Peer Victimization and Well-being as a Function of Same-Ethnicity Classmates and Classroom Social Norms: Revisiting Person × Environment Mismatch Theory

Abstract

Person × environment mismatch theory has been applied to understanding how the classroom social ecology moderates associations between peer victimization and socioemotional well-being. In 2004, Bellmore et al. applied this theory to the ethnic composition and social climate of the classroom. The current study tested whether their findings replicate with a slightly younger, rural sample from the Southeastern United States, whether associations held longitudinally, and whether child ethnicity moderated effects. Participants were 4^th-grade and 5^th-grade students from 13 elementary schools (N = 1,448; M_age = 10.13 years; 701 girls; 37.9% Black; 4.40% Latina/o, 57.7% White). Measures included peer-reports of peer victimization, teacher-reports of loneliness, social withdrawal, and anxiety, and self-reports of prosocial peer treatment. Classroom social disorder was assessed using teacher-reports of aggressive behavior and peer victimization. Evidence that having a large percentage of same-ethnicity peers amplifies peer victimization-adjustment linkages was limited. Although the exact nature of identified interactive effects somewhat varied from Bellmore et al., findings similarly underscored the benefits of low social disorder and ethnically diverse classrooms. Together, these findings point to a need to understand the proximal socio-structural impact of ecological factors when studying the consequences of peer victimization.

Peer victimization has emerged as one of the primary interpersonal stressors of concern among researchers studying peer relationships in childhood and adolescence (Troop-Gordon, 2017). The correlates of peer victimization are so vast (e.g., academic and health problems, depression; McDougall & Vaillancourt, 2015) that the effects of peer harassment can seem ubiquitous. However, whether peer-victimized children evidence maladjustment is contingent upon their interpretation of their peer experiences (Graham & Juvonen, 1998) and the interpersonal resources available to them (Troop-Gordon & Gerardy, 2012; Troop-Gordon & Kuntz, 2013). Thus, understanding the impact of peer victimization requires not only assessment of the frequency, intensity, and developmental timing of the victimization, but also the social ecology in which the victimization takes place and the meaning imposed by the child.

Characterological self-blame (i.e., blaming internal, stable traits) is a key contributor to the socioemotional problems evidenced by peer-victimized children (Graham & Juvonen, 1998). However, the extent to which peer-victimized children engage in categorical self-blame likely depends on whether the victimization can alternatively be attributed to external factors. When peer victimization deviates from group norms, external attributions are difficult to formulate, and self-blame and emotional distress are likely amplified, a proposition derived from person × environment mismatch theory (Wright et al., 1986). In a seminal study, Bellmore et al. (2004) tested this theory in relation to two contextual features: (a) a high percentage of same-ethnicity peers, a context in which peer victimization would be non-normative (Verkuyten & Thijs, 2002), and (b) a peer group characterized by high levels of classroom disorder (i.e., aggression, misbehavior, and peer victimization). Their findings supported the person × environment mismatch theory with regard to same-ethnicity peers, but not for classroom disorder.

Bellmore et al.’s findings have important implications for understanding the psychological mechanisms underlying the distress caused by peer-victimization. The findings are also relevant in light of broader sociopolitical changes. The U.S. is becoming increasingly diverse (U.S. Census, 2020a); yet schools are remaining re-segregated, particularly in the American South (Amsterdam, 2017; Stroub, & Richards, 2013). Resources (e.g., school counselors) to help disruptive and aggressive students (Carrell & Carrell, 2006) are stretched thin (American School Counselor Association, 2020), a problem likely to be compounded due to budget cuts associated with COVID-19. Thus, it is imperative that we understand how classrooms’ ethnic composition and social climate impact children’s socioemotional development. Bellmore et al. (2004) presented important findings, but few attempts have been made to replicate their results and test their robustness across varying methodologies and samples. Thus, the goal of the current study was to attempt to replicate their results with a diverse group of older elementary school students residing in the Southeastern United States.

Peer Victimization, Same-ethnicity Peers, and the Person × Environment Mismatch Theory

The underlying rationale for person × environment mismatch theory (Wright et al., 1986) is that the negative ramifications of a behavior are dependent on whether there is a “misfit” between the child’s actions and the peer group culture. Applied to peer victimization, youth are hypothesized to attribute greater characterological self-blame for their victimization when aggression is uncommon in the peer group or the victimization cannot otherwise be attributed to external factors or mutable characteristics (Graham & Juvonen, 1998; Schacter & Juvonen, 2015). This provides a compelling explanation for research showing that peer victimization is associated with heightened emotional maladjustment (Huitsing & Veenstra, 2012) and lower peer preference (Sentse et al., 2007) in peer groups with low levels of peer victimization.

Graham and Juvonen (2002) posited that the ethnic composition of a peer group could similarly impact the risk posed by peer victimization. For children in the ethnic majority, being peer-victimized is non-normative, as members of the ethnic majority hold greater social power than those in the minority (Graham & Juvonen, 2002). Consequently, children in the ethnic majority who are peer victimized would presumably engage in more characterological self-blame and evidence greater emotional maladjustment than victimized youth in the ethnic minority. Utilizing data from a single school, the researchers found that peer victimization was more strongly associated with maladjustment among Black students, who were in the ethnic majority in the school, than among students of other ethnic backgrounds (Graham & Juvonen, 2002).

Building on this earlier work, Bellmore et al. (2004) conducted a more stringent test of the person × environment mismatch theory using data collected from 1,630 6^th-grade students during the fall of their first year in middle school. Students were recruited from 99 classrooms in 11 schools located in urban, working-class communities in Los Angeles, California. The sample was 46% Latina/o, 29% Black, 9% Asian, 9% White, and 7% other ethnic background. The researchers found that peer victimization was more positively associated with self-reported social anxiety and loneliness when children were in a classroom with a higher percentage of same-ethnicity classmates, consistent with the person × environment mismatch theory. In addition, low levels of peer victimization were associated with less social anxiety and loneliness when youth had a higher percentage of same-ethnicity classmates, suggesting being consonant with group norms may confer protective benefits. We refer to this person × environment pattern of low peer-victimization conferring benefits in putatively protective environments as a person × environment congruence effect (for a graphical representation of mismatch and congruence effects see Figure S1 in the supplement). Graham et al. (2009) followed-up this research by presenting evidence of an indirect effect of peer victimization in the fall of 6^th grade on maladjustment the following spring through heightened characterological self-blame, but only when youth were of the majority ethnicity in a classroom low in ethnic diversity.

Perhaps because of the unique properties of the data set utilized by Bellmore et al. (2004; e.g., a large number of schools and classrooms, substantial sample diversity, variable classroom-level diversity), attempts to replicate these findings with different populations or at different developmental periods have been scarce. Mehari and Farrell (2013) attempted to replicate the Bellmore et al. (2004) findings with a large sample of Latina/o, Black, and White 6^th-graders from 36 schools located primarily in urban areas across the U.S. Unlike Bellmore et al., Mehari and Farrell examined ethnic composition at the school-level, not the classroom-level. They did not find that the association between peer victimization and life satisfaction varied across schools, and, therefore, determined that tests of moderation (i.e., accounting for between-school variance in peer victimization-adjustment linkages) were unwarranted. Having failed to replicate Bellmore et al.’s findings, they proposed that whether ethnic majority membership moderates the effects of peer victimization may depend on the socio-historical culture in which schools are embedded, calling for further research on youth from diverse backgrounds.

