behaviors, 36(12), 1120–1126. https://doi.org/10.1016/j.addbeh.2011.07.032

Khan, L., Parsonage, M., & Stubbs, J. (2015). Investing in children’s mental health. A review of evidence on the costs and benefits of increased service provision. Center for Mental Health.

Knoth, L., wanner, P., & He, L. (2019). Washington state’s aggression replacement training for juvenile court youth: Outcome evaluation (Document Number 19-06-1201). washington State Institute for Public Policy.

Kristjansson, A. L., James, J. E., Allegrante, J. P., Sigfusdottir, I. D., & Helgason, A. R. (2010). Adolescent substance use, parental monitoring, and leisure-time activities: 12-year outcomes of primary prevention in Iceland. Preventive Medicine, 51(2), 168–171. https://doi.org/10.1016/j.ypmed.2010.05.001

Latimer, J., Dowden, C., & Muise, D. (2005). The effectiveness of restorative justice practices: A meta-analysis. The Prison Journal, 85(2), 127–144. https://doi.org/10.1177/0032885505276969

Leiber, M. J., & Peck, J. H. (2012). Race in juvenile justice and sentencing policy: An overview of research and policy recom- mendations. Law & Inequality, 31(2), 331–339.

18 CRIMINAL JUSTICE AND BEHAVIOR

Lutz, w., Schiefele, A. K., wucherpfennig, F., Rubel, J., & Stulz, N. (2016). Clinical effectiveness of cognitive behavioral therapy for depression in routine care: A propensity score based comparison between randomized controlled trials and clinical practice. Journal of Affective Disorders, 189(1), 150–158. https://doi.org/10.1016/j.jad.2015.08.072

Mackinnon, S. P. (2012). Perceived social support and academic achievement: Cross-lagged panel and bivariate growth curve analyses. Journal of Youth and Adolescence, 41(4), 474–485. https://doi.org/10.1007/s10964-011-9691-1

McIvor, G. (2016). what is the impact of community service? In F. McNeill, I. Durnescu, & R. Butter (Eds.), Probation: 12 essential questions (pp. 107–128). Palgrave Macmillan.

Miller, M., Fumia, D., & He, L. (2015). The King County Education and Employment Training (EET) program: Outcome evaluation and benefit-cost analysis (Doc. No. 15-12-3901). washington State Institute for Public Policy.

Morris, N., & Tonry, M. (1991). Between prison and probation: Intermediate punishments in a rational sentencing system. Oxford University Press.

Murray, J., & Farrington, D. P. (2010). Risk factors for conduct disorder and delinquency: Key findings from longitudinal studies. The Canadian Journal of Psychiatry, 55(10), 633–642.

O’Neill, M. E., & Mathews, K. (2000). Theory & methods: A weighted least squares approach to Levene’s test of homogeneity of variance. Australian & New Zealand Journal of Statistics, 42(1), 81–100. https://doi.org/10.1111/1467-842X.00109

Peterson, A. (2017). Functional Family Therapy in a probation setting: Outcomes for youths starting treatment January 2010–September 2012. Center for Court Research, Administrative Office of the Courts.

Phelps, M. S. (2018). Ending mass probation. The Future of Children, 28(1), 125–146. Rubin, D., & Thomas, N. (1996). Matching using estimated propensity scores: Relating theory to practice. Biometrics, 52(1),

249–264. https://doi.org/10.2307/2533160 Rutherford, A. (2001). Introducing ANOVA and ANCOVA: A GLM approach. SAGE. Ryals, J. S. (2004). Restorative justice: New horizons in juvenile offender counseling. Journal of Addictions and Offender

Counseling, 25(1), 18–25. https://doi.org/10.1002/j.2161-1874.2004.tb00190.x Sha, T. (2006). Optimism, pessimism and depression; the relations and differences by stress level and gender. Acta

Psychologica Sinica, 38(6), 886–901. Spengler, M., Damian, R. I., & Roberts, B. w. (2018). How you behave in school predicts life success above and beyond

family background, broad traits, and cognitive ability. Journal of Personality and Social Psychology, 114(4), 620–635. https://doi.org/10.1037/pspp0000185

Strang, H., & Braithwaite, J. (Eds.). (2017). Restorative justice: Philosophy to practice. Routledge. Suzuki, M., & wood, w. R. (2018). Is restorative justice conferencing appropriate for youth offenders? Criminology &

Criminal Justice, 18(4), 450–467. https://doi.org/10.1177/1748895817722188 Tanner-Smith, E. E., wilson, S. J., & Lipsey, M. w. (2013). The comparative effectiveness of outpatient treatment for adoles-

cent substance abuse: A meta-analysis. Journal of Substance Abuse Treatment, 44(2), 145–158. https://doi.org/10.1016/j. jsat.2012.05.006

Taylor, A. S., LoSciuto, L., Fox, M., Hilbert, S. M., & Sonkowsky, M. (1999). The mentoring factor: Evaluation of the across ages’ intergenerational approach to drug abuse prevention. Child and Youth Services, 20(1), 77–99. https://doi. org/10.1300/J024v20n01_07

Tolan, P. H., Henry, D. B., Schoeny, M. S., Lovegrove, P., & Nichols, E. (2014). Mentoring programs to affect delinquency and associated outcomes of youth at risk: A comprehensive meta-analytic review. Journal of Experimental Criminology, 10(2), 179–206. https://doi.org/10.1007/s11292-013-9181-4

Valdebenito, S., Eisner, M., Farrington, D. P., Ttofi, M. M., & Sutherland, A. (2019). what can we do to reduce disciplinary school exclusion? A systematic review and meta-analysis. Journal of Experimental Criminology, 15(3), 253–287. https:// doi.org/10.1007/s11292-018-09351-0

Van der Put, C. E., Creemers, H. E., & Hoeve, M. (2014). Differences between juvenile offenders with and without substance use problems in the prevalence and impact of risk and protective factors for criminal recidivism. Drug and Alcohol Dependence, 134(1), 267–274. https://doi.org/10.1016/j.drugalcdep.2013.10.012

watts, A. S. (2016). Probation in-depth: The length of probation sentences. Robina Institute of Criminal Law & Criminal Justice. wood, w. R. (2012). Correcting community service: From work crews to community work in a juvenile court. Justice

Quarterly, 29(5), 684–711. https://doi.org/10.1080/07418825.2011.576688

Abere Sawaqdeh Church, PhD, is a staff psychologist at the Orlando VA Healthcare System. She received her Ph.D. in clini- cal psychology from washington State University in 2019 and completed her dissertation based on this research study.

david K. marcus, PhD, is a professor and chair of the Department of Psychology at washington State University. His research interests include psychopathy, spitefulness, and other dark personality traits.

Zachary K. hamilton, PhD, is an associate professor in the School of Criminology and Criminal Justices and the associate director of the Nebraska Center for Justice Research at the University of Nebraska Omaha. His main research focus is risk and needs assessment for criminal justice populations. He has published more than 50 peer-reviewed journal articles, chapters, and books on risk and needs assessment, evidence-based practices, and program efficacy.