Bullying Experiences and Compromised Academic Performance Across Middle School Grades

23
http://jea.sagepub.com/ Adolescence The Journal of Early http://jea.sagepub.com/content/31/1/152 The online version of this article can be found at: DOI: 10.1177/0272431610379415 September 2010 2011 31: 152 originally published online 2 The Journal of Early Adolescence Jaana Juvonen, Yueyan Wang and Guadalupe Espinoza Across Middle School Grades Bullying Experiences and Compromised Academic Performance Published by: http://www.sagepublications.com can be found at: The Journal of Early Adolescence Additional services and information for http://jea.sagepub.com/cgi/alerts Email Alerts: http://jea.sagepub.com/subscriptions Subscriptions: http://www.sagepub.com/journalsReprints.nav Reprints: http://www.sagepub.com/journalsPermissions.nav Permissions: http://jea.sagepub.com/content/31/1/152.refs.html Citations: What is This? - Sep 2, 2010 Proof - Sep 8, 2010 Proof - Dec 29, 2010 Version of Record >> by muhammad ardi on October 13, 2011 jea.sagepub.com Downloaded from

Transcript of Bullying Experiences and Compromised Academic Performance Across Middle School Grades

http://jea.sagepub.com/Adolescence

The Journal of Early

http://jea.sagepub.com/content/31/1/152The online version of this article can be found at:

 DOI: 10.1177/0272431610379415

September 2010 2011 31: 152 originally published online 2The Journal of Early Adolescence

Jaana Juvonen, Yueyan Wang and Guadalupe EspinozaAcross Middle School Grades

Bullying Experiences and Compromised Academic Performance  

Published by:

http://www.sagepublications.com

can be found at:The Journal of Early AdolescenceAdditional services and information for     

  http://jea.sagepub.com/cgi/alertsEmail Alerts:

 

http://jea.sagepub.com/subscriptionsSubscriptions:  

http://www.sagepub.com/journalsReprints.navReprints:  

http://www.sagepub.com/journalsPermissions.navPermissions:  

http://jea.sagepub.com/content/31/1/152.refs.htmlCitations:  

What is This? 

- Sep 2, 2010Proof  

- Sep 8, 2010Proof  

- Dec 29, 2010Version of Record >>

by muhammad ardi on October 13, 2011jea.sagepub.comDownloaded from

Journal of Early Adolescence31(1) 152 –173

© The Author(s) 2011Reprints and permission:

sagepub.com/journalsPermissions.navDOI: 10.1177/0272431610379415

http://jea.sagepub.com

379415 JEA

1University of California, Los Angeles

Corresponding Author:Jaana Juvonen, Department of Psychology, University of California, Los Angeles, CA 90095 Email: [email protected]

Bullying Experiences and Compromised Academic Performance Across Middle School Grades

Jaana Juvonen1, Yueyan Wang1, and Guadalupe Espinoza1

Abstract

The goal of the study was to examine whether bullying experiences are asso-ciated with lower academic performance across middle school among urban students. The ethnically diverse sample was drawn from a longitudinal study of 2,300 sixth graders (44% Latino, 26% African American, 10% Asian, 10% White, and 10% mixed) from 11 public middle schools. Results of multilevel mod-els (MLMs) showed that grade point averages and teacher-rated academic engagement were each predicted by both self-perceptions of victimization and peer nominations of victim reputation, controlling for demographic and school-level differences as well as overall declines in academic performance over time. Further MLM analyses suggested that most of the victimization effect was due to between-subject differences, as opposed to within-subject fluctuations, in victimization over time. The results of the study suggest that peer victimization cannot be ignored when trying to improve educational outcomes in urban middle schools.

Keywords

bullying, academic performance, middle school, longitudinal research

by muhammad ardi on October 13, 2011jea.sagepub.comDownloaded from

Juvonen et al. 153

Since the late 1990s we have learned that schools are not necessarily safe havens for learning and achievement but institutions where substantial num-ber of students are worried about getting beat up or harassed by their peers. For example, in California, only about half of middle school students are reporting feeling “safe” or “very safe” in their school (WestEd, 2007), and close to half of students encounter at least one incident of bullying during their 1st year in middle school (Nishina & Juvonen, 2005). Although only a relatively small proportion of students (e.g., 6%-9%) are chronic victims of bullying (e.g., Kochenderfer & Ladd, 1996; Nylund, Bellmore, Nishina, & Graham, 2007), even temporary maltreatment by peers may compromise students’ school achievement. The question guiding the current study is whether vic-timization by peers is consistently associated with low academic performance over the course of 3 years of middle school.

A recent meta-analysis of 33 studies concluded that bullied students are more likely to earn lower grades and score lower on standardized achievement tests (Nakamoto & Schwartz, 2009). Although the findings from this meta-analysis highlight the negative association between peer victimization and academic achievement, there are less than handful of longitudinal investiga-tions on this topic. Moreover, we are aware of only one published investiga-tion (Kochenderfer-Ladd & Wardrop, 2001) on this topic that capitalized on more than two data points and bridged a time span greater than 1 year. Hence, although bullying experiences and academic performance are concurrently related, the existing research does not allow us to determine whether this association is robust across multiple years. Moreover, only a few investiga-tions have tested the temporal sequence and the mechanisms that might account for such an association. We briefly review the findings of the key studies that help us better understand what is currently known about the links between peer victimization and academic performance.

Temporal Sequence and Direct Versus Indirect EffectsA few studies have examined the direct and indirect links between victimization and achievement in elementary school over time. In one of the earliest investi-gations on this topic, Kochenderfer and Ladd (1996) showed that peer victim-ization experiences served as a precursor of school adjustment problems (e.g., academic achievement, school avoidance, loneliness) across the kindergarten year. Self-reported peer victimization among this predominately European American sample also predicted changes in school avoidance (i.e., desire not to come to school) and loneliness from fall to spring, whereas no support was

by muhammad ardi on October 13, 2011jea.sagepub.comDownloaded from

154 Journal of Early Adolescence 31(1)

obtained for school adjustment problems predicting changes in peer victimiza-tion. In a subsequent study, Ladd, Kochenderfer, and Coleman (1997) showed that even when controlling for peer acceptance and friendships, kindergartners who were bullied in the fall reported greater school avoidance in the spring, but school adjustment problems in the fall did not predict subsequent peer maltreat-ment. These findings are consistent with the social stressor model that presumes bullying experiences to affect victims’ adaptive functioning.