In a recent study of middle school students from urban areas, Espinoza et al. (2019) examined whether the relation between peer victimization in the 6^th grade and school adjustment in the 7^th grade varies as a function of the percentage of same-ethnicity classmates in their grade. The researchers found that for Latina/o youth only low levels of peer victimization were associated with better school adjustment in schools with fewer same-ethnicity peers, a pattern inconsistent with a person × environment mismatch or congruence effect. Thus, efforts to replicate the person × environment mismatch theory with regard to the peer group’s ethnic composition has yet to be successful, but examinations have been scarce. To help fill this gap in the literature, the primary goal of this study was to test percentage of same-ethnicity classmates as a moderator of the link between peer victimization and socioemotional well-being and determine whether findings support a person × environment mismatch and/or congruence effect.

Peer Victimization, Classroom Disorder, and the Person × Environment Mismatch Theory

Bellmore et al. (2004) also examined whether classroom social disorder moderated the link between peer victimization and psychological distress. They did not find a moderating effect for loneliness. For anxiety, however, findings indicated a protective effect of low social disorder at low levels of peer victimization (i.e., a congruence effect).

The little subsequent research conducted on this issue has yielded evidence of person × environment mismatch and congruence processes. Morrow et al. (2019) found that verbal peer victimization was associated with low levels of perceived social competence in 5^th-grade classrooms with low versus high levels of aggression, consistent with mismatch theory. In a study of middle school youth, Totura et al. (2006) found that low levels of perceived school aggression and delinquency protected against internalizing problems at low levels of peer victimization and served as a risk factor for internalizing problems at high levels of peer victimization, supporting both person × environment mismatch and congruence outcomes (see Figure S1 in the supplement, for a graphical representation).

Thus, the few attempts to replicate Bellmore et al.’s (2004) findings have yielded mixed results, suggesting that low levels of classroom social disorder may provide protection at low level of peer victimization and heightened risk at high levels of peer victimization. The second objective of this research was to test whether classroom disorder moderates the link between peer victimization and psychosocial adjustment with an understudied population, and, if the findings replicate the congruence effect as found by Bellmore et al., a person × environment mismatch effect (Morrow et al., 2019), or both (Totura et al., 2006).

The Moderating Role of Classroom Ethnicity

As part of their analytic strategy, Bellmore et al. (2004) controlled for the level of ethnic diversity in the classroom and its interaction with peer victimization. Although this was not pertinent for testing person × environment mismatch theory, the findings remain an important contribution to the literature, as the benefits of interracial friendships and within-ethnicity support (Benner & Crosnoe, 2011; Chen & Graham, 2017) are critical lines of investigation within developmental psychology. Bellmore et al. found that at low levels of peer victimization, youth evidenced lower social anxiety if they were in a classroom high in ethnic diversity. In contrast, at high level of peer victimization, youth evidenced higher social anxiety regardless of the ethnic diversity of the classroom. Research by Juvonen et al. (2006) provides an explanation for this pattern of results. By distributing power across ethnic groups, classroom ethnic diversity may reduce students’ fears of being the target of aggression, thereby reducing social anxiety. However, consistent with the person × environment congruence hypothesis, the protective benefits of classroom ethnic diversity may extend only to children infrequently peer-victimized.

Few studies have attempted to replicate this finding. Using data from a Canadian middle school attended by primarily White students, Hogland and Hosan (2012) found that for Aboriginal youth, peer ethnic victimization was associated with heightened depression/anxiety at high levels of classroom ethnic diversity. For Asian students, ethnic victimization was associated with higher levels of aggression at low levels of classroom diversity. Given the study’s homogenous sample from a single school and mixed results, there is need for further research testing Bellmore et al.’s (2004) finding that ethnic diversity benefits low victimized youth. The data collected for the present study allowed for such test.

The Present Study – Replication Analyses

In sum, Bellmore et al. (2004) presented an important set of findings regarding the role of peer group ethnic composition and norms in the socioemotional adjustment of peer-victimized youth. However, the lack of studies aimed at replicating their results, and the inconsistent findings among those studies that have been conducted, call into question the extent to which their results can be replicated or generalized to other populations and developmental periods. It should be noted that efforts at exact replication within social sciences are extremely difficult, as the circumstances and methodology of the original study can almost never be reproduced (Zwaan et al., 2018). However, there is great value in efforts to replicate findings using alternative methods (i.e., conceptual or constructive replications, Lykken, 1968). Support for a theory should not rely on a highly specific set of study parameters, and efforts at replication can point to the conditions in which an effect occurs, contributing to theory refinement (Zwann et al., 2018). Thus, the objective of the current study was to utilize an extant data set to test the replicability of Bellmore et al.’s findings, using a sample and methodology that diverge from those used by Bellmore et al., but for which the person × environment mismatch theory should be applicable.

Sampling choice.

Participants for this study came from rural communities within two counties of a single state in the Southeastern United States. Three ethnic groups with the greatest representation were of interest -- White, Black, and Latina/o. White and Black residents dominate the two communities (County One: 57.1% White, 40.0% Black; County Two: 70.9% White and 23.1% Black), with families that often go back generations within the state or region. Although sharing a strong sense of community, in the shadow of slavery, Jim Crow laws, and the Civil Rights Movement, as well as close proximity to the Southern Black Belt, racial disparities and racism remain pervasive (Schuman et al., 1997). For example, lifelong White residents of the South report greater racial resentment than White adults from other regions (Carter & Carter, 2014). Poverty rates also are higher among Black residents than White residents (County One: 20.92% Black vs. 12.58% White; County Two: 29.39% Black vs. 18.76% White), and racial disparities in educational resources persist (Alabama State Department of Education, 2020).

The Latina/o population is much smaller (County One: 2.6%; County Two: 3.8%) and predominantly reflects families that had recently moved to the U.S. or had migrated to the South during the late 20^th century (Pew Research Center, 2005). The state has a culture of anti-immigrant, anti-Latina/o sentiment, perhaps best exemplified in 2011 by the passage of House Bill 56, designed to force the Latina/o community to “self-deport.” This resulted in large-scale out-migration of the state’s Latina/o community (Mohl, 2016). Even after many of the most punitive provisions of the bill were blocked in 2013, the state’s Latina/o/Hispanic community remains small (a 2.7% increase from 2010 to 2019; U.S. Census, 2020b, 2020c). In light of the racial tension that pervades the Southeast, ethnic identity likely plays a substantial role in defining in-group and out-group membership within the participating communities and schools.

Developmental stage.

Whereas Bellmore et al. (2004) focused on 6^th-graders attending middle schools, this study focused on 4^th-graders and 5^th-graders in elementary schools. There are a number of reasons to assume that the person × environment effects identified by Bellmore et al. would apply to this slightly younger age group. The pernicious effects of peer victimization during elementary school are well-documented (Reijntjes et al., 2010, 2011). Ethnic identity is already salient (Pfeifer et al., 2007; Rogers et al., 2012), and peer victimization and discrimination based on ethnicity are already observed (Verkuyten & Thijs, 2002). Furthermore, classroom ethnic composition may take on an even more prominent role in elementary school classrooms, as children spend the majority of the day with the same classmates.

