ArticlePDF Available

Does high self-esteem cause better performance, interpersonal success, happiness, or healthier lifestyles?

Authors:

Abstract

Self esteem has become a household word. Teachers, parents, therapists, and others have focused efforts on boosting self-esteem, on the assumption that high self-esteem will cause many positive outcomes and benefits-an assumption that is critically evaluated in this review. Appraisal of the effects of self-esteem is complicated by several factors. Because many people with high self-esteem exaggerate their successes and good traits, we emphasize objective measures of outcomes. High self-esteem is also a heterogeneous category, encompassing people who frankly accept their good qualities along with narcissistic, defensive, and conceited individuals. The modest correlations between self-esteem and school performance do not indicate that high self-esteem leads to good performance. Instead, high self-esteem is partly the result of good school performance. Efforts to boost the self-esteem of pupils have not been shown to improve academic performance and may sometimes be counterproductive. Job performance in adults is sometimes related to self-esteem, although the correlations, vary widely, and the direction of causality has not been established. Occupational success may boost self-esteem rather than the reverse. Alternatively, self-esteem may be helpful only in some job contexts. Laboratory studies have generally failed to find that self-esteem causes good task performance, with the important exception that high self-esteem facilitates persistence after failure. People high in self-esteem claim to be more likable and attractive, to have better relationships, and to make better impressions on others than people with low self-esteem, but objective measures disconfirm most of these beliefs. Narcissists are charming at first but tend to alienate others eventually. Self-esteem has not been shown to predict the quality or duration of relationships. High self-esteem makes people more willing to speak up in groups and to criticize the group's approach. Leadership does not stem directly from self-esteem, but self-esteem may have indirect effects. Relative, to people with low self-esteem, those with high self-esteem show stronger in-group favoritism, which may increase prejudice and discrimination. Neither high nor low self-esteem is a direct cause, of violence. Narcissism leads,to increased aggression in retaliation, for wounded pride. Low self-esteem may contribute to externalizing behavior and delinquency, although some studies have found that there are no effects or that the effect of self-esteem vanishes when other variables are controlled. The highest and. lowest rates of cheating and bullying are found in different subcategories of high self-esteem. Self-esteem has a strong relation to happiness. Although the research has not clearly established causation, we are persuaded that high self-esteem does lead to greater happiness., Low self-esteem is more likely than high to lead to depression under some circumstances. Some studies support the buffer hypothesis, which is that high self-esteem mitigates the effects of stress, but other studies come to the opposite conclusion, indicating that I the negative effects of low self-esteem are, mainly felt in good times. Still others find that high self-esteem leads to happier outcomes regardless of stress or other circumstances. High self-esteem does not prevent children from smoking, drinking, taking drugs, or engaging in early sex. If anything high self-esteem fosters experimentation, which may increase early sexual activity or drinking, but in general effects of self-esteem are negligible. One important exception is that high self-esteem reduces the chances of bulimia in females. Overall, the benefits of high self-esteem fall into two categories: enhanced initiative and pleasant feelings. We have not found evidence that boosting self-esteem (by therapeutic, interventions or school programs) causes benefits. Our findings do not support continued widespread efforts to boost self-esteem in the hope that it will by itself foster improved outcomes. In view of the heterogeneity of high self-esteem, indiscriminate praise might just as easily promote narcissism, with its less desirable consequences. Instead, we recommend using praise to boost self-esteem as a reward for socially desirable behavior and self-improvement.
PSYCHOLOGICAL SCIENCE IN THE PUBLIC INTEREST
VOL. 4, NO. 1, MAY 2003
Copyright © 2003 American Psychological Society
1
DOES HIGH SELF-ESTEEM CAUSE BETTER
PERFORMANCE, INTERPERSONAL SUCCESS,
HAPPINESS, OR HEALTHIER LIFESTYLES?
Roy F. Baumeister,
1
Jennifer D. Campbell,
2
Joachim I. Krueger,
3
and Kathleen D. Vohs
4
1
Florida State University;
2
University of British Columbia, Vancouver, British Columbia, Canada;
3
Brown University;
and
4
University of Utah
Summary—Self-esteem has become a household word.
Teachers, parents, therapists, and others have focused efforts
on boosting self-esteem, on the assumption that high self-
esteem will cause many positive outcomes and benefits—an
assumption that is critically evaluated in this review.
Appraisal of the effects of self-esteem is complicated by
several factors. Because many people with high self-esteem
exaggerate their successes and good traits, we emphasize ob-
jective measures of outcomes. High self-esteem is also a het-
erogeneous category, encompassing people who frankly accept
their good qualities along with narcissistic, defensive, and
conceited individuals.
The modest correlations between self-esteem and school
performance do not indicate that high self-esteem leads to
good performance. Instead, high self-esteem is partly the re-
sult of good school performance. Efforts to boost the self-
esteem of pupils have not been shown to improve academic
performance and may sometimes be counterproductive. Job
performance in adults is sometimes related to self-esteem, al-
though the correlations vary widely, and the direction of cau-
sality has not been established. Occupational success may
boost self-esteem rather than the reverse. Alternatively, self-
esteem may be helpful only in some job contexts. Laboratory
studies have generally failed to find that self-esteem causes
good task performance, with the important exception that
high self-esteem facilitates persistence after failure.
People high in self-esteem claim to be more likable and at-
tractive, to have better relationships, and to make better im-
pressions on others than people with low self-esteem, but
objective measures disconfirm most of these beliefs. Narcis-
sists are charming at first but tend to alienate others eventu-
ally. Self-esteem has not been shown to predict the quality or
duration of relationships.
High self-esteem makes people more willing to speak up in
groups and to criticize the group’s approach. Leadership
does not stem directly from self-esteem, but self-esteem may
have indirect effects. Relative to people with low self-esteem,
those with high self-esteem show stronger in-group favorit-
ism, which may increase prejudice and discrimination.
Neither high nor low self-esteem is a direct cause of vio-
lence. Narcissism leads to increased aggression in retaliation
for wounded pride. Low self-esteem may contribute to external-
izing behavior and delinquency, although some studies have
found that there are no effects or that the effect of self-esteem
vanishes when other variables are controlled. The highest and
lowest rates of cheating and bullying are found in different sub-
categories of high self-esteem.
Self-esteem has a strong relation to happiness. Although the
research has not clearly established causation, we are per-
suaded that high self-esteem does lead to greater happiness.
Low self-esteem is more likely than high to lead to depression
under some circumstances. Some studies support the buffer hy-
pothesis, which is that high self-esteem mitigates the effects of
stress, but other studies come to the opposite conclusion, indi-
cating that the negative effects of low self-esteem are mainly
felt in good times. Still others find that high self-esteem leads to
happier outcomes regardless of stress or other circumstances.
High self-esteem does not prevent children from smoking,
drinking, taking drugs, or engaging in early sex. If anything,
high self-esteem fosters experimentation, which may increase
early sexual activity or drinking, but in general effects of self-
esteem are negligible. One important exception is that high
self-esteem reduces the chances of bulimia in females.
Overall, the benefits of high self-esteem fall into two cate-
gories: enhanced initiative and pleasant feelings. We have not
found evidence that boosting self-esteem (by therapeutic in-
terventions or school programs) causes benefits. Our findings
do not support continued widespread efforts to boost self-
esteem in the hope that it will by itself foster improved out-
comes. In view of the heterogeneity of high self-esteem, indis-
criminate praise might just as easily promote narcissism, with
its less desirable consequences. Instead, we recommend using
praise to boost self-esteem as a reward for socially desirable
behavior and self-improvement.
Most people feel that self-esteem is important. It is difficult,
if not impossible, for people to remain indifferent to informa-
tion that bears on their own self-esteem, such as being told that
they are incompetent, attractive, untrustworthy, or lovable. In-
creases and decreases in self-esteem generally bring strong
emotional reactions. Moreover, these fluctuations are often co-
incident with major successes and failures in life. Subjective
Address correspondence to Roy F. Baumeister, Department of Psychology,
Florida State University, Tallahassee, FL 32306-1270.
PSYCHOLOGICAL SCIENCE IN THE PUBLIC INTEREST
Benefits of Self-Esteem
2
VOL. 4, NO. 1, MAY 2003
experience creates the impression that self-esteem rises when
one wins a contest, garners an award, solves a problem, or gains
acceptance to a social group, and that it falls with corresponding
failures. This pervasive correlation may well strengthen the im-
pression that one’s level of self-esteem is not just the outcome,
but indeed the cause, of life’s major successes and failures.
But is self-esteem a cause of important consequences in
life? In this monograph, we report the results of a survey of
major research findings bearing on this question. Our mission
was to conduct a thorough review of empirical findings—em-
phasizing the most methodologically rigorous research stud-
ies—to ascertain whether high self-esteem is in fact a cause of
positive or negative outcomes. We anticipated we would find
that self-esteem has positive value for bringing about some hy-
pothesized benefits, but not others. Such a pattern would pre-
sumably allow an accurate and nuanced understanding of just
what high self-esteem is good for. This would be beneficial
both for theory (in that it would promote a better understanding
of self-esteem as well as the outcomes it predicts) and for prac-
tical applications—and even for determining whether efforts at
boosting self-esteem are worth undertaking in order to solve
particular social problems.
Self-esteem is literally defined by how much value people
place on themselves. It is the evaluative component of self-
knowledge. High self-esteem refers to a highly favorable glo-
bal evaluation of the self. Low self-esteem, by definition, refers
to an unfavorable definition of the self. (Whether this signifies
an absolutely unfavorable or relatively unfavorable evaluation
is a problematic distinction, which we discuss later in connec-
tion with the distribution of self-esteem scores.) Self-esteem
does not carry any definitional requirement of accuracy what-
soever. Thus, high self-esteem may refer to an accurate, justi-
fied, balanced appreciation of one’s worth as a person and
one’s successes and competencies, but it can also refer to an in-
flated, arrogant, grandiose, unwarranted sense of conceited su-
periority over others. By the same token, low self-esteem can
be either an accurate, well-founded understanding of one’s
shortcomings as a person or a distorted, even pathological
sense of insecurity and inferiority.
Self-esteem is thus perception rather than reality. It refers to
a person’s belief about whether he or she is intelligent and at-
tractive, for example, and it does not necessarily say anything
about whether the person actually is intelligent and attractive.
To show that self-esteem is itself important, then, research
would have to demonstrate that people’s beliefs about them-
selves have important consequences regardless of what the un-
derlying realities are. Put more simply, there would have to be
benefits that derive from believing that one is intelligent, re-
gardless of whether one actually is intelligent. To say this is not
to dismiss self-esteem as trivial. People’s beliefs shape their ac-
tions in many important ways, and these actions in turn shape
their social reality and the social realities of the people around
them. The classic study
Pygmalion in the Classroom
, by
Rosenthal and Jacobson (1968), showed that teachers’ false,
unfounded beliefs about their students later became objective,
verifiable realities in the performance of those students. In the
same way, it is quite plausible that either high or low self-
esteem, even if initially false, may generate a self-fulfilling
prophecy and bring about changes in the objective reality of the
self and its world.
Then again, self-esteem might not bring about such changes.
Many researchers, clinicians, teachers, parents, and pundits
have taken it as an article of faith that high self-esteem will
bring about positive outcomes. Such an assumption was per-
haps reasonable several decades ago, given the lack of firm
data either way and the anecdotal impressions and theoretical
bases for assuming that self-esteem has strong effects. It is par-
ticularly understandable that practitioners would accept this as-
sumption without proof, because they cannot generally afford
to admonish their suffering clients to hang on for a few decades
until needed research is conducted. They must use the best evi-
dence available at the time to design their interventions.
By now, however, the excuse of inadequate data is begin-
ning to wear thin. The fascination with self-esteem that began
to spread during the 1970s infected researchers too, and in the
past couple of decades, a number of methodologically rigor-
ous, large-scale investigations on the possible effects of self-
esteem have been conducted. We do not think all the final an-
swers are in, but many of them are taking shape. There is no
longer any justification for simply relying on anecdotes, im-
pressions, and untested assumptions about the value of self-
esteem.
WHY STUDY SELF-ESTEEM?
In the heady days of the 1970s, it might have seemed possi-
ble to assert that self-esteem has a causal effect on every aspect
of human life, and by the 1980s, the California legislature
might well have been persuaded that funding a task force to in-
crease the self-esteem of Californians would ultimately pro-
duce a huge financial return because reducing welfare dependency,
unwanted pregnancy, school failure, crime, drug addiction, and
other problems would save large amounts of taxpayers’ money.
However, as Karl Marx, Sigmund Freud, and other grand
thinkers could assert if they were alive today, even the most
elaborate and persuasive theories about human behavior do not
generally receive empirical support in all aspects. Thus, we
note at the outset that we did not expect all the extravagant
claims of the self-esteem movement to be supported.
Even if the self-esteem movement was wrong in crucial re-
spects, its positive aspects and contributions deserve to be rec-
ognized and celebrated. The self-esteem movement showed
that the American public was willing to listen to psychologists
and to change its institutional practices on the basis of what
psychology had to teach. It would not be in psychology’s best
interest to chastise the American public for accepting the ad-
PSYCHOLOGICAL SCIENCE IN THE PUBLIC INTEREST
R.F. Baumeister et al.
VOL. 4, NO. 1, MAY 2003
3
vice of psychologists. If errors were committed, perhaps psy-
chologists should reduce their own self-esteem a bit and
humbly resolve that next time they will wait for a more thor-
ough and solid empirical basis before making policy recom-
mendations to the American public. Regardless of the outcome
of the self-esteem movement, it showed that there is a voice for
psychology in public policy and discourse. If psychology uses
that voice judiciously, it may still be able to make a major con-
tribution to the well-being of society.
The Appeal of Self-Esteem
As self-aware and self-reflective creatures, many people in-
tuitively recognize the importance of self-esteem. Not surpris-
ingly, a great deal of psychological theorizing has focused on
the motivation to protect and, if possible, enhance self-esteem.
Research is showing that even psychodynamic defense mecha-
nisms, which Freud originally understood as ways of keeping
threatening sexual and aggressive impulses at bay, serve as
strategies to bolster self-esteem (for a review, see Baumeister,
Dale, & Sommer, 1998).
But the desire to feel good about oneself is certainly not the
only self-related motive at play. Having to cope with reality,
people are also motivated to perceive themselves accurately
and admit awareness of their undesirable characteristics (Swann,
Stein-Seroussi, & Giesler, 1992; Trope, 1986). Nevertheless,
people would rather learn positive things about themselves
than negative things (Sedikides, 1993). Although they may
want to know whether they are good or not, they much prefer
to learn that they are good.
Over the past few decades, the need for high self-esteem has
risen from an individual to a societal concern. North American
society in particular has come to embrace the idea that high
self-esteem is not only desirable in its own right, but also the
central psychological source from which all manner of positive
behaviors and outcomes spring. This strong psychological
claim has begun to permeate popular beliefs. Its corollary, the
idea that low self-esteem lies at the root of individual and thus
societal problems and dysfunctions, has obvious implications
for interventions on both the individual and the societal level.
The hope that such interventions might work has sustained an
ambitious social movement. Nathaniel Branden, a leading fig-
ure in the self-esteem movement, stated categorically that
“self-esteem has profound consequences for every aspect of
our existence” (Branden, 1994, p. 5), and, more pointedly, that
he “cannot think of a single psychological problem—from anx-
iety and depression, to fear of intimacy or of success, to spouse
battery or child molestation—that is not traceable to the prob-
lem of low self-esteem” (Branden, 1984, p. 12). Other advo-
cates of the movement have endorsed this sentiment. Andrew
Mecca, for example, is cited as saying that “virtually every so-
cial problem can be traced to people’s lack of self-love”
(Davis, 1988, p. 10).
Academic and professional psychologists have been more
hesitant to endorse strong categorical claims. Eminent clinical
psychologist Albert Ellis, for example, is convinced that “self-
esteem is the greatest sickness known to man or woman be-
cause it’s conditional” (cited in Epstein, 2001, p. 72). Accord-
ing to Ellis, people would be better off if they stopped trying to
convince themselves that they are worthy. Others believe that
concerns about self-esteem are a peculiar feature of Western in-
dividualist cultures. According to this perspective, the search
for high self-esteem is not a universal human motive, but a cul-
tural or ideological artifact. Indeed, such a motive is difficult to
detect in collectivist cultures, and especially in Japan (Heine,
Lehman, Markus, & Kitayama, 1999). Even in Western culture,
the need for high self-esteem seems to be a rather recent devel-
opment (Baumeister, 1987). The Judeo-Christian tradition has
long considered modesty and humility as virtues conducive to
spiritual growth. In this tradition, high self-esteem is suspect
because it opens the door to sentiments of self-importance.
Medieval theologians considered pride or vainglory to be par-
ticularly satanic and thus a deadly sin. To combat it, religious
devotees cultivated an unattractive appearance (e.g., shorn hair,
no makeup, unfashionable clothes, no jewelry), spoke with
self-effacement, and submitted to degrading exercises (e.g.,
begging, prostrations, self-flagellations).
Such practices are but a faint memory in contemporary pop-
ular culture, in which high self-esteem seems to reign supreme.
Prodded by Assemblyman John Vasconcellos, the then gover-
nor of California George Deukmeijian agreed in 1986 to fund a
Task Force on Self-Esteem and Personal and Social Responsi-
bility with a budget of $245,000 per annum for several years.
Vasconcellos argued that raising self-esteem would help solve
many of the state’s problems, including crime, teen pregnancy,
drug abuse, school underachievement, and pollution. At one
point, he expressed the hope that raising self-esteem would
help balance the state’s budget because people with high self-
esteem earn more money than people with low self-esteem and
therefore pay more taxes (Winegar, 1990). It is easy to dismiss
and satirize such claims (Dawes, 1994). However, Vasconcellos
and the task force also speculated astutely about the possibility
that self-esteem might protect people from being overwhelmed
by life’s challenges and thus reduce failures and misbehaviors,
much as a vaccine protects against disease.
Concurrent with its activities in the field, which included
creating self-esteem committees in many California counties,
the task force assembled a team of scholars to survey the rele-
vant literature. The results were presented in an edited volume
(Mecca, Smelser, & Vasconcellos, 1989). Echoing Branden
(1984), Smelser (1989) prefaced the report by stating that
“many, if not most, of the major problems plaguing society
have roots in the low self-esteem of many of the people who
make up society” (p. 1). But the findings did not validate the
high hopes of the task force, and Smelser had to acknowledge
that “one of the disappointing aspects of
every
[italics added]
PSYCHOLOGICAL SCIENCE IN THE PUBLIC INTEREST
Benefits of Self-Esteem
4
VOL. 4, NO. 1, MAY 2003
chapter in this volume . . . is how low the associations between
self-esteem and its [presumed] consequences are in research to
date” (p. 15). Given that the correlations were so low, the ques-
tion of whether low self-esteem in fact caused the societal
problems did not even arise.
The lack of supportive data created a dilemma. Should a no-
tion as attractive as self-esteem be abandoned and replaced
with more promising concepts, or should the validity of the ev-
idence be questioned? The editors and the authors opted for a
mix of these two strategies. Some retreated to a defense of self-
esteem on a priori grounds. Undeterred, Smelser (1989) main-
tained that
diminished self-esteem stands as a powerful
independent variable
(condition, cause, factor) in the genesis of major social problems. We
all know this to be true, and it is really not necessary to create a special
California task force on the subject to convince us. The real problem
we must address—and which the contributors to this volume ad-
dress—is how we can determine that it is scientifically true. (p. 8)
Others, however, acknowledged the limitations of the findings
and called for additional study, or tried to fit more complex the-
oretical models of self-knowledge to the data. Our report is fo-
cused primarily on studies conducted since the review by the
California task force. Instead of examining the merits of the
more complex models of self, we have retained the hypothesis
that global self-esteem causes desirable, adaptive, and benefi-
cial behaviors. There is a certain beauty to this hypothesis be-
cause it is simple, clear, and testable. There have also been
sufficient methodological advances in study design and statisti-
cal analysis that warrant a fresh look at the evidence.
Meanwhile, the self-esteem movement was not deterred by
the disappointing findings of the task force. After it was dis-
banded in 1995, the National Council for Self-Esteem inherited
its mandate, which was subsequently taken on by the National
Association for Self-Esteem, or NASE. Vasconcellos (now a
member of the California Senate) and Jack Canfield (
Chicken
Soup for the Soul
) are on NASE’s advisory board, and such
media personalities as Anthony Robbins (
Unlimited Power
),
Bernie Siegel (
Love, Medicine, and Miracles
), and Gloria
Steinem (
A Revolution From Within: A Book of Self-Esteem
)
are members of a “Masters Coalition,” created by NASE. The
mission statement of NASE minces no words about the pre-
sumed benefits of self-esteem. Its goal is to “promote aware-
ness of and provide vision, leadership and advocacy for
improving the human condition through the enhancement of
self-esteem” (NASE, 2000). The goal of the Masters Coalition
is no less ambitious. “It is hoped that the Master Coalition can,
in a meaningful way, facilitate the actualization of society and
lead to the amelioration, if not elimination, of various negative
influences which have operated in part to trivialize and demean
the human condition” (NASE, 2000).
It is hard not to conclude that the self-esteem movement has
ignored its own major scholarly document (i.e., the Mecca et
al., 1989, volume). In the quest for enhanced self-esteem, any
tool in the psychological—and pseudopsychological—box is
thrown into the fray, including
disparate psychological models that have given rise to such popular
notions as the “inner child”; the “self-image”; principles of proper
grieving; “super learning”; “community networking”; “relaxation
techniques” and their effects on overall mental and physical well-be-
ing; the principles of “neuro-linguistic programming”; and the well-
founded scientific basis for the connection between the body and the
mind and the effect of this interface on overall wellness. (NASE,
2000)
Even a contributor to the volume edited by Mecca et al. (1989)
argued that self-esteem must be enhanced, although its causal
role is far from established. “To abandon the search for esteem-
related solutions . . . is to admit defeat before exploring all our
options” (Covington, 1989, p. 74).
Was it reasonable to start boosting self-esteem before all the
data were in? Perhaps. We recognize that many practitioners
and applied psychologists must deal with problems before all
the relevant research can be conducted. Still, by now there are
ample data on self-esteem. Our task in this monograph is to
take a fresh look and provide an integrative summary.
An Epidemic of Low Self-Esteem?
A key assumption of the self-esteem movement is that too
many people have low self-esteem. Under this assumption,
raising self-esteem becomes a meaningful goal. But what does
“too many” mean? Self-esteem scales are designed to capture
valid individual differences that exist in a population. Thus, a
good measure will yield a distribution of scores from low to
high. However, unlike some other measurement instruments,
such as IQ tests, that are constructed to yield symmetrical dis-
tributions centered around an arbitrary mean (e.g., 100), self-
esteem scales allow skewed distributions to emerge. The
average score typically lies far above the midpoint of the scale,
often by more than a standard deviation (Baumeister, Tice, &
Hutton, 1989). The fact that most people score toward the high
end of self-esteem measures casts serious doubt on the notion
that American society is suffering from widespread low self-
esteem. If anything, self-esteem in America is high. The aver-
age person regards himself or herself as above average.
The skewed distribution of self-esteem scores raises two
methodological issues. First, when researchers split samples at
the median to distinguish between respondents with high ver-
sus low self-esteem, the range of scores among respondents
classified as having low self-esteem is much greater than the
range of scores among respondents classified as having high
self-esteem. A good number of respondents in the low self-es-
teem category have scores above the midpoint of the scale. In
other words, the classification of a person as someone with low
self-esteem has no longer an absolute, but only a relative mean-
ing. Second, correlations involving variables with skewed dis-
PSYCHOLOGICAL SCIENCE IN THE PUBLIC INTEREST
R.F. Baumeister et al.
