Sex Roles, Vol. 38, Nos.

9/10, 1998

Brief Report

Gender Differences in the Structure of

Narcissism : A Multi-Sam ple An alysis of the
1
Narcissistic Person ality Inventory
Brian T. Tschan z2
University of Utah

Carolyn C. Morf
University of Toron to

Charles W. Tu rn er
University of Utah

Mu lti- sam ple an alysis of th e E m m on s factor m o del of th e Narcissistic

Person ality Inventory (NPI) was used to test the prediction that exploitive
tendencies an d feelin gs of entitlem ent are less central to the con struct of
narcissism am on g fem ales than they are am on g m ales. As predicted, the
h yp o th e s i s o f c ro s s - ge n d e r e q u i v a le n c e o f th e E m m o n s fa c to r
varian ce/covarian ce m atrix was rejected. Follow-up analyses confirm ed that this
hypothesis was rejected prim arily because Exploitiveness/En titlem ent showed
weaker correlation s with the other Em m on s factors amon g females that it did
am ong m ales. Results are discu ssed in terms of such factors as norm s regarding
appropriate sex role condu ct, as well as the power differential between m ales
an d females.

The Narcissistic Personality Inve ntory (NPI, Raskin & Hall, 1979) , was de-
velope d to measure individual diffe rences in the extent to which a grandi-
ose se nse of self and a grandiose fantasy life combine with hype rsensitivity,
exhibitionism, fee lings of e ntitle ment, interpersonal exploitive ness, and a
lack of e mpathy for othe rs to form dominant themes of an individual ’s per-
1
We would especially like to thank Fre d Rhodewalt for his invaluable advice and comments
on this paper.
2
To whom correspondence and reque sts for reprints should be addre ssed at Unive rsity of
Utah, Departmen t of Psychology, 390 S., 1530 E., Room 502, Salt Lake City, UT 84112-0211.

863

0360-0025/98/0500-0863$ 15.00/0 Ó 1998 Plenum Publishing Corporation

864 Tsch an z et al.

sonality. It is the most widely used measure in the area and has de mon-
strate d ample re liability and construct validity (e.g., Emmons, 1984, 1987;
Raskin & Terry, 1988) . In addition, factor analyse s of the NPI (Emmons,
1987; Raskin & Terry, 1988) sugge st that it measure s the key constitue nts
of the syndrome . For e xample , Emmons (1987) found that the NPI con-
taine d four factors: Le ade rship/Authority, Self-absorption/Se lf-admiration,
Supe riority/Arrogance , and Exploitive ness/Entitle ment.
Howe ver, some the orists have questione d whether the type of narcissism
that is purporte dly asse ssed by the NPI can be validly ge neralize d to both
male and female e xperie nce (e.g., Akthar & Thompson, 1982; Haake n, 1983;
Philipson , 1985) . Although only a fe w inve stigators (e .g., Carroll, 1987;
McCann and Biaggio, 1989) have reporte d a ge nde r-specific patte rn of re-
sults in e mpirical rese arch with the NPI, there is theore tical and empirical
reason to be lieve that the hypothe sis of gende r diffe rences in narcissism, par-
ticularly as it is asse ssed by the NPI, may be at least in part correct.
Spe cifically, past research sugge sts that exploitive tende ncies and open
displays of feelings of e ntitle ment will be less integral to narcissism for fe-
male s than for male s. For female s such displays may carry a greate r possibility
of negative social sanctions be cause they would violate ste reotypical gender-
role e xpectancie s for women, who are e xpe cted to e ngage in such positive
social behavior as being tende r, compassionate , warm, sympathe tic, sensitive ,
and unde rstanding (Martin, 1987). Indee d, it has be en found that women in
le ade rship positions are e valuate d negative ly if they violate these e xpe ctan-
cie s by be ing autocratic and directive (Eagly, Makhijani, & Klonsky, 1992),
or by occupying leade rship positions which typically require the ability to di-
rect and control people (Butler & Geis, 1990). Moreover, it appe ars that in
orde r to influe nce men, women must appe ar to be sociable , like able people ;
where as men, irrespective of the gender of the targe t of pe rsuasion, merely
must appe ar to be competent (Carli, Lafleur, & Loebe r, 1995).
The above describe d the ory and research le d to the prediction that
Exploitive ness/Entitle ment would be a le ss well-inte grate d compone nt of
the narcissistic syndrome (as measure d by the NPI) for fe male s than it
would be for males. Spe cifically, it was predicted that the corre lations of
the Exploitive ne ss/Entitle ment factor with the Le ade rship/Authority, Self-
absorption/Se lf-adm iration, and Supe riority/Arrogance factors of the NPI
would be lower for fe male s than for male s.3
3
Raskin and Terry (1988) identified an alternative 7 factor structure for the NPI which in-
clude d the factors Authority, Exhibitionism, Supe riority, V anity, E xploitiveness, E ntitlement,
and Self-Sufficiency. We did not use this factor structure in the present research primarily
be cause the large data set on which the analyses were conducted did not include all of the
items comprising these factors. Moreover, the Emmons factor structure, due to its greate r
simplicity, lends itself to a more tractable and straightforward test of the hypothesis unde r
consideration.
Gend er an d Narcis sism 865