However, there are also reasons one might anticipate an attenuation of effects among a younger sample. Most research on characterological self-blame focuses on adolescents, and evidence indicates that links between characterological self-blame and emotional well-being strengthen with age (Cole et al., 1996). Furthermore, during early adolescence, bullying increases despite fewer youth experiencing high levels of peer victimization. This suggests that aggression becomes increasingly targeted at a small group of youth (see Troop-Gordon, 2017), who likely view their victimization as non-normative. Thus, although there are arguments for expecting Bellmore et al.’s findings to generalize to a younger age group, empirical tests of this assumption are needed.

Measurement differences.

Although the available data set allowed for testing person × environment mismatch theory as it applies to peer victimization, measurement differences between Bellmore et al. (2004) and the current study are worth noting. Bellmore et al. used a limited peer nomination procedure to assess physical (“gets pushed around”), verbal (“gets put down or made fun of by others”), and relational victimization (“other kids spread nasty rumors about them”). The current study used a peer-report rating-scale to measure physical, verbal, and relational victimization (i.e., “get hit or pushed,” “called bad names or [kids] say other mean things to him or her,” and “get left out of things that kids are doing (kids don’t let him or her play with them”), providing a strong measure of classroom consensus. Bellmore et al. measured classroom social disorder using teacher-ratings of each student on the items “starts fights,” “argues,” “gets in trouble,” and “picked on.” Three items from the current study were used – “starts fight with peers,” “gets into verbal arguments,” and “gets teased by another kid.” Although an item assessing general misbehavior was not available, the items did assess a peer group culture in which aggression and victimization are normative, the key contextual features that would indicate whether peer victimization could best be attributed to peer culture versus characterological self-attributes.

In comparison to our assessment of peer victimization and classroom disorder, which shared method source and conceptual overlap with the measures used by Bellmore et al. (2004), there was strong divergence in the assessments of emotional well-being. Bellmore et al. used self-reports of loneliness and social anxiety. Only teacher-reports of emotional well-being were available for this study, due to relying on a data set whose primary purpose did not include assessing children’s psychological health. Teachers completed 16 items from the Anxious/Depressed subscale of the Teacher Report Form (TRF; Achenbach, 1991), a well-validated measure with this age group and with diverse populations (see Achenbach et al., 2008). From this subscale, one item, “complains of loneliness,” was used to assess loneliness, and eight previously validated items (Kendall et al., 2007; Read et al., 2015) were used to measure generalized anxiety. As social anxiety is distinct from generalized anxiety, teacher-reports of social withdrawal using the Asocial with Peers subscale from the Child Behavior Scale (CBS; Ladd & Profilet, 1996) were also included. Although asocial behavior is distinct from anxious/fearful behavior, the two are moderately correlated (Ladd et al., 2009). Moreover, as peer victimization is predictive of increased preference for solitude (Ladd et al., 2019), and social withdrawal engenders downstream costs to academic and emotional functioning (Hughes & Coplan, 2010), it is an important correlate of peer victimization.

As the greatest inconsistency between this study and Bellmore et al. (2004) was the reliance on teacher-reports of adjustment, failure to replicate findings could point to a study boundary in which Bellmore et al.’s findings are applicable (i.e., self-reported well-being). To test whether Bellmore et al.’s findings are more replicable when self-reports of adjustment are utilized, self-reported positive peer treatment was included. Prosocial treatment from peers and peer victimization are distinct constructs, each uniquely predictive of emotional well-being (Troop-Gordon & Unhjem, 2018). However, as perceptions of positive peer treatment are conceptually different from loneliness and social anxiety, this study provides only modest evidence regarding the requisiteness of self-reported adjustment for testing person × environment mismatch.

Despite these methodological differences, we hypothesized that our findings would parallel those reported by Bellmore et al. (2004). That is, we anticipated finding support for person × environment mismatch and congruence effects when testing moderation by percentage of same-ethnicity classmates and support for person × environment congruence effects when testing moderation by classroom social disorder and ethnic diversity.

The Present Study – Extension Analyses

The current work also built on Bellmore et al.’s findings in two important ways. First, we examined whether the hypothesized interactive effects would predict trajectories of adjustment. Previous work suggests that person × environment misfit processes underlie the prospective links between peer victimization and subsequent well-being (Graham et al., 2009). By utilizing data collected at three time points during the school year, the current study allowed for testing whether person × environment interactions account for concurrent-only (i.e., fall only), emerging (spring-only), or sustained (fall and spring) maladjustment among peer-victimized children.

Second, we examined whether findings differed between White, Black, and Latina/o children. Investigators have previously found ethnic subgroup differences, attributing them to sociocultural differences (Espinoza et al., 2019; Hoglund & Hosan, 2012). Majority-minority status within the larger community, racial tensions, and immigrant status could affect the attributions children make for peer victimization separate from majority-minority status in the classroom. For example, White children may not expect to be peer-victimized when in the ethnic majority due to the privilege and relative power afforded to White members of their community, and peer-victimized Latina/o children may evidence greater maladjustment when in a classroom with few same-ethnicity peers due to increased feelings of social isolation given their underrepresentation in the schools and region more broadly.

Method

Participants

Data for this study came from 1,448 children (M_age = 10.13 years; SD = .67; 701 girls) in the 4^th or 5^th grade in 13 public elementary schools in the Southeastern United States. All 13 schools were located within small, rural communities. Data were collected in three waves, approximately three months apart, beginning in the fall of the school year (i.e., Wave 1 = fall; Wave 2 = winter; Wave 3 = spring). At the beginning of the school year, all children in the schools’ 91 classrooms were invited to participate. Of these children, 1,564 (76.1%; SD_{across classrooms} = .13) received parental permission; all of whom provided written assent to participate.

Children whose ethnicities were not Black, Latina/o, or White were underrepresented in the current study (n = 76), and, therefore, not included in analyses. One teacher did not provide social disorder data, resulting in data from 20 children being excluded. Twenty children received parental permission after the fall data collection. Therefore, the final sample included 1,448 children (37.9% Black; 4.40% Latina/o, 57.7% White) from 90 classrooms. The percentage of children in each classroom who were White ranged from 0 to 92.9% (M = 55.4%), Black ranged from 5.3 to 95% (M = 35.6%), and Latina/o ranged from 0 to 30.8% (M = 4.2%). Using a cut off of > 60%, 53 (58.9%) classrooms were predominantly White (61.5 – 92.9%), 15 (16.7%) were predominantly Black (61.5% - 95%), and 22 (24.4 %) were diverse (Black:White ratios ranged from 60:40 to 20:60). It should be noted that due to missing adjustment data at all three waves, data from eight children were not included in the latent growth curve analyses. Participants came from primarily low-income families. The percentage of children at each school receiving reduced or free lunches ranged from 53.1% to 93.6%, (M = 71.31%). The teachers were predominantly White (White = 78.9%; Black = 17.8%; Latina/o = 3.3%).

Measures

Independent Variables at the Level of the Child

Demographic Characteristics

Students’ gender (0 = boys; 1 = girls) and ethnicity were obtained from school records. Two dummy-coded variables were created to indicate Black or Latina/o ethnicity with White ethnicity as the reference group.

Peer Victimization

Peer victimization was assessed using three peer-report items tapping physical, verbal, and relational victimization. Children rated participating classmates on a four-point scale (1 = never; 2 = once or twice; 3 = sometimes; 4 = a lot). The average rating received for each item was calculated, and the item-level scores were averaged to create a composite score (α = .81).

