graham -the moral stereotypes of liberals and conservatives

The Moral Stereotypes of Liberals and Conservatives:Exaggeration of Differences across the Political SpectrumJesse Graham1*, Brian A. Nosek2, Jonathan Haidt3

1 Psychology Department, University of Southern California, Los Angeles, California, United States of America, 2 Psychology Depatment, University of Virginia,

Charlottesville, Virginia, United States of America, 3 Stern School of Business, New York University, New York, New York, United States of America

Abstract

We investigated the moral stereotypes political liberals and conservatives have of themselves and each other. In reality,liberals endorse the individual-focused moral concerns of compassion and fairness more than conservatives do, andconservatives endorse the group-focused moral concerns of ingroup loyalty, respect for authorities and traditions, andphysical/spiritual purity more than liberals do. 2,212 U.S. participants filled out the Moral Foundations Questionnaire withtheir own answers, or as a typical liberal or conservative would answer. Across the political spectrum, moral stereotypesabout ‘‘typical’’ liberals and conservatives correctly reflected the direction of actual differences in foundation endorsementbut exaggerated the magnitude of these differences. Contrary to common theories of stereotyping, the moral stereotypeswere not simple underestimations of the political outgroup’s morality. Both liberals and conservatives exaggerated theideological extremity of moral concerns for the ingroup as well as the outgroup. Liberals were least accurate about bothgroups.

Citation: Graham J, Nosek BA, Haidt J (2012) The Moral Stereotypes of Liberals and Conservatives: Exaggeration of Differences across the Political Spectrum. PLoSONE 7(12): e50092. doi:10.1371/journal.pone.0050092

Editor: Liane Young, Boston College, United States of America

Received August 8, 2012; Accepted October 16, 2012; Published December 12, 2012

Copyright: ! 2012 Graham et al. This is an open-access article distributed under the terms of the Creative Commons Attribution License, which permitsunrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.

Funding: This work was supported by Project Implicit. The funders had no role in study design, data collection and analysis, decision to publish, or preparation ofthe manuscript.

Competing Interests: The authors have declared that no competing interests exist.

* E-mail: [email protected]

Introduction

‘‘The national Democratic Party is immoral to the core. Any

American who would vote for Democrats is guilty of

fostering the worst kind of degeneracy. The leaders of this

party are severely out of touch with mainstream, traditional

American values. They are crusaders for perversion, for

licentiousness, for nihilism and worse.’’

—Joseph Farah [1], World Net Daily

‘‘Republicans don’t believe in the imagination, partly

because so few of them have one, but mostly because it

gets in the way of their chosen work, which is to destroy the

human race and the planet. Human beings, who have

imaginations, can see a recipe for disaster in the making;

Republicans, whose goal in life is to profit from disaster and

who don’t give a hoot about human beings, either can’t or

won’t.’’

—Michael Feingold [2], Village Voice

For as long as there have been political rivalries there have beenunflattering stereotypes painted by each side about the other.These stereotypes go far beyond cliches about latte liberals andgun-rack conservatives; as the quotations above show, they ofteninclude the claim that the other side is immoral or downright evil.

Of course, evil is in the eye of the beholder, and liberal andconservative eyes seem to be tuned to different wavelengths ofimmorality. For conservatives, liberals have an ‘‘anything goes’’

morality that says everything should be permitted for the sake ofinclusion and diversity, no matter how bizarre or depraved (e.g.,[3]). For liberals, conservatives lack basic moral compassion,especially for oppressed groups, and take a perverse joy in seeingthe rich get richer while innocents suffer in poverty (e.g., [4]).These views may be caricatures, but they suggest that accusationsof immorality may differ in content depending on the ideologies ofthe source and the target. In this paper we use Moral FoundationsTheory [5] to investigate liberals’ and conservatives’ moral stereotypesof themselves and each other—that is, their expectations abouthow strongly typical partisans would endorse values related toeach of five intuitive moral foundations. Our study was designed toanswer three questions: 1. How accurate are these moralstereotypes? 2. Are they exaggerations of real differences in moralvalues? 3. Where on the political spectrum do we find the greatestaccuracy? Rather than examining general beliefs about theimmorality of the other side, we sought a finer resolution of themoral domain to provide the first identification of patterns ofinaccuracy for moral concerns.

