English Assignment

School Effects on Psychological Outcomes During Adolescence
Eric M. Anderman
University of Kentucky
Data from the National Longitudinal Study of Adolescent Health were used to examine school-level
differences in the relations between school belonging and various outcomes. In Study 1, predictors of
belonging were examined. Results indicated that belonging was lower in urban schools than in suburban
schools, and lower in schools that used busing practices than those that did not. In Study 2, the relations
between belonging and psychological outcomes were examined. The relations varied depending on the
unit of analysis (individual vs. aggregated measures of belonging). Whereas individual students’
perceptions of belonging were inversely related to depression, social rejection, and school problems,
aggregated belonging was related to greater reports of social rejection and school problems and to higher
grade point average.
Research on school-level differences during adolescence often
has focused on nonpsychological outcomes, such as academic
achievement and behavioral issues, instead of on psychological
outcomes (Roeser, 1998). Indeed, research on school-level differences in nonacademic variables is quite rare. The purpose of the
present research was to examine school-level differences in a
variety of psychological outcomes, using a large nationally representative sample of adolescents.
School Effects on Student Outcomes
Although there is an abundant literature on effective schools,
most of the research in this literature has focused on academic
variables, such as achievement, dropping out, and grade point
average (GPA; e.g., Edmonds, 1979; Miller, 1985; Murphy, Weil,
Hallinger, & Mitman, 1985). This literature generally indicates
that schools that are academically effective have certain recognizable characteristics.
Some of these studies have examined differences between public schools and other types of schools. For example, some research
indicates that students who attend public schools achieve more
academically than do students who attend other types of schools
(e.g., Coleman & Hoffer, 1987). Other research suggests that there
may be a benefit in terms of academic achievement for students
who attend Catholic schools compared with non-Catholic schools
(Bryk, Lee, & Holland, 1993). Lee and her colleagues (Lee,
Chow-Hoy, Burkam, Geverdt, & Smerdon, 1998) found that students who attended private schools took more advanced math
courses than did students who attended public schools. However,
they also found specific benefits for Catholic schools: Specifically,
in Catholic schools, there was greater school influence on the
courses that students took, and the social distribution of course
enrollment was found to be particularly equitable.
In recent years, psychologists have started to become interested
in the effects of schooling on mental health outcomes (e.g., Boekaerts, 1993; Cowen, 1991; Roeser, Eccles, & Strobel, 1998;
Rutter, 1980). However, little research to date has examined
school-level differences in mental health outcomes. One of the
areas that has received considerable attention has been the study of
dropping out. Rumberger (1995) found that perceptions of schools’
fair disciplinary policies by students are related to lower drop-out
rates. A recent study using data from the National Education
Longitudinal Study (NELS) found that after controlling for student
characteristics, drop-out rates were higher in public schools than in
private schools (Goldschmidt & Wang, 1999). Goldschmidt and
Wang (1999) also found that a school’s average family socioeconomic status (SES) was related to drop-out rates. Specifically, in
both middle schools and high schools, drop-out rates were higher
This research is based on data from the Add Health project, a program
project designed by J. Richard Udry (Principal Investigator) and Peter
Bearman and funded by National Institute of Child Health and Human
Development Grant P01-HD31921 to the Carolina Population Center,
University of North Carolina at Chapel Hill, with cooperative funding
participation by the following institutions: the National Cancer Institute;
the National Institute of Alcohol Abuse and Alcoholism; the National
Institute on Deafness and Other Communication Disorders; the National
Institute on Drug Abuse; the National Institute of General Medical Sciences; the National Institute of Mental Health; the National Institute of
Nursing; the Office of AIDS Research, National Institutes of Health (NIH);
the Office of Behavior and Social Science Research, NIH; the Office of the
Director, NIH; the Office of Research on Women’s Health, NIH; the Office
of Population Affairs, Department of Health and Human Services (DHHS);
the National Center for Health Statistics, Centers for Disease Control and
Prevention, DHHS; the Office of Minority Health, Office of Public Health
and Science, DHHS; the Office of the Assistant Secretary for Planning and
Evaluation, DHHS; and the National Science Foundation. Persons interested in obtaining data files from the Add Health study should contact
Joyce Tabor, Carolina Population Center, 123 West Franklin Street, Chapel
Hill, North Carolina 27516-3997.
This research was supported by a Research Committee grant from the
Vice President of Research and Graduate Studies at the University of
Kentucky. Portions of this article were presented as an invited address at
the annual meeting of the American Psychological Association, Boston,
Massachusetts, August 1999. I am grateful to Lynley Anderman, Fred
Danner, and Skip Kifer for comments on earlier versions of this article. I
am also grateful to Dawn Johnson and Barri Crump for assistance with this
research.
Correspondence concerning this article should be addressed to Eric
M. Anderman, Department of Educational and Counseling Psychology, University of Kentucky, Lexington, Kentucky 40506-0017. E-mail:
[email protected]
Journal of Educational Psychology Copyright 2002 by the American Psychological Association, Inc.
2002, Vol. 94, No. 4, 795–809 0022-0663/02/$5.00 DOI: 10.1037//0022-0663.94.4.795
795
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
when there were high numbers of low-SES children attending the
school.
Perceptions of School Belonging
In recent years, a small but important literature on school
belonging has emerged. Results of a variety of studies converge on
the consistent finding that perceiving a sense of belonging or
connectedness with one’s school is related to positive academic,
psychological, and behavioral outcomes during adolescence. Although different researchers operationalize and study belonging in
various ways, there is a general consensus among a broad array of
researchers that a perceived sense of belonging is a basic psychological need and that when this need is met, positive outcomes
occur.
Baumeister and Leary (1995) have discussed belonging as a
construct that is important to all aspects of psychology. Specifically, they have argued that the need to belong is a fundamental
human motivation, that individuals desire to form social relationships and resist disruption of those relationships, and that individuals have the need to experience positive interactions with others
and these interactions are related to a concern for the well being of
others. In addition, they have demonstrated that when individuals
are deprived of belongingness, they often experience a variety of
negative outcomes, including emotional distress, various forms of
psychopathology, increased stress, and increased health problems
(e.g., effects on the immune system). Baumeister and Leary argued
that belonging is a need rather than a want because it has been
related to these and other outcomes; that is, if an individual is
deprived of such a need (as opposed to something that the individual wants), then negative outcomes (e.g., stress, health problems) may occur (Baumeister & Leary, 1995, p. 520).
Deci and colleagues (Deci, Vallerland, Pelletier, & Ryan, 1991),
in their discussion of self-determination theory, have included the
concept of relatedness as one of the basic psychological needs
inherent to humans (the other two needs are the need for competence and the need for autonomy). Deci et al. argued that social–
contextual influences that support students’ relatedness lead to
intrinsic motivation if the individuals who provide support to the
student are also supportive of the student’s autonomy.
Finn (1989) related the concept of belonging to drop-out behavior. Finn developed the participation–identification model to attempt to explain this behavior. Finn’s model posits that students
who identify with their schools develop a perception of school
belonging. It is this perception of belonging that facilitates the
students’ academic engagement and commitment to schooling.
When a sense of belonging is not nurtured in students, they may
become more likely to drop out.
Some programs of research have examined belonging (and
related variables) specifically in relation to school learning environments. Most of these studies indicate that when students experience a supportive environment in school, they are more likely to
experience positive outcomes. For example, Newman, Lohman,
Newman, Myers, and Smith (2000) interviewed urban adolescents
making the transition into ninth grade. One of the factors distinguishing successful from nonsuccessful transitions was that highachieving middle-school students who made a successful transition
into high school reported having friends who supported their
academic goals. This notion of peer support of goals is an important component of many operational definitions of school
belonging.
Battistich and colleagues (Battistich, Solomon, Watson, &
Schaps, 1997) have demonstrated that the presence of a “caring
school community” often is associated with positive outcomes for
students. Battistich et al. agreed with the tenets of Deci et al.
(1991) regarding students’ needs for belonging. However, Battistich et al. argued that when the school environment facilitates
student participation in a caring community, students’ needs for
belonging (as well as for autonomy and competence) are met. The
results of Battistich et al.’s program of research indicates that a
sense of community is related to a variety of positive outcomes for
students, such as improved social skills, motivation, and achievement (Battistich et al., 1997).
Goodenow (1993b) developed a measure of the psychological
sense of school membership for use with adolescents. The scale
originally was developed and validated on samples of early adolescents from suburban and urban schools. Students’ reported
perceptions of school membership were found to be related positively to teachers’ projected year-end grades in English classes and
to expectancies for success, the subjective value of school work,
and academic achievement (see also Goodenow & Grady, 1993).
Similar research on classroom belonging indicates that the relation
between belonging and motivation (expectancies and values) declines as students progress through the sixth and eighth grades
(Goodenow, 1993a).
Roeser, Midgley, and Urdan (1996) examined the relations
between perceived school belonging and academic achievement in
a sample of early adolescents. They found, when controlling for
prior achievement, demographics, personal achievement goals,
perceptions of school goal stresses, and perceptions of the quality
of teacher–student relationships, that school belonging positively
predicted end-of-year grades.
L. H. Anderman and Anderman (1999) examined changes in
personal task and ability goal orientations over the middle-school
transition. After controlling for demographics, perceptions of
classroom goal orientations, and social relationship variables, they
found that a perceived sense of school belonging was related to
changes in personal achievement goals. Specifically, school belonging was related to an increase in personal task goals and to a
decrease in personal ability goals across the middle-school
transition.
