Original article

High self-esteem among adolescents: Longitudinal trends, sex

differences, and protective factors
Susan Birndorf, D.O., M.P.H.
a,
*, Sheryl Ryan, M.D.
b
, Peggy Auinger, M.S.
b
, and
Marilyn Aten, R.N., Ph.D.
c
a
Afliation at time of study: Division of Adolescent Medicine at the University of Rochester, Rochester, New York
b
Department of Pediatrics, Rochester General Hospital, Rochester, New York
c
School of Nursing at the University of Rochester, Rochester, New York
Manuscript received February 2, 2004; Manuscript accepted August 15, 2004
Abstract Purpose: To better characterize self-esteem and to identify factors in early adolescence associated
with high self-esteem in later adolescence.
Methods: The National Education Longitudinal Study provided a random, nationally representa-
tive sample of 8th grade adolescents who completed surveys in grades 8, 10, and 12 (n 16,489).
We measured changes in self-esteem, and high versus low self-esteem. By using the Developmental
Assets Model, predictor variables were selected to include intrinsic developmental assets (e.g.,
decision making) and extrinsic assets (e.g., safety). Logistic regression with separate modeling for
sex identied assets that were predictive of high self-esteem, controlling for sociodemographic
factors.
Results: Boys versus girls were more likely to report high self-esteem in all grades (all p .001).
For girls, factors predictive of high self-esteem in grade 12 were as follows: African-American (odds
ratio [OR] 2.06) or Hispanic (OR 1.52) race/ethnicity, positive family communication (OR
3.03), safety (OR 1.50), and high self-esteem (OR 3.03). For boys, factors were family income
being above the federal poverty level (OR 1.50), positive family communication (OR 2.40), safety
(OR 1.41), religious community (OR 1.21), and high self-esteem (OR 3.01) (all p .05).
Conclusions: Fewer adolescent girls than boys report high self-esteem. Adults may foster self-
esteem in adolescents by providing positive communication through supportive and caring
relationships. 2005 Society for Adolescent Medicine. All rights reserved.
Keywords: Self-esteem; Adolescents; Adolescence
Yonder puny fellow, however, whom everyone can beat,
suffers no chagrin about it, for he has long ago aban-
doned the attempt to carry that line, as the merchants
say, of self at all. With no attempt there can be no
failure; with no failure no humiliation. So our self-
feeling in this world depends entirely on what we back
ourselves to be and do. It is determined by the ratio of
our actualities to our supposed potentialities; a fraction
of which our pretensions are the denominator and the
numerator our success: thus, self-esteem success/
pretensions.
William James, 1893 [1]
The notion that self-esteem affects wellness is not new.
In the original description of self-esteem, psychologist Wil-
liam James wrote, So, our feeling in this world depends
entirely on what we back ourselves to be and do [1]. High
self-esteem has been shown to be associated with better
mental health outcomes, such as a better ability to cope, and
also to be associated with a lower incidence of depression in
both adolescents and adults [25]. Beyond its role in mental
health, however, self-esteem has proven to be disappointing
both as a predictor of physical disease and as a predictor of
*Address correspondence to: Susan Birndorf Roberts, D.O., M.P.H.,
Kaiser Permanente, 7300 North Fresno Street, Fresno, CA 93720.
E-mail address: roberts1626@comcast.net
Journal of Adolescent Health 37 (2005) 194201
1054-139X/05/$ see front matter 2005 Society for Adolescent Medicine. All rights reserved.
doi:10.1016/j.jadohealth.2004.08.012
risk behaviors, such as smoking initiation and alcohol use,
which predispose adolescents and adults to poor physical
health [6,7]. These conicting results are surprising consid-
ering that external factors such as social supports and social
networks have been shown to affect longevity positively
and to protect against morbidities such as cerebrovascular
disease among adults [8,9]. Just as these external factors are
presumed to buffer and mediate stressful life events, high
self-esteem has been theorized to provide adults with added
individual strength. This results in a better ability to cope
and thus a lower incidence of disease [10,11].
