Anda di halaman 1dari 25

Political Psychology, Vol. 26, No.

2, 2005

The Dimensionality of Right-Wing Authoritarianism:


Lessons from the Dilemma between Theory
and Measurement
Friedrich Funke
Friedrich-Schiller-University, Jena, Germany

The RWA Scale (Altemeyer, 1981, 1988, 1996) is commonly regarded as the best measure
of right-wing authoritarianism. The one-dimensional instrument assesses the covariation
of three attitudinal clusters: authoritarian submission, authoritarian aggression, and con-
ventionalism. The incongruence between the implicit conceptual dimensionality on the one
hand and methodological operationalization on the other makes room for discussion about
whether it would be advantageous to measure the 3 facets of RWA separately. I rely on
three arguments: (1) confirmatory factor analyses showing that three-dimensional scales
fit the data better than the conventional one-dimensional practice; (2) the dimensions
showing a considerable interdimension discrepancy in their capability to explain valida-
tion criteria; and (3) the dimensions showing an intradimensional discrepancy which is
dependent upon the research question. The argumentation is illustrated by empirical evi-
dence from several Web-based studies among German Internet users.
KEY WORDS: authoritarianism; operationalization; scale development

Seven decades after the attempts of European scholars to explain the psy-
chological roots of mass support for Fascism (Fromm, 1936; Horkheimer, Fromm,
& Marcuse, 1936; Reich, 1933) and fifty years after the Berkeley researchers gave
this phenomenon a name (Adorno, Frenkel-Brunswik, Levinson, & Sanford,
1950), authoritarianism continues to be one of the most influential concepts in
political psychology. The construct outlived its architects and survived numerous
crises (Christie & Jahoda, 1954; Hyman & Sheatsley, 1954; Kirscht & Dillehay,
1967, for reviews cf. Martin, 2001; Meloen, 1994; Stone, Lederer, & Christie,
1993). In authoritarianism research not much has remained from the Freudo-
Marxist methodology of dialectical, materialist hermeneutics. The common

195
0162-895X © 2005 International Society of Political Psychology
Published by Blackwell Publishing. Inc., 350 Main Street, Malden, MA 02148, USA, 9600 Garsington Road, Oxford, OX4 2DQ
196 Funke

denominator, however, seems to be that something measurable “exists,” as lan-


guage unintentionally expresses the material structure of social reality (cf. Buck-
Morss, 1977). As in most fields of modern psychology, qualitative methods play
the role of the outsider or even the outcast; the orthodox way to assess authori-
tarianism is to present the respondents with a self-report measure, which in the
broadest sense resembles an F-Scale derivate (cf. Christie, 1991, for a review).
As a logical consequence, a gigantic body of empirical evidence has been
accumulated in correlational studies with authoritarianism both as a dependent as
well as an independent variable, cause and effect. Among the latter are all vari-
ants of narrow-mindedness, including such different targets of discrimination as
homosexuals (Haddock & Zanna, 1998; Hunsberger, 1996; Hunsberger, Owusu,
& Duck, 1999; Whitley & Ægisdóttir, 2000), feminists (Haddock & Zanna, 1994;
Smith & Winter, 2002), drug addicts (Kohn & Mercer, 1973), or whistle-blowers
(Natoli, 2000). Virtually all aspects of social life seem to be related—or at least
relatable—to authoritarianism, e.g., physician-assisted suicide (Kemmelmeier,
Burnstein, & Peng, 1999), sportspersonship (McCutcheon, 1999), attitudes toward
Gary Larson’s Far Side humor (Lefcourt, Davidson, Shepherd, & Phillips, 1997)
and the space program (Mischel & Schopler, 1959; Peterson, Doty, & Winter,
1993), to name only a few.
Two lessons can be learned from this all-encompassing variety of correlates
towards authoritarian attitudes: first, the character of authoritarianism is funda-
mental for the description of the individuals’ social reality. Second, I deduce that
authoritarianism is multifaceted rather than monolithical, and in certain spheres
only certain aspects of the whole are responsible for the correlations.

The RWA Scale

The International Encyclopedia of the Social & Behavioral Sciences names


the RWA Scale as the state-of-the-art instrument in that field (Feldman, 2001,
p. 971); the Handbook of Personality and Social Psychological Attitudes promotes
Altemeyer’s scale(s) as “the best current measure of the essence of what the
authors of TAP were attempting to measure” (Christie, 1991, p. 552, cf. also
Eckhardt, 1991).

Reduction

There is little dissent about the conception of authoritarianism as a complex,


multifaceted construct. The Berkeley research group (Adorno et al., 1950) listed
nine traits as the components of the authoritarian syndrome: conventionalism,
authoritarian submission, authoritarian aggression, anti-intraception, superstition
and stereotypy, power and “toughness,” destructiveness and cynicism, projectiv-
ity, and, finally, sex (which is central to the theory). Altemeyer (1996, 1981, 1988)
has reduced the nine facets to the “Big Three,” the common denominator: his
Dimensionality of RWA 197

RWA Scale measures the “covariation of three attitudinal clusters in a person”:


authoritarian submission, authoritarian aggression, and conventionalism (1996,
p. 6).

Dimensionality

The Berkeley researchers as well as Altemeyer conceptualized authoritarian-


ism as a complex construct. The “concern with highly specific, statistically ‘pure’
factors was put aside, in favor of an attempt to gain a dependable estimate of an
over-all system” (Adorno et al., 1950, p. 14).
Altemeyer has “a definite hypothesis about the factorial structure of the RWA
Scale, namely, that it is essentially unidimensional” (1996, p. 53). He explains the
failure of three-factor solutions in Exploratory Factor Analysis with the very
process of scale construction: “The answer is, that the three components are thor-
oughly intertwined among the items on the test. All of the items tap at least two
(. . .). It would therefore be nigh impossible to find, say an authoritarian submis-
sion factor in such a tangle (but the “tangle” is consistent with the definition of
right-wing authoritarianism as the covariation of the three)” (p. 320, footnote 6).
From the psychometric point of view this assertion might be regarded as
counterintuitive. It is consistent, however, with the implicit perception of author-
itarianism as something complex, multifaceted, but nevertheless holistic.
However, it might be interesting to distinguish analytically between the
facets, precisely because they are part of a whole. A numerical thought experi-
ment helps to exemplify the problem. Consider three items (A; S; C) measuring
one facet of authoritarianism each on a 7-point Likert scale. The lowest possible
global (sum) score would be three (1; 1; 1), the highest 21 (7; 7; 7). One can infer
the combination of the subdimensions one-to-one from the global score, but only
at the extremes of the codomain. However, with a global score of four instead of
three (resp. 20 instead of 21) this score-combination projection becomes ambigu-
ous with three combinations each (e.g., 1; 1; 2 or 1; 2; 1 or 2; 1; 1). The theoret-
ical mean of the global sum score in our example is 12, a figure that can
be obtained by every one out of 37 (!) possible combinations (totaling 73 = 343
combinations)1.
The theoretical implication is that the commonly used global RWA-score
conceals the genuine combination of subdimensions. This is particularly critical
in the middle range between very high and very low scores, where the density of
the distribution is maximal. Taking into consideration that the lion’s share of sur-
veyed samples in authoritarianism research has such moderate RWA-scores, it

