The Relation between Personality and Prejudice: A Variable

European Journal of Personality
Eur. J. Pers. 17: 449–464 (2003)
Published online 19 October 2003 in Wiley InterScience (www.interscience.wiley.com). DOI: 10.1002/per.494
The Relation between Personality and Prejudice:
A Variable- and a Person-Centred Approach
BO EKEHAMMAR* and NAZAR AKRAMI
Department of Psychology, Uppsala University, Uppsala, Sweden
Abstract
The relationship between Big Five personality (measured by the NEO-PI) and prejudice
was examined using a variable- and a person-centred approach. Big Five scores were
related to a generalized prejudice factor based on seven different prejudice scales (racial
prejudice, sexism, etc). A correlation analysis disclosed that Openness to Experience and
Agreeableness were significantly related to prejudice, and a multiple regression analysis
showed that a variable-centred approach displayed a substantial cross-validated
relationship between the five personality factors and prejudice. A cluster analysis of the
Big Five profiles yielded, in line with previous research, three personality types, but this
person-centred approach showed a low cross-validated relationship between personality
and prejudice, where the overcontrolled type showed the highest prejudice and the
undercontrolled the lowest, with the resilient falling in between. A head-to-head
comparison sustained the conclusion that, based on people’s Big Five personalities, their
generalized prejudice could be predicted more accurately by the variable- than the personcentred approach. Copyright # 2003 John Wiley & Sons, Ltd.
INTRODUCTION
Several explanations have been suggested for why some people are more prejudiced than
others (see e.g. Allport, 1954; Duckitt, 1992, for reviews). There are roughly two major
lines of explanation (see e.g. Reynolds, Turner, Haslam, & Ryan, 2001): differences in
prejudice are due to differences in people’s personalities (personality psychology) or due
to differences in people’s group membership (social psychology). The social psychological approach, often employing experimental research methods, has probably been the
most influential in prejudice research during the last decades. Examples of the social
psychological proposition are theories and research based on social cognition paradigms
(see e.g. Fiske, 1998) or on social identity and self-categorization in the spirit of Tajfel (see
e.g. Tajfel & Turner, 1979; Turner & Onorato, 1999; Turner & Reynolds, 2001).
*Correspondence to: Bo Ekehammar, Department of Psychology, Uppsala University, Box 1225, SE-751 42
Uppsala, Sweden. E-mail: [email protected]
Contract/grant sponsor: Bank of Sweden Tercentenary Foundation; contract/grant number: 2000-0282.
Copyright # 2003 John Wiley & Sons, Ltd.
Received 18 September 2002
Accepted 8 April 2003
450
B. Ekehammar and N. Akrami
The personality approach, on the other hand, is often carried out using correctional
methods and is based on the contention that prejudice is not solely a function of the social
environment, social-group membership, or social identity, but rather a function of internal
attributes of the individual. Thus, stereotypical beliefs and prejudicial attitudes can be
explained by factors within the individual rather than by characteristics of the social
context. Probably the best known example of the personality approach is the classical
study on authoritarian personality theory (Adorno, Frenkel-Brunswik, Levinson, &
Sanford, 1950). Adorno et al. (1950) disclosed a broad ethnocentric pattern that involved
generalized prejudice towards various outgroups. This pattern could be explained by basic
factors within the individual’s personality, for example, conventionalism, authoritarian
submission and aggression, and power and toughness. A modern variant of this thinking
can be found in the theory of right-wing authoritarianism (Altemeyer, 1988, 1998). One
idea that can be taken as evidence for the personality explanation is the empirical findings
of Adorno et al. (1950) and Allport (1954), and more recently of Altemeyer (1998) and
McFarland (manuscript submitted for publication), that attitudes to various out-groups
seem to be highly correlated among people irrespective of their social background. Thus,
this generalized prejudice (cf. Allport, 1954) can be seen as deriving from one or more
basic personality traits.
Emphasizing the personality approach, one can conclude that quite a few recent
empirical studies (e.g. Altemeyer, 1998; Doty, Winter, Peterson, & Kemmelmeier, 1997;
Duriez & Van Hiel, 2002; Hillin & McFarland, 2002; Lippa & Arad, 1999; McFarland,
manuscript submitted for publication; McFarland & Mattern, 2002; Sidanius & Pratto,
1999; Van Hiel & Mervielde, 2002; Verkuyten & Hagendoorn, 1998; Whitley, 1999;
Whitley & Ægisdottir, 2000; Whitley & Lee, 2000) have examined the relationship
between certain individual difference or personality variables on the one hand and one
or several types of prejudice, or political ideology (often related to prejudice), on the
other. Two individual difference variables have been examined in all these studies: RightWing Authoritarianism (RWA; see e.g. Altemeyer, 1988) and/or Social Dominance
Orientation (SDO; see e.g. Pratto, Sidanius, Stallworth, & Malle, 1994; Sidanius & Pratto,
1999).
RWA, according to Altemeyer (1981), is a construct that can be seen as composed of
conventionalism, authoritarian submission, and authoritarian aggression, which means
that high-RWA people tend to favour traditional values, are submissive to authority figures,
and can be expected to act aggressively toward outgroups. SDO, according to Pratto et al.
(1994), can be seen as ‘a general attitudinal orientation toward intergroup relations,
reflecting whether one generally prefers such relations to be equal, versus hierarchical’
(p. 742). Thus, high-SDO people tend to rank social groups in a superior–inferior
hierarchy. The main outcome in the empirical research referred to above is that RWA and
SDO have been shown to be two powerful predictors of prejudice.
The definitions of RWA and SDO imply that these individual difference variables are to
a great extent focused on intra- and intergroup perceptions and attitudes and thus they
could rather be placed in the interface between personality and social psychology than in
the centre of personality psychology. Thus, RWA and SDO are narrowly defined traits that
are conceptually rather close to the prejudice construct. Consequently, high correlations of
RWA and SDO with prejudice are to be expected. On the other hand, broad and general
personality dimensions, such as the Big Five personality factors, are higher-order traits
that are conceptually distant from prejudice. What, then, is the relation of the Big Five with
prejudice? Not many empirical studies, if any, have had the examination of this
Copyright # 2003 John Wiley & Sons, Ltd.
