ArticlePDF Available

Differential Effects of Right Wing Authoritarianism and Social Dominance Orientation on Outgroup Attitudes and Their Mediation by Threat From and Competitiveness to Outgroups

June 2006
Personality and Social Psychology Bulletin 32(5):684-96

DOI:10.1177/0146167205284282

Source
PubMed

Authors:

University of Auckland

A dual-process model of individual differences in prejudice proneness proposes that Right Wing Authoritarianism (RWA) and Social Dominance Orientation (SDO) will influence prejudice against particular outgroups through different motivational mechanisms. RWA should cause negative attitudes toward groups seen as threatening social control, order, cohesion, and stability, such as deviant groups, and negativity toward these groups should be mediated through perceived threat from them. SDO should cause negative attitudes toward groups that activate competitiveness over relative dominance and superiority, such as socially subordinate groups low in power and status, and negativity toward these groups should be mediated through competitiveness toward them. Findings from four student samples that assessed attitudes toward seven social groups selected as likely to vary systematically in social threat and social subordination supported these predictions. The findings have implications for reconciling intergroup and individual difference explanations of prejudice and for interventions to reduce prejudice.

A causal model of the impact of personality, social situation, and worldview on the two ideological attitude-value dimensions of Right Wing Authoritarianism (RWA) and Social Dominance Orientation (SDO) and their impact on outgroup attitudes mediated through perceived threat from the outgroup or competitiveness toward the outgroup.

…

Standardized maximum likelihood coefficients for the structural equation model of direct and mediated (through competitiveness and perceived threat) effects of Right Wing Authoritarianism (RWA) on attitudes to drug dealers (n = 98). NOTE: To simplify, manifest variables and the paths from latent to manifest variables are not shown. The model has a correlated error between threat and competitiveness (basic model). The beta in parentheses from RWA to perceived threat was computed for the same model substituting a direct path from competitiveness to threat for the correlated error (controlling for the effect of competitiveness on threat), whereas the beta in parentheses from RWA to competitiveness was computed for the same model substituting a direct path from threat to competitiveness for the correlated error (controlling for the effect of threat on competitiveness). The fit indices for the basic model were χ 2 = 67.8, df = 48, χ 2 /df = 1.41, root mean square error of approximation (RMSEA) = .065, standardized root mean square residual (SRMR) = .059, Comparative Fit Index (CFI) = .98. Nonsignificant paths are shown as broken arrows. *p < .05. **p < .01.

…

Standardized maximum likelihood coefficients for the structural equation model of direct and mediated (through competitiveness and perceived threat) effects of Right Wing Authoritarianism (RWA) on attitudes to rock stars (n = 129). NOTE: To simplify, manifest variables and the paths from latent to manifest variables are not shown. The model has a correlated error between threat and competitiveness (basic model). The beta in parentheses from RWA to perceived threat was computed for the same model substituting a direct path from competitiveness to threat for the correlated error (controlling for the effect of competitiveness on threat), whereas the beta in parentheses from RWA to competitiveness was computed for the same model substituting a direct path from threat to competitiveness for the correlated error (controlling for the effect of threat on competitiveness). The fit indices for the basic model were χ 2 = 67.8, df = 48, χ 2 /df = 1.41, root mean square error of approximation (RMSEA) = .065, standardized root mean square residual (SRMR) = .055, Comparative Fit Index (CFI) = .98. Nonsignificant paths are shown as broken arrows. *p < .05. **p < .01.

…

Figures - uploaded by John Duckitt

Content may be subject to copyright.

Content uploaded by John Duckitt

Content may be subject to copyright.

10.1177/0146167205284282PERSONALITY AND SOCIAL PSYCHOLOGY BULLETINDuckitt / DIFFERENTIAL EFFECTS OF RWA AND SDO

Differential Effects of Right Wing Authoritarianism

and Social Dominance Orientation on Outgroup

Attitudes and Their Mediation by Threat From

and Competitiveness to Outgroups

John Duckitt

University of Auckland, New Zealand

A dual-process model of individual differences in prejudice

proneness proposes that Right Wing Authoritarianism (RWA)

and Social Dominance Orientation (SDO) will influence preju-

dice against particular outgroups through different motiva-

tional mechanisms. RWA should cause negative attitudes

toward groups seen as threatening social control, order, cohe-

sion, and stability, such as deviant groups, and negativity

toward these groups should be mediated through perceived threat

from them. SDO should cause negative attitudes toward groups

that activate competitiveness over relative dominance and supe-

riority, such as socially subordinate groups low in power and sta-

tus, and negativity toward these groups should be mediated

through competitiveness toward them. Findings from four stu-

dent samples that assessed attitudes toward seven social groups

selected as likely to vary systematically in social threat and social

subordination supported these predictions. The findings have

implications for reconciling intergroup and individual

difference explanations of prejudice and for interventions to

reduce prejudice.