VOL. 4, NO. 1, MAY 2003
5
tributions tend to be smaller than correlations involving variables
with symmetric distributions. Moreover, when self-esteem is
raised selectively for those respondents with the lowest initial
values, correlations between self-esteem and relevant outcome
variables shrink further, not necessarily because the elevation
of self-esteem had the desired causal effect, but simply because
of the restriction in the range of scores. It is always necessary
to ask whether relevant outcomes also changed in the desired
direction.
The standard finding that most self-esteem scores are high
raises the possibility that at least some scores are affected by
deliberate or unwitting self-enhancement (Krueger, 1998).
Brown (1986), for example, found that people high in self-es-
teem were also most likely to rate themselves more positively
than they rated other people. Because self-enhancement may
involve invalid and undesirable distortions of the self-concept,
it is unwarranted to rush to boost everyone’s self-esteem.
In short, we find no evidence that modern Western societies
are suffering from an epidemic of low self-esteem. If anything,
self-esteem seems generally high in most North American sam-
ples. Regardless of their race, gender, or socioeconomic status,
Americans already appear to live in a “culture of self-worth”
(Twenge & Campbell, 2001, p. 325). Indeed, levels of self-
esteem increased at a time when the self-esteem movement be-
moaned the lack of self-love. Disturbingly, academic performance
decreased at the same time (Twenge & Campbell, 2001).
PROBLEMS AND CHALLENGES
Measurement of Self-Esteem
Many scales are available for measuring self-esteem, and
different investigations have used different ones, which com-
pounds the difficulty of comparing results from different inves-
tigations (especially if the results are inconsistent). Blascovich
and Tomaka (1991) reviewed multiple measures and found
them of uneven quality, giving high marks to only a few (such
as Fleming & Courtney’s, 1984, revision of Janis & Field’s,
1959, scale, and Rosenberg’s, 1965, global self-esteem mea-
sure). In essence, self-esteem scales ask people to rate them-
selves in response to questions such as “Are you a worthwhile
individual?” “Are you good at school or work?” “Do people
like you?” and “Are you reliable and trustworthy?” When re-
searchers check self-esteem measures against the so-called lie
scales (also called measures of social desirability, because they
assess tendencies to give distorted, even unrealistic answers
just to make a good impression), they conclude that self-esteem
scores are somewhat contaminated by people’s efforts to make
themselves look good. These measures also obscure needed
distinctions between defensive, inflated, narcissistic, and so-
called genuine high self-esteem. (We discuss different varieties
of high self-esteem in the next section.) Unfortunately, there is
no objective criterion against which to compare self-reported
self-esteem, because of the nature of the construct: Self-esteem
essentially consists of how a person thinks about and evaluates
the self. In the case of intelligence, for example, self-ratings
can be compared against objective performance on intellectual
tests, and the results can (and often do) show that people’s self-
reports of their own intelligence are wrong. But there is no
known basis for saying that certain people really have more or
less self-esteem than they think they have.
To overcome these measurement problems, some research-
ers measure
implicit
, or unfakeable, self-esteem by using a va-
riety of subtle methods, such as reaction times to good and bad
thoughts that can be paired with the self (Greenwald & Farn-
ham, 2000). Though promising, this research has only recently
begun, and it therefore does not play a significant role in this
review. Despite the potential pitfalls of
explicit
(i.e., self-report)
measures, the fact that scores on different scales are positively
correlated (e.g., Greenwald & Farnham, 2000) is an indication
that they can be used with some confidence. Even more signifi-
cantly, the Rosenberg scale, which is by far the most popular
among researchers, has been shown to be highly reliable (e.g.,
if a person completes the scale on two occasions, the two
scores tend to be similar). As a measure of global self-esteem,
this scale is unidimensional (Gray-Little, Williams, & Han-
cock, 1997; Robins, Hendin, & Trzesniewski, 2001). Indeed,
its reliability is so high that a single item (“I have high self-
esteem”) may be sufficient (Robins et al., 2001).
Heterogeneity of High Self-Esteem
The high internal consistency of self-esteem measures may
mask the possibility that a variety of psychological processes
contribute to high (or low) scores. One approach to studying
the heterogeneity of self-esteem is to examine the pattern of
scores across multiple measurement instruments. Schneider
and Turkat (1975) suspected that some people’s high self-esteem
is defensive rather than genuine, and that these individuals
could be identified if they also scored high on the Marlowe-
Crowne Social Desirability Scale (Crowne & Marlowe, 1960).
The concept of defensive self-esteem has recently been refined
by the distinction between deception of others (i.e., impression
management) and deception of the self (see Paulhus, 2002, for
a review). High self-esteem is considered defensive if it is cou-
pled with high scores on a self-deception scale (which has
items such as “I always know why I do things”).
Taking a different approach, Kernis and his colleagues (see
Kernis & Waschull, 1995, for a review) measure both the over-
all level and the temporal stability of self-esteem. In many
studies, the stability of self-esteem, either by itself or in combi-
nation with level of self-esteem, has been shown to predict behav-
ioral outcomes. Baumeister and his colleagues (e.g., Baumeister,
1993; Baumeister, Smart, & Boden, 1996) found that behaviors
and outcomes are often more variable for people high in self-
PSYCHOLOGICAL SCIENCE IN THE PUBLIC INTEREST
Benefits of Self-Esteem
6
VOL. 4, NO. 1, MAY 2003
esteem than for people low in self-esteem. For example, studies
often fail to yield a significant correlation between aggression
or violence and self-esteem, in part because high self-esteem is
associated with both the presence and the absence of aggres-
sion. Kernis’s work leads to similar conclusions. People with
high but unstable self-esteem score higher on measures of hos-
tility than do people with low self-esteem (whether stable or
unstable), whereas people with high but stable self-esteem are
the least hostile (Kernis, Grannemann, & Barclay, 1989). In
other words, people with high self-esteem are found at the ex-
tremes of both high and low hostility.
A third approach is to measure narcissism along with, or
even instead of, self-esteem. The construct of narcissism in-
volves highly favorable, even grandiose views of self, a sense
of being special or unique, fantasies of personal brilliance or
beauty, and the belief that one is entitled to privileges and ad-
miration by others (see American Psychiatric Association,
2000). In normal populations, scores on the Narcissistic Per-
sonality Inventory (Emmons, 1984; Raskin & Hall, 1981) cor-
relate substantially with self-esteem (Rhodewalt & Morf, 1995;
Sinha & Krueger, 1998). The heterogeneity of high self-esteem
is indicated by the finding that some people with high self-
esteem are not narcissistic, whereas others are; the reverse is
not true (i.e., there are very few narcissists with low self-
esteem). Bushman and Baumeister (1998) measured both self-
esteem and narcissism in laboratory studies of aggression. Nar-
cissism, but not self-esteem, predicted aggression, and only
when the target was someone who had previously insulted the
participant. The idea that narcissistically high self-esteem is
unhealthy is further supported by the finding that narcissists’
high self-esteem tends to be unstable (Rhodewalt, Madrian, &
Cheney, 1998) and self-defensive (Paulhus, 1998).
Taken together, these lines of research suggest that some
low correlations between self-esteem and socially relevant be-
haviors or outcomes conceal the tendency for different categories
of high self-esteem to produce opposite responses. Researchers
who believe in the value of so-called genuine or healthy self-
esteem may find that they can obtain more valid correlations
with desirable outcomes if they control variables such as nar-
cissism, self-deception, or temporal stability. The broader im-
plication is, however, that the category of people with high
self-esteem is a mixed bag that contains individuals whose self-
opinions differ in important ways.
Theorists and researchers have linked low self-esteem to
other constructs, generally focusing on the links between low
self-esteem and pathologies such as depression. Recently, low
self-esteem has begun to be associated with more general con-
cepts such as emotional lability (a tendency for strong emotions
to occur in both directions) and low internal locus of control (a
generalized belief that the self is not in control of what happens).
Although there are moderate associations between low self-
esteem and pathology (see the section on Coping and Depres-
sion), there is also evidence against the notion that the con-
structs of low self-esteem and depression are isomorphic.
Correlations between self-esteem and depression are of only
moderate strength, ranging from .4 to .6 (e.g., Joiner, Alfano, &
Metalsky, 1992). Also, a theoretical standpoint indicates that
although the psychological processes associated with self-
esteem and depression may overlap somewhat (e.g., Abramson,
Metalsky, & Alloy, 1989), they are not identical. Rather, the
concept of depression has been characterized by a constellation
of symptoms, of which low self-esteem is one (Roberts &
Monroe, 1999). But low self-esteem is neither necessary nor
sufficient for depression.
A recent analysis of the interrelations among self-esteem,
neuroticism, locus of control, and feelings of being effective
points to a more serious methodological problem regarding their
overlap. Judge, Erez, and Bono (2002) found that these con-
structs are highly related and reflect one overarching construct
(which they said was neuroticism, broadly defined). Judge et al.
concluded that the ability of any of these constructs to uniquely
(i.e., independently) predict outcomes is quite poor, and they
urged psychological scientists to begin thinking of each con-
struct in tandem with the others. We concur with this suggestion
and hope that researchers will include more of these measures in
studies of the effect of self-esteem on objective outcomes.
Global Versus Domain-Specific Self-Esteem
The heterogeneous nature of people who score high on self-
esteem measures is not the only reason why the predictive
power of global measures of self-esteem is limited. Another
reason is that it is difficult to detect a correspondence between
a global attitude and specific behaviors or outcomes (Fishbein
& Ajzen, 1975). It is not to be expected, for example, that a
global sense of being worthy, competent, and popular will pre-
dict performance on a trigonometry quiz. Many people may
consider mathematical ability to be irrelevant to global self-
appraisal, and so their self-esteem could be utterly irrelevant to
how well they can perform numerical calculations. Predictions
improve when self-esteem is measured for the domain of inter-
est and among people who consider this domain to be person-
ally important (Crocker & Wolfe, 2001). If relevant domains
are hierarchically organized, it is important to measure self-
esteem at the appropriate level of specificity. If the domain be-
comes too narrow, the assessment of specific self-esteem may
yield only trivial results. For example, a high correlation between
people’s success at doing long division and their self-evaluation
for this task may simply result from people’s awareness of their
ability in this domain. If so, any attempts to improve perfor-
mance by way of enhancing self-esteem would fail.
The difficulties of relating global self-esteem to specific be-
haviors may be overcome, in part, by aggregating behaviors
into bundles. But aside from differences in levels of specificity
of measurement, there is also a difference in affectivity. Global
self-esteem is heavily invested with feelings about the self,
whereas specific facets of self-esteem include a variety of self-
related thoughts (Brown, 1998; Rosenberg, Schooler, Schoen-
PSYCHOLOGICAL SCIENCE IN THE PUBLIC INTEREST
R.F. Baumeister et al.
VOL. 4, NO. 1, MAY 2003
7
bach, & Rosenberg, 1995). It is perhaps because of its high af-
fective relevance that global self-esteem has been the preferred
target for well-intentioned interventions. Interventions at the
level of domain-specific self-esteem would arguably become
both more fragmented and less affectively charged. For these
reasons, we focus our review on global self-esteem.
Floccinaucinihilipilification
The word
floccinaucinihilipilification
is allegedly the longest
word in the Oxford English Dictionary, and it is defined as “the
action or habit of estimating as worthless.” The definition of low
self-esteem involves making a disparaging or low-worth judg-
ment about the self, yet there is mounting evidence that people
with low self-esteem are not merely negative about themselves.
Rather, they express a generally negative attitude toward many
events, circumstances, people, and other realities. That is, they
have a tendency toward floccinaucinihilipilification.
The problem this raises for the researcher is twofold. First,
it is difficult to distinguish the general negativity from the spe-
cific low self-esteem. A good example of this problem emerged
in the research literature on prejudice. In early work, research-
ers assumed that people with low self-esteem would have more
prejudice than others toward out-groups, and studies in which
people simply rated other groups seemingly confirmed that
people with low self-esteem expressed more negative attitudes
toward them than did people with higher self-esteem (see
Wills, 1981, for review). But Crocker and Schwartz (1985)
pointed out that if people rate themselves and their in-groups
negatively, it is hardly fair to label them as prejudiced for rating
out-groups negatively, too. Such general floccinaucinihilipilifi-
cation can be corrected by subtracting the ratings of self or the
in-group from the ratings of the out-group, and when this was
done, the finding was reversed: People with high self-esteem
showed greater in-group bias and greater prejudice than people
with low self-esteem (see also Aberson, Healy, & Romero,
2000; Crocker & Luhtanen, 1990; Crocker, Thompson, McGraw,
& Ingerman, 1987).
Second, floccinaucinihilipilification raises the very substan-
tial danger that self-reports will show spurious correlations be-
tween self-esteem and other outcomes. A general tendency
toward floccinaucinihilipilification may cause certain people to
describe themselves in disparaging terms, thus generating low
scores on self-esteem, as well as to describe their lives and out-
comes in disparaging terms, thus furnishing the appearance of
maladaptive behaviors and pathology. This is one reason why it
is important to obtain objective outcome measures.
Need for Objective Measures
All too often, results involving self-esteem are based on the
self-report of outcome variables. The problem this raises can
be illustrated by the case of a researcher who asks people,
“How good are you at juggling red beanbags?” and then takes
their answers as indications of their self-esteem in the beanbag-
juggling sphere. If the researcher also decides to use self-reports
to measure performance, almost exactly the same question
could be used as a measure of performance. The resulting high
correlation between the measures of self-esteem and perfor-
mance would reflect nothing more than the fact that the same
question was asked to measure different constructs.
It is essential to keep in mind that self-esteem is measured
almost exclusively (and unavoidably) by self-report. People
score high in self-esteem because they respond to a question-
naire by endorsing favorable statements about themselves. The
habit of speaking well of oneself does not abruptly cease when
the respondent turns from the self-esteem scale to the question-
naire asking for self-report of other behaviors. People who like
to describe themselves in glowing terms will be inclined to re-
port that they get along well with others, are physically attrac-
tive, do well in school and work, refrain from undesirable
actions, and the like. That is how they get high scores in self-
esteem, but researchers may easily mistake this identical ten-
dency as evidence that self-esteem predicts or even causes a
broad range of positive outcomes. Over and over during our
survey of the literature, we found that researchers obtained
more impressive evidence of the benefits of self-esteem when
they relied on self-reported outcomes than when they relied on
objective outcomes. The research on the relationship between
self-esteem and physical attractiveness provides a good exam-
ple of such a discrepancy. Most self-esteem scales do not con-
tain items that specifically ask whether respondents consider
themselves physically attractive, but they do measure the glo-
bal tendency to speak well of oneself. It seems plausible that
people who speak well of themselves in general would rate
their physical attractiveness more highly than others. Then
again, it is plausible that physically attractive people would end
up with higher self-esteem than other people, if only because
attractive people are treated more favorably than unattractive
ones throughout life—they are more popular, more sought af-
ter, more valued by lovers and friends, and so forth.
Several studies have explored correlations between global
self-esteem and self-rated attractiveness, generally finding very
strong positive relationships. Harter (1993) described results
from multiple studies indicating that the correlation was around
.85. This is a remarkably strong connection, indicating that
people’s physical attractiveness accounted for more than 70%
of the variance in their self-esteem. If this result is correct, it
means that people’s self-esteem is mainly based on their ap-
pearance. But one cannot easily rule out the possibility that the
correlation received an unfair boost by the general tendency to
speak well of oneself. People who score high on self-esteem by
claiming to be wonderful people in general may claim to be
physically attractive, and people who rate themselves relatively
poorly overall may derogate their appearance.
Hence, it is important to obtain more objective measures of
physical appearance, to match up with the subjective measures
of self-esteem. Diener, Wolsic, and Fujita (1995) obtained self-
esteem scores from a broad sample of individuals and then
PSYCHOLOGICAL SCIENCE IN THE PUBLIC INTEREST
Benefits of Self-Esteem
8
VOL. 4, NO. 1, MAY 2003
photographed everyone. The photographs were shown to other
judges, who rated them for attractiveness, thus providing an
objective measure of physical attractiveness unaffected by self-
report. The ratings of attractiveness based on full-length pic-
tures showed a trivial correlation of .06 with self-esteem—not
even a significant correlation. Head-and-shoulders photos fared
only slightly better, with a correlation of .14, which suggests
that physical attractiveness accounted for less than 2% of self-
esteem. Even this figure may be inflated, because it could be
that people with high self-esteem take particular care to present
themselves in a favorable light, such as by wearing attractive
clothing and jewelry. When these props were removed to pro-
duce unadorned photos of the participants’ faces, the correla-
tion of judge-rated attractiveness with self-esteem was .00. In
that same investigation, however, self-reported physical attrac-
tiveness was found to have a hefty (
r
.59) correlation with
self-esteem. Thus, people’s ratings of their own good looks ac-
counted for 35% of the variance in their self-esteem, whereas
objective ratings of their looks had a negligible relationship to
their self-esteem. People with high self-esteem are gorgeous in
their own eyes, but objective observers do not see any difference.
Similar findings were reported by Gabriel, Critelli, and Ee
(1994), who obtained peer ratings of attractiveness from photo-
graphs. The correlations with self-esteem were negligible:
r
.01 for males and
r
.04 for females. Again, though, peo-
ple with high self-esteem rated themselves as more attractive
than those with low self-esteem.
The discrepancy is sobering. What seems at first to be a
powerful relationship between physical good looks and high
self-esteem turns out to be nothing more than a pattern of con-
sistency in how favorably people rate themselves. Those who
say they are very good overall tend to say that they look good,
too. When other people are brought in to judge unadorned pho-
tographs, however, people with high self-esteem do not emerge
as any more attractive than people with low self-esteem.
Inflated views of one’s own attributes are not confined to
physical attractiveness. Gabriel et al. (1994) also asked partici-
pants to rate their own intelligence and then gave them an
intelligence (IQ) test. People with high self-esteem rated them-
selves as significantly more intelligent than people with low
self-esteem (
r
.35). But the results of the objective IQ test
did not justify these favorable claims, for there was no signifi-
cant relationship between self-esteem and IQ scores (
r
.07).
The authors also reported that self-esteem was significantly
correlated (
r
.38) with the discrepancy between self-rated in-
telligence and objectively measured intelligence. This finding
confirms that people with high self-esteem exaggerate their in-
telligence more than people with low self-esteem.
Weight has long been associated with self-esteem, espe-
cially in modern Western cultures that glorify slender, young-
looking bodies (especially for women). Hence, a lighter body
weight should be associated with high self-esteem, whereas
obesity should be linked to low self-esteem. Consistent with
that view, a meta-analysis (which statistically combines the re-
sults of multiple studies) by C.T. Miller and Downey (1999)
found a significant correlation (
r
.24) between actual body
weight and self-esteem. But the correlation of self-esteem with
self-rated body weight was much stronger (
r
.72). Thus,
people with high self-esteem are a little slimmer than others,
but not nearly as much as they think.
The broad implication of these examples is that self-reports
are likely to contain substantial biases that can easily yield mis-
leading empirical findings. People with high self-esteem claim
to be successful, attractive, and wonderful in many respects.
Objective evidence sometimes paints a much different picture,
and many of the ostensible (self-reported) advantages claimed
by people with high self-esteem are clearly disconfirmed.
The systematic discrepancies between objective and self-
reported outcomes led us to conclude that we should set up our
literature search and review to emphasize objective measures
of outcomes whenever possible. This vastly reduced the
amount of material we could use. But the material that re-
mained is far more reliable and convincing insofar as the re-
sults are based on objective measures.
To be sure, objective measures are not always possible to
obtain, and reliance on self-report is thus inevitable in some
spheres. For example, in the case of happiness, there is almost
no alternative to self-report. Even when it is necessary to use
self-report, however, we urge researchers to emphasize the
most objective, concrete, and verifiable data possible. It seems
likely, for example, that global self-ratings of intelligence are
more vulnerable to bias and inflation than self-reports of grades
in specific courses or scores on specific tests. Behavioral self-
reports (e.g., “How many cigarettes did you smoke yester-
day?”) should be more reliable than vague or “in general” rat-
ings of one’s own behavior (e.g., “How much do you smoke?
Very much, some, not very much, or not at all?”).
The bias in self-report may well be partly responsible for
the popularity of self-esteem among teachers, parents, thera-
pists, and others who seek to intervene in people’s lives in a
positive fashion. A rise in self-esteem may well cause a person
to honestly believe that he or she is doing better in many
spheres, even if these beliefs are utterly false and stem from the
positive illusions that attend high self-esteem. If both teacher
and student believe that boosting the student’s self-esteem has
led to improvements in academic performance, the entire enter-
prise of boosting self-esteem is likely to be marked by a seduc-
tive feeling of success and efficacy. The unfortunate fact that
academic performance may not have changed at all by objec-
tive measures might be overlooked in the context of the mutu-
ally sustained social reality of the belief that performance has
improved.
Direction of Causality
A great many researchers have proceeded by administering
a self-esteem scale and some behavioral measures and then
computing the correlation. Such an approach can establish
PSYCHOLOGICAL SCIENCE IN THE PUBLIC INTEREST
R.F. Baumeister et al.
VOL. 4, NO. 1, MAY 2003
9
links between self-esteem and other variables. Correlations
cannot establish causality, however, and so the meaning of such
correlations remains highly ambiguous. As anyone who has
taken a research methodology course knows, a correlation be-
tween X and Y could mean that X causes Y, that Y causes X, or
that some other variable causes both. Thus, if self-esteem cor-
relates with good school performance, self-esteem might be ei-
ther the cause or the result of good performance, or both
outcomes could derive from something else, such as a privi-
leged upbringing.
The question of causality goes to the heart of the debate
about the self-esteem movement and interventions aimed at
boosting self-esteem. If self-esteem causes an assortment of
positive outcomes, then it may well be worth considerable ef-
fort and expense to instill high self-esteem into children, the
mentally ill, stigmatized populations, and other vulnerable or
at-risk groups. In contrast, if the correlations mean simply that
self-esteem is an outcome of success and good behavior—
which is, after all, at least as plausible as the hypothesis that
self-esteem is a cause of success—then there is little to be
gained by raising self-esteem.
If self-esteem is indeed an outcome rather than a cause, then
it is even plausible that raising self-esteem could backfire and
produce undesirable effects. Suppose, for example, that work-
ing hard in school leads to good grades, and good grades lead
to high self-esteem. Assume also that high self-esteem feels
good, and so the rise in self-esteem could operate as an impor-
tant reinforcer for the hard work that leads to academic suc-
cess. If a school program intervenes directly to boost self-
esteem regardless of academic performance, then students can
enjoy the rewards of self-esteem without making the effort.
One major incentive to work hard would thereby be eliminated,
and students might reduce their effort, leading to poorer aca-
demic performance. Such speculations indicate the need to at-
tend to the causal processes involved in self-esteem and raise
substantial questions about the impact of boosting self-esteem.
The claim that correlational findings can shed no light on
causal processes is overstated. Two patterns of correlational
findings are especially helpful for evaluating causality. First, if
A causes B, or B causes A, then A and B will show up as corre-
lated in most data sets. If A fails to show a significant correla-
tion with B, then it is fair to conclude that there is no causal
relationship in either direction (assuming that measurement er-
ror or methodological shortcomings are not responsible for the
null correlation). Put another way, null correlations falsify
causal hypotheses.
Second, causes generally precede consequences in time, and
so correlations across time are often used to make causal infer-
ences. If self-esteem on one occasion (Time 1) predicts school
performance at a later time (Time 2), whereas school perfor-
mance at Time 1 does not predict self-esteem at Time 2, then
one can reasonably conclude that self-esteem is the cause and
school performance is the outcome. Likewise, if self-esteem at
Time 1 predicts school performance at Time 2 after differences
in school performance at Time 1 are controlled statistically (in
effect, self-esteem at Time 1 predicts change in school perfor-
mance between Times 1 and 2), then causal inferences may be
made, albeit somewhat tentatively.
