Content uploaded by Richard Robins
All content in this area was uploaded by Richard Robins on Nov 05, 2018
Content may be subject to copyright.
An emerging epidemic of narcissism or much ado about nothing?
M. Brent Donnellan
, Kali H. Trzesniewski
, Richard W. Robins
Department of Psychology, Michigan State University, East Lansing, MI 48823, United States
Department of Psychology, The University of Western Ontario, Canada
Department of Psychology, University of California, Davis, CA 95616-8686, United States
Available online 27 December 2008
We evaluated secular increases in narcissism in light of issues raised by Twenge and Foster (2008)
[Twenge and Foster (2008). Mapping the scale of the narcissism epidemic: Increases in narcissism
2002–2007 within ethnic groups. Journal of Research in Personality]. Year of data collection (1996–
2008) and narcissism scores were weakly correlated in a sample of 30,073 participants (rs ranged from
0.019 to 0.044). The effect sizes were well below the conventional threshold for a ‘‘small” effect. Based
on these and previous ﬁndings, we concluded that there is little evidence of a secular increase in narcis-
sism in our datasets.
Ó2008 Elsevier Inc. All rights reserved.
Twenge and colleagues have recently raised concerns about an
epidemic of narcissism among today’s college students and
Twenge (2006) has argued that this increase is an unintended ef-
fect of the self-esteem movement. In fact, Twenge (2006) has sug-
gested that, ‘‘the self-esteem movement has created an army of
little narcissists” (p. 223). In a series of articles, including a recent
one entitled, ‘‘Mapping the Scale of the Narcissism Epidemic”
(Twenge & Foster, 2008), Twenge and colleagues claim that narcis-
sism levels have been rising dramatically over the past decade or
two (Twenge, 2008; Twenge, Konrath, Foster, Campbell, & Bush-
man, 2008a; Twenge, Konrath, Foster, Campbell, & Bushman,
2008b). However, the evidentiary basis of this claim has been the
subject of much controversy within the scientiﬁc literature (Trzes-
niewski, Donnellan, & Robins, 2008a; Trzesniewski, Donnellan, &
Robins, 2008b) as well as in the popular media (e.g., Associated
Press, February 27, 2007; Rosenbloom, January 17, 2008). The
aim of this brief report is to re-examine the evidence for secular in-
creases in narcissism in light of the issues, criticisms, and ﬁndings
presented in Twenge and Foster (2008).
1. Previous research on secular changes in narcissism
In a previous article (Trzesniewski et al., 2008a), we presented
data that failed to support Twenge’s (2006) assertion that ‘‘narcis-
sism is much more common in recent generations” (p. 69). Specif-
ically, we found nearly the same average score on the 40-item
Narcissistic Personality Inventory (NPI; Raskin & Terry, 1988) for
data collected on large samples of college students: (a) from
1979 to 1985 at the University of California (UC) Berkeley and
UC Santa Cruz; (b) in 1996 at UC Berkeley; and (c) from 2002–
2007 at UC Davis. In a response to Trzesniewski et al. (2008a);
Twenge et al. (2008b) re-analyzed secular trends in their meta-
analytic database after restricting their analyses to the seven sam-
ples from California universities (all but one sample were from UC
campuses) and replicated our original ﬁnding that college students
in California have not shown an increase in narcissism. In an at-
tempt to explain this null result, Twenge and Foster (2008) specu-
lated that the secular increase in narcissism that they believe is
occurring in society as a whole has been obscured at California uni-
versities because of countervailing increases in the proportion of
Asian–American students, who tend to score lower on measures
of narcissism than White students.
Whether secular trends vary across racial/ethnic groups is an
important question, albeit one that has been largely ignored by
all of the participants in this debate (e.g., Twenge et al. 2008a)
did not evaluate ethnicity as a moderator of secular trends in their
meta-analysis). In Trzesniewski et al. (2008a), we reported that
Asian–Americans students scored lower on the NPI than White stu-
dents, but we found no evidence that ethnicity moderated the sec-
ular trends. However, based on a reanalysis of our UC Davis data
(the only data for which ethnicity was available to them), Twenge
and Foster (2008) concluded that narcissism levels actually in-
creased from 2002 to 2007 for Whites and Asian Americans (but
not African Americans or Latinos/as). They interpreted this result
as supporting their contention that there is an epidemic of narcis-
sism among today’s college students.