Percentage Same Ethnicity

For each child, a score was computed representing the number of same-ethnicity children in the classroom. Specifically, for each ethnic group, we calculated the number of children of that ethnicity in the classroom and divided it by the number of children in the classroom minus one. Children were assigned the percentage ethnicity score for their ethnicity.

Independent Variables at the Level of the Classroom

Ethnic Diversity

The amount of ethnic diversity in each classroom was calculated using Simpson’s index (Simpson, 1949). The index was based on the ethnicities of all participating students (White, Black, Latino/a, Asian, Native American/Alaskan Native, Native Hawaiian/Other Pacific Islander, Multiethnic, and Other). Scores ranged from 0 to 1.00. Larger values indicate greater diversity and a lower probability that two children in the classroom share the same ethnicity.

Social Disorder

Classroom social disorder was measured using three teacher-report items. All items were rated on a five-point scale (1 = Never to 5 = All the Time). Children received an average score on these three items (α = .81), and within-class averages were calculated.

Dependent Variables: Emotional Adjustment

Loneliness and Anxiety

Teachers completed 16 items from the TRF Anxious/Depressed subscale (Achenbach, 1991) on a three-point scale (1 = Not True As Far As You Know; 2 = Somewhat or Sometimes True; 3 = Very True or Often True). One item, “complains of loneliness,” was used to assess children’s loneliness, allowing for a conceptually similar measure to the one used by Bellmore et al. (2004), which outweighs the limitations of relying on a single-item indicator.

Eight items were used to construct a subscale of anxiety. Previous research has supported the use of subsets of TRF items to assess anxiety distinct from depression (Kendall et al., 2007; Read et al., 2015). The items scores were averaged to create a composite anxiety score (α = .86, .87, and .87, for the fall, winter, and spring, respectively).

Social Withdrawal

Teachers completed the six-item Asocial with Peers subscale from the CBS (Ladd & Profilet, 1996) using a three-point scale (1 = Not True As Far As You Know; 2 = Somewhat or Sometimes True; 3 = Very True or Often True). Item scores were averaged to create a composite social withdrawal score (α = .93, 92, and .92, for the fall, winter, and spring, respectively).

Perceived Prosocial Treatment from Peers

Children completed the five-item Receipt of Prosocial Acts subscale from the Social Experience Questionnaire (Crick & Grotpeter, 1996), which measures perceptions of receiving emotional support (e.g., “cheer you up when you feel upset”), aid (e.g., “give you help when you need it”), and kindness (e.g., “do something that makes you feel happy”). Ratings were made on a five-point scale (1 = Never to 5 = All the Time). Items were averaged to create a composite score (α = .79, .82, and .84, for the fall, winter, and spring, respectively).

Procedures

Data were collected in two consecutive years. The first cohort of five schools participated in the 2017–2018 school year. The second cohort of eight schools participated in the 2018–2019 school year. Schools included 91 4th-grade and 5th-grade classes with the exception of one school that went only to 4^th grade. Despite this, the distribution of classes across grades was fairly even (49 4^th-grade classes, 53.8%). All schools were participating in an evaluation of a novel activity to increase defending behavior in response to bullying. Prior to initiating consent procedures or data collection, all schools were randomly assigned to either the novel intervention or a control intervention (55 novel intervention classrooms; 37 control intervention classrooms).

Fall data were collected prior to the intervention activity. In all schools, children participated in the 45-minute intervention activity approximately two weeks after the fall data collection. Therefore, the intervention could not have influenced concurrent associations in the fall. The winter data collection took place approximately three months after the intervention activity, and the spring data collection took place three months after that. Multi-group analyses were conducted to determine whether intervention condition moderated any of the findings from the latent growth curve analyses. No differences as a function of intervention condition emerged.

All forms were group administered in the classrooms. An undergraduate or graduate research assistant read the instructions and each question aloud, and two or more assistants were present to provide additional help. The questionnaires took approximately 50–55 minutes to complete. This study was approved by the Institutional Review Board of Auburn University, Protocol #17–092 MR 1703, Project Title, “Using Deviance Regulation to Combat Bullying.”

Plan of Analyses

All analyses were conducted using Mplus (Muthén & Muthén, 1998–2017) with full information maximum likelihood (Enders & Bandalos, 2001), allowing for the inclusion of data from participants with missing data. Following an examination of the distributional properties of all study variables and bivariate correlations, we conducted analyses parallel to those performed by Bellmore et al. (2004). First, the proportion of between-classroom and within-classroom variance was estimated for each dependent variable. Next, four multilevel models, one for each dependent variable, was estimated using the fall data. At the within-classroom level, the predictors included gender, the two dichotomous ethnicity variables, peer victimization, percentage of same-ethnicity classmates, and the peer victimization × gender and peer victimization × percentage same-ethnicity classmates interactions. At the between-classroom level, classroom social disorder and ethnic diversity were included as predictors of the intercept (i.e., main effects on the dependent variable) and the peer victimization slope (i.e., cross-level interactions). All continuous predictors at the within-classroom and between-classroom levels were grand-centered. All cross-level interactions were decomposed by estimating and plotting simple slopes at +/− 1 SD for the classroom-level predictor (Bauer & Curran, 2005).

Next, multilevel linear latent growth curve models were estimated separately for each dependent variable. The same within-classroom variables were included as predictors of the trajectory intercept and slope. At the between-classroom level, classroom social disorder and ethnic diversity were included as predictors of: (a) the latent intercept (i.e., the main effect on the dependent variable in the fall), (b) the latent slope (i.e., the main effect on linear change in the dependent variable, (c) the association between peer victimization and the latent intercept, and (d) the association between peer victimization and the latent slope. Significant interactions were decomposed by estimating the linear trajectory at +/− 1 SD for peer victimization and the classroom-level predictor. The model was then re-estimated with the linear slope parameterized such that the intercept represented the dependent variable in the spring (i.e., −2, −1, 0), allowing for testing whether moderating effects remained or emerged at the end of the school year.

In light of the strong theoretical foundation for Bellmore et al.’s findings, an alpha-level of .10 was applied to parameter estimates consistent with the original findings (i.e., focal to testing the original study’s hypotheses and in the same direction as the original result), analogous to the one-tailed tests often employed in replication studies.

A final set of analyses was conducted to examine whether the effects were moderated by child’s ethnicity. Cross-sectional models were re-estimated with the inclusion of percentage same-ethnicity classmates × Black ethnicity, percentage same-ethnicity classmates × Hispanic ethnicity, and three-way interactions between peer victimization, percentage same-ethnicity classmates, and each ethnicity at the within-classroom level. Cross-sectional models were then estimated setting the between-classroom variance for the Black ethnicity × peer victimization and Hispanic ethnicity × peer victimization interactions to random and predicted by classroom social disorder and ethnic diversity (i.e., cross-level interactions were created). This process was then repeated for the growth curve analyses (i.e., within-classroom interactions were included as predictors of the latent intercept and slope; the slopes of the ethnicity × peer victimization terms were set as random and predicted by classroom disorder and ethnic diversity). An alpha-level of .05 was used as these analyses were exploratory. A figure summarizing the findings from Bellmore et al. (2004) and the current study can be found in the supplement (see Figure S2).