Insisting that self-esteem must predict achievement at Time
2 after controlling for achievement at Time 1 could obscure
some actual causal relationships, so it should be regarded as a
highly conservative way of testing the hypothesis. Suppose, for
example, that self-esteem does consistently contribute to achieve-
ment—and that self-esteem is fairly stable. Achievement at
Time 1 is thus a result of prior self-esteem (say, at Time
1). If
one controls for achievement at Time 1 when computing the
correlation between Time 1 self-esteem and Time 2 achieve-
ment, one may be throwing a very large baby out with the sta-
tistical bathwater. What in reality was a replication of causal
processes (Time
1 self-esteem causing Time 1 achievement,
and Time 1 self-esteem causing Time 2 achievement) ends up
looking like no relationship at all.
In short, care must be taken to avoid either overstating or
understating the causal influence of self-esteem.
METHOD OF SEARCHING THE LITERATURE
The research covered in the subsequent sections of this re-
view was located in the following manner. On several occa-
sions from January 2001 through October 2001, we searched
the PsychINFO database and obtained a list of all articles con-
taining “self esteem” in the abstract. Our searches yielded a to-
tal of 15,059 publications. We also went to the “subject”
category in PsychINFO and searched for “self esteem,” finding
a total of 11,860 articles. Because these two lists overlapped
substantially, we decided to rely mainly on one of them; we
chose the former because it included more entries.
The 15,000-article list was too long to work with, so we nar-
rowed it down using a series of criteria. First, we did not retain
work that had not been published and whose sole source was
Dissertation Abstracts International
. This biased our results in
favor of positive effects of self-esteem, insofar as null results
are difficult to publish. Second, we excluded studies designed
to show what causes self-esteem. In other words, we sought to
understand the consequences of self-esteem rather than its de-
terminants. Third, we generally eliminated secondary sources
in order to obtain original data. Many book chapters, books,
and other writings talk about self-esteem, but our goal was to
see what the original data indicate. We therefore included sec-
ondary sources only if they presented meta-analyses or other
authoritative literature reviews.
Three other sources of information beyond the PsychINFO
list were utilized. First, we have all been reading and writing
about self-esteem for many years, and we all felt it appropriate
to contribute any additional work from our own knowledge.
Second, sources cited in the articles from the PsychINFO
search were also included. This was especially helpful for find-
ing null effects, because researchers who investigate self-esteem
PSYCHOLOGICAL SCIENCE IN THE PUBLIC INTEREST
Benefits of Self-Esteem
10
VOL. 4, NO. 1, MAY 2003
and fail to confirm predictions often cite previous work to note
that their own failure is not unique. Third, in August 2001, we
sent out a request on the Listserve of the Society for Personal-
ity and Social Psychology asking for any new or unpublished
articles on objective outcomes of self-esteem. This request
elicited an additional five articles.
Finally, we sought to narrow our list to include only work
on outcomes with broad social relevance. We searched for spe-
cific relationships between self-esteem and variables of inter-
est, such as health, sexual behavior, financial status, grades,
intelligence, job performance and satisfaction, and interper-
sonal relations.
We then read the remaining articles carefully. At this point,
our overriding goal was to obtain studies that included objec-
tive evidence of the effects of self-esteem. As we have already
noted, some variables (such as happiness) can be assessed only
via self-report. Whenever possible, however, we wanted objec-
tive or third-party evidence. We also favored studies that
sought to establish causal relationships, either by longitudinal
designs or by laboratory experimentation.
Given our restrictive criteria, relatively few studies survived
from the initial set. Although many thousands of publications
refer to self-esteem, relatively few of them report studies that
used highly rigorous methods to examine the causal impact of
self-esteem on personal and social problems. These few gener-
ally represent vast amounts of work and deserve careful attention.
We summarize them individually rather than meta-analytically.
SCHOOL PERFORMANCE
The self-esteem movement has been especially influential in
American schools, and part of the reason for this is the assump-
tion that raising self-esteem will lead to improvements in chil-
dren’s academic performance. There are plausible reasons for
thinking that high self-esteem will lead to good schoolwork.
People with high self-esteem may set higher aspirations than
people with low self-esteem. They may be more willing to per-
sist in the face of initial failure and less likely to succumb to
paralyzing feelings of incompetence and self-doubt. Learning,
by definition, involves acquiring information and skills that one
does not initially have, and high self-esteem may help prevent
the recognition of one’s initial incapability from producing a
sense that the cause is hopeless. High self-esteem may foster
the confidence to tackle difficult problems and enable people to
derive satisfaction from progress and success.
Correlational Findings
Many studies have found that self-esteem is positively cor-
related with academic performance. In an early review, Wylie
(1979) concluded that the correlation between self-esteem and
students’ grade point averages was about .30. She added that
similar or slightly stronger relationships had been reported be-
tween self-esteem and scores on various achievement tests.
Creativity, however, was not consistently related to any form of
self-regard.
The most definitive compilation was Hansford and Hattie’s
(1982) meta-analysis of 128 studies involving more than
200,000 participants. These studies explored a variety of mea-
sures of self-regard (mostly self-esteem) and a variety of objec-
tive performance measures, most of which were achievement
tests. The correlations reported varied widely, from
.77 to
.96, and averaged between
.21 and
.26 (depending on
how the average was computed, how studies were weighted,
etc.). Hansford and Hattie concluded that overall there is a sig-
nificant positive relationship between self-esteem and aca-
demic performance, with self-esteem accounting for between 4
and 7% of the variance in academic performance.
More recent studies have yielded similar conclusions. Using
standard achievement tests, Davies and Brember (1999) found
significant though weak positive relationships between self-
esteem and academic performance in a large (
N
3,001) Brit-
ish sample. The correlations ranged from .10 to .13, and averaged
.12. A somewhat stronger relationship was found by Bowles
(1999), who showed that self-esteem correlated at .29 with stu-
dents’ most recent semester grades in mathematics and En-
glish. Kugle, Clements, and Powell (1983) found that scores on
a reading achievement test correlated .18 with level of self-
esteem. However, when these authors controlled for ethnicity, the
effect of self-esteem was no longer significant. Thus, these re-
cent studies also indicate that self-esteem goes with doing well
in school, although the relationship is weaker than one might
have expected in a society that values doing well in school.
As already noted, people with high self-esteem report their
intelligence to be high, although there is no relationship be-
tween self-esteem and scores on objective IQ tests (Gabriel et
al., 1994). Simon and Simon (1975) found scores on self-esteem
to be correlated significantly (
r
.33) with scores on academic
achievement tests and also with IQ test scores (thus contradict-
ing the null result obtained by Gabriel et al.). In general,
though, there is very little evidence that self-esteem correlates
with IQ or other academic abilities.
Self-esteem also predicts performance in minority and at-
risk samples. Ortiz and Volloff (1987) found significant corre-
lations between self-esteem and tests of IQ and school abilities,
using a limited sample of Hispanic students in grades 3 to 6
who had been nominated for testing as candidates for gifted
classes. Howerton, Enger, and Cobbs (1994) studied at-risk
Black male students, and found that self-esteem predicted
grades and school achievement. Although Howerton et al. used
objective measures (school records) of achievement, a draw-
back of the study is that the sample was very small (
N
42).
Different authors have drawn very different conclusions
from correlations between self-esteem and school perfor-
mance. Among the most optimistic was that of Zimmerman,
Copeland, Shope, and Dielman (1997): “Efforts either to pre-
vent and stabilize decreasing self-esteem or to build self-esteem
PSYCHOLOGICAL SCIENCE IN THE PUBLIC INTEREST
R.F. Baumeister et al.
VOL. 4, NO. 1, MAY 2003
11
may have vital effects on . . . outcomes for youth” (p. 137). In
their own study, Zimmerman et al. measured performance by
asking students for general ratings of their grades (e.g., “Are
your grades mostly A’s, mostly A’s and B’s . . . ?”), so their re-
sults were vulnerable to subjective bias. In addition, the corre-
lations with self-esteem were very weak. Their results do not
seem to support their confidence that high self-esteem leads to
better grades.
Other authors have inferred that significant correlations be-
tween self-esteem and school performance support the oppo-
site causal conclusion, namely, that good work in school leads
to high self-esteem. Bowles (1999) specifically measured self-
esteem after the semester for which he obtained school grades,
so the positive correlation he found seems most consistent with
the view that self-esteem is a result, not a cause, of doing well
in school. This conclusion was further supported by path analy-
sis (a statistical technique for testing theories about complex
chains of causes), which indicated that there was no direct
causal path from self-esteem to achievement.
Still other researchers have concluded that the correlations
between self-esteem and school performance, albeit signifi-
cant, are so small as to be not worth pursuing. Rubin, Dorle,
and Sandidge (1977) found that self-esteem was significantly
correlated with all their measures of achievement, as well as
with teachers’ ratings of students’ behavior and performance.
Yet statistical analysis showed that taking self-esteem into ac-
count barely improved the accuracy of predictions of achieve-
ment that were based on socioeconomic status and intelligence
(IQ) alone. Rubin et al. wrote, “While these increases were sig-
nificant, their practical significance is negligible” (p. 506), and
they concluded that the links between self-esteem and aca-
demic performance are based on “common underlying factors
such as ability and background” (p. 503).
These and other findings generally point to a positive but
weak and ambiguous relationship between self-esteem and
school performance. Students with high self-esteem generally
have done somewhat better in school and on school achieve-
ment tests than students with low self-esteem. The correla-
tional findings do not indicate whether self-esteem is a cause or
a result of school performance. They do, however, furnish one
possible explanation for the continuing belief that self-esteem
may be beneficial for school performance. Teachers, parents,
and others may observe that high self-esteem and good school
performance go together and infer that self-esteem plays a
causal role. Unfortunately, impressions—even when backed up
by significant correlations—do not justify causal conclusions.
We now turn to studies that have investigated whether there is a
causal relationship between self-esteem and academic perfor-
mance.
Investigating Causality
Several studies have investigated the time course of the pos-
itive relation between self-esteem and academic performance,
as a way of establishing causal priority. Some of these studies have
also investigated whether third variables, such as socioeco-
nomic status or intelligence (IQ), could be responsible for the
correlations between self-esteem and academic performance.
An early and still well respected study by Bachman and
O’Malley (1977) used data from a nationwide longitudinal
study that tracked more than 1,600 young men from 1966,
when they were in 10th grade, up through 1974. All partici-
pants in this study completed a modified version of the Rosen-
berg (1965) self-esteem scale at several points during this
period. Although Bachman and O’Malley found that self-es-
teem correlated with school performance, their more sophisti-
cated statistical tests (i.e., path analyses) did not point to any
causal role for self-esteem. Instead, they concluded that shared
prior causes, including family background, ability, and early
school performance, affect self-esteem and later educational at-
tainment and were responsible for the correlation between the
two. They also concluded that occupational success caused
self-esteem to rise, whereas obtaining higher education had a
negligible impact on self-esteem. Of all their findings, the one
that gives the most credence to the view that self-esteem is an
important cause of successful outcomes is that self-esteem in
high school predicted eventual level of educational attainment
(final degree earned), but in their path analysis the direct link
from high school self-esteem to later educational attainment
was only .072. Its link to eventual occupational status was sim-
ilar, at .061. These numbers indicate that the relationship is ex-
tremely weak, if it exists at all. Neither link was significant.
The findings of Bachman and O’Malley (1977) are important
for several reasons. First, these researchers were perhaps the first
to conduct such a thorough and sophisticated study of the impact
of self-esteem. Second, they plainly had hoped to find that self-
esteem played a causal role, and they favored this hypothesis in
their initial exposition. Indeed, they proposed that high self-es-
teem fosters high aspirations and persistence, which lead ulti-
mately to better academic performance and career success.
Third, most subsequent work has generally confirmed their con-
clusions that self-esteem is a result rather than a cause, and that
any correlations between self-esteem and achievement are likely
to be due to third variables such as family background.
Another milestone study was conducted by Maruyama, Ru-
bin, and Kingsbury (1981). This research is an important com-
plement to Bachman and O’Malley’s (1977) study because it
focused on a much younger age: Maruyama et al. followed a fi-
nal sample of more than 700 students from age 4 to age 15.
Achievement was measured using academic achievement tests,
including the Stanford Achievement Test and, later, the Wide
Range Achievement Test, which emphasizes spelling, vocabu-
lary, and arithmetic. Like Bachman and O’Malley, Maruyama
et al. found that self-esteem and academic achievement are cor-
related, but concluded that there is no causal relationship be-
tween those variables. Instead, they argued that ability (IQ) and
social class are the underlying causal factors that affect the lev-
els of both self-esteem and academic achievement.
PSYCHOLOGICAL SCIENCE IN THE PUBLIC INTEREST
Benefits of Self-Esteem
12
VOL. 4, NO. 1, MAY 2003
A follow-up by Bachman and O’Malley (1986) confirmed
their previous findings. In this follow-up, they sought to im-
prove their methodological rigor in several ways. One of these
was to confine analyses to White males in White-majority
schools. Although the exclusion of minorities is not considered
methodologically desirable today, Bachman and O’Malley had
quite persuasive reasons for deciding that such inclusion might
have weakened the chance for self-esteem to emerge as caus-
ally significant. Specifically, in their original sample, African
American students scored lower than White students on achieve-
ment and ability tests but higher on self-esteem. Combining data
from White and African American students could therefore ob-
scure a positive relationship between self-esteem and achieve-
ment. The authors also acknowledged that the low achievement
scores of African American students might indicate some lack
of validity of those measures for such a sample. Any lack of
validity would make self-esteem look less effective than it ac-
tually is.
Despite their effort to increase the potential for their study
to demonstrate effects of self-esteem, Bachman and O’Malley
(1986) found that global self-esteem had a negligible relation-
ship to eventual educational attainment. Family background
(socioeconomic status), ability (IQ), and early school grades
predicted eventual level of educational attainment, and self-
esteem added little to the accuracy of prediction. Self-esteem
was correlated with actual ability, although self-rated ability
was consistently inflated: People thought they were smarter
than they actually were. In short, these findings support the role
of illusion in self-esteem, but they contradict the view that self-
esteem causes long-term educational success.
An article titled “Is There a Causal Relation Between Self-
Concept and Academic Achievement?” was published by Pot-
tebaum, Keith, and Ehly (1986). Their answer was a rather
blunt negative: “The results suggest that there is no significant
causal relation between self-concept and academic achieve-
ment [in either direction], but rather that the observed relation
is the result of one or more uncontrolled and unknown third
variables” (p. 142). A great deal of methodologically sophisti-
cated work went into producing that conclusion. Pottebaum et
al. used a sophisticated research design, testing a very large
sample of high school students (more than 23,000) in the 10th
grade and again in the 12th grade. Self-esteem in 10th grade
predicted academic achievement in 12th grade quite weakly (
r
.11). Conversely, academic achievement in 10th grade pre-
dicted self-esteem in 12th grade only trivially better (
r
.12).
The authors noted that a reciprocal causal relationship between
self-esteem and academic performance could produce the pat-
tern of findings they obtained—but only if the two variables
cause each other with about the same amount of power, which
seemed a priori implausible. Hence, Pottebaum et al. con-
cluded it is more likely that self-esteem and academic perfor-
mance are both the result of a third variable (or set of
variables). This conclusion is clearly consistent with what
Bachman and O’Malley (1977) and Maruyama et al. (1981)
found, despite different methods and somewhat different nu-
merical results.
A slightly different conclusion was reached by Rosenberg,
Schooler, and Schoenbach (1989). They analyzed data from the
Youth in Transition longitudinal study that was also the basis
for Bachman and O’Malley’s (1977) work. They used the 10th-
and 12th-grade data for nearly 1,900 boys, including measure-
ments of achievement that relied on having the students report
their grade point average in school. This sort of measure is in
the middle of the span of methodological rigor that we have
outlined: It is not fully objective, but it is somewhat specific
and verifiable. A slight tendency of people with high self-
esteem to furnish self-flattering reports is to be expected, al-
though the scope for such inflation may be more limited than if
respondents are asked a general self-rating question, such as
“How good are you at school?”
Rosenberg et al. (1989) did find significantly positive, although
weak, correlations between self-esteem and self-reported grades,
r
.24 in 10th grade and
r
.25 in 12th grade. Of greater in-
terest were the findings across time. These supported the con-
clusion that self-esteem is the result of grades, rather than the
cause. There was a modest causal relationship (.15) leading
from grades to self-esteem, but the causal relationship leading
from self-esteem to grades was only .08, which was not signifi-
cantly different from zero. In other words, there was no solid
evidence that self-esteem had any effect on grades, despite the
fact that even a weak relationship would likely have been sig-
nificant because Rosenberg et al. tested so many people.
The view that self-esteem is an outcome rather than a cause
of good school performance was further supported in another
large and sophisticated study, by Skaalvik and Hagtvet (1990).
Their sample consisted of 600 Norwegian schoolchildren in
two cohorts, one in third grade (about age 9) and the other in
sixth grade at the start of the study. A second set of data was
obtained a year and a half later. Achievement was measured by
teachers’ ratings, which furnish a good measure although erro-
neous perceptions by the teacher could affect a student’s self-
esteem (e.g., if the teacher treats the student as a genius or dull-
ard). Skaalvik and Hagtvet found evidence that doing well in
school one year led to higher self-esteem the next year,
whereas high self-esteem did not lead to performing well in
school. In fact, high global self-esteem in grade 6 predicted
lower academic achievement in grade 7.
Skaalvik and Hagtvet (1990) also measured students’ self-
concept of their academic ability. Although our focus is not on
domain-specific measures of self-esteem in this review, the
findings are of interest. These researchers concluded that self-
concept of ability mediates the relation between academic per-
formance and global self-esteem. Specifically, doing well in
school leads to thinking of oneself as good at schoolwork,
which in turn can boost global self-esteem. There was also
some evidence for a causal influence of academic self-concept
on school performance, which is thus one finding that suggests
thinking well of oneself can lead to better schoolwork (al-
PSYCHOLOGICAL SCIENCE IN THE PUBLIC INTEREST
R.F. Baumeister et al.
VOL. 4, NO. 1, MAY 2003
13
though this relationship involves thinking of oneself as good at
schoolwork rather than as good overall). This causal influence
was found in some analyses but not others and was not strong.
Still, the fact that it was demonstrated at all helps dispel wor-
ries that some kind of methodological or measurement problem
renders it impossible to verify causal effects of self-concept.
This in turn lends further credence to the conclusion that global
self-esteem is not a cause of school performance.
Interventions
We found relatively little evidence on how self-esteem pro-
grams or other interventions affect self-esteem. Such interven-
tions are practiced in many schools and other places, but it is
common for them to target not only self-esteem but also study
skills, citizenship, conflict reduction, and other variables. Obvi-
ously, if a program that attempts to boost self-esteem and im-
prove study skills ends up producing an improvement in
grades, it is hardly safe to conclude that self-esteem is respon-
sible for the improvement. Furthermore, given that the studies
investigating causality have not demonstrated that self-esteem
has an impact on academic achievement, it seems likely that
any attempt to collect solid data on the impact of boosting self-
esteem would end up with null results, and the evaluators
would not be anxious to publish their results even if they could.
An impressive review of research on such programs was
published by Scheirer and Kraut (1979). The title, “Increased
Educational Achievement via Self-Concept Change,” sounds
promising with respect to the benefits of self-esteem, but the
findings were not. Scheirer and Kraut covered evidence from
both published and unpublished evaluations of school-based
interventions and programs, including Head Start, the Early
Training Project, and Upward Bound. By and large, the evalua-
tions of these studies “generally failed to find an association
between self-concept change and academic achievement” (Scheirer
& Kraut, 1979, p. 140). When there was an association, it
tended to be temporary (i.e., it was not maintained beyond the
end of the program), or it indicated that self-esteem was the re-
sult of academic achievement rather than the cause. Programs
that targeted factors other than self-esteem (such as by encour-
aging parents to become involved in their children’s school-
work) seemed to get better results.
Scheirer and Kraut (1979) carefully considered a variety of
factors that could have led to the general pattern of null results,
including poor measurement, methodological problems, and
failure to implement interventions properly, among others. Yet
as far as they could tell, these factors were not sufficient to ex-
plain the broad pattern of results. For example, they noted that
some interventions that did not target self-esteem had produced
measurable gains in academic performance. Hence, they con-
cluded that the most likely explanation of the disappointing re-
sults was that the basic theoretical hypothesis—namely, that
improving self-esteem will lead to better academic perfor-
mance—was wrong. They said that self-esteem may be an out-
come of academic achievement, but it does not appear to be a
cause or a mediating variable.
A nicely controlled field experiment by Forsyth and Kerr
(1999) provided converging evidence using an adult (college
student) sample. This investigation was conducted in connec-
tion with a regular course. Students who received a C, D, or F
on the first examination were targeted to receive weekly e-mail
messages from the professor. Each message contained a review
question pertaining to that week’s assignment. In the control
condition, the review question was all that was included in the
e-mail. Students who were randomly assigned to two other
groups, however, received either a message aimed at boosting
their sense of personal control and responsibility for their own
performance or a message aimed at boosting and maintaining
their positive sense of self-worth. These manipulations had no
effect on the C students, but the D and F students who received
the self-esteem boost performed significantly worse on subse-
quent tests than the D and F students in the other conditions.
Forsyth and Kerr’s (1999) study is of interest because it in-
volved full random assignment and because it found significant
differences as a function of treatments aimed at self-esteem.
Unfortunately for the self-esteem movement, the findings sug-
gest that the intervention aimed at boosting self-esteem was
counterproductive in its impact on academic performance. We
have already noted some evidence linking high self-esteem to
subsequently poorer performance (e.g., Skaalvik & Hagtvet,
1990), although the preponderance of findings suggests that
self-esteem is positively linked to academic achievement. Still,
Forsyth and Kerr used an intervention rather than simple mea-
surement of current self-esteem, and they aimed their interven-
tion specifically at low-performing students. Their results are
consistent with the view that self-esteem is an important rein-
forcer for good academic performance and that supplying the
reward indiscriminately (i.e., not linking it to good perfor-
mance) may deflate its reward value. In plainer terms, students
may ordinarily work hard in order to be permitted to feel good
about themselves, and an intervention that encourages them to
feel good about themselves regardless of work may remove the
reason to work hard—resulting in poorer performance.
Conclusion
The impact of self-esteem on school performance has been
studied more carefully and thoroughly than any other outcome
we discuss in this review. We were able to find studies that used
longitudinal designs and other methods that can address causa-
tion, used objective methods rather than relying on self-report,
and attempted to untangle self-esteem from other variables.
The results do not support the view that self-esteem has a
strong causal effect on school achievement. Indeed, most of the
evidence suggests that self-esteem has no impact on subse-
quent academic achievement. The few studies suggesting any
positive causal impact of self-esteem generally found only tiny
PSYCHOLOGICAL SCIENCE IN THE PUBLIC INTEREST
Benefits of Self-Esteem
14
VOL. 4, NO. 1, MAY 2003
effects. Some findings even point (again weakly) in the oppo-
site direction, suggesting that high or artificially boosted self-
esteem may detract from subsequent performance.
There were in fact some reasons to hope that self-esteem
would be more potent. On a theoretical basis, self-esteem
seemed likely to enhance academic strivings and persistence.
Early empirical findings may also have encouraged the belief
that self-esteem is helpful, because simple correlations be-
tween self-esteem and academic performance have often been
positive and significant (as noted by Wylie, 1979). Unfortu-
nately, those correlations appear to be due to processes other
than self-esteem causing good performance. First, good perfor-
mance in school may sometimes lead to higher self-esteem, in-
stead of the reverse. (Even that tendency is disappointingly
weak, however.) Second, self-esteem overlaps with other vari-
ables, and when these are controlled in the statistical analyses,
the proportion of variance in performance accounted for by
self-esteem dwindles rapidly.
JOB AND TASK PERFORMANCE
Applied to the world of work, the self-esteem hypothesis
suggests that people who feel better about themselves perform
better. Early reviews of studies on the relationship between self-
esteem and work performance noted the high variability of the
reported size and nature of this relationship (Brockner, 1983;
Tharenou, 1979). Judge and Bono (2001) surveyed the results
of 40 studies with more than 5,000 participants and found that
most of them showed weak positive relationships. In some stud-
ies, self-esteem was measured with global indices, such as the
Rosenberg scale, whereas in others, it was measured more spe-
cifically as organization-based self-esteem. Similarly, there were
differences in the way performance was measured. Some stud-
ies employed self-perception measures, whereas others used su-
pervisors’ ratings or objective measures. Whether the association
between self-esteem and performance was moderated by such
differences in measurement was not reported.