A major limitation to the generalizability of ndings that
relate self-esteem to health outcomes is a lack of informa-
tion characterizing self-esteem across time and between
genders for all age groups. In addition, little information
exists that explains which factors contribute to the develop-
ment and maintenance of high self-esteem over time for
children and adolescents. Further information on the self-
esteem construct would aid in our understanding of whether
or not communities, families, schools, and clinicians may be
able to affect self-esteem, considering its possible mediating
effect on mental health outcomes such as emotional distress
and depression.
One model that holds promise as a framework for as-
sessing the relationship between internal characteristics
such as self-esteem and external factors such as social
support is the Developmental Assets Model [12]. This
model proposes that potential protective assets for children
and adolescents consist of the intrinsic strengths of an in-
dividual within the context of the environmental inuences
of community, school, and family. Forty developmental
assets are thought to contribute collectively to the success
and wellness of children and adolescents. The assets are
divided into 2 broad categories: intrinsic and extrinsic. In-
trinsic assets are characterized by qualities that adolescents
may possess inherently, such as social competency or pos-
itive identity. Extrinsic assets are those characterized by
environmental aspects of a young persons life such as
family and school support, boundaries, and expectations.
The objective of our study was to characterize self-
esteem better in adolescents across time for both boys and
girls. We also examined those modiable community and
individual factors that may predict or inuence high self-
esteem and the maintenance of high self-esteem in adoles-
cents over time. Based on existing qualitative and cross-
sectional research [1316], we hypothesized that adolescent
girls would report lower self-esteem than boys at each grade
level, African-American race would predict higher self-
esteem, and Hispanic ethnicity would predict lower self-
esteem. In addition, we predicted that several developmen-
tal assets from both the internal and external categories
would be associated with the maintenance of high self-
esteem in adolescence.
Methods
Study design and data source
This study was a prospective cohort design following-up
a group of teenage students over a 5-year period. Informa-
tion on adolescents was obtained by using the 1988 National
Education Longitudinal Survey, a nationally representative
sample of public school students in grades 8 through 12 that
was designed primarily to assess educational outcomes [17].
Participating schools were selected randomly and included
815 public and 237 private schools. A total of 26,432
students within the 1,052 schools then were selected ran-
domly to provide an eligible sample of 24,599 students in
the 8th grade who participated in the rst administration of
the survey in 1988. In addition, 22,651 parents (1 parent per
student), 5,193 teachers, and 1,035 school administrators
completed questionnaires in 1988. Students answered a
broad range of questions regarding topics including school
work, aspirations, and social relationships. Students in-
cluded in the core 1988 survey were tracked longitudinally
and completed similar surveys again in 1990 and 1992. Our
main predictor and outcome variables of interest were ob-
tained through adolescent self-reports from all 3 survey
waves. Information on demographics was obtained through
both the adolescent and parent surveys.
Sample
Of the 24,599 total eligible students in the rst survey in
8th grade, 16,489 students completed subsequent surveys in
10th and 12th grade and were included in the analysis. A
total of 8,110 (33%) students in the 8th grade did not
complete 1 or more follow-up surveys and were considered
lost to follow-up evaluation. The 2 groups differed statisti-
cally along a number of demographic characteristics and
developmental assets (Table 1). Of note is that the percent
of adolescents reporting high self-esteem in the group lost to
follow-up evaluation compared with those followed-up for
the entire period did not differ.