Keywords: Right Wing Authoritarianism; Social Dominance Orien

tation; intergroup threat; intergroup competition; preju

dice; intergroup attitudes

Two approaches have dominated inquiry into the

causes of prejudice and intergroup hostility. One ap

proach sees prejudice as a group phenomenon and ex

plains it in terms of intergroup processes (Sherif, 1967;

Tajfel & Turner, 1979). A second, which has been rela

tively neglected for the past half century, sees prejudice

as an attitude held by individuals and explains it in terms

of stable individual differences in personality or social

attitudes (Adorno, Frenkel-Brunswick, Levinson, &

Sanford, 1950). This second approach is the focus of the

present investigation.

Individual Difference Explanations of Prejudice

Empirical support for the individual differences ap-

proach to explaining prejudice comes from research

showing that persons who are less favorable to one

outgroup or minority are less favorable to others as well,

seemingly irrespective of the characteristics of these

outgroups or their relationship to the ingroup (Adorno

et al., 1950; Allport, 1954; Altemeyer, 1988). This “gener-

ality of prejudice” has supported theories focusing on

relatively stable characteristics of individuals that make

them prone to hold prejudiced intergroup attitudes in

general.

The two most prominent individual difference theo

ries of prejudice have been the authoritarian personality

and social dominance orientation approaches. An au

thoritarian personality dimension was originally de

scribed by Adorno et al. (1950), who measured this con

struct using their famous but seriously flawed F scale.

The construct was later refined by Altemeyer (1981,

1988, 1998), who developed his Right Wing Authoritari

anism (RWA) scale from those items of the original F

scale that did covary strongly enough to define a uni

684

Author’s Note: I would like to thank Nickola Overall and Boris Bizumic

for their helpful comments on this article. Correspondence concern

ing this article should be addressed to John Duckitt, Department of

Psychology, University of Auckland, Private Bag 92019, Auckland, New

Zealand; e-mail: j.duckitt@auckland.ac.nz.

PSPB, Vol. 32 No. 5, May 2006 684-696

DOI: 10.1177/0146167205284282

dimensional construct. Altemeyer (1981) suggested that

these items assessed three covarying traits of convention

alism, authoritarian aggression, and authoritarian sub

mission. Social dominance orientation (SDO) was con

ceptualized and measured by Sidanius and Pratto (1993,

1999) as a “general attitudinal orientation toward inter

group relations, reflecting whether one generally pre

fers such relations to be equal versus hierarchical” and

the “extent to which one desires that one’s ingroup

dominate and be superior to outgroups” (Pratto,

Sidanius, Stallworth, & Malle, 1994, p. 742).

Numerous studies have shown that the RWA and SDO

scales powerfully and independently predict a wide

range of political, ideological, and intergroup phenom

ena and are particularly powerful predictors of general

ized negativity to minorities and outgroups (Altemeyer,

1988, 1998; McFarland, 1998; McFarland & Adelson,

1996; Sidanius & Pratto, 1999). The issue of what exactly

the RWA and SDO scales are measuring has caused de

bate. Adorno et al. (1950), and more recently Altemeyer

(1981, 1998), assumed that the items of their F and RWA

scales, although formulated as statements of social atti-

tude, actually measured an underlying dimension of per-

sonality that causes a need for prejudice. However, this

assumption has never been empirically supported. In ad-

dition, research has shown that both the RWA and SDO

scales correlate most powerfully with attitude and value

measures (Duriez & Van Hiel, 2002; Heaven & Conners,

2001; Saucier, 2000) and are both significantly reactive

to situational manipulations (Altemeyer, 1988; Duckitt &

Fisher, 2003; Guimond, Dambrun, Michinov, & Duarte,

2003; Sales, 1973; Schmitt, Branscombe, & Kappen,

2003). These findings suggest that the RWA and SDO

scales are measuring social attitude dimensions of a

broadly ideological nature rather than personality.