In another 1-year longitudinal study of an ethnically diverse sample of third- and fourth-grade students, Schwartz, Gorman, Nakamoto, and Toblin (2005) also found support for the social stressor model. Relying on peer- and teacher-reports of peer victimization, the findings showed that greater bully-ing predicted lower levels of standardized achievement scores and grade point averages (GPAs). No support was obtained for academic indicators pre-dicting victimization over this time period. In addition, Schwartz et al. (2005) demonstrated that the predictive associations between peer victimization and academic difficulties were partly accounted for by increased levels of depres-sion. Similar findings were obtained in a study with an ethnically diverse sam-ple of middle school students: peer victimization in the fall of sixth grade was associated with psychosocial difficulties and somatic problems, which in turn predicted maladaptive school functioning (i.e., lower grades and higher rates of school absences) by the end of the school year (Nishina, Juvonen, & Witkow, 2005). Thus, bullying experiences can have both direct and indirect effects on the achievement outcomes.

The most frequently tested indirect (mediational) model presumes that emotional distress caused by negative peer encounters inhibits learning and performance (Graham, Bellmore, & Mize, 2006; Juvonen, Nishina, & Graham, 2000; Nishina et al., 2005; Schwartz et al., 2005). In other words, it is assumed that a student who is victimized by peers becomes worried about getting ridi-culed or beaten up and therefore stops participating in class (Buhs & Ladd, 2001) or has trouble concentrating on the academic tasks because of compro-mised self-regulation. This assumption is supported by experimental data on college students. Baumeister, Twenge, and Nuss (2002) showed that even a brief manipulation of social exclusion adversely affects college students’ per-formance on a challenging test. Because social exclusion did not affect performance on an easy test, the authors concluded that the demands of the cognitively difficult tasks exceeded participants’ efforts to suppress emotional distress caused by negative social experiences. Based on these findings, it appears then that emotional distress elicited by bullying encounters likely impedes students’ ability to concentrate and do well in exams that largely deter-mine their academic grades in school.

by muhammad ardi on October 13, 2011jea.sagepub.comDownloaded from

Juvonen et al. 155

Although it is important to examine the underlying mechanism that can help us understand why and how adverse peer interactions are related to com-promised academic performance, from the policy or intervention perspective, it is also vital to test whether experiences of bullying and academic achieve-ment are directly related. If victims of bullying do worse academically than other students consistently over time, educators and school administrators are more compelled to intervene with bullying than when the academic problems are attributed to emotional stress (purview of mental health professionals).

Current StudyTo extend past research on peer victimization and academic outcomes, we test two multilevel models (MLMs) examining the direct links between peer victimization and academic performance. The first model expands on past research that relies mainly on concurrent analyses to examine the associa-tions between peer victimization and academic performance across 3 years of middle school. We rely on up to six data points to examine these asso-ciations. To assess the robustness of the model across specific measures, two indicators of peer victimization (self-reports and peer nominations) are used to predict two academic outcomes (academic GPA and teacher-rated aca-demic engagement). Consistent with past research, we hypothesized that a higher level of victimization across the 3 years would be associated with lower grades and lower academic engagement during this time span.

Assuming the findings of the first model support our hypothesis, we then want to find out whether this general victimization “effect” across all partici-pants and all data points is mainly due to between-subject differences versus within-subject variations in victimization over time. (We use the term effect in the statistical sense and do not imply causality). The first model combines these two effects. To understand whether it is mainly the most victimized students whose academic performance is compromised or whether any tem-poral fluctuations in individual students’ victimization experiences are related to lower concurrent academic performance, the effect of peer victimization was decomposed. Separating the between-subject differences and within-subject variations in peer victimization is important because it is one thing to assume that the academic risks are a plight of a few persistently victimized students (Buhs, Ladd, & Herald, 2006) and another to conclude that even temporary peer maltreatment is associated with compromised academic func-tioning (Nishina & Juvonen, 2005). As far as we know, the current study is the first to partial these two effects apart.

by muhammad ardi on October 13, 2011jea.sagepub.comDownloaded from

156 Journal of Early Adolescence 31(1)

The current study complements the research on elementary students (Buhs et al., 2006, Ladd et al., 1997; Schwartz et al., 2005), inasmuch as we focus on the 3 years of middle school that are critical in shaping the subsequent trajectories of school engagement and achievement (Juvonen, Le, Kaganoff, Augustine, & Constant, 2004). Relying on multiple sources of data (self-reports, peer nominations, teacher-ratings, school records), up to six time points, a large sample, and statistical methods that take into account the nested struc-ture of the data, the study is methodologically stronger than most prior research on this topic. Moreover, we examine the links between peer victimization and academic performance among relatively low-achieving ethnic minority stu-dents in urban public middle schools where about one third of students are estimated to drop out of high school (Orfield, 2005). Although we do not claim to explain low achievement by focusing on bullying, the current study findings allow us to understand whether peer relationship problems are related to low achievement and academic disengagement consistently across the 3 years of middle school.

MethodParticipants

Participants were drawn from a longitudinal study of approximately 2,300 mid-dle school students (46% boys, 54% girls). The self-reported ethnic composi-tion of the sample was ethnically diverse with 44% Latino, 26% African American, 10% Asian, 10% White, and 10% Other/mixed. During the fall of sixth grade, students were recruited from 11 public middle schools located in metropolitan Los Angeles. The schools were chosen based on their ethnic composition ranging from ethnically nondiverse to diverse (Juvonen, Nishina, & Graham, 2006). Students from 99 classrooms were recruited based on the teachers’ willingness to let them to take part in the study. All of the schools were eligible for Title I compensatory education funding and served pre-dominately low socioeconomic status communities.

ProceduresDuring the fall of sixth grade when students were recruited, they were pro-vided with information letters and consent forms in English and Spanish to take home that explained the study. To increase the return rate of consent forms, students were entered into a raffle if they returned the form with a

by muhammad ardi on October 13, 2011jea.sagepub.comDownloaded from

Juvonen et al. 157

parent signature, regardless of whether their parent granted them permission to participate. Seventy-five percent of students returned the signed parent consent form with 89% of these returned forms granting students permission to participate. Before each survey administration, students signed an assent form and they were reminded about the confidentiality of their responses. Students completed the survey (approximately 45 minutes in length) in a classroom setting during the fall and spring of each year. A graduate or under-graduate student researcher read the survey items aloud as students privately marked their responses and another student was available to individually answer questions.