Outcome and explanatory variables
The main outcome variable, high (vs. low) self-esteem,
was measured using adolescent self-report on 7 items from
Rosenbergs [18] self-esteem scale. Rosenbergs [18] scale
has been validated extensively for high school students and
is the most widely accepted self-esteem scale for this pop-
ulation [18,19]. The internal consistency of this measure in
the study sample using Cronbachs was .81, which is
consistent with reported reliabilities in diverse populations
ranging from an value of .72 to .87. High self-esteem was
dened as respondents scoring in the highest tertile of the
self-esteem scale at each survey wave. Low self-esteem was
dened as the remainder of the sample. Although not the
most common way to express this scale, it remains consis-
tent with Rosenbergs [18] ndings that subjects reporting
195 S. Birndorf et al. / Journal of Adolescent Health 37 (2005) 194201
high and medium self-esteem group together clinically in
terms of subjective reports of unhappiness, gloom, and
discouragement, as well as possessing an outward appear-
ance of gloominess.
To guide the selection of explanatory variables that po-
tentially affect the development and maintenance of high
self-esteem, we chose the Developmental Assets Model as a
theoretical framework [12]. By using the National Educa-
Table 1
Percent of baseline characteristics in our sample compared with subjects lost to follow-up evaluation
Subject characteristics Study
sample
n 16,489
Subjects lost to follow-up evaluation
n 8,110 p value
Gender
Male 50.2 51.2 .31
Female 49.8 48.8
Race
Hispanic 10.4 11.8* .001
African-American 13.2 19.1
Other 4.7 6.1
Caucasian 71.7 63.0
Parental education
High school education 20.7 20.8 .17
Less than high school education 10.8 12.0
More than high school education 68.4 67.2
Poverty
At or above federal poverty level 84.8 81.7* .001
Below federal poverty level 15.2 18.3
Household composition
Single parent 21.1 28.7* .001
Two parents 78.9 71.3
Internal assets
Hours spent on homework
5.5 h/wk 67.8 69.3 .09
5.5 h/wk 32.2 30.7
Planning and decision making
Has plans after high school graduation 10.1 11.9* .001
Plans to graduate high school 88.4 86.1
Will not graduate high school 1.5 2.0
Reading for pleasure
None 20.3 22.1 .07
12 h/wk 53.1 51.6
3 h/wk 26.6 26.3
High self-esteem, grade 8 33.3 33.8 .59
External assets
Positive family communication 96.3 94.3 .054
Other adult relationships 67.7 64.4* .001
Caring school climate 97.0 95.8* .001
Parent involvement in schooling 98.7 98.0* .02
Safety 88.3 85.6* .001
Family boundaries
Often 42.6 42.8* .003
Sometimes 31.2 28.7
Never 26.3 28.5
School boundaries 99.6 99.3* .01
Adult role models 81.2 78.4* .001
Religious community 37.0 32.8* .001
Creative activities 57.5 52.5* .001
Youth programs
None 15.8 19.2* .001
110/y 80.7 76.0
10/y 3.5 4.8
Youth as resources
Work 20 h/wk 93.6 93.7 .86
Work 20 h/wk 6.4 6.3
* p .05.
196 S. Birndorf et al. / Journal of Adolescent Health 37 (2005) 194201
tion Longitudinal Survey, we were able to capture adoles-
cent self-report of 16 of the 40 potential assets elucidated by
this theoretical model. Asset variables were dened either
by a single item or a composite of items from the adolescent
responses to the survey questions (Figure 1). The predictive
effects of social and economic demographics also were
examined. These variables included gender, race/ethnicity
(self-report of Caucasian, African-American, Hispanic, or
other, which included Asian Pacic-Islander, American-
Indian, and multiple races), household composition, respon-
dents parents level of education, family size, and family
income above versus below the federal poverty level.
Analyses

2
analyses were performed to assess bivariate relation-
ships (signicant at p .10) between assets, sociodemo-
graphic characteristics, and self-esteem. Those relationships
signicant at the .10 level then were included in logistic
regression analyses modeling the effects of both internal
and external assets on the outcome of high versus low
self-esteem, controlling for sociodemographic factors. Sep-
arate modeling for boys and girls was performed to account
for known gender differences in self-esteem. SUDAAN
software (Research Triangle Park, NC) was used to account
for the complex sampling design of the survey [20].