In summary, a variety of studies have identified the construct of
belonging as being an important psychological variable. When an
individual’s need for belonging is met, positive outcomes occur.
Within schools, a perceived sense of school belonging is related to
enhanced motivation, achievement, and attitudes toward school.
School-Level Differences in Perceived School Belonging
An extensive review of the literature has not uncovered any
studies that have examined school-level differences in perceived
belonging. Nevertheless, there is reason to suspect that belonging
varies as a function of school characteristics. In particular, school
size, school grade configuration, and urbanicity are three schoollevel variables that theoretically should be related to a student’s
sense of belonging.
796 ANDERMAN
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
School Size
It is plausible that students may develop a greater sense of
belonging in smaller sized schools than in larger sized schools.
Specifically, when schools are small in size, students are more
likely to get to know their teachers and their classmates on a more
interpersonal level. Because it may be easier to form social relationships both with students and teachers in a smaller sized school
environment, the need for belonging may be more easily satisfied
in a smaller school (see Baumeister & Leary, 1995).
There is some research evidence that indicates that smaller sized
schools are more effective than are larger sized schools. Lee and
Smith (1995) examined the effects of school size and restructuring
on gains in academic achievement and engagement in high school
students. They found that students who attended small-sized
schools and students who attended schools that used specific
reform practices (e.g., keeping the same homeroom throughout
high school, interdisciplinary teaching, schools-within-schools)
learned more and were more academically engaged than students
who attended other schools. In addition, they found that gains in
achievement were more equitably distributed (in terms of SES) in
schools that used restructuring practices (see Lee & Smith, 1995,
for a full description of such practices). A subsequent study that
used additional data from later in students’ high school careers
confirmed many of these findings (Lee, Smith, & Croninger,
1997).
Nevertheless, not all evidence points to negative effects of large
school size. One recent study using NELS data (Rumberger &
Thomas, 2000) examined school effects on dropping out. Results
indicated, after student characteristics were controlled, that dropping out was related to several variables. Specifically, characteristics of schools with high drop-out rates included low SES, high
student–teacher ratios, perceptions of poor quality of teaching, and
low teacher salaries. Public schools had significantly higher dropout rates than did Catholic schools or other private schools. However, the results concerning school size were surprising. Specifically, large-sized schools had lower drop-out rates than did smaller
sized schools.
Pianta (1999) noted that student–teacher ratios must be considered when examining relationships between students and teachers
in schools. Specifically, Pianta argued that in both regular and
special education classrooms, lower student–teacher ratios lead to
better communication and more positive interactions between
teachers and students and to closer monitoring of student progress
by teachers. In addition, from a Vygotskian perspective, Pianta
also argued that the teacher is more effectively able to operate
within individual children’s zones of proximal development when
student–teacher ratios are low.
Grade Configuration
Although there have been no studies to date that have examined
specifically the relations between grade configuration and perceived school belonging, it is plausible that certain configurations
are more conducive to the development of a sense of belonging
than are others. Specifically, some research indicates that schools
with larger grade spans and schools that educate both young
children and older adolescents simultaneously may be conducive
to more positive outcomes for adolescents than other types of
schools. In addition, some research suggests that feelings of belonging may be particularly low in typical middle-grade schools.
For example, there is some evidence that schools that contain
multiple grades and that also educate elementary school children
along with adolescents tend to be more developmentally appropriate for adolescents. For example, Simmons and Blyth (1987) found
that girls who attended schools with kindergarten–eighth-grade
configurations made a healthier transition into high school than did
girls who attended more typical middle schools (e.g., schools with
a Grade 6–8 configuration). Eccles and Midgley (1989) found that
typical middle schools (e.g., Grades 6–8 or 7–9) were associated
with declines in academic motivation for many adolescents.
E. M. Anderman and Kimweli (1997) found that adolescents
who attended schools with a kindergarten–Grade 8 or a kindergarten–Grade 12 type of grade configuration were less likely to report
being victimized, less likely to report getting into trouble for bad
behavior, and less likely to perceive their school as unsafe, compared with students in more traditional Grade 6–8 or 7–9 configuration schools. Other research (e.g., National Institute of Education, 1978) has demonstrated that violent behavioral problems
among students in the seventh–ninth grades are fewer when those
students are in schools with configurations of seventh–12th grade,
compared with more traditional middle-school grade configurations. Blyth, Thiel, Bush, and Simmons (1980) found that students
were victimized more often in schools with seventh–ninth-grade
configurations than in schools with kindergarten–eighth-grade
configurations. However, other studies examining other types of
outcomes have found the opposite pattern (e.g., Simmons & Blyth,
1987).
Urbanicity
Some research indicates that students in urban, rural, and suburban schools may have different types of educational experiences.
For example, some studies indicate that the academic achievement
of students in urban schools is lower than the achievement of
students in other schools (e.g., Eisner, 2001; National Assessment
of Educational Progress, 2001).
There has been some school-level research on nonacademic
outcomes comparing students in urban, rural, and suburban regions. E. M. Anderman and Kimweli (1997) found that students in
urban schools reported being victimized and perceiving their
schools as unsafe more than did students in suburban schools; they
also found that students in rural schools perceived their school
environments as more unsafe than did students in suburban
schools. Other research (e.g., Rumberger & Thomas, 2000) has
indicated that drop-out rates may be lower in urban schools than in
suburban schools.
A limited amount of research has specifically examined perceptions of belonging across these settings with mixed results. For
example, some research (e.g., Trickett, 1978) suggests that students who attend urban schools report a greater sense of belonging
or relatedness than do students who attend rural schools. However,
results of a recent comparative study by Freeman, Hughes, and
Anderman (2001) using an adapted version of Goodenow’s
(1993b) measure of belonging compared adolescents’ perceptions
of belonging in urban and rural schools. Results indicated that
perceptions of belonging were higher in rural schools than in urban
schools.
SCHOOL EFFECTS 797
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
School Contexts in Educational Psychology
In the present research, the relations of perceived school belonging to a variety of other psychological outcomes were examined. Reviews of the literature suggest that psychological phenomena are seldom examined contextually across different school
environments. To verify this observation, in addition to reviewing
all of the literature on school belonging, I examined all studies
published in the Journal of Educational Psychology and Contemporary Educational Psychology over a 5-year period (between
1995 and 1999) to explore the frequency of studies of children and
adolescents in educational psychology that incorporated more than
one school in their design. I did not examine the frequency of
studies that included institutions of higher education because the
present study only concerned students in kindergarten–12th-grade
schools.
The search indicated that a total of 428 articles were published
between 1995 and 1999 in those journals. Specifically, 135 articles
were published in Contemporary Educational Psychology, and 293
articles were published in the Journal of Educational Psychology.
An examination of the methodology sections of those studies
revealed that 105 of the 428 studies (24.5%) were studies of
children or adolescents that incorporated at least two or more
schools in the design of the study. Consequently, it appears that in
the field of educational psychology, researchers do examine phenomena across multiple school contexts in about 25% of published
studies; however, the relations of perceived school belonging to
various phenomena to date have not been examined across multiple school contexts.
The present series of studies were designed to examine schoollevel differences in perceived school belonging. Both studies used
data from the National Longitudinal Study of Adolescent Health
(Add Health). Study 1 was an examination of school-level differences in perceived school belonging. Specifically, characteristics
of schools that might be predictive of a perceived sense of belonging, after controlling for student characteristics, were examined.
Study 2 examined school-level differences in the relations between
school belonging and a variety of outcomes. The analyses focused
on psychological outcomes that have been identified as being
highly prevalent or problematic during adolescence, including
social rejection (e.g., Asher & Coie, 1990), depression and optimism (e.g., Hogdman, 1983; Peterson & Bossio, 1991; Reynolds,
1984), and behavioral problems (e.g., Caspi, Henry, McGee, Moffitt, & Silva, 1995).
For Study 1, the hypothesis that perceived school belonging
would be greater in schools with specific sizes, grade configurations, and locations was examined. Specifically, it was predicted
that after controlling for individual differences, a greater sense of
belonging would be associated with schools that were small in
size, with schools that used a kindergarten–Grade 8 or kindergarten–Grade 12 type of configuration, and with schools that were not
located in urban regions. In Study 2, the relations of school
belonging to other psychological outcomes were examined, controlling for student and school-level variables. Specifically, it was
predicted that the relations between perceived belonging and other
psychological outcomes would vary by school. In addition, it was
hypothesized that aggregated school belonging, grade configurations, school size, and urbanicity would be significant school-level
predictors of the outcomes and of the relations between belonging
and psychological outcomes.
Study 1
The purpose of Study 1 was to examine individual and schoollevel predictors of perceived school belonging. Although a variety
of studies have examined the positive relations of school belonging
with a variety of outcomes (e.g., L. H. Anderman & Anderman,
1999; Goodenow, 1993a, 1994b; Roeser et al., 1996), no studies to
date have examined school-level differences in belonging.
Method
Sample
Data for both studies came from Add Health. Data for Add Health were
collected from several sources, from 1994 through 1996. Initially, 132
schools that served adolescents were selected for participation. From those
schools, a large sample of students (N 90,118) completed in-school
questionnaires. In addition, a subsample of 20,745 students were interviewed in their homes in 1995 (14,738 were reinterviewed in 1996).
Administrators from the 132 schools also completed a school-administrator
survey describing various school characteristics.