Fall survey administration did not take place until several weeks after the beginning of each school year to ensure that there would be sufficient famil-iarity among peers to nominate classmates and that teachers could complete ratings of student school engagement. School record data on academic grades were obtained at the end of each semester.

At the end of the middle school (i.e., spring of eighth grade), the participa-tion rate was 75% of the original sample, which exceeds the retention rate of 59% that is expected when taking into account average student mobility within the 11 participating middle schools. This retention rate is comparable to other longitudinal studies among urban, ethnic minority youth (e.g., Gutman & Eccles, 2007). Analyses were conducted to examine if the retained sample of students across the middle school years is comparable to those students who dropped out of the study before the spring of eighth grade. Independent sample t tests compared fall of sixth-grade responses to those at the spring of eighth grade for self-perceived victimization, peer nomination peer victim-ization, GPA, and school engagement. Students who dropped out of the study received more peer nominations for victimization, had lower GPA’s, and accord-ing to teacher reports were less engaged in school (t’s > 5.00, p’s < .001) in the fall of sixth grade, suggesting that some of the most vulnerable students were not retained. Yet because of our analysis method, we were able to include students with even just one wave of data (see results).

MeasuresWe relied on both self-report and peer nomination data to assess victimization. Self-perceptions of victimization are available for each of the six time points or waves, whereas peer nomination data are available only for five of the six waves (no data for the fall of seventh grade are available due to experimenta-tion with different procedures). Two measures of academic achievement were used: school record GPA and teacher-rated academic engagement.

by muhammad ardi on October 13, 2011jea.sagepub.comDownloaded from

158 Journal of Early Adolescence 31(1)

Self-perceived victimization. A modified six-item version of the Peer Victim-ization Scale (Neary & Joseph, 1994) was used to assess self-perceptions of victimization by peers (Graham, Bellmore, & Juvonen, 2003). The wording of this scale is similar to that used in Harter’s (1987) Self-Perception Profile for Children designed to decrease social desirability bias. Each item describes two sets of hypothetical students and participants determine which type of student is more like them (e.g., “Some kids are often picked on by other kids, BUT other kids are not picked on by other kids”) and indicate whether that option is “really true for me” or “sort of true for me.” Based on a 4-point scale, the items were averaged and higher scores indicated higher levels of peer victimization (α = .82). For this measure, 2,294; 2,152; 1,928; 1,775; 1,542; and 1,092 students, respectively, have at least 1, 2, 3, 4, 5, and 6 waves of data.

Peer nominations. During the fall and spring of sixth grade (i.e., Waves 1 and 2), students were given a classroom roster organized alphabetically and by gender and were asked to list the names of up to four classmates (both same- and other-sex nominations were allowed) who fit each of the three victimization descriptions. The three items depicted physical, verbal, and relational victimization (i.e., “who gets pushed around,” “gets put down or made fun of by others,” and “other kids spread nasty rumors about them”) and were summed for each student. Consistent with prior research (Bellmore & Cillessen, 2006), the intercorrelations of the three types of victimization of the classroom-based nominations was high (average r = .90). To increase the reliability of peer nominations, only classrooms with more than 50% par-ticipation were included in the analysis. This criterion allowed us to include 80 classrooms for Wave 1 and 74 classrooms for Wave 2 when nominations were obtained at the classroom level.

After sixth grade, a similar procedure was followed, except that a random-ized list was generated for each student with the names of 50 participating grade mates instead of a classroom roster from which to choose names. The classroom roster could no longer be used when students were in seventh and eighth grade because students were not placed in teams where they had the same set of classmates for classes. Similarly to the class rosters, each list with 50 randomly generated names was organized alphabetically and by gender. Participants were asked to cross out names of students they did not know on the roster. The nominations across the three types of victimization were again highly intercorrelated (average r = .65). Analyses comparing these two meth-ods of obtaining peer nominations (classroom-based and random-list based) suggest that the nominations function similarly in terms of their temporal stability and concurrent relations to other measures (Bellmore, Jiang, &

by muhammad ardi on October 13, 2011jea.sagepub.comDownloaded from

Juvonen et al. 159

Juvonen, 2010). Peer nominations were transformed into percentage scores to control for difference in the number of possible nominators. The peer nomi-nated variable used for the analyses for Waves 1 and 2 was the total number of nominations/the total number of classmates present at the data collection × 100; whereas for Waves 4 through 6, it was the total number of nominations/the total number of times the student’s name appeared on a list and not crossed out × 100. The number of students with peer-nominated victimization data ranged from 2,298; 2,199; 1,907; 1,709 to 1,468 students with at least 1, 2, 3, 4, and 5 waves of data, respectively.

GPA. Students’ grades were collected from report cards at the end of each semester. Based on the grades earned in academic classes (English, Mathe-matics, Science, Social Studies), the GPA of each student was scored on a 5-point scale ranging from 0 (F) to 4 (A) and averaged to create a GPA com-posite for each student.

Academic engagement. Teachers completed six items from the Teacher Report of Engagement Questionnaire (Wellborn & Connell, 1991) to assess students’ level of academic engagement. Sample items include, “In my class this student concentrates on doing his/her work” and “In my class this student likes to figure things out for him/herself.” Teachers rated whether each item was “not at all characteristic of this student” to “very characteristic of this student” on a 4-point Likert-type scale such that higher values indicated greater academic engagement. The items were averaged; the measure was internally very consistent (α = .89).

ResultsThe results section is divided into three main parts. In the first section, we examine intercorrelations over time and across our main variables. The sec-ond section includes the results when we rely on self-perceptions of victim-ization as a predictor. Both a general MLM (Model 1) and a model that is designed to separate between-subjects and within-subjects effects (Model 2) are presented separately for GPA and academic engagement. The third sec-tion contains the two models for the two academic outcomes when relying on peer nominations assessing victim reputation.

IntercorrelationsTemporal consistency of the scores was estimated by computing correlation coefficients across time. Both self-perception scores and peer nominations of peer victimization showed that the rank order of participants remained rather

by muhammad ardi on October 13, 2011jea.sagepub.comDownloaded from

160 Journal of Early Adolescence 31(1)

consistent. The correlations ranged between .38 and .66 for self-perceptions of victimization. Peer nominations were particularly consistent over time (r’s ranging from .55 to .75), suggesting that regardless of the nominating base (different raters at different time points), reputations are maintained over time. For both set of indicators, the consistency of the rank order of participant scores decreases slightly across wider spans of time, as expected. For example, the correlation coefficients for self-perceptions of victimization ranged from r = .38 from the fall of sixth grade to the spring of eighth grade, while the strongest coefficient (r = .66) was recorded between the fall and spring of eighth grade.