Exaggeration and Accuracy in StereotypesAlthough the literature on stereotypes has tended to concentrate

on biases and inaccuracies, several reviews have noted theaccuracy of many social stereotypes in terms of real groupdifferences [6,7,8]. The notion that stereotypes could be exagger-ations of actual group differences was popularized by Allport [9] inThe Nature of Prejudice: ‘‘a stereotype is an exaggerated beliefassociated with a category’’ (p.191). Stereotypes have long beenthought of as motivated exaggerations both of stereotypical

PLOS ONE | www.plosone.org 1 December 2012 | Volume 7 | Issue 12 | e50092

characteristics (Irish people are drunk every day) and in overgen-eralization (Every Irish person is drunk every day).

A review by McCauley [10], however, found only weak supportfor stereotypes-as-exaggeration as a general cognitive process. Forinstance, McCauley & Stitt [11] found general accuracy with someunderestimation of group differences when White students wereasked to estimate characteristics of Black students. But in the casesof racial, gender and occupational groups McCauley [10] reviews,there may be motives to appear unprejudiced against outgroups,and these motives might counteract exaggeration tendencies. Incases where one does not wish to hide signs of intergroup hostility,motivational factors may have the opposite effect, increasingexaggeration and stereotyping.

This brings us to politics, where people are quite willing toreport their preferences for ingroups over outgroups (e.g., [12]),and sometimes even relish the opportunity. Social identity theory[13] has been applied to political partisans, positing a motivationto maximize distinctions between the political ingroup andoutgroup based on identifications with one’s own political party[14]. Examining the accuracy of stereotypes about the issuepositions of Democrats and Republicans, Judd and Park [15]found more exaggeration in the outgroup (vs. ingroup) stereotypesof either side; outgroup stereotype exaggeration was strongest forthose most identified with their ingroup, suggesting that partisansof either side exaggerate more than moderates and centrists.Although this work made use of moral issues, we have found nostudies looking specifically at the content of moral stereotypes, andhow such stereotypes might be driven by processes beyond simplepartisan outgroup derogation.

Moral Stereotyping along Five FoundationsMoral Foundations Theory was created to identify the moral

content areas most widely discussed in the anthropological andevolutionary literatures. The theory posits five best candidates forbeing the psychological ‘‘foundations’’ upon which moral virtuesand institutions can be socially constructed. The first twofoundations are Harm/care (involving intuitions of sympathy,compassion, and nurturance) and Fairness/reciprocity (includingnotions of rights and justice). These two foundations are generallyconcerned with the protection and fair treatment of individuals;they are therefore called the two ‘‘individualizing’’ foundations.The other three foundations, in contrast, are called the ‘‘binding’’foundations because they underlie moral systems in which peopleare bound into larger groups and institutions. (These labels are notmeant to imply that welfare and fairness concerns can never begroup-focused, or that the others can never be individual-focused;see [16]). These foundations are Ingroup/loyalty (supportingmoral obligations of patriotism and ‘‘us vs. them’’ thinking);Authority/respect (including concerns about traditions andmaintaining social order) and Purity/sanctity (including moraldisgust and spiritual concerns about treating the body as a temple).

Graham, Haidt, and Nosek [17] found that liberals endorsedthe individualizing foundations (Harm, Fairness) more thanconservatives did, whereas conservatives endorsed the bindingfoundations (Ingroup, Authority, Purity) more than liberals did.This pattern has been observed across a variety of samples andmethods, including self-report measures of (un)willingness toviolate the foundations for money, text analyses of sermons inliberal and conservative churches, content coding of life narratives,and facial muscle movements [18,19,20,21].

If this pattern is found so consistently, are people aware of thesedifferences? Research on partisan stereotypes [15], as well asresearch on naıve realism and the culture war [22], suggests thatthe two sides will overestimate their differences on specific issues.

Might they likewise exaggerate differences in fundamental moralconcerns, stereotyping their opponents as immoral/amoral mon-sters? Would these moral stereotypes be characterized by generalderogation of outgroup morality, or would there be morecomplexity or asymmetry to the stereotypes?