Ultimately, the results regarding the link between self-esteem
and job performance seem to echo what has been found with
school performance: The simple search for correlations yields a
modest positive relationship, such that high self-esteem goes
with better performance. However, the reported link between
self-esteem and job performance is highly variable, in contrast
to the more consistent findings regarding school performance. A
likely explanation is that the relevance of self-esteem to work
performance varies substantially because the demands and re-
wards of different occupations are highly variable.
The most important point, however, is that correlational
findings do not permit causal inferences. Self-esteem might
improve job performance. Then again, occupational success
might well boost self-esteem, whereas failure at one’s job may
deflate it. As in the case of school performance, time-lagged
longitudinal studies are useful for determining whether self-
esteem is a result or a cause of good performance, but these
have not yet been done.
A quite different approach to studying how self-esteem may
influence performance is to investigate the relationship in the
laboratory. Such studies typically measure self-esteem as a pre-
existing trait and then investigate task performance in the labo-
ratory. In most cases, researchers manipulate other variables
such as initial failure or size of reward to investigate how they
interact with self-esteem.
Performance Quality
Relatively few studies on the effect of self-esteem on perfor-
mance quality have been published, and this leads us to suspect
that the results are typically nonsignificant. We can point to ev-
idence from our own laboratories. Wallace and Baumeister
(2002) had students solve arithmetic problems under varying
levels of challenge and performance pressure. There was no
effect of self-esteem on any of the performance measures.
Baumeister, Heatherton, and Tice (1993) examined performance
at a video game that was presented as a test of nonverbal intel-
ligence. Across multiple studies, there was essentially no effect
of self-esteem on performance. The control conditions (which
indicate whether there is a link between self-esteem and perfor-
mance in the absence of special interventions) typically showed
no difference in performance as a function of self-esteem,
whereas ego threat in the form of critical feedback on a prior
test caused people with high self-esteem to perform worse than
those with low self-esteem.
In contrast, J.D. Campbell and Fairey (1985) found that
people with high self-esteem performed better than those with
low self-esteem on an anagram test when participants were first
asked to write explanations for a hypothetical failure (there
were no self-esteem differences when participants first wrote
explanations for a hypothetical success or in a control condi-
tion). Thus, when the prospect of a possible failure is salient,
higher levels of self-esteem may benefit task performance. In
general, though, self-esteem seems to have little or no direct re-
lationship to task performance.
Persistence at Tasks
Conventional wisdom claims that persistence is a vital as-
pect of performance (at both school and work). The research on
persistence has generally been consistent with the notion that a
core characteristic of people with high self-esteem is their ten-
dency to persist in the face of failure (e.g., McFarlin, Baumeis-
ter, & Blascovich, 1984; Perez, 1973; Shrauger & Sorman,
1977, Study 1).
The conventional wisdom notwithstanding, blind persis-
tence is not an ideal self-regulatory strategy; there is also value
in knowing when to quit. Some studies have not found a corre-
lation between greater persistence and high self-esteem, but
these seeming failures to replicate suggest that individuals with
PSYCHOLOGICAL SCIENCE IN THE PUBLIC INTEREST
R.F. Baumeister et al.
VOL. 4, NO. 1, MAY 2003
15
high self-esteem may make better use of cues telling them
whether it is prudent to persist or withdraw. For example, Jan-
off-Bulman and Brickman (1982) and McFarlin (1985) told
some participants that some of the tasks that they would work
on did not have solutions. This information caused high self-
esteem participants to persist less (compared with the control
participants who did not receive this information), but did not
affect the persistence of low self-esteem participants. In a simi-
lar vein, Sandelands, Brockner, and Glynn (1988) advised
some participants that persistence was a wise strategy and oth-
ers that it was a less prudent strategy. People with high self-
esteem persisted more in the former condition than in the latter,
whereas those with low self-esteem were relatively unaffected
by the manipulation.
Thus, when people are given advice or cues that persistence
may not be a good strategy, those with high self-esteem persist
less than those with low self-esteem. Results were rather differ-
ent, however, when McFarlin et al. (1984) directly told partici-
pants what to do when they encountered a difficult problem
(quit or persist). In this case, low self-esteem participants were
responsive to the instructions, whereas high self-esteem partic-
ipants were not. Taken together, these studies suggest that peo-
ple with high self-esteem make better use of situational cues in
deciding for themselves the appropriate course of action, but
that people with low self-esteem may be more responsive to di-
rections that simply tell them what to do.
Although these experiments are interesting, in the real
world individuals are rarely (if ever) provided with this kind of
information. Indeed, persistence is a ubiquitous phenomenon
precisely because people are not given such direct advice. The
experiments also differ from natural conditions in that experi-
mental participants experience only a single instance of failure,
and there are no opportunities to pursue alternative goals or
tasks (Bandura, 1989). Under these conditions, persistence at
the failed task seems to be a reasonable strategy. In contrast, if
one has experienced multiple failures on a task and, as is typi-
cally the case outside the laboratory, there are opportunities to
pursue alternative goals, it seems more adaptive to abandon the
original task in favor of an alternative.
Di Paula and Campbell (2002) recently examined the rela-
tionship between level of self-esteem and knowing when to
quit. First, they conducted a laboratory experiment in which
they manipulated the degree of failure (a single, initial failure
or three instances of failure) and the availability of an alterna-
tive goal. Compared with low self-esteem participants, high
self-esteem participants persisted more after a single failure,
but less after repeated failure when an alternative was avail-
able. In a second correlational study, participants enumerated
10 specific goals at the beginning of an academic year and at
the end of the year indicated, for each goal, their perceived
progress and behavioral persistence, as well as how much they
had ruminated about the goal. Self-esteem was positively cor-
related with achieving more goals, more satisfaction with
progress toward goals, more behavioral pursuit of goals, and
less rumination. More important, compared with participants
with low self-esteem, those with high self-esteem showed
higher correlations between their ratings of perceived progress
and behavioral persistence across the 10 goals.
Taken together, this research supports the notion that high
self-esteem individuals utilize more adaptive self-regulatory
strategies than low self-esteem individuals; these strategies
may, in fact, be a factor in their higher levels of reported happi-
ness and satisfaction (reviewed in a later section). The fact that
high self-esteem people have more functional responses to fail-
ure may also convey slight advantages in performance, and
thus contribute to the (weak) associations between self-esteem
and performance.
Conclusion
High self-esteem has value in causing people to persist
longer in the face of failure, at least when there is no viable al-
ternative task or goal to pursue and when no cues suggest that
persistence is not a good strategy. When there is an alternative
goal or information that persistence may be a poor strategy,
high self-esteem is also associated with knowing when to quit.
Thus, in performance contexts, high self-esteem people appear
to use better self-regulation strategies than low self-esteem
people.
It is difficult to draw firm causal conclusions about self-
esteem and other aspects of job and task performance. Overall,
there are weak positive correlations between job performance
and self-esteem, but these may be due in whole or in part either
to successful performance improving self-esteem or to self-
report biases. There is no strong evidence indicating specifi-
cally that high self-esteem leads to improved performance on the
job. The links between self-esteem and job performance de-
serve more systematic and careful study, using objective mea-
sures and longitudinal designs. However, if high self-esteem
consistently produced better performance in laboratory tasks,
this would be well established by now, and the lack of such ev-
idence suggests to us that self-esteem has little impact on task
performance.
INTERPERSONAL RELATIONS
Another major sphere of life activity is interpersonal rela-
tions. The apparent failure of self-esteem to contribute directly
to improvements in academic and job performance would eas-
ily be offset if self-esteem helped people to get along better
with others.
It is certainly plausible that high self-esteem would improve
interpersonal relations. People with high self-esteem claim to
be more popular than people with low self-esteem (Battistich,
Solomon, & Delucchi, 1993), although as already noted, self-
reports are not reliable guides to objective realities. High
self-esteem might well make a person more likable insofar as
people may prefer to interact with confident, enterprising indi-
PSYCHOLOGICAL SCIENCE IN THE PUBLIC INTEREST
Benefits of Self-Esteem
16
VOL. 4, NO. 1, MAY 2003
viduals and to avoid interacting with people who suffer from
self-doubts and insecurities. The reverse causal relationship is
also quite plausible, of course, and indeed it forms the center-
piece of the sociometer theory of self-esteem put forward by
Leary and his colleagues (Leary, Tambor, Terdal, & Downs,
1995; see also Leary & Baumeister, 2000; Leary & Downs,
1995). According to this view, self-esteem operates as an inter-
nal measure of one’s interpersonal appeal and success. Popu-
larity would therefore cause self-esteem to rise, whereas social
rejection would cause it to drop. It is also plausible that high
self-esteem elicits dislike, insofar as people reject conceited or
arrogant individuals.
Friendship Patterns and Interpersonal Interactions
As already noted, people with high self-esteem claim to be
especially popular (Battistich et al., 1993). They also rate their
friendships as being of higher quality compared with people
with low self-esteem (Keefe & Berndt, 1996). People low in
self-esteem report more negative, aversive social interactions,
more stressful life events, and less social support than people
high in self-esteem (Lakey, Tardiff, & Drew, 1994). In the
workplace, people with high self-esteem claim to get along bet-
ter with their co-workers (although not with their supervisors),
compared with people with low self-esteem (Frone, 2000).
Thus, self-reports indicate that the social lives of people with
high self-esteem are far better, richer, and more satisfying than
the lives of people with low self-esteem. But are these claims
justified?
In a large sociometric study by Bishop and Inderbitzen
(1995), 542 ninth-grade students nominated their most-liked
and least-liked peers. On the basis of these votes, each person
was put into one of five categories along a span of degree of
popularity (i.e., popular, average, controversial, neglected, re-
jected). Perhaps surprisingly, these categories showed no corre-
lation with self-esteem whatsoever. This finding contradicts
both the view that high self-esteem is socially appealing and the
theory that self-esteem results from peer approval. One finding
that offered minimal support for the general self-esteem hypoth-
esis was that people who had no reciprocal friendships at all ex-
hibited lower self-esteem than people who had at least one.
Beyond one reciprocal friendship, however, self-esteem ap-
peared to be irrelevant to friendship status and popularity.
Glendinning and Inglis (1999) categorized students into
four distinct social groups on the basis of their self-reported at-
titudes about peer relations and their own popularity and then
related group membership to the students’ self-esteem levels.
The social groups were called “peer oriented” (i.e., high status,
popular with peers), “conventional” (i.e., middle status, accept-
ing of authority), “socially isolated” (i.e., hard to make and
keep friends), and “disaffected” (i.e., troublesome, disrespect-
ful of authority). Results showed that adolescents with low
self-esteem were relatively more common in the socially iso-
lated and disaffected categories, whereas those with high self-
esteem were more often found in the conventional and peer-
oriented categories. Unfortunately, because these findings were
based on students’ self-rated popularity rather than peer rat-
ings, they may reflect the tendency of people with high self-
esteem to rate themselves favorably. Without objective mea-
sures, it is hard to know whether some students were actually
more popular or merely believed themselves to be so.
A somewhat better approach was taken by Dolcini and
Adler (1994), who asked a subsample of eighth graders to indi-
cate which peer group, or “crowd,” every other student in the
eighth grade belonged to, and then obtained self-esteem scores
from all the students directly. Global self-esteem did not differ
among the different crowds, although there were some effects
for domain-specific self-esteem. For example, the academically
oriented “smart” crowd scored higher than the jocks on aca-
demic self-esteem, and the athletic crowd scored higher than
the bookworms on athletic self-esteem.
Adams, Ryan, Ketsetzis, and Keating (2000) used teachers’
ratings as the basis for evaluating the interpersonal behavior and
social desirability of 300 schoolchildren in grades 4 and 7 in
Ontario, Canada. Two main types of desirable behavior were the
focus of the ratings: peer sociability, which included popularity
and good relationships, and desirable classroom behavior,
which included following rules and behaving properly in class.
Neither popularity nor classroom behavior varied as a function
of self-esteem. Several background factors and child character-
istics did successfully predict these outcomes, though (e.g., high
scores on assertiveness were correlated with higher popularity
but also higher levels of rule breaking in class). Thus, the null
findings for self-esteem do not appear to reflect a general failure
of measurement or a high amount of random fluctuation. Self-
esteem did show some correlations with some of the back-
ground variables, such as family cohesion, but even though
these background variables predicted popularity and classroom
behavior, self-esteem failed both as a possible mediator and as
an independent predictor of the outcomes. The authors con-
cluded glumly that their findings “call into question whether
self-esteem is a potential causal agent or even a correlate of in-
terpersonal relations or classroom citizenship” (p. 247).
A smattering of methodologically sound studies with adult
college students has examined the social interaction patterns of
people high in self-esteem. An important study by Buhrmester,
Furman, Wittenberg, and Reis (1988) examined a variety of in-
terpersonal behaviors. All the students were rated both by them-
selves and by their roommates, so the method offered a valuable
opportunity to compare subjective and external perspectives. As
usual, people high in self-esteem rated themselves as signifi-
cantly better than other people on all interpersonal skills: They
claimed to be substantially better at initiating relationships (r
.63), better at disclosing things about themselves (r .41), bet-
ter at asserting themselves in connection with objectionable be-
haviors by others (r .40), better at providing emotional
support to others (r .30), and significantly better even at man-
aging interpersonal conflicts (r .20).
PSYCHOLOGICAL SCIENCE IN THE PUBLIC INTEREST
R.F. Baumeister et al.
VOL. 4, NO. 1, MAY 2003 17
The roommates’ ratings told a very different story, however.
For four of the five interpersonal skills, the correlation between
self-rated self-esteem and roommate-rated skill fell short of
significance, ranging from .01 for conflict management to .15
for assertion in the face of objectionable behaviors. In the case
of initiating new social contacts and friendships, the link to
self-esteem (r .38) remained significant but was substan-
tially lower than the link between self-esteem and self-rated
skill at making new friends. A troublesome aspect of this
study’s findings was that roommate-rated self-esteem corre-
lated better than self-rated self-esteem with roommate-rated
social skills, and the correlations between roommate-rated self-
esteem and roommate-rated interpersonal skills were quite
similar to the correlations between self-rated self-esteem and
self-rated interpersonal skills. Self-esteem almost certainly has
to be more accurate when reported by the self than when esti-
mated by others, and so these findings raise the worrisome pos-
sibility that the correlations were due to some kind of response
bias that resulted in answers to questions about self-esteem be-
ing similar to answers to questions about interpersonal skills.
The findings of Buhrmester et al. (1988) thus suggest that
people high in self-esteem regard themselves as having an im-
pressive range of superior social skills, but that these skills are
not readily discerned or confirmed by others. By and large,
roommates did not find their high self-esteem roommates to be
any more socially skilled than their low self-esteem room-
mates. The only significant exception involved skill in initiat-
ing new relationships and interactions. This does seem to be
one sphere in which confidence could come into play: People
who think that they are desirable and attractive might well be
willing to strike up conversations with strangers, possibly be-
cause they do not expect rejection. In contrast, people with low
self-esteem might shy away from initiating such contacts.
Interpersonal interactions were studied in the laboratory by
Brockner and Lloyd (1986). All subjects first rated their self-
esteem and then had a 10-min “get acquainted” conversation
with a member of the opposite sex. Care was taken to ensure
that conversational partners did not know each other prior to
the session. Afterward, they answered questions about their im-
pressions of each other. In addition to these peer ratings, self-
ratings of likability were obtained. People with high self-es-
teem estimated that their partners had liked them much more
than people with low self-esteem, but partners’ actual ratings
did not show even a hint of this correlation. It is also notewor-
thy that people with low rather than high self-esteem appear to
have been responsible for the discrepancy between self-ratings
and partner ratings: People low in self-esteem estimated that
their partners liked them much less than the partners actually
indicated, whereas people high in self-esteem estimated their
partners’ ratings quite accurately. These results dovetail with
those of Buhrmester et al. (1988): High self-esteem makes for
thinking oneself popular and likable, but in reality people with
high and low self-esteem are equally likable.
Similar findings were reported by J.D. Campbell and Fehr
(1990). After a 15-min get-acquainted conversation, people es-
timated how their partner and an observer would evaluate them
and rated how favorably they evaluated their partner. Ratings
were also obtained from uninvolved observers. People with
high self-esteem estimated that their partners and observers
would rate them more positively than people with low self-
esteem, but there was no difference in the actual ratings by
partners or by observers. As in the study by Brockner and
Lloyd (1986), people high in self-esteem estimated quite accu-
rately how much their partners liked them (low self-esteem
people underestimated how much their partners liked them).
However, because the observers rated everyone more harshly
than did interaction partners, low self-esteem people were
more accurate in their estimates of the observers’ ratings.
A more complex laboratory investigation of self-esteem and
interpersonal liking was recently published by Heatherton and
Vohs (2000). They reasoned that the null findings of Brockner
and Lloyd (1986) might have been due in part to the relatively
neutral context, so they introduced a manipulation of ego
threat. Specifically, prior to the conversational interaction, half
of the participants took a difficult intellectual test and received
esteem-threatening feedback that they had performed far worse
than their peers. The other participants took a neutral version of
the same test, which was labeled as a pilot task, and received
no feedback. Then all participants had a structured, 20-min
conversation with a naive participant involving topics of low,
medium, and high intimacy. Finally, the participants rated how
much they liked each other.
Heatherton and Vohs’s (2000) results showed that self-esteem
interacted with ego threat. In the control condition, in which
there was no blow to anyone’s pride or sense of competence,
people were liked equally regardless of their level of self-esteem.
In that respect, the results replicated the findings of Brockner
and Lloyd (1986) and J.D. Campbell and Fehr (1990). In the
ego-threat condition, however, differences did emerge. People
with high self-esteem were given lower ratings than people
with low self-esteem. These findings directly contradict the
view that high self-esteem leads to better interpersonal liking.
(They also contradict the so-called buffer hypothesis, which
holds that people with high self-esteem cope better with stress
and adversity. We discuss this hypothesis further in the section
on Coping and Depression.) Additional evidence suggested
that the decreased liking for people with high self-esteem in the
ego-threat condition compared with the control condition was
due to their being perceived as antagonistic. In contrast, and
perhaps counterintuitively, low self-esteem people were liked
significantly more in the ego-threat condition than in the con-
trol condition.
These findings were corroborated by Vohs and Heatherton
(in press) in a naturalistic study of 372 college men and women
who were rated by their peers. Participants’ global self-esteem
was assessed during the spring of their senior year in high
school and then again midway through their 1st year at an Ivy
League college. Those whose self-esteem dropped substan-
PSYCHOLOGICAL SCIENCE IN THE PUBLIC INTEREST
Benefits of Self-Esteem
18 VOL. 4, NO. 1, MAY 2003
tially (i.e., by 1 SD), presumably because the move to college
shook their positive sense of self, were placed in the ego-threat
group, and compared with students whose self-esteem did not
drop. This study supported Heatherton and Vohs’s (2000) pre-
vious studies in showing that high self-esteem men and women
who were threatened were liked least by their peers. Once
again, being seen as antagonistic mediated the effect.
Vohs and Heatherton (2001) looked into possible mecha-
nisms underlying these results. They found that high and low
self-esteem people think about themselves differently after
threat, such that high self-esteem people become more inde-
pendent and less interdependent (i.e., they emphasize self-reli-
ance and believe that they really do not need other people),
whereas the opposite is true for low self-esteem people (i.e.,
they emphasize interpersonal relations, even at the expense of
the self). In the absence of threat, however, high and low self-
esteem people show no differences in concern for the self ver-
sus interpersonal relations. Further, Vohs and Heatherton found
that it is seeing the self as separate versus connected to others
that drives differences in likability and personality perceptions
between high and low self-esteem people who have experi-
enced an ego threat.
Converging evidence was recently provided by Bonanno,
Field, Kovacevic, and Kaltman (2002). They found that un-
trained observers disliked people prone to self-enhancement.
Among the same group of participants, the size of the tendency
toward self-enhancement was associated with the self-reported
ability to cope with stress and trauma (see the section on Cop-
ing and Depression). Put another way, holding an inflated view
of themselves helped people feel good and recover from
trauma, but also tended to alienate others.
Other evidence, however, suggests that people with high
self-esteem are liked better if certain conditions are met. Joiner
et al. (1992) conducted a 5-week study of same-gender room-
mates, measuring roommates’ liking using a modified version
of the Rosenberg scale. Thus, if roommates rated people as
they rated themselves, people low in self-esteem would appear
to be disliked, and people with high self-esteem would appear
to be liked. This procedure therefore made it extra likely that
people with high self-esteem would emerge as well liked. Even
so, people with low self-esteem were liked less than others
only if they were male, depressed, and prone to reassurance
seeking (an annoying pattern of interpersonal tactics designed
to elicit positive, supportive comments from others). The re-
sults for females, nondepressed individuals, and other partici-
pants did not show greater dislike or disparagement of people
with low self-esteem than people with high self-esteem.
Romantic and Intimate Relationships
The role of self-esteem in romantic relationships has re-
ceived fairly little attention. In particular, little is known about
whether self-esteem predicts the durability of romantic rela-
tionships. One study with a very small sample (N 30) found
that couples with low self-esteem were more likely than cou-
ples with high self-esteem to break up over a 1-month period
(S.S. Hendrick, Hendrick, & Adler, 1988). Data on love styles
and self-esteem support this finding, showing that low self-
esteem is related to feelings of manic love, which is character-
ized by extreme feelings of both joy and anguish over the love
object (W.K. Campbell, Foster, & Finkel, 2002). High self-
esteem is related to passionate, erotic love, which is marked by
the escalation of erotic feelings for the love object. These find-
ings are consistent with other studies showing that, compared
with people with high self-esteem, those with low self-esteem
experience more instances of unrequited love (Dion & Dion,
1975) and more intense feelings of love for others (C. Hendrick
& Hendrick, 1986).
Several findings indicate that relationship behavior differs
as a function of self-esteem. Murray, Rose, Bellavia, Holmes,
and Kusche (2002) found that people low in self-esteem en-
gage in a variety of potentially destructive behaviors. They tend
to distrust their partners’ expressions of love and support, and
so they act as though they are constantly expecting their part-
ners to reject and abandon them. Thus far, however, these patterns
have not translated into any evidence that the relationships are
actually more likely to dissolve.
Thus, despite the relationship problems caused by low self-
esteem, relationships are no more likely to break up if a partner
has low self-esteem than if a partner has high self-esteem. Pos-
sibly the reason for this is that high self-esteem leads to rela-
tionship problems, too. Rusbult, Morrow, and Johnson (1987)
examined four types of responses to problems within close rela-
tionships, and found that self-esteem produced the largest dif-
ference in the active-destructive (“exit”) category of responses.
People with high self-esteem were significantly more likely
than others to respond to problems and conflicts by deciding to
leave the relationship, seeking other partners, and engaging in
other behaviors that would actively contribute to the deteriora-
tion of the relationship. These results were based on responses
to hypothetical scenarios, which share many of the drawbacks
of self-report measures. However, as the authors noted, it seems
unlikely that their findings can be attributed to a simple re-
sponse bias because people with high self-esteem were admit-
ting to more undesirable, rather than desirable, behaviors.
Shackelford (2001) found that self-esteem was intertwined
with a variety of patterns in marriage, although he did not pro-
vide evidence as to whether high self-esteem affects the dura-
bility of marriages. Spouses showed similar levels of self-
esteem, with global self-esteem of spouses correlating at .23
and physical self-esteem (including self-rated attractiveness)
correlating fairly strongly at .53. Significantly, Shackelford re-
garded self-esteem as an outcome rather than a cause of marital
interactions, although his data were correlational. Wives’ fidel-
ity was the strongest predictor of husbands’ self-esteem. This
might indicate that men with high self-esteem cause their wives
to remain faithful, or—as Shackelford speculated—that cuck-
olded husbands experience a loss of self-esteem.