Both indicators of academic achievement also showed consistency in the rank order of scores over time. Correlations ranged from .40 to .70 for GPA, with the strongest associations within the academic years (e.g., between fall and spring of seventh grade, r = .70, and the fall and spring of eighth grade, r = .61) and the lowest between the fall of sixth grade and the spring of eighth grade (r = .40). Teacher ratings of academic engagement showed particularly high consistency across grade levels. Academic engagement in the fall of sixth grade was correlated r = .61 even with the eighth-grade spring score, whereas the eighth-grade fall-spring correlation was r = .87. All correlation coefficients within the victimization and academic achievement indicators were significant at the p < .001 level.

Consistent with previous findings (e.g., Juvonen, Nishina, & Graham, 2001), the two victimization indicators obtained at the same wave were moderately correlated; the correlations ranged from r = .20 to .26 across the five time points. The two academic performance indicators, in turn, were strongly cor-related, ranging from r =.62 to .71 across the six time points.

MLMsThe data were analyzed with MLMs using the PROC MIXED procedure in SAS Version 9.1.3 Software (SAS Institute). MLM was chosen because of the nested nature of the data (students nested within schools and time points nested within students). MLM has its advantage that it takes into account the dependency among the observations from the same individual as well as the similarity between the students who share the same environment (school). For a longitudinal study, MLM has an additional advantage of handling miss-ing data: not all participants need to have data available for all time points (Singer & Willett, 2003). Furthermore, MLM models with full information maximum likelihood (FIML) estimation produces unbiased and efficient parameter estimates when data are missing at random (MAR; Raudenbush &

by muhammad ardi on October 13, 2011jea.sagepub.comDownloaded from

Juvonen et al. 161

Bryk, 2002). MAR means the missingness is not related to values that are not observed or not included in the model (Little & Rubin, 1987). Within the current sample, data are missing mainly for academic performance and vic-timization at earlier time points. Because these variables are included in the model, it is therefore reasonable to assume the data are missing at random. For the current analyses, any participant with at least one wave of complete data was included. Unstructured covariance matrix was used to estimate the covariances among observations.

A series of three-level models are estimated to examine changes in aca-demic performance as a function of time, victimization, gender, and ethnicity. The three-level model enables us to follow individuals within groups (schools) over time. For each outcome variable, GPA and teacher-rated school engage-ment, two models were examined.

Ytij

= γ000

+ γ100

ttij

+ γ200

ttij

2 + γ300

Victtij

+ γ010

Sexij + γ

020Eth1

ij +

γ030

Eth2ij + γ

040Eth3

ij + γ

050Eth4

ij + u

00j + r

0ij + r

1ijttij

+ εtij

(1)

Model 1 (see Equation 1 above) was designed to examine the overall effects of peer victimization (Vict

t ij) on GPA and academic engagement across

the 3 years of middle school.We included both linear (ttij

) and quadratic (ttij

2) time effects at Level 1 to describe the trend of change in academic perfor-mance throughout middle school. Time was indicated by wave of data collec-tion, ranging from 0 to 5 (or 4), indicating the six (or 5) waves of data collection. We relied on the wave rather than a specific date of data collection because the length of time was approximately 6 months between the fall and spring data collection (variation from the 6 month interval was minimal).

At Level 2, we included time-invariant student characteristics: sex (Sexij)

and ethnicity (Eth1ij – Eth4

ij). Each variable was dummy coded: the reference

group for sex was male, and the reference group for ethnicity was Latino, who comprised the largest ethnic group in the study.1 The raw victimization variable (Vict

tij) was entered at Level 1 as a time-varying predictor. The inter-

cept and slope for time were allowed to vary randomly across students (r0ij

). We decided not to include random slope for victimization at Level 1 in our final model because this component is statistically not important and the model is more parsimonious this way. We examined the effect of including random slope for the victimization at level 1 through deviance change tests of nested models, but none were significant except for the Model 1 where GPA was predicted by self-perceived victimization (χ2 = 16.7, df = 3, p < .001). At Level 3 depicting school level effects, no predictors were included, but the intercept was allowed to vary randomly across schools (u

00j). That is, we

by muhammad ardi on October 13, 2011jea.sagepub.comDownloaded from

162 Journal of Early Adolescence 31(1)

allowed the model to account for the fact that students from the same school are more similar than students from different schools.

Ytij

= γ000

+ γ100

ttij

+ γ200

ttij

2 + γ300

CVicttij

+ γ010

Sexij +

γ020

Eth1ij + γ

030Eth2

ij + γ

040Eth3

ij + γ

050Eth4

ij (2)

+ γ060

MVictij + u

00j + r

0ij + r

1ijttij

+ εtij

Model 2 (see Equation 2 above) was similar to Model 1 except that the raw victimization variable was replaced with mean level of victimization (MVict

ij) at Level 2 (i.e., capturing between-subject differences across all

time points) and centered victimization (CVicttij

) score (i.e., capturing fluc-tuations from the individual’s mean across the time points) at Level 1. By separating these two scores, the overall effect of victimization was decom-posed into between-subjects effect and within-subjects effect. This decompo-sition was realized through group-mean centering (e.g., Dotterer, McHale, & Crouter, 2009). First, each individual’s mean level of victimization (MVict

ij)

was calculated across the 3 years (six or five data points depending on the victimization indicator). This mean level of victimization corresponds to the between-subjects variation in victimization. Second, the raw scores of vic-timization at each time point were transformed into deviations from each individual’s mean level of victimization. This centered victimization (CVict

tij =

Victtij

– MVictij) variable therefore represents within-subject fluctuations

across time.The same models were tested first by relying on self-ratings of victimiza-

tion and second on peer nominations that capture victim reputations. We present the Model 2 findings for each academic outcome following Model 1 results. Table 1 depicts the Model 1 results, whereas Table 2 includes the results for Model 2.

Self-Perceptions of VictimizationGPA. The left side of Table 1 depicts the Model 1 findings regarding GPA.

The significant time effects show that GPA generally declined from the beginning to the end of middle school (γ

100 = –.097, p < .001), but the rate of

decrease gets smaller over time (γ200

= .011, p < .05). Girls had higher aca-demic GPAs throughout middle school than boys (γ

010 = .438, p < .001).