1
The numerical example was chosen for the sake of better comprehensibility. The illustration is
oversimplified, however, since the analytical dimensions are not orthogonal and therefore do not
simply sum up. The correlation between the subscales would demand a weighted sum with weights
below 1.
198 Funke

becomes obvious that relying on a single score potentially masks important


aspects of authoritarianism.
The inconsistency between multifaceted theory and one-dimensional meas-
urement has been difficult to determine, but so far it has not attracted explicit sci-
entific interest, especially not from an explicitly methodological perspective. It is
crucial to emphasize that Altemeyer himself on the one hand denies the necessity
of a three-dimensional measurement; on the other hand he devoted several exper-
iments (and paragraphs) to the examination of the separate dimensions (submis-
sion: Altemeyer, 1981, pp. 189–192, 224–227; 1988, p. 270ff.; 1996, pp. 20–22;
aggression: 1981, 232–238; 1988, pp. 112–114; 1996, pp. 22–24; conventional-
ism: 1988, p. 311ff., 1996, pp. 31–38, Altemeyer & Hunsberger, 1992; Leak &
Randall, 1995).
Some colleagues presume that counterintuitive empirical findings (such as a
return of small or even negative correlations when substantial effects had been
expected) can be attributed to the fact that only one dimension is responsible for
the global correlation with the validation criterion rather than all three dimen-
sions. Examples of this implicit treatment of RWA as a multidimensional con-
struct can be found in Duncan, Peterson, and Winter (1997), Smith and Winter
(2002), or—according to Duckitt’s group authoritarianism—in Stellmacher and
Petzel (this issue). Only recently has this problem been explicitly identified
(Funke, 2000, 2002, 2003). The parallel and independent development of an
empirical approach to this dilemma (Duckitt & Bizumic, 2002) illustrates the
emergent awareness of the scientific community.

Inconsistency between Theory and Operationalization

The major thesis of this article is the ill fit between theory and methodology,
specifically the claim that the theory could be read as implying three facets, while
the existing RWA- and F-Scales are constructed one-dimensionally. This article
demonstrates the necessity or at least the benefit of items capable of assessing the
three subdimensions of authoritarianism separately.
The empirical part of my argumentation relies on the following line of rea-
soning: (1) with the use of structural equation modelling, it can be demonstrated
that three-dimensional scales fit the data better than the conventional one-
dimensional practice; (2) the three subscales show a considerable interdimension
discrepancy in their capability to explain a validation criterion of interest; and (3)
depending upon the research question, the dimensions show an intradimensional
discrepancy.
Enhanced Model Fit. Given a scale comprising items that measure the three
dimensions separately—which is not the case with the traditional RWA Scale—
it should be feasible to show that three-dimensional measurement models fit
the data better than the conventional one-dimensional practice. In terms of the
superiority of three-dimensional scales, it can be considered a necessary (but
Dimensionality of RWA 199

insufficient) condition that confirmatory analytical factor models with three cor-
related constructs describe empirical data better than restrictive models with only
one underlying factor.
In addition it is advantageous to take the wording direction of the items into
account (Eid, 2000; Marsh, 1989; Steyer & Eid, 2001). This is necessary because
of certain response styles with regard to positively versus negatively formulated
items. This general bias can be found in virtually all self-reporting measures (e.g.,
self-esteem, Salgado & Iglesias, 1995; Tomás & Oliver, 1999). Consequently the
Exploratory Factor Analyses on RWA responses commonly result in two princi-
pal components (e.g., Altemeyer, 1996, 1998; Tarr & Lorr, 1991).
Divergent Validity of the Dimensions—Interdimensional Discrepancy. The
relative benefit of a three-dimensional measurement could be falsified by showing
that the dimensions do not differ from one another in explaining validation crite-
ria (e.g., prejudice, punitiveness). In other words, in order to illustrate the supe-
riority of multidimensional measurement models, one has to provide evidence for
a certain degree of discriminant or divergent validity (Campbell & Fiske, 1959;
Kenny & Kashy, 1992; Schwarzer, 1986).
In the conventional approach of using a single aggregate (e.g., sum, mean,
median) of RWA item responses, logically there is only one regression coefficient
b between the RWA score x and the dependent validation criterion y. Supposing
we had three dimensions x1 to x3 instead of one, we could consequently examine
three regression coefficients, b1 to b3. If the equality of those parameters could be
falsified, this would make room for a discussion about the differential predictive
potential of the separate dimensions.
Moreover if the three predictors differ from one another both in magnitude
or even in direction, this implies that at least one of them might have a different
statistical impact on the dependent variable from what the traditional single RWA
score would indicate. Such an empirical finding would support the interpretation
that the usual one-dimensional approach is blind to differential effects of the sub-
dimensions, since the global score per definitionem masks potential differences.
Intradimensional Discrepancy Depending on the Explained Construct. The
previously mentioned interdimensional divergence justifies the analytical distinc-
tion between the subdimensions. Even more convincing evidence for the benefit
of multidimensional models can be provided by showing intradimensional dis-
crepancy. By this I mean a different predictive potential of the very same dimen-
sion depending on the explained construct.
By imagining two studies with different research questions a possible finding
might be that the regression coefficient from the global RWA score on one depend-
ent construct equals the association with the second validation criterion. Instead,
the use of three predictors might reveal that in one context authoritarian submis-
sion is the strongest predictor, while in others it is conventionalism.
Theoretical Implications. There are real theoretical advantages to conceiving
and measuring authoritarianism as multidimensional. Operationalizing the three
200 Funke

facets with dimensionally pure items enables research to explore at least three
classes of hypotheses that could not be subject to research with the common one-
dimensional approach: (1) multidimensional assessment allows the identification
of several phenotypes of authoritarian attitudes (especially among “Moderates”);
(2) in experimental studies it becomes possible to clarify the dialectical dynam-
ics linking the dimensions (e.g., the systemic shift from an individual’s submis-
sive to aggressive behavior when a dominant high status in-group member has
left the group); and (3) the proposed approach creates the opportunity to test the
(qualitative) intraindividual stability of authoritarianism. All those hypotheses
have far reaching consequences for authoritarianism theory as a whole, as they
touch axiomatically introduced postulations that are not testable with a single
scale score.
Assessing authoritarianism multidimensionally is a necessity, but it is not a
sufficient prerequisite to clarify differences between followers and leaders (see
my first point in the preceding paragraph), who have been treated equally as
authoritarians. This approach is compatible with recent developments of treating
Social Dominance Orientation as “the other authoritarianism” (Altemeyer, 1998;
Heaven & Bucci, 2001; Sidanius & Pratto, 1999; Six, Wolfradt, & Zick, 2001;
Whitley, 1999).
While the supposition of these types of authoritarians reflects an under-
standing of authoritarianism as a relatively stable trait, the multidimensional
measurement also could explore interactionist approaches. Researchers especially
interested in the dialectics of submission and aggression could benefit from an
analytical separation of the subdimensions. A promising research question could
be the clarification of the role of threat in the process of bridging the gap between
social stressors and the activation of authoritarian behavior (Duckitt, 1992;
Feldman, 2003; Feldman & Stenner, 1997; Jost, Glaser, Kruglanski, & Sulloway,
2003). The hypothesis could be that submission shifts to open aggression under
certain circumstances. Again, this hypothesis is only testable with separate
subscales.