To examine the moral stereotypes that liberals and conserva-tives hold about each other, we took advantage of a methodintroduced by Dawes, Singer, and Lemons [23] of havingpartisans indicate the values of ‘‘typical’’ partisan group members,allowing comparison of these projections with the partisans’ actualanswers. Participants completed multiple versions of the MoralFoundations Questionnaire (MFQ; [16]). One version askedparticipants for their own responses; we refer to these as the‘‘actual’’ scores. The other two versions asked participants tocomplete the MFQ as a ‘‘typical liberal’’ would, or as a ‘‘typicalconservative’’ would; we refer to these as the ‘‘moral stereotype’’scores. These versions allow us to assess moral stereotypes aboutliberals and conservatives, and to quantify their accuracy bycomparing them to the responses people gave for themselves.

Regarding our first research question (Are moral stereotypesaccurate?), because of the pervasiveness of the actual liberal-conservative differences, we predicted that participants would, onaverage, correctly guess that liberals value the individualizingfoundations more than conservatives do, and that conservativesvalue the binding foundations more than liberals do. Regardingour second question (Are these stereotypes exaggerations of realgroup differences?), although McCauley [10] found only weakevidence for a general cognitive process of stereotypes-as-exaggeration, we expected that the hostility between liberals andconservatives could create motivations to exaggerate the existinggroup differences. It is even possible that liberals and conservativeswould exaggerate the moral concerns of their own group, not justthe outgroup, perhaps as motivation to further distinguish theirgroup from the other [14]. Regarding our third question (Who ismost accurate?) we find reasons in the literature to generate threehypotheses, among which we hoped to adjudicate:

1) Moderates most accurate. Studies on ideological polarization (e.g.,[24,25]), the ideological extremity hypothesis [e.g., 26,27,28],and naıve realism [22] suggest a symmetrical exaggeration ofdifferences when liberals and conservatives try to look at theworld through the eyes of the other. Partisans should distortequally (presumably by underestimating their opponents’moral concerns) because both sides think the other side doesnot truly care about morality. On this view, politicalmoderates should be the most accurate, morally stereotypingliberals and conservatives the least.

2) Liberals most accurate. Social psychological work on conserva-tism [see 29, 30 for meta-analytic reviews] shows relationsbetween conservatism or authoritarianism and mentalrigidity, intolerance, and close-mindedness. Similarly, Carteret al. [31] found that acceptance of stereotyping was highestin individuals with conservative traits. These findings suggestthat conservatives might be more threatened and less able tosee the world from an alternate moral standpoint, andtherefore more motivated to stereotype liberals than vice-versa.

3) Conservatives most accurate. Moral Foundations Theory suggeststhat liberals may have a harder time understandingconservatives’ morality than vice-versa. If liberals don’tintuitively feel what could be considered moral about Ingroup(racism?), Authority (oppression?), and Purity (sexual Puri-tanism?), then they may be forced to conclude thatconservatives simply don’t care about morality—specifically,

Moral Stereotypes of Liberals and Conservatives


Kevin Deutsch

Kevin Deutsch

Kevin Deutsch

that conservatives don’t care about Harm and Fairness,because they support policies that seem to hurt and cheatpeople for no morally good reason.

Following the existing stereotype literature, we consider the firsthypothesis to be the default prediction: if the results only showoutgroup derogation by partisans about each other, then moralstereotypes are no different than other forms of stereotyping.However, if the results show asymmetrical inaccuracies (hypoth-eses 2 and 3), inaccuracies about the ingroup as well as theoutgroup, or overestimations as well as underestimations of moralvalues, then this would suggest that moral stereotypes involvenovel psychological processes beyond the well-understood inter-group stereotyping processes driving exaggeration of outgroupcharacteristics.

Methods

ParticipantsThe participants were 2,212 visitors (62% female; median age

28; only U.S. residents or citizens) to ProjectImplicit.org, wherethey were randomly assigned to this study. All participants in theresearch pool had previously filled out demographic information,including sex, age, and political identity (7-point scale, stronglyliberal to strongly conservative). 1,174 participants self-identifiedusing one of the three liberal options, 538 chose the ‘‘moderate’’midpoint, and 500 chose one of the three conservative options.Data from 77 participants were excluded because of high ratingson the catch item of the MFQ; removal of these participants didnot significantly alter any of the results. The study was approvedby the University of Virginia Institutional Review Board.

MaterialsThe MFQ consists of two parts, moral relevance and moral

judgments. In the relevance part, participants indicate the moralrelevance of foundation-related concerns (e.g., ‘‘whether or notsomeone did something disgusting,’’ for Purity) on a 6-point scale,from never relevant to always relevant. In the judgments section,participants rated their agreement with foundation-related state-ments (e.g., ‘‘It is more important to be a team player than toexpress oneself,’’ for Ingroup) on a 6-point scale, from stronglydisagree to strongly agree. Alphas for the foundation subscaleswere .71 (Harm), .69 (Fairness), .58 (Ingroup), .73 (Authority), and.83 (Purity).