PSYCHOLOGICAL SCIENCE IN THE PUBLIC INTEREST
R.F. Baumeister et al.
VOL. 4, NO. 1, MAY 2003 19
Most important, women complained more about husbands
with low than with high self-esteem. Low self-esteem men
were derided by their wives as jealous, possessive, inconsider-
ate, moody, prone to abuse alcohol, and emotionally con-
stricted. Again, the direction of causality is difficult to
determine. Possibly, husbands’ low self-esteem elicits negative
perceptions among wives. Conversely, being disrespected or
despised by his wife may lower a man’s self-esteem. Yet an-
other possibility is that having a variety of bad traits leads both
to low self-esteem and to being disrespected by one’s wife.
Meanwhile, the self-esteem of wives was unrelated to their
husbands’ complaints about them, except that husbands who
criticized or insulted their wives’ appearance were generally
married to wives with low self-esteem, and indeed Shackelford
(2001) found that this was the most consistent predictor of low
self-esteem among wives.
Group Behavior and Leadership
Behavior in groups is an important complement to behavior
in intimate dyadic relationships. We found relatively little work
that shows objective behavioral effects of self-esteem in group
settings. One notable exception is an impressive longitudinal
study by LePine and Van Dyne (1998). These researchers com-
piled a broad sample of work groups, consisting in total of
nearly 600 employees in 21 different organizations. They fol-
lowed the groups for 6 months and obtained peer ratings of the
behavior of each participant. The main variable of interest was
“voice behavior,” defined as the willingness to speak within the
group, criticizing its operation or proposing alternative routes of
action. Voice behavior is vital to an organization’s long-term
success, because innovation depends on it, as do the group’s
chances for stopping itself from proceeding down a destructive
path (as in “groupthink”; Janis, 1982). Because speaking up and
criticizing the group require initiative and a willingness to devi-
ate from the group’s apparent consensus, the authors predicted
that high self-esteem would lead to greater voice behavior.
This prediction was confirmed, although the effect was only
marginally significant. Across the 6-month period, LePine and
Van Dyne (1998) found that high self-esteem led to more voice
behavior, though it accounted for only 1% of the differences
between people on that dimension. Moreover, it was indeed
people high in self-esteem who were most willing to speak out.
An additional 3% of the variance in voice behavior was ac-
counted for by interactions of self-esteem with other variables,
especially group size. As one might expect, as groups became
larger, fewer people were willing to speak out against the con-
sensus or criticize the group, but this pattern of declining initia-
tive was less pronounced for people high in self-esteem than
for those low in self-esteem. Put another way, people with high
self-esteem were less intimidated by the prospect of speaking
out in a large group.
Thus, high self-esteem was found to have one benefit, and it
was a benefit that fits well with theoretical assumptions. Taking
initiative and criticizing a group would seemingly require hav-
ing some confidence in one’s own views, and sure enough, peo-
ple with high self-esteem were rated by their peers as more
willing and more likely to do those things. It is in fact surpris-
ing that the effect was so small.
Group performance was also investigated in a longitudinal
study by Paulhus (1998). He had 4-person and 5-person groups
meet once per week for 7 weeks to discuss readings for their
psychology class. Group members rated each other after the
first and last weeks. Self-enhancement, defined as rating one-
self more favorably than group members did, was positively
correlated with narcissism at the end of the study (whereas if
anything, narcissists came across as fairly modest at the initial
assessment). In contrast, self-esteem showed a weak, but posi-
tive correlation with self-enhancement at both times. The most
relevant outcome variable, however, was the rating of each in-
dividual’s contribution to group performance. People with high
self-esteem were rated more positively than people with low
self-esteem at both times (r .25 and r .28, respectively),
indicating good contribution to the group’s work. The hetero-
geneity of high self-esteem was evident, however, in the ratings
of narcissists. Initially, narcissists were rated quite positively as
contributing to the group, but by the end of the 7th week, the
favorable impression of narcissists had disappeared completely
(r .01). In a second study, similar results were found, and
in addition, narcissists went from generally positive peer rat-
ings initially to being rated as cold, arrogant, inclined to brag,
prone to overestimate self, and hostile by the end of the study.
The second study also solicited ratings by friends prior to
the series of group meetings, which allowed Paulhus (1998) to
calculate a second, pretest, measure of self-enhancement, that
is, each person’s tendency to rate himself or herself more favor-
ably than friends did. The pattern of ratings for participants
with high scores on this pretest measure of self-enhancement
was quite similar to the pattern found for narcissists—peers’
ratings of their contribution to the group were significantly
positive initially, but this positive impression was gone in the
final ratings. The ratings at the end of the study depicted self-
enhancers as confident, but cold, arrogant, prone to brag, de-
fensive, and hostile.
Leadership is another important aspect of group behavior,
and most experts would say that the causes and determinants of
good leadership form a set of questions that psychology has not
yet managed to answer to everyone’s satisfaction. Intuition
suggests that leadership ought to correlate with self-esteem, al-
though the causal arrow could point in either direction. High
self-esteem might give a person the confidence and initiative to
take charge of a group, make decisions under pressure or un-
certainty, and expect that others will do his or her bidding. But
it is also plausible that becoming a leader engenders a sense of
personal superiority and brings about a rise in self-esteem. Cer-
tainly, the careers of many exceptional leaders in history sug-
gest that they must have had some high degree of self-
confidence in order to make it to the top and that, conversely,
PSYCHOLOGICAL SCIENCE IN THE PUBLIC INTEREST
Benefits of Self-Esteem
20 VOL. 4, NO. 1, MAY 2003
over time their positions of power encouraged them to become
narcissistic and sometimes megalomaniacal (Baumeister, 1989).
An impressive study of leadership among military cadets
was recently published by Chemers, Watson, and May (2000).
The cadets were followed over time, including during a special
summer program devoted to leadership training. The measures
included ratings by peers and by military science professors, as
well as objective performance on two leadership tasks and two
nonleadership tasks (rifle marksmanship and land navigation).
Simple correlational analyses showed some significant effects
of self-esteem. It predicted peer ratings, professors’ ratings,
and performance on the second (though not the first) leadership
exercise, as well as performance on the two control nonleader-
ship tasks.
Self-esteem, optimism, and leadership efficacy (feelings of
being a good leader) proved to be intercorrelated, however, so
the authors conducted more rigorous statistical analyses that
disentangled the effects of these variables. In these analyses,
global self-esteem did not have a significant correlation with
leadership ratings obtained either from the professors or from
the cadet peers. Self-esteem also failed to yield a significant
contribution to performance on the leadership exercises (as did
optimism), but leadership efficacy succeeded. The fact that the
effects of self-esteem disappeared in these more sophisticated
analyses indicates that self-esteem overlapped with other fac-
tors (e.g., leadership efficacy) that may have taken precedence
in predicting the outcome measures assessed. Chemers et al.
found a unique association between self-esteem and perfor-
mance only on the land navigation exercise. High self-esteem
might help people avoid getting lost (or, conversely, finding
one’s way home might raise self-esteem).
Group behavior also involves intergroup relations. In gen-
eral, people favor their own groups in thought and deed. They
prefer to assign greater rewards to members of their own group
than to out-groups, they rate their own group more favorably
than out-groups, and they tend to direct prejudice and discrimi-
nation toward out-groups. Early theorizing suggested that such
in-group favoritism is stronger among people with low self-
esteem than among those with high self-esteem (e.g., Ehrlich,
1973). It seemed plausible that people low in self-esteem are
most inclined to boost themselves by means of disparaging and
discriminating against members of other groups. Indeed, the
opportunity to discriminate against a member of a negatively
stereotyped group can bolster self-esteem (Fein & Spencer, 1997).
Crocker et al. (1987) pointed out, however, that one cannot
infer the presence of prejudice when people with low self-
esteem give low ratings to out-groups. People with low self-
esteem seem generally negative about many events and people,
including themselves. A more appropriate measure of bias is
the difference between how one rates oneself or one’s own group
versus how one rates the out-group. By that measure, Crocker
et al. found that prejudice was stronger among people with
high rather than low self-esteem (see also Crocker & Luhtanen,
1990).
In a meta-analysis, Aberson et al. (2000) found that people
high in self-esteem rate groups to which they belong more fa-
vorably than they rate groups to which they do not belong. Peo-
ple low in self-esteem show no such difference. Recent work
on social categorization suggests a simple process can account
for this effect. When people predict the attributes of their in-
groups, they rely heavily on their own self-concepts, but when
they predict the attributes of out-groups, they do not engage in
such self-anchoring (or projection) (Cadinu & Rothbart, 1996;
Clement & Krueger, 2002). Because positive attributes out-
number negative ones in most people’s self-concepts, projec-
tions to in-groups yield positive outcomes. By not being subject
to projection, descriptions of out-groups remain more neutral.
The process of differential projection also explains why in-
group bias typically reflects in-group favoritism rather than
out-group derogation (Brewer & Kramer, 1985).
Conclusion
Popular wisdom in the age of self-esteem holds that loving
oneself is a prerequisite for loving others (e.g., Crooks & Baur,
1999). Efforts to boost self-esteem in schools, homes, and else-
where would be well justified if they resulted in significant im-
provements in how people got along with one another.
The evidence suggests that the superior social skills and in-
terpersonal successes of people with high self-esteem exist
mainly in their own minds. People with high self-esteem claim
to be more popular and socially skilled than others, but objec-
tive measures generally fail to confirm this and in some cases
point in the opposite direction. When rated by peers, teachers,
or laboratory interaction partners, people with high self-esteem
are not liked any better than people with low self-esteem, and
following a blow to their pride they may become antagonistic
and elicit negative reactions. People who have elevated or in-
flated views of themselves tend to alienate others.
The sociometer model proposed by Leary and his colleagues
(Leary et al., 1995) suggests that the level of self-esteem should
be an outcome of interpersonal successes and failures. This view
raises a very plausible alternative explanation for any findings
that might link high self-esteem to social success. But as al-
ready noted, there are relatively few such findings. Bishop and
Inderbitzen (1995) found that children without any friends
tended to have low self-esteem, and it seems quite plausible
that this lack of friends was a cause rather than a result of low
self-esteem.
The results of studies on group behavior do not provide a
ringing endorsement of the importance of self-esteem. High
self-esteem does not appear to have much to contribute to lead-
ership, especially after controlling for other variables. High
self-esteem appears to foster a small but significant tendency
toward greater voice behavior in work groups. This tendency to
speak up may also be responsible for people with high self-
esteem receiving somewhat higher ratings for contributing to
the group, compared with people with low self-esteem. People
PSYCHOLOGICAL SCIENCE IN THE PUBLIC INTEREST
R.F. Baumeister et al.
VOL. 4, NO. 1, MAY 2003 21
with high self-esteem also generally show greater bias in favor
of their own groups, which may be beneficial for the in-group
but may also produce greater discrimination and conflict be-
tween groups.
Self-esteem does appear to have a variety of interpersonal
consequences. People with low self-esteem differ from those
with high self-esteem in how they conduct their interpersonal
relationships and how they choose and pursue social goals.
Still, these differences do not appear to translate into substan-
tially different relationship outcomes.
The most promising results involve social initiative. The
tendency to initiate interpersonal contacts and relationships (as
rated by roommates) was the only interpersonal skill that dif-
ferentiated participants with high and low-esteem in the study
by Buhrmester et al. (1988). In groups, people with high self-
esteem are more willing than those with low self-esteem to
speak up to criticize the group or propose directions for action.
Likewise, people high in self-esteem take more initiative in ex-
tricating themselves from unhappy relationships (Rusbult et al.,
1987). At present the evidence is sparse, but both theoretical
considerations and the few available findings suggest that self-
esteem may have value in promoting initiative.
Self-esteem per se may not be the best variable for explor-
ing interpersonal processes. Narcissists (many of whom have
high self-esteem) exhibit a variety of interpersonally problem-
atic and alienating behaviors. Some depressed people (a group
that includes some people with low self-esteem) constantly
seek interpersonal reassurance and alienate others for different
reasons.
AGGRESSION, VIOLENCE, DELINQUENCY, AND
ANTISOCIAL BEHAVIOR
For decades, psychologists believed that low self-esteem
was an important cause of aggression. This view appears to
have emerged from clinical impressions rather than any single
theoretical formulation or line of empirical evidence. It was
challenged by Baumeister et al. (1996), who reviewed assorted
bodies of findings and concluded that perpetrators of aggres-
sion generally hold favorable and perhaps even inflated views
of themselves. They proposed that aggression results from
what they called “threatened egotism,” namely, a positive view
of self that is questioned or attacked by other people. Two
points need to be noted. First, that review did not contain labo-
ratory experiments linking measured self-esteem to measured
aggression, and indeed many of the findings about violent
groups (such as incarcerated murderers) relied on interviews
and impressions rather than standard measures of self-esteem.
Second, Baumeister et al. firmly embraced the notion of the
heterogeneity of high self-esteem. They noted that many peo-
ple high in self-esteem are not aggressive, even though others
are highly aggressive. If that view is correct, then carefully
controlled and systematic studies attempting to measure both
self-esteem and aggression might generally find null results,
insofar as the aggressive and nonaggressive people in the high
self-esteem category would cancel each other out.
Aggression
Among children, bullying is an important form of aggres-
sion. The traditional view that bullies have low self-esteem has
recently come under attack. Olweus (1990, 1994) was one of
the first to dispute the notion that bullies suffer from insecuri-
ties and self-doubts that underlie a tough exterior. Although Ol-
weus did not measure self-esteem directly, he did show that
bullies reported less anxiety and were more sure of themselves
than other children. Analyzing self-reports of bullying behavior
from more than a thousand Australian schoolchildren, Slee and
Rigby (1993) found no correlation between bullying and self-
esteem. Clearly, though, self-reports of bullying are suspect,
insofar as people who score high on self-esteem generally
claim to perform a broad range of socially desirable behaviors
more often than other people do (and indeed they claimed to
perform more prosocial behavior in this sample, too).
Reliable information about bullying requires confirmation
from external sources, preferably peers. Such data were ob-
tained by Salmivalli, Kaukiainen, Kaistaniemi, and Lagerspetz
(1999), who had a sample of more than 300 Finnish adoles-
cents (ages 14 and 15) fill out trait measures, including mea-
sures of self-esteem and defensive egotism, and rate all their
classmates on a variety of behavioral and trait measures. In ad-
dition to calculating correlations between self-esteem ratings
and other measures, Salmivalli et al. used the measures of self-
rated and peer-rated self-esteem to categorize the participants
according to whether they were best characterized as having
defensive high self-esteem, genuine high self-esteem, low self-
esteem, a tendency to belittle themselves, or humble pride. By
using these categories, the authors were able to allow for the
possible heterogeneity of high self-esteem.
The correlation between each person’s self-esteem and the
classmates’ ratings of that person as prone to bully people was
not significant. However, people categorized as having defen-
sive high self-esteem (characterized by very high scores on de-
fensive egotism along with above-average scores on self-rated
and peer-rated self-esteem) were significantly more likely than
others to be named by peers as bullies. Salmivalli et al. (1999)
noted that this finding confirmed both the heterogeneity of high
self-esteem and the threatened-egotism hypothesis put forward
by Baumeister et al. (1996). Of further interest was that the cat-
egory of defensive high self-esteem covered not only the bul-
lies themselves but their cronies, too. People in this category
were rated by peers as prone to assist bullies in their victimiza-
tion of others, such as by joining in once a bully had begun at-
tacking someone, or by catching and holding the victim so the
bully could continue tormenting him or her. People with defen-
sive high self-esteem were also rated as more likely than others
to reinforce bullying, such as by coming over to watch the bul-
lying, shouting out encouragement to the bully, laughing at the
PSYCHOLOGICAL SCIENCE IN THE PUBLIC INTEREST
Benefits of Self-Esteem
22 VOL. 4, NO. 1, MAY 2003
spectacle, and making approving comments either to the bully
or to other bystanders.
The heterogeneity of high self-esteem was confirmed by
Salmivalli et al. (1999) in another way. Adolescents with genu-
ine (as opposed to defensive) high self-esteem were rated by
their peers as significantly more likely than others to defend the
victims of bullying. Upon reflection, this finding is not surpris-
ing, because coming to the aid of a victim requires some degree
of confidence. One might combine all these findings to suggest
that high self-esteem appears to support initiative, whether for
good or ill: People with high (defensive) self-esteem were
more likely than others to engage in bullying, and people with
high (so-called genuine) self-esteem were more likely to fight
against bullying. Meanwhile, people with low self-esteem were
more likely than others to be the victims of bullies.
The work by Salmivalli et al. (1999) is the most rigorous
and methodologically sophisticated study of bullying that we
have found, because it incorporated objective (peer-rated) mea-
sures of bullying behavior and multiple measures of self-
esteem that allowed for heterogeneity in this construct. The main
drawback of this study is that it was cross-sectional and corre-
lational, so that causality cannot be determined from it. Does
bullying other people cause (defensive) self-esteem to rise, or
does defensive high self-esteem lead to bullying? The link be-
tween low self-esteem and victimization can also quite plausi-
bly be construed in either causal direction: Bullies may choose
victims on the basis of perceived lack of self-esteem, or being
victimized may be detrimental to self-esteem.
Probably the best evidence about the causal links between
aggression and self-esteem would come from programs specifi-
cally designed to alter self-esteem. A pair of studies of this type
was recently reported by Murphy, Stosny, and Morrel (2001),
who tracked domestic offenders through therapy programs that
included interventions aimed at boosting self-esteem. Their re-
search has several drawbacks that must be kept in mind. One
drawback is that the sample consisted of men who were referred
by courts to therapy on the basis of spouse abuse. It is generally
believed that men in therapy for domestic violence differ from
other domestic offenders, not least because they know well that
the situation demands that they express remorse and criticize
themselves. Another drawback of these studies is that the ther-
apy could hardly focus on self-esteem alone, but also included
interventions directly aimed at reducing violent behavior, and
so one must be cautious in attributing any reduction in violent
behavior directly to increases in self-esteem.
Nonetheless, the two studies are noteworthy because of their
focus on real-world adult violence, and because in both cases,
therapy led to significant increases in self-esteem. These in-
creases were accompanied by significant self-reported reduc-
tions in spouse abuse. Both before and after treatment, self-
esteem correlated negatively with aggression, such that higher
self-esteem was associated with less self-reported aggression,
although the correlations reached significance in only one of
the two studies.
Self-reported violence is a problematic measure, however.
Murphy et al. (2001) addressed this problem by obtaining re-
ports from the spouses and partners of the domestic offenders
during the 6 months after the therapy. These data alone meet
the standard of objectively confirming behavioral outcomes,
and they provide no confirmation at all of the benefits of boost-
ing self-esteem. There was no significant correlation between
men’s self-esteem or change in self-esteem and their partner-
reported abusive behavior, and if anything the trend was in the
wrong direction: Higher post-therapy self-esteem and greater
increases in self-esteem from the beginning to the end of ther-
apy were both correlated with more rather than less subsequent
spouse abuse (rs .08, n.s.).
Another fairly rigorous approach to examining the impact of
self-esteem on aggression is to conduct laboratory experi-
ments. Assuming causes precede effects, self-esteem measured
at the start of a laboratory session could not be the result of ag-
gressive behavior that is observed later in the session, although
in principle a chronic tendency toward aggression might be the
underlying cause of both the aggression during the experiment
and the self-esteem level. Baumeister et al. (1996) were unable
to find any reports of laboratory experiments linking self-esteem
to aggression. Given the wide popularity of both self-esteem
and aggression as research topics, it did not seem plausible that
the possible link between them had never been tested. This
pointed tentatively toward the conclusion that studies had been
conducted but remained unpublished because of null results.
A pair of laboratory experiments by Bushman and Baumeis-
ter (1998) did provide a direct test of the role of self-esteem in
aggressive behavior. The experiments investigated aggression
in direct retaliation for an insult, displaced aggression (i.e., par-
ticipants were insulted by one person and then had a chance to
aggress against someone else), and unprovoked aggression
(i.e., participants were praised rather than insulted and then had
a chance to aggress against the praiser). The two studies used
different measures of self-esteem. Aggression was measured
by the loudness and duration of blasts of aversive, stressful
noise participants delivered to an opponent when the opponent
lost a trial in a competitive game.
There were no significant correlations between levels of
self-esteem and aggression in any of the three situations stud-
ied. This confirmed the suspicion that the absence of previ-
ously published findings probably indicated that studies had
been done but obtained null results.
However, Bushman and Baumeister (1998) did find that nar-
cissism was a significant predictor of aggression. Specifically,
narcissists were more aggressive than other individuals when
provoked by an insult, but did not show elevated rates of dis-
placed or unprovoked aggression. More recent studies by Stucke
and Sporer (2002) demonstrated high aggression among nar-
cissists who had been criticized and further suggested that the
combination of an unclear or confused self-concept with nar-
cissism is especially likely to lead to aggressive retaliation. The
impact of narcissism in those studies provided further support
PSYCHOLOGICAL SCIENCE IN THE PUBLIC INTEREST
R.F. Baumeister et al.
VOL. 4, NO. 1, MAY 2003 23
for the heterogeneity of high self-esteem, insofar as narcissists
may be considered a category of high self-esteem people—
people who not only think well of themselves, but also view
themselves as being superior to others.
Delinquency
Juvenile delinquency is of both theoretical and practical in-
terest. It combines outright aggression, sometimes extending to
assault and murder, with antisocial behavior such as stealing
and vandalism. The assumption that low self-esteem causes ag-
gression has traditionally offered one explanation for why
young people turn to delinquency, namely, that they lack self-
esteem. However, evidence for the role of self-esteem in delin-
quency is quite contradictory. In this section, we summarize
some of the more important findings.
Neumark-Sztainer, Story, French, and Resnick (1997) ob-
tained self-reports of delinquent behavior by having adoles-
cents report how frequently during the past 12 months they had
damaged or destroyed property, struck or beaten another per-
son, or stolen from a store. Self-esteem showed a significant
negative correlation with self-reported delinquent behavior
among both boys (r .22) and girls (r .26), but in a re-
gression analysis that controlled for family cohesion, school in-
volvement, prior victimization, risk-taking disposition, and
emotional well-being, self-esteem failed to emerge as a signifi-
cant independent predictor, despite the exceptionally large
sample (N 123,132). The lack of independent prediction
does not necessarily undermine the value of self-esteem: Self-
esteem may have important effects on delinquency that happen
to be mediated by other variables (e.g., perhaps low self-esteem
people are more likely to engage in delinquency because they
are more likely to disengage from school and take risks).
A more recent study by Trzesniewski, Donnellan, Robins,
Moffitt, and Caspi (2002, Study 1) with 292 fifth- and eighth-
grade students yielded significant correlations (rs ranged from
.26 to .35) between self-reported delinquency and three
different measures of self-esteem (two self-report measures
and a teacher report). In this case, the relation remained signif-
icant (standardized coefficient .28) in a complex statistical
analysis (latent structural equation model) that controlled for
both supportive parenting and academic achievement.
One of the most ambitious and intriguing studies of self-
esteem and delinquency was published by Rosenberg et al.
(1989). Their nationwide sample included nearly 2,000 adoles-
cent males who were measured twice. The study’s longitudinal
design permits some causal inference, although its exclusive
reliance on self-reports of delinquent behavior means that the
results may be compromised by self-report biases. The simple
correlations between self-esteem and self-reported delinquency
were quite weak in this study, rs .09 and .07 at the first
and second assessments, respectively. Rosenberg et al. pro-
posed, however, that this seemingly null result might actually
conceal contrary trends. By examining changes across time,
they concluded that there were actually two significant links
between self-esteem and delinquency in the data. First, low
self-esteem was associated with an increase in delinquent be-
havior (r .19). Second, delinquent behavior was associated
with an increase in self-esteem (r .08, marginally signifi-
cant). They also found that these effects interacted with social
class. Low self-esteem was most prone to increase delinquency
in the more affluent group, possibly because delinquent behav-
ior was more normative among the lower classes, so the num-
bers were already so high that there was not much room for
increase. Meanwhile, the boost in self-esteem deriving from
delinquent participation was greater among the young men of
low rather than high (or medium) socioeconomic status.
