American Sociological Association is collaborating with JSTOR to digitize, preserve and extend access to
American Sociological Review.
http://www.jstor.org
Carmi Schooler
Carrie Schoenbach
Florence Rosenberg
In this paper, we attempt to shed light on the nature of relevance of, and
relationship between global self-esteem and specific self-esteem. We marshal evidence that the two types of self-esteem may have strikingly different
consequences, global self-esteem being more relevantto psychological wellbeing, and specific self-esteem being more relevant to behavior. We use linear structural equation causal modeling to test this hypothesisfor the case
of global self-esteem (Rosenberg 1979) and specific (academic) self-esteem.
Ourfindings show that, while global self-esteem is more strongly related to
measures of psychological well-being, specific (academic) self-esteem is a
much better predictor of school performance. Otherfindings indicate that
the degree to which specific academic self-esteem affects global self-esteem,
particularly the positive component of global self-esteem, is a-function of
how highly academic performance is personally valued.
Looking at the general body of research on
self-esteem today, it is evident that most of
this literature deals with global self-esteem,
that is, the individual's positive or negative
attitude toward the self as a totality. In the
last decade, however, a number of writers
have stressed the importanceof studying specific self-esteem, as well (e.g., Harter 1985;
Marsh 1986; Marsh and Shavelson 1985;
Swann 1987). As Marsh (1990) expresses it:
"More recently, self-concept theory has
stressed the multi-dimensionality of self*Direct all correspondence to Carmi Schooler, concept, and empirical studies have identiLaboratory of Socio-environmental Studies,
fied distinct, a priori facets of self-concept"
NIMH, NIH, Rm BIA-14, Federal Building, 7550
(p. 107).
The
MD
20892.
Wisconsin Avenue, Bethesda,
The aim of this paper is to shed light on
authors presented an earlier version of this paper
the
nature and relevance of global and speto Research Committee 42, Social Psychology, at
the 12th World Congress of Sociology, Madrid, cific self-esteem and their relationship to
Spain, July 1990. This paper, which represents a each other. We begin by focusing on two
very substantial revision of the earlier one and general features of attitudes to clarify the
includes new analyses, was completed after distinctions we make regarding self-esteem.
Morris Rosenberg's untimely death in 1992. First, the study of any attitude, and self-esAmong the analyses completed after Rosenberg's
teem is an attitude, must take account of the
death are those that separatethe positive and negahave attitudes both totive aspects of self-esteem and those that examine fact that people may
how valuing academic performancerelates to spe- ward an object as a whole (global or general)
cific academic self-esteem and global self-esteem. and toward specific "facets" of that object
141
142
143
144
cause the relationships reportedin the literature between self-esteem and other variables
are often weaker than might be expected.
This is true whether self-esteem is treated as
cause or outcome. For example, Wylie
(1979) and others reported that sociodemographic variables show no better than modest success in predicting self-esteem. (Some
reasons for these results have been discussed
by McCarthy and Yancey 1971; Rosenberg
and Pearlin 1978; Rosenberg and Simmons
1972). Although there have been some successes (e.g., Menaghan and Parcel 1990), in
general, self-esteem has not proved to be an
impressive predictorof behavioraloutcomes.
One reason for these comparatively weak
relationships between self-esteem and behavior has been the failure to recognize that global and specific self-esteem are both important, but that they are importantfor different
reasons and are relevant in different ways. A
central hypothesis of this paper is that specific self-esteem is most relevant to behavior, whereas global self-esteem is most relevant to psychological well-being. Much
self-esteem research, we believe, relates precisely the wrong type of self-esteem to the
outcome variable (Scheff, Retzinger, and
Ryan 1989), examining the relationship between global self-esteem and specific behaviors or behavioral outcomes. We believe, on
the contrary, that a specific behavior is best
predicted by a specific self-esteem that is in
some way connected to that behavior,
whereas psychological well-being is best
predicted by global self-esteem. Let us consider some of the evidence on which our hypothesis is based.
knowing that someone has high global selfesteem will tell us little about that person's
assessment of his or her competence as a
pole-vaulter or as a writer of sonnets. Conversely, the fact that I consider myself totally
inept as a pole-vaulter in itself offers little
indication of my overall feeling of selfworth. We do not suggest that global self-esteem is totally unrelatedto behavior, but that
this relationship is likely to be weaker than
the relationship of specific self-esteem to a
relevant behavior or performance.
Although evidence is limited, the results of
both experimental and correlational studies
support this view. In an experimental study,
Shrauger (1972) asked female undergraduates to specify what percentage of undergraduates would perform better than themselves on a "concept attainment task." The
percentage score served as the measure of
specific self-esteem. General (global) selfesteem was based on a modification of a selfesteem measure used extensively by Diggory
(1966). The concept attainment task was a
modification of a measure developed by
Weick (1964). Shrauger found that specific
self-esteem was a significant predictorof actual performance on the concept attainment
test, but that global self-esteem was not.