Results

Descriptive Statistics

Means and standard deviations for all study variables are presented in Table 1. On average, approximately half of children’s classmates were of the same ethnicity (M_{%SameEthnicity} = .53). Percentage of same-ethnicity classmates ranged from 0 to 86% for White children, 0 to 90% for Black children, and 0 to 23% for Latina/o children. The average Simpson’s Index was .43. These values are highly comparable to those found in Bellmore et al.’s sample (M_{%SameEthnicity} = .53; M_{Simpson’sIndex} =.47). Bivariate correlations among all study variables are presented in Table 2. Peer victimization was significantly associated with all study variables, including a small positive association with percentage same ethnicity. Stability coefficients were significant but moderate. Boys received higher peer victimization scores than girls, t(1,446) = 8.00, and had higher loneliness in the fall, t(1,397) = 3.73, and social withdrawal in the fall, t(1,398) = 4.08, winter, t(1,348) = 3.97, and spring, t(1,349) = 3.52, all ps < .001. Girls reported higher levels of prosocial peer treatment in the fall, t(1,413) = −10.44, winter, t(1,373) = −8.41, and spring, t(1,342) = −9.78, all ps < .001. The effect sizes for these differences were small (d ≤ .14) with the exception of perceived prosocial peer treatment (ds ranged from .23 to .28).

Table 1.

Descriptive Statistics

Variable	M	SD
Individual-level
Fall peer victimization	1.56	0.34
Percentage same ethnicity	0.53	0.23
Fall loneliness	1.11	0.36
Winter loneliness	1.11	0.38
Spring loneliness	1.14	0.38
Fall anxiety	1.26	0.35
Winter anxiety	1.27	0.37
Spring anxiety	1.24	0.35
Fall social withdrawal	1.14	0.33
Winter social withdrawal	1.14	0.33
Spring social withdrawal	1.12	0.32
Fall prosocial treatment	3.50	0.87
Winter prosocial treatment	3.40	0.88
Spring prosocial treatment	3.40	0.89
Classroom-level
Classroom disorder	1.90	0.50
Classroom ethnic diversity	0.43	0.12

Table 2.

Bivariate Correlations

	1.	2.	3.	4.	5.	6.	7.	8.	9.	10.	11.	12.	13.
1. Victimization	---
2. % same ethnicity	.06*	---
3. Fall loneliness	.19***	.04	---
4. Winter loneliness	.20***	−.04	.49***	---
5. Spring loneliness	.17***	.04	.46***	.51***	---
6. Fall anxiety	.09***	.06*	.43***	.35***	.28***	---
7. Winter anxiety	.07**	.05	.31***	.50***	.33***	.61***	---
8. Spring anxiety	.12***	.03	.33***	.36***	.51***	.52***	.58***	---
9. Fall withdrawal	.17***	.04	.32***	.29***	.22***	.40***	.26***	.27***	---
10. Winter withdrawal	.12***	.03	.19***	.40***	.23***	.28***	.38***	.25***	.55***	---
11. Spring withdrawal	.15***	.03	.24***	.31***	.38***	.25***	.26***	.44***	.55***	.57***	---
12. Fall prosocial treatment	−.14***	−.01	−.04	−.05	−.05	.00	.02	.01	−.09***	−.03	−.05	---
13. Winter prosocial treatment	−.19***	−.03	−.07**	−.12***	−.07**	−.04	−.03	−.04	−.08**	−.05	−.06*	.54***	---
14. Spring prosocial treatment	−.20***	−.05	−.06*	−.10***	−.07**	−.02	−.05	−.05	−.09***	−.09***	−.08**	.48***	.64***

^*p < .05.

^**p < .01.

^***p < .001.

Two-tailed.

Missing data ranged across variables from 3.3% to 7.2%. Missing data primarily were due to children moving during the year or nonresponse from children or teachers. Children with missing data had higher fall peer victimization (missing data: M = 1.65, SD = .38; No missing data: M = 1.55, SD = .33, t(1,446) = −4.18, p < .001), higher fall social withdrawal (missing data: M = 1.19, SD = .39; No missing data: M = 1.14, SD = .33, t(1,398) = −2.10, p = .04), higher spring social withdrawal (missing data: M = 1.21, SD = .40; No missing data: M = 1.12, SD = .30, t(1,349) = −3.42, p < .001), and lower fall perceived prosocial peer treatment (missing data: M = 3.34, SD = 1.05; No missing data: M = 3.49, SD = .84, t(1,413) = 2.21, p = .03).

Within and Between-Class Variability in Adjustment

We first examined the within-class and between-class variability in the dependent variables. For loneliness, the within-class variance was .10, and the between class variance was .03, p = .02. The Intraclass Correlation Coefficient (ICC) was .20, indicating that 20% of the variance was between classrooms. This was higher than that found by Bellmore et al. (2004; ICC = .05). For anxiety, the within-class variance was .09, and the between class variance was .04, p < .001, ICC = .34. For withdrawal, the within-class variance was .10, and the between class variance was .01, p = .01, ICC = .10. The ICC for social anxiety in the Bellmore et al. study was .05, suggesting higher between-classroom variance in the current study. For perceived prosocial peer treatment, the within-class variance was .75, and the between class variance was .01, n.s., ICC = .01, indicating that variance was almost entirely within classrooms.

Prediction of Fall Adjustment

Four multilevel models were estimated to determine whether the association between peer victimization in the fall and children’s fall adjustment was moderated by the percentage of same-ethnicity classmates, the level of classroom social disorder, and the ethnic diversity of the classroom. Parameter estimates can be found in Table 3. A main effect of Black ethnicity indicated that Black children had lower anxiety and social withdrawal scores than White children. As was the case for Bellmore et al. (2004), an interaction emerged between peer victimization and Black ethnicity, such that the association between peer victimization and loneliness was weaker for Black children than for White children. Unlike Bellmore et al., no interactions were found between peer victimization and Latina/o ethnicity, but main effects indicated that Latina/o children had lower social withdrawal and perceived prosocial peer treatment than White children. Consistent with the t-tests, a main effect of gender indicated that girls had lower social withdrawal and prosocial peer treatment scores than boys.

Table 3.

Beta Coefficients and Standard Errors for Fall Multilevel Models Predicting Loneliness, Anxiety, Withdrawal, and Perceived Prosocial Treatment

	Loneliness		Anxiety		Withdrawal		Prosocial treatment
Variable	b	SE	b	SE	b	SE	b	SE
Intercept	1.20	.03	1.30***	.02	1.21***	.02	3.25***	.04
Ethnic diversity	−.18	.20	.03	.16	−.14	.15	−.34	.27
Classroom disorder	.12*	.05	.22***	.05	.10*	.05	.03	.05
Peer victimization	.25***	.07	.15**	.04	.17**	.06	−.30**	.12
Ethnic diversity on slope	.16	.37	.51b	.31	.77*	.36	−.95	.71
Classroom disorder on slope	.14a	.08	.08	.05	.10*	.05	−.09	.15
Gender	−.03	.02	−.00	.02	−.06***	.02	.42***	.05
Black ethnicity	−.07**	.03	−.05*	.02	−.07***	.02	.06	.06
Latina/o ethnicity	−.12	.08	−.07	.04	−.21***	.05	−.47**	.17
% same ethnicity	−.08	.07	.04	.05	−.10	.07	−.12	.15
Victimization × gender	.02	.08	−.10	.06	−.01	.07	−.06	.14
Victimization × Black ethnicity	−.10*	.04	−.08	.06	−.10	.07	.12	.15
Victimization × Latina/o ethnicity	−.22	.21	.07	.14	−.22	.13	−.71	.56
Victimization × % same ethnicity	.05	.12	.04	.14	.04	.19	−.34	.30

^ap = .09.