Results
Prole of sample
Of the total sample, 50.2% were boys; 71.7% reported
Caucasian (non-Hispanic) race, 10.4% reported Hispanic,
13.2% reported African-American, and 4.7% identied as
other. The mean ( standard error) number of assets re-
ported by students in the sample was 12.4 .04 (range
116). A signicantly higher percentage of boys reported
high self-esteem as compared with girls at all grade levels
(Table 2). In grade 8, 39.2% of boys versus 27.4% of girls
reported high self-esteem (relative risk 1.4; 95% con-
dence interval [CI] 1.31.5). In grades 10 and 12, boys
remained signicantly more likely to report high self-es-
teem (relative risk 1.2; 95% CI 1.11.3), although the
disparity between boys and girls decreased at each subse-
quent survey wave.
Bivariate analyses
Bivariate relationships among assets, sociodemographic
characteristics, and self-esteem that were found to be sig-
nicant at the p less than .10 level for either boys or girls
using
2
analyses included the following: hours spent on
homework, planning and decision making, high self-esteem,
positive family communication, other adult relationships,
caring school climate, parental involvement in school, fam-
ily boundaries, school boundaries, adult role models, safety,
and religious community. These factors were included in the
nal logistic regression models. Factors not signicant for
either gender, and therefore not included, were reading for
pleasure, youth as resources, creative activities, and youth
programs.
Sociodemographic factors predictive of high self-esteem
For girls, those of Hispanic ethnicity or African-Ameri-
can race were 1.52 (95% CI 1.152.02) and 2.06 (95% CI
1.502.85) times more likely, respectively, to report high
self-esteem in grade 12, as compared with Caucasian female
adolescents (Table 3). Boys (odds ratio [OR] 1.50; 95%
CI 1.142.00) and girls (OR 1.31; 95% CI 1.02
1.69) at or above the federal poverty level were signicantly
more likely to report high self-esteem. Sociodemographic
variables that were not signicant predictors (p .05) of
grade 12 high self-esteem included household composition,
parent level of education, and family size.
Developmental assets predictive of high self-esteem
(logistic regression)
For both adolescent boys and girls, several developmen-
tal assets perceived in grade 8 independently predicted re-
ports of high self-esteem in grade 12 (Table 3). Factors
common to both boys and girls included positive family
communication, baseline self-esteem, and feelings of safety.
Boys and girls in grade 8 who reported positive family
communication were signicantly more likely to report high
self-esteem in grade 12 (OR for boys 2.4, 95% CI
1.553.73; OR for girls 3.03, 95% CI 1.934.74). Both
boys and girls reporting high self-esteem in grade 8 were 3
times more likely to report high self-esteem in grade 12
(95% CI for boys 2.533.59; 95% CI for girls 2.53
3.63). Those boys and girls who reported feeling safe in
their school in grade 8 were 1.41 and 1.50 times, respec-
tively, more likely to report high self-esteem in grade 12
than those not feeling safe (95% CI for boys 1.041.91;
95% CI for girls 1.082.08). In addition, adolescent males
who reported involvement in their religious community (in
comparison with those who did not) were 1.21 times more
likely to report high self-esteem in grade 12 (95% CI
1.011.43).
Discussion
This was a nationally representative longitudinal study
characterizing self-esteem among male and female adoles-
cents, and the ndings offer information on the role of
developmental assets that are predictive of high self-esteem.
The ndings that fewer girls than boys report high self-
esteem at all grade levels is consistent with existing litera-
ture [13,21], and our study extends this knowledge to broad
populations. In addition, we found that female report of
African-American race or Hispanic ethnicity was an inde-
pendent predictor of high self-esteem over time. This nd-
ing in African-American girls is as we predicted and con-
197 S. Birndorf et al. / Journal of Adolescent Health 37 (2005) 194201
Figure 1. Developmental assets identied as predictor variables and their composition.