For Study 1, the Add Health in-school survey data were used, with a
subsample size of 58,653 students from 132 schools. On the basis of
the suggestions of Raudenbush, Bryk, Cheong, and Congdon (2000),
listwise deletion of data at the student level was used; consequently, the
student sample in this data set had full data on all variables. The sample is
evenly divided in terms of gender (48.8% male, 51.2% female). The
sample is diverse in terms of ethnicity, with 1.5% of the sample being
Native American, 5.6% Asian–Pacific Islander, 15.0% African American,
and 6.3% being of other non-White racial groups. Some ethnic minority
groups were oversampled, but the oversampling of those groups is corrected through the use of weights. In addition, 14.0% of the sample
indicated that they were of Hispanic or Spanish origin. In terms of grade
level, 10.9% of the sample were in the seventh grade, 11.6% were in the
eighth grade, 20.1% were in the ninth grade, 20.8% were in the 10th
grade, 19.2% were in the 11th grade, and 17.4% were in the 12th grade.
The schools included in this study represent an array of diverse characteristics. Schools were divided among urban (32.6%), suburban (54.7%),
and rural (12.8%) locations. Most schools in the sample (90.1%) were
public schools. With regard to school size, 22.7% of the schools were small
sized (1–400 students), 45.3% were medium sized (401–1,000 students),
and 32.0% were large sized (1,001–4,000 students). In addition, 16.0% of
the participating schools (n 23) reported using busing practices (i.e.,
busing students to schools in other neighborhoods).
Measures
Scales were developed to measure perceived school belonging and
self-concept. Principal-components analyses with varimax rotations guided
all scale construction. All scales displayed good reliability. All items and
descriptive statistics are listed in full in Table 1.
Several demographic measures were included. Gender was coded as a
dummy variable, where 0 male and 1 female. Ethnicity was coded as
several dummy variables, where 0 not a member of ethnic group and 1
member of ethnic group. Dummy variables were created for African
American, Asian–Pacific Islander, Native American, and other race (European American served as the comparison group). Grade-level was represented by five dummy variables, with 12th grade serving as the comparison group. In subsequent hierarchical linear modeling (HLM) analyses,
these dummy-level variables were grand-mean centered (as were all other
predictor variables); consequently, the coefficients for the dummy vari798 ANDERMAN
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
ables in the HLM analyses are interpreted as the mean difference between
each group and the omitted group (e.g., European Americans). GPA was
the mean of students’ grades for English, mathematics, social studies, and
science, where 1 A, 2 B, 3 C, and 4 D or lower. GPA data were
omitted for students who did not take a particular subject or who indicated
that they did not know their current grades in that subject domain. All items
assessing GPA were reverse coded, so that a high GPA was indicative of
receiving high grades. Participants also indicated how many years they had
been a student at their present school (1 this is my 1st year, 2 this is
my 2nd year,…5 this is my 5th year, 6 I have been here more than 5
years). All continuous variables were transformed into z scores across
schools so that results could be reported as standard deviation units.
Construction of School-Level Variables
School-level variables were created from a school-administrator survey
that was completed by an administrator at each site. Several general
demographic variables were created. School size was coded as small
(1–400 students), medium (401–1,000 students), and large (1,001–4,000
students) on the basis of a priori categories. Dummy variables were created
for small- and large-sized schools (medium-sized schools served as the
comparison group). Class size was the actual average class size in whole
numbers, as reported by a school administrator. In addition, schools were
classified as urban, rural, or suburban. Dummy variables were created for
urban and rural schools, with suburban schools serving as the comparison
group. In addition, a dummy variable was created to compare Catholic
schools with other types of schools (i.e., public, private, other parochial)
because research suggests that Catholic schools often operate in a more
equitable manner than do other schools (Bryk et al., 1993). A dummy
variable also was included indicating whether or not the school used any
types of busing practices (0 does not use busing, 1 does use busing).
Schools were identified as using busing practices if the school administrator reported that the school assigned students from several geographic areas
to achieve a desired racial and/or ethnic composition of students or if the
school used busing practices to allow for transfers.
For the present study, schools were classified into two groups on the
basis of grade configurations. The first group (n 21) included schools
that educated young children in addition to adolescents; specifically, it
contained schools with a configuration of kindergarten–Grade 12 (n 14)
and kindergarten–Grade 8 (n 7). The other group consisted of all other
types of grade configurations. These included schools that served early
adolescents (n 51), schools with configurations of Grades 6–12 (n
15), and high schools with grade configurations of Grades 9–12 (n 70)
and Grades 10–12 (n 5).
Results and Discussion
Scale Development
The School Belonging items were analyzed using a principalcomponents analysis with a varimax rotation. One of the items
(“The students at this school are prejudiced”) did not load on the
School Belonging factor, so that item was dropped. The remaining
factor exhibited an eigenvalue of 2.71 and explained 45.21% of the
variance in the items. The items and descriptive statistics are
presented in Table 1. The scale displayed good internal consistency (Cronbach’s .78).
A self-concept scale was constructed from six items (see Table
1). A principal-components analysis indicated that the six items
formed one factor, explaining 58.95% of the variance in the items
(eigenvalue 3.54). The scale displayed good reliability (Cronbach’s .86). Because items were anchored with a scale where
1 strongly agree and 5 strongly disagree, the six items were
reverse coded so that a high score on the scale represented a
positive self-concept.
Descriptive statistics and correlations are presented for studentlevel variables in Table 2. Perceived school belonging was correlated positively with self-concept (r .57, p .01), GPA (r
.20, p .01), and parental education (r .09, p .01).
Multilevel Regressions
HLM (Bryk & Raudenbush, 1992) was used to examine the
nested structure of school belonging. HLM analyses proceeded in
three steps. First, the intraclass correlation (ICC), or betweenschools variance in perceived school belonging, was examined.
Second, student-level predictors of school belonging were examined (similar to a traditional ordinary least squares multiple regresTable 1
Items and Descriptive Statistics for Scales
Scale and item M SD Loading
School Belonging .78
I feel like I am part of this school. 2.49 1.22 .84
I am happy to be at this school. 2.47 1.25 .81
I feel close to people at this school. 2.48 1.15 .77
I feel safe in my school. 2.33 1.09 .64
The teachers at this school treat
students fairly. 2.62 1.15 .56
Self-Concept .86
I have a lot to be proud of. 4.11 0.95 .80
I like myself just the way I am. 3.83 1.11 .78
I feel loved and wanted. 3.94 1.00 .77
I feel socially accepted. 3.76 0.98 .76
I feel like I am doing everything
just right. 3.32 1.04 .76
I have a lot of good qualities. 4.16 0.85 .74
Table 2
Descriptive Statistics for In-School Sample
Variable M SD 1 2 3 45
1. Belonging 3.56 0.83 —
2. Self-concept 3.86 0.74 .57** —
3. Years at current school 2.50 1.39 .01 .01* —
4. GPA 2.84 0.79 .20** .12** .05** —
5. Parent education 4.28 1.50 .09** .06** .01 .24** —
Note. GPA grade point average.
* p .05. ** p .01.
SCHOOL EFFECTS 799
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
sion). Third, school-level variables were added to the model to
examine school-level predictors of perceived school belonging
while controlling for individual differences. The appropriate student weights were used in all HLM analyses; thus, results are
generalizable to the population of American adolescents. All predictor variables were grand-mean centered, as suggested by a number
of methodologists (e.g., Bryk & Raudenbush, 1992; Snijders &
Bosker, 1999). By grand-mean centering the predictor variables,
the intercept can be interpreted as the expected value for an
average student rather than for students who are coded as zero. In
HLM analyses, all continuous variables were standardized using z
scores prior to their inclusion in the HLM models. Consequently,
coefficients should be interpreted as standard deviation units,
similar to the interpretation of a beta in a traditional ordinary least
squares regression.
ICCs. As a first step, the variance between schools in perceived school belonging was examined. For this step, perceived
school belonging was entered into the HLM analysis as a dependent variable, with no predictors in the model. Results indicated
that a significant portion of the variance in perceived school
belonging lies between schools. Specifically, 7.95% of the variance occurs between schools, 2
(137, N 58,653) 4,225.44,
p .01.
Student-level model. A student-level model was run with characteristics of students as predictors of perceived school belonging.
The model is expressed by the following equation:
Individual school belonging 0j 1j gender
2j GPA 3j self-concept
4j years at present school 5j Hispanic ethnicity
6j African American 7j Asian–Pacific Islander
8j Native American 9j other race
10j Grade 7 11j Grade 8 12j Grade 9
13j Grade 10 14j Grade 11 ij.
The intercept was allowed to vary between schools. The slopes
for grade level and for ethnicity–race were fixed, whereas all other
slopes were allowed to vary randomly between schools.1 Results
are displayed in Table 3.
The strongest student-level predictor of perceived belonging
was self-concept ( .56, p .01). The gamma coefficient of .56
indicates that a 1-unit increase in self-concept produces a .56
standard deviation increase in perceived belonging. Other results
indicated that African American ( .24, p .01) and Native
American students ( .13, p .05) perceived less belonging
than did European American students. Seventh ( .25, p .01),
eighth ( .16, p .01), ninth ( .19, p .01), and 10th (
.09, p .01) graders reported greater perceptions of belonging
than did seniors. School belonging was related to gender, with girls
perceiving stronger senses of belonging than boys ( .07, p
.01). Belonging also was related positively to GPA ( .09,
p .01).