ProcedureTo keep the study session brief and repetitiveness to a

minimum, we capitalized on the power of a large sample with aplanned missingness design [32]. Participants were randomlyassigned to complete four of six possible questionnaires: 2 (moralrelevance or moral judgments), by 3 (answered as oneself, as a‘‘typical liberal’’, or as a ‘‘typical conservative’’). (Results forquestionnaires answered as oneself are reported in Graham, et al.[17], Study 2. Participants also completed an Implicit AssociationTest that is not relevant for this report.) Because participantscompleted four out of the six possible measures, all of our 2,212participants completed two to four measures as they thought atypical political partisan would complete them.

Instructions for the liberal [and conservative] versions of themoral relevance measures read as follows:

When A TYPICAL LIBERAL [CONSERVATIVE] de-cides whether something is right or wrong, to what extent

are the following considerations relevant to the liberal’s

[conservative’s] thinking? Remember, instead of selecting

your own answers, answer all questions as a typical liberal

[conservative].

Instructions for the moral judgments measure read as follows:

Please read the following statements and indicate the extent

to which A TYPICAL LIBERAL [CONSERVATIVE]

would agree or disagree. Remember, instead of selecting

your own answers, answer all questions as a typical liberal

[conservative].

Comparison datasetsTo gauge the accuracy of participants’ predictions of ‘‘typical’’

liberal and conservative responses, we needed a standard ofcomparison. The most obvious comparisons were the actual ratingsprovided by the liberals and conservatives in our sample, whenthey were asked to answer as themselves. This was indeed our firstcomparison. However, it is not ideal because our sample is notrepresentative of the national population. For instance, our sampleof conservatives contains a higher proportion of self-describedslight conservatives than a representative population would. Wetherefore created a second comparison dataset by selecting theactual responses of self-reported extreme liberals and conservatives(the two endpoints of our 7-point politics measure). If the moralstereotypes are equivalent or stronger than these extremes, thenthey are likely to be exaggerations compared to the average liberalor conservative in the general population. A second reason forincluding the extreme comparisons is that people may imagine a‘‘typical’’ liberal/conservative to be a party-line prototype ratherthan an average partisan, and so accuracy may be better measuredin terms of extremes than averages.

To further increase confidence in our exaggeration interpreta-tions, we also obtained scores for a short-form MFQ collectedfrom a nationally-representative sample [33]. This dataset is theresult of a random-digit-dialing survey given to 1,001 individualsby Knowledge Networks. The two samples had four items incommon for every foundation except Ingroup, which had one itemin common. Comparisons between the moral stereotypes and thisnationally-representative dataset include only the items commonto both datasets.

Results

We measured and analyzed accuracy at the level of moralfoundations subscores, aggregates of 4–5 items each; this allowedus to capture accuracy in terms of mean foundational concerns, aswell as relative rankings of the five foundations. For the ten MFQsubscores (five foundations measured by relevance and judgmentssubscales) we compared predicted (moral stereotype) scoresanswered as a typical liberal or typical conservative to fourcriteria: (a) the actual liberal/conservative means in the currentsample, (b) the actual means for extreme liberals/conservatives inthe current sample, (c) the actual liberal/conservative means in therepresentative sample, and (d) the actual means for extremeliberals/conservatives in the representative sample.

As an example, the mean Harm-relevance score for allparticipants answering as a typical conservative was 2.46, with astandard deviation of 1.11 (see Table 1). The actual mean ofconservatives in the sample was 3.43 (SD .95), meaning thatpeople on average underestimated how morally relevant conser-vatives would find Harm concerns, t(477.53) = 213.52, p,.001,



d = 21.24. We compared such overall moral stereotype scores(using the entire sample) as well as the moral stereotypes held byliberals, moderates, and conservatives separately. Means for allcomparisons (ten foundation subscores answered as a typicalliberal and typical conservative, compared to the four comparisoncriteria) are available in Table 1, and the comparison statistics areavailable in Table 2 and a supplemental spreadsheet; thespreadsheet calculates t, df, and d for each comparison (seeexample above) using formulas that do not assume equal variancesor Ns. Below we present meta-analytic summaries of thesecomparisons. We organize the results around answers to ourthree central questions.