198 S. Birndorf et al. / Journal of Adolescent Health 37 (2005) 194201
sistent with existing literature [1316]. The same nding in
girls of Hispanic ethnicity is not consistent with existing
literature and may reect a general trend of increasing
self-esteem over time among the Hispanic population rela-
tive to Caucasians in the United States [22]. Another con-
sideration is the large variability in the degree of difference
in reported self-esteem among Hispanics relative to Cauca-
sians depending on the particular self-esteem measure used
[22]. This variability in self-reports of esteem with different
measures is not as pronounced among African-American
and Asian samples [22]. Clearly, more research is needed to
validate existing theoretical explanations and to broaden our
understanding of possible emerging trends among Hispanic
and Asian teenagers through improved consistency with
self-esteem measures and improved precision in dening
ethnic subgroups.
Our ndings that fewer boys reported high self-esteem in
grades 10 and 12 as compared with grade 8 are consistent
with reported uctuations in self-esteem over time for high
schoolage boys [13,21]. This may reect the effect of
cognitive development on self-reporting styles from middle
to late adolescence, although differences in self-esteem be-
tween genders are most likely multifactorial and currently
are not well characterized.
Because self-esteem usually has begun to change by the
age of 12, the context within which the adolescent is devel-
oping already may have promoted (or discouraged) the
development of high self-esteem by the age of 14 [23]. This
Table 2
High self-esteem within gender (bivariate analyses)
Grade Male (%) Female (%) p value Relative risk 95% CI
8th 39.2 27.4 .001 1.43 1.331.54
10th 36.8 30.3 .001 1.21 1.131.31
12th 37.4 30.6 .001 1.22 1.141.32
Table 3
Grade 8 independent predictors of high self-esteem in grade 12 (logistic regression analyses)
Sociodemographics Males Females
OR 95% CI OR 95% CI
Race/ethnicity
Hispanic 1.23 (.961.58) 1.52** (1.152.02)
African-American 1.38 (1.021.88) 2.06*** (1.502.85)
Other 1.13 (.821.56) .95 (.631.43)
Caucasian
a
1.00 1.00
Parents education
Less than High school .95 (.631.41) 1.16 (.821.65)
More than High school 1.16 (.901.49) 1.17 (.921.48)
High school graduate
a
1.00 1.00
At or above federal poverty level 1.50** (1.142.00) 1.31* (1.021.69)
Household composition 1.00 (.771.24) 1.00
Internal assets
Hours spent on homework
5.5 h/wk 1.12 (.951.34) .87 (.721.05)
5.5 h/wk
a
1.00 1.00
Planning and decision making
Has plans after graduation 1.77 (.684.61) 2.69 (.818.90)
Plans to graduate high school 2.06 (.815.24) 2.78 (.878.92)
Will not graduate high school
a
1.00 1.00
High self-esteem, grade 8 3.01*** (2.533.59) 3.03*** (2.533.63)
External assets
Positive family communication 2.40*** (1.553.73) 3.03*** (1.934.74)
Other adult relationships 1.16 (.941.43) 1.14 (.931.40)
Caring school climate 1.12 (.681.85) .69 (.341.39)
Parent involvement in schooling 1.67 (.574.86) 2.07 (.2318.2)
Safety 1.41* (1.041.91) 1.50* (1.082.08)
Family boundaries
Often .92 (.741.14) 1.11 (.891.37)
Sometimes .93 (.751.15) 1.18 (.961.46)
Never
a
1.00 1.00
School boundaries .77 (.212.83) 1.32 (.364.88)
Adult role models 1.27 (.991.64) 1.00 (.771.28)
Religious community 1.21* (1.011.43) .97 (.821.14)
* p .01, ** p .001; *** p .05.
a
Referent group.
199 S. Birndorf et al. / Journal of Adolescent Health 37 (2005) 194201
may explain why many of the developmental assets mea-
sured in 8th grade such as school climate and other adult
role models and relationships were not predictive of high
self-esteem in 12th grade. As hypothesized, we did nd that
several developmental assets predicted high self-esteem in
our sample. These included positive family communication,
feeling safe in school, and religious community (for boys).