The goal of this brief report is to revisit this question of whether
there is an epidemic of narcissism among today’s youth by
0092-6566/$ - see front matter Ó2008 Elsevier Inc. All rights reserved.
E-mail address: email@example.com (M.B. Donnellan).
The ﬁrst and second authors contributed equally to this article.
Journal of Research in Personality 43 (2009) 498–501
Contents lists available at ScienceDirect
Journal of Research in Personality
journal homepage: www.elsevier.com/locate/jrp
examining changes in narcissism from 1996 to 2008 as a function
of ethnicity using potentially more meaningful and easier to
understand effect size measures. In addition, we examined
whether any observed trends held across gender and across several
speciﬁc facets of narcissism.
NPI scores were drawn from prescreening sessions conducted
at U.C. Berkeley in 1996 and at U.C. Davis from 2002 to 2008.
We restricted the analyses to the 30,073 participants between
18 and 24 years of age who completed the NPI and self-identi-
ﬁed their racial/ethnic background. The dataset used for some
of these analyses differs from the dataset analyzed by Twenge
and Foster (2008) with additional data collected in 2007
(n= 2025) and 2008 (n= 2405) and with ethnicity data analyzed
for the 1996 sample (n= 571).
Given the number of comparisons
and the large sample sizes we set the alpha level to 0.01 to eval-
uate statistical signiﬁcance.
There are two general ways to analyze secular trends in nar-
cissism. First, mean scores can be computed each year and then
correlated with year of data collection. This analysis produces
what is referred to as an ‘‘ecological correlation” (the Nin these
analyses is the number of years that data were collected). Sec-
ond, the individual participants’ scores can be correlated with
the year each score was collected (the Nin these analyses is
the number of participants). This kind of analysis produces the
kind of correlation coefﬁcient that is commonly used in person-
ality research (i.e. the unit of analysis is individual scores rather
than yearly averages). Both approaches presumably reﬂect the
degree of association between narcissism levels and year of
assessment, and thus tell us something about secular trends.
As we will show, however, the two approaches produce dramat-
ically different effect sizes.
The method that Twenge and Foster (2008) used to obtain
their most impressive effects sizes was the ecological approach.
Speciﬁcally, to test their assumption that ethnicity confounds
the general trend, they predicted yearly means broken down
by ethnicity from the 6 years of data collection (2002–2007) re-
ported in Trzesniewski et al. (2008a). From this analysis, they re-
ported large standardized regression coefﬁcients (i.e.
correlations) for each ethnic group (b= 0.96, 0.68, 0.72, and
0.56 for Whites, Asian Americans, Latinos/as, and African Amer-
icans, respectively). One problem with the ecological approach is
that the effect size can be quite large because individual-level
variability is not factored into the standardization of the regres-
When individual–level effects were computed using the second
approach, the resulting standarized coefﬁcients (i.e., correlations)
were considerably smaller (b= 0.045, 0.002, 0.021, and 0.048 for
Whites, Asian Americans, Latino/a, and African Americans, respec-
In contrast to Twenge and Foster’s analysis of the same data,
these coefﬁcients suggest that little secular change has occurred,
even when the trends were analyzed within ethnic groups. Twenge
and Foster (2008) suggest that these small effects could amount to
potentially substantial changes if they played out at the same rate
over 24 or so years. However, such an extrapolation might be unwar-
ranted given the restricted time frame moreover, it is unclear from a
‘‘generational change” perspective why college students in 2002 or
2003 should be much different from college students in 2006 or
Based on these concerns, we evaluated within-group secular
changes in narcissism by extending the sample analyzed by
Twenge and Foster (2008) to include data from 1996, Fall 2007,
and Winter and Spring of 2008. Speciﬁcally, our goal was to exam-
ine secular trends within ethnic group, within gender, and across
the NPI subscales for this longer interval. Table 1 displays mean
NPI scores as a function of ethnicity and year of data collection.