^bp = .099.

^*p < .05.

^**p < .01.

^***p < .001.

Although Bellmore et al. (2004) found support for the person × environment mismatch and congruence processes when testing percentage of same-ethnicity classmates, these data did not reveal any significant interactions between peer victimization and percentage of same-ethnicity classmates and, therefore, did not support either person × environment processes.

With regard to classroom social disorder, Bellmore et al. (2004) found a negative interaction between classroom social disorder and peer victimization, reflecting person × environment congruence processes. In the current study, a marginally significant interaction emerged for loneliness, and a significant interaction emerged for social withdrawal. However, these were positive. Simple slopes revealed that peer victimization was more strongly positively associated with loneliness and social withdrawal at higher levels of classroom social disorder (high classroom disorder: loneliness: b = .32, p < .001; social withdrawal: b = .22, p < .001; low classroom disorder: loneliness: b = .18, p = .01; social withdrawal: b = .12, p = .06). Plots of these interactions are presented in Figure 1. The findings supported a person × environment exacerbation of risk effect (i.e., see Figure S1 in the supplement for a graphical representation).

Figure 1.

Plots of the Moderating Effect of Classroom Social Disorder on the Association Between Peer Victimization and (a) Loneliness and (b) Social Withdrawal.

With regard to classroom ethnic diversity, Bellmore et al. (2004) found a positive interaction with peer victimization, reflecting person × environment congruence processes. In this study, positive interactions emerged for anxiety and social withdrawal. Peer victimization was more strongly associated with higher levels of anxiety at higher levels of classroom ethnic diversity, (b = .18, p < .001) than at low levels of diversity (b = .10, p = .04). As shown in Figure 2a, higher levels of anxiety were estimated at high levels of peer victimization for classrooms with greater ethnic diversity. In contrast, the pattern for social withdrawal replicated Bellmore et al.’s finding of a congruence effect. Peer victimization was associated with greater social withdrawal at high levels of ethnic diversity (b = .26, p < .001), but not at low levels of diversity (b = .07, p = .36). As shown in Figure 2b, lower levels of social withdrawal were estimated at low levels of peer victimization in classrooms with greater diversity.

Figure 2.

Plots of the Moderating Effect of Classroom Ethnic Diversity on the Association Between Peer Victimization and (a) Anxiety and (b) Social Withdrawal.

Prediction of Adjustment Trajectories

Four multilevel latent growth curve models were estimated to determine whether the association between peer victimization in the fall and children’s adjustment trajectories across the school year was moderated by the percentage of same-ethnicity classmates, classroom social disorder, and the ethnic diversity of the classroom. Parameter estimates can be found in Table 4. Overall, associations between the intercept of adjustment (i.e., the fall assessment) and gender, ethnicity, and the interaction between peer victimization and gender and ethnicity were the same as when these links were tested in the cross-sectional analyses. In addition, girls evidenced greater loneliness than boys over the school year. Fall peer victimization predicted increases in perceived prosocial peer treatment (likely due to regression to the mean), as well as a more negative slope for anxiety among Latina/o children than White children.

Table 4.

Results from Multilevel Latent Growth Curve Analyses

	Loneliness		Anxiety		Withdrawal		Prosocial treatment
Variable	b	SE	b	SE	b	SE	b	SE
	Within-Level
Intercept
Gender	−.03	.02	−.01	.02	−.07***	.01	.27***	.05
Victim	.28***	.06	.12**	.04	.16**	.05	−1.19***	.15
Black ethnicity	−.07**	.03	−.04*	.02	−.07**	.02	−.07	.06
Latina/o ethnicity	−.15*	.07	−.06	.04	−.19**	.06	.24	.14
% same ethnicity	−.12	.07	.03	.05	−.10	.07	.10	.15
Victim × gender	.01	.08	−.09	.06	−.05	.07	.13	.15
Victim × Black ethnicity	−.14**	.05	−.07	.06	−.10	.07	−.21	.20
Victim × Latina/o ethnicity	−.35	.19	−.02	.13	−.13	.15	.28	.46
Victim × % same ethnicity	−.07	.13	.02	.14	.12	.17	−.07	.57
Slope
Gender	.02*	.01	.01	.01	.01	.01	−.01	.03
Victim	−.01	.04	.01	.02	−.02	.02	.17*	.07
Black ethnicity	.00	.01	−.01	.01	−.01	.01	−.01	.04
Latina/o ethnicity	−.02	.04	−.05	.03	−.01	.02	.07	.09
% same ethnicity	−.01	.04	−.02	.03	−.01	.02	.11	.09
Victim × gender	.06	.03	.05	.04	.05	.03	−.01	.07
Victim × Black ethnicity	−.03	.04	−.01	.02	.00	.03	−.01	.09
Victim × Latina/o ethnicity	−.17	.10	−.20*	.08	−.03	.05	.02	.23
Victim × % same ethnicity	−.26*	.12	−.06	.07	−.07	.08	.03	.25
	Between-Level
Intercept
Classroom disorder	.12*	.05	.24***	.05	.14**	.04	.16	.16
Ethnic diversity	−.18	.21	.04	.17	−.13	.15	.52	.62
Slope
Classroom disorder	−.03	.02	−.04	.02	−.02	.02	−.08	.07
Ethnic diversity	−.06	.08	.03	.06	−.00	.04	.29	.23
Intercept on victim
Classroom disorder	.15a	.08	.03	.06	.03	.05	.24	.23
Ethnic diversity	.18	.35	.28	.32	.78*	.32	−1.81	1.24
Slope on victimization
Classroom disorder	−.02	.25	−.00	.15	−.01	.03	−.21*	.10
Ethnic diversity	−.21	.04	−.14	.03	−.23	.13	−.03	.46

^ap= 0.08.

^*p < .05.

^**p < .01.

^***p < .001.

The interaction between peer victimization and percentage of same-ethnicity classmates predicted the slope of loneliness. However, none of the estimated trajectories were significantly different than 0 (low victim, low percentage same ethnicity: slope = −.01, p = .75; low victim, high percentage same ethnicity: slope = .02, p = .12; high victim, low percentage same ethnicity: slope = .02, p = .32; high victim, high percentage same ethnicity: slope = −.02, p = .26). Thus, as was the case for the cross-sectional analyses, percentage of same-ethnicity classmates did not moderate the association between peer victimization and adjustment, inconsistent with Bellmore et al.’s (2004) findings supporting person × environment mismatch and congruence theories.

Classroom social disorder moderated (p = .08) the association between peer victimization and the intercept for loneliness (i.e., fall loneliness). Estimated trajectories at high and low levels of peer victimization and classroom social disorder are presented in Figure 3a. At high levels of peer victimization, children evidenced greater loneliness in the fall if they were in classrooms with higher disorder (b = .17, p < .001). Although none of the estimated slopes were significantly different than 0, the estimated trajectories indicated a small increase in loneliness at high levels of peer victimization and low levels of classroom social disorder. By the spring, children who experienced high levels of peer victimization in the fall no longer differed in loneliness a function of classroom social disorder (b = .10, p = .11). Thus, the findings supported a temporary exacerbation of risk effect, not a congruence effect as found by Bellmore et al.

Figure 3.

Estimated Trajectories at High and Low Levels of Peer Victimization for (a) Loneliness as a Function of Classroom Social Disorder, an (b) Perceived Prosocial Peer Treatment as a Function of Classroom Social Disorder.