Not surprisingly, high self-esteem in the baseline survey
also was highly predictive of high self-esteem in grade 12.
Several limitations warrant further discussion. The data-
set is relatively old because it spans the years from 1988 to
1992. The National Education Longitudinal Survey was
chosen at the onset of our investigation despite this because
it was the largest study that provided such a rich and
detailed amount of pertinent longitudinal data from a very
large and diverse sample size. Another limitation to the
generalizability of our ndings results from the loss of
subjects over time caused by various factors including drop-
out, absenteeism, and suspended/expelled status. Of note is
the fact that there were no differences in the rst 8th-grade
survey in self-esteem between our sample and the subjects
lost to follow-up evaluation. Although almost a third of our
sample was lost to follow-up evaluation, a retention rate of
66% across the 5 years of survey waves is substantial and
better than many longitudinal studies of this magnitude and
duration. In fact, a major strength of this study is its longi-
tudinal design and its ability to capture information as
adolescents span critical periods in development.
Another limitation is the fact that the rst measurement
of self-esteem in our sample occurred in 8th grade, or at
approximately age 14 for most students. It has been shown
in cross-sectional studies that changes in the self-esteem
construct begin to occur earlier, at around the age of 11 or
12 [21]. We are limited by our dataset in this regard, and we
were not able to capture the changes in self-esteem before
8th grade in our subjects. In addition, our surrogate mea-
sures of developmental assets were not standardized or
tested for internal consistency. They were, however, tested
for association before inclusion in the multivariate model.
Although we were able to capture only 16 of the 40
existing assets in this dataset, the selection of assets repre-
sents a majority of 10 external and intrinsic assets that are
thought to capture community impact best as it relates to the
development of positive social, educational, and health out-
comes. These 10 core assets have been found in aggregates
of cross-sectional surveys from over 250,000 adolescents to
be associated with lower levels of antisocial behavior [24].
Our results are consistent with theoretical models and ex-
isting literature that suggests adults from a teenagers fam-
ily, school, and community have the potential to impact
adolescent development positively by providing supportive
relationships and safe environments for adolescents [25]. In
addition, our surrogate measures of developmental assets
are not standardized or tested for internal consistency. They
were, however, tested for association before inclusion in the
multivariate model.
Although this study does not examine the relationship of
self-esteem to functional outcomes, existing literature also
suggests that in addition to characterizing self-esteem fur-
ther, additional studies are needed to clarify the relationship
of self-esteem to functional outcomes such as the develop-
ment of depression and suicidality in adolescents.
In conclusion, the study ndings suggest that the con-
struct of self-esteem must be treated with care in future
studies that either further characterize self-esteem in ado-
lescents, or examine the predictive role of self-esteem in
functional outcomes. Not only are there major differences
across time in reports of self-esteem from 8th to 12th grade,
but self-esteem differs considerably between genders and
among race/ethnicity groups for girls. These ndings sug-
gest that the inuences of self-esteem on different groups
vary, and thus the role of self-esteem regarding health and
disease most likely is complex. Finally, our ndings also
suggest that clinicians and community leaders may promote
self-esteem in their adolescents by encouraging families,
schools, and community organizations to engage teenagers
in positive communication and to provide safe and nurturing
environments.
References
[1] James W. The Principles of Psychology. New York: Henry Holt and
Company, 1890310.
[2] Shamir B. Self-esteem and the psychological impact of unemploy-
ment. Soc Psychol Q 1986;49:6172.
[3] Kaplan HB, Robbins C, Martin SS. Antecedents of psychological
distress in young adults: self-rejection, deprivation of social support,
and life events. J Health Soc Behav 1983;24:23044.
[4] Turner RJ, Roszell P. Psychosocial resources and the stress process.