Step I: Deconstructing the Double-Barreled Items

Most RWA items are tapping—by design—at least two or even all three sub-
dimensions. Item 7 from the 1996 RWA Scale can be considered a prototypical
example. It reads: “Our country will be great if we honor the ways of our fore-
fathers [conventionalism], do what the authorities tell us to do [submission], and
get rid of the ‘rotten apples’ who are ruining everything [aggression]” (Altemeyer,
1996, p. 13).
From a psychometric point of view the confusion of several constructs in
one item is commonly regarded as a professional error or at least a peccadillo
(DeVellis, 1991; Robinson, Shaver, & Wrightsman, 1991). The main problem
about such so-called double-barreled items is the ambiguity and logical inability
Dimensionality of RWA 201

to decide which part of the item forced the respondent to her or his final answer.
In the case of RWA items it is impossible to deduce the theoretical contribution
of each dimension from the single aggregated mean score, as pointed out in the
numerical example above.
In a first step the original RWA items had to be evaluated concerning their
degree of “double-barreledness.” Therefore, a team of raters (18 undergraduates)
informed about the definition of RWA was asked to assess each item with regard
to its ability to measure the isolated dimensions. Three strategies have been fol-
lowed: the first instruction was to make a crystal-clear decision by placing the
items in one of the three possible mutually exclusive categories: aggression (A),
submission (S), or conventionalism (C). As the motive of this procedure was to
disentangle the items with reference to more than one construct, it was plausible
that hardly any item fell in such a clear-cut category. Thus, in a second step the
experts were asked to put the items in one of seven categories comprising the pure
types A, S, and C, and, in addition, all mixed types AS, AC, SC, and ASC.
The most painstaking but also promising approach was independently to rate
each item concerning its capacity to measure every one of the three subdimen-
sions on a 5-point rating scale ranging from 0 (worthless as a measure for this
construct) to 4 (tapping exactly this construct). As a result of this modus operandi
every rater described each item with three integers [A; S; C] ranging from 0 to 4.
A hypothetically pure item measuring only authoritarian aggression would have
received the description [4; 0; 0], abstractly speaking, a high absolute value on
one dimension and simultaneously a relatively high value with regard to the other
facets.
In order to sharpen the contrast between the dimensions the raw evaluations
were reparameterized according to Equation 1 in order to receive results with high
positive values on one and at the same time with negative values on the two other
dimensions.
Adist = (a - s) + (a - c)
(1)
= 2 a - (s + c)
In our illustration, [4; 0; 0] would have been transformed to [8; -4; -4] and
would be reading as “a pure aggression item, not measuring submission nor
conventionalism.”
As the item evaluation was based on multiple measurements, each element
of the vectors [A; S; C] has a certain sampling distribution with a specific central
tendency (arithmetic mean) and a particular shape (variance, skewness, kurtosis).
While the former parameter has an obvious interpretation, the latter can be deci-
phered as a proxy for the degree of consensus between the experts.
The 30-plus-four items of the RWA Scale (Altemeyer, 1996) in their German
translation (Funke, 1996) were subject to this procedure. The hope was to find a
high proportion of pure items, balanced with regard to the three dimensions and—
nested within them—wording direction. The described quantitative approach
202 Funke

revealed that 11 RWA items are (perceived as) double barreled; an additional three
items tap all three facets. The extensive presentation of those results goes beyond
the scope of this article and can be requested from the author.
More dramatic, however, is the fact that even if we allow for mixed items (as
explicitly intended by Altemeyer), they are seriously imbalanced, with most
aggression items (pure or mixed) being “protraits,” and almost all “contraits”
tapping conventionalism or a combination of conventionalism and submission. It
would therefore be conceivable that putatively formal differences (like the coding
direction of protraits and contraits) obscure the conceptual specifics of certain
items.
From these findings we learned the bitter but salutary lesson that it will not
be sufficient to use the original double-barreled RWA items and simply analyze
them with modified sophisticated measurement models (e.g., allowing for double
loadings or interval restrictions). If we want to learn more about the relative
impact of the separate dimensions, we have to construct a three-dimensional scale
with pure items for each facet.

Step II: Constructing a Three-Dimensional Scale

The logical consequence was the construction of a three-dimensional scale


with items purely measuring one of the three constructs. However, it was not
intended to create an entirely new scale from scratch. In order to gain a broad
acceptance in the scientific community, the modified scale should be as pioneer-
ing as necessary but as conservative as possible.
The items of the 1996 RWA Scale (Altemeyer, 1996, p. 12ff.) in their German
adaptation (Funke, 1996) have been grammatically simplified and occasionally
split up in items that are pure with respect to the underlying theoretical dimen-
sions. In some cases item reversals have been constructed in order to fit in the
3 ¥ 2 ¥ 2 (Dimension ¥ Wording Direction ¥ Item) framework. This procedure
resulted in a 12-item short-scale (see the appendix) balanced with respect to
theoretical and methodological aspects and still maintaining the character of the
original Altemeyerian RWA Scale.

Study 1: Enhanced Measurement Models

Study 1 was a Confirmatory Factor Analysis with the objective to demon-


strate that three-dimensional scales fit the data better than the conventional one-
dimensional practice.

Method

Participants. Confirmatory Factor Analysis (CFA) is a large sample method.


In order to meet the data requirements you need multivariate, normally distrib-
Dimensionality of RWA 203

uted data from several hundreds of cases (Bollen, 1986, 1989; Nachtigall,
Kroehne, Funke, & Steyer, 2003; Tanaka, 1987). Therefore it was sensible to test
the measurement model using a pooled sample from nine different studies.
All surveys used for the present purpose were aimed at the same population,
namely adult Internet users. In the space available here the studies will not be
documented in detail. Further information is available from http://www.uni-
jena.de/~sff/research/.
The total sample of 3,095 persons included 1,429 women (48.5%) and 1,515
men (51.5%). One hundred and fifty-one individuals did not disclose their gender.
The percentage of female participants has been growing over the years, indicat-
ing an increasing representativeness (cf. Best, Krueger, Hubbard, & Smith, 2001).
The mean age among the participants was 26.3 years (Mdn = 24, SD = 8.1).
Procedure. The framework conditions of data collection were essentially
comparable. Internet users were informed about the publication of a Web-based
questionnaire about current political issues (e.g., Parliamentary elections for the
German Bundestag 1998 and 2002, military actions in Kosovo, Chechnya,
Afghanistan, and Iraq, attitudes towards torture). All studies consisted of several
thematic blocks, one of them containing the 12 items of the RWA3D Scale. It was
left up to the participants to answer the questions or to skip some of them.
Anonymity was guaranteed. After the automatic data storage the participants were
debriefed about the background of the study.
Data Analysis. The purpose of data analysis was to contrast enhanced-theory-
based CFA models to the traditional practice using the single RWA score. How
can the calculation of the mean (or sum) score be translated in the notation of
structural equation modeling (SEM)? Strictly speaking this would result in a direct
formative model (Edwards & Bagozzi, 2000): all factorial loadings (from the
manifest indicators to the aggregate) are equal, there is no residual, and intercor-
relations between the items are allowed. The mean (or sum) is a perfect linear
combination of the unweighted items.
This model, however, is saturated (i.e., has zero degrees of freedom) and is
therefore not testable, as it fits the data perfectly. For that reason a similar, but
restrictive, model was chosen as the reference model. All loadings on a single
factor and all error variances were restricted to be equal (i.e., a single-trait model
of parallel tests, subsequently called single-factor model).
According to the theoretical assumptions during the scale construction
process, at least three characteristics of a preferential alternative model have to
be guaranteed: (1) the covariance structure among the RWA3D items is explained
by three intercorrelated factors; (2) the item reliabilities may differ from one
another; and (3) the wording direction of protraits and contraits is adequately
taken into account (Eid, 2000; Marsh, 1989; Marsh & Grayson, 1995).
The extensive description here will be limited primarily to the two mentioned
models at the theoretical extremes, although there are 10 hierarchically nested
models between the most liberal and most restrictive one (Funke, 2003). For
204 Funke

model parsimony and robustness considerations 6-item parcels of two items each
(Bandalos & Finney, 2001; Hall, Snell, & Foust, 1999; Schallow, 2000) have been
used as indicators.
The models were subjected to Maximum Likelihood Estimation as imple-
mented in LISREL 8.54 (Jöreskog & Sörbom, 1996) using the covariance matrix
(based on listwise deletion) and the asymptotic covariance matrix (Satorra, 1989).