In addition, classroom social disorder significantly moderated the association between peer victimization and the slope of perceived prosocial peer treatment (see Figure 3b). Results did not replicate Bellmore et al.’s (2004) finding of a congruence effect. Perceived prosocial peer treatment increased during the school year when fall levels of peer victimization were high, and this increase was stronger at low levels of classroom social disorder (slope = .36, p = .002) than at high levels of classroom social disorder (slope = .21, p = .002). Consistent with person × environment mismatch theory, perceived prosocial peer treatment in the fall was somewhat lower at high levels of peer victimization in classrooms relatively low in social disorder (b = .24, p = .10). However, estimated levels of perceived prosocial peer treatment increased steadily over the school year at high levels of peer victimization and low levels of social disorder and were similar to those found at high levels of social disorder by the spring (b = −.06, p = .70).

Finally, classroom ethnic diversity moderated the association between peer victimization and the social withdrawal intercept. Estimated scores revealed the same pattern found in Figure 2b and replicated Bellmore et al.’s (2004) congruence effect. Low levels of peer victimization were associated with lower levels of social withdrawal in classrooms with higher ethnic diversity (b = −.40, p = .003), and this difference was sustained through the spring (b = −.27, p = .046).

Moderation by Child Ethnicity

The final set of analyses tested for three-way interactions with child ethnicity (see the supplement, Tables S1–S4, for parameter estimates). No interactions emerged for loneliness or social withdrawal. A peer victimization × classroom disorder × Hispanic ethnicity three-way interaction emerged when predicting fall anxiety in both the cross-sectional analysis, b = .44, p = .01, and the latent growth curve analysis, b = .43, p = .008. However, in both analyses, the peer victimization × classroom disorder two-way interaction was not significant for White or Hispanic children, and estimated means at high and low levels of peer victimization and classroom disorder were highly similar across both ethnicities. In addition, a peer victimization × percentage same-ethnicity classmates × Hispanic ethnicity interaction predicted the anxiety latent slope, b = 1.71, p = .02. However, none of the estimated slopes at high and low levels of peer victimization and percentage same-ethnicity classmates were significantly different than 0.

For perceived prosocial treatment, the peer victimization × percentage same-ethnicity classmates × Black ethnicity interaction predicted the slope, b = .24, p = .049 (see Figure 4). At low levels of fall peer victimization, both Black and White children reported high levels of positive peer treatment. In addition, Black children reported increasingly greater perceptions of positive peer treatment (slope = .21, p = .005) at high levels of same-ethnicity classmates. By the end of the school year, they reported more positive peer treatment if they were in classrooms with a higher percentage of same-ethnicity classmates, b = .74, p = .02, supporting a person × environment congruence effect similar to that found by Bellmore et al. (2004).

Figure 4.

Estimated Trajectories of Perceived Prosocial Peer Treatment for White and Black Children at High and Low Levels of Percent Same-Ethnicity Classmates and at (a) Low Levels of Peer Victimization and (b) High Levels of Peer Victimization

At high levels of fall peer victimization, Black and White children reported an increase in perceptions of positive peer treatment across the school year regardless of the percentage of same-ethnicity classmates (slopes ranged from .15 to .25, all ps ≤ .02). White children reported less positive peer treatment in the fall, b = −.70, p = .06, and spring, b = −.65, p = .032, if they had a higher percentage of same ethnic classmates, supporting a person × environment mismatch effect. Black children, in contrast, reported more positive peer treatment in the fall, b = .73, p = .002, and spring, b = 1.19, p < .001, if they had a higher percentage of same-ethnicity classmates, supporting a person × environment exacerbation of risk effect.

Discussion

The goal of this study was to replicate Bellmore et al.’s (2004) groundbreaking work on the consequences of peer victimization in varying ethnic and social classroom contexts with a slightly younger sample from a different geographical region. This study also built on Bellmore et al.’s work by examining differences longitudinally and by ethnicity. A number of person × environment effects emerged, but whether those interactions replicated Bellmore et al. varied. Support for person × environment mismatch theory was found when testing the moderating effect of having a high percentage of same-ethnicity classmates, but only when testing the one self-reported outcome and was limited to White children. In addition, although both studies underscore the protective role of classrooms low in social disorder, Bellmore et al. found these salutary effects at low levels of peer victimization, and the current study found them at high levels of peer victimization. Moreover, both studies provided evidence that classroom ethnic diversity may benefit low peer-victimized children. Thus, we identified important potential boundary conditions for the person × environment mismatch hypothesis and found further evidence supporting the benefits of low disorder, ethnically diverse classrooms.

Percentage of Same-Ethnicity Classmates

Bellmore et al. (2004) found heightened emotional distress (i.e., loneliness and social anxiety) among peer-victimized children if they were in a classroom with a high percentage of same-ethnicity peers. In the current study, lower perceptions of positive peer treatment were reported at high levels of peer victimization when children were in a classroom with a high percentage of same-ethnicity classmates, but this finding held only for White children. Interestingly, this elevated risk was evident even at the end of the school year. This finding is consistent with the notion that children are less likely to anticipate peer victimization when in the ethnic majority and, therefore, attribute peer victimization to self-related characterological factors, leading to heightened maladjustment (Graham & Juvonen, 2002; Graham et al., 2009).

These results point to potential qualifiers of the original findings. A person × environment mismatch process emerged only for White children, suggesting that such effects may be specific to children who experience power and privilege in their broader community (Dulin-Keita et al., 2011), and, therefore, do not anticipate in-group negative treatment. In addition, the findings point to the need to test person × environment mismatch theory using assessments that are sensitive to internal processes accessible to the child. Although we relied on well-validated measures of internalizing problems, teacher-reports of emotional problems correlate only modestly with self-reports (Achenbach et al., 2008), presumably due to teachers not being privy to their students’ internal, affective experiences. Additionally, teacher are not often present when students are victimized by peers (Vaillancourt et al., 2010). It is possible, therefore, that teachers rely on biased assumptions (e.g., that a White student with a large number of White classmates is doing well) when assessing internalizing distress among their students in the ethnic majority.

Central to Graham and Juvonen’s (2002) application of person × environment mismatch theory is the contention that children in the ethnic majority experience lower levels of peer victimization than those in the ethnic minority. In the current study, the correlation between percentage of same-ethnicity classmates and peer victimization was quite small (r = .06), even when broken down by ethnicity (for White children r = −.08, p = .03; for Black children r = .23, p < .001; for Latina/o children r = −.13, p = .29). This raises the issue of whether it is the actual protection afforded by same-ethnicity classmates or the expectation of that protection that shapes children’s attributions for their peer victimization experiences.

The nature of peer victimization to which children are exposed also warrants consideration. Black and Latina/o children are often the target of bias-based peer victimization (Hoglund & Hosan, 2012; Verkuyten & Thijs, 2002). In the current study, they were almost always in the ethnic minority or in an ethnically diverse classroom. Same-ethnicity peers may have provided social and emotional support for these children if they were the target of biased-based peer victimization (McGill et al., 2012). This would explain the finding that Black children reported less prosocial peer treatment if they were in a classroom with few same-ethnicity peers.

Classroom Social Disorder

The second goal was to examine whether classroom social disorder moderates victimization-adjustment linkages. Although Bellmore et al. (2004) hypothesized that person × environment mismatch theory would apply to classroom social disorder, they found that low classroom social disorder conferred benefits (i.e., less social anxiety) at low levels of peer victimization (i.e., a person × environment congruence effect). Bellmore et al. suggested that low classroom social disorder may be a “protective but reactive” attribute (Luthar et al., 2000), such that a low social disorder classroom may be advantageous unless children are under substantial stress, such as being victimized by peers.