Eur. J. Pers. 17: 449–464 (2003)
Personality and prejudice
451
relationship as their primary focus. However, recently, Heaven and Bucci (2001) presented
a detailed analysis of the relations of RWA and SDO on the one hand and the Big Five
personality factors and facets on the other. They found that RWA and SDO were aligned
with different Big Five personality traits. Further, Peterson, Smirles, and Wentworth
(1997) examined the relation between the Big Five personality factors and authoritarianism and found that authoritarianism correlated negatively with Openness to Experience but
was unrelated to other Big Five factors. However, measures of prejudice were not included
in any of these two studies. Similarly, Van Hiel, Kossowska, and Mervielde (2000)
analysed the relation of one Big Five factor (Openness to Experience) and its facets with
political ideology, but they did not include any measures of prejudice either.
To the best of our knowledge, only two studies (McFarland, manuscript submitted for
publication; Saucier & Goldberg, 1998) have reported data on the relationship between
Big Five personality factors and prejudice, although these studies did not have the BigFive-personality versus prejudice relationship as their major concern. Saucier and
Goldberg (1998) analysed, among other things, self-ratings on prejudice-related adjectives
(clustered as Prejudiced–Bigoted) and found that this cluster was only weakly related to
the Big Five personality factors, with significant but low correlations for Agreeableness
(negative relation) and Neuroticism (positive relation) only. McFarland (manuscript
submitted for publication, Study 4) analysed the contribution of the Big Five factors,
together with RWA, SDO, and empathy, when predicting prejudice. He found that
generalized prejudice (a composite measure based on scores from various types of
prejudice) correlated significantly with four of the Big Five factors: Openness to
Experience (negative relation), Agreeableness (negative relation), Conscientiousness
(negative relation), and Neuroticism (positive relation). Thus, although not fully
consistent, these results indicate that there are relationships between at least some Big
Five personality factors and some aspects of prejudice.
Against the background outlined above, the main aim of the present study was to make a
close examination of the relation between the Big Five personality factors and prejudice.
This examination was carried out in two ways, one using a variable-centred and one a
person-centred approach.
The traditional variable-centred approach, as used by McFarland (manuscript submitted
for publication) and Saucier and Goldberg (1998), for example, means that the simple
correlations between personality and prejudice variables are examined, and that a linear
multiple regression equation is computed to study the overall predictive power and the
significant contributions of the Big Five factors in the prediction of prejudice, assuming
that personality causally precedes prejudice.
The (untraditional) person-centred approach has recently attracted a lot of interest in,
especially Big Five, personality research, and especially in Europe (see Asendorpf, 2003;
Asendorpf, Borkenau, Ostendorf, & van Aken, 2001; Asendorpf & van Aken, 1999;
Barbaranelli, 2002; Boehm, Asendorpf, & Avia, 2002; Caspi, 1998; De Fruyt, Mervielde,
& Van Leeuwen, 2002; Mervielde & Asendorpf, 2000; Pulkkinen, 1996; Rammstedt,
Riemann, Angleitner, & Borkenau, in press; Schnabel, Asendorpf, & Ostendorf, 2002). A
common person-centred methodology, also applied in the present study, is to carry out
cluster analyses to form homogeneous subgroups of people with similar profiles across
variables, in this case the Big Five personality factors, and then relate cluster membership
to an external variable, in this case prejudice. To the best of our knowledge, no previous
research has examined the relation between personality and prejudice using a personcentred approach.
Copyright # 2003 John Wiley & Sons, Ltd.
Eur. J. Pers. 17: 449–464 (2003)
452
B. Ekehammar and N. Akrami
In addition, we compared the predictive power of the variable- versus the person-centred
approach and, in line with Costa, Herbst, McCrae, Samuels, and Ozer (2002), we
performed a head-to-head comparison of the obtained personality cluster types versus the
Big Five personality factors in predicting prejudice.
METHOD
Participants
The sample comprised 156 non-psychology students, 77 women and 79 men, aged between
18 and 57 years (M ¼ 23.8 years), at Uppsala University and at the local-authorityadministered adult education. They received cinema vouchers for their participation. The
students represented various academic disciplines, such as social science, behavioral science,
medicine, economics, technology, and dentistry.
Personality instrument and variables
NEO-PI
An official Swedish translation of the NEO-PI (Costa & McCrae, 1985) was used. The
personality factors, together with the number of items (n) and the Cronbach alpha
reliabilites () in the present sample, were Neuroticism, n ¼ 48, ¼ 0.93; Extraversion,
n ¼ 48, ¼ 0.89; Openness to Experience, n ¼ 48, ¼ 0.86; Agreeableness, n ¼ 18,
¼ 0.88; and Conscientiousness, n ¼ 18, ¼ 0.74. All scale reliabilities for the present
sample were regarded as satisfactory.
Prejudice instruments and variables
The Modern and Classical Racial Prejudice Scales
The Modern Racial Prejudice scale was constructed for measuring modern (covert, subtle,
symbolic) racial/ethnic prejudice in a Scandinavian context by Akrami, Ekehammar, and
Araya (2000) following the item contents of McConahay’s (1986) Modern Racism Scale
and based on Sears’ (1988) classification system. Sears characterized modern (symbolic)
prejudice by three components: denial of continued discrimination, antagonism toward
minority group demands, and resentment about special favours for minority groups.
Reliability and validity data for the scale are provided by Akrami et al. (2000). The scale
contains nine items (example: Discrimination against immigrants is no longer a problem
in Sweden).
The Classical Racial Prejudice scale has the same origin as that above (see Akrami et al.,
2000 for reliability and validity information) and was constructed to measure classical
(blatant, ‘old-fashioned’) racial/ethnic prejudice in a Scandinavian context and in line with
McConahay’s (1986) ‘old-fashioned’ racism items. The scale includes eight items
(example: Immigrants do not keep their homes tidy).
The Swedish Modern and Classical Sexism Scales
The Swedish Modern Sexism scale was constructed for measuring attitudes toward women
in a Scandinavian context by Ekehammar, Akrami, and Araya (2000). It was based on
Sears’ (1988) classification system and the Modern Sexism Scale items by Swim, Aikin,
Hall, and Hunter (1995). For reliability and validity information, see Ekehammar et al.