This view of RWA and SDO as social attitude dimen

sions raises two questions that had been obscured by the

original personality assumption. First, what are the social

and psychological bases of RWA and SDO? Second, how

or through what mechanisms do RWA and SDO influ

ence prejudice? A dual-process, cognitive-motivational

model of individual differences in prejudice attempts to

answer both of these questions (Duckitt, 2001; Duckitt,

Wagner, du Plessis, & Birum, 2002), but only research

pertaining to the first of these two questions has been re

ported thus far. The next section briefly describes this

model and research testing it before outlining its predic

tions pertaining to the second question, which is inves

tigated in the research reported here.

A Cognitive-Motivational Model of

Personality, Ideology, and Prejudice

The dual-process model proposes that the two social

attitude dimensions of RWA and SDO express motiva

tional goals made chronically salient for individuals by

their personalities and social worldviews. RWA expresses

the value or motivational goal of societal or group secu

rity and order (obtained through establishing and main

taining societal or group control, stability, and cohesion)

generated by a view of the social world as dangerous and

threatening. The predisposing personality dimension is

that of social conformity.

In the case of SDO, the model proposes that the un

derlying personality dimension is that of tough-mind

edness, characterized by traits of being hard, tough,

ruthless, and unfeeling to others, as opposed to compas

sionate, generous, caring, and altruistic. Toughminded

personalities tend to adopt a view of the world as a ruth

lessly competitive jungle in which the strong win and the

weak lose. This “social Darwinist” view of the world

makes salient the value or motivational goals of group

power, dominance, and superiority over others, which is

expressed in the social attitudes of high SDO.

These two worldviews (of the social world as danger

ous and threatening or as a social Darwinist competitive

jungle) should generally be relatively stable reflections

of individuals’ personalities and socialization but also

should be influenced by their real social situations.

Thus, social situations that become markedly more dan-

gerous and threatening could shift individuals’

worldviews in that direction and increase authoritarian-

ism. Social situations characterized by high levels of in-

equality and competition over power and status should

produce a competitive-jungle or social Darwinist

worldview and so increase social dominance. This causal

model of personality, social situation, worldview, ideo

logical social attitudes, and prejudice is summarized in

Figure 1. As depicted, the influences of personality, the

social situation, and individuals’ worldviews on inter

group attitudes are indirect and mediated through their

impact on the two ideological value-attitude dimensions

of RWA and SDO.

A series of studies using structural equation modeling

with latent variables showed excellent overall fit for the

causal relationships proposed among the two personal

ity dimensions, two worldview dimensions, two ideologi

cal attitude dimensions, and intergroup attitudes for

large samples in New Zealand, South Africa, and the

United States (Duckitt, 2001; Duckitt et al., 2002). An ex

periment also showed that manipulating societal threat

(using future scenarios) increased RWA, but not SDO,

with the effect being entirely mediated through an in

creased perception of the social world as dangerous and

threatening (Duckitt & Fisher, 2003).

However, the second half of this model, which pro

poses an explanation for how RWA and SDO influence

intergroup attitudes, has not yet been systematically

Duckitt / DIFFERENTIAL EFFECTS OF RWA AND SDO 685

tested (see right half of Figure 1). This portion of the

model is the focus of the present research.

Explaining the Effect of RWA and

SDO on Intergroup Attitudes

The model suggests that RWA and SDO should pre-

dict negative attitudes toward different outgroups. RWA

expresses threat-driven motivation to establish and

maintain social or group security in the form of social

control, order, cohesion, and stability. Consequently,

persons high in RWA should dislike groups that seem to

threaten societal or group security. SDO, on the other

hand, expresses competitively driven motivation to

maintain or establish group dominance and superiority

so that persons high in SDO would dislike and devalue

outgroups that aroused their competitiveness over inter

group status or power differentials. These could be di

rectly competing outgroups, which would activate com

petitive desires to establish dominance, or subordinate

outgroups, which would activate competitive desires to

justify and maintain relative dominance (see Figure 1).

In practice, RWA and SDO will not always predict prej

udice against different groups because the same groups

often may be seen as both socially threatening and so

cially subordinate so that both RWA and SDO will pre

dict negative attitudes to them, although for different

reasons. This is typically the case for ethnic minority

groups because they are invariably low in power and sta

tus and also often deviate from majority group values

and norms. Directly competing outgroups (enemy or ri

val groups) also should elicit both RWA and SDO moti-

vated prejudice because direct intergroup competition

should activate both competitive desires to establish

dominance and threat perceptions.