1. Are the moral stereotypes accurate with regard to the direction of liberal-conservative differences in the foundations? Yes. For both relevance andjudgment items, answers as a typical liberal yielded higher scoreson Harm and Fairness than answers as a typical conservative(ts.23.83, ps,.001, ds.1.00), and lower scores on Ingroup,Authority, and Purity (ts,215.76, ps,.001, ds.0.65). Theseshowed directional accuracy compared to the real groupdifferences found both in this study (see below) and in previousresearch: liberals endorse individual-focused moral concerns morethan conservatives do, and conservatives endorse group-focusedmoral concerns more than liberals do.

2. Are these stereotypes exaggerations of the real group differences? Yes.Figure 1 shows the average conservative-liberal differences foreach foundation, comparing the moral stereotypes (answered astypical partisans) to the actual differences found in our fourcomparison criteria (current sample means, current sampleextremes, representative sample means, representative sampleextremes). For all of the measures, foundation differences weresimilar across formats (relevance and judgments), and so for clarityof presentation the two MFQ subscales are combined in Figure 1.Differences were calculated as follows: the overall moral founda-tion means for answered-as-typical-liberal versions were subtractedfrom the overall means for the same scores answered as a typicalconservative. Differences between the actual means of liberals andconservatives, and between the actual means of extreme liberalsand extreme conservatives, were calculated the same way (seeTable 3).

As both the top panel (comparisons to current sample) andbottom panel (comparisons to representative sample) of Figure 1show, moral stereotypes exaggerated the liberal-conservativedifferences in line with Moral Foundations Theory. Not onlywere the moral stereotypes about typical partisans more differentfrom each other (average absolute difference 1.41, range 0.58–2.12) than the actual MFQ scores of liberals and conservatives(average absolute difference = 0.57, range 0.12–1.41), they were asdifferent or even moreso than the actual scores of extremepartisans (average absolute difference = 0.98, range 0.06–1.91; seeFigure 1 and Table 3). That is, participants’ beliefs about the‘‘typical’’ liberal and conservative were even more polarized thanthe actual polarization between extreme liberals and conservatives.

3. Who is most accurate? It depends on the type of morality. Comparisonsto actual group means were also made separately for the moralstereotypes held by liberals, moderates, and conservatives. Thisallows us to address our third research question about who is mostaccurate when answering as a typical liberal or typical conserva-tive. Statistics and effect sizes for each of these comparisons (thethree groups’ moral stereotypes about typical liberals andconservatives compared to the four actual group criteria, for fivefoundations, gauged by relevance and judgments measures) werecalculated (see Table 2). Here we meta-analytically summarize thecomparisons using ranges and averages of effect sizes, gaugingaccuracy in terms of differences from the current sample means

and (using only items common to both datasets) the representativesample means.

3a. Conservatives were most accurate about the individual-focused moralconcerns of either side, and liberals were least accurate. Compared to actualgroup means of either data set, moral stereotypes about the typicalconservative showed substantial underestimation of conservatives’Harm and Fairness concerns. Liberals tended to underestimate themost (average d = 2.98, 21.50#ds#2.41), followed by moderates(average d = 2.48, 2.79#ds#2.08); conservatives underestimatedthe individualizing concerns of the typical conservative the least(average d = 2.34, 2.55#ds#2.11), but they too underestimatedtheir own group’s Harm and Fairness concerns in everycomparison with actual conservative scores.

Stereotypes about the Harm and Fairness concerns of thetypical liberal tended to be more accurate as compared to actualliberal scores in the two datasets. Here again conservatives werethe most accurate, only slightly underestimating liberal individu-alizing concerns (average d = 2.08, 2.66#ds#.26), followed bymoderates, who underestimated slightly more (average d = 2.12,2.61#ds#.30). Liberals were the least accurate about their owngroup’s individualizing concerns, overestimating them on average(average d = .40, 2.11#ds#.80).

3b. Moderates were most accurate about the group-focused moral concerns ofeither side, and liberals were least accurate. Stereotypes about theIngroup, Authority, and Purity concerns of the typical conserva-tive tended to be overestimations compared to the actual groupmeans in both datasets. Here again liberals were the least accurate,overestimating conservative binding concerns the most (averaged = .55, .03#ds#1.01), followed by conservatives, who alsooverestimated their own group’s binding concerns (averaged = .34, 2.22#ds#.70); moderates were the most accurate(average d = .28, 2.14#ds#.66), but they too overestimated thebinding concerns when answering as a typical conservative.