There were also ethnic differences in GPA across all six waves of data, with White (γ

020 = .407, p < .001) and Asian (γ

040 = .690, p < .001) students

obtaining higher grades in academic subjects than Latino youth, and African American youth obtaining lower grades (γ

030 = –.309, p < .001) than Latino

by muhammad ardi on October 13, 2011jea.sagepub.comDownloaded from

Juvonen et al. 163

students. Controlling for these demographic differences and overall decline in GPA across time, stronger perceptions of victimization were associated with lower GPA (γ

300 = –.028, p < .05) across all six waves of data. That is,

the more bullied the students perceived themselves, the lower grades they obtained.

To decompose the self-perceived victimization effects, we turn to Model 2, depicted in Table 2. As shown in the left hand side of Table 2, the control variables had very similar effects as just documented showing overall declines in GPA across time; girls, White, and Asian students obtained higher aca-demic GPAs than did boys and Latino students, whereas African American students received lower grades compared to their Latino classmates. Controlling

Table 1. Multilevel Model 1 Estimates for Academic Performance Predicted by Victimization

Self-perceived victimization Peer nomination victimization

GPAaSchool

engagementb GPAcSchool

engagementd

Term Coeff. (SE) Coeff. (SE) Coeff. (SE) Coeff. (SE)

Fixed effects

Intercept, γ000

2.154* (.070) 2.621* (.043) 2.132* (.068) 2.523* (.037)

Time, γ100

-0.097* (.011) -0.107* (.012) -0.070* (.012) -0.085* (.013)

Time2, γ200

0.011* (.002) 0.010* (.002) 0.006* (.002) 0.009* (.003)

Victimization, γ300

-0.028* (.011) -0.057* (.011) -0.002* (.001) -0.003* (.001)

Sex (vs. males), γ010

0.438* (.036) 0.327* (.026) 0.414* (.037) 0.309* (.026)

Ethnicity (vs. Latino)

White, γ020

0.407* (.069) 0.151* (.048) 0.414* (.071) 0.178* (.050)

African, γ030

-0.309* (.050) -0.267* (.035) -0.312* (.051) -0.253* (.036)

Asian, γ040

0.690* (.065) 0.430* (.045) 0.677* (.067) 0.428* (.046)

Other/Mixed, γ050

-0.018 (.065) -0.041 (.047) -0.013 (.067) -0.037 (.048)

Variance components

Intercept (school), τβ00

0.038* (.017) 0.006* (.003) 0.040* (.018) 0.007* (.003)

Intercept (student), τπ00

0.734* (.027) 0.406* (.019) 0.747* (.029) 0.421* (.021)

Time slope (student), τπ11

0.020* (.001) 0.010* (.001) 0.022* (.001) 0.011* (.001)

Covariance, τπ01

-0.043* (.005) -0.038* (.004) -0.048* (.005) -0.042* (.004)

Residual, σe2 0.222* (.004) 0.273* (.005) 0.216* (.005) 0.277* (.006)

Note: GPA = grade point average. Coeff. = coefficient.a. n = 10,416 observations within 2,243 participants.b. n = 9,553 observations within 2,250 participants.c. n = 8,367 observations within 2,211 participants.d. n = 7,907 observations within 2,219 participants.*p < .05.

by muhammad ardi on October 13, 2011jea.sagepub.comDownloaded from

164 Journal of Early Adolescence 31(1)

for these demographic differences and overall declines in GPA across time, higher mean level of self-perceived victimization across all six time points was associated with lower GPA (γ

060 = –.296, p < .001). Within-subject fluc-

tuations (i.e., centered scores) were not, however, related to GPA (γ300

= .006, p > .05). These findings suggest that more victimized students obtain lower grades than less victimized students and that variations in levels of victimiza-tion over time for an individual student are not associated with lower grades in academic courses.

For both Models 1 and 2, the significant intercept variances at school level (τ

β00) and at student level (τ

π00) indicate GPA differs across schools as well

as across students at the beginning of the study (see lower parts of left sides of Tables 1 and 2). The significant slope variance at the student level (τ

π11)

suggests that the rate of change differs across students. The intercept and slope

Table 2. Multilevel Model 2 Estimates for Academic Performance Predicted by Victimization

Self-perceived victimization Peer nomination victimization

GPASchool

engagement GPASchool

engagement

Term Coeff. (SE) Coeff. (SE) Coeff. (SE) Coeff. (SE)

Fixed effects Intercept, γ

000 2.675* (.086) 2.955* (.057) 2.215* (.067) 2.570* (.038)

Time, γ100

-0.094* (.011) -0.103* (.012) -0.072* (.012) -0.089* (.013) Time2, γ

200 0.011* (.002) 0.010* (.002) 0.007* (.002) 0.009* (.003)

Centered victimization, γ300

0.006 (.012) -0.003 (.013) -0.001 (.001) -0.001 (.001) Sex (vs. males), γ

010 0.411* (.036) 0.308* (.025) 0.384* (.037) 0.292* (.026)

Ethnicity (vs. Latino) White, γ

020 0.395* (.068) 0.133* (.048) 0.457* (.071) 0.197* (.050)

African, γ030

-0.300* (.049) -0.261* (.035) -0.290* (.050) -0.241* (.036) Asian, γ

040 0.688* (.064) 0.422* (.045) 0.690* (.066) 0.432* (.046)

Other/Mixed, γ050

-0.013 (.064) -0.045 (.046) 0.006 (.067) -0.029 (.048) Mean victimization, γ

060-0.296* (.032) -0.230* (.023) -0.006* (.001) -0.005* (.001)

Variance components Intercept (school), τ

β00 0.027* (.013) 0.005* (.003) 0.036* (.016) 0.006* (.003)

Intercept (student), τπ00

0.712* (.026) 0.396* (.018) 0.737* (.029) 0.416* (.021) Time slope (student), τ

π11 0.020* (.001) 0.010* (.001) 0.021* (.001) 0.011* (.001)

Covariance, τπ01

-0.043* (.005) -0.037* (.004) -0.047* (.005) -0.042* (.004) Residual, σ

e2 0.221* (.004) 0.273* (.005) 0.215* (.005) 0.277* (.006)

Note. GPA = grade point average; Coeff. = coefficient. Sample sizes for each outcome are the same as those presented in Table 1.*p < .05.

by muhammad ardi on October 13, 2011jea.sagepub.comDownloaded from

Juvonen et al. 165

at the student level have a negative covariance (τπ01

), suggesting that students who started with a lower GPA in the fall of sixth grade demonstrates a smaller rate of change in GPA across the 3 years of middle school.