Copyright # 2003 John Wiley & Sons, Ltd.
Eur. J. Pers. 17: 449–464 (2003)
Personality and prejudice
453
(2000). The scale comprises eight items (example: Discrimination against women is no
longer a problem in Sweden).
The Classical Sexism Scale has the same origin as that above (see Ekehammar et al.,
2000, for reliability and validity data) and was constructed for measuring classical
(blatant, ‘old-fashioned’) sexism in a Scandinavian context. The scale contains seven
items (example: Women are generally not very talented).
The Modern and Classical Attitudes toward Intellectually Disabled Individuals Scales
The modern scale was developed by Akrami, Ekehammar, Sonnander, and Claesson
(manuscript submitted for publication) to measure attitudes and prejudice toward mentally
disabled persons, taking into account the distinction between classical (blatant, ‘oldfashioned’) and modern prejudice, based on Sears’ (1988) classification system. The scale
includes 11 items (example: Mentally disabled persons are getting too demanding in their
push for equal rights).
The classical version of the scale has the same origin as that above (Akrami et al.,
manuscript submitted for publication) and was developed to measure the classical (blatant,
‘old-fashioned’) aspect of prejudice toward mentally disabled people. The scale contains
eight items (example: Mentally disabled people often commit crimes).
The Attitude to Homosexuality Scale
The scale was constructed for a Scandinavian context by Ekehammar and Akrami
(manuscript in preparation) to measure attitudes towards homosexuals in general and
separate attitudes towards lesbians and gay men. The total score (general homophobia) can
be seen as an index of general attitude to homosexuality. The scale contains 28 items
(example: Homosexuality should be banned).
Procedure
All participants were tested individually on a computer. All instructions and item wordings
were presented on the computer screen. The participant’s responses were automatically
stored in a file. All items were randomly mixed within the domains of personality and
prejudice, respectively. The item order was randomized individually. Appropriate items
were reversed when coded. The items of all instruments were answered on five-step Likerttype scales ranging from strongly disagree (1) to strongly agree (5). For all scales, the
average score across items were computed for all participants. For the prejudice domain,
higher scores indicate more negative attitudes, that is, a higher (negative) prejudice.
RESULTS
Analyses of prejudice scale scores
Before the main analyses, the participants’ scores on the seven prejudice scales presented
above were analysed by computing pair-wise correlations (r) and Cronbach alpha
coefficients. As shown in Table 1, the scale reliabilities can be regarded as satisfactory,
with exception of that for the Classical Attitudes Towards Disabled Persons Scale. Further,
all scale intercorrelations are in general high (Mr ¼ 0.36), and highly statistically
significant ( p < 0.01) as well. The largest correlation coefficient (r ¼ 0.68) was obtained
for the relation between modern racial prejudice and modern sexism, and the smallest
(r ¼ 0.20) for the relation between attitudes to homosexuality and classical racial
prejudice.
Copyright # 2003 John Wiley & Sons, Ltd.
Eur. J. Pers. 17: 449–464 (2003)
454
B. Ekehammar and N. Akrami
Table 1. Correlations among the prejudice scale scores and Cronbach alpha reliabilities for each
scale in the main diagonal (in boldface)
Scalea
1.
2.
3.
4.
5.
6.
7.
Homosexuality
Modern racism
Classical racism
Modern sexism
Classical sexism
Modern disabled
Classical disabled
1
2
3
4
5
6
7
0.91
0.35
0.20
0.42
0.40
0.24
0.38
0.85
0.58
0.68
0.50
0.58
0.43
0.78
0.39
0.55
0.40
0.43
0.82
0.52
0.56
0.37
0.76
0.45
0.37
0.73
0.43
0.68
All correlation are significant at p < 0.01 (N ¼ 156).
a
For the complete names and general contents of the scales, see the Method section.
Table 2. Factor loadings on the generalized
prejudice factor
Scalea
Homosexuality
Modern racism
Classical racism
Modern sexism
Classical sexism
Modern disabled
Classical disabled
Factor loading
0.56
0.83
0.71
0.79
0.76
0.73
0.66
a
For the complete names and general contents of the
scales, see the Method section.
To arrive at a smaller number of general prejudice factors, we performed a principal
components analysis on the correlation matrix. There was only one factor with an
eigenvalue greater than unity (4.24), explaining 53% of the total variance, and a scree plot
indicated one factor as well. Following Allport’s (1954) terminology, we denoted this
factor generalized prejudice. As shown in Table 2, the loadings on this factor are high,
varying between 0.56 and 0.83, and the two largest loadings were obtained for modern
racial prejudice (0.83) and modern sexism (0.79), and the two smallest for attitudes to
homosexuality (0.56) and classical attitudes toward disabled persons (0.66). Finally,
standardized factor scores (M ¼ 0, SD ¼ 1) for the generalized prejudice factor were
calculated for each participant, with higher scores indicating higher generalized prejudice.
These scores were used as the dependent variable in the main analyses below.
The relation between Big Five personality and prejudice
The variable-centred approach
To make possible a traditional cross-validation, from here on we split the sample randomly
into two gender-matched halves (sample 1, N ¼ 78; sample 2, N ¼ 78). As shown in
Table 3, the correlations (r) of the Big Five personality factor scores with the generalized
prejudice factor score were very similar in samples 1 and 2. In both samples, two of the five
personality factors showed substantial (0.44 to 0.45), and highly significant, negative
Copyright # 2003 John Wiley & Sons, Ltd.
Eur. J. Pers. 17: 449–464 (2003)
Personality and prejudice
455
Table 3. Correlations of Big Five personality factors with
generalized prejudice
Big Five scale
Neuroticism
Extraversion
Openness to Experience
Agreeableness
Conscientiousness
Sample 1
Sample 2
0.10
0.20
0.45*
0.45*
0.09
0.09
0.16
0.44*
0.45*
0.01
*p < 0.001.
correlations with generalized prejudice: Openness to Experience and Agreeableness. A
similar analysis between the Big Five factors and each of the seven prejudice scales
revealed the same pattern of results. In both samples, Openness to Experience (r varying
between 0.28 and 0.43) and Agreeableness (r varying between 0.21 and 0.45) were
consistently and significantly ( p at least 0.01) negatively correlated with each of the
prejudice scales. There were no systematic differences in correctional pattern, either
between the two samples or between the Big Five and prejudice scales. Thus, it seems that
the more open and agreeable the person is, the less generalized and specific prejudice does
he or she display.