Attitudes to these two kinds of outgroups have been

most studied in research on prejudice. This is not sur-

prising because it is precisely these groups that would ex-

perience most prejudice and discrimination. As a result,

most research on prejudice has supported the widely

held conclusion that both RWA and SDO predict out

group prejudice (Altemeyer, 1998, Duriez & Van Hiel,

2002; McFarland, 1998; McFarland & Adelson, 1996;

Pratto et al., 1994). The current model, however, sug

gests that outgroups that elicit either perceived threat or

competitiveness over relative superiority, and not both,

will reveal theoretically meaningful exceptions to this

pattern by showing differential effects for RWA and SDO

on outgroup prejudice.

The model also implies a differential mediation hy

pothesis. The effect of RWA on negative outgroup atti

tudes should be mediated by perceived threat from

outgroups, whereas the effect of SDO should be medi

ated by competitiveness over relative dominance and su

periority. Again, there has been little relevant research,

although several studies have reported findings that

seem consistent with this hypothesis. Esses, Haddock,

and Zanna (1993) found that the effect of RWA on anti

gay prejudice was mediated by perceived threat, and re

search by Esses, Jackson, and Armstrong (1998) found

that the effect of SDO on negative attitudes to a bogus

686 PERSONALITY AND SOCIAL PSYCHOLOGY BULLETIN

Figure 1 A causal model of the impact of personality, social situation, and worldview on the two ideological attitude-value dimensions of Right

Wing Authoritarianism (RWA) and Social Dominance Orientation (SDO) and their impact on outgroup attitudes mediated through per

ceived threat from the outgroup or competitiveness toward the outgroup.

immigrant group was mediated by perceived economic

competition from that group (see also Esses, Dovidio,

Jackson, & Armstrong, 2001; Jackson & Esses, 2000).

However, because these studies examined either RWA

and threat or SDO and competitiveness separately, they

could not show if perceived intergroup threat and

competitiveness mediated the effects of RWA and SDO

differentially.

Objectives of the Research

This research set out to investigate the differential ef

fect and differential mediation hypotheses for RWA and

SDO on outgroup attitudes. To test these hypotheses, it

was necessary to assess attitudes to social groups that

were likely to vary systematically in the degree to which

they would be seen as socially deviant (i.e., threatening

conventional norms, values, and traditions) and as so

cially subordinate (i.e., low in social status and power).

A list of social groups was prepared drawing largely

from the groups used by Fiske, Cuddy, Glick, and Xu

(2002) in their study of social stereotype content. Two

small ad hoc groups of postgraduate social psychology

students (total N = 22) were used as expert judges to rate

the extent to which these social groups or categories

were likely to be seen by the average person in New Zea

land as (a) “likely to threaten, disrupt, or violate main-

stream New Zealand society’s conventional norms,

values, and traditions” and (b) as “being, or represent-

ing, social groups or categories of persons likely to be

generally viewed as socially disadvantaged, subordinate,

low in power, influence, and prestige in New Zealand

society.”

Ratings were on a 9-point scale with 1 denoting defi

nitely not, the midpoint 5 denoting unsure or intermediate,

and 9 denoting definitely yes. The intention was to obtain

at least one and ideally two groups that were rated high

on both dimensions (taken as group mean ratings that

were statistically significantly higher than the midpoint

of 5 on the rating scale), low on both dimensions (taken

as group mean ratings that were statistically significantly

lower than the midpoint of 5 on the rating scale), high

on deviance and low on subordination, and low on devi

ance and high on subordination. Eight of the social

groups that had been rated in this pilot study fell clearly

into these four categories and were selected as target

groups for the research. Their mean ratings are shown in

the first five columns of Table 1.

These mean ratings suggested that two groups were

likely to be seen as socially threatening but not socially

subordinate: rock stars and drug dealers; three were

likely to be seen as socially subordinate but as not as so

Duckitt / DIFFERENTIAL EFFECTS OF RWA AND SDO 687

TABLE 1: Ratings of Likelihood of Target Groups Being Seen as “Socially Threatening/Deviant” and “Socially Subordinate” in the Pilot Study

(N = 22) and Research Hypotheses

Ratings of

Threat Subordination Research Hypotheses for Groups

Group Negativity

Target Groups MSD M SD Predicted by: Effect Mediated by:

Socially deviant/threatening, but not

subordinate, groups RWA Perceived threat

Rock stars 6.36

1.68 2.96

1.84

Drug dealers 7.23

1.66 3.50

1.97

Socially subordinate, but not deviant/

threatening, groups SDO Competitiveness

Physically disabled 2.09

1.82 7.27

1.78

Housewives 2.64

2.01 7.18

1.50

Unemployed 3.86

2.46 7.36

1.59

Both socially subordinate and deviant/

threatening group RWA & SDO Threat & competitiveness

Feminists 6.45

2.58 6.64

2.08

Neither subordinate nor deviant/

threatening groups No effect No effect

Business leaders 2.68

1.99 2.27

1.70

Medical specialists 2.72

1.70 2.86

1.67

NOTE: RWA = Right Wing Authoritarianism; SDO = Social Dominance Orientation.

a. Denotes mean was significantly above rating scale midpoint.

b. Denotes mean was significantly below rating scale midpoint.

cially threatening: housewives, physically disabled per

sons, and unemployment beneficiaries; one group

seemed likely to be seen as both socially subordinate and

threatening: feminists; and two groups were seen as nei

ther high in social threat nor as socially subordinate:

business leaders and medical specialists. One of these

target groups—medical specialists—was ultimately not

used in the research because there was too little time for

the section of the questionnaire dealing with it to be

completed. These seven groups, however, seemed to

provide an adequate sampling of social groups varying in

social threat and social subordination to test the hypoth

eses for this research. These hypotheses are summarized

in the last two columns of Table 1.

METHOD

Samples and Questionnaires

Four questionnaires were administered at the end of

laboratory classes to four samples of undergraduate stu

dents at Auckland University during 2001 and 2002. The

questionnaires used the same items, with each question-

naire assessing RWA and SDO, and then perceived

threat from, competitiveness toward, and attitudes to-

ward two of the target groups. Four different question-

naires were needed because pilot testing had shown that

questionnaires with more than two target groups were

too long to be to be completed in the time available for

the research. The target groups and samples for the four

questionnaires were as follows:

1. Attitudes toward housewives and drug dealers adminis

tered to 98 students with a mean age of 21.7 years (SD =

6.1), of whom 65.3% were female.

2. Attitudes toward business leaders and rock stars admin

istered to 129 students with a mean age of 20.6 years

(SD = 5.5), of whom 66.7% were female.

3. Attitudes toward feminists administered to 117 students

with a mean age of 20.6 years (SD = 5.1), of whom 67.4%

were female.

4. Attitudes toward unemployment beneficiaries and

physically disabled persons administered to 134 stu

dents with a mean age of 19.8 years (SD = 4.3), of whom

67.2% were female.

Measures of RWA and SDO

The short-balanced RWA scale consisted of 12 items

randomly sampled from Altemeyer’s (1996) RWA scale

so that there were equal numbers of positively and nega

tively worded items. The alpha coefficients in the four

samples ranged from .76 to .86.

The short-balanced SDO scale consisted of 10 items

randomly sampled from Pratto et al.’s (1994) SDO scale,

again with equal numbers of positively and negatively

worded items. The alpha coefficients in the four samples

ranged from .75 to .83.

Measures of Perceived Outgroup Threat

and Outgroup Competitiveness

Two measures were developed to assess perceived so

cial threat from and competitiveness toward outgroups.

Both measures had eight Likert items with equal num

bers of positively and negatively worded items.

Perceived social threat. The perceived social threat items

were selected or written to assess the perception of

threat from any group or category of persons with the

target group designated in the instructions and a head

ing. High threat items described the target group as un

dermining important social values, norms, and tradi

tions and threatening security and stability in society

(e.g., “They seem to want to destroy or harm what is good

in our society”), whereas low threat items described the

target group as strengthening important social norms,

values, and traditions and therefore making society safer,

stronger, and more united (e.g., “They strengthen val

ues, norms, and traditions that are important to people

like me”). Initial item pools were written according to

this construct definition with items adapted mainly from

the symbolic and realistic threat scales used by Stephan,

Ybarra, Martinez, Schwarzwald, and Tur-kaspa (1998)

but formulated at a higher level of generality. These item

pools were then content validated by two independent

judges for fidelity to the construct definition, and those

content-validated items finally were tested on small stu-

dent samples to select those with the highest item reli-

abilities for inclusion in the final balanced eight-item

perceived social threat scale. These eight items are

shown in the appendix. The alpha coefficients for per

ceived threat from the seven target groups ranged from

.67 to .90, indicating adequate internal consistency re

liabilities for a measure of this length.