In sum, self-perceptions of peer victimization predicted lower GPA across the six time points over the 3 years of middle school. Considering the mean-ing of the Model 2 findings, a one point higher mean level of self-perceived victimization (1-4 scale) across all six time points is associated with a 0.3 decrease in average GPA across time for a student. If this effect was all due to just one of the academic subjects included in the GPA, this means that peer victimization can account for up to an average of 1.5 letter grade decrease in the subject across the 3 years of middle school.

Academic engagement. The second column from the left of Table 1 depicts the findings regarding teacher-rated academic engagement. Similar effects were obtained to those just documented in terms of GPA. Students became less engaged across the middle school grades (γ

100 = –.107, p < .001), and the

rate of decrease gets smaller over time (γ200

= .010, p < .05). Girls were more academically engaged throughout middle school than boys (γ

010 = .327, p <

.001). Comparisons among ethnic groups revealed that compared to Latino stu-dents, White (γ

020 = .151, p < .001) and Asian (γ

040 = .430, p < .001) students were

perceived as more engaged and African American youth less engaged (γ030

= –.267, p < .001) by their middle school teachers. Controlling for all these differ-ences, stronger self-perceived victimization was associated with less academic engagement (γ

300 = –.057, p < .001) across all six waves of data.

When decomposing the effects of victimization into between-subject and within-subject effects, similar effects for all the control variables were obtained as in Model 1 (see left-side of Table 2). Controlling for the overall declines in academic engagements, gender, and ethnic differences, higher mean levels of self-perceived victimization across the six time points were associated with lower engagement (γ

060 = –.230, p < . 001). No effects were obtained at the

within-subject level.2

Similarly to GPA, for both Models 1 and 2, variances components show that the expected academic engagement differs across schools (τ

β00) as well

as across students (τπ00

) at the beginning of the study (see Tables 1 and 2, lower sections). In addition, the rate of change differs across students (τ

π11),

and students who started with lower academic engagement have a smaller rate of change (τ

π01).

In sum, individual differences, but not within-subject changes over time, in self-perceptions of victimization across time were consistently related to the two academic indicators. Students with stronger sense of being bullied were likely not only to obtain lower grades but were also rated by their

by muhammad ardi on October 13, 2011jea.sagepub.comDownloaded from

166 Journal of Early Adolescence 31(1)

teachers as less academically engaged consistently across all data points within the 3 years of middle school.

Peer NominationsGPA. As shown in the third column of Table 1, the changes over time, gen-

der, and ethnic differences were all similar to those obtained in the analyses relying on self-perceptions of victimization. Controlling for the overall declines in GPA and demographic differences, stronger victim reputation was associ-ated with lower GPA (γ

300 = –.002, p < .001) across all five time points.

When decomposing the effects of victimization into between-subject and within-subject effects (third column of Table 2), the control variables again had very similar effects as documented above. Controlling for these changes and demographic differences in GPA, stronger mean level of victim reputa-tion across all five time points was associated with lower GPA (γ

060 = –.006,

p < .001). Within-subject fluctuations were not, however, related to GPA.In sum, much like self-perceptions, victim reputation predicted lower GPA

consistently across the 3 years of middle school. Overall, however, the effects of peer nominations of victimization on GPA were smaller in magnitude than those obtained for self-perceptions of victimization.

Academic engagement. Model 1 effects of peer nominations on teacher-rated engagement are shown on the right side of Table 1. The general decline over time, gender, and ethnic differences in academic engagement were all similar to those reported above. Controlling for these effects, stronger victim reputation was associated with lower academic engagement (γ

300 = –.003,

p < .001) across all five waves of data.The far right column of Table 2 captures the results for teacher ratings of

academic engagement for Model 2, designed to decompose the victimization effects. Controlling for general declines in engagement over time, gender, and ethnic differences, stronger mean level of victim reputation across all five time points was associated with lower academic engagement (γ

060 =

–.005, p < .001). Although there was a trend, the within-subject fluctuations in victim reputation over time were not significantly related to lower aca-demic engagement.

In sum, individual differences, but not within-subject changes over time, in victim reputation were consistently related to the two academic indicators. Students who had stronger reputations as victims across the 3 years of middle school were likely to be rated by their teachers as less engaged and obtain lower academic grades compared to other students.

by muhammad ardi on October 13, 2011jea.sagepub.comDownloaded from

Juvonen et al. 167

Discussion

The current study findings demonstrate robust direct associations between peer victimization and compromised academic performance over time. Our main findings are consistent regardless of whether victimization relied on self-assessments or peer nominations and whether we predicted GPA or teacher-rated academic engagement. In spite of school, gender, and ethnic differences in the academic performance indicators, the results also replicated across 11 large urban middle schools and across both five and six time points. As far as we know, this study is the first one examining the links between bullying experiences and academic performance across this many data points over one entire phase of schooling with a sample of mainly ethnic minority youth in urban settings. We were also able to show that the link between peer maltreatment and compromised academic performance is largely due to indi-vidual differences in bullying experiences. This finding suggests that high level of bullying by school mates is consistently related to academic disen-gagement and poor grades across the 3 years of middle school.

The magnitude or practical significance of the findings is substantial, inas-much as one point higher mean on self-perceived victimization score on the 4-point scale across the 6 time points predicted .3 reduction in GPA. Projecting this effect on just one of the academic subjects included in the GPA, this means that peer victimization can account for up to an average of 1.5 letter grade decrease in one academic subject (e.g., math) across the 3 years of middle school. Although our correlational findings do not allow us to draw causal conclusions, we believe this finding may be one of the strongest dem-onstrations of the how social stressors and academic performance indicators are linked in middle school.

The empirical findings documented are especially impressive in light of the high consistency of the academic indicators across the 3 years of middle school. Although there was a general decline in GPA and academic engage-ment from the first fall to the last spring of middle school, the rank order of the close to 2,000 students remained rather consistent across the 3 years. To be able to show that any social stressor is consistently related to such stable academic indicators suggests that the association is indeed a very robust one. Our findings reveal that students who were generally more bullied were likely to fall into the low range of the rank order, receiving lower grades and engag-ing less in academic tasks than did other students.