Results

Single-Factor Model. The covariance matrix of 3 ¥ 2 (dimension ¥ wording)


item parcels was subjected to a CFA with a single factor, equal loadings, and equal
error variances of the indicators. This (parallel) model is close to the common
practice of calculating a global mean score for the RWA Scale. All completely
standardized loadings have been estimated as .58 implying a low-squared multi-
ple correlation (SMC) of .34 for the indicators. The model does not fit the data
with c2 (19, N = 3,095) = 2,448.4, p = .000, also the descriptive indices fail to
support it (e.g., RMSEA = .20, GFI = .79, CFI = .58). The necessary rejection of
the model cannot be defended by the large sample size, as the Critical N (Hoelter,
1983) lies drastically under the actual sample size (cN = 42).
Simple Three-Dimensional Model. The proposed model explains the covari-
ance structure of the 6-item parcels by three correlated factors. All loadings and
error variances have been estimated without restrictions, except for scaling pur-
poses (congeneric model). The model showed acceptable descriptive fit indices
(GFI = .98, NFI, CFI, IFI = .97) but had to be rejected by strict criteria with c2
(6, N = 3,095) = 226.0, p = .000. Because of skewed distributions on some indi-
cators data have been normalized within PRELIS 2.54, and the covariance matrix
has been supplemented by the asymptotic covariance matrix in order to estimate
the Satorra-Bentler Scaled c2 (Chou, Bentler, & Satorra, 1991; Satorra & Bentler,
1988). This resulted in an improved fit of c2 (6, N = 3,095) = 172.2, p = .000.
Three-Dimensional Model with Method Factor. Half of the item parcels rep-
resent responses on contraits keyed negatively. As those variables share the
commonality of the same “method” (i.e., wording direction), it is theoretically
reasonable to take the coding direction into account (Eid, 2000; Marsh, 1989).
This could have been done by introducing a t-equivalent method factor with load-
ings on the (recoded) contrait indicators. As item reversal might affect the three
dimensions in diverse ways, it is justified to specify a more liberal model allow-
ing for error covariances between the contraits (CTCU—correlated-traits-
correlated-uniqueness-model; cf. Marsh, Byrne, & Craven, 1992). This model
modification is strictly theory driven from the methodological perspective and
does not represent data-driven “post-hockery.”
The estimation process resulted in a low RMSEA = .031 and a Scaled c2 (3,
N = 3,095) = 11.7, p = .009, both indicating a close fit (pclose = .95; Browne &
Cudeck, 1993; Steiger, 2000). The descriptive indices range from .99 (AGFI, RFI,
Dimensionality of RWA 205

NNFI) to the theoretical maximum of 1.0 (GFI, CFI, NFI, IFI). Because of the
large sample (N = 3,065), c2 is still significant, but Critical N is only slightly under
the actual sample size (cN = 2,805). The large sample size allows for cross vali-
dation. The Expected Cross-Validation Index is a measure of the discrepancy
between the fitted covariance matrix in the analyzed sample and the expected
covariance matrix that would be obtained in another sample of the same size
(Browne & Cudeck, 1989). It was estimated with ECVI = .015. Comparing one
random sample (50%) of the original dataset to the other fraction yielded a cross-
validation index of CVI = .0045, indicating a stable parameter estimation. The
completely standardized solution is displayed in Figure 1. The Lisrel-input file is
available from http://www.uni-jena.de/~sff/JPP/.
The omnibus evaluation of the model reveals an acceptable fit. The local
evaluation of particular parameters shows systematically lower loadings for the
negatively worded items. To a certain extent this is the logical consequence of
higher error variances, as respondents with an acquiescence tendency bias the
estimations.

Figure 1. Completely standardized solution of the congeneric three-dimensional model with


method factor. Scaled c2 (3, N = 3,095) = 11.7, p = .008, RMSEA = .031.
206 Funke

The attention should also be concentrated on the latent disattenuated corre-


lation between the subdimensions. The highest association was found between
aggression and submission. This finding is attributable to the dialectical interplay
of authoritarian aggression and authoritarian submission as two sides of the (same)
medal: people favoring aggression against deviants implicitly demand their
submission.
The covariance between submission and conventionalism also is substantial,
as conventionalism can be interpreted as a special case of submission (under a
social norm), whereas the theoretical distance of the concepts aggression and con-
ventionalism is represented by the lowest correlation.
Model Comparison. Both proposed three-dimensional models fit the empiri-
cal data better than the parallel one-factor model, supporting the notion that a fine-
grained resolution in the measurement model is superior to the common practice
of aggregating item responses to a single indicator (i.e., sum, mean). As the pre-
sented models are organized in a hierarchical structure (nested), it is possible to
compare them by the c2-difference test2 (see Table 1). If the difference is statisti-
cally significant, the null hypothesis has to be rejected. Given the assumption that
the liberal model was accepted, this indicates that the more restrictive model leads
to an inferior correspondence of empirical and model-implied covariance
structures.
A single-factor model fits the data significantly worse than both presented
three-factor models. Introducing error covariances between negatively worded
subscales improves the model fit.

Discussion

In order to show the superiority of fine-grained measurement models, a


theory-driven sequence of hierarchically nested models was tested. The most
effective way to improve model fit in comparison with the common practice was
to allow for three factors instead of one. This supports the notion that the three
dimensions of authoritarianism are correlated but nevertheless show discriminant
validity. Moreover, the model can be improved by introducing a congeneric
method factor, accounting for the specifics of the contrait items. This fact has the-
oretical implications going beyond methodological details. This poses the crucial
question whether the contraits are really measuring the opposite of authoritarian-
ism. The problem, however, is not an empirical one. We have to be aware of the

2
Since the data depart from multivariate normality, the Satorra-Bentler Scaled c2 statistic is superior
to the ordinary Normal Theory Weighted Least Squares c2. However, it is a little known fact
that the mainstream way of simply replacing c2 with the Satorra-Bentler (SB) c2 (e.g., Byrne &
Campbell, 1999) is illegitimate, as the difference between two SB-Scaled c2 test statistics does
not necessarily correspond to the scaled c2-difference statistic. Satorra and Bentler (2001) have
formulated a correction permitting an adequate test. The model comparison in Table 1 is based on
this appropriate adjustment.
Dimensionality of RWA 207

Table 1. Fit Indices for Nested Sequence of Measurement Models for the RWA3D (N = 3,065),
best-fitting model in bold.

Model df c2 Scaled c2SB RMSEA GFI corrected (ordinary)


(90% CI) Dc 2SB
Ddf
original skewed data
Null Model 15 6,491.24***
.190
Single-Factor 19 2,448.40*** 2,221.44*** (.190; .200) .79 most constraints
.096 168.2
Three Factors 6 225.97*** 175.41*** (.084; .110) .98 (174.8)
Three Factors with
Error Covariance .037 46.4
(CTCU) 3 16.56*** 15.46** (.020; .056) 1.00 (53.3)
normalized data (PRELIS 2.54)
Null Model 15 6,445.91***
.200
Single-Factor 19 2,423.83*** 2,314.57*** (.190; .200) .79 most constraints
.095 172.3
Three Factors 6 200.70*** 172.23*** (.053; .080) .98 (164.8)
Three Factors with
Error Covariance .031 49.6
(CTCU) 3 12.48** 11.70** (.014; .050) 1.00 (53.5)
Note. The c2-difference test was based on the Scaled Satorra-Bentler c2 with the obligatory
correction (Satorra & Bentler, 2001). Restrictive models are compared to the next more liberal
model, not only to the most liberal baseline model. ***p ≤ .001, **p ≤ .01, *p ≤ .05, RMSEA =
root-mean-square error of approximation with confidence interval in parentheses, GFI = goodness-
of-fit index.

fact that “opposing authoritarian attitudes” is a construct in its own right. The
results provide evidence that future research has to elucidate the polarity problem
of authoritarianism (as a whole) as well as its subscales.