Group competitiveness. The group competitiveness

items were selected or written to assess competitiveness

toward any social group or category of persons, with the

target group designated in the instructions and a head

ing. High competitiveness items were written to express

a zero-sum competitive orientation toward the target

group (e.g., “If they make economic gains, people like

me will be worse off”) and low competitiveness items to

express a cooperative, altruistic orientation to the target

group (e.g., “It would be to our advantage for them to get

more resources”). These definitions were used to write

items or guide selection from existing measures of inter

group zero-sum competitive orientation, particularly

those used by Esses et al. (1998) and Fiske et al. (2002).

These items were then content validated and empirically

tested on small student samples to obtain a balanced and

reliable eight-item scale and are shown in the appendix.

The alpha coefficients for competitiveness to each of the

688 PERSONALITY AND SOCIAL PSYCHOLOGY BULLETIN

seven target groups ranged from .75 to .89, indicating

adequate internal consistency reliabilities.

Measure of Group Attitude

Sixteen Likert items were used to assess group evalua

tion or attitude, with equal numbers of items expressing

positive and negative evaluation. This set of items was ini

tially intended to try and assess two aspects of group eval

uation, with a balanced set of eight items to assess re

specting versus disrespecting and another balanced set

of eight items to assess liking versus disliking, with the

two sets of items conceptualized as the attitudinal equiva

lents of the two stereotype dimensions of competence

and warmth described by Fiske, Xu, Cuddy, and Glick

(1999).

For respect versus disrespect, positively worded items

expressed respect and admiration for the possession of

abilities related to social success and achievement (e.g.,

“I admire them for their real strength and personal

power”) and negatively worded items expressed disre

spect and disdain for characteristics associated with fail

ure, weakness, and incompetence (e.g., “They don’t

seem to have much initiative and energy”). For liking

versus disliking, positively worded items expressed liking

and the attribution of positive moral and interpersonal

qualities (e.g., “I like them for their honesty and gener-

osity”), whereas negatively worded items expressed dis-

like based on the attribution of negatively valued moral

and interpersonal qualities (e.g., “I believe they are

often bad people”).

These two item pools were content validated by two

independent judges and tested on small student sam-

ples. This testing indicated that the two sets of items were

highly correlated and scaled as a single group attitude di

mension. These items were therefore used as a single

measure of group attitude, with alpha coefficients for

the seven target groups ranging from .84 to .92.

RESULTS

Expectation maximization (Schafer, 1997) was used

to replace isolated missing values in the four data sets

(less than 1% in each data set) so that the full samples

were used for all the analyses.

Correlational Analyses

The correlations between the five measures for the

seven target groups are shown in Tables 2 through 5. Be

fore considering the hypothesized effects of RWA and

SDO on attitudes to each group, two other sets of vari

able intercorrelations merit comment; first, the correla

tions of perceived threat from and competitiveness to

ward each group with attitudes toward the group and,

second, the intercorrelations between perceived threat

and competitiveness for each of the target groups.

Duckitt / DIFFERENTIAL EFFECTS OF RWA AND SDO 689

TABLE 2: Intercorrelations of RWA, SDO, Threat From, Competi

tiveness to, and Attitude to Housewives (below the diago

nal) and Drug Dealers (above the diagonal) for Sample 1

(n = 98)

Measures 1 2 3 4 5

RWA — .32** .53** .41** –.49**

SDO .32** — .06 .22** –.11

Threat .10 .18 — .45** –.67**

Competitiveness .18 .62** .26* — –.40**

Group attitude –.06 –.47** –.37** –.64** —

NOTE: RWA = Right Wing Authoritarianism; SDO = Social Dominance

Orientation.

*p < .05. **p < .01.

TABLE 3:

Intercorrelations of RWA, SDO, Threat From, Competi

tiveness to, and Attitude to Business Leaders (below the di

agonal) and Rock Stars (above the diagonal) for Sample 2

(n = 129)

Measures 1 2 3 4 5

RWA — .09 .51** .18* –.31**

SDO .09 — –.17 .13 .04

Threat –.04 –.14 — .42** –.61**

Competitiveness –.11 –.02 .66** — –.55**

Group attitude –.07 .08 –.73** –.68** —

NOTE: RWA = Right Wing Authoritarianism; SDO = Social Dominance

Orientation.

*p < .05. **p < .01.

TABLE 4:

Intercorrelations of RWA, SDO, Threat From, Competi-

tiveness to, and Attitude to Unemployment Beneficiaries

(below the diagonal) and Physically Disabled Persons

(above the diagonal) for Sample 3 (n = 134)

Measures 1234 5

RWA — .15 .16 .19* –.07

SDO .15 — .24** .43** –.35**

Threat .11 .06 — .58** –.69**

Competitiveness .19* .27** .43** — –.68**

Group attitude .01 –.27** –.57** –.54** —

NOTE: RWA = Right Wing Authoritarianism; SDO = Social Dominance

Orientation.