To test this pre diction, we employe d multi-sample analysis, a method

by which rese archers may te st for cross-population invariance in the op-
e ration of a give n me asuring instrum e nt (J ö re skog & S ö rbom, 1993) .
Brie fly, multi-sample analysis utilizing LISREL approache s the question of
cross-population invariance by testing a se ries of orde red hypothe se s con-
cerning the structure of the data. Only one such hypothe sis, the hypothe sis
of e quivale nt factor variance s/covariance s, is directly re le vant to the ques-
tion unde r conside ration in the prese nt study and the refore this hypothe sis
(to be de scribed in greate r detail be low) is the only one re porte d.

METHOD

Participan ts
Data on the NPI were colle cted from stude nts from se veral introduc-
tory psychology classes at the Unive rsity of Utah. These stude nts partici-
pate d in mass te sting sessions conducte d at the be ginning of the quarte r
during the 1992-93 and 1993-94 acade mic years. Among the se stude nts,
only those who had no missing data on any of the NPI items and who had
indicate d the ir ge nder were use d in the analyse s re porte d in this study.
O ur sample thus consiste d of 1029 females and 1060 male s. Although we
do not have exact data about the precise e thnic make-up of our sample ,
the introductory psychology classe s tend to be representative of the state
as a whole . Spe cifically, the 1990 Unite d State s Bureau of the Ce nsus re-
ports that 93.8% of Utah ’s population is White , no more than 5% of whom
are Hispanic. The re mainde r of the population is large ly compose d on non-
White Hispanics, Asians and Pacific Islande rs, Native Americans, and Af-
rican Americans.
All of the se participants re ceive d credit towards the ir introductory psy-
chology class for the ir participation.

Instrum ent

The particular ve rsion of the NPI containe d the 37 items which had
loadings of .35 or greater on at le ast one of Emmons ’ (1987) 4 factors
(Rhode walt & Morf, 1995) .

Statistical Analyses

We e valuate d the cross gende r e quivale nce in the factor structure of

the NPI by conducting multi-sample analysis using LISREL 8 (Jö reskog &
866 Tsch an z et al.

Table I. Means and Standard Deviations of the Emmons Factors

and the NPI as a Function of Ge nder a
Gende r

Fe males Males

M SD M SD
L/A 5.49 2.05 5.76 1.98
S/S 4.22 1.62 4.49 1.64
S/A 3.40 1.72 3.98 1.88
E /E 2.28 1.73 2.58 1.82
NPI 15.40 4.84 16.82 5.06
a
L/A, Leade rship/Authority; S/S, Self-absorption/Self-admiration; S/A,
Supe riority/Arrogance ; E/E, E xploitiveness/Entitlement; NPI, Narcis-
sistic Personality Inventory. All mean ge nder differences, using the
Tuke y te st, were significant at p < .01.

S ö rbom, 1993) . Inte r-item polychoric coe fficients served as the basis for
the variance -covariance matrice s generated by PRELIS (Jö reskog & S ö r-
bom, 1990) . The parame te rs of LISREL’s factor analyse s were estimated
by the method of maximum like lihood using the variance -covariance ma-
trices as input. The metric of the Emmons factor model was fixed by setting
the ite m for which he reporte d the highe st factor loading equal to one on
its associate d factor.
The proce dure s for te sting the ge nde r invariance hypothe ses e ntail
comparing a mode l in which certain parame te rs are constraine d to be e qual
across gende rs with a le ss restrictive mode l in which the se parame ters are
fre e to take on any value . Be cause the more restrictive models (i.e ., the
factor variance s/covariance s were constraine d to be equal across ge nder)
were neste d within the le ss re strictive mode ls (i.e ., the factor variance s/co-
variance s were not constraine d to be e qual across ge nde r), diffe re nces in
chi-square betwe en the two models could be used to te st the null hypothe -
ses that the restrictions were true in the population. A more restrictive
mode l neste d within a le ss restrictive mode l is reje cted if the diffe re nce in
chi-square were significant.