Another recent longitudinal study, however, failed to confirm
the conclusions of Rosenberg et al. (1989). In this study, Jang
and Thornberry (1998) followed nearly a thousand students
(starting in seventh and eighth grade) for 4 1/2 years. They dis-
tinguished between delinquent activities and associating with de-
linquent friends, although self-reports were the main source of
information about both. Jang and Thornberry found no tendency
for low self-esteem to predict subsequent delinquency, whether
measured by criminal and antisocial behavior or having delin-
quent friends. They also failed to find that delinquent activity
boosted self-esteem, and, if anything, self-reported delinquent
activity led to a decrease in self-esteem. Forming friendships
with delinquent peers did, however, boost self-esteem. Thus, any
apparent benefits from delinquency appeared to be due to form-
ing friendships and engaging in social comparison with delin-
quent peers, as opposed to committing violent or antisocial acts.
Most recently, an ambitious longitudinal study by Trzes-
niewski et al. (2002, Study 2) yielded fairly strong evidence
that self-esteem may indeed play a causal role in externalizing
behavior, a constellation of behaviors that includes delinquency
plus additional antisocial misbehavior not generally considered
delinquent. (The term externalizing behavior reflects the as-
sumption that the person is dumping his or her problem onto
others.) This study followed students from age 11 (n 809;
48% female) to age 13 (n 726; 48% female) and assessed ex-
ternalizing behavior with three relatively objective measures:
teacher report, parent report, and an interview report (at age 13
only) based on the American Psychiatric Association’s symp-
tom criteria for Conduct Disorder. Self-esteem was significantly
correlated with externalizing behavior at both time points (rs
ranged from .16 to .25), and self-esteem at age 11 pre-
dicted externalizing behavior at age 13 (rs ranged from .19 to
.21). Moreover, the contribution of self-esteem was found to
be independent of relationships to parents, relationships with
peers, IQ, and socioeconomic status. Given the large sample,
longitudinal design, and multimethod measurement of external-
izing behavior, this study provides the best available evidence
for a positive link between low self-esteem and subsequent delin-
quent behavior.
Overall, there is some support for the traditional view that
low self-esteem may predispose a person to participate in anti-
PSYCHOLOGICAL SCIENCE IN THE PUBLIC INTEREST
Benefits of Self-Esteem
24 VOL. 4, NO. 1, MAY 2003
social behavior. The correlation between self-esteem and delin-
quency varies among studies (from nearly zero to around
.30), but it is almost always negative, suggesting that the ef-
fect is present, although probably quite weak. There is also
some suggestion that the effect of self-esteem on delinquency
may be mediated by variables such as school involvement and
choosing to associate with delinquent friends.
Antisocial Behavior
Antisocial behavior is a companion to delinquency. By anti-
social behavior, we mean behavior that is socially undesirable
but not necessarily violent. An experiment by Lobel and Le-
vanon (1988) examined cheating in relation to individual dif-
ferences in self-esteem and need for approval. The participants
were 12-year-old boys and girls in school in Israel. Their task
was to trace line drawings without lifting the pencil or retrac-
ing lines. After two easy problems, three unsolvable problems
were presented, and the children could claim to have solved
them only if they cheated, which thereby allowed the research-
ers to obtain an objective measure of cheating. Self-esteem did
not have a direct effect on cheating, but interacted with need
for approval to determine the amount of cheating. The results
were consistent with the view that self-esteem is a heteroge-
neous construct, in that the children who cheated the most and
who cheated the least both had high self-esteem. More pre-
cisely, those who were high in both self-esteem and need for
approval cheated the most, whereas those who were high in
self-esteem but had a low need for approval cheated least. The
authors proposed that a distinction between true and defensive
high self-esteem could account for their results.
School disciplinary problems and related antisocial behav-
iors were the focus of a large survey by DeWit et al. (2000). All
students in four secondary schools in Ontario, Canada, were in-
cluded in the study, for a total of a thousand participants. Al-
though the authors clearly favored the self-esteem hypothesis,
self-esteem was not a significant predictor for three of the five
outcome variables. Most notably, self-esteem did not contrib-
ute to the prediction of disciplinary problems in school, sub-
stance abuse, or conduct problems.
DeWit et al. (2000) relied exclusively on self-reports of anti-
social behavior. As we noted earlier, all self-reports are not
equal, with reports of concrete and verifiable misbehavior pre-
sumably being less subject to distortion and bias than global es-
timates. The reports of school disciplinary problems seem the
most concrete and verifiable measure DeWit et al. used, insofar
as students were asked how frequently they had been sus-
pended, served detentions, been sent to the principal, or been
truant in the past 6 weeks. The measures of substance abuse
likewise seem reasonably concrete, although less readily verifi-
able, and on these measures self-esteem yielded no significant
effects. In contrast, oppositional-defiant disorder (a recurring
pattern of negative, hostile, disobedient, and defiant behavior)
was related to low self-esteem. Indications of this disorder were
measured largely by an attitude survey consisting of questions
such as, “How often in general do you get annoyed by others?”
The fact that self-esteem had its effects on the less concrete
measures suggests some potential role for self-report bias.
Last, a set of interview and laboratory studies by Colvin,
Block, and Funder (1995) looked specifically at people with in-
flated self-esteem. This was done by comparing how favorably
people rated themselves with how favorably they were rated by
others, and the people whose self-ratings were higher than the
peer ratings were classified as having inflated favorable views
of self. Participants were then invited into the laboratory, and
their interactions with randomly assigned peers were observed.
As compared with controls, people who had inflated self-es-
teem exhibited a high rate of antisocial and socially undesir-
able behaviors. They were more prone to interrupt others
during the conversation. They expressed more hostility and
were found to annoy and irritate their partners. Their conversa-
tional style was seen as talking at rather than talking with
someone. Despite their favorable self-ratings, they came across
as socially awkward. In a second study, people of this sort were
found to be prone to psychological maladjustment and to have
poor social skills.
Conclusion
Many researchers have sought to link self-esteem to vio-
lence, aggression, and antisocial tendencies. The results are
mixed at best. Psychologists who wish to study or reduce ag-
gression might be well advised to focus on factors other than
self-esteem or, at least, to respect the heterogeneity of high
self-esteem and therefore consider additional variables. In the
case of aggression, simple measures of self-esteem have gener-
ally failed to predict objective behavior, whereas high narcis-
sism (presumably one subcategory of high self-esteem) does
lead to aggressive retaliation, a finding consistent with the
threatened-egotism model. Inflated self-esteem, in the sense of
regarding oneself more favorably than one’s peers do, is also
associated with socially undesirable behaviors. Although most
large studies have found self-esteem to be weak at predicting
delinquency, school disciplinary problems, and related antiso-
cial activities, and in fact some findings suggest that people
with some varieties of high self-esteem are most prone to act
antisocially, one methodologically strong recent study has sug-
gested that low self-esteem may predispose individuals toward
externalizing behavior.
Although further research is needed, one impression that
emerges from these data is that self-esteem simply intensifies
both prosocial and antisocial tendencies. As the study by Salm-
ivalli et al. (1999) showed, people who think well of them-
selves are overrepresented among both the perpetrators of
bullying and the people who stand up to bullies and defend vic-
tims. By the same token, Lobel and Levanon (1988) found that
people high in self-esteem made up both the highest-cheating
and the lowest-cheating groups. Quite possibly, the actual ef-
PSYCHOLOGICAL SCIENCE IN THE PUBLIC INTEREST
R.F. Baumeister et al.
VOL. 4, NO. 1, MAY 2003 25
fect of high self-esteem per se is to support initiative and confi-
dent action, for good or ill.
HAPPINESS, COPING, DEPRESSION
We turn now to subjective outcomes, such as happiness. The
pursuit of happiness is one of the overarching goals in life for
most people, and feeling happy is one form of positive out-
come whose desirability is difficult to dispute. Along with hap-
piness, we consider two other related phenomena. One of these
is depression, which is often defined (at least in part) by perva-
sive unhappiness. The other is coping with stress and trauma.
Even if self-esteem were to fail to contribute to a better, hap-
pier life overall, it might accomplish a great deal by helping
people to recover from aversive events. Misfortunes, problems,
and setbacks can be experienced as blows to one’s favorable
view of self (e.g., Janoff-Bulman, 1992; Taylor, 1983), and it is
plausible that high self-esteem enables people to recover faster
or more thoroughly from such events.
Before we present findings, we must acknowledge that studies
of happiness and related variables seem almost inevitably to rely
on self-reports. For other outcomes, we were routinely suspicious
and critical of self-report data, and whenever possible we sought
objective measures. Yet it is not clear what sort of objective mea-
sure could replace self-reports of happiness. It is possible to prove
that people are less (or more) intelligent, attractive, or likable than
they think they are, but in what sense could researchers conclude
that people are less (or more) happy than they think they are? If
someone estimates his or her happiness to be at 60 on a scale run-
ning from 1 (very low happiness) to 100 (very high happiness),
there is no objective criterion, or “happiness meter,” that could
show that the person’s true happiness is higher or lower. There is
not even any meaningful or realistic way to assert that someone
can be mistaken about his or her happiness. As a result, studies in-
vestigating any possible link between self-esteem and happiness
will necessarily have to rely on self-reports for both measures, de-
spite the pitfalls and drawbacks of self-reports in connection with
establishing the effects of self-esteem.
Reliance on self-report carries the danger of distorting the
role of self-esteem. As we reported in preceding sections, peo-
ple with high self-esteem repeatedly emerge as better than oth-
ers in self-report data, whereas objective data often depict such
people to be no better than others. A response bias toward pre-
senting the self in favorable, socially desirable ways appears to
characterize people high in self-esteem. Hence, one might well
expect that people with high self-esteem claim to be happier
than other people. However, there does not seem to be any way
around this problem, given the lack of objective measures (in-
deed the ineluctable subjectivity) of happiness.
Happiness
A major international study of self-esteem and happiness
was reported by Diener and Diener (1995). The data came from
more than 13,000 college students from 49 different universi-
ties, 31 countries, and five continents. High self-esteem emerged
as the strongest of several predictors of life satisfaction overall.
The simple correlation between self-esteem and happiness was
quite significant at .47. In short, self-esteem and happiness are
substantially interrelated.
The correlation between self-esteem and happiness (life sat-
isfaction) varied somewhat across countries. In particular, it
was stronger in individualistic countries than in collectivistic
ones. Diener and Diener (1995) proposed that in individualistic
countries, people are socialized to attend to their own internal
attributes, and so these become important in determining over-
all happiness, whereas collectivistic cultures encourage people
to focus on groups and relationships.
In a sample of 406 young people (aged 14–28), Furnham
and Cheng (2000) measured a number of potential correlates of
happiness, including recalled parental rearing styles, scores on
the Eysenck Personality Questionnaire, and self-esteem. They
reported that self-esteem was the most dominant and powerful
predictor of happiness. Shackelford (2001) examined the hap-
piness of young to middle-aged couples (ages ranged from 17
to 41) who had been married within the past year and found
that for both husbands and wives, self-esteem was significantly
correlated with happiness in the form of global, sexual, and
emotional satisfaction.
Data on older adults were provided recently by Lyubomir-
sky and Lepper (2002), and these nicely complement the em-
phasis on college students in Diener and Diener’s (1995) study.
Lyubomirsky and Lepper obtained data from more than 600
adults ranging in age from 51 to 95. Once again, happiness and
self-esteem were highly correlated (r .58). It is important to
note that in both this study and Diener and Diener’s work, self-
esteem and happiness had very different patterns of correla-
tions with other predictor variables, which supports the conclu-
sion that happiness and self-esteem are in fact distinct
constructs that can be measured separately despite their rela-
tively high intercorrelation. In Lyubomirsky and Lepper’s
study, self-esteem was more strongly correlated than happiness
with hopelessness, optimism, and sense of mastery, whereas
happiness was more strongly correlated than self-esteem with
energy level, overall health, loneliness, mood and emotion, and
purpose in life.
A meta-analysis of the relation between 137 personality
traits and happiness (subjective well-being) was published by
DeNeve and Cooper (1998). They found that “private collec-
tive self-esteem” (a term they did not explain) was one of the
strongest predictors of happiness (r .31). Other forms of col-
lective self-esteem yielded weaker results in their analysis.
They do not appear to have included the more typical measure
of self-esteem, nor did they define their terms, but in any case
their results provide further evidence that self-esteem (of some
sort) is consistently correlated with happiness.
Taken together, these findings uniformly indicate that self-
esteem and happiness are strongly interrelated. They suggest
PSYCHOLOGICAL SCIENCE IN THE PUBLIC INTEREST
Benefits of Self-Esteem
26 VOL. 4, NO. 1, MAY 2003
that high self-esteem may pay off handsomely for the individ-
ual in terms of subjective happiness. Our main note of caution
is that the primitive data resemble the simple and previously
promising correlations between self-esteem and school achieve-
ment. Further research with longitudinal designs (or other means
of testing causal relationships), controls for third variables, and,
if possible, alternatives to self-report measures would greatly
strengthen the case.
Coping and Depression
Studies of coping look specifically at how people respond to
harmful, threatening, traumatic, or otherwise unpleasant events.
The hypothesis favored by most theorists is that high self-
esteem operates as a kind of resource that enables people to
suffer less or to recover more rapidly from such events (e.g.,
Arndt & Goldenberg, 2002). According to this hypothesis, un-
der low stress, people would fare reasonably well regardless
of self-esteem, but in the wake of stress, people with high self-
esteem would fare better than those with low self-esteem.
Sometimes this hypothesis is labeled the buffer hypothesis, be-
cause it asserts that high self-esteem operates as a buffer
against stress, trauma, and misfortune.
To test the buffer hypothesis, Murrell, Meeks, and Walker
(1991) interviewed a sample of more than a thousand adults at
6-month intervals for several years. Contrary to their predic-
tions, self-esteem failed to operate as a buffer, in that the effect
of self-esteem on depression was essentially unchanged whether
stressful events were included in the statistical analysis or not.
Instead, high self-esteem predicted lower depression overall,
regardless of stress. In other words, low self-esteem predis-
posed individuals toward depression in both good times and
bad, and actual circumstances made little difference.
A similar finding was obtained by Robinson, Garber, and
Hilsman (1995). Their original sample was composed of nearly
400 sixth-grade students, almost 300 of whom were inter-
viewed again in the fall of seventh grade (and more than 200
were surveyed again in the spring of that year). This timing is
of particular interest because many students move from ele-
mentary to junior high schools between sixth and seventh
grade, and the change of school can be a potent source of
stress. Self-esteem had a direct effect on depression but not on
externalizing behaviors. There was no interaction between self-
esteem and stress in predicting either dependent variable. Thus,
as Murrell et al. (1991) found, low self-esteem predicted de-
pression regardless of life stress.
It must be noted that Robinson et al. (1995) did find a three-
way interaction between stress, self-esteem, and attributional
style. That is, depression was increased by a combination of
external stresses and a tendency to blame oneself for failure,
especially among people low in self-esteem. Put another way,
low self-esteem alone did not create a vulnerability to becom-
ing depressed in response to stress, but low self-esteem com-
bined with a pessimistic and self-blaming style of thinking did.
Another pair of studies did find interactions between self-
esteem and life stress predicted participants’ increases in de-
pressive feelings. Whisman and Kwon (1993) tested a sample
of 80 undergraduates on two occasions separated by 3 months.
They found that increases in dysphoria (a term for depression
that is sometimes preferred because it avoids the connotation of
clinical pathology) were predicted by self-esteem, by life
stress, and by the interaction between the two. As in the studies
already discussed, low self-esteem led to greater depression.
Unlike in those studies, high life stress also led to greater de-
pression, and the interaction between stress and self-esteem
was also a significant predictor.
However, the interaction Whisman and Kwon (1993) found
did not conform to the pattern of the buffer hypothesis. The
buffer hypothesis holds that high self-esteem helps people cope
with life stress, so differences between people with high and
low self-esteem should be found mainly under conditions of
high life stress. Instead, Whisman and Kwon found the oppo-
site: The self-esteem differences were most pronounced under
conditions of low life stress. Moreover, Whisman and Kwon
reported that life stress appeared to have the biggest effect on
people with high (rather than low) self-esteem. That is, people
with high self-esteem were happy in good times but unhappy
during stressful times, whereas the degree of life stress appar-
ently made less difference to people low in self-esteem.
A larger study along the same lines was conducted by Ralph
and Mineka (1998), who surveyed a sample of 141 undergrad-
uates more than a week before their midterm examination,
again after the exam but before grades were issued, the day af-
ter the students received their grades, and 3 days later. This
procedure allowed the researchers to track reactions to a partic-
ular, well-defined stressful experience with a positive or a neg-
ative outcome. The distress of people with high self-esteem
tended to decrease after they received their grades, regardless
of what the grades were. In contrast, the responses of low self-
esteem people depended on how their grades compared with
what they had sought and expected. (The researchers had asked
the students to indicate the lowest acceptable grade, and the re-
lationship between that answer and the actual grade was used
to classify each student’s outcome as a personal success or fail-
ure.) Contrary to the buffer hypothesis, the greatest increases in
generalized distress occurred among people who had low self-
esteem (and also high pessimism) and received grades that
were acceptable to them. Thus, it was the reaction of people
with low self-esteem to a relatively good outcome that was
most problematic and revealing.
The studies by Whisman and Kwon (1993) and Ralph and
Mineka (1998) produced findings that contradicted the buffer
hypothesis: Differences between people with high versus low
self-esteem emerged under relatively positive, benign condi-
tions, rather than under stressful conditions. Moreover, low
self-esteem seems to poison the good times. In contrast, the
buffer hypothesis did receive support from DeLongis, Folk-
man, and Lazarus (1988) in a study of 75 married couples as-
PSYCHOLOGICAL SCIENCE IN THE PUBLIC INTEREST
R.F. Baumeister et al.
VOL. 4, NO. 1, MAY 2003 27
sessed 20 times over 6 months. They measured daily hassles,
physical health and symptoms, and mood. Perhaps surpris-
ingly, self-esteem did not moderate the impact of stressful has-
sles on mood, but it did moderate the link between stressful
hassles and physical symptoms. Participants who were low in
self-esteem showed a stronger link than others between the
amount of stress they experienced on a particular day and
whether they had physical illness on the same day and on the
following day. Moreover, the buffering effect of self-esteem re-
mained significant even in analyses that controlled for the par-
ticipants’ emotional support and the size of their social
networks. Thus, people with low self-esteem were apparently
more prone than others to get sick or suffer other physical
problems in connection with stressful daily events.
Metalsky, Joiner, Hardin, and Abramson (1993) also found
support for the buffer theory, although only under certain con-
ditions. They assessed self-esteem and other factors in a sam-
ple of more than a hundred students prior to midterm grade
reports and then took several measures on the day midterm
grades were received and for 5 consecutive days thereafter.
Subjective failure on the midterm was defined as getting a
grade below the level the student had previously stated as ac-
ceptable. Increases in depression at first depended solely on
subjective failure, regardless of other factors. Over time, how-
ever, depressive reactions were found to be a product of a
three-way interaction. Substantial increases in depression were
found only among people who fit three criteria: subjective fail-
ure, negative (self-blaming) attributional style, and low self-
esteem. (These results are quite consistent with the pattern of
findings reported by Robinson et al., 1995, described earlier in
this section.) Thus, low self-esteem, especially in combination
with a self-blaming attributional style, made people more vul-
nerable to feeling depressed in response to a real-life stressor
over a short period of time.
A similar investigation by Roberts and Monroe (1992)
reached a somewhat different conclusion. They, too, tracked a
sample (N 216) of college students, and measured self-esteem
before midterm exams and depression afterward. Academic
stress was assessed by calculating the difference between a stu-
dent’s hoped-for grade and actual grade, and then multiplying
this difference by the subjective importance (to the student) of
academic achievement. There was a small independent effect
of self-esteem on depression, such that students with lower
self-esteem just before the examination reported more depres-
sion after getting their grades, regardless of what their grades
were. This effect is interesting, despite its small size, because it
shows low self-esteem leading to increased depression over a
short period of time. Unfortunately, the effect was not main-
tained in the later days of the study, which is puzzling and casts
doubt on the impact of self-esteem on depression.
Level of self-esteem did not interact with academic stress to
predict depression in Roberts and Monroe’s (1992) study, con-
trary to the buffer theory. However, unstable self-esteem (mea-
sured in terms of within-person variance in self-esteem scores
during the preexamination, baseline period) did. Depression in-
creased most among students who were initially not depressed,
had unstable self-esteem, and subjectively failed on the exami-
nation. It also increased among those who were initially de-
pressed, had unstable self-esteem, and did well on the exam.
Unstable self-esteem also independently predicted an increase
in depression regardless of other factors, although when statis-
tical controls were entered into a more complex analysis, this
effect vanished.
Building on this work and other findings, Roberts and Mon-
roe (1994) proposed a general theoretical account of the role of
self-esteem in depression. They acknowledged that low self-
esteem has often been proposed as a risk factor that creates a
vulnerability to depression, but concluded that in research,
level of self-esteem has failed to emerge as a robust predictor
of the onset of depression. They proposed that vulnerability to
depression accompanies unstable self-esteem (i.e., self-esteem
that is prone to fluctuate across time), as well as self-esteem
based on relatively few and unreliable sources.
The buffer theory was supported in Corning’s (2002) re-
search on reactions to perceived discrimination among women.
As in most other studies, there was a general tendency for peo-
ple with high self-esteem to feel better than those with low self-
esteem regardless of events. However, self-esteem interacted with
perceived discrimination. Women with low self-esteem reacted
to being the target of discrimination by becoming distressed,
whereas the effects of discrimination were much weaker on
women with high self-esteem. Corning examined a variety of
bad outcomes, including anxiety, depression, and physical symp-
toms, rather than focusing exclusively on distress.
Recent work by Roese and Pennington (2002) suggests a
possible reason why people with high self-esteem apparently
cope better with discrimination than people with low self-
esteem do. People with high self-esteem tend to have a gap be-
tween their perception of discrimination against their group
and their perception of discrimination against themselves. That
is, they believe that their group is often the victim of discrimi-
nation, but that they themselves have not been victims.
Laboratory experiments by Greenberg et al. (1992) also
supported the buffering hypothesis, although with regard to
anxiety rather than depression. They sought to boost self-esteem
by first giving people a personality test and then providing bo-
gus feedback that conveyed a very positive evaluation. After-
ward, the participants were exposed to one of several stressors,
including having to watch upsetting images of gruesome death
and receiving bad feedback on an IQ test. The stressors tended
to produce increases in self-reported anxiety and signs of phys-
ical distress, but these effects were attenuated among people
whose self-esteem had been boosted.
The buffer hypothesis also received support from a recent
pair of studies by Bonanno et al. (2002), although they used a
measure of self-enhancement (i.e., whether participants rated
themselves as better than average and more highly than peers
rated them on various dimensions) rather than a standard self-
PSYCHOLOGICAL SCIENCE IN THE PUBLIC INTEREST
Benefits of Self-Esteem
28 VOL. 4, NO. 1, MAY 2003
esteem measure. Bonanno et al. examined reactions to stress in
two very different populations, residents of Bosnia during the
recent civil war and young to middle-aged residents of the San
Francisco Bay area whose spouse had died within 3 years of
marriage. The full buffer hypothesis could not be tested be-
cause these studies did not include low-stress control groups,
but in both studies, self-enhancers reported better coping than
non-self-enhancers. Moreover, the findings for the self-report
data were supported by more objective data: The good adjust-
ment of the self-enhancers was confirmed in mental health ex-
perts’ ratings based on clinical interviews.