Correlational findings are consistent with
the experimental results. In Bachman's
(1970) study of tenth-gradeboys, the correlation between global self-esteem and school
marks was .23, whereas the correlation between school marks and self-concept of academic ability was .48. Reviewing a number
of studies dealing with the relationship between self-esteem and academic performance, Wylie (1979) concluded that the correlation between global self-esteem and grade
point average is usually about .30, whereas
the association between specific self-esteem
(academic self-concept) and grade point average is more likely to be in the range of .45
to .70. Wylie (1979:698-700) also reporteda
number of other instances in which specific
self-esteem is correlatedwith specific behavior but global self-esteem is not.
Self-Esteem and Psychological Well-Being
Although global self-esteem is less likely
than specific self-esteem to be a good predictor of behavior or performance, there is
145
146
use Bachman's School Ability Self-Concept Index, which is based on subjects' responses
to three questions. Two questions are coded on a
scale from 1 through 6-far below average to far
above average: (1) "How do you rate yourself in
school ability compared with those in your grade
in school?"; (2) "How intelligent do you think
you are compared to others your age?" The third
question, (3) "Comparedto others your age, how
important is it to you to be able to use your intelligence?" is coded from 1 through 5-much less
important than average to much more important
than average.
RESULTS
Table 1 presents the zero-order correlation
coefficients for each of these 10 well-being
indices with global self-esteem on the one
hand, and with specific (academic) self-esteem on the other.3
Table 1 clearly shows that global self-esteem is more strongly related to most measures of psychological well-being (depression, anomie, general anxiety, resentment,
anxiety-tension, irritability, life satisfaction,
happiness, and negative affective states)
than is specific self-esteem. The mean
strength of the relationship of global self-esteem to all 10 measures is .337, whereas the
mean strength relationship of academic selfesteem to these measures is .079. The single
exception is the guilt variable, which is unrelated to global self-esteem but is significantly related to academic self-esteem. On
the -other hand, when we consider our specific behavioral outcome (school performance), then specific (academic) self-esteem turns out to be much more highly correlated than is global self-esteem (.488 versus .253). This pattern of correlations indicates that self-esteem is significantly related
to a number of other variables, but only if
2
Global
Self-Esteem
Psychological Well-Being
Depression
-.432***
Academic
Self-Esteem
-.103***
Anomie
Generalanxiety
Resentment
Anxiety/tension
-.409***
-.232***
-.346***
-.319***
-. 154***
-.010
-.120***
-.023
Irritability
-.353***
-.044
Life satisfaction
Guilt
-.033
Happiness
Negative affective
states
Behavioral Outcome
School marks(GPA)
.318***
.052*
.087***
.499***
-.424***
.105***
-.093***
.253***
.488***
147
148
ables at both points in time are estimated. In addition, significant correlations of residuals between different variables are included where suggested by an examination of the first-order partial derivatives. In all structural equations discussed, we allow errors in the causal equations to
be correlated.
effect should only be indirect, and the instrument must be reasonably correlated with the
variable it is not allowed to affect directly. In
our model, cross-lagged effects are omitted
to provide a source of instrumentationfor the
reciprocal effects that are modeled for wave
2. These wave 2 effects are viewed as outcomes of long-term processes and thus represent the sum of the lagged and contemporaneous effects.
Additional instrumentalvariables are also
employed. Instrumentsfor the global to academic self-esteem relationship are positive
family relationships, numberof best friends,
physical appearance,and complexion. All of
these instrumental variables could be expected to directly affect global self-esteem.
None of them would be expected to have a
direct effect on academic self-esteem; any
effect would be indirect through global selfesteem's effect on academic self-esteem. Instrumentsfor the academic to global self-esteem relationship are hours spent on homework, positive school attitudes, negative
school attitudes, and valuation of academic
achievement. All of these instrumental variables would be expected to affect academic
self-esteem without directly affecting global
/
self-esteem.
We introduce the following variables as
controls affecting both global self-esteem
and academic self-esteem: race, age, intact
family, mother's education, father's education, family socioeconomic status, father's
occupational status, mother's occupational
status, and numberof siblings.
The model of the reciprocal effects of global self-esteem and academic self-esteem fits
the data reasonably well, with a chi-square/
d.f. ratio of 4.05. This shows that global selfesteem and academic self-esteem affect one
another significantly, but that academic selfesteem has a more powerful effect on global
self-esteem than the other way around (.21
versus .1 1). Both effects are significant at the
p < .05 level.
A second question is: Does the same pattern of findings appear when we consider a
specific facet of the self-concept that rests on
a higher level of abstraction,namely, self-estimate of intelligence? Facets of the self-concept, as Marsh and Shavelson (1985) have
noted, exist at different levels of specificity.
Some facets may be highly specific, others
149
more general. This is why Marsh and Shavelson conceptualized the self-concept as a hierarchical structure. Consider, for example,
facets of the self-concept that exist in the intellectual realm. At a highly specific level
might be a child's assessment of his or her
reading ability or math ability (Marsh 1986).