Raw-score linear regression equations for predicting generalized prejudice from the Big
Five personality factors were computed for sample 1 and sample 2 separately (see Table 4).
The prediction equations are very similar, and in both cases only Openness to Experience
and Agreeableness give statistically significant contributions to the prediction. The
predictive power (R) before cross-validation was 0.60 and 0.58 ( p < 0.001) in samples 1
and 2, respectively. The regression equation obtained on sample 1 was applied on sample 2
for cross-validation. The correlation (R) between obtained and predicted generalized
prejudice scores was found to be R ¼ 0.553 (R2 ¼ 0.306), p < 0.001. Further, the regression
equation obtained on sample 2 was applied on sample 1 for cross-validation. The
correlation (R) between obtained and predicted generalized prejudice scores was found to
be R ¼ 0.563 (R2 ¼ 0.317), p < 0.001. Thus, the cross-validated relationship (multiple
correlation) between the Big Five personality factors and generalized prejudice using a
variable-centred approach was found to be R ¼ 0.558 (R2 ¼ 0.312), the mean of the two
cross-validated values presented above.
Table 4. Raw-score multiple regression equations in
sample 1 and sample 2 for predicting generalized prejudice
from the Big Five factors
Regression coefficient (b)
Big Five scale
Neuroticism
Extraversion
Openness to Experience
Agreeableness
Conscientiousness
Sample 1
Sample 2
0.14
0.02
0.32*
0.42*
0.09
0.10
0.08
0.35*
0.41*
0.10
The intercept in the regression equations was 6.80 in sample 1 and
7.43 in sample 2.
*p < 0.001.
Copyright # 2003 John Wiley & Sons, Ltd.
Eur. J. Pers. 17: 449–464 (2003)
456
B. Ekehammar and N. Akrami
The person-centred approach
We conducted a cluster analysis of the Big Five personality profiles on sample 1 and
sample 2 separately. In accord with suggestions in statistical handbooks (e.g. Hair,
Anderson, Tatham, & Black, 1998) and with previous empirical studies (e.g. Asendorpf
et al., 2001; Barbaranelli, 2002; Boehm et al., 2002; Costa et al., 2002; Schnabel et al.,
2002), we (i) applied Ward’s hierarchical clustering procedure on the individual Big Five
personality scale scores on the basis of the squared Euclidean distances (SEDs) between
profiles, and (ii) used the resulting two-, three-, and four-cluster solutions from Step 1 as
initial values in a non-hierarchical K-means cluster analysis. To test the congruence of the
cluster solutions between sample 1 and sample 2, we calculated SEDs between scores on
corresponding Big Five factors in the different solutions. All solutions (two, three, and four
clusters) showed satisfactory congruence between corresponding clusters in sample 1 and
sample 2. The mean SEDs were 0.005 for the two-cluster, 0.083 for the three-cluster, and
0.089 for the four-cluster solution.
Focusing on the three-cluster solution, which has been shown to be the most meaningful
and replicable in previous research (e.g. Asendorpf et al., 2001; Asendorpf & van Aken,
1999; Boehm et al., 2002; Rammstedt et al., in press; Schnabel et al., 2002), we present the
Big Five profile means (expressed as z-scores for comparison) for this solution only. As
shown in Figure 1, and in accord with the research referred to, the three personality types
could be labelled Resilient (48% of the sample), Overcontrolled (24% of the sample), and
Undercontrolled (28% of the sample), respectively.
To study the relation between Big Five personality type and generalized prejudice we
performed multiple regression analyses with dummy-coding of the two-, three-, and fourcluster solutions. In parallel with the variable-centred approach above, linear regression
equations for predicting generalized prejudice from the Big Five personality types were
computed for sample 1 and sample 2 separately (see Table 5). Thus, the regression
equation based on sample 1 was applied on sample 2 for cross-validation, and vice versa.
As shown in Table 5, the cross-validated relations between personality type and
generalized prejudice are very low for all cluster solutions but in any case higher for the
three-cluster (R ¼ 0.18) than for the two-cluster (R ¼ 0.07) and four-cluster (R ¼ 0.09)
solutions. Also, the relation between the three-cluster solution and generalized prejudice
was marginally significant before cross-validation (R ¼ 0.20, p ¼ 0.053), and marginally
significant ( p ¼ 0.07) in one of the cross-validations (see Table 5) as well. Thus, one may
conclude that the three-personality-type solution displayed the highest relationship with
generalized prejudice. What, then, does this relation look like? The answer is given in
Figure 2, where the generalized prejudice factor-score means are depicted for each of the
three personality types. As shown there, the overcontrolled personality type expressed
the highest degree of generalized prejudice and the undercontrolled the lowest, with the
resilient personality type falling in between. A Newman–Keuls post hoc test showed only
one significant difference between the types, that between the overcontrolled and the
undercontrolled ( p ¼ 0.02).
The person- versus the variable-centred approach when predicting prejudice
As shown above, the variable-centred approach based on the Big Five personality factors
linearly combined revealed a highly significant ( p < 0.001) cross-validated relationship
with generalized prejudice, expressed as a multiple correlation coefficient of 0.558 (R2 ¼
0.312). In contrast, the person-centred approach showed a cross-validated relationship,
Copyright # 2003 John Wiley & Sons, Ltd.
Eur. J. Pers. 17: 449–464 (2003)
Figure 1. Personality types based on the Big Five factors for the three-cluster solution (N ¼ Neuroticism, E ¼ Extraversion, O ¼ Openness to Experience, A ¼ Agreeableness,
C ¼ Conscientiousness).
Personality and prejudice
Copyright # 2003 John Wiley & Sons, Ltd.
457
Eur. J. Pers. 17: 449–464 (2003)
458
B. Ekehammar and N. Akrami
Table 5. Cross-validated multiple correlations (R ¼ eta) for the relation between cluster membership (dummy coded) and generalized prejudice
Number of clusters
2
3
4
Sample 1a
Sample 2b
Mean cross-validated R (R2)
0.090 ( p ¼ 0.43)
0.153 ( p ¼ 0.18)
0.149 ( p ¼ 0.19)
0.055 ( p ¼ 0.63)
0.208 ( p ¼ 0.07)
0.031 ( p ¼ 0.79)
0.073 (0.005)
0.181 (0.033)
0.090 (0.008)
a
Based on the regression equation for sample 2.