How can all these findings on the relation between depres-
sion and self-esteem be integrated? First, nearly every study
did find that people low in self-esteem felt more depressed or
otherwise worse than people high in self-esteem. Second, the
buffer hypothesis (namely, the hypothesis that low self-esteem
creates vulnerability to stress) received its best support in stud-
ies that used outcome measures other than depression, such as
physical illness or anxiety. Third, when low self-esteem did in-
teract with life stress, the findings were about evenly split as to
whether they supported the buffer hypothesis or the opposite
pattern—that is, half the studies confirmed that low self-esteem
increased vulnerability to stress but was irrelevant in good
times, whereas the rest found that low self-esteem poisoned the
good times rather than exacerbating vulnerability to the bad
times. Last, some findings favor a more complicated version of
the buffer hypothesis that incorporates a third variable, namely,
self-blame. In other words, the combination of low self-esteem,
self-blame, and stressful or aversive events causes depression
to increase.
Conclusion
The link between self-esteem and happiness is strong. Peo-
ple with high self-esteem are significantly, substantially hap-
pier than other people. They are also less likely to be depressed,
either in general or specifically in response to stressful, trau-
matic events. Many studies have confirmed this link.
Before it is safe to conclude that self-esteem causes happi-
ness, however, further research must address the methodologi-
cal shortcomings of the work that has been done so far. First,
causation needs to be established, and potential third-variable
causes need to be ruled out. It seems plausible that self-esteem
causes happiness, but no research has shown this. The strong
correlation between self-esteem and happiness is just that—a
correlation. It is possible that occupational, academic, or inter-
personal successes cause both happiness and high self-esteem,
whereas corresponding failures cause both unhappiness and
low self-esteem. It is even possible that happiness, in the sense
of a temperament or disposition to feel good, causes high self-
esteem. In this regard, the research on academic performance is
informative: The correlations between self-esteem and good
grades do not seem to derive from any causal impact of self-
esteem on schoolwork, but rather other background causes lead
to both, and to some extent good school performance leads to
higher self-esteem. Happiness seems to us to be more promis-
ing than academic performance as a potential consequence of
self-esteem, but the alternative causal pathways need to be con-
sidered with equal rigor and care.
Second, happiness and depression have been studied mainly
by means of self-report. As we have seen, the tendency of some
people toward general floccinaucinihilipilification may pro-
duce both their low self-esteem and unfavorable reports about
other aspects of life. Clearly, objective measures are much
more difficult to obtain for happiness and depression than for
academic performance. But the lessons from academic perfor-
mance, interpersonal skill, physical attractiveness, and the like
are sobering with regard to self-reports.
Support for the buffer hypothesis is uneven, but there are
several strong and supportive findings. The correlations between
high self-esteem and happiness and between low self-esteem
and depression may conceal some influence of circumstances.
Yet further research is needed to untangle these processes.
Some studies supported the buffer hypothesis: Low self-esteem
produced its detrimental effects mainly in bad times. Others
found the opposite: Low self-esteem seemed to poison the
good times, whereas in bad times everyone suffered. Still oth-
ers found circumstances to be irrelevant, and low self-esteem
simply predicted worse outcomes in both good times and bad.
The final resolution of this theoretical controversy is not in
sight, and we recommend that researchers continue to conduct
tests and analyses to illuminate this question, all the while be-
ing cognizant of conflicting findings. It seems fair to assume
that the ultimate answer will be more complex than the original
buffer hypothesis. Nonetheless, the findings consistently sug-
gest that low self-esteem leads to poorer outcomes, including
depression and possibly physical illness, under some circum-
stances.
HEALTH, SMOKING, ALCOHOL AND OTHER
DRUGS, SEX, EATING
There are a number of reasons to believe that high levels of
self-esteem may have a beneficial impact on physical health.
For example, as we noted earlier, high self-esteem individuals
perceive that they have more social support, and perceptions of
social support have well-documented relations with numerous
health outcomes, including length of life (see Ornish, 1998, for
a summary).
Another reason to surmise that high self-esteem might have
benefits for health lies in the vast literature demonstrating the
impact of stress (especially chronic stress) on health. Stress is,
of course, a highly subjective response, and depends critically
on both the individual’s interpretation of events and his or her
beliefs about coping resources that are available (Lazarus &
Folkman, 1984). Because high self-esteem people interpret
PSYCHOLOGICAL SCIENCE IN THE PUBLIC INTEREST
R.F. Baumeister et al.
VOL. 4, NO. 1, MAY 2003 29
negative events more benignly, perceive that they have greater
control, and are generally more optimistic about their coping
abilities, it seems reasonable that they would experience less
stress and therefore enjoy better levels of physical health than
people with low self-esteem.
The adverse health consequences of stress are primarily me-
diated by activation of stress hormones (cortisol). Thus, the
best indirect evidence for the supposition that high self-esteem
benefits health comes from studies demonstrating that people
who differ in self-esteem show differential cortisol reactivity in
response to stressors experienced in the laboratory. In general,
these studies have shown that when the stressor involves actual
failure or the strong potential for failure (e.g., for elderly peo-
ple, a challenging test of ability to drive a car), cortisol re-
sponses are higher in participants with low self-esteem than in
those with high self-esteem (Kirschbaum et al., 1995; Pruess-
ner, Hellhammer, & Kirschbaum, 1999; Seeman et al., 1995).
Nonetheless, there are only a few studies that provide any
direct evidence linking self-esteem and objective health out-
comes. In a study of 502 adults who were over 60 years of age
and had cardiovascular disease, high self-esteem predicted bet-
ter physical and psychosocial functioning 12 months later
among female, but not male, participants (Forthofer, Janz,
Dodge, & Clark, 2001). And an earlier study of more than
1,300 Helsinki policemen showed that among those whose
electrocardiograms had signs of coronary heart disease, lower
self-esteem was related to a higher risk of dying from myocar-
dial infarction over the next 10 years (Nirkko, Lauroma, Silt-
anen, Tuominen, & Vanhala, 1982).
Several studies have found that people high in self-esteem
rate their overall health better than people with low self-esteem
(e.g., Glendinning, 1998). James (1997) found that collective
self-esteem (such as racial pride) predicted self-reported health
in a small sample of African American workers but not White
workers. In a large sample of 1,700 Scottish youth, aged 14 to 16,
Glendinning (1998) found that self-esteem was linked to self-
ratings of health, but not to self-reports of health-relevant be-
haviors. Still, these latter findings were not verified by objective
measures, and could be distorted by the usual self-report bias.
Although we regard it as premature to draw any conclusions
about the impact of self-esteem on physical health, this does
seem an area that deserves further study (and any null findings
should be published). In particular, further investigation through
longitudinal designs that control for potential mediators or
moderators of the relation would be desirable. It is conceivable
that the benefits of high self-esteem, including feeling good,
lead directly to better health. Then again, the studies we sur-
veyed all found self-esteem to be correlated with better health
only in some groups and not in others, and they differed as to
which groups benefited. It is also possible that underlying bio-
logical causes, such as temperament or good physical condition,
contribute to both self-esteem and health, in which case inter-
ventions aimed at raising self-esteem would likely have no effect
on health. In any case, more and better research is warranted.
Unlike overall health, many health-relevant behaviors have
been studied in relation to self-esteem. Foremost among these
are several sets of destructive or risky behaviors that young
people engage in. The public interest in self-esteem is moti-
vated in part by the hope that boosting self-esteem will help
young people abstain from abusing alcohol and other drugs,
smoking cigarettes, succumbing to pathological eating pat-
terns, and engaging in risky sexual practices. We summarize
the findings regarding these behaviors in this section.
A major shortcoming of most of these studies is that they
rely primarily on self-reports. To be sure, it would be neither
ethical nor practical to attempt to induce youngsters to take il-
legal drugs or engage in risky sex during a laboratory experi-
ment, and attempting to observe these covert activities in the
world outside the laboratory is also not very practical. Hence,
some degree of reliance on self-report is difficult to avoid. Still,
interpretation of the findings should be tempered with recogni-
tion that people high in self-esteem will tend to report more so-
cially desirable patterns of behavior than people with low self-
esteem. Also, self-reporting of specific behaviors is presum-
ably more reliable than vague, general self-ratings.
Smoking
Smoking is widely recognized as detrimental to health. If
high self-esteem could help prevent smoking, even just among
young people, that would be a valuable contribution to individ-
ual welfare and society at large. It is also possible that smoking
affects self-esteem rather than the reverse, because the self-
esteem of smokers who feel stigmatized may suffer. Several large
investigations have explored possible links between self-esteem
and smoking, although smoking has generally been measured
only by self-report, and findings that people with high self-esteem
smoke less than others might be an artifact of self-report bias.
Three studies have suggested that low self-esteem is linked
to smoking, but unfortunately none of them shows that low
self-esteem clearly leads to subsequent smoking. Pederson,
Koval, McGrady, and Tyas (1998) surveyed more than 1,600
Canadian eighth graders, distinguishing multiple categories of
smokers. The current smokers had lower self-esteem than the
adolescents who had never smoked. The self-esteem of ex-
smokers was in between, and experimental smokers (i.e., those
who tried out smoking occasionally without becoming regular
smokers) had nearly the same level of self-esteem as the eighth
graders who had never smoked.
In a younger sample of nearly 1,500 third-grade and fifth-
grade students in central North Carolina, Jackson, Henriksen,
Dickinson, and Levine (1997) found that children with lower
self-esteem were more likely to admit to having smoked on oc-
casion.
A longitudinal study by Andrews and Duncan (1997)
tracked more than 400 adolescents, whose age at the start of
the study was between 11 and 15, for 3 years. The researchers
PSYCHOLOGICAL SCIENCE IN THE PUBLIC INTEREST
Benefits of Self-Esteem
30 VOL. 4, NO. 1, MAY 2003
were particularly interested in whether self-esteem and other
variables would mediate the relation between academic moti-
vation and cigarette smoking (among other variables). Self-
esteem did not mediate this relation, but it was an independent
predictor of smoking. When self-esteem and smoking were mea-
sured at the same time, the average correlation between them
was .18. Lagged correlations across time yielded no clear
pattern as to whether self-esteem predicted subsequent smok-
ing or smoking predicted subsequent self-esteem.
Two additional studies have found that the link between
self-esteem and smoking depends on gender. Lewis, Harrell,
Bradley, and Deng (2001) surveyed 1,200 10- to 15-year-olds
living in three tobacco-producing counties in North Carolina.
Among girls, smokers had significantly lower self-esteem than
nonsmokers. Among boys, the smoking and nonsmoking
groups did not differ in self-esteem. A sample of more than
8,000 children (which shrank to 6,530 by the 4th year of the
study) in Calgary, Alberta, Canada, was studied by Abernathy,
Massad, and Romano-Dwyer (1995). These researchers found
that for girls, low self-esteem in grade 6 predicted a substan-
tially greater likelihood of smoking by grade 9. Indeed, girls
with low self-esteem were about 3 times more likely to try cig-
arettes than other girls. Among boys, self-esteem had no rela-
tionship to smoking. This study is notable both for its large
sample and for its prospective, longitudinal design.
In contrast, several large studies found no relationship what-
soever. The possible link between self-esteem and smoking
was the specific and primary focus of research by Glendinning
and Inglis (1999). They reported findings from two Scottish
surveys (one national, one local), which included a total of
nearly 3,000 young people aged 13 and 14. Noting weakness
and inconsistency in past findings, they included several mea-
sures of smoking, such as frequency of smoking and having
given up smoking versus never having smoked. They also
sorted their respondents into various social categories, such as
socially isolated, conventional, and peer oriented. Despite
these efforts, the data showed no significant relation between
self-esteem and smoking, even when social categories were
taken into consideration.
A longitudinal study of a thousand young people in New
Zealand was reported by McGee and Williams (2000). The re-
searchers assessed self-esteem at age 12 and then inquired
about an assortment of health-compromising behaviors at age
15. There was a slight trend toward greater smoking among in-
dividuals low in self-esteem, but it failed to reach significance,
and given the large sample, a lack of significance must be taken
as an indication that the relation was negligible.
As part of a prospective, longitudinal study of health pat-
terns in a cohort of a thousand young Scots, West and Sweeting
(1997) reported on the participants when they were 15. There
was a slight tendency for the young people with the highest
self-esteem to report less smoking than others, but this failed to
reach significance despite the large sample. In another large
sample (1,700) of rural Scottish youth, ages 14 and 16,
Glendinning (1998) found that self-esteem had no correlation
with self-reported smoking.
A Canadian sample of more than 1,500 sixth graders was
studied by Koval and Pederson (1999). In view of the generally
weak findings from previous research on self-esteem and
smoking, they measured stress in order to look for possible in-
teractions (e.g., perhaps the combination of stress plus low
self-esteem would lead to the most smoking, a pattern consis-
tent with the predictions of the buffer hypothesis). Self-esteem
did not have any significant relationship to smoking, either by
itself or in interaction with stress. This investigation is notable
because several other variables (including stress, rebellious-
ness, being in control of other aspects of one’s life, and
whether the child’s mother smoked) did yield significant pre-
dictions, and so the failure of self-esteem is not due to any
problem in measuring smoking.
A smaller sample of 183 eighth graders was surveyed by
Dolcini and Adler (1994). They examined not only self-esteem
but also membership in various so-called crowds. Global self-
esteem was not related to cigarette smoking. Similarly, in a
young adult (college student) sample, Sharp and Getz (1996)
found no correlation between self-esteem and smoking.
Our study-by-study review suggests the need for a meta-
analysis of the literature on self-esteem and smoking. There
may be a small positive association moderated by gender.
Given that the largest study (Abernathy et al., 1995) found a
longitudinal effect for females, the possibility that low self-esteem
is a risk factor for smoking in girls remains a hypothesis wor-
thy of further study. Still, there was a preponderance of null
findings, and the few positive findings could conceivably re-
flect nothing more than self-report bias, or could even indicate
that recognizing oneself as a smoker leads to lower self-esteem.
Self-esteem is either irrelevant to the causation of smoking or
at best a weak risk factor in girls.
Alcohol and Other Drugs
Abuse of alcohol and other drugs is one of the most worri-
some behaviors among young people, and there was ample rea-
son to hope that boosting self-esteem would prevent it (e.g.,
California Task Force, 1990). In this view, people turn to drugs
for solace because they feel bad, and low self-esteem is often
essentially a matter of feeling bad about oneself. Boosting self-
esteem would therefore remove the need for the escapist abuse
of these substances.
However, other predictions are plausible as well. High self-
esteem might contribute to drug use, too. Gerrard, Gibbons,
Reis-Bergan, and Russell (2000) summarized evidence that
young people with high self-esteem may become vulnerable to
drug and alcohol abuse, albeit for reasons quite different from
those of their peers with low self-esteem. As we have already
noted, people with high self-esteem are prone to a variety of bi-
ases in how they interpret and understand events, and these bi-
ases conspire to make them feel better about themselves. Gerrard
PSYCHOLOGICAL SCIENCE IN THE PUBLIC INTEREST
R.F. Baumeister et al.
VOL. 4, NO. 1, MAY 2003 31
et al. showed that adolescents with high self-esteem systemati-
cally distort their perceptions of how their parents judge their
drinking so as to justify their actions. That is, for all adoles-
cents, perceived approval by parents is an important correlate
of drinking, but the direction of causation differs as a function
of self-esteem. For teenagers with low self-esteem, perceived
parental approval leads to more drinking. For those with high
self-esteem, however, more drinking leads to perceiving greater
parental approval.
Gerrard et al. (2000) also showed that individuals with high
self-esteem tend to minimize their own vulnerability. They em-
ploy a variety of cognitive strategies to convince themselves
that bad things will not or cannot happen to them, and ignore
disagreeable information. Thus, they distance themselves from
the potentially harmful consequences of risky behavior. All
these patterns may make people with high self-esteem more
prone to drink, take drugs, and engage in other risks (such as
unsafe sex, a point to which we return later in this section). In
short, these cognitive strategies might at least offset the factors,
such as influenceability and need to escape from bad feelings,
that could induce people with low self-esteem to do the same
things, and as a result people with low versus high self-esteem
might not differ in their tendency to abuse drugs.
The data do not consistently show that adolescent self-
esteem causes or even predicts abuse of alcohol or other drugs.
For example, the large-scale study by McGee and Williams
(2000) found no relationship between self-esteem (measured at
age 9 to 13) and drinking or drug use at age 15. A major longi-
tudinal study of heavy drinking in Finland by Poikolainen,
Tuulio-Henriksson, Aalto-Setälä, Marttunen, and Lönnqvist
(2001) tracked 700 young people from late adolescence (age
15–19 at the start of the study) into adulthood 5 years later.
They measured both alcohol intake and heavy drinking, with
the latter defined as having 13 or more drinks on one occasion.
Self-esteem as measured at the start of the study did not predict
general alcohol consumption 5 years later (r .08 for males,
r .01 for females), nor did it predict heavy drinking.
Likewise, a 7-year longitudinal study of 125 children (aged
7 to 13 at the start of the study) by Hill, Shen, Lowers, and
Locke (2000) found no effect for self-esteem. In this study, al-
cohol intake was measured by both self-report and interview-
ers’ assessments. The focus was on whether self-esteem
(among other factors) might serve to buffer the individual
against stresses, and Hill et al. carefully compared families
with a history of alcoholism against other families without that
risk factor. Self-esteem did not emerge as a significant predic-
tor or moderator.
Even when findings do show a link between alcohol and
self-esteem, they are mixed and inconclusive. The large study
by Glendinning (1998; described earlier, in connection with
smoking) found that high self-esteem was associated with
more (rather than less) frequent alcohol consumption among
young men but not women, and self-reported episodes of drunk-
enness were unrelated to self-esteem. Likewise, in a 4-year study
of 700 students, beginning in middle school, Scheier, Botvin,
Griffin, and Diaz (2000) found that students with higher self-
esteem in grade 7 reported more rapid rise in alcohol use over
the subsequent years, although this effect appeared only in
more complex analyses; simple analyses yielded no relation-
ship between self-esteem and alcohol use. These authors were
surprised to find that alcohol use was linked to high rather than
low self-esteem, and they noted that several other patterns in
their data rendered the issue even more obscure. In particular,
across the entire group, the several years of middle school were
marked by declining self-esteem and increased drinking, and
these trends were interrelated such that more rapid decreases in
self-esteem were accompanied by more severe increases in
drinking. Hence, the apparent link between high self-esteem
and escalating alcohol abuse might reflect the fact that the stu-
dents with the highest initial self-esteem and the lowest initial
drinking had the largest scope for change. It is a common sta-
tistical illusion for extreme scorers to tend to become less ex-
treme over time, and it often does not indicate any genuine
difference.
A cross-sectional survey of college students by Sharp and
Getz (1996) found that those who reported having used alcohol
in the past month scored higher on self-esteem than those who
reported no alcohol use in the past month. The authors inter-
preted this to mean that using alcohol was a social strategy for
making a good impression on peers and therefore in turn re-
sulted in higher self-esteem, but given the correlational nature
of their data, the findings could equally well mean that high
self-esteem leads to greater willingness to drink alcohol.
The study of third and fifth graders by Jackson et al. (1997;
described earlier) did find that self-rated self-confidence pre-
dicted lesser self-reported alcohol use. With teachers’ ratings
of self-confidence, no effect was found. Moore and Li (1998)
found a weak correlation between low self-esteem and illicit
drug use, which remained significant after controlling for other
variables, in a cross-sectional survey of people with disabili-
ties.
Alcohol use showed little relation to self-esteem in the study
(cited earlier) by Andrews and Duncan (1997). Self-esteem
did, however, predict marijuana use, and in particular was
found to be a significant mediator between academic motiva-
tion (the study’s main focus) and marijuana use. Essentially,
the results showed that a decrease in academic motivation led
to a reduction in self-esteem (possibly because poor school-
work led to social rejection) and a subsequent increase in mari-
juana use. The relationship between self-esteem and marijuana
use was rather weak, however.
Wills (1994) challenged earlier findings claiming that low
self-esteem leads to drinking or drug abuse. He proposed that
perceived control over one’s life is more relevant and powerful.
Because feeling oneself to be in control of one’s life might
overlap with both self-esteem and drug use, researchers might
find correlations between self-esteem and drug use and mistak-
enly conclude that self-esteem, rather than feelings of control,
PSYCHOLOGICAL SCIENCE IN THE PUBLIC INTEREST
Benefits of Self-Esteem
32 VOL. 4, NO. 1, MAY 2003
was the important factor. Wills surveyed a large sample of
nearly 1,800 students in eighth grade and then a year later. His
measure of perceived control was, in fact, substantially corre-
lated with self-esteem (r .67). He then conducted statistical
analyses to determine which was more powerful at predicting
use of alcohol, tobacco, and marijuana. Because self-esteem
and perceived control overlapped with each other, it was neces-
sary to sort the data statistically in order to establish the unique
contribution of each (i.e., the effect of each one independent of
the other). The unique contribution of perceived control was
more than 6 times as large as the unique contribution of self-es-
teem. Also, the only significant predictor of increase in sub-
stance abuse over time was lower levels of perceived control.
Wills concluded that any findings linking self-esteem to sub-
stance abuse may well be inflated because at bottom they sim-
ply reflect the contribution of perceived control.
Interpretation of the findings regarding drinking and drug
abuse is probably complicated by the fact that very different
phenomena are lumped together. Some people experiment with
drugs recreationally, whereas others become addicted. Some
may approach the experience out of curiosity or thrill seeking,
whereas others may use it to cope with or escape from chronic
unhappiness. Emler (2001) concluded that high self-esteem
leads to greater willingness to take physical risks, which may
account for some of the findings that high self-esteem leads to
more use of alcohol and other drugs.
Sexual Behavior
Sexual behavior is another sphere that seems to have poten-
tial for showing links with self-esteem. Some participation in
sexual behavior, especially at a vulnerable young age, can be
the result of succumbing to influence or even exploitation by
others. High self-esteem might enable people to resist such in-
fluence.
Such simple predictions do not take account of possible
complicating factors, however. For example, a person’s level of
self-esteem may be the result rather than the cause of sexual
activity. Third variables may also play an important role. For
example, being attractive and popular may lead to both high
self-esteem and early or frequent sexual activity. Last, as we
noted earlier, people high in self-esteem tend to discount vari-
ous health risks (Gerrard et al., 2000), and this can certainly ex-
tend to the risks associated with sex. Smith, Gerrard, and
Gibbons (1997) found that women with high self-esteem
tended to dismiss and discount risks of unwanted pregnancy
more than other women, and being reminded of the risks only
increased the tendency for women with high self-esteem to be-
lieve that such misfortunes could not befall them. Similar con-
clusions were reached by Burger and Burns (1988), who
showed that sexually active women with high self-esteem rated
their chances of becoming pregnant as lower than those of
other women (including women who were not sexually ac-
tive!). Burger and Burns concluded that high self-esteem is
characterized by a pervasive illusion that undesirable conse-
quences will not happen to the self.
In the California self-esteem project, Crockenberg and Soby
(1989) looked for evidence whether self-esteem had any rela-
tionship to teenage sexual activity in general and pregnancy in
particular. They found only one longitudinal and three cross-
sectional studies. The longitudinal study found that high self-
esteem led to more sexual activity among males but not fe-
males (Jessor & Jessor, 1975). Another study found a weak
link between low self-esteem and sexual activity, but this was
limited to active Mormons (B.C. Miller, Christensen, & Olson,
1987). A third study found no relationship at all (Cvetkovich &
Grote, 1980), and the last found that high self-esteem was cor-
related with more permissive attitudes toward sex (Herold &
Goodwin, 1979). Crockenberg and Soby did, however, think
there was some evidence to support the notion that high self-
esteem facilitates use of contraceptives and hence reduces teen
pregnancy, although the evidence was mixed. They found only
one longitudinal study that measured self-esteem and investi-
gated its relation to subsequent pregnancy, and this study found
no relationship (Vernon, Green, & Frothingham, 1983). An-
other investigation, however, used a measure of self-derogation
instead of a standard self-esteem measure, and two different
analyses (Kaplan, Smith, & Pokorny, 1979; Robbins, Kaplan,
& Martin, 1985) found that the girls who derogated themselves
more during the seventh grade were more likely to have a sub-
sequent pregnancy.