Study 2: Interdimensional Discrepancy

Having shown that congeneric three-dimensional models with (or without)


method factors are superior to the common practice of aggregating all items to a
single score is a necessary, but not sufficient argument in my line of reasoning.
The relative benefit of a three-dimensional measurement could be falsified by
showing that the dimensions do not differ from one another in explaining valida-
tion criteria.
208 Funke

Method

Participants. From September 2000 until March 2001, 416 persons took part
in a Web-based survey about “Political Attitudes and Personality.” The sample
included 165 women (valid 45.6%) and 197 men (valid 54.4%) with a mean age
of 24.9 years (Mdn = 23, SD = 7.3).
Material. The questionnaire consisted of several parts, including the RWA3D
Scale and the Big-Five-Inventory NEO-FFI (Borkenau & Ostendorf, 1991; Costa
& McCrae, 1992). Furthermore the participants of the study were asked to put
themselves in the position of a prosecutor and impose a prison sentence in four
criminal cases presented in the form of (fictive) newspaper articles (e.g., sexual
abuse, professional misconduct of policemen). The length of imprisonment (in
years) was the dependent variable. This part of the study was intended to test
the association of authoritarianism and punitiveness. Results of the sexual abuse
scenario are presented here exemplarily. The conclusions for the other fictive
criminal cases are analogous with the exception of the police misconduct
scenario. As policemen are part of the establishment, a plausible authoritarian
reaction is to defend their behavior, thus the association between authoritarianism
and punitiveness is suppressed.
Data Analysis. Because of the relatively small sample size, classical methods
have been favored instead of SEM. Simple regression with the global RWA3D
score as the only predictor has been contrasted to multiple regression with the
three dimensions as the independent variables.

Results

Global RWA Model. The zero-order correlation of the global RWA3D score
and imposed term of detention in the sexual abuse case was r = .073 (n.s.), con-
sequently with an insignificant adjusted R2 = .002, F(1, 308) = 1.65, p = .20. This
result would support the questionable null hypothesis of no linear association
between authoritarianism and punitiveness against delinquents.
Three-Dimensional Model. Entering the three dimensions in the regression
instead reveals a significant relationship (adj. R2 = .118, F(3, 304) = 14.7, p < .00).
Stepwise regression nominated “aggression” as the strongest predictor (R2 = .073),
followed by “submission” (DR2 = .039, p < .00) and “conventionalism” (DR2 =
.014, p = .03). The final model with all three predictors was (1) significantly better
than the one-predictor model and (2) revealed notable differences between the
dimensions (Table 2).
Obviously the unexpected null effect of global authoritarianism is an artifi-
cial result of a mutual “canceling out” of the separate effects of the three subdi-
mensions. The negative effects of submission and conventionalism suppress the
positive effect of aggression.
The differential effect was exemplified only for the case of sexual abuse. A
comprehensive discussion of the results concerning other criminal cases (espe-
Dimensionality of RWA 209

Table 2. Regression Analysis Summary for One Global RWA-Score versus Three Dimensions of
Authoritarianism Predicting Term of Detention (Sexual Abuse Case) in Years (N = 308).

Variable B SE B b zero-order r
(Constant) 11.89 1.45
RWA3D global .63 .49 .07 n.s. .073
(Constant) 12.20 1.38
RWA3D aggression 2.26 .36 .38*** .270
RWA3D submission -1.32 .57 -.16** -.038
RWA3D conventionalism -0.88 .39 -.14** -.125
Note.
One-Predictor Model: Adjusted R2 = .002, F(1, 308) = 1.7, p = .20.
Three-Predictor-Model: Adjusted R2 = .118, F(3, 304) = 14.7, p < .00.

cially professional misconduct of policemen) goes beyond the focus of this article,
but one point is noteworthy: it could be shown that authoritarianism is negatively
related to punitiveness toward lawbreakers who are themselves authorities, includ-
ing policemen (Feather, 1998; cf. Henkel, Sheehan, & Reichel, 1997, for contra-
dicting results). The present study revealed that people favoring exemption from
punishment are significantly more authoritarian than respondents demanding
imprisonment of at least one year: respondents favoring acquittal are more con-
ventional (mean difference .91), more submissive (.82), and more aggressive
(.77).

Discussion

The traditional approach would not have been able to detect the theoretically
expected association between (global) authoritarianism and punitiveness against
offenders. Higher levels of authoritarian aggression, however, find their concrete
expression in longer imposed terms of imprisonment. This plausible and in a way,
tautological, association is totally suppressed by the negative correlation of con-
ventionalism and submission with punitiveness. Presumably respondents high in
authoritarian submission and conventionalism tend to reinterpret the sexual abuse
situation to the disadvantage of the sufferer (“blaming the victim”) in order to sta-
bilize their worldview (just-world beliefs, cf. Connors & Heaven, 1987; Montada
& Lerner, 1998; Rubin & Peplau, 1975). As an expression of their conventional-
ism and conservative attitudes towards sexuality, those respondents implicitly
allege that the victim might have provoked the offender and is therefore com-
plicit. Authoritarian submission leads to the tendency to excuse the (authority of
the) physician.
This interpretation is supported by a separate analysis for men and women.
The effect of authoritarian aggression is virtually the same in both subsamples,
while the negative association of conventionalism and submission with punitive-
ness is only found among (less empathetic) men.
210 Funke

This example supports the postulation of an interdimensional discrepancy and


moreover demonstrates impressively that significant (positive and negative) cor-
relations of the three dimensions can be completely masked by the traditionally
applied global RWA score.

Study 3: Intradimensional Discrepancy

Even more convincing evidence for the benefit of multidimensional models


could be provided by showing a differential predictive potential of the subdi-
mensions for other validation criteria (intradimensional discrepancy) in addition
to the above-mentioned interdimensional divergence. In order to show that, I
discuss a further study providing evidence that aggression is often, but not always,
the most predictive dimension.

Method

Participants. In March and April 2003 after the 2002 Parliamentary elections
for the German Bundestag, Internet users were asked to take part in a Web-based
study about general and current political issues. The valid sample size of 513 con-
sisted of 209 women (valid 50.9%) and 202 men (valid 49.1%) with a mean age
of 29.4 (Mdn = 26, SD = 10.4). As some respondents dropped out before the end
of the questionnaire, the effective sample size for the presented analyses was 368.
Material. Among questions about the Iraq War more general topics had been
touched, such as attitudes toward different acculturation styles (Berry, 1997). The
survey included short scales assessing the promotion of segregation, integration,
and assimilation as well as the Capitalist Value Scale (McClosky & Zaller, 1984).
Data Analysis. Again simple regression (correlation) has been contrasted to
multiple regression with the three dimensions as predictors.

Results

Global RWA model. Individuals with higher scores on the global RWA3D
Scale similarly support capitalist values(r = .45), segregation (r = .46), and assim-
ilation (r = .47) and oppose integration of immigrants (r = -.40).
Three-dimensional model. Entering the separate dimensions of authoritarian-
ism instead revealed diverse patterns of association. While assimilation was
significantly predicted by all three dimensions, conventionalism had no impact
on integration and segregation. Likewise submission failed to predict support of
capitalist values. Standardized regression coefficients along with the zero-order
correlations are reported in Table 3.
Discussion. It can be shown that the association between the global RWA3D
score and typical validation criteria in authoritarianism research masks differen-
tial effects of the separate dimensions. While in some cases all three dimensions
Dimensionality of RWA 211

Table 3. Differential Effect of Three Dimensions of Authoritarianism on Different Validation


Criteria. (Standardized Partial Regression Coefficients, Zero-Order Correlations in Parentheses)

Variable Acculturation Capitalist


Value Scale
Assimilation Integration Segregation
3
RWA D Aggression .24*** -.25*** .35*** .32***
(.41) (-.37) (.46) (.41)
RWA3D Submission .14* -.41* .14* -.04 n.s.
(-.16) (-.33) (.37) (.29)
RWA3D .18*** -.09 n.s. .05 n.s. .26***
Conventionalism (.38) (-.29) (.30) (.39)
Note. Pay attention to the distinction between the highly significant zero-order correlations and the
partial regression coefficients.