*p < .05. **p < .01.

TABLE 5:

Intercorrelations of RWA, SDO, Threat From, Competi

tiveness to, and Attitude to Feminists for Sample 4 (n = 117)

Measures 1 2 3 4 5

RWA —

SDO .10 —

Threat .36** .38** —

Competitiveness .22* .38** .81** —

Group attitude –.36** –.34** –.71** –.73** —

NOTE: RWA = Right Wing Authoritarianism; SDO = Social Dominance

Orientation.

*p < .05. **p < .01.

Correlations of perceived threat and competitiveness with

group attitudes. Theoretically perceived threat from and

competitiveness to outgroups should be negatively cor

related with attitudes to those groups (i.e., associated

with more negative attitudes). The findings supported

this for each target group with these correlations varying

from moderate (–.37 for perceived threat with attitudes

to housewives) to strong (–.73 for competitiveness with

attitudes to feminists).

Correlations between perceived threat and competitiveness.

Perceived threat from and competitiveness to each

group should be positively correlated because perceiv

ing threat from an outgroup also should generate com

petitive feelings to that outgroup, and feeling competi

tive to an outgroup also should generate perceptions of

threat from that outgroup. The findings supported this,

with all seven correlations positive and significant. How

ever, the magnitude of these effects (cf. Cohen, 1988)

varied, being weak for housewives (.26); moderate for

rock stars (.42), unemployed persons (.43), and drug

dealers (.45); strong for physically disabled persons

(.58); and very strong for business leaders (.66) and

feminists (.81).

The magnitude of these correlations when corrected

for attenuation due to unreliability (Lord & Novick,

1968) suggested that these two measures were clearly

tapping different dimensions for all target groups ex-

cept for the two groups, business leaders and feminists,

that had been rated as likely to be seen high or low on

both dimensions. In these two instances, the correlations

corrected for attenuation suggested that these two mea-

sures either approached unidimensionality (business

leaders, r = .80) or were effectively unidimensional (fem

inists, r = .90). This was confirmed for the two measures

for feminists using confirmatory factor analysis, with the

difference (3.32, df =1,p > .05) indicating no signifi

cant difference between a more parsimonious one-

factor, item-level solution (χ

= 204.4, df = 87) and a two-

factor, item-level solution (χ

= 202.1, df = 86). For this

reason, the measures of perceived threat from and com

petitiveness to feminists were summed to give a compos

ite threat-competitiveness measure to be used for the

mediational analyses for that group. No mediational

analysis was conducted for business leaders because nei

ther RWA nor SDO correlated significantly with attitudes

to this group, indicating that there could be no mediated

effect (Baron & Kenny, 1986).

Correlations of RWA and SDO with group attitudes. The

correlations of RWA and SDO with group attitudes (Ta

bles 2-5) were consistent with the differential effect hy

pothesis. RWA was significantly correlated with nega

tivity to the socially threatening or deviant groups, drug

dealers and rock stars, whereas SDO was not. SDO was

significantly correlated with negativity to the socially sub

ordinate groups, housewives, physically disabled

persons, and unemployment beneficiaries, whereas

RWA was not. Both RWA and SDO were significantly as

sociated with negativity to feminists (socially threatening

and socially subordinate), whereas neither RWA nor

SDO were significantly associated with negativity to busi

ness leaders (neither socially threatening nor socially

subordinate).

Mediational Analyses Using

Structural Equation Modeling (SEM)

A series of maximum likelihood SEM analyses were

used to test (a) if the significant effects obtained for RWA

on attitudes to drug dealers and rock stars were medi

ated by perceived threat and not competitiveness; (b) if

the significant effects for SDO on attitudes to housewives,

physically disabled persons, and unemployment benefi

ciaries were mediated through competitiveness and not

perceived threat; and (c) if the significant effects of both

RWA and SDO on attitudes to feminists were mediated

through the composite perceived threat-competitiveness

measure. The path coefficients and fit indices for these

six models are shown in Figures 2 through 7. Because a

significant relationship between predictor and criterion

variables is an essential precondition for mediation

(Baron & Kenny, 1986), only the significant predictor

variable or variables (i.e., RWA, SDO, or both) were in-

cluded in these SEM analyses testing for mediation.