In contrast to the between-subject differences in victimization across the 3 years, the level of within-subject fluctuations in victimization over time did not contribute to the decline in academic performance in this study. It is

by muhammad ardi on October 13, 2011jea.sagepub.comDownloaded from

168 Journal of Early Adolescence 31(1)

important to interpret these findings in light of the high consistency of the victimization measures over time, however. For example, the peer nomina-tion data suggest that victim reputations “stick,” inasmuch the correlations coefficients of the nominations received remained strong across time in spite of the fact that the nominating grade mates varied across the different time points and that different methods were used to obtain the nominations (classroom-based vs. random list-based nomination pools). Thus, once a stu-dent is perceived as a victim by a few classmates, it may be difficult to change that label even if bullying decreases. Because the rank order of the self-perceptions of victimization was somewhat less consistent over time than peer nominations, they may better lend themselves to the test of within-subject effects of victimization over time. Yet different self-report questions or scales might better capture the fluctuations than those used in the current study. By relying on daily reports of specific incidents across 5 days over a 2-week period, Nishina and Juvonen (2005) demonstrated that on days when students reported being bullied, they showed stronger sense of school aver-sion (e.g., disliking school). Whether the number of specific incidents might predict changes in academic engagement over longer periods of time is a question that should be tested in future research.

Although our findings suggest that most of the victimization “effects” on academic performance are due to variations in the degree to which students are bullied, the current findings do not reveal whether there are discrete vic-timization trajectories variations. Whereas some students may be consistently victimized across all three middle school grades, others might be extremely victimized only for a year, for example. To be able to answer this question, latent class analyses that classify students into categories based on their tra-jectories of victimization are informative. Nylund et al. (2007) identified three classes based on self-reports of victimization of the same sample as used in the current study. The smallest class consisted of students who were most victimized over time, the largest group consisted of nonvictimized stu-dents, and the medium category included students who were high only on verbal forms (both name calling and rumors) of victimization. The most vic-timized students felt less safe concurrently and reported highest rates of depression one semester later. Thus, latent class analyses inform us about the discrete types of plights that may help us gain further insights about the lack of academic progress.

The biggest limitation of the current study is that we cannot make causal inferences about victimization experiences affecting academic performance based on the analyses presented. We decided not to test whether victimization is negatively associated with subsequent academic performance for two main

by muhammad ardi on October 13, 2011jea.sagepub.comDownloaded from

Juvonen et al. 169

reasons. First, as described in the introduction, the experimental evidence on young adults suggests that the emotional distress elicited by exclusion inhibits performance on cognitively demanding (but not easy) tasks (Baumeister et al., 2002). Thus, the “cognitive load” explanation suggests it is the concurrent competing resources needed to complete the difficult task and to try to sup-press the distress associated with the aversive social experience that then com-promise test performance. The second reason for not conducting time-lagged analyses is based on the empirical evidence of longitudinal studies on peer victimization and mainly emotional distress. Research conducted both in ele-mentary and middle school settings shows that temporary victims recover from their adverse social experiences between the fall and spring of the school year (Juvonen et al., 2000; Kochenderfer & Ladd, 1996). When relying on five data points across initial school entry and third grade, Kochenderfer-Ladd and Wardrop (2001) found that although some former victims “bounced back,” others continued to show high levels of loneliness, while yet another group showed not even concurrent distress. These findings suggest that although there are individual differences in ability to cope with bullying encounters, many youth are resilient. Based on all these considerations, we presumed that it is more critical to test whether victimization experiences are consistently associated with compromised academic performance across time rather than whether earlier experiences predict subsequent academic performance. This latter question is nevertheless ripe for additional analyses.

Although we did not test the direction of the effects, two specific findings suggest that academic problems may increase the risk of bullying. In Model 2, the significant intercept effect of peer victimization shows that more victim-ized students were less academically engaged and had lower GPAs already from the outset. In addition, the difference between highly victimized stu-dents and less victimized students largely stayed the same across all the data points. Although the association between the two constructs is likely to be more similar than different across all grades in middle schools, this does not mean that bullying experiences could not have cumulative effects over time, however. It is possible that academic performance declines faster for the most chronically bullied students over time. To study this question, it would be interesting to compare victimized and nonvictimized students whose initial academic performance is the same in the beginning of middle school. Alternatively, latent class analysis could be used to study the academic per-formance of the chronically bullied students.

Even if academic problems not only result from but also increase the risk of peer victimization, the current findings suggest that bullying cannot be

by muhammad ardi on October 13, 2011jea.sagepub.comDownloaded from

170 Journal of Early Adolescence 31(1)

ignored when trying to improve educational outcomes in urban middle schools. Yet, in many urban areas where academic pressures are greatest with the low-est performing students (and schools), administrators and educators are solely focused on improving achievement scores. Ignoring or not being able to “afford” to address social-emotional issues, such as bullying, may be a very short-sighted view of educational progress. In light of the research evidence presented in this study, the connection between students’ peer relationships and their academic performance is irrefutable.

Declaration of Conflicting Interests

The author(s) declared no potential conflicts of interest with respect to the authorship and/or publication of this article.

Funding

This research was supported by the National Science Foundation and the William T. Grant Foundation.

Notes

1. We initially tested sex and ethnicity also as predictors of the time slope. Because they were not significant predictors of the time slope for most models according to deviance change tests, and because excluding them had minimal influence on the parameter estimates related to our hypotheses, our final models only include sex and ethnicity as predictors of the intercepts but not the slopes.

2. We also tested whether the association between mean level of victimization and the academic indicators varied as a function of time. Both the linear and qua-dratic effects by time were assessed. Two of the four analyses revealed signifi-cant time moderator effects. To assess the meaningfulness of these effects, two procedures were used. First, proportional reduction in variance (O’Connell & McCoach, 2008) was computed comparing the mean level of victimization and its interactions with Time. The results indicated that the interaction had a smaller effects (0.00%~0.86% reduction in slope variance), compared to the mean level of victimization on the intercept (3.15%~9.30% reduction in intercept variance). Second, including or excluding these smaller than 1% effects—that are con-sidered small (Cohen, Cohen, West, & Aiken, 2003)—from the model did not change the inferences regarding any of the other parameters. In addition, because the documented slightly less steep decline in academic outcomes for the most victimized students could reflect possible floor effects and because such effects were obtained only for two of the four tests, we decided not to include these Mean level of victimization × Time interactions in the final model.

by muhammad ardi on October 13, 2011jea.sagepub.comDownloaded from

Juvonen et al. 171

References

Baumeister, R. F., Twenge, J. M., & Nuss, C. K. (2002). Effects of social exclusion on cognitive processes: Anticipated aloneness reduces intelligent though. Journal of Personality and Social Psychology, 83, 817-827.

Bellmore, A., & Cillessen, T. (2006). Reciprocal influences of victimization, per-ceived social preference, and self-concept in adolescence. Self and Identity, 5, 209-229.