Averages were based on a 0–40 metric for scoring the NPI. To quan-
tify changes over time, we calculated the individual–level correla-
tions between NPI scores and year (1996–2008). This information
is reported in row labeled r
. For example, the association be-
tween individual scores for Whites and year of data collection was
0.019 (p= 0.067) and the correlation between year of data collec-
tion and NPI score in the total sample was 0.024 (p< 0.01). These
correlations between year and NPI scores were quite small
(range = 0.019–0.044) and they provided little indication of a
meaningful secular increase in narcissism from 1996 to 2008.
We next analyzed the trends separately for men and women of
each ethnicity. This is an instructive analysis because Twenge et al.
(2008a) found that only women showed a statistically signiﬁcant
increase in NPI scores. The correlations for women are reported
in the row labeled r
and the correlations for men are reported
in the row labeled r
For example, the correlation between indi-
vidual scores for White women and year of data collection was
0.021 (p= 0.111), whereas the corresponding correlation for White
men was 0.024 (p= 0.184). All of the correlations were small and
none exceeded 0.10, the conventional threshold for a ‘‘small” ef-
fect. As another illustration of the magnitude of the effect, year
of data collection accounted for 0.1% of the variance, whereas gen-
der accounted for 1.2% (i.e., 12 times as much) in a regression anal-
ysis predicting NPI scores.
It is possible that only certain facets of narcissism have been
increasing over time. To explore this possibility, we computed
correlations between the seven NPI subscales (Raskin & Terry,
1988) and year of data collection, separately in the total sample
and for each ethnic group. The vast majority of the 42 correla-
tions were below 0.05 (median r= 0.017) and none were above
0.10 (see Table 2). Of note, the weakest correlations, hovering
around zero (e.g., r= 0.005 in the total sample, p= 0.351), were
found for the Entitlement subscale. Konrath, Bushman, and
Campbell (2006) identiﬁed Entitlement as the best predictor of
laboratory aggression, even outperforming the overall NPI score.
Thus, this particularly socially toxic aspect of narcissism showed
no increase from 1996–2008, contrary to the characterization of
the current generation of college students as ‘‘Generation Me”
(e.g., Twenge, 2006).
The 1996 data were not included in the Twenge and Foster (2008) analysis
because ethnicity information was not readily available when requested by Dr. Foster
in February of 2008. We tracked these down in light of the Twenge and Foster (2008)
analysis and provided them to their team in September of 2008.
We are not claiming the Twenge and her colleagues are committing the ecological
fallacy with these analyses (i.e., the error of assuming that effects at an aggregated
level generalize to a lower level). Rather, we are pointing out that correlating two
variables at an aggregated level yields an effect size that is quite different from the
traditional correlation used in personality research, which is based on individual level
data (see Robinson, 1950; Rosenthal, Rosnow, & Rubin, 2000, p. 2). Ecological
correlations are typically much larger than individual-level correlations (and can even
reverse signs) and it is not clear how personality researchers should interpret the
magnitude of these effects.
These results are based on analyses conducted using the exact same set of data
analyzed by Twenge and Foster (2008), which consists of the 20,627 participants
between 18 to 24 years of age who completed the NPI between 2002 and Spring of
2007 and self-identiﬁed their racial/ethnic background as African American (n= 553),
Asian American (n= 9969), Latino/Latina (n= 2,358), or White (n= 7,747). These
regression results were also provided to Dr. Twenge in Spring of 2008.