Based on the regression equation for sample 1.
b
expressed as a multiple correlation coefficient (R ¼ eta), with generalized prejudice of
only 0.073 (R2 ¼ 0.005), 0.181 (R2 ¼ 0.033), and 0.090 (R2 ¼ 0.008), for the two-, three-,
and four-cluster solutions, respectively. Thus, the variable-centred approach appears to be
clearly more powerful than the person-centred approach when predicting people’s
generalized prejudice from their Big Five personalities. However, it has been argued
(Asendorpf, 2003) that this comparison is unfair as the number of degrees of freedom is
larger in the variable-centred (df ¼ 4) than in the person-centred case (df ¼ 2). We argue,
on the other hand, that the comparison is not unfair as the same five factors form the basis
for prediction in both cases but in the person-centred case only three personality types have
been found, in this and previous studies, to be meaningful whereas in the variable-centred
case, all five personality factors are regarded as meaningful. If one does not agree with this
argument, one solution (see Ozer & Howell, 2002) could be to adjust the predictive power
with the number of degrees of freedom (R/df). For the present data, this would give an
adjusted R of 0.140 in the variable-centred case and 0.091 in the person-centred case.
Thus, although we regard this adjustment as highly questionable, it shows in any case a
predictive advantage of the variable-centred over the person-centred approach in the
present case. Evidently, the comparison between the variable- and the person-centred
approach could have been carried out in other ways, for example, by including the
interaction aspect as products among Big Five factors in the regression equation under the
Figure 2. Mean generalized prejudice scores for the three Big Five personality types in the three-cluster
solution.
Copyright # 2003 John Wiley & Sons, Ltd.
Eur. J. Pers. 17: 449–464 (2003)
Personality and prejudice
459
Table 6. Regression summaries of model 1 (step 1: the three Big Five personality types; step 2: the
Big Five personality traits added) and model 2 (step 1: the Big Five personality traits; step 2: the three
Big Five personality types added) using generalized prejudice as dependent variable (total sample,
N ¼ 156, without cross-validation)
Model 1
2
R
R2 adjusted
Model 2
Step 1 (types)
Step 2 (traits)
Step 1 (traits)
0.038 ( p ¼ 0.053)
0.025
0.311 ( p ¼ 0.000)
0.292
0.334 ( p ¼ 0.000)
0.312
Step 2 (types)
0.014 ( p ¼ 0.332)
0.006
Note: Cluster types are dummy coded.
variable-centred approach. However, this would hardly change the main conclusion of our
comparison and, further, our procedure follows closely that presented by, for example,
Costa et al. (2002).
The conclusion above is also sustained in our final head-to-head comparison where we,
in line with Costa et al. (2002), conducted multiple regression analyses on the total sample
according to two models: (1) on the first step, entering the three personality types (dummy
coded), and in the second, the five personality factors, and (2) on the first step, entering the
five personality factors, and in the second, the three personality types (dummy coded). As
can be seen in Table 6, entering the personality types first results in a marginally significant
and low amount of explained variance (3.8%), whereas the addition of the personality
factors increases the amount of explained variance substantially (31.1%) and with high
statistical significance. When entering the personality factors first, these are shown to give
a highly significant variance contribution (33.4%), whereas the personality types give a
negligible (1.4%) and statistically non-significant contribution to the explained variance in
generalized prejudice.
DISCUSSION
To sum up, our study has shown that various types of prejudice could be reduced to a
generalized prejudice factor, and, using a variable-centred approach, this factor was shown
to be significantly and substantially related to the Big Five personality factors Openness to
Experience and Agreeableness. A person-centred approach based on cluster analysis of the
Big Five personality profiles disclosed three personality types that were related to
generalized prejudice. However, this relation was not strong and showed a marginal
statistical significance, with the overcontrolled as the most prejudiced personality type and
the undercontrolled as the least, with the resilient personality profile falling in between.
This outcome is perhaps not surprising taking into consideration that the Big Five
personality types do not differ substantially in Openness to Experience and Agreeableness,
which were shown to be the single two best predictors of generalized prejudice in the
variable-centred approach. A multiple regression analysis with prejudice as the dependent
variable showed that the variable-centred approach displayed a substantial and highly
significant cross-validated relationship between the Big Five personality factors and
prejudice, whereas the person-centred approach showed a low and marginally significant
cross-validated relationship between personality type and generalized prejudice. A major
conclusion is that there appears to be a substantial relation between Big Five personality
Copyright # 2003 John Wiley & Sons, Ltd.
Eur. J. Pers. 17: 449–464 (2003)
460
B. Ekehammar and N. Akrami
and prejudice, at least when examining the relationship using a variable-centred approach.
Evidently, our results need to be replicated, especially taking into account the relatively
small sample size, which was, however, compensated for by the double cross-validation
procedure and the similar outcome in the two sub-samples. A further extension of our
study would be to analyse the Big Five factors on a facet level rather than the higher order
level we preferred in the present case.
Interestingly, the analyses of the various prejudice scales disclosed that these were
strongly correlated, and through factor analysis it was shown that they formed one factor
only. In accord with Allport (1954; see also Altemeyer, 1988; McFarland, manuscript
submitted for publication), we labelled this factor generalized prejudice. Thus, in spite of
the fact that our prejudice instruments were based on either a classical or a modern
definition of prejudice (see the Method section) and that they covered various types of
prejudice (racial prejudice, sexism, homophobia, and prejudice toward disabled people),
they could all be reduced to one and the same general factor. In fact, this finding was
discovered by Hartley (1946), and emphasized by Adorno et al. and Allport during the
1950s, and since then by some other researchers as well (e.g. Duckitt, 1992; McFarland,
manuscript submitted for publication). Further, and important in the present context,
Allport’s conclusion from the findings of highly correlated prejudice scales was that
prejudice can be seen not only as a generalized attitude but also as a personality trait: ‘The
evidence we have reviewed constitutes a very strong argument for saying that prejudice is
basically a trait of personality. When it takes root in life it grows like a unit. The specific
object of prejudice is more or less immaterial’ (Allport, 1979, p. 73).