Full model. For the full model, school characteristics from the
Add Health school-administrators’ surveys were added to the
model as predictors of the intercept. This allowed for an examination of the relations between both student and school-level
characteristics and school belonging. School-level predictors were
not incorporated as predictors of other Level 1 parameters.
Several sets of school characteristics were examined. First,
schools with grade configurations of kindergarten–Grade 12 (i.e.,
schools that contained both young children and older students)
were compared with all other types of schools. Second, dummy
variables were included, comparing public schools and Catholic
schools with all other types of schools (e.g., private, parochial).
Third, dummy variables representing busing and geographic location of the school were included (urban and rural, with suburban as
the comparison). Fourth, several indices of school size were incorporated, including a measure of the average class size and
dummy variables representing school size (large and small, with
medium as the comparison). The between-schools model is expressed by the following equation:
0j 00 01 urban 02 rural 03 large
04 small 05 busing
04 kindergarten–Grade 12 configuration
05 average class size.
Results are presented in Table 4.
The variables representing Catholic and public schools were
dropped because neither were significant in the analysis. After
controlling for student-level variables, I found that belonging was
lower in schools that reported using busing practices compared
1 Grade level and ethnicity were fixed because some schools did not
contain large enough populations of certain ethnicities to estimate effects.
In addition, not all schools contained all grade levels. These parameters
were fixed to maximize the number of schools used to compute chi-square
statistics.
Table 3
Student-Level Hierarchical Linear Model Predicting School
Belonging Using In-School Survey With Design Weights
Variable SE
Intercept .05* .02
Gender .07** .01
Grade point average .09** .01
Self-concept .56** .01
Parental education .02** .01
Years at present school .01† .01
Hispanic–Latino American .01 .02
African American .24** .02
Asian–Pacific Islander .03 .02
Native American .13* .06
Other race .05* .02
Grade 7 .25** .04
Grade 8 .16** .03
Grade 9 .19** .03
Grade 10 .09** .02
Grade 11 .02 .02
Note. For gender, 0 male, 1 female; for all measures of ethnicity,
0 not a member of ethnic group, 1 member of ethnic group, with
European American as the comparison group.
† p .10. * p .05. ** p .01.
800 ANDERMAN
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
with those that did not ( .13, p .01). In addition, belonging
was lower in urban schools than in suburban schools ( .07,
p .01). Attending schools with the kindergarten–Grade 12 type
of configuration was modestly related to belonging ( .12, p
.10). School size was unrelated to perceived belonging. The model
explained 36.67% of the between-schools variance in the intercept.
Summary. In summary, results of Study 1 indicate that perceived school belonging does vary across schools. Perceived
school belonging is related to several individual difference variables. Specifically, higher perceived school belonging is associated with high self-concept. Ethnicity emerged as a predictor of
belonging for African Americans and Native Americans, each of
whom reported lower levels of perceived belonging than did
European Americans.
Several school-level characteristics are related to perceived
school belonging. The practice of busing was related to lower
levels of perceived belonging. Perceived belonging was significantly lower in urban schools than in suburban schools. Attending
a kindergarten–Grade 12 type of school was modestly related to
belonging, once other variables were controlled.
One of the questions that remains is whether perceived school
belonging is related to lower levels of psychological distress
among adolescents. More importantly, the significant ICC found in
the present study leads to the question of whether the relations
between belonging and other outcomes vary between schools.
Those questions are addressed in Study 2.
Study 2
The purpose of Study 2 was to examine the relations of perceived school belonging to various psychological outcomes. For
this study, the in-home interview portion of the Add Health study
was used (N 20,745 students, N 132 schools).
Method and Measures
The outcome variables included measures of depression, optimism,
social rejection, school problems, and GPA. Scaled predictors included
perceived school belonging and self-concept.
Items for scales are presented in Table 5. The Depression, Social
Rejection, and Optimism scales were anchored with four response categories (0 never or rarely, 1 sometimes, 2 a lot of the time, and 3
most of the time or all of the time). For the Self-Concept scale, participants
indicated how much they agreed with a series of statements (1 strongly
agree, 3 neither agree nor disagree, and 5 strongly disagree). For the
scale measuring school problems, students indicated how often during the
current school year they had trouble with various issues (e.g., getting along
with teachers, getting homework done). That scale was anchored with five
response categories (0 never, 1 just a few times, 2 about once a
week, 3 almost everyday, and 4 everyday). The items measuring
perceived school belonging were identical to those used in Study 1. Most
demographic items were treated identically to those in Study 1. Gender and
ethnicity were treated as dummy variables. For gender, 0 male and 1
female. For the measures of ethnicity, dummy variables were created for
African American, Native American, Asian–Pacific Islander, and other
race categories, with European Americans serving as the omitted comparison group (0 not a member of ethnic group, 1 member of ethnic
group). GPA was calculated the same way as in Study 1 (items were
identical across the two data sets). Five grade-level dummy variables were
included for all grades except the 12th grade (0 not in the grade, 1 in
the grade).
Parent education was the mean level of education for both resident
parents (if data were available for only one resident parent, then those data
were used). Parent education was recoded so that 0 never went to school,
1 eighth-grade education or less, 2 more than eighth-grade education
but did not graduate high school (or attended vocational or trade school
instead of high school), 3 high school graduate or completed a graduate
equivalency diploma, 4 went to business or trade school or some college,
5 graduated from a college or a university, and 6 professional or
training beyond a 4-year college or university. This measure is similar to
measures used in other large-scale research (e.g., Johnston, O’Malley, &
Bachman, 1992). Data on parental income were only provided for a
subsample of students; consequently, parental education was used because
it was the best available measure that would maximize the sample size.
To assess school absenteeism in the in-home interviews, respondents
indicated how many times they were absent from school for a full day with
an excuse. Response categories included 0 never, 1 one or two times,
2 3 to 10 times, and 3 more than 10 times.
During the in-home interview portion of the study, all participants
completed the Peabody Picture Vocabulary Test (Dunn & Dunn, 1997).
Scores on this test were included as a covariate.
All predictors were grand-mean centered in HLM analyses, as they were
in Study 1. Therefore, all intercepts may be interpreted as the mean level
for average students rather than as the value when all predictors are coded
as zero. Effects for dummy-level variables are interpreted as the mean
difference between each group represented by a dummy variable and the
omitted group. All continuous variables were transformed into z scores
across schools so that results could be reported as standard deviation units.
Table 4
Full Hierarchical Linear Model Predicting School Belonging
Using Full In-School Survey and Administrator Survey With
Design Weights
Variable SE
Intercept .05** .02
School-level predictors
Urban .07** .03
Rural .03 .04
Large .02 .03
Small .07 .06
Busing .13** .05
Kindergarten–Grade 12 configuration .12† .07
Average class size .03 .02
Student-level predictors
Gender .07** .01
Grade point average .09** .01
Self-concept .56** .01
Years at present school .01† .01
Hispanic–Latino American .02 .02
African American .24** .02
Asian–Pacific Islander .02 .02
Native American .12* .06
Other race .05† .02
Grade 7 .24** .03
Grade 8 .16** .03
Grade 9 .19** .03
Grade 10 .09** .02
Grade 11 .02 .02
Note. For gender, 0 male, 1 female; for all measures of ethnicity,
0 not a member of ethnic group, 1 member of ethnic group, with
European American as the comparison group. For the final model, 2
(124,
N 58,653) 1,797.67, p .01.
† p .10. * p .05. ** p .01.
SCHOOL EFFECTS 801
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
Results and Discussion
Scaling of Measures
Factor analyses were run to verify the uniqueness of the scaled
variables. All of the psychological measures were submitted to a
single analysis to examine the discriminate validity of the measures. Items were transformed into z scores for these analyses. A
principal-components analysis with a varimax rotation yielded a
six-factor solution. The unique factors that emerged from the
analysis represented Perceived School Belonging, School Problems, Depression, Optimism, Social Rejection, and Self-Concept.
The factors, eigenvalues, percentage of explained variance, loadings, reliability estimates for scales, and items are presented in
Table 5.
The Self-Concept and School Belonging scales were identical to
those used in Study 1. However, for the Self-Concept scale, one
additional item was added from the in-home interview data. That
item assessed participants’ perceptions of how physically fit they
perceived themselves to be. Internal consistency for the SelfConcept scale remained high (Cronbach’s .86).
Preliminary Analyses
Descriptive statistics and correlations are presented in Table 6.
Perceived school belonging was related positively and significantly ( p .01) to optimism (r .28), self-concept (r .36), and
GPA (r .21). Perceived belonging was related negatively and
significantly ( p .01) to depression (r .28), social rejection
(r .27), school problems (r .34), and absenteeism (r
.13).
Most of the scaled predictors and outcomes were distributed
normally. Two of the variables were somewhat skewed (depression skew 1.63 and social rejection skew 1.56) but not
enough to significantly affect results of the HLM models.
Multilevel Regressions
ICCs. First, intraclass correlations were calculated for the
outcomes tested in the HLM analyses as well as for perceived
school belonging. Listwise deletion of data was used, resulting in a sample size of n 15,457. Results are presented in
Table 7, adjusted for the reliability of the estimates. ICCs
for the outcomes ranged from a low of .027 to a high of
.102. All chi-square statistics were significant at p .01,
indicating that all of these outcomes varied significantly between schools. Consequently, complete HLM models were developed to examine student- and school-level predictors of the
outcomes.