M.B. Donnellan et al. / Journal of Research in Personality 43 (2009) 498–501 499
Based on the overall pattern of results in this report, we are not
convinced that there has been either a widespread or a substantial
secular increase in narcissism. All of the effect sizes were well be-
low the conventional threshold for a ‘‘small” effect and, even in our
very large sample, there was no statistically detectable increase for
men (p= 0.105) or for Latinos and African Americans (p= 0.252 and
p= 0.253, respectively). Although very small effects can be highly
meaningful in some contexts, in the present context, we prefer to
follow recommendations by Hyde (2005) who has argued that Co-
hen’s criteria (1988) (e.g., 1988) for ‘‘small” (r= 0.10), ‘‘medium”
(r= 0.30), and ‘‘large” (r= 0.50) effects are reasonable for interpret-
ing contentious ﬁndings such as the presence of gender differences
(see pp. 586–587). In this case, the observed correlations are even
smaller than ‘‘small,” which argues for erring on the side of caution
before making pronouncements about an entire generation of
Moreover, it is occasionally overlooked that the secular increase
in narcissism reported by Twenge et al. (2008a) held only for wo-
men; the association between year and mean NPI scores was not
statistically detectable for men whereas it was statistically detect-
able for women based on their sub-analyses of the 44 studies that
had information separately reported by gender (out of 85 total
studies). The present analyses further hinted at this gender differ-
ence, although the trend was miniscule even for women. Thus,
even if we were to accept Twenge et al.’s claim that narcissism lev-
els are rising, this trend seems to be restricted to women. Such an
apparent boundary condition raises questions about how to best
interpret the reported secular effect as scores on the NPI can reﬂect
heightened social potency as well as more socially toxic aspects of
personality (see Trzesniewski et al. 2008b).
Finally, we emphasize that there is nothing in any of these data
that suggests an epidemic of narcissism. Why? The NPI is mea-
sured on an arbitrary metric (e.g., Blanton & Jaccard, 2006) and
there is no basis for declaring when an average score indicates
‘‘excessively” high levels of the characteristic in question. In fact,
even in the most recent time point (2008), when the epidemic
was presumably raging, participants on average endorsed fewer
than half (about 40%) of the narcissistic responses; whether this
value should be considered high or low in an absolute sense is
impossible to determine with the arbitrary metric of the NPI. In-
deed, claims about an ‘‘epidemic of narcissism” essentially boil
down to quite small changes over a fairly restricted range of scores
all of which typically average below the midpoint of the scale.
Based on these considerations, we do not believe that ‘‘epidemic”
is an appropriate adjective in this context.
We continue to believe that a conservative approach is war-
ranted with respect to the empirical status of generational in-
creases in narcissism. Our extensive analyses of data from over
30,000 participants showed uniformly small secular trends, which
failed to reach statistical signiﬁcance for men, for two of the four
ethnic groups, and for three of the facets of narcissism. Moreover,
all of the existing NPI evidence is based on convenience (i.e., non-
probability) samples, and such samples are highly problematic for
making inferences about trends in the general population. Thus,
evidence for an increase in narcissism is far from clear-cut and
there are good reasons why researchers should be wary of labeling
any apparent changes in this multifaceted dimension of personal-
ity as evidence of an epidemic.
Associated Press. (2007). College students think they’re so special. from <http://
www.msnbc.msn.com/id/17349066/> Retrieved 10.05.2007.
Blanton, H., & Jaccard, J. (2006). Arbitrary metrics in psychology. American
Psychologist, 61, 27–41.
Means, standard deviations, and correlations of NPI Scores with year of data collection (1996–2008), separately for each ethnic group.
Year African–American Asian–American Latino/a White ‘‘Other” Total sample
MSD NM SD NMSD NMSD N M SD N M SD
1996 18.24 5.98 28 13.53 6.84 254 15.82 6.22 82 17.78 6.71 163 14.12 5.84 44 15.35 6.86
2002 17.38 6.55 70 13.26 6.67 1133 15.51 6.82 252 15.30 6.59 1107 16.64 7.03 535 14.85 6.84
2003 17.38 6.89 109 13.79 6.92 1818 16.32 6.65 420 15.58 6.66 1644 16.57 6.93 817 15.18 6.90
2004 17.99 7.30 92 13.64 6.87 1884 16.23 6.70 438 15.64 6.88 1505 16.54 6.90 830 15.10 6.98
2005 18.58 6.23 92 13.96 6.84 1769 16.12 6.89 437 15.96 6.87 1332 16.45 7.03 777 15.31 6.97
2006 18.24 6.48 119 13.70 6.77 2080 16.17 6.61 487 15.97 6.96 1348 16.48 6.86 924 15.19 6.94
2007 18.57 6.47 125 14.34 6.98 2171 16.40 6.61 528 16.56 7.15 1287 16.76 7.11 967 15.68 7.10
2008 18.65 5.92 48 14.33 6.73 1049 16.32 6.69 267 16.00 7.00 585 17.18 6.81 456 15.58 6.89
Overall 18.12 6.56 683 13.87 6.85 12158 16.18 6.68 2911 15.88 6.87 8971 16.61 6.95 5350 15.27 6.96
African–American Asian–American Latino/a White ‘‘Other” Total sample
0.021 .019 0.027 0.024
0.015 .021 0.042 0.030
0.001 0.028 0.047 .024 0.013 0.016
is the correlation for both men and women, whereas r
is the correlation for women and r
is the correlation for men.