The variable-centred approach, based on correlation and multiple regression analyses,
showed that two of the Big Five factors (Openness to Experience and Agreeableness)
displayed rather strong (r ¼ 0.44 to 0.45) negative zero-order relationships with
generalized prejudice. Also, McFarland (manuscript submitted for publication) found that
Openness to Experience (r ¼ 0.38) and Agreeableness (r ¼ 0.33) displayed the largest
correlations with his generalized prejudice measure, and he also found that Conscientiousness and Neuroticism were significantly related to generalized prejudice, but these
correlation coefficients were smaller than those above. Saucier and Goldberg (1998) used
quite another approach to the measurement of prejudice, and they obtained only small
correlations with this measure and the Big Five traits, two of which (Agreeableness,
r ¼ 0.23; Neuroticism, r ¼ 0.25) were statistically significant, however. In conclusion,
when studying simple correlations, it seems as if Agreeableness and Openness to
experience are those single Big Five factors most closely associated with generalized
prejudice. This is not unexpected. Openness to Experience, for example, includes
components that have to do with nonconformity and unconventionality (cf. John &
Srivastava, 1999), which is also the interpretation of this factor in some European samples
(De Raad, Perugini, Hrebickova, & Szarota, 1998). Further, Openness to Experience has
been shown to be inversely related to authoritarianism (McCrae & Costa, 1997; Peterson
et al., 1997; Trapnell, 1994) and positively related to liberal social and political values
(McCrae, 1996; Van Hiel et al., 2000). All these characteristics of Openness to Experience
would predict a negative relation to prejudice. Also, Agreeableness, as the opposite of
antagonism, includes components such as tender-mindedness and altruism (cf. John &
Srivastava, 1999) as well as non-hostility, empathy, and prosocial behavior (cf. Graziano &
Eisenberg, 1997), which could be expected to relate negatively to prejudice as well.
When using all Big Five factors for predicting generalized prejudice in a linear multiple
regression analysis (MRA), we arrived at a fairly high predictive power (cross-validated
Copyright # 2003 John Wiley & Sons, Ltd.
Eur. J. Pers. 17: 449–464 (2003)
Personality and prejudice
461
R ¼ 0.56) with Agreeableness and Openness to experience showing, as expected, the
largest regression weights. McFarland (manuscript submitted for publication), examining
primarily the predictive power of RWA, SDO, and empathy versus generalized prejudice,
made a similar analysis in his Study 4, where he included the Big Five factors in a first
MRA step as well. Like our study, only Openness and Agreeableness gave significant
contributions in the regression equation and the predictive power obtained in his study
(R ¼ 0.52, not cross-validated) was similar to ours. When adding RWA, SDO, and empathy
(the Big Three predictors of prejudice, according to McFarland (manuscript submitted for
publication)), the predictive power was substantially improved (from 27 to 61%).
However, we argue that a fairer treatment of the Big Five factors in this context would be
to carry out a causal modelling analysis rather than a strict prediction study. It seems clear
from previous research (e.g. Altemeyer, 1998; McFarland, manuscript submitted for
publication; Whitley, 1999) that, in a prediction study, RWA and SDO would most
probably come out as the single strongest predictors of prejudice with a combined
predictive power of around 40%. However, as shown by Heaven and Bucci (2001),
RWA and SDO are in their turn, and in different ways, associated with the Big Five traits.
As RWA and SDO were not included in the present study, we suggest that future research
in this area would investigate this issue applying a causal modelling approach including
these two variables together with the Big Five and prejudice factors. In this way, in
addition to the direct effects, also the indirect effects of the Big Five factors on prejudice,
through RWA and SDO, would have a chance to be examined and estimated. Of course,
other models concerning causal order could be investigated and tested in such an approach.
In that context one could also include measures of social desirability in order to test
whether the relationship between personality and prejudice disclosed in the present study
could have been inflated by social desirability aspects. Thus, social desirability has been
found to be correlated with at least the Big Five factor Agreeableness (see e.g. Stöber,
2001).
Our person-centred analysis of the personality–prejudice relationship showed, in the
first hand, that the only personality typology that had any notable relation with prejudice
was that comprising three types. Further, these three personality types could, in accord
with previous research, be labelled resilient, overcontrolled, and undercontrolled. The
undercontrolled were shown to be the least prejudiced, which is no surprise when
considering their relatively high levels of Agreeableness and Openness to Experience,
whereas the overcontrolled, with their relatively low levels of Agreeableness and
Openness, consequently were shown to be the most prejudiced. However, a prediction of
the relationship between personality type and prejudice based on the original definitions
(Block & Block, 1980) of ego-control and ego-resiliency would seem to be more difficult.
According to Block and Block (1980), ego-resiliency refers to the tendency to responding
in a flexible rather than a rigid way to external demands, whereas ego-control has to do
with the tendency to withhold (high ego-control ¼ overcontrolled people) or express (low
ego-control ¼ undercontrolled people) motivational and emotional impulses. From these
general definitions, one might have predicted that the overcontrolled personality type
would be the less prejudiced because of their control of expressing negative explicit
prejudice whereas the undercontrolled type could be expected to be the most prejudiced
because of their lack of this control. However, our empirical results showed quite the
reverse outcome. Finally, we could not find any previous study using the person-centred
approach for studying the personality–prejudice relationship, so comparison with previous
research is not possible on this point.
Copyright # 2003 John Wiley & Sons, Ltd.
Eur. J. Pers. 17: 449–464 (2003)
462
B. Ekehammar and N. Akrami
When comparing the utility of the variable- versus person-centred approach for
predicting people’s generalized prejudice from their Big Five personalities, our results
were quite conclusive. The predictive power was markedly higher when using the variablecentred (cross-validated R ¼ 0.56) than the person-centred approach (cross-validated
R ¼ 0.18). Further, a head-to-head comparison of the relative utility of types and traits in
line with that presented by Costa et al. (2002) showed that when Big Five types were
entered on the first step in a multiple regression analysis, the predictive power versus
generalized prejudice was low and marginally significant. However, when the Big Five
traits were entered on the second step, there was a marked and significant increase in
predictive power. When the roles were reversed, the Big Five types contributed nothing
above and beyond what was already predicted by the traits. In fact, the power of the
person-centred approach when predicting generalized prejudice was even lower than those
figures presented by Costa et al. (2002) for predicting various personality disorders and
other behaviours from the three Big Five types. Thus, when predicting people’s prejudice,
like various other behaviours, from their Big Five personalities, it seems as if the variablecentred approach is superior to the person-centred one.