A person × environment congruence effect for classroom social disorder was not found in the current study. Rather, findings pointed to lower levels of fall perceived prosocial treatment from peers among peer-victimized children in low social disorder classrooms (a person × environment mismatch effect) and greater fall loneliness and social withdrawal among peer-victimized children in high social disorder classrooms (person × environment exacerbation of risk effect). Peer victimization may increase risk for psychosocial distress in high and low disorder classrooms, but the mechanisms underlying this distress may differ in the two contexts. Characterological self-blame may underlie maladjustment due to peer victimization in low social disorder classrooms (see also, Morrow et al., 2019; Totura et al., 2006). In high social disorder classrooms, peer victimized youth may experience heightened maladjustment due to harassment from a greater proportion of classmates. Furthermore, emotional distress may be more readily detected in classrooms with high levels of social disorder. For example, teachers may be more attentive of the emotional difficulties of their students when social disorder is prevalent.

Growth curve analyses revealed that these effects were limited to the fall due to increases in perceived prosocial peer treatment and loneliness during the school year for children at high levels of peer victimization and low levels of classroom social disorder. To account for these seemingly contradictory findings, research is needed that tracks children’s peer interactions and relationships across a school year, particularly in low social disorder classrooms. For example, peer-victimized children may still receive support within the larger peer group (Troop-Gordon & Unhjem, 2018). In low social disorder classrooms, peer victimization and positive peer treatment may become increasingly decoupled, as norms for prosociality are established. However, as peer-victimized children often have difficulty forming friendships and are excluded from peer activities (Ellis & Zarbatany, 2007; Wang et al., 2010), loneliness may also increase.

Classroom Ethnic Diversity

The last goal of this study was to examine the moderating role of classroom ethnic diversity on peer victimization-adjustment linkages. Bellmore et al. (2004) reported a protective effect of diverse classrooms on social anxiety, but only for children infrequently peer-victimized (i.e., a person × environment congruence effect). The concurrent and longitudinal analyses in this study replicated this finding when examining social withdrawal. However, the cross-sectional analysis yielded a slightly different pattern when anxiety was tested such that higher levels of classroom diversity amplified associations with peer victimization.

Juvonen et al. (2006) postulated that classroom diversity benefits students by providing a less hierarchical social structure in which no single ethnic group is dominant. In ethnically diverse classrooms, infrequently peer-victimized children may have greater opportunities to socialize with differing subgroups of students, reducing reliance on one set of classmates for social engagement, preventing social withdrawal. For chronically peer-victimized children, having ethnically diverse classmates may do little to alleviate the desire to withdraw when with peers. Furthermore, as same ethnicity is a predictive factor in friendship formation (Updegraff et al., 2002), peer-victimized children, who often have few friends (Hodges et al., 1997), may feel anxious at the beginning of the school year in an ethnically diverse class, particularly if they depend on friends to protect them (Huitsing & Veensta, 2012).

Limitations, Strengths, and Future Directions

By studying person × environment mismatch theory with a sample of rural, low-income elementary school children, we were able to extend our knowledge of the limits to which ethnic context mitigates or exacerbates socioemotional difficulties caused by peer victimization. However, differences between this study and Bellmore et al. (2004) hinder our ability to identify the study factors responsible for failures to replicate original findings. To address this limitation, data sets are needed that allow for identifying the specific circumstances in which the presence of same-ethnicity peers and classroom social disorder serve as referents for determining “mismatch” with the larger social environment.

A more interesting question is why differences in sample or methodology lead to discordant results. The reductionist approach to ethnicity taken here, and in previous studies, impedes our ability to understand how socio-cultural context interacts with peer victimization in determining developmental outcomes (Gunaratnam, 2003; Stanfield, & Dennis, 1993). Ethnic status and diversity need to be studied as organizing features of social environments. Studies would be strengthened by taking into account children’s same- and other-ethnicity attitudes, the extent to which same-ethnicity contributes to friendship formation and maintenance, and the attributions children make with regard to ethnicity as a source of victimization. Mixed methods designs would further allow for a nuanced understanding of how inter-ethnic relations have historically unfolded within communities and impact children’s within-school relationships.

Developmental differences may also account for discrepant results. Cognitive immaturity may curtail the characterological self-blame theorized to underlie peer-victimized children’s emotional distress (Cole et al., 1996), and age-related changes have been documented in within-classroom segregation by race and ethnicity (Updegraff et al., 2002). Furthermore, the current study examined elementary school children, who were the oldest in their school and were with the same set of classmates throughout the day. Bellmore et al. (2004) studied young adolescents who had just transitioned to a middle school where they were the youngest and likely had a large set of new peers. They also examined the moderating effect of a single classroom context despite their students likely having classes with different subsets of peers. Mehari and Farrell (2013) argued that when studying adolescents, school-level context may be more important than classroom-level context when testing for peer group moderators of victimization-adjustment linkages. For example, being in a high social disorder classroom may be more distressful for elementary school children than middle schoolers who may have only one or two high-disorder classes during the day. On the other hand, being in the ethnic minority in one’s middle school classes may be a greater determinant of well-being for peer-victimized children than majority-minority status at the school-level, or both may uniquely moderate victimization-adjustment linkages. Thus, studies are needed to uncover how age-related changes, both intrapersonal and contextual, impact the meaning of ethnicity and social disorder for peer-victimized children.

A number of important nuances were not taken into account in the current study as well. A high percentage of same-ethnicity classmates may have different consequences depending on whether one is in the ethnic majority or minority. Therefore, future studies should test three-way interactions between ethnic diversity, percentage of same- ethnicity classmates, and peer victimization. The type and source of peer victimization may impact whether a child attributes the victimization to being a “misfit” within a classroom (Hoglund & Hosan, 2012). For example, relational victimization (e.g., social exclusion) from same-ethnicity peers may lead to greater emotional distress among children in the ethnic majority than ethnicity-based peer victimization from children of a different ethnicity. In addition, the stability of the peer group should be taken into account in future research. It is possible that external attributions or behavioral self-blame comes readily when experiencing peer victimization in a novel peer group. However, over time, chronic victimization may lead to greater characterological self-blame as efforts to improve peer relationships fail and same-ethnicity peers do not experience similar maltreatment from classmates. Finally, only internalizing symptoms were examined. It is possible that person × environment mismatch processes were occurring, but maladjustment manifests differently (e.g., externalizing problems, school avoidance) for children in the rural Southeastern United States.

Conclusion

The findings presented here underscore the importance of the peer group ethnic composition and culture in determining the impact of peer victimization on psychosocial health. Support for person × environment mismatch processes was found, but only for same-ethnicity classmates and only for White children. In contrast, for Black children, the congruence model was supported, such that low victimized children benefited from being in classrooms with a high percentage of same-ethnicity classmates. Consistent with Bellmore et al., results pointed, on the whole, to the benefits of low disorder, ethnically diverse classrooms, although primarily through congruence and exacerbation of risk mechanisms. To fully understand the discrepancies across these studies, large data sets from diverse geographical regions and varying developmental periods are needed, as well as a more thorough account of the meaning and socio-structural implications of ethnic-group identity in children’s respective communities and peer groups.