Student-level models. Student-level HLM models were run for
all of the psychological outcomes (depression, optimism, social
rejection, and school problems). The within-school model is represented by the following equation:
Table 5
Factor Analysis and Reliability Analyses for Psychological Measures
Scale Eigenvalue % variance Item Loading
School Belonging .76 2.189 7.30 You feel like you are part of your school. .79
You feel close to people at your school. .78
You are happy to be at your school. .75
You feel safe in your school. .57
School Problems .69 1.375 4.58 Since the school year started, how often have you had trouble…
Paying attention in school? .77
Getting your homework done? .64
Getting along with your teachers? .64
Getting along with other students? .53
Depression .84 6.619 23.40 You felt depressed. .78
You felt you could not shake off the blues, even with help from
your family and friends.
.75
You felt sad. .72
You felt lonely. .67
You were bothered by things that usually don’t bother you. .62
You didn’t feel like eating, your appetite was poor. .54
You thought your life had been a failure. .51
You felt fearful. .50
You felt life was not worth living. .48
Optimism .71 1.480 4.93 You felt hopeful about the future. .75
You felt that you were just as good as other people. .67
You were happy. .62
You enjoyed life. .62
Social Rejection .67 1.154 3.85 People were unfriendly to you. .76
You felt that people disliked you. .73
Self-Concept .86 2.428 8.09 You have a lot to be proud of. .76
You like yourself just the way you are. .74
You have a lot of good qualities. .72
You feel like you are doing everything just about right. .68
You feel loved and wanted. .67
You feel socially accepted. .66
You feel physically fit. .63
802 ANDERMAN
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
Psychological outcome 0j
1j individual school belonging 2j gender
3j African American 4j Native American
5j Asian–Pacific Islander 6j other race
7j Hispanic ethnicity 8j parent education
9j Grade 7 10j Grade 8 11j Grade 9
12j Grade 10 13j Grade 11 14j absenteeism
15j Peabody Picture Vocabulary Test score
16j GPA 17j self-concept ij.
In addition, a fifth model predicting GPA was included to
compare the prediction of psychological outcomes with the prediction of a more traditional academic outcome. The within-school
model for GPA is represented by the following equation:
GPA 0j 1j individual school belonging
2j gender 3j African American
4j Native American 5j Asian–Pacific Islander
6j other race 7jHispanic ethnicity
8j parent education 9j Grade 7
10j Grade 8 11j Grade 9 12j Grade 10
13j Grade 11 14j absenteeism
15j Peabody Picture Vocabulary Test score
16j self-concept ij.
Background characteristics (ethnicity, parent education, grade
level, and gender) were controlled in all models, as were academic
and psychological characteristics (absenteeism, GPA, Peabody
Picture Vocabulary Test score, and self-concept). All parameters
were allowed to vary between schools, except ethnicity and the
grade-level dummy variables, which were fixed to maximize the
number of schools used in chi-square analyses. All variables were
grand-mean centered, as they were in Study 1. Results are displayed in Table 8.
Results indicate that perceived school belonging was related to
all outcomes: Higher levels of belonging were associated with
lower reported levels of depression ( .12, p .01), social
rejection ( .19, p .01), and school problems ( .25,
p .01), whereas higher levels of belonging were associated with
reports of greater optimism ( .10, p .01) and higher GPA
( .15, p .01).
Most background characteristics were unrelated to the outcomes, although girls reported higher levels of depression (
.21, p .01) and higher GPAs ( .36, p .01) and lower levels
of social rejection ( .06, p .05) and school problems (
.22, p .01) than did boys. Ethnicity was, for the most part,
Table 6
Correlations and Descriptive Statistics for In-Home Interview Data
Variable M SD 1 2 3 4 5 6 7 8 9 10
1. Belonging 3.74 0.80 —
2. Depression 0.41 0.44 .28 —
3. Optimism 2.00 0.68 .28 .45 —
4. Social rejection 0.41 0.55 .27 .45 .24 —
5. School problems 1.03 0.73 .34 .30 .19 .26 —
6. GPA 2.76 0.77 .21 .16 .23 .10 .34 —
7. Parental education 3.60 1.31 .06 .11 .16 .06 .00† .21 —
8. Absences from school 1.63 0.86 .13 .12 .05 .04 .14 .16 .04 —
9. PPVT 64.50 11.09 .04 .16 .23 .10 .01† .27 .31 .01† —
10. Self-concept 4.07 0.59 .36 .42 .46 .27 .23 .15 .09 .11 .02 —
Note. All correlations are statistically significant ( p .01), except those noted with a dagger (†). GPA grade point average; PPVT Peabody Picture
Vocabulary Test.
Table 7
Intraclass Correlations for In-Home Interview Dependent Variables
Variable 2 Reliability ICC 2
(126, N 15,547)
Belonging .056 .926 .850 .066 912.13**
Depression .027 .891 .747 .039 532.43**
Optimism .037 .949 .786 .047 633.55**
Social rejection .017 .947 .641 .027 393.31**
School problems .028 .969 .738 .038 518.93**
Grade point average .094 .918 .902 .102 1,426.66**
Note. ICC intraclass correlation.
** p .01.
SCHOOL EFFECTS 803
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
unrelated to the outcomes, although African American students
reported higher levels of depression ( .15, p .01) and social
rejection ( .12, p .01) and lower GPAs ( .19, p .01)
than did European American students. Asian–Pacific Islander students reported higher GPAs than did European American students
( .46, p .01).
Scores on the Peabody Picture Vocabulary Test were related
negatively and weakly to depression ( .11, p .01) and
social rejection ( .06, p .01), whereas the scores were
related positively to optimism ( .16, p .01), school problems
( .03, p .05), and GPA ( .26, p .01). Self-concept was
related positively to optimism ( .40, p .01) and GPA (
.10, p .01) and negatively to depression ( .32, p .01),
social rejection ( .22, p .01), and school problems (
.13, p .01).
Full models. School-level characteristics were modeled on the
intercept and on the school belonging slope; school-level variables
were not modeled on other Level 1 parameters. Urbanicity (rural,
urban, and suburban), school size (small, medium, and large),
average class size, busing practices, grade configuration (kindergarten–Grade 12 compared with others), and aggregated school
belonging were modeled on the intercept and school belonging
slope.2
The between-schools model for the intercept is expressed by the
following equation:
0j 00 01 aggregated school belonging
02 urban 03 rural 04 busing
05 kindergarten–Grade 12 configuration.
The between-schools model for variation in individual perceptions
of school belonging as a predictor of each outcome is expressed by
the following equation:
1j 10 11 aggregated school belonging
12 urban 13 rural 14 large-sized school
15 small-sized school 16 average class size.
Results are presented in Table 9 and are discussed separately for
each outcome.
Depression. Depression was higher in schools that reported
using busing practices than in those that did not use busing
practices ( .08, p .01). Aggregated school belonging was not
significantly related to depression.
The school belonging slope was related negatively to depression
( .12, p .01). However, the relation between individual
students’ perceived belonging and depression varied between
schools. Specifically, that effect was diminished in schools with
higher aggregated belonging ( .16, p .01). The negative
relation between perceived belonging and depression was less
strong in large schools than it was in medium-sized schools (
.07, p .01).
The strongest student-level predictors of depression were gender
( .21, p .01), with girls reporting greater levels of depression
than boys, and grade level, with seventh graders in particular
2 School size was dropped from the intercept model because it was not
significant in any of the models; the kindergarten–Grade 12 configuration
was dropped as a predictor of school belonging slope because it was not
significant in any of the models.
Table 8
Student-Level Hierarchical Linear Models Predicting Psychological Outcomes and Grade Point Average (GPA)
Variable
Depression Optimism Social rejection School problems GPA
SE SE SE SE SE
Intercept .04** .01 .04** .01 .02 .01 .01 .02 .08** .02
Background characteristic
Gender .21** .02 .01 .02 .06* .03 .22** .02 .36** .02
African American .15** .03 .06 .04 .12** .04 .08† .04 .19** .03
Native American .18† .11 .02 .13 .05 .11 .09 .13 .10 .13
Asian–Pacific Islander .08* .04 .10* .05 .04 .05 .02 .06 .46** .06
Other race .00 .05 .02 .05 .04 .05 .05 .06 .05 .06
Hispanic ethnicity .09* .04 .03 .04 .10** .04 .03 .06 .14** .05
Parent education .02* .01 .04** .01 .03** .01 .06** .01 .14** .01
Grade 7 .20** .03 .10** .04 .09** .04 .13** .04 .10* .04
Grade 8 .13** .03 .06† .04 .04 .03 .14** .04 .11** .04
Grade 9 .09** .03 .05 .03 .01 .03 .07* .03 .23** .04
Grade 10 .07† .04 .05 .04 .01 .04 .01 .03 .20** .03
Grade 11 .02 .03 .06* .03 .01 .04 .04 .03 .19** .04
Academic & psychological control
Individual School belonging .12** .01 .10** .01 .19** .01 .25** .01 .15** .01
Absenteeism .06** .01 .02† .01 .00 .01 .08** .01 .12** .01
PPVT score .11** .01 .16** .01 .06** .01 .03* .01 .26** .02
GPA .04** .01 .10** .01 .02 .01 .29** .01
Self-concept .32** .01 .40** .01 .22** .01 .13** .01 .10** .01
Note. PPVT Peabody Picture Vocabulary Test.
† p .10. * p .05. ** p .01.
804 ANDERMAN
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
reporting feeling less depressed than did seniors ( .20, p
.01). Students with higher self-concepts were less likely to report
feeling depressed ( .32, p .01). The model explained 27.14% of the between-schools variance in depression.