ACKNOWLEDGEMENTS
This research was supported by Grant No. 2000-0282 from the Bank of Sweden
Tercentenary Foundation to Bo Ekehammar. A preliminary version of this paper was
presented at the 11th European Conference on Personality (July 2002), Jena.
REFERENCES
Adorno, T. W., Frenkel-Brunswik, E., Levinson, D. J., & Sanford, R. N. (1950). The authoritarian
personality. New York: Norton.
Akrami, N., Ekehammar, B., & Araya, T. (2000). Classical and modern racial prejudice: A study of
attitudes toward immigrants in Sweden. European Journal of Social Psychology, 30, 521–532.
Allport, G. W. (1954). The nature of prejudice. Reading, MA: Addison-Wesley.
Allport, G. W. (1979). The nature of prejudice (25th anniversary ed.). Cambridge, MA: Perseus.
(Original work published 1954.)
Altemeyer, B. (1981). Right-wing authoritarianism. Manitoba: University Press.
Altemeyer, B. (1988). Enemies of freedom: Understanding right-wing authoritarianism. San
Francisco: Jossey-Bass.
Altemeyer, B. (1998). The other ‘authoritarian personality.’ In L. Berkowitz (Ed.), Advances in
experimental social psychology (Vol. 30, pp. 47–92). Orlando, FL: Academic.
Asendorpf, J. B. (2003). Head-to-head comparison of the predictive validity of personality types and
personality dimensions. European Journal of Personality, 17, in press.
Asendorpf, J. B., Borkenau, P., Ostendorf, F., & van Aken, M. A. G. (2001). Carving personality
description at its joints: Confirmation of three replicable personality prototypes for both children
and adults. European Journal of Personality, 15, 169–198.
Asendorpf, J. B., & van Aken, M. A. G. (1999). Resilient, overcontrolled, and undercontrolled
personality prototypes in childhood: Replicability, predictive power, and the trait-type issue.
Journal of Personality and Social Psychology, 77, 815–832.
Barbaranelli, C. (2002). Evaluating cluster analysis solutions: An application of the Italian NEO
Personality Inventory. European Journal of Personality, 16, S43–S56.
Block, J. H., & Block, J. (1980). The role of ego-control and ego-resiliency in the organization of
behavior. In W. A. Collins (Ed.), Minnesota symposium on child psychology (Vol. 13, pp. 39–101).
Hillsdale, NJ: Erlbaum.
Copyright # 2003 John Wiley & Sons, Ltd.
Eur. J. Pers. 17: 449–464 (2003)
Personality and prejudice
463
Boehm, B., Asendorpf, J. B., & Avia, M. D. (2002). Replicable types and subtypes of personality:
Spanish NEO-PI samples. European Journal of Personality, 16, S25–S42.
Caspi, A. (1998). Personality development across the life course. In W. Damon (Ed.) and
N. Eisenberg (Vol. Ed.), Handbook of child psychology: Social, emotional, and personality
development (5th ed., Vol. 3, pp. 311–388). New York: Wiley.
Costa, P. T., Jr., Herbst, J. H., McCrae, R. R., Samuels, J., & Ozer, D. J. (2002). The replicability and
utility of three personality types. European Journal of Personality, 16, S73–S88.
Costa, P. T., Jr., & McCrae, R. R. (1985). The NEO Personality Inventory manual. Odessa, FL:
Psychological Assessment Resources.
De Fruyt, F., Mervielde, I., & Van Leeuwen, K. (2002). The consistency of personality type
classification across samples and five-factor measures. European Journal of Personality, 16,
S57–S72.
De Raad, B., Perugini, M., Hrebickova, M., & Szarota, P. (1998). Lingua franca of personality:
Taxonomies and structures based on the psycholexical approach. Journal of Cross-Cultural
Psychology, 29, 212–232.
Doty, R. G., Winter, D. G., Peterson, B. E., & Kemmelmeier, M. (1997). Authoritarianism and
American students’ attitudes about the Gulf War, 1990–1996. Personality and Social Psychology
Bulletin, 23, 1133–1143.
Duckitt, J. (1992). The social psychology of prejudice. New York: Praeger.
Duriez, B., & Van Hiel, A. (2002). The march of modern fascism: A comparison of social dominance
orientation and authoritarianism. Personality and Individual Differences, 32, 1199–1213.
Ekehammar, B., Akrami, N., & Araya, T. (2000). Development and validation of Swedish classical
and modern sexism scales. Scandinavian Journal of Psychology, 41, 307–314.
Fiske, S. T. (1998). Stereotyping, prejudice, and discrimination. In D. T. Gilbert, S. T. Fiske, & G.
Lindzey (Eds.), Handbook of social psychology (4th ed., pp. 357–411). New York: McGraw-Hill.
Graziano, W. G., & Eisenberg, N. H. (1997). Agreeableness: A dimension of personality. In
R. Hogan, J. A. Johnson, & S. R. Briggs (Eds.), Handbook of personality psychology (pp. 795–
824). San Diego, CA: Academic.
Hair, J. F., Jr., Anderson, R. E., Tatham, R. L., & Black, W. C. (1998). Multivariate data analysis
(5th ed.). Upper Saddle River, NJ: Prentice-Hall.
Hartley, E. L. (1946). Problems in prejudice. New York: Kings Crown.
Heaven, P. C. L., & Bucci, S. (2001). Right-wing authoritarianism, social dominance orientation and
personality: An analysis using the IPIP measure. European Journal of Personality, 15, 49–56.
Hillin, S. J., & McFarland, S. G. (2002, January). Authoritarianism, social dominance orientation,
empathy and ingroup favoritism in minimal groups. Poster session presented at the annual meeting
of the Society for Personality and Social Psychology, Savannah, GA.