contribute to the prediction of dependent variables, frequently authoritarian


aggression is the strongest predictor or even the only one. Occasionally one or
more dimensions fail to have a significant linear regressive effect on the depend-
ent variable. This is not attributable to dissimilar circumstances of data collection
or sampling, as the presented data stem from the same study.
While the successful demonstration of differential effects of the separate
dimensions is the main focus of this article, it can be only the first step in future
research. With a more reliable English version of a comprehensive RWA3D-R
Scale, it will be possible to draw consistent theoretical consequences. First, empir-
ical results support the plausible viewpoint that authoritarian aggression is the
strongest predictor for an attitude towards immigrants’ need for assimilation as a
prerequisite for their acceptance or toleration. As the demand for assimilation is
theoretically related to the submission to authorities and norms (conventionalism),
the results are concordant with the theoretical assumptions (Table 3). The demand
for segregation, however, has by definition a much more aggressive component.
Theoretically it should not be positively related to conventionalism, as social
norms in modern societies rather oppose the active discrimination of minorities.
Both theoretical suppositions are reflected in the empirical results.

General Discussion

The Benefit of Three-Dimensional Measurement

Using three separately operationalized indicators instead of one global score


leads to better fitting models.3 The dimensions of authoritarianism are—by
3
The comparison of a one-predictor versus a multiple-predictor approach reminds one of the contro-
versy about g-factors in intelligence research (e.g., Ree, Earles, & Teachout, 1994). The empirical
decision about the significance of special factors is not straightforward, as they are a statistical dis-
advantage and underestimated when entered after a global score. The presented approach of multi-
ple regression can on its part overestimate the parameters, but at least every effect is controlled for
the influence of the other predictors.
212 Funke

definition—highly correlated, but nevertheless show notable discriminant valid-


ity (interdimension discrepancy) that may differ depending on the construct under
examination (intradimension discrepancy). The presented results provide evi-
dence for the assumption that three-dimensional measurement of authoritarianism
is (1) possible and (2) promising, as scientific answers can be found to questions
that could not be asked before. This should be understood, however, only as the
first step on a long (and rough) road ahead.
Next Steps. The purpose of this article was to give theoretical reasons for the
benefit of fine-grained analyses in the quantitative assessment of authoritarianism.
The RWA3D Scale used to illustrate my argumentation is definitely suitable for
research, but has a few limitations that have to be overcome in the future. The
scales should be (1) longer (and consequently more reliable), (2) more sensitive
to changes over time (with trait-like as well as state-like subscales for experi-
mental designs), and (3) the item wording and content has to be updated in order
to change item characteristics from dramatically skewed to symmetric distribu-
tions. This process of construction, adaptation, and validation is in progress.
Cautionary Note. It is crucial to emphasize that the three dimensions are dis-
tinguishable but nevertheless highly correlated. This has methodological conse-
quences that are more than solely a mathematical detail. While in Confirmatory
Factor Analyses three-dimensional models are appropriate, this does not imply
that those correlated factors should be used as simultaneous exogenous variables
in complex structural models, as we are facing the problem of multicollinearity.
Because of their intercorrelation the three dimensions are collinear and partially
convey similar information. Therefore, it would be absolutely misleading to inter-
pret regression coefficients as measures of absolute association (as zero-order
correlations). They are measures of net contribution, controlling for the impact of
the other dimensions. Being aware of that fact helps to understand seemingly
paradox findings.
Is One-dimensional Measurement Obsolete? “A High is a High is a High.”
It would be a complete misreading of my argumentation as a plea for leaving
behind the Altemeyerian definition of RWA as the covariation of the three dimen-
sions submission, aggression, and conventionalism. The prototypical authoritar-
ian attitude continues to imply a high level of compliance with established
authorities (S), especially when those authorities oppress out-group members of
any kind (A), and both generalized attitudes are embedded in a personal ideology
that favors old-fashioned social norms perceived to be endorsed by the establish-
ment (C).
The covariation of all three dimensions is not only an empirical reality but is
also theoretically plausible due to the dialectical interplay of the dimensions.
There is neither a necessity nor a need to oppose this position. At least at high
levels of RWA, all of the three dimensions are necessary conditions for the iden-
tification of authoritarianism.
However, it could be illustrated in the aforementioned numerical example
that there is little “room to move” for combinations of the dimensions only at the
Dimensionality of RWA 213

extremes of low and high global RWA scores. In the wide-ranging grey area of
intermediate RWA-levels, however, far more combinations of the dimensions are
imaginable, implying totally different “phenotypes” of authoritarians (e.g., a dom-
inant person who likes to boss others around versus an opportunistic, yet agree-
able turncoat doing U-turns all the time but being unable to hurt a fly).
The consequence is that the unweighted sum (or arithmetic mean) of all item
responses (aggregated over all three subscales) continues to be the best available
approximation of a single indicator for authoritarianism. When researchers are
interested in the impact of the separate dimensions, however, they are encouraged
to use the subdimensions in the analyses.

Conclusion

The consequence of my argumentation can be generalized beyond the limits


(and limitations) of quantitative authoritarianism research with the RWA Scale. It
is vital for authoritarianism research in the next decade to bridge the gap between
reductionist personality psychology on the one hand and social psychology on the
other. In order to cope with those challenges it is helpful to clear up the scientific
workbench and to put a fine-grained instrument in the toolbox.

ACKNOWLEDGMENTS

The author thanks George Marcus and three anonymous reviewers for their
critical comments and is most grateful for the encouragement by numerous col-
leagues—especially Thomas Pettigrew and John Duckitt—to pursue the chal-
lenging idea of the presented position. Correspondence may be addressed to
Friedrich Funke (Social Psychology, Humboldtstr. 26, 07743 Jena/Germany).
E-mail: sff@uni-jena.de
214 Funke

APPENDIX: English Translation of the 12 German RWA3D-Items.

Order Item M SD
Aggression 2 What our country really needs instead of more 2.62 1.67
“civil rights” is a good stiff dose of law and
order.
8 What our country really needs is a strong, 3.11 1.88
determined Chancellor which will crush the
evil and set us on our right way again.
5 There is no such crime to justify capital 2.85 2.24
punishment.*
11 It is important to protect the rights of radicals 4.00 1.82
and deviants in all ways.*
Submission 6 Obedience and respect for authority are the 2.75 1.68
most important values children should learn.
12 The real keys to the “good life” are obedience, 2.34 1.43
discipline, and virtue.
3 The days when women are submissive should 1.37 .93
belong strictly in the past. A “woman’s place”
in society should be wherever she wants to be.*
9 It is good that nowadays young people have 2.05 1.18
greater freedom “to make their own rules”
and to protest against things they don’t like.*
Conventionalism 4 The withdrawal from tradition will turn out to 2.97 1.66
be a fatal fault one day.
10 Being virtuous and law-abiding is in the long 3.12 1.63
run better for us than permanently challenging
the foundation of our society.
1 People should develop their own personal 3.11 1.79
standards about good and evil and pay less
attention to the Bible and other old, traditional
forms of religious guidance.*
7 Homosexual long-term relationships should be 2.44 1.94
treated as equivalent to marriage.*
Note. Contrait items are flagged. Their mean values have been inverted. Items have been reorganized
for purposes of presentation. The actual presenting order is indicated.

REFERENCES

Adorno, T. W., Frenkel-Brunswik, E., Levinson, D. J., & Sanford, R. N. (1950). The authoritarian
personality. New York: Harper and Row.
Altemeyer, B. (1988). Enemies of freedom: Understanding right-wing authoritarianism. San
Francisco: Jossey-Bass.
Altemeyer, B. (1981). Right-Wing Authoritarianism. Winnipeg: University of Manitoba Press.
Altemeyer, B. (1998). The other “authoritarian personality.” Advances in Experimental Social Psy-
chology, 30, 47–92.
Dimensionality of RWA 215

Altemeyer, B. (1996). The authoritarian specter. Cambridge: Harvard University Press.