Optimism. Only two of the school-level predictors emerged as
being modestly related to optimism: Students reported lower levels
of optimism in schools that reported using busing practices (
.06, p .10) and slightly greater levels of optimism in kindergarten–Grade 12 types of schools ( .07, p .10).
Students’ individual perceived belonging emerged as a predictor
of optimism: Students who personally reported perceiving that
they belong in their schools reported being more optimistic (
.09, p .01). However, that relation varied by school. Students
reported being less optimistic when they attended urban schools
compared with suburban schools ( .06, p .01). In addition,
optimism was slightly higher when average class sizes within the
school were higher ( .03, p .01).
The only student-level predictor that stood out as a strong
predictor of optimism was self-concept ( .40, p .01):
Students who reported higher levels of self-concept reported being
more optimistic. The model explained 30.96% of the betweenschools variance in optimism.
Social rejection. Students reported experiencing greater social
rejection in schools with higher aggregated school belonging (
.13, p .05). In addition, the use of busing practices was associated with greater perceptions of social rejection ( .07, p .01).
Student-level self-reported school belonging emerged as a negative predictor of social rejection in the model ( .19, p
.01). However, that relation varied between schools. Specifically,
that relation was diminished in large-sized schools ( .10, p
.01). African American students reported feeling greater social
rejection than did European American students ( .12, p .01).
Students of Hispanic origin reported lower levels of social rejection than did majority students ( .11, p .01). Self-concept
was related negatively to social rejection ( .22, p .01). The
model explained 17.97% of the between-schools variance in social
rejection.
School problems. Aggregated school belonging also emerged
as a predictor of self-reported school problems ( .14, p .05).
Students who attended schools with greater aggregated school
Table 9
Full Hierarchical Linear Models Predicting Psychological Outcomes and Grade Point Average (GPA)
Variable
Depression Optimism Social rejection School problems GPA
SE SE SE SE SE
Intercept .05** .01 .04** .01 .02 .02 .02 .02 .08** .02
Aggregated belonging .03 .05 .00 .05 .13* .05 .14* .07 .34** .08
Urban .03 .04 .01 .03 .01 .03 .02 .04 .03 .05
Rural .03 .04 .03 .04 .01 .03 .02 .03 .02 .05
Busing .08** .03 .06† .04 .07* .03 .03 .05 .05* .08
Kindergarten–Grade 12 .02 .03 .07† .04 .04 .04 .02 .05 .03 .08
Individual School belonging .12** .01 .09** .01 .19** .01 .25** .01 .14** .01
Aggregated belonging .16** .04 .08† .04 .09 .06 .04 .05 .00 .05
Urban .01 .02 .06** .04 .03 .03 .04 .03 .05† .03
Rural .03 .03 .00 .03 .00 .04 .01 .04 .02 .03
Large .07** .02 .02 .03 .10** .03 .09** .03 .00 .03
Small .02 .02 .03 .03 .05 .04 .04 .03 .03 .03
Average class size .01 .01 .03** .01 .01 .02 .02 .01 .02* .01
Background characteristic
Gender .21** .02 .02 .02 .06* .03 .22** .02 .36** .02
African American .15** .03 .06 .04 .12** .04 .08† .04 .19** .03
Native American .18† .11 .03 .13 .04 .11 .10 .13 .09 .13
Asian–Pacific Islander .08* .03 .10† .05 .04 .05 .03 .06 .46** .06
Other race .00 .05 .02 .05 .04 .05 .05 .06 .06 .06
Hispanic ethnicity .09 .04 .02 .04 .11** .04 .03 .06 .14** .05
Parent education .02* .01 .04** .01 .03** .01 .06** .01 .14** .01
Grade 7 .20** .03 .11** .04 .07† .04 .11** .04 .12** .04
Grade 8 .14** .03 .07† .04 .02 .04 .13** .04 .13** .04
Grade 9 .09** .03 .05 .03 .01 .03 .07* .03 .23** .04
Grade 10 .06 .04 .05 .04 .01 .04 .01 .03 .20** .03
Grade 11 .02 .03 .06† .03 .01 .04 .04 .03 .19** .04
Academic & psychological control
Absenteeism .06** .01 .02† .01 .00 .01 .08** .01 .12** .01
PPVT score .11** .01 .16** .01 .06* .01 .03* .01 .26** .02
GPA .04** .01 .09** .01 .02 .01 .29** .01
Self-concept .32** .01 .40** .01 .22** .01 .13** .01 .10** .01
Final model 2
(119, N 15,457) 255.77** 331.95** 299.94** 355.90** 651.54**
% between-school variance
explained in intercept 27.14 30.96 17.97 29.67 24.66
Note. PPVT Peabody Picture Vocabulary Test.
† p .10. * p .05. ** p .01.
SCHOOL EFFECTS 805
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
belonging reported having more problems in school than did
students attending schools with lower levels of aggregated perceived belonging. In addition, self-reported school belonging was
related negatively to school problems ( .25, p .01). This
effect was less strong in large-sized schools compared with
medium-sized schools ( .09, p .01). Girls reported fewer
school problems than did boys ( .22, p .01). GPA was
related negatively to school problems ( .29, p .01). The
model explained 29.67% of the between-schools variance in
school problems.
GPA. The final model examined predictors of GPA; consequently, GPA was eliminated as a predictor in this model. A strong
effect emerged for aggregated school belonging: In schools with
higher aggregated perceived belonging, the GPAs of individual
students were significantly higher ( .34, p .01). Comparatively, this was the strongest effect of aggregated belonging in any
of the models. GPA was lower in schools that used busing practices ( .05, p .05).
Individual-level belonging also was related positively to GPA
( .14, p .01). That relation was slightly stronger in schools
with greater average class sizes ( .02, p .05). Female
students reported higher GPAs than did male students ( .36,
p .01). African American students ( .19, p .01) and
Hispanic students ( .14, p .01) reported lower GPAs than
did European American students, whereas Asian American students reported higher GPAs than did European American students
( .46, p .01). GPA was related positively to students’
Peabody Picture Vocabulary Test scores ( .26, p .01) and to
parent education ( .14, p .01). The model explained 24.66%
of the between-schools variance in GPA.
General Discussion
The present studies used data from two substudies of the Add
Health. The goal of Study 1 was to examine school-level variables
that were related to perceived school belonging. The goal of
Study 2 was to examine the relations of perceived school belonging to a variety of psychological outcomes across different types of
schools.
In Study 1, it was predicted that after controlling for other
variables, school size, grade configuration, and urbanicity would
be related to perceptions of belonging. The results partially supported this prediction. First, after other variables were controlled
for, school size was unrelated to perceptions of belonging. Other
studies (e.g., Lee & Smith, 1995) have found that attending smallsized schools is related to other positive outcomes, such as increased academic achievement. Nevertheless, other research (e.g.,
Rumberger & Thomas, 2000) indicates potential benefits of attending large-sized schools, such as lower drop-out rates. One
potential explanation for these findings may be that school size is
related to other types of outcomes (e.g., achievement, drop-out
rates) but that size in and of itself is unrelated to belonging. Indeed,
it may be that within larger sized schools, students are able to
develop a sense of belonging within smaller communities or subpopulations within the school, thus making aggregate measures of
school-size inappropriate as determinants of individual belonging.
The predictions about grade configuration were partially confirmed. The hypothesis that attending a school with a kindergarten–Grade 8 or Grade 12 type of configuration would be conducive
to greater perceived school belonging was found, albeit weakly
( p .10). Thus, students attending kindergarten–Grade 12 types
of schools reported a slightly greater sense of belonging than did
students attending other types of schools. Some research suggests
that schools with kindergarten–Grade 8 configurations are more
conducive to early adolescent development (e.g., Simmons &
Blyth, 1987). Nevertheless, results of the present study suggest that
after other variables have been controlled, grade configurations of
this nature are only weakly related to perceived belonging. As
suggested elsewhere (e.g., Anderman & Maehr, 1994), it is the
actual practices used within a school that are related to studentlevel outcomes and not the grade configuration per se. Nevertheless, because certain practices are associated with certain configurations (Simmons & Blyth, 1987), it is essential to control such
configurations in school-level studies.
The results for urbanicity indicated that after other student- and
school-level variables were controlled for, students’ perceived
sense of belonging was lower in urban schools than in suburban
schools. Whereas other studies have yielded similar results (e.g.,
Freeman et al., 2001), the present study is the first to examine
school belonging and urbanicity using nationally representative
data. Although it is possible that students in urban schools may
report lower levels of belonging because they may be drawn from
diverse regions within a city, the present results were found after
busing practices were controlled.
In Study 2, school belonging was examined as a predictor of
several psychological outcomes. School belonging was incorporated into the analyses both as the aggregated measures of school
belonging for each school (as a school-level variable) and as an
individual measure of belonging for each student (as a studentlevel variable).
Aggregated school belonging emerged as a predictor of the
outcomes in some of the models. Specifically, aggregated belonging was related positively to social rejection, school problems, and
GPA. The fact that higher levels of aggregated belonging are
related to increased reports of social rejection and school problems
is troubling. This suggests that in schools in which many students
feel that they do belong, those who do not belong may experience
more social rejection and problems in school. Research clearly
indicates that being accepted by one’s peers is fundamental to
healthy psychological development (e.g., Parker, Rubin, Price, &
DeRosier, 1995). However, results of the present study suggest
that when the overall level of belonging in a school is high, there
is an increased reporting of social rejection and problematic behaviors by individual students. Thus, when a school environment
is perceived of as supportive by many of its students, that supportive environment may be related to problematic psychological
outcomes for those students who do not feel supported.