The longitudinal study of New Zealand adolescents by Mc-
Gee and Williams (2000) found no relationship between self-
esteem at age 12 and self-reports of sexual intercourse by the
age of 15. Paul, Fitzjohn, Herbison, and Dickson (2000) fol-
lowed an equally large (N 1,020) sample of New Zealanders
from age 3 up to age 21. Self-esteem was measured at age 11.
At age 21, the participants were asked whether they had had
sexual intercourse before the age of 15. Simple analyses
showed no relationship between self-esteem and early sexual
behavior. More sophisticated (multivariate) analyses still found
no link between self-esteem and early sexual initiation for
males. For females, there was a significant effect, although in
the opposite direction from what the advocates of self-esteem
might have hoped: Girls with higher self-esteem at age 11 were
more (rather than less) likely to start having sex by the age of
15. This effect was especially remarkable because the other
predictors of early sexual intercourse tended to be socially un-
desirable or disadvantageous, such as being in trouble at
school, planning to leave school early, and smoking cigarettes
at an early age.
Thus, prospective work suggests that self-esteem is irrele-
vant to becoming sexually active at an early age or, if anything,
high self-esteem leads to early sex among girls. In contrast, a
survey of a thousand girls and women by Kalil and Kunz
(1999) found that the unmarried teenage mothers were less
likely than other girls and women to have high self-esteem. In
this case, self-esteem was measured after the person had en-
PSYCHOLOGICAL SCIENCE IN THE PUBLIC INTEREST
R.F. Baumeister et al.
VOL. 4, NO. 1, MAY 2003 33
gaged in sex and borne a child; in view of the prospective find-
ings, Kalil and Kunz’s results suggest that becoming an unwed
teen mother causes a reduction in self-esteem.
However, in a large longitudinal survey of young people,
Oates (1997) found no relationships between fertility in gen-
eral and self-esteem. His sample included more than 10,000
young people, aged 14 to 21 at the start of the project and 23 to
30 at the end of it, and he sought to test both the idea that low
(or high) self-esteem predisposes individuals toward having
children and the converse hypothesis that having children af-
fects self-esteem. No significant relationships were found. In
particular, there was no relationship between self-esteem and
the number of offspring. Oates acknowledged the theory that
some people may have children as a way of boosting their self-
esteem, but he concluded that even if that is their goal, these
anticipated gains do not seem to materialize.
Some work suggests that self-esteem may be relevant to
sexual behavior only in selected populations. Berry, Shilling-
ton, Peak, and Hohman (2000) used data from a national
(United States) longitudinal study of risk factors for adolescent
pregnancy. Self-esteem of 5,000 young women was measured
during early adolescence, and 8 years later the women were
asked whether they had become pregnant before age 19. High
self-esteem reduced the odds of self-reported pregnancy among
Black and Hispanic females, but not among Whites and Ameri-
can Indians. Even among Blacks and Hispanics, the effect was
characterized by the authors as slight.
At-risk youth were also the focus of Langer and Tubman
(1997), who studied a special sample of 120 substance-abusing
adolescents. They were specifically interested in whether alco-
hol and other drug use would predict risky sexual behavior and
whether factors such as self-esteem might moderate the ef-
fect—such as if it was mainly adolescents with low self-esteem
who succumbed to engaging in risky behavior when intoxi-
cated. Self-esteem had no effect.
Beyond these findings, a smattering of confusing results has
emerged from different studies. Martin and Knox (1997) found
that unstable self-esteem, unlike level of self-esteem, was cor-
related with gay and bisexual men’s reports of engaging in un-
safe anal sex.
People with high self-esteem report more frequent use of
condoms, which might indicate a greater propensity to practice
safe sex (Hollar & Snizek, 1996). Then again, they also report
greater frequency of unsafe sex (Hollar & Snizek, 1996). There
is also some evidence that they may be more likely to have
abortions (Plotnick, 1992). These seemingly contradictory find-
ings may be reconciled by proposing that they simply have
more sex in general, a possibility that is supported by other
findings (Strouse & Buerkel-Rothfuss, 1987). Walsh (1991)
found that virginity was linked to low self-esteem among men
but not among women. He also found that self-esteem was pos-
itively correlated with number of sex partners; the correlation
was fairly strong for men (r .25) and weak but significant for
women (r .12). As Walsh noted, the direction of causality re-
mains unclear, and both directions are plausible, although he
favored the view that high self-esteem makes a person more
willing to initiate sexual encounters because the threat of rejec-
tion is less daunting. Walsh also noted that changes in sexual
norms may produce dramatic changes in the link between self-
esteem and sexual activity, depending on whether sexual purity
or sexual experience is more highly valued by the society.
Homosexuality is generally stigmatized in our society, and
one might expect that homosexuals would end up with lower
self-esteem than heterosexuals, although in this case the level
of self-esteem would be the result rather than the cause. Lerner
and Galambos (1998) found no difference in self-esteem be-
tween gay and heterosexual adolescents. In contrast, Stokes,
Damon, and McKirnan (1997) found that bisexual men with
high self-esteem were more likely to shift toward a full homo-
sexual identification than were other bisexual men. This sug-
gests a causal role for self-esteem, presumably indicating that
some bisexual men may be resisting homosexuality because of
social stigma and other factors, and that the greater self-accep-
tance of people high in self-esteem makes them more willing to
accept their homosexuality. When homosexuals are attacked or
victimized for being gay, however, their self-esteem drops af-
terward (Waldo, Hesson-McInnis, & D’Augelli, 1998). Thus,
being victimized leads to lower self-esteem, which is consis-
tent with the findings on bullying we reported earlier. However,
Kendall-Tackett, Williams, and Finkelhor (1993) reviewed
many empirical studies on sexual abuse of children and did not
find evidence that low self-esteem consistently resulted from
such abuse (cf. Emler, 2001). They noted that many theories
about the impact of sexual abuse on children emphasize dam-
age to self-image or self-esteem, but research findings had not
substantiated that such damage occurs.
Some findings suggest that high self-esteem may reduce
sexual inhibitions. Herold and Way (1983) found that women
with high self-esteem were more likely to report that they per-
formed fellatio. Hurlbert and Whittaker (1991) found that
women who masturbated had higher self-esteem than other
women. Herold, Corbesi, and Collins (1994) found that women
with high self-esteem were more likely than other women to go
topless on Australian beaches. Although these findings are cor-
relational, we regard it as a priori unlikely that high self-esteem
is the result of performing fellatio, masturbating, and showing
off one’s breasts in public, and so a causal role for self-esteem
seems more plausible. Conversely, low self-esteem is corre-
lated with erectile dysfunction among men (Herer & Holz-
apfel, 1993), but insofar as most erectile dysfunction is now
considered to have biological causes, self-esteem seems more
likely to be the result than the cause in this case.
All in all, the results do not support the simple view that low
self-esteem predisposes people to more or earlier sexual activ-
ity. If anything, people with high self-esteem are less inhibited,
more willing to disregard risks, and more prone to engage in
sex. Many of these findings could stem from the greater popu-
larity of people with high self-esteem (popularity could cause
PSYCHOLOGICAL SCIENCE IN THE PUBLIC INTEREST
Benefits of Self-Esteem
34 VOL. 4, NO. 1, MAY 2003
both high self-esteem and more sex). They may also reflect
greater initiative among people high in self-esteem. Not sur-
prisingly, however, bad sexual experiences and unwanted preg-
nancies may well lower self-esteem.
Eating Disorders
The literature on eating disorders very consistently docu-
ments that individuals with eating disorders typically have very
low levels of self-esteem. A recent and particularly impressive
demonstration of this association was reported by French et al.
(2001), who examined the correlates of binge-purge and
weight-loss behaviors among a sample of more than 48,000
girls and 47,000 boys in grades 6 through 12 who lived in 213
cities and towns in the United States. For both genders, self-
esteem and similar variables were the strongest predictors of eat-
ing disorders. Similar results have been obtained in many other
studies, including one by Williams et al. (1993), who found that
anorexic and bulimic patients have lower levels of self-esteem
than obese dieters, nonobese dieters, and normal controls.
Anorexics’ thinking is pervaded by a discrepancy between
their own view of themselves and other people’s view of them:
They frequently have many accomplishments, abilities, and in-
terests, but according to anorexia theorists, they are unaware of
their successes and capabilities. Bruch (1962, 1975, 1978)
identified in these patients a paralyzing underlying sense of in-
effectiveness that pervades their thinking and activities. Bers
and Quinlan (1992) showed empirically that this is true, stat-
ing, “This disparity between interests and perceived abilities
supports the common clinical observation that anorexics gener-
ally feel incompetent and unable to perform well, although
they pursue many activities and objectively could claim many
accomplishments” (p. 428).
Perhaps because bulimia is a much more pervasive problem
than anorexia, more research has been done to investigate pos-
sible causes of bulimia. Mintz and Betz (1988) investigated
self-esteem differences in a sample of undergraduate women
who were categorized in one of the following categories: nor-
mals, bulimics, bingers (women who eat large amounts of food
but do not engage in compensatory behaviors), purgers (those
who purge but do not really eat huge amounts of food), chronic
dieters, and subthreshold bulimics. They found that bulimics
had much lower self-esteem than any of the other groups. Thus,
the impact of self-esteem may be tied specifically to the pattern
of bingeing and purging.
Bulik, Wade, and Kendler (2000) studied the relation be-
tween self-esteem and bulimia in monozygotic (“identical”)
twins. Although monozygotic twins have the same genetic pre-
disposition to bulimia, it is sometimes the case that one twin
has the disorder and the other does not. In a sample of such
twin pairs, Bulik et al. found that the twin with bulimic symp-
toms had significantly lower self-esteem than the other. In a
laboratory study of chronic dieters who were given a high-cal-
orie milkshake to drink, those with low (but not high) self-
esteem ate considerably more ice cream later (Polivy, Heather-
ton, & Herman, 1988). However consistent these findings are, a
recent study calls into question whether raising self-esteem
might be useful for treating already-established eating disorder
patterns. In this study (Safer, Lively, Telch, & Agras, 2002),
women who were binge eaters (though not purgers) received
behavior therapy for 20 weeks and then 6 months later were re-
assessed. At the follow-up, the difference in self-esteem scores
between women who relapsed and those who did not was non-
significant, and, indeed, those who relapsed had slightly higher
self-esteem.
Although researchers have found fairly consistent associa-
tions between low self-esteem and eating pathology, there is
some debate as to whether low self-esteem is a cause or conse-
quence of disordered eating. There are some prospective stud-
ies that suggest self-esteem may play a causal role. For example,
Button, Sonuga-Barke, Davies, and Thompson (1996) assessed
self-esteem in 400 schoolgirls aged 11 to 12 and found that
those with lower levels of self-esteem were significantly more
likely to have developed an eating disorder at age 15 to 16.
There is also some evidence that self-esteem may affect prog-
nosis; in a 4-year prospective study of bulimic patients, low
levels of self-esteem at admission were predictive of poorer
outcomes (van-der-Ham, van-Strein, & van-Engeland, 1998).
Heatherton and Polivy’s (1992) spiral model of binge eat-
ing, however, posits a bidirectional relationship between diet-
ing and self-esteem. They stated that
because each dietary failure may produce lower self-esteem and be-
cause lower self-esteem may, in turn, make dietary failure more likely,
individuals who undertake chronic dieting may enter a spiral in which
each failure at dieting produces greater negative affect and precludes
either successful acceptance or successful alteration of their bodies.
(p. 139)
Another factor that must be considered is that women who
become bulimic have additional problems that contribute to
their vulnerability, and to the extent that these other problems
are correlated with low self-esteem, the importance of self-
esteem per se may be exaggerated when they are not taken into
account. The best way to examine this hypothesis is to assess
the relations between several vulnerability factors (including
self-esteem) and the continuum of bulimic symptoms among
women who have not sought treatment for bulimia. If this can
be done in a prospective study, the causal direction of any link
between low self-esteem and bulimic symptoms can be tested.
This method was employed by Vohs and her colleagues
(Vohs, Bardone, Joiner, Abramson, & Heatherton, 1999; Vohs
et al., 2001; for a summary, see Vohs et al., 2002), who con-
ducted two prospective studies with a sample totaling more
than 400. Vohs et al. (1999) measured self-esteem, perfection-
ism, feeling overweight, and bulimic symptoms at the initial
assessment and then measured bulimic symptoms again an av-
erage of 9 months later. They found that although self-esteem
PSYCHOLOGICAL SCIENCE IN THE PUBLIC INTEREST
R.F. Baumeister et al.
VOL. 4, NO. 1, MAY 2003 35
scores were significantly correlated with bulimic symptoms
(.52 and .36 at the first and second assessments, respec-
tively), self-esteem was not a significant predictor of change in
bulimic symptoms in a statistical model that also included the
effects of perfectionism and feeling overweight. Bulimia was
not predicted by the combination of level of self-esteem with
level of perfectionism or with feeling overweight. Analyses
showed, however, that a combination of all three factors was a
significant predictor of increases in bulimic symptoms. Specifi-
cally, the combination of high perfectionism and feeling over-
weight was predictive of increased bulimic symptoms across 9
months, but only among low self-esteem women. This finding
suggests that low self-esteem is a risk factor for bulimic symp-
tomatology, but only when it is accompanied by exceedingly
high standards for oneself (perfectionism) and a feeling that
those standards are not being met (feeling overweight).
A second study, across a 5-week period, showed the same
pattern of results (Vohs et al., 2001). In this study, there were
simple associations between self-esteem scores and bulimic
symptoms, but the only statistical factor to predict increased bu-
limic symptoms was the combination of having low self-esteem,
having high body dissatisfaction, and being highly perfectionis-
tic. This study also examined the development of anxiety and
depressive symptoms, and analyses showed that self-esteem did
not, on its own, predict change in depression or anxiety when
perfectionism and body dissatisfaction were taken into account.
However, the results for depression mirrored the pattern seen
for bulimia: The greatest increase in depressive symptoms oc-
curred among women who had high perfectionism scores, high
body dissatisfaction, and low self-esteem.
Taken together, these studies suggest that self-esteem does
play a role in various eating problems. However, it is also im-
portant to consider that the role of self-esteem may be much
more complex than is suggested by the simple correlations be-
tween constructs. Bulimics, in particular, may have low self-
esteem, although low self-esteem may be both a predisposing
cause and a consequence of the disordered eating. A recent
study of obese women supports this idea. Matz, Foster, Faith,
and Wadden (2002) found that self-esteem was second only to
being teased in the ability to predict degree of body-image dis-
satisfaction. However, we again note that the influence of self-
esteem on pathological eating patterns may be a consequence
of the presence of other risk factors (such as perfectionism and
body dissatisfaction) or of the interaction between low self-
esteem and these other factors.
Conclusion
Most studies reviewed in this section relied on self-report,
and the possibility of response bias is therefore substantial.
Even so, the studies do not provide a great deal of evidence that
high self-esteem can prevent undesirable outcomes. The most
promising possibility is that high self-esteem might prevent bu-
limia, and there are some links to longevity and physical health
that seem well worth further study.
Most studies on self-esteem and smoking have failed to find
any significant relationship, even with very large samples and
the correspondingly high statistical power. The studies that
have found a relationship have tended to find it only in some
analyses, such as only among girls and not boys, or only for
some ages, or only for current smokers and not for ex-smokers
or experimental smokers.
Self-esteem also does not seem relevant to use of alcohol
and other drugs, although we did not find much highly rigorous
data on use of illegal drugs, and it is conceivable that further
studies might paint a different picture. Large, longitudinal in-
vestigations have tended to yield no relationship between self-
esteem and either drinking in general or heavy, problem drink-
ing in particular. The few studies that have found links point in
different directions, with some linking alcohol use to higher
self-esteem and others to lower self-esteem. Moreover, even if
some correlations between low self-esteem and substance
abuse could be found, they might well turn out to be based on
third variables, such as perceived control (Wills, 1994). What-
ever the causes of alcohol abuse and drug addiction, low self-
esteem per se does not appear to be one of them.
Self-esteem does not appear to prevent early sexual activity
or teen pregnancy. Some studies have found self-esteem to be
unrelated to sexuality. Others have yielded small effects that
sometimes point in contrary directions. One promising pattern
suggests that high self-esteem reduces sexual inhibitions, en-
abling women to engage in various sexual practices more
freely and enabling people to accept their homosexual tenden-
cies. Still, the causal inference is speculative, and it is quite
plausible that greater sexual freedom leads to popularity, which
in turn boosts self-esteem, as the sociometer model might pre-
dict. Indeed, there are various signs supporting the hypothesis
that popularity is the hidden variable behind the correlations
between high self-esteem and greater sexual activity.
The data do, however, suggest a role for self-esteem with re-
spect to disordered eating patterns. Concurrent and prospective
findings indicate that low self-esteem is a risk factor in disor-
dered eating, with evidence suggesting that the development of
bulimic symptoms may be affected both directly by the pres-
ence of low self-esteem and indirectly by the interaction of low
self-esteem with other factors (e.g., Vohs et al., 1999, 2001).
However, eating disorders are usually preceded by chronic di-
eting and body dissatisfaction, which themselves are related to
low self-esteem (see, e.g., Heatherton & Polivy, 1992; Matz et
al., 2002). And it is also important to note that some women
with eating disorders also have other disorders, such as anxiety
and depression, that are also related to low self-esteem. In
short, there is a complex set of relationships through which
self-esteem has its effects on vulnerability to eating disorders.
Nonetheless, on the whole, we conclude that low self-esteem is
a concurrent and prospective risk factor for eating disorder
symptoms.
PSYCHOLOGICAL SCIENCE IN THE PUBLIC INTEREST
Benefits of Self-Esteem
36 VOL. 4, NO. 1, MAY 2003
GENERAL DISCUSSION AND CONCLUSIONS
Our goal was to survey the research literature in order to as-
sess whether self-esteem has important consequences. The task
was complicated by four main factors: (a) the many thousands
of articles alluding to self-esteem; (b) the tendency of people
high in self-esteem to rate themselves as superior on many di-
mensions (and the floccinaucinihilipilification among people
with low self-esteem); (c) the difficulty of establishing the di-
rection of causal relationships; and (d) the heterogeneity of
high self-esteem. When possible, we restricted our search to
studies with objective measures, and we have emphasized large
longitudinal studies that offered some opportunity to assess the
direction of causality.
Our particular interest was in the possible benefits of high
self-esteem and the corresponding costs of low self-esteem. As
we note later, this is not the only possible theoretical perspec-
tive, but it is the one that has characterized the self-esteem
movement and indeed our own initial interest in self-esteem.
Does high self-esteem make life better? Should parents, teach-
ers, or anybody else seek to boost self-esteem whenever possi-
ble? We begin with a summary of our main findings.
Summary of Findings
High self-esteem does not reliably cause any improvement
in academic performance. If anything, high self-esteem may be
the result (but only weakly) of doing well in school. Other fac-
tors may underlie both self-esteem and academic performance.
People with high self-esteem do better than other people on
some jobs and tasks, although most laboratory studies and many
field studies have found no difference. Quite possibly, occupa-
tional success leads to high self-esteem rather than the reverse.
High self-esteem does improve persistence in the face of
failure, especially when persistence is an adaptive strategy.
People with high self-esteem are more willing than others to
choose their own strategies, and they are more responsive to
situational cues indicating when to persist and when to move
on to a more promising alternative.
People high in self-esteem regard themselves as better liked
and more popular than others, but most of these advantages ex-
ist mainly in their own minds, and objective data (such as rat-
ings by peers) generally fail to confirm them. In some cases,
such as after an ego threat, people with high self-esteem are ac-
tually disliked more than others.
There is relatively little known about the impact of self-esteem
on close relationships. The evidence that is available indicates
both advantages and disadvantages. Thus, wives complain more
about husbands with low than high self-esteem, although the di-
rection of causality is unclear. People with high self-esteem are
more willing than others to terminate a relationship, but there is
no sign that their relationships actually suffer more or end
sooner.
People with high self-esteem sometimes perform better than
people with low self-esteem in groups. They speak up more
and are recognized by peers as contributing more. There are
some weak simple correlations between self-esteem and lead-
ership, but analyses that control for other factors have found
that self-esteem has little in the way of direct and unique causal
impact. The possibility remains, however, that self-esteem has
indirect effects that are mediated by other factors, such as lead-
ership efficacy.
People with high self-esteem have a stronger tendency than
those with low self-esteem to judge and treat their own groups
more favorably than out-groups. (This may be regarded either
positively, as in supporting group pride, or negatively, as in
contributing to prejudice and discrimination.)
Self-esteem is essentially unrelated to aggression, although
this overall pattern may conceal divergent trends for different
kinds of high self-esteem. Some categories of high self-esteem
(such as defensiveness or narcissism) are associated with height-
ened aggression. High self-esteem predicts being a bully and sup-
porting a bully–but a different category of high self-esteem
predicts defending victims against bullies. High self-esteem may
thus amplify both prosocial and antisocial tendencies. Low self-
esteem is associated with victimization, although whether the
low self-esteem is the cause or consequence is not known.
Self-esteem has at best a weak relationship to delinquency.
Many findings are tainted by possible self-report bias and by
inconsistencies across studies. Some studies suggest that low
self-esteem predisposes people to engage in more delinquent
behavior later on (at least, by their own report), whereas other
studies have found no effect. However, one recent study pro-
vided good evidence that low self-esteem leads to delinquency
(Trzesniewski et al., 2002).
Likewise, there is little to connect self-esteem directly to anti-
social behavior. If anything, some subcategories of high self-
esteem are associated with antisocial behavior. Again, self-esteem
may magnify both prosocial and antisocial tendencies.
Happiness appears to be the most desirable correlate of high
self-esteem. Although research is needed to establish causality
and to control for other variables, it seems quite possible that
high self-esteem contributes to making people happy. Low self-
esteem is linked to depression and may be a risk factor for it.
However, the relationship is weak, inconsistent, and condi-
tional on other variables.
The buffer hypothesis proposes that high self-esteem helps
people cope with stress and adversity. There are some positive
findings to support this view. Other findings, however, indicate
that self-esteem is more relevant under low than high stress
(low self-esteem poisons the good times). Yet other findings in-
dicate no effect of self-esteem in either direction. Still, no find-
ings show worse outcomes or poorer coping among people
with high self-esteem than among those with low self-esteem.
Thus, self-esteem may contribute to coping and adjustment af-
ter stress or trauma, although the precise nature of the relation-
ship may be complicated and may depend on other factors.
Self-esteem is largely irrelevant to smoking. At best, to a
small degree high self-esteem may lead to less smoking among
PSYCHOLOGICAL SCIENCE IN THE PUBLIC INTEREST
R.F. Baumeister et al.
VOL. 4, NO. 1, MAY 2003 37
young females. Given the null findings in some large studies,
self-esteem is probably not worth considering as a cause of
smoking. Use of alcohol or other drugs also does not show a
consistent relationship to self-esteem. If anything, young peo-
ple with high self-esteem may be more willing than others to
experiment.
High self-esteem does not prevent early, extensive, or risky
sexual activity. If anything, people with high self-esteem dis-
play fewer inhibitions, more disregard for risks, and greater
sexual initiative.
Low self-esteem is a risk factor that can contribute to eating
disorders, especially bulimia. Its relation to bulimia may be es-
pecially strong in the presence of other factors, such as body
dissatisfaction and perfectionism.
We conclude our summary of findings with some general
points.
With the exception of the link to happiness, most of the ef-
fects are weak to modest. Self-esteem is thus not a major
predictor or cause of almost anything (again, with the possi-
ble exception of happiness).
Moreover, the effects of self-esteem become weaker as the
criteria for evidence become more objective. It is perhaps no
accident that the strongest apparent benefit of self-esteem
has been found for the most subjective outcome, namely,