John, O. P., & Srivastava, S. (1999). The Big Five trait taxonomy: History, measurement, and
theoretical perspectives. In L. A. Pervin, & O. P. John (Eds.), Handbook of personality: Theory
and research (2nd ed., pp. 102–138). New York: Guilford.
Lippa, R., & Arad, S. (1999). Gender, personality, and prejudice: The display of authoritarianism and
social dominance in interviews with college men and women. Journal of Research in Personality,
33, 463–493.
McConahay, J. B. (1986). Modern racism, ambivalence, and the Modern Racism Scale. In J. F.
Dovidio, & S. L. Gaertner (Eds.), Prejudice, discrimination and racism (pp. 91–126). New York:
Academic.
McCrae, R. R. (1996). Social consequences of experiential openness. Psychological Bulletin, 120,
323–337.
McCrae, R. R., & Costa, P. T., Jr. (1997). Conceptions and correlates of openness to experience.
In R. Hogan, J. A. Johnson, & S. R. Briggs (Eds.), Handbook of personality psychology
(pp. 825–847). San Diego, CA: Academic.
McFarland, S. G., & Mattern, K. (2002, January). Generalized explicit and implicit prejudice. Poster
session presented at the annual meeting of the Society for Personality and Social Psychology,
Savannah, GA.
Mervielde, I., & Asendorpf, J. B. (2000). Variable-centred and person-centred approaches to
childhood personality. In S. E. Hampson (Ed.), Advances in personality psychology (Vol. 1, pp.
37–76). Hove: Psychology.
Copyright # 2003 John Wiley & Sons, Ltd.
Eur. J. Pers. 17: 449–464 (2003)
464
B. Ekehammar and N. Akrami
Ozer, D. J., & Howell, R. (2002, July). Evaluating personality types. Paper presented at the biannual
European Conference on Personality, Jena.
Peterson, B. E., Smirles, K. A., & Wentworth, P. A. (1997). Generativity and authoritarianism:
Implications for personality, political involvement, and parenting. Journal of Personality and
Social Psychology, 72, 1202–1216.
Pratto, F., Sidanius, J., Stallworth, L. M., & Malle, B. F. (1994). Social dominance orientation: A
personality variable predicting social and political attitudes. Journal of Personality and Social
Psychology, 72, 741–763.
Pulkkinen, L. (1996). Female and male personality styles: A typological and developmental analysis.
Journal of Personality and Social Psychology, 70, 1288–1306.
Rammstedt, B., Riemann, R., Angleitner, A., & Borkenau, P. Resilient, overcontrollers, and
undercontrollers: Replicability of the three personality prototypes across informants. European
Journal of Personality, in press.
Reynolds, K. J., Turner, J. C., Haslam, S. A., & Ryan, M. K. (2001). The role of personality and
group factors in explaining prejudice. Journal of Experimental Social Psychology, 37, 427–434.
Saucier, G., & Goldberg, L. R. (1998). What is beyond the Big Five? Journal of Personality, 66,
495–524.
Schnabel, K., Asendorpf, A. J., & Ostendorf, F. (2002). Replicable types and subtypes of personality:
German NEO-PI-R versus NEO-FFI. European Journal of Personality, 16, S7–S24.
Sears, D. O. (1988). Symbolic racism. In P. A. Katz, & D. A. Taylor (Eds.), Eliminating racism:
Profiles in controversy (pp. 53–84). New York: Plenum.
Sidanius, J., & Pratto, F. (1999). Social dominance: An intergroup theory of social hierarchy and
oppression. New York: Cambridge University Press.
Stöber, J. (2001). The Social Desirability Scale-17 (SDS-17): Convergent validity, discriminant
validity, and relationship with age. European Journal of Psychological Assessment, 17, 222–232.
Swim, J. K., Aikin, K. J., Hall, W. S., & Hunter, B. A. (1995). Sexism and racism: Old-fashioned and
modern prejudice. Journal of Personality and Social Psychology, 68, 199–214.
Tajfel, H., & Turner, J. C. (1979). An integrative theory of intergroup conflict. In W. G. Austin, &
S. Worchel (Eds.), The social psychology of intergroup relations (pp. 33–47). Pacific Grove, CA:
Brooks–Cole.
Trapnell, P. D. (1994). Openness versus intellect: A lexical left turn. European Journal of
Personality, 8, 273–290.
Turner, J. C., & Onorato, R. S. (1999). Social identity, personality, and the self-concept: A selfcategorization perspective. In T. R. Tyler, R. Kramer, & O. John (Eds.), The psychology of the
social self (pp. 11–46). Mahwah, NJ: Erlbaum.
Turner, J. C., & Reynolds, K. J. (2001). The social identity perspective in intergroup relations:
Theories, themes, and controversies. In R. Brown, & S. Gaertner (Eds.), Blackwell handbook of
social psychology: Intergroup processes (Vol. 4, pp. 133–152). Oxford: Blackwell.
Van Hiel, A., Kossowska, M., & Mervielde, I. (2000). The relationship between openness to
experience and political ideology. Personality and Individual Differences, 28, 741–751.
Van Hiel, A., & Mervielde, I. (2002). Explaining conservative beliefs and political preferences: A
comparison of social dominance orientation and authoritarianism. Journal of Applied Social
Psychology, 32, 965–976.
Verkuyten, M., & Hagendoorn, L. (1998). Prejudice and self-categorization: The variable role of
authoritarianism and in-group stereotypes. Personality and Social Psychology Bulletin, 24, 99–
110.
Whitley, B. E., Jr. (1999). Right-wing authoritarianism, social dominance orientation, and prejudice.
Journal of Personality and Social Psychology, 77, 126–134.
Whitley, B. E., Jr., & Ægisdottir, S. (2000). The gender belief system, authoritarianism, social
dominance orientation, and heterosexuals’ attitudes toward lesbians and gay men. Sex Roles, 42,
947–967.
Whitley, B. E., Jr., & Lee, S. E. (2000). The relationship of authoritarianism and related constructs to
attitudes toward homosexuality. Journal of Applied Social Psychology, 30, 144–170.
Copyright # 2003 John Wiley & Sons, Ltd.
Eur. J. Pers. 17: 449–464 (2003)