Stereotypes about the typical liberal, on the other hand, tendedto underestimate the binding moral concerns actual liberalsreported. Here again liberals were the least accurate, underesti-mating their own binding concerns the most (average d = 2.62,21.19#ds#2.11), followed by conservatives (average d = 2.46,2.90#ds#.18). Moderates were the most accurate (averaged = 2.17, 2.63#ds#.43), but also underestimated the bindingconcerns when answering as a typical liberal.

3c. Liberals exaggerate moral differences the most. Means for the threegroups’ moral stereotypes about the typical liberal and typicalconservative are shown compared to the real group means (solidblack lines) in Figure 2. As both of the top panels (current samplecomparison) and both of the bottom panels (representative samplecomparison) show, participants across the political spectrumtended to exaggerate the liberal-conservative differences, asevidenced by the steeper slopes of the prediction lines as comparedto the actual lines. This exaggeration of differences is an effect ofoverestimating liberals’ individualizing concerns and underesti-mating their binding concerns, and overestimating conservatives’binding concerns and underestimating their individualizingconcerns. All four panels of Figure 2 show that liberals exaggeratedifferences the most (lines representing moral stereotypes held byliberals have the steepest slopes); the figure also shows that thelargest inaccuracies were liberal underestimations of the individ-ualizing concerns of the typical conservative. Overall exaggerationof moral differences (operationalized as overestimating conserva-tive binding concerns, underestimating conservative individualiz-ing concerns, and doing the opposite for liberals) is plotted acrossthe full ideological spectrum in Figure 3.



Kevin Deutsch

Kevin Deutsch

Kevin Deutsch

Kevin Deutsch

Kevin Deutsch

Kevin Deutsch

Ta

ble

1.

Pre

dic

ted

and

actu

alM

ora

lFo

un

dat

ion

sQ

ues

tio

nn

aire

sub

sco

rem

ean

san

dst

and

ard

dev

iati

on

s.

An

swe

rsa

sT

yp

ica

lL

ibe

ral

Act

ua

lL

ibe

ral

an

swe

rsA

nsw

ers

as

Ty

pic

al

Co

nse

rva

tiv

eA

ctu

al

Co

nse

rva

tiv

ea

nsw

ers

All

Lib

era

lsE

xtr

em

eL

ibe

rals

All

Co

nse

rva

tiv

es

Ex

tre

me

Co

nse

rva

tiv

es

Me

an

SD

Me

an

SD

Me

an

SD

Me

an

SD

Me

an

SD

Me

an

SD

Rel

evan

ce:

Har

m3.

770.

934.

000.

804.

110.

802.

461.

113.

430.

953.

231.

11

Fair

nes

s3.

870.

903.

780.

824.

000.

762.

291.

073.

110.

913.

111.

12

Ing

rou

p2.

530.

892.

600.

942.

400.

953.

120.

902.

720.

982.

461.

17

Au

tho

rity

2.36

0.86

2.44

0.84

2.26

0.86

3.41

0.84

2.88

0.77

3.04

0.87

Pu

rity

2.06

1.00

2.33

0.97

2.14

1.02

3.38

0.97

3.02

0.95

3.14

1.21

Jud

gm

ents

:H

arm

3.72

0.92

3.53

0.88

3.64

0.90

2.72

1.10

2.98

0.92

2.92

0.95

Fair

nes

s3.

790.

873.

760.

794.

000.

752.

551.

023.

050.

782.

900.

91

Ing

rou

p1.

810.

811.

850.

881.

730.

842.

960.

892.

320.

812.

520.

95

Au

tho

rity

2.05

1.09

2.39

0.99

2.05

0.97

4.08

0.94

3.58

0.74

3.93

0.74

Pu

rity

1.62

1.12

1.63

1.09

1.38

1.13

3.74

1.12

3.04

1.01

3.27

1.03

An

swe

rsa

sT

yp

ica

lL

ibe

ral

Act

ua

lL

ibe

ral

an

swe

rsA

nsw

ers

as

Ty

pic

al

Co

nse

rva

tiv

eA

ctu

al

Co

nse

rva

tiv

ea

nsw

ers

All

Lib

era

lsE

xtr

em

eL

ibe

rals

All

Co

nse

rva

tiv

es

Ex

tre

me

Co

nse

rva

tiv

es

Me

an

SD

Me

an

SD

Me

an

SD

Me

an

SD

Me

an

SD

Me

an

SD

Rel

evan

ce:

Har

m3.