Altemeyer, B., & Hunsberger, B. E. (1992). Authoritarianism, religious fundamentalism, quest, and
prejudice. International Journal for the Psychology of Religion, 2, 113–133.
Bandalos, D. L., & Finney, S. J. (2001). Item parceling issues in structural equation modeling. In G.
A. Marcoulides & R. E. Schumacker (Eds.), New developments and techniques in structural
equation modeling (pp. 269–296). Mahwah, NJ: Lawrence Erlbaum Associates.
Berry, J. W. (1997). Immigration, acculturation, and adaptation. Applied Psychology: An International
Review, 46, 5–34.
Best, S. J., Krueger, B., Hubbard, C., & Smith, A. (2001). An assessment of the generalizability of
Internet surveys. Social Science Computer Review, 19, 131–145.
Bollen, K. A. (1986). Sample size and Bentler and Bonett’s nonnormed fit index. Psychometrika, 51,
375–377.
Bollen, K. A. (1989). Structural equations with latent variables. New York: John Wiley and Sons.
Borkenau, P., & Ostendorf, F. (1991). Ein Fragebogen zur Erfassung fünf robuster Persönlichkeits-
faktoren. Diagnostica, 37, 29–41.
Browne, M. W., & Cudeck, R. (1993). Alternative ways of assessing model fit. In K. A. Bollen &
J. S. Long (Eds.), Testing structural equation models (pp. 136–162). Newbury Park: SAGE.
Browne, M. W., & Cudeck, R. (1989). Single sample cross-validation indices for covariance struc-
tures. Multivariate Behavioral Research, 24, 445–455.
Buck-Morss, S. (1977). The Adorno legacy. Personality and Social Psychology Bulletin, 3, 707–713.
Byrne, B. M., & Campbell, T. L. (1999). Cross-cultural comparisons and the presumption of equiva-
lent measurement and theoretical structure: A look beneath the surface. Journal of Cross
Cultural Psychology, 30, 555–574.
Campbell, D. T., & Fiske, D. W. (1959). Convergent and discriminant validation by the multitrait-
multimethod matrix. Psychological Bulletin, 56, 81–105.
Chou, C. P., Bentler, P. M., & Satorra, A. (1991). Scaled test statistics and robust standard errors
for non-normal data in covariance structure analysis: A Monte Carlo study. British Journal of
Mathematical and Statistical Psychology, 44, 347–357.
Christie, R. (1991). Authoritarianism and related constructs. In J. P. Robinson & P. R. Shaver (Eds.),
Measures of personality and social psychological attitudes (pp. 501–571). San Diego:
Academic Press.
Christie, R., & Jahoda, M. (1954). Studies in the scope and method of “The Authoritarian Personal-
ity.” Glencoe, IL: Free Press.
Connors, J., & Heaven, P. C. (1987). Authoritarianism and just world beliefs. Journal of Social
Psychology, 127, 345–346.
Costa, P. T., & McCrae, R. R. (1992). The five-factor model of personality and its relevance to
personality disorders. Journal of Personality Disorders, 6, 343–359.
DeVellis, R. F. (1991). Scale development: Theory and applications. Newbury Park, CA: Sage
Publications.
Duckitt, J. H. (1992). Threat and authoritarianism: Another look. Journal of Social Psychology, 132,
697–698.
Duckitt, J. H., & Bizumic, B. (2002). Items to measure the three dimensions of Right-Wing
Authoritarianism. Unpublished manuscript, University of Auckland, New Zealand.
Duncan, L. E., Peterson, B. E., & Winter, D. G. (1997). Authoritarianism and gender roles: Toward a
psychological analysis of hegemonic relationships. Personality and Social Psychology Bulletin,
23, 41–49.
Eckhardt, W. (1991). Authoritarianism. Political Psychology, 12, 97–124.
Edwards, J. R., & Bagozzi, R. P. (2000). On the nature and direction of relationships between
constructs and measures. Psychological Methods, 5, 155–174.
216 Funke

Eid, M. (2000). A multitrait-multimethod model with minimal assumptions. Psychometrika, 65,


241–261.
Feather, N. T. (1998). Reactions to penalties for offences committed by the police and public citizens:
Testing a social-cognitive process model of retributive justice. Journal of Personality and Social
Psychology, 75, 528–544.
Feldman, S. (2001). Authoritarianism. In N. J. Smelser & P. B. Baltes (Eds.), International
Encyclopedia of the Social & Behavioral Sciences (p. 971). Amsterdam: Elsevier.
Feldman, S. (2003). Enforcing social conformity: A theory of authoritarianism. Political Psychology,
24, 41–74.
Feldman, S., & Stenner, K. (1997). Perceived threat and authoritarianism. Political Psychology,
18, 741–770.
Fromm, E. (1936). Sozialpsychologischer Teil. In M. Horkheimer, E. Fromm, & H. Marcuse (Eds.),
Studien über Autorität und Familie (pp. 77–135). Paris: Felix Alcan.
Funke, F. (1996). RWA96-G—German Adaptation of the Right-Wing Authoritarianism Scale by
Altemeyer. Retrieved 4 August 2001, from http://www.uni-jena.de/~sff/research/rwa96.pdf.
Funke, F. (2000). The three-dimensional structure of authoritarianism—necessity, sense and
Gestalt of adequate measurement models. Paper presented at the European Conference on
Personality, Krakov/Poland. Retrieved March 3, 2004, from http://www.uni-jena.de/~sff/
research/KRAKOW_A3_sec.pdf.
Funke, F. (2002). The three-dimensional structure of Right-Wing Authoritarianism. Paper presented
at the 25th Annual Scientific Meeting of the International Society of Political Psychology, Berlin.
Retrieved March 5, 2004, from http://www.uni-jena.de/~sff/research/ISPPnew.pdf.
Funke, F. (2003). Die dimensionale Struktur von Autoritarismus [The dimensional Structure of Author-
itarianism]. Doctoral dissertation, Friedrich-Schiller-University at Jena, Germany.
Haddock, G., & Zanna, M. P. (1998). Authoritarianism, values, and the favorability and structure of
antigay attitudes. In G. M. Herek (Ed.), Stigma and sexual orientation: Understanding preju-
dice against lesbians, gay men, and bisexuals (pp. 82–107). Thousand Oaks: Sage Publications.
Haddock, G., & Zanna, M. P. (1994). Preferring “housewives” to “feminists”: Categorization and the
favorability of attitudes toward women. Psychology of Women Quarterly, 18, 25–52.
Hall, R. J., Snell, A. F., & Foust, M. S. (1999). Item parcelling strategies in SEM: Investigating
the subtle effects of unmodeled secondary constructs. Organizational Research Methods, 2,
233–256.
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.
Henkel, J., Sheehan, E. P., & Reichel, P. (1997). Relation of police misconduct to authoritarianism.
Journal of Social Behavior and Personality, 12, 551–555.
Hoelter, J. W. (1983). The analysis of covariance structures: Goodness-of-fit-indices. Sociological
Methods and Research, 11, 325–344.
Horkheimer, M., Fromm, E., & Marcuse, H. (1936). Studien über Autorität und Familie. Paris: Felix
Alcan.
Hunsberger, B. E. (1996). Religious fundamentalism, right-wing authoritarianism, and hostility toward
homosexuals in non-Christian religious groups. International Journal for the Psychology of
Religion, 6, 39–49.
Hunsberger, B. E., Owusu, V., & Duck, R. (1999). Religion and prejudice in Ghana and Canada: Reli-
gious fundamentalism, right-wing authoritarianism and attitudes toward homosexuals and
women. International Journal for the Psychology of Religion, 9, 181–194.
Hyman, H., & Sheatsley, P. B. (1954). The Authoritarian Personality—a methodological
critique. In R. Christie & M. Jahoda (Eds.), Studies in the scope and method of “The
Authoritarian Personality.” Continuities in social research. (pp. 50–122). Westport, CT:
Greenwood Press.
Dimensionality of RWA 217