In: Avison WR, Gotlib IH, eds. Stress and Mental Health. New York:
Plenum Press, 1994:179210.
[5] Benson PL. All Kids Are Our Kids. San Francisco: Jossey Bass,
1997:73.
[6] White V, Hill D, Hopper J. The outgoing, the rebellious and the
anxious: are adolescent personality dimensions related to the uptake
of smoking? Psychol Health 1996;12:7385.
[7] Simon TR, Sussman S, Dent CW, et al. Prospective correlates of exclu-
sive or combined adolescent use of cigarettes and smokeless tobacco: a
replication-extension study. Addict Behav 1995;20:51724.
[8] House JS, Landis KR, Umberson D. Social relationships and health.
Science 1988;241:54045.
[9] Kawachi I, Colditz GA, Ascherio A, et al. A prospective study of
social networks in relation to total mortality and cardiovascular dis-
ease in men in the USA. J Epidemiol Community Health 1996;50:
24551.
[10] Thoits P. Stress, coping, and social support processes: where are we?
What next? (extra issue) J Health Soc Behav 1995:5379.
[11] Cohen S. Psychosocial models of the role of social support in the
etiology of physical disease. Health Psychol 1988;7:26997.
[12] Benson PL. All Kids Are Our Kids. San Francisco: Jossey Bass,
1997:2753.
[13] Benson PL. All Kids Are Our Kids. San Francisco: Jossey Bass,
1997:66.
200 S. Birndorf et al. / Journal of Adolescent Health 37 (2005) 194201
[14] Porter JR, Washington RE. Sex differences in the self-concept in ado-
lescence. In: Rosenberg M, Kaplan HB, eds. Social Psychology of the
Self-Concept. Arlington Heights, IL: Harlan Davidson, 1982:22434.
[15] Brown KM, McMahon RP, Biro FM, et al. Changes in self-esteem in
black and white girls between the ages of 9 and 14 years. J Adolesc
Health 1998;23:719.
[16] Carlson C, Uppal S, Prosser EC. Ethnic differences in processes
contributing to the self-esteem of early adolescent girls. J Early
Adolesc 2000;20:4467.
[17] Curtin TR, Ingels SJ, Wu S, Heuer R. National Education Longitu-
dinal Study of 1988: Base-Year to Fourth Follow-up Data File Users
Manual (NCES 2002-323). Washington, D.C.: U.S. Department of
Education, National Center for Education Statistics.
[18] Rosenberg M. Society and the Adolescent Self-Image. Princeton, NJ:
Princeton University Press, 1965; 1636.
[19] Wylie R. Measures of Self-Concept. Lincoln, NE: University of
Nebraska Press, 1989; 2435.
[20] Shah BV, Barnwell BG, Bieler GS. SUDAAN Users Manual, Re-
lease 7.5. Research Triangle Park, NC: Research Triangle Institute,
1997.
[21] Rosenberg FR, Simmons RG. Sex differences in the self-concept in
adolescence. In: Rosenberg M, Kaplan HB, eds. Social Psychology of
the Self-Concept. Arlington Heights, IL: Harlan Davidson, 1982:
24250.
[22] Twenge JM, Crocker J. Race and self-esteem: meta-analyses comparing
whites, blacks, Hispanics, Asians, and American Indians and comment
on Gray-Little and Hafdahl. Psychol Bull 2002;128:371408.
[23] Simmons RG, Rosenberg F, Rosenberg M. Disturbance in the self-
image at adolescence. Am Soc Rev 1973;38:55368.
[24] Benson PL. All Kids Are Our Kids. San Francisco: Jossey Bass,
1997:74.
[25] Resnick MD, Bearman PS, Blum RW, et al. Protecting adolescents
from harm: ndings from the National Longitudinal Study on Ado-
lescent Health. JAMA 1997;278:82332.
201 S. Birndorf et al. / Journal of Adolescent Health 37 (2005) 194201