811.

033.

770.

964.

310.

532.

351.

223.

431.

103.

571.

16

Fair

nes

s3.

851.

033.

570.

963.

920.

702.

251.

223.

221.

043.

351.

15

Ing

rou

p2.

361.

272.

931.

232.

921.

383.

121.

303.

181.

333.

391.

22

Au

tho

rity

2.34

1.13

2.95

0.95

2.48

1.15

3.81

0.98

3.30

1.03

3.54

1.00

Pu

rity

1.73

1.16

2.53

1.22

2.17

1.40

3.52

1.12

3.16

1.19

3.54

1.13

Jud

gm

ents

:H

arm

3.73

1.15

3.23

1.03

3.50

1.11

2.32

1.38

2.78

1.16

2.84

1.17

Fair

nes

s4.

181.

013.

950.

814.

130.

773.

071.

303.

630.

993.

411.

17

Au

tho

rity

2.39

1.27

3.40

0.96

2.96

1.17

4.26

0.98

4.00

0.79

4.11

0.95

Pu

rity

1.72

1.38

2.71

1.18

2.10

1.41

3.64

1.21

3.53

0.98

4.01

0.95

Not

e.To

pp

anel

sho

ws

pre

dic

ted

and

actu

alan

swer

sfo

rth

est

ud

ysa

mp

le,a

nd

bo

tto

mp

anel

sho

ws

the

sam

efo

rth

eco

mp

aris

on

toth

en

atio

nal

ly-r

epre

sen

tati

ved

atas

et,u

sin

go

nly

item

sco

mm

on

tob

oth

dat

aset

s(n

oit

ems

inco

mm

on

for

Ing

rou

pju

dg

men

ts).

Sam

ple

size

sfo

rea

chst

atis

tic,

asw

ell

asp

red

icte

d‘‘t

ypic

al’’

answ

ers

bro

ken

do

wn

by

liber

als,

mo

der

ates

,an

dco

nse

rvat

ives

,ca

nb

efo

un

din

the

sup

ple

men

t.d

oi:1

0.13

71/j

ou

rnal

.po

ne.

0050

092.

t001



Ta

ble

2.

Stat

isti

cal

com

par

iso

ns

of

actu

alan

dp

red

icte

dan

swer

sfo

rlib

eral

s,m

od

erat

es,

and

con

serv

ativ

es.

Ty

pic

al

Lib

era

la

nsw

ers

com

pa

red

tosa

mp

lea

ve

rag

e:

Ty

pic

al

Co

nse

rva

tiv

ea

nsw

ers

com

pa

red

tosa

mp

lea

ve

rag

e:

Lib

Mo

dC

on

Lib

Mo

dC

on

td

fd

td

fd

td

fd

td

fd

td

fd

td

fd

Rel

evan

ce:

Har

m2

1.89

1281

.15

20.

112

6.33

427.

362

0.61

26.

4738

5.75

20.

662

17.9

357

1.72

21.

502

9.08

532.

892

0.79

25.

1253

5.73

20.

44

Fair

nes

s5.

2212

70.7

70.

292

2.65

435.

512

0.25

0.14

418.

730.

012

15.9

958

0.92

21.

332

7.44

526.

332

0.65

23.

8153

2.36

20.

33

Ing

rou

p2

3.44

1253

.77

20.

191.

2252

0.94

0.11

0.49

481.

330.

047.

7246

9.87

0.71

5.12

546.

990.

441.

8854

3.74

0.16

Au

tho

rity

21.

9112

64.1

72

0.11

1.09

464.

440.

102

3.85

441.

872

0.37

9.82

572.

810.

827.

4152

7.65

0.65

6.35

541.

440.

55

Pu

rity

24.

7212

78.1

52

0.26

20.

7749

1.63

20.

072

7.33

451.

772

0.69

6.53

531.

090.

572.

3053

9.96

0.20

3.74

545.

000.

32

Jud

gm

ents

:H

arm

6.51

1286

.48

0.36

0.45

462.