At a somewhat broader level would be the
child's academic self-esteem, which would
be based in part on math and reading ability
and in part on other facets. Still broader
would be the child's assessment of his or her
general level of intelligence, which would be
affected by academic self-esteem but also by
other facets. When we consider the reciprocal effects of global and specific self-esteem,
then, it is importantto consider the level of
specificity or generality of the specific selfconcept facet.
We addressed this question by examining
the reciprocal effects of global self-esteem
and the students' estimates of their own intelligence. Self-estimate of intelligence was
measuredby responses to the following item:
"How intelligent do you think you are compared with other boys your age?" Six categories of response were provided ranging from
"Far below average (bottom 10 percent)" to
"Far above average (top 10 percent)." Note
that this item was also included in the academic self-esteem scale; however, compared
to the other items in that scale, it is general
and makes no specific reference to the school
experience.
To identify the path from global self-esteem to self-estimate of intelligence, the following instrumental variables were used:
respondent's height, weight, number of
friends, level of social support, closeness to
father, closeness to mother, parents' tendency to reason with the child in disciplinary situations, parentalpunitiveness, and stability of global self-esteem. All of these
would be expected to affect global self-esteem directly and self-estimate of intelligence only indirectly. Instruments for the
path from perceived intelligence to global
self-esteem include certain performance
measures (scores on a job information test
and on a test of political knowledge), positive school attitudes, negative school attitudes, and stability of self-assessment of intelligence. All of these variables would be
expected to have a direct effect on self-as-
150
sessment of intelligence and only indirect effects on global self-esteem. Again, crosslagged effects were constrained to 0 in order
to provide a source of instrumentation.
This model was found to fit the data reasonably well, as shown by a chi-square/d.f.
ratio of 3.45. Again, we find that the specific
facet of intelligence has a more powerful effect on global self-esteem than the other way
around. Both variables produce significant
effects upon one another,but the effect of intelligence on global self-esteem is .24,
whereas the effect of global self-esteem on
intelligence is .11.
Intellectual ability, then, whether narrowly
conceived as academic ability or broadly
conceived as general intelligence, produces
a more powerful effect on global self-esteem
than the other way around. Note that this effect appears despite the fact that only single
items are used to measure the two more narrowly defined concepts (academic self-esteem and self-estimate of intelligence), while
global self-esteem is modeled with multiple
indicators based on a widely used measure.
High school boys' general feelings of selfworth appear to be significantly affected by
their judgments of their intellectual ability,
but their general feeling of self-worth has a
weak effect on their opinions of their intellectual ability.
We do not suggest, of course, that specific
self-esteem is necessarily more likely to affect global self-esteem than the other way
around. Intellectual ability may not be typical of other facets of the self. Our prediction
would be that the relative effects of various
facets of the self-concept on global self-esteem would depend on the degree of
"schematization" of the specific facets
(Markus 1977)-if a particularfacet is firmly
established or solidly crystallized and is bolstered by clear evidence, then it is likely to
be fairly independentof the individual's general feeling of self-worth. For example, if
many experiences of failure have taught a
pupil through clear evidence that he or she is
incompetent in school, then his or her general feeling of self-worth will do little to
change this conviction. But if a particular
facet is nonschematic (uncrystallized), then
global self-esteem may exercise a more powerful effect on that specific facet. For example, if people are asked to judge their
151
152
153
154
155
156
Purkey, William W. 1970. Self-Concept and
School Achievement. Englewood Cliffs, NJ:
Prentice-Hall.
Rosenberg, Morris. 1979. Conceiving the Self.
Malabar, FL: Robert E. Krieger.
. 1985. "Self-Concept and Psychological
Well-Being in Adolescence." Pp. 205-46 in
The Development of the Self, edited by R. L.
Leahy. Orlando, FL: Academic Press.
. 1989. Society and the Adolescent Self-Image. Rev. ed. Middletown, CT: Wesleyan University Press.
Rosenberg, Morris and Leonard I. Pearlin. 1978.
"Social Class and Self-Esteem among Children
and Adults." American Journal of Sociology
84:53-77.
Rosenberg, Morris, Carmi Schooler, and Carrie
Schoenbach. 1989. "Self-Esteem and Adolescent Problems: Modeling Reciprocal Effects."
American Sociological Review 54:1004-18.
Rosenberg, Morris and Roberta G. Simmons.
1972. Black and White Self-Esteem. Washington, DC: American Sociological Association.
Scheff, Thomas, Suzanne M. Retzinger, and
Michael T. Ryan. 1989. "Crime, Violence, and
Self-Esteem: Review and Proposals." Pp. 16599 in The Social Importance of Self-Esteem,
edited by A. M. Mecca, N. J. Smelser, and J.
Vasconcellos. Berkeley, CA: University of
California Press.