Bellmore, A., Jiang, X. L., & Juvonen, J. (2010). Utilizing peer nominations in middle school: A longitudinal comparison between complete classroom list and random list methods. Journal of Research on Adolescence, 20, 538-550.

Buhs, E. S., & Ladd, G. W. (2001). Peer rejection in kindergarten: Relational pro-cesses mediating academic and emotional outcomes. Developmental Psychology, 37, 550-560.

Buhs, E. S., Ladd, G. W., & Herald, S. L. (2006). Peer exclusion and victimization: Processes that mediate the relation between peer group rejection and children’s classroom engagement. Journal of Educational Psychology, 98, 1-13.

Cohen, J., Cohen, P., West, S. G., & Aiken, L. S. (2003). Applied multipleregression/correlation analysis for the behavioral sciences. Mahwah, NJ: Lawrence Erlbaum.

Dotterer, A. M., McHale, S. M., & Crouter, A. C. (2009). The development and corre-lates of academic interests from childhood through adolescence. Journal of Edu-cational Psychology, 101, 509-519.

Graham, S., Bellmore, A., & Juvonen, J. (2003). Peer victimization in middle school: When self- and peer views diverge. Journal of Applied School Psychology, 19, 117-137.

Graham, S., Bellmore, A. D., & Mize, J. (2006). Peer victimization, aggression, and their co-occurrence in middle school: Pathways to adjustment problems. Journal of Abnormal Child Psychology, 34, 363-378.

Gutman, L. M., & Eccles, J. S. (2007). Stage-environment fit during adolescence: Trajectories of family relations and adolescent outcomes. Developmental Psychol-ogy, 43, 522-537.

Harter, S. (1987). Manual for the Self-Perception Profile for Children. Denver, CO: University of Denver.

Juvonen, J., Le, V. N., Kaganoff, T., Augustine, C., & Constant, L. (2004). Focus on the wonder years: Challenges facing the American middle school. Santa Monica, CA: Rand.

Juvonen, J., Nishina, A., & Graham, S. (2000). Peer harassment, psychological adjust-ment, and school functioning in early adolescence. Journal of Educational Psy-chology, 92, 349-359.

by muhammad ardi on October 13, 2011jea.sagepub.comDownloaded from

172 Journal of Early Adolescence 31(1)

Juvonen, J., Nishina, A., & Graham, S. (2001). Self-views versus peer perceptions of victim status among early adolescents. In J. Juvonen & S. Graham (Eds.), Peer harassment in school: The plight of the vulnerable and victimized (pp. 105-124). New York: Guilford.

Juvonen, J., Nishina, A., & Graham, S. (2006). Ethnic diversity and perceptions of safety in urban middle schools. Psychological Science, 17, 393-400.

Kochenderfer, B. J., & Ladd, G. W. (1996). Peer victimization: Cause or consequence of school maladjustment, Child Development, 67, 1305-1317.

Kochenderfer-Ladd, B. J., & Wardrop, J. L. (2001). Chronicity and instability of chil-dren’s peer victimization experiences as predictors of loneliness and social satis-faction trajectories. Child Development, 72, 134-151.

Ladd, G. W., Kochenderfer, B. J., & Coleman, C. C. (1997). Classroom peer accep-tance, friendship, and victimization: Distinct relational systems that contribute uniquely to children’s school adjustment? Child Development, 68, 1181-1197.

Little, R. J. A., & Rubin, D. B. (1987). Statistical analysis with missing data. New York: Wiley.

Nakamoto, J., & Schwartz, D. (2009). Is peer victimization associated with academic achievement? A meta-analytic review. Social Development, 19, 221-242.

Neary, A., & Joseph, S. (1994). Peer victimization and its relationship to self-concept and depression among schoolgirls. Personality and Individual Differences, 16, 183-186.

Nishina, A., & Juvonen, J. (2005). Daily reports of witnessing and experiencing peer harassment in middle school. Child Development, 76, 435-440.

Nishina, A., Juvonen, J., & Witkow, M. R. (2005). Sticks and stones may break my bones, but names will make me feel sick: The psychosocial, somatic, and scho-lastic consequences of peer harassment. Journal or Clinical Child and Adolescent Psychology, 34, 37-48.

Nylund, K., Bellmore, A., Nishina, A., & Graham, S. (2007). Subtypes, severity, and structural stability of peer victimization: What does latent class analysis say? Child Development, 78, 1706-1722.

O’Connell, A. A., & McCoach, D. B. (Eds.). (2008). Multilevel modeling of educa-tional data. Greenwich, CT: Information Age Publishing.

Orfield, G. (2005). Confronting the graduation rate crisis in California: Civil Rights Project. Cambridge, MA: Harvard University.

Raudenbush, S. W., & Bryk, A. S. (2002). Hierarchical linear models: Applications and data analysis methods (2nd ed.). Thousand Oaks, CA: SAGE.

Schwartz, D., Gorman, A. H., Nakamoto, J., & Toblin, R. L. (2005). Victimization in the peer group and children’s academic functioning. Journal of Educational Psychology, 97, 425-435.

by muhammad ardi on October 13, 2011jea.sagepub.comDownloaded from

Juvonen et al. 173

Singer, J. D., & Willett, J. B. (2003). Applied longitudinal data analysis: Modeling change and event occurrence. New York: Oxford University Press.

Wellborn, J. G., & Connell, J. P. (1991). Engagement versus disaffection: Moti-vated patterns of action in the academic domain. Rochester, NY: University of Rochester.

WestEd. (2007). California Healthy Kids Survey. Aggregated California data (Tech-nical Report, 2005-2006 & 2006-2007). San Diego, CA: San Diego County Office of Education.

Bios

Jaana Juvonen, PhD, is a professor of developmental psychology program at University of California, Los Angeles (UCLA). Her area of expertise is in young adolescent peer relationships and school adjustment. She has coauthored and coedited books, including Focus on the Wonder Years: Challenges Facing the American Middle School; Peer Harassment in School: The Plight of the Vulnerable and Victimized.

Yueyan Wang is a doctoral student in quantitative psychology at UCLA. Her research interests lie in the applications of multilevel models to longitudinal or nested data in psychology and improving the statistical methods to address substantive research questions.

Guadalupe Espinoza is currently a doctoral student in developmental psychology at UCLA. Her research interests focus on examining how social contexts impact school functioning and peer relationships (e.g., bullying), primarily among Latino adolescents.

by muhammad ardi on October 13, 2011jea.sagepub.comDownloaded from