Item parceling was used to create manifest indicators

for the analyses. Item parceling provides more reliable

indicators than individual items, enables the use of bal

anced indicators that control direction of wording ef

fects, and requires the estimation of fewer parameters,

making it more suitable for the relatively small samples

used here (Little, Cunningham, Shahar, & Wdamen,

2002). The use of at least three manifest indicators per

latent variable has been recommended to avoid prob

lems with identification, nonconvergence, and negative

variance estimates (Bollen, 1989). Consequently, the

items for each of the scales were randomly assigned to

one of three manifest indicators with equal numbers of

positively and negatively worded items in each indicator.

The manifest indicators for each scale were allowed to

relate only to their one latent variable in the analyses.

Model fit was assessed using Hu and Bentler’s (1999) rec

ommendations that good fit would be indicated by val

ues close to or better than .06 for root mean square error

of approximation (RMSEA), .08 for standardized root

mean square residual (SRMR), .95 for Comparative Fit

Index (CFI), and Bollen and Long’s (1993) suggestion

that the χ

/df ratio should be close to or less than 2.

To test for mediation, each model included both di

rect paths from RWA/SDO to group attitudes and indi

690 PERSONALITY AND SOCIAL PSYCHOLOGY BULLETIN

rect paths through both perceived threat and competi-

tiveness. Ideally, the model should have included recip

rocal causal impacts of perceived threat and competitive

ness on each other because this would then control for

the effects of each of these variables on each other, elimi

nating spuriously significant mediational effects. For ex

ample, if RWA had a significant path to perceived threat,

the positive relationship between perceived threat and

competitiveness could produce a spurious effect of RWA

on competitiveness that might attain significance.

Unfortunately, modeling reciprocal causal paths be

tween perceived threat and competitiveness made the

models overly complex, resulting in nonidentification

and/or failures to converge. Because of this, a correlated

error was modeled between perceived threat and com

petitiveness in the basic model. This would not control

for the effects of threat and competitiveness on each

other because it merely correlated the residuals so that

spurious mediational effects would be possible. To assess

whether mediational effects might be spurious, each

analysis was repeated twice, substituting unidirectional

causal paths between threat and competitiveness for the

correlated error. Modeling a unidirectional causal path

from competitiveness to threat would therefore com

pute the path coefficients from SDO/RWA to threat with

the effects of competitiveness on threat controlled,

whereas modeling a unidirectional causal path from

threat to competitiveness would compute the path

coefficients from SDO/RWA to competitiveness with

threat controlled.

Figures 2 through 7 show the path coefficients ob

tained for the basic model (i.e., with the correlated error

between threat and competitiveness) and also show in

parentheses the path coefficients from SDO or RWA to

threat with competitiveness controlled and from SDO or

RWA to competitiveness with threat controlled. Signifi

cant path coefficients from SDO or RWA to threat or

competitiveness for the basic model (with the correlated

error) would suggest mediational effects that might or

might not be spurious. If the corresponding coefficients

in parentheses were not significant, this would suggest

that the effect had indeed been spurious and due to the

effects of threat or competitiveness on each other.

Duckitt / DIFFERENTIAL EFFECTS OF RWA AND SDO 691

Figure 3 Standardized maximum likelihood coefficients for the

structural equation model of direct and mediated (through

competitiveness and perceived threat) effects of Right

Wing Authoritarianism (RWA) on attitudes to rock stars (n =

129).

NOTE: To simplify, manifest variables and the paths from latent to

manifest variables are not shown. The model has a correlated error be

tween threat and competitiveness (basic model). The beta in parenthe

ses from RWA to perceived threat was computed for the same model

substituting a direct path from competitiveness to threat for the corre

lated error (controlling for the effect of competitiveness on threat),

whereas the beta in parentheses from RWA to competitiveness was

computed for the same model substituting a direct path from threat

to competitiveness for the correlated error (controlling for the effect

of threat on competitiveness). The fit indices for the basic model were

= 67.8, df = 48, χ

/df = 1.41, root mean square error of approxi-

mation (RMSEA) = .065, standardized root mean square residual

(SRMR) = .055, Comparative Fit Index (CFI) = .98. Nonsignificant

paths are shown as broken arrows.

*p < .05. **p < .01.

Figure 2 Standardized maximum likelihood coefficients for the

structural equation model of direct and mediated (through

competitiveness and perceived threat) effects of Right

Wing Authoritarianism (RWA) on attitudes to drug dealers

(n = 98).