Correlations of NPI sub-scales with year of data collection (1996–2008), separately for each ethnic group.
Authority Exhibitionism Superiority Entitlement Exploitativeness Self-sufﬁciency Vanity
African–American 0.053 0.029 0.038 0.017 0.008 0.070 0.004
Asian–American 0.017 0.016 0.044
Latino/a 0.002 0.003 0.056
0.003 0.012 0.039 0.002
White 0.011 0.013 0.022 0.008 0.037
Other 0.021 0.003 0.040
0.002 0.017 0.026 0.029
Total sample 0.010 0.009 0.031
500 M.B. Donnellan et al. / Journal of Research in Personality 43 (2009) 498–501
Cohen, J. (1988). Statistical power analysis for the behavioral sciences (2nd ed.).
Hillsdale, NJ: Lawrence Erlbaum.
Hyde, J. S. (2005). The gender similarities hypothesis. American Psychologist, 60, 581–592.
Konrath, S., Bushman, B. J., & Campbell, W. K. (2006). Attenuating the link between
threatened egotism and aggression. Psychological Science, 17, 995–1001.
Raskin, R., & Terry, H. (1988). A principal-components analysis of the Narcissistic
Personality Inventory. Journal of Personality and Social Psychology, 54, 890–902.
Robinson, W. S. (1950). Ecological correlations and the behavior of individuals.
American Sociological Review, 15, 351–357.
Rosenbloom, S. (2008). Generation me vs. you revisited.<http://www.nytimes.com/
Rosenthal, R., Rosnow, R. L., & Rubin, D. B. (2000). Contrasts and effect sizes in
research: A correlational approach. New York: Cambridge University Press.
Trzesniewski, K. H., Donnellan, M. B., & Robins, R. W. (2008a). Do today’s young
people really think they are so extraordinary? An examination of secular
changes in narcissism and self-enhancement. Psychological Science, 19,
Trzesniewski, K. H., Donnellan, M. B., & Robins, R. W. (2008b). Is ‘‘Generation Me”
really more narcissistic than previous generations? Journal of Personality, 76,
Twenge, J. M. (2006). Generation Me: Why today’s young Americans are more
conﬁdent, assertive, entitled–and more miserable than ever before. New York: Free
Twenge, J. M. (2008). Generation Me, the origins of birth cohort differences in
personality traits, and cross-temporal meta-analysis. Social and Personality
Twenge, J. M., & Foster, J. D. (2008). Mapping the scale of the narcissism epidemic:
Increases in narcissism 2002-2007 within ethnic groups. Journal of Research in
Personality, 42, 1619–1622.
Twenge, J. M., Konrath, S., Foster, J. D., Campbell, W. K., & Bushman, B. J. (2008a).
Egos inﬂating over time: a cross-temporal meta-analysis of the Narcissistic
Personality Inventory. Journal of Personality, 76, 875–901.
Twenge, J. M., Konrath, S., Foster, J. D., Campbell, W. K., & Bushman, B. J. (2008b).
Further evidence of an increase in narcissism among college students. Journal of
Personality, 76, 919–927.
M.B. Donnellan et al. / Journal of Research in Personality 43 (2009) 498–501 501