The fact that aggregated school belonging emerged as a predictor in these models is important, particularly when examined in
conjunction with the student-level measure of school belonging.
Specifically, individual student-level reports of belonging were
used as predictors as well and were allowed to vary between
schools. As expected, results indicated that the individual measure
of perceived school belonging was related positively to adaptive
outcomes (e.g., optimism and GPA) but negatively to maladaptive
outcomes (e.g., depression, social rejection, and school problems).
It is interesting to note that the findings for depression were
moderated by aggregated school belonging. The negative (and
806 ANDERMAN
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
beneficial) relation between individual perceived belonging and
depression is essentially wiped out in schools with higher aggregated school belonging (i.e., the negative effect of individual
belonging [.12, p .01] essentially is canceled out by the
positive effect of aggregated belonging [.16, p .01]); thus, the
potential positive buffering effects of an individual student’s perception of belonging matter less in schools in which many students
report a high sense of belonging.
Individual difference predictors such as gender, grade level,
ethnicity, and achievement were significantly related to the outcomes in some of these models, but most of their effects were
modest once school-level variables were included as well. This
suggests that psychological phenomena during adolescence are
related in important ways to both school-level and individual-level
variables and that individual difference variables may not be as
predictive of these outcomes as once thought once school-level
phenomena have been controlled. This is an important finding
because school-level variables are, for the most part, contextual in
nature, and, thus, many of those variables may be altered through
reform efforts. The fact that aggregated school belonging emerged
as a predictor of many of these outcomes is encouraging from a
school-reform perspective because research has demonstrated that
school environments and climates can be altered to meet the
developmental needs of adolescents (e.g., E. M. Anderman,
Maehr, & Midgley, 1999; Battistich et al., 1997; Maehr & Midgley, 1996). Thus, policies and practices that are aimed at developing environments that foster a sense of belonging may lead to
the improved psychological well being and achievement of some
adolescents. However, those involved with school reform must be
cautious because an increased sense of belonging for some students at the exclusion of other students may lead to detrimental
outcomes for some students, such as greater levels of perceived
social rejection and greater reports of problems in school. In
addition, because the present research is correlational in nature,
future studies with longitudinal designs may better help to explain
these findings.
The results of Study 2 confirmed that the relation of an individual student’s perceived school belonging to psychological and
academic outcomes varies across schools. School size emerged as
the most consistent predictor of the relation between school belonging and the other psychological outcomes.
School size emerged as a predictor of the relations between
belonging and depression, belonging and social rejection, and
belonging and school problems. Specifically, the relations between
belonging and these outcomes were diminished in large-sized
schools compared with medium-sized schools. Although school
size emerged as a significant predictor, the advantageous effects
for smaller schools are probably due to the fact that school size is
related to other process variables. Lee, Bryk, and Smith (1993)
have argued that effects of school size are probably indirect.
Although school size often emerges as a significant predictor in
multilevel studies of academic achievement (e.g., Lee & Smith,
1995), it is possible that size is influencing other variables (e.g.,
social organization of the school) that ultimately are related to
student outcomes. Lee et al. (1997) suggested the formation of
schools-within-schools as a possible solution to the problems
associated with larger schools.
Other School Effects
One particularly salient finding of the present studies is that
mental health variables (e.g., depression, optimism) vary between
schools and are related to some school-structure variables.
Whereas psychological phenomena such as depression have been
linked to variables such as a family history of depression (e.g.,
Beardslee, Keller, & Klerman, 1985), genetic factors (e.g., Pike,
McGuire, Hetherington, Reiss, & Plomin, 1996), and parenting
behaviors (e.g., Ge, Best, Conger, & Simons, 1996), studies to date
have not demonstrated that these phenomena vary by school.
The present research does not examine processes that occur
within these differing schools that may lead to the observed
variations in psychological phenomena. Indeed, the purpose of the
present study was to investigate the relations of perceived school
belonging to these outcomes while controlling for individual and
school characteristics. Nevertheless, the use of statistical tools
such as HLM allowed for the analysis of school effects on several
psychological outcomes. It is most probable that the observed main
effects for school characteristics are a result of complex dynamics
between individual and contextual variables that were not captured
in the present study. Nevertheless, the present study represents a
first attempt to demonstrate that individual-level psychological
phenomena can be predicted by school-level characteristics during
adolescence.
Limitations
The present study has a number of limitations. First, the data
that were collected were part of a larger study of adolescent health
and well being. Thus, it was not possible to include some of the
measures that would have yielded greater insights into the processes that might explain some of the observed findings. For
example, whereas school size emerged as a predictor, it was not
possible to examine the more intricate processes in large- and
small-sized schools that might be responsible for the observed
patterns. Second, the student-level data were self-reported. The
reliability and validity of student self-report data always are somewhat problematic in school-based research. However, because the
in-home interviews were conducted using computer-assisted interviews that allowed for confidentiality of responses, reliability and
validity are probably not greatly affected in Study 2. Third, results
of all HLM analyses were reported in standard deviation units (in
z-score format) to provide a standardized metric (z score) for
interpretation. Nevertheless, as noted by Pedhazur (1997), there is
still much debate about whether it is best to report results of
regression-based analyses in standardized or unstandardized formats. In the present study, standardized coefficients were chosen
so that readers could make basic comparisons among the predictors. The use of unstandardized coefficients in HLM studies is
further complicated by the multiple units of analysis (i.e., student
and school). Therefore, a decision was made to use standardized
coefficients in the present study. Nevertheless, because the size of
standardized coefficients is influenced by the variances and covariances of both other variables in the model as well as variables
not included in the model, the use of standardized coefficients to
compare the relative effects of predictors is somewhat limited.
Fourth, the percentage of variance in the outcomes that lies between schools was not very large for some of the outcomes in
SCHOOL EFFECTS 807
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
Study 2. Nevertheless, all ICCs were statistically significant, and
the final models did explain reasonable amounts of the variance
between schools (from a low of 17.97% to a high of 36.67%).
Although the percentage of between-schools variance in some
outcomes was low, this is the first study to attempt to examine
these psychological phenomena from both student- and schoollevel perspectives. The fact that school-level predictors emerged as
significant in Study 2 and between-schools variance was explained
by the HLM models suggests that school-level phenomena do
contribute to students’ psychological well being and certainly are
worthy of further study. Fifth, it is possible that belonging could be
a consequence of the other psychological variables rather than a
possible protective factor against those outcomes. Because these
data are correlational, this possibility could not be examined directly. However, there is strong support in the educational psychology literature for the idea that the promotion of developmentally and psychologically appropriate learning environments (e.g.,
environments that foster a sense of belonging among students) lead
to adaptive psychological and mental health outcomes (e.g., Battistich et al., 1997). In addition, some quasi-experimental research
(e.g., E. M. Anderman et al., 1999; Maehr & Midgley, 1996;
Weinstein, 1998) indicates that reform efforts aimed at changing
schools’ psychological environments can lead to adaptive outcomes, including enhanced motivation, for adolescents.
Nevertheless, the present studies do make several unique contributions to the literature. First, both studies incorporated appropriate design weights; consequently, results are generalizable to
the full population of American adolescents. Second, these studies
used appropriate statistical techniques (HLM) to examine schoollevel differences in belonging and belonging’s relation to various
outcomes. Whereas numerous other studies have demonstrated that
the perception of belonging in schools is related to positive outcomes for students, this is the first study to do so using multilevel
statistical techniques that separate within and between-schools
variance. Third, the present research indicates that studies of
school belonging must acknowledge both students’ individual perceptions of belonging as well as aggregated belonging within
particular schools because when measured at different levels, these
variables relate in different ways to psychological outcomes.
Results of the present studies offer several implications for
practitioners. First, educators must realize that psychological phenomena can vary as a function of school environments. Thus,
issues such as depression or social rejection truly may be more
salient in some schools than in others. Second, the issue of school
size must be examined in greater detail. Whereas it often is
difficult to build smaller schools because of financial considerations, the problems associated with large schools cannot be ignored. Some research (e.g., Lee & Smith, 1995) indicates that the
use of schools-within-schools may be a reform that is both plausible and cost effective. Third, the issue of busing must be examined in greater depth. Whereas some research indicates positive
effects of busing (e.g., Mathews, 1998), results of the present
research indicate that busing may be related to increased psychological distress for some adolescents. Fourth, the present studies
emphasize the importance of individual perceived school belonging as a possible defense against negative psychological outcomes
for adolescents. If school personnel can do more to create caring
communities for adolescents (e.g., Battistich et al., 1997), then
students might be less likely to experience various forms of negative affect and psychological distress. Finally, results of the
present study suggest that when many students in a school experience a perception of belonging (i.e., there is a high aggregated
perception of belonging), some students who do not feel supported
in those environments may experience additional social rejection
and school problems.
References
Anderman, E. M., & Kimweli, D. (1997). Victimization and safety in
schools serving early adolescents. Journal of Early Adolescence, 17,
408–438.
Anderman, E. M., & Maehr, M. L. (1994). Motivation and schooling in the
middle grades. Review of Educational Research, 64, 287–309.
Anderman, E. M., Maehr, M. L., & Midgley, C. (1999). Declining motivation after the transition to middle school: Schools can make a difference. Journal of Research and Development in Education, 32, 131–147.