RESULTS AND DISCUSSION

Before testing the hypothe sis of inte re st in this study, factor means
and standard deviations, as well as the mean and standard deviation of the
NPI, were calculate d for each ge nde r. As can be se en in Table I, although
our large sample size lent us sufficie nt powe r to de te ct that all of the means
were significantly highe r for the male s than the y were for the fe male s, the
diffe re nces be twee n the means across ge nde rs were none thele ss small.
Gend er an d Narcis sism 867

Table II. Matrix of Correlations among E mmons NPI Factors as a Function of Gende ra
Factors

Factors L/A S/S S/A E /E

L/A 1.00
F M
S/S .44 .50 1.00
F M F M

S/A .82 .83 .58 .50b 1.00

F M F M F M
c c
E /E .38 .53 .25 .44 .62 .73c 1.00
a
L/A, Le adership/Authority; S/S, Self-absorption/Self-admiration; S/A, Supe riority/Arrogance ;
E/E, E xploitiveness/Entitlement. F, Fe males; M, Males.
b
The gende r difference betwe en the correlations was significant at p .05.
c
The ge nder difference between the correlations was significant at p < .01.

More important for this research was that the standard deviations for
each of the factors were also of similar magnitude across ge nders. There-
fore, it is unlike ly that any gende r diffe rence s found in the covariations of
Exploitive ness/Entitle ment with the othe r factors could be due to a popu-
lation-spe cific restriction in range on one or more of the factors.

Test for the Cross-G ender Invarian ce of the Factor Covarian ces

For the test of this hypothe sis, the model was e stimate d using starting
value s for the fe male s, and the equality constraints were subse quently im-
posed upon the male s. Also factor loadings, e rror variance s/covariance s,
and factor variance s/covariance s were constraine d to be e qual across ge n-
ders. We the n looke d for a change in c 2 from a mode l in which the factor
loadings and error covariance s were constraine d to be equal across gende rs.
Consiste nt with predictions, the hypothe sis of cross-ge nde r invariance of
factor covariance s prove d to be unte nable , D c 2 (10) = 21.23, p < .02. In
orde r to examine the nature of the ge nder diffe re nces for covariance s be-
tween factors, corresponding factor corre lations were compare d across ge n-
der using Fishe r’s r to Z transformation. Table II presents the correlations
among the factors for e ach ge nde r. As can be se en in this table , there were
significant ge nder differe nces on all the covariance s that involve d the Ex-
ploitive ness/Entitle ment factor.4
4
An unexpecte d contribution to the re jection of the hypothesis of cross-gende r invariance of
NPI factor covariance s appeared to be the significant tende ncy for Se lf-absorption/self-admi-
ration to corre late more strongly with Supe riority/Arrogance among female s than among
males.
868 Tsch an z et al.

Thus the main hypothe sis that Exploitive ness/Entitle ment would be
le ss well-inte grate d with the othe r compone nts of narcissism fe male s than
it would be for male s was supporte d.5

Conclu sions

As stated in the introduction, we suspe ct that the reason Exploitive -

ness/Entitle ment is not as well-inte grate d into the narcissistic syndrome for
females as it is for males is because such behaviors are not sanctione d when
displaye d by fe male s because the y violate culturally he ld e xpe ctations re-
garding appropriate fe male be havior (Butler & Geis, 1990; Eagly et al.,
1992; Martin, 1987) .
Howeve r, it must be emphasize d that the ge ne rality of our results may
be limite d to narcissism as measure d by the NPI. Also, the y may not ge n-
eralize to the factor model of the NPI propose d by Raskin and Terry (1987) .
Moreove r, it is possible that the strong influe nce of the Mormon church
(which discourage s females from assuming positions of leade rship) on the
people in our sample produce d greater gende r differe nce s than might oth-
erwise be found among the American population in ge ne ral. Notwithstand-
ing, the findings on ge nder diffe rence s in e ffective leade rship style (e.g.,
Eagly e t al., 1992) , obtaine d from non-Mormon populations and on which
our predictions were base d, sugge st that our re sults may be re plicable in
othe r populations.
Despite these caveats, there are some intriguing aspe cts of our find-
ings. First, male and female narcissists in gene ral showed striking similari-
ties in the manne r in which most of the face ts of narcissism were inte grate d
with e ach othe r (note the ove rall cross-ge nde r similaritie s in factor corre-
lations displaye d in Table II). Se cond, standing out as figure against the
ground of the se similaritie s, were the pre dicted and theoretically important
ge nde r diffe re nces in the manne r and extent to which exploitive ne ss and
entitle ment were inte grate d with the othe r facets of narcissism. In light of
5
In the process of conducting a multi-sample analyses in LISREL 8, we were able to obtain
indices of the goodness-of-fit of the Emmons model. The se indices tended to indicate that
the model was a relatively poor-fitting model. For example, the values for the adjusted good-
ness-of-fit index, the non-normed fit index, and the comparative fit index were all .60 or
lower, or substantially less that what has been suggested as a criterion (i.e., .90) for a rea-
sonable fit to the data (Bentle r & Bonne t, 1980). However, it may be that the poor fit of
the Emmons model simply reflects that a structure that allows for a number of secondary
loadings, rather than a simple structure, should have been specified. Despite the weak fit,
however, it is a theore tically interesting model of narcissism. First, Emmons (1984, 1987) has
demonstrate d this factor structure is re plicable, and, second, Rhodewalt and Morf (1995)
provide d some e vide nce for the construct validity of the E xploitiveness/Entitlement factor.
Thus, we believe that the reported gende r differences represe nt me aningful ge nder differ-
ence s in the structure of narcissism.
Gend er an d Narcis sism 869