Jöreskog, K. G., & Sörbom, D. (1996). LISREL 8 User’s reference guide. Chicago: Scientific Soft-
ware International.
Jost, J. T., Glaser, J., Kruglanski, A. W., & Sulloway, F. J. (2003). Political conservatism as motivated
social cognition. Psychological Bulletin, 129, 339–375.
Kemmelmeier, M., Burnstein, E., & Peng, K. (1999). Individualism and authoritarianism shape atti-
tudes toward physician-assisted suicide. Journal of Applied Social Psychology, 29, 2613–2631.
Kenny, D. A., & Kashy, D. A. (1992). Analysis of the multitrait-multimethod matrix by confirmatory
factor analysis. Psychological Bulletin, 112, 165–172.
Kirscht, J. P., & Dillehay, R. C. (1967). Dimensions of Authoritarianism: A review of research and
theory. Lexington: University of Kentucky Press.
Kohn, P. M., & Mercer, G. W. (1973). Authoritarianism, rebelliousness, permissiveness about drugs,
and person perception. Proceedings of the Annual Convention of the American Psychological
Association, 309–310.
Leak, G. K., & Randall, B. A. (1995). Clarification of the link between right-wing authoritarianism
and religiousness: The role of religious maturity. Journal for the Scientific Study of Religion, 34,
245–252.
Lefcourt, H. M., Davidson, K., Shepherd, R., & Phillips, M. (1997). Who likes “Far Side” humor?
Humor: International Journal of Humor Research, 10, 439–452.
Marsh, H. W. (1989). Confirmatory factor analyses of multitrait-multimethod data: Many problems
and a few solutions. Applied Psychological Measurement, 13, 335–361.
Marsh, H. W., Byrne, B. M., & Craven, R. (1992). Overcoming problems in confirmatory factor analy-
ses of MTMM data: The correlated uniqueness model and factorial invariance. Multivariate
Behavioral Research, 27, 489–507.
Marsh, H. W., & Grayson, D. (1995). Latent variable models of multitrait-multimethod data. In R. H.
Hoyle (Ed.), Structural equation modelling: Concepts, issues, and applications (pp. 177–198).
Thousand Oaks: Sage Publications.
Martin, J. L. (2001). The authoritarian personality, 50 years later: What lessons are there for political
psychology? Political Psychology, 22, 1–26.
McClosky, H., & Zaller, J. (1984). The American ethos. Cambridge: Harvard University Press.
McCutcheon, L. E. (1999). The multidimensional sportspersonship orientations scale has psychome-
tric problems. Journal of Social Behavior and Personality, 14, 439–444.
Meloen, J. D. (1994). A critical analysis of forty years of authoritarianism research: Did theory testing
suffer from Cold War attitudes? In R. F. Farnen (Ed.), Nationalism, ethnicity, and identity: Cross-
national and comparative perspectives (pp. 127–165). New Brunswick, NJ: Transaction
Publishers.
Mischel, W., & Schopler, J. (1959). Authoritarianism and reactions to “sputniks.” Journal of Abnor-
mal and Social Psychology, 59, 142–145.
Montada, L., & Lerner, M. J. (1998). Responses to victimizations and belief in a just world. New York:
Plenum Press.
Nachtigall, C., Kroehne, U., Funke, F., & Steyer, R. (2003). (Why) Should we use SEM?—Pros and
cons of Structural Equation Modelling. Methods of Psychological Research Online, 8, 1–21.
Natoli, V. J. (2000). Authoritarianism as a correlate of retaliation against whistle-blowers. Disserta-
tion Abstracts International Section A: Humanities and Social Sciences, 61, 1938.
Peterson, B. E., Doty, R. M., & Winter, D. G. (1993). Authoritarianism and attitudes toward contem-
porary social issues. Personality and Social Psychology Bulletin, 19, 174–184.
Ree, M. J., Earles, J. A., & Teachout, M. S. (1994). Predicting job performance: Not much more than
g. Journal of Applied Psychology, 79, 518–524.
Reich, W. (1933). Massenpsychologie des Faschismus. Berlin/Kopenhagen: Verlag für Sexualpolitik.
Robinson, J. P., Shaver, P. R., & Wrightsman, L. S. (1991). Measures of personality and social
psychological attitudes. San Diego: Academic Press.
218 Funke

Rubin, Z., & Peplau, L. A. (1975). Who believes in a just world? Journal of Social Issues, 31, 65–89.
Salgado, J. F., & Iglesias, M. (1995). Estructura factorial de la Escala de Autoestima de Rosenberg:
Un análisis factorial confirmatorio. Pysichologica, 16, 441–454.
Satorra, A., & Bentler, P. M. (1988). Scaling corrections for chi-square statistics in covariance struc-
ture analysis. In American Statistical Association (Ed.), 1988 proceedings of the business and
economics section (pp. 308–318). Alexandria, VA: American Statistical Association.
Satorra, A. (1989). Alternative test criteria in covariance structure analysis: A unified approach.
Psychometrika, 54, 131–151.
Satorra, A., & Bentler, P. M. (2001). A scaled difference chi-square test statistic for moment structure
analysis. Psychometrika, 66, 507–514.
Schallow, J. R. (2000). A comparison of three approaches to constructing item parcels to improve
subject-to-parameter ratios in confirmatory factor analysis. Multivariate Experimental Clinical
Research, 12, 29–41.
Schwarzer, R. (1986). Evaluation of convergent and discriminant validity by use of structural equa-
tions. In A. Angleitner & J. S. Wiggins (Eds.), Personality assessment via questionnaires. Current
issues in theory and measurement (pp. 191–213). New York: Springer.
Sidanius, J., & Pratto, F. (1999). Social dominance: An intergroup theory of social hierarchy and
oppression. New York: Cambridge University Press.
Six, B., Wolfradt, U., & Zick, A. (2001). Autoritarismus und Soziale Dominanzorientierung als
generalisierte Einstellungen. Zeitschrift für Politische Psychologie, 9, 23–40.
Smith, A. G., & Winter, D. G. (2002). Right-wing authoritarianism, party identification, and attitudes
toward feminism in student evaluations of the Clinton-Lewinsky Story. Political Psychology, 23,
355–383.
Steiger, J. H. (2000). Point estimation, hypothesis testing, and interval estimation using the RMSEA:
Some comments and a reply to Hayduck and Glaser. Structural Equation Modeling, 7, 149–162.
Steyer, R., & Eid, M. (2001). Messen und Testen. [Measurement and Testing]. (2nd ed.) Berlin,
Germany: Springer.
Stone, W. F., Lederer, G., & Christie, R. (1993). Strength and weakness. The Authoritarian Person-
ality today. New York: Springer.
Tanaka, J. S. (1987). “How big is big enough?” Sample size and goodness of fit in structural equa-
tion models with latent variables. Child Development, 58, 134–146.
Tarr, H., & Lorr, M. (1991). A comparison of right-wing authoritarianism, conformity and conser-
vatism. Personality and Individual Differences, 12, 307–311.
Tomás, J. M., & Oliver, A. (1999). Rosenberg’s self-esteem scale: Two factors or method effects.
Structural Equation Modelling, 6, 84–98.
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., & Ægisdóttir, S. (2000). The gender belief system, authoritarianism, social domi-
nance orientation, and heterosexuals’ attitudes toward lesbians and gay men. Sex Roles, 42,
947–967.

Anda mungkin juga menyukai