Anderman, L. H., & Anderman, E. M. (1999). Social predictors of changes
in students’ achievement goal orientations. Contemporary Educational
Psychology, 25, 21–37.
Asher, S. R., & Coie, J. D. (Eds.). (1990). Peer rejection in childhood. New
York: Cambridge University Press.
Battistich, V., Solomon, D., Watson, M., & Schaps, E. (1997). Caring
school communities. Educational Psychologist, 32, 137–151.
Baumeister, R. F., & Leary, M. R. (1995). The need to belong: Desire for
interpersonal attachments as a fundamental human motivation. Psychological Bulletin, 117, 497–529.
Beardslee, W. R., Keller, M. B., & Klerman, G. D. (1985). Children of
parents with affective disorders. International Journal of Family Psychiatry, 6, 283–299.
Blyth, D. A., Thiel, K. S., Bush, D. M., & Simmons, R. G. (1980). Another
look at school crime. Youth and Society, 11, 369–388.
Boekaerts, M. (1993). Being concerned with well-being and with learning.
Educational Psychologist, 28, 149–167.
Bryk, A. S., Lee, V. E., & Holland, P. B. (1993). Catholic schools and the
common good. Cambridge, MA: Harvard University Press.
Bryk, A. S., & Raudenbush, S. W. (1992). Hierarchical linear models:
Applications and data analysis methods. Newbury Park, CA: Sage.
Caspi, A., Henry, B., McGee, R. O., Moffitt, T. E., & Silva, P. A. (1995).
Temperamental origins of child and adolescent behavior problems: From
age three to age fifteen. Child Development, 66, 55–68.
Coleman, J. S., & Hoffer, T. (1987). Public and private high schools: The
impact of communities. New York: Basic Books.
Cowen, E. L. (1991). In pursuit of wellness. American Psychologist, 46,
404–408.
Deci, E. L., Vallerland, R. J., Pelletier, L. G., & Ryan, R. M. (1991).
Motivation and education: The self-determination perspective. Educational Psychologist, 26, 325–346.
Dunn, L. M., & Dunn, L. M. (1997). Peabody Picture Vocabulary Test.
Circle Pines, MN: AGS Publishing.
Eccles, J. S., & Midgley, C. (1989). Stage/environment fit: Developmentally appropriate classrooms for early adolescents. In R. E. Ames & C.
Ames (Eds.), Research on motivation in education (Vol. 3, pp. 139–
186). New York: Academic Press.
Edmonds, R. (1979). Effective schools for the urban poor. Educational
Leadership, 37, 15–24.
Eisner, C. (2001). Making the grade: A report on SAT I results in the
nation’s urban schools. Washington, DC: Council of the Great City
Schools.
Finn, J. (1989). Withdrawing from school. Review of Educational Research, 59, 117–142.
Freeman, T., Hughes, H., & Anderman, L. H. (2001, April). Changes in
students’ school belonging in rural and urban middle schools. Poster
808 ANDERMAN
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
presented at the annual meeting of the American Educational Research
Association, Seattle, WA.
Ge, X., Best, K. M., Conger, R. D., & Simons, R. L. (1996). Parenting
behaviors and the occurrence and co-occurrence of adolescent depressive symptoms and conduct problems. Developmental Psychology, 32,
717–731.
Goldschmidt, P., & Wang, J. (1999). When can schools affect dropout
behavior? A longitudinal multilevel analysis. American Educational
Research Journal, 36, 715–738.
Goodenow, C. (1993a). Classroom belonging among early adolescent
students: Relationships to motivation and achievement. Journal of Early
Adolescence, 13, 21–43.
Goodenow, C. (1993b). The psychological sense of school membership
among adolescents: Scale development and educational correlates. Psychology in the Schools, 30, 79–90.
Goodenow, C., & Grady, K. E. (1993). The relationship of school belonging and friends’ values to academic motivation among urban adolescent
students. Journal of Experimental Education, 62, 60–71.
Hogdman, C. H. (1983). Current issues in adolescent psychiatry. Hospital
and Community Psychiatry, 34, 514–521.
Johnston, L. D., O’Malley, P. M., & Bachman, J. G. (1992). The 1991
survey of drug use by American high school and college students. Ann
Arbor, MI: University of Michigan, Institute of Social Research.
Lee, V. E., Bryk, A. S., & Smith, J. B. (1993). The effects of high school
organization on teachers and students. In L. Darling-Hammond (Ed.),
Review of research in education (Vol. 19, pp. 171–268). Washington,
DC: American Educational Research Association.
Lee, V. E., Chow-Hoy, T. K., Burkam, D. T., Geverdt, D., & Smerdon,
B. A. (1998). Sector differences in high school course taking: A private
school or Catholic school effect? Sociology of Education, 71, 314–335.
Lee, V. E., & Smith, J. B. (1995). Effects of high school restructuring and
size on early gains in achievement and engagement. Sociology of Education, 68, 241–270.
Lee, V. E., Smith, J. B., & Croninger, R. G. (1997). How high school
organization influences the equitable distribution of learning in mathematics and science. Sociology of Education, 70, 128–150.
Maehr, M. L., & Midgley, C. (1996). Transforming school cultures.
Boulder, CO: Westview Press.
Mathews, J. (1998). Breaking down barriers. School Administrator, 55(8),
34–36.
Miller, S. K. (1985). Research on exemplary schools: A historical perspective. In G. Austin & H. Gatber (Eds.), Research on exemplary schools
(pp. 3–30). New York: Academic Press.
Murphy, J., Weil, M., Hallinger, P., & Mitman, A. (1985). School effectiveness: A conceptual framework. Educational Forum, 49, 361–374.
National Assessment of Educational Progress. (2001). The Nation’s report
card: Fourth-grade reading, 2000. Washington, DC: National Center for
Education Statistics.
National Institute of Education. (1978). Violent schools–safe schools: The
safe school study report to the Congress, Volume I. Washington, DC:
United States Department of Health, Education, and Welfare.
Newman, B. M., Lohman, B. J., Newman, P. R., Myers, M. C., & Smith,
V. L. (2000). Experiences of urban youth navigating the transition to
ninth grade. Youth and Society, 31, 387–416.
Parker, J. G., Rubin, K. H., Price, J. M., & DeRosier, M. E. (1995). Peer
relationships, child development, and adjustment: A developmental psychopathology perspective. In D. Cicchetti & D. J. Cohen (Eds.), Developmental psychopathology, Vol. 2: Risk, disorder, and adaptation (pp.
96–161). New York: Wiley.
Pedhazur, E. J. (1997). Multiple regression in behavioral research: Explanation and prediction (3rd ed.). Fort Worth, TX: Harcourt Brace.
Peterson, C., & Bossio, L. M. (1991). Health and optimism. New York:
Free Press.
Pianta, R. C. (1999). Enhancing relationships between children and teachers. Washington, DC: American Psychological Association.
Pike, A., McGuire, S., Hetherington, E. M., Reiss, D., & Plomin, R. (1996).
Family environment and adolescent depressive symptoms and antisocial
behavior: A multivariate genetic analysis. Developmental Psychology, 32, 590–603.
Raudenbush, S. W., Bryk, A. S., Cheong, Y., & Congdon, R. T. (2000).
HLM 5: Hierarchical linear and nonlinear modeling. Chicago: Scientific
Software International.
Reynolds, W. M. (1984). Depression in children and adolescents: Phenomenology, evaluation, and treatment. School Psychology Review, 13,
171–182.
Roeser, R. W. (1998). On schooling and mental health: Introduction to the
special issue. Educational Psychologist, 33, 129–133.
Roeser, R. W., Eccles, J. S., & Strobel, K. R. (1998). Linking the study of
schooling and mental health: Selected issues and empirical illustrations
at the level of the individual. Educational Psychologist, 33, 153–176.
Roeser, R. W., Midgley, C., & Urdan, T. (1996). Perceptions of the
psychological environment and early adolescents’ psychological and
behavioral functioning in school: The mediating role of goals and
belonging. Journal of Educational Psychology, 88, 408–422.
Rumberger, R.W. (1995). Dropping out of middle school: A multilevel
analysis of students and schools. American Educational Research Journal, 32, 583–625.
Rumberger, R. W., & Thomas, S. L. (2000). The distribution of dropout
and turnover rates among urban and suburban high schools. Sociology of
Education, 73, 39–67.
Rutter, M. (1980). School influences on children’s behavior and development: The 1979 Kenneth Blackfan Lecture, Children’s Hospital Medical
Center, Boston. Pediatrics, 65, 208–220.
Simmons, R. G., & Blyth, D. A. (1987). Moving into adolescence: The
impact of pubertal change and school context. New York: Aldine de
Gruyter.
Snijders, T. A. B., & Bosker, R. J. (1999). Multilevel analysis: An introduction to basic and advanced multilevel modeling. Thousand Oaks,
CA: Sage.
Trickett, E. J. (1978). Toward a social–ecological conception of adolescent
socialization: Normative data on contrasting types of public school
classrooms. Child Development, 49, 408–414.
Weinstein, R. S. (1998). Promoting positive expectations in schooling. In
N. M. Lambert & B. L. McCombs (Eds.), How students learn: Reforming schools through learner-centered education (pp. 81–111). Washington, DC: American Psychological Association.
Received October 31, 2000
Revision received December 19, 2001
Accepted December 19, 2001
SCHOOL EFFECTS 809
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.