our findings, we recommend that future attempts be made to explicate the

diffe re ntial proce sse s by which male s and fe male s e ithe r do or do not in-
corporate fe elings of e ntitle ment and exploitive te nde ncies into the narcis-
sistic syndrome . For example , though this spe culation goe s a bit beyond
our curre nt findings, we suspe ct that research on narcissism looking at more
intern al and unde rlying psychological phe nomenology (e .g., lack of inter-
nalize d se lf, proble ms with self-este em re gulation, etc.) is like ly to obtain
similar outcome s for both gende rs. However, re search involving behavioral
m anifestation s of e xploitive ne ss and e ntitle ment should expe ct to obtain
ge nde r diffe rences.
Finally, we would also like to sugge st that our research de monstrate s the
value of using multi-sample analyse s or othe r similar covariance structure
analyse s for e xamining gender diffe rence s among various theoretically inte r-
related psychological constructs and processes. For example , an inspe ction of
the gender difference s on the means on the Emmons factors would lead one
to conclude that males are uniformly more narcissistic than female s (se e Ta-
ble I). However, our multi-sample analysis sugge sts that the matter is more
comple x: males and females are like ly to show both similar and different pat-
terns of relationships be tween and among the various face ts of narcissism.

REFERENCES

Akhtar, S., & Thompson, J. (1982) . Ove rview: Narcissistic Pe rsonality Disorder. Am erican Jour-
nal of Psychiatry, 139, 12-20.
Bentler, P. M., & Bonett, D. G. (1980) . Significance tests and goodness of fit in the analysis
of covariance structures. Psychological Bulletin, 56, 588-606.
Butler, D., & Ge is, F. L. (1990) . Nonverbal affect responses to male and fe male leaders: Im-
plications for leadership e valuations. Journal of Person ality and Social Psychology, 58, 48-
59.
Carli, L. L., Lafleur, S. J., & Loebe r, C. C. (1995) . Nonve rbal behavior, gender, and influence.
Journal of Personality and Social Psychology, 68, 1030-1041.
Carroll, L. (1987) . A study of narcissism, affiliation, intimacy,and power motives among stu-
dents in business administration. Psychological Reports, 61, 355-358.
Eagly, A. H., Makhijani, M. G., & Klonsky, B. G. (1992) . Ge nder and the e valuation of lead-
ers: A meta-analysis. Psychological Bulletin, 111, 3-22.
Emmons, R. A. (1984) . Factor analysis and construct validity of the Narcissistic Personality
Inventory. Journ al of Personality Assessm ent, 48, 291-300.
Emmons, R. A. (1987) . Narcissism: Theory and me asurement. Journal of Personality and Social
Psychology, 52, 11-17.
Haake n, J. (1983) . Sex differences in narcissistic disorders. Am erican Journal of Psychoanalysis,
43, 315-324.
J ö reskog, K. G., & S ö rbom, D. (1990) . SPSS LISREL 7 and PRELIS: User’s G uide and Ref-
eren ce. Chicago, IL: SPSS Inc.
J ö reskog, K. G., & S ö rbom, D. (1993) . SPSS LISREL 8: Users reference guide. Chicago, IL:
SPSS Inc.
McCann, J. T., & Biaggio, M. K.(1989). Narcissistic personality features and se lf-reported an-
ge r. Psychological R Reports, 64, 55-58.
Philipson, I.(1985) . Sex and narcissism. Psychology of Wom en Quarterly, 9, 213-228.
870 Tsch an z et al.

Raskin, R., & Hall, C. S. (1979) . A narcissistic personality inve ntory. Psychological Reports,
40, 590.
Raskin, R., & Te rry, H. (1988) . A principle components analysis of the Narcissistic Personality
Inventory and further e vidence for its construct validity. Journal of Personality and Social
Psychology, 54, 890-902.
Rhodewalt, F., & Morf, C. C. (1995) . Self and interpersonal correlates of the Narcissistic
Pe rsonality Inve ntory: A re view and new findings. Journ al of Research in Personality, 29,
1-23.
United States Bureau of the Census (1990). United States population estim ates: 1990. Wash-
ington, DC, U.S. Governmen t Printing Office.