670.

041.

2847

5.18

0.12

25.

1662

6.68

20.

412

0.90

524.

022

0.08

21.

2550

5.71

20.

11

Fair

nes

s6.

0412

79.0

40.

342

3.74

428.

032

0.36

23.

7348

1.93

20.

342

9.37

700.

272

0.71

24.

4751

2.56

20.

392

3.67

503.

772

0.33

Ing

rou

p2

3.79

1270

.49

20.

211.

5257

6.13

0.13

2.06

551.

540.

1812

.13

580.

371.

017.

5952

9.00

0.66

5.83

521.

190.

51

Au

tho

rity

28.

5312

84.7

02

0.48

20.

3544

5.33

20.

032

4.96

448.

142

0.47

8.66

670.

830.

674.

7747

5.30

0.44

7.31

523.

640.

64

Pu

rity

23.

8712

78.1

92

0.22

4.74

494.

390.

431.

1046

8.42

0.10

10.5

158

0.14

0.87

4.47

509.

340.

407.

9952

3.15

0.70

Ty

pic

al

Lib

era

la

nsw

ers

com

pa

red

tosa

mp

lee

xtr

em

e:

Ty

pic

al

Co

nse

rva

tiv

ea

nsw

ers

com

pa

red

tosa

mp

lee

xtr

em

e:

Lib

Mo

dC

on

Lib

Mo

dC

on

td

fd

td

fd

td

fd

td

fd

td

fd

td

fd

Rel

evan

ce:

Har

m2

3.01

326.

802

0.33

25.

7847

8.58

20.

532

5.87

461.

512

0.55

26.

0436

.17

22.

012

2.93

41.8

42

0.91

21.

2040

.97

20.

38

Fair

nes

s0.

1133

9.27

0.01

24.

6247

5.86

20.

422

2.60

460.

962

0.24

26.

3735

.97

22.

122

3.21

41.1

92

1.00

21.

6840

.26

20.

53

Ing

rou

p0.

3730

4.77

0.04

3.19

434.

430.

312.

5943

3.84

0.25

4.96

34.9

31.

683.

8038

.78

1.22

2.36

38.4

00.

76

Au

tho

rity

1.50

308.

120.

172.

9545

6.37

0.28

20.

7644

7.19

20.

072.

8536

.41

0.95

2.27

42.6

00.

691.

8240

.53

0.57

Pu

rity

20.

8331

6.98

20.

091.

3944

0.46

0.13

23.

5944

0.19

20.

341.

9935

.28

0.67

0.44

39.5

40.

141.

0838

.00

0.35

Jud

gm

ents

:H

arm

2.96

352.

950.

312

0.96

484.

702

0.09

20.

3548

2.81

20.

032

1.99

55.7

42

0.53

20.

0967

.70

20.

022

0.27

67.9

32

0.07

Fair

nes

s0.

2137

1.54

0.02

25.

8549

1.58

20.

532

6.25

490.

832

0.56

23.

2455

.83

20.

872

1.22

64.7

22

0.30

20.

8263

.53

20.

21

Ing

rou

p2

0.89

352.

202

0.09

2.90

451.

040.

273.

2846

4.37

0.30

4.15

52.3

51.

152.

5160

.93

0.64

1.62

59.5

30.

42

Au

tho

rity

21.

6237

7.07

20.

173.

1749

3.39

0.29

20.

3149

4.19

20.

031.

5058

.22

0.39

0.14

82.6

60.

031.

0865

.07

0.27

Pu

rity

0.34

338.

810.

046.

0647

1.20

0.56

3.17

483.

640.

293.

5354

.63

0.96

1.06

72.2

40.

253.

0258

.37

0.79



Ta

ble

2.

Co

nt.

Ty

pic

al

Lib

era

la

nsw

ers

com

pa

red

tore

pre

sen

tati

ve

sam

ple

av

era

ge

:T

yp

ica

lC

on

serv

ati

ve

an

swe

rsco

mp

are

dto

rep

rese

nta

tiv

esa

mp

lea

ve

rag

e:

Lib

Mo

dC

on

Lib

Mo

dC

on

td

fd

td

fd

td

fd

td

fd

td

fd

td

fd

Ty

pic

al

Lib

era

la

nsw

ers

com

pa

red

tore

pre

sen

tati

ve

sam

ple

av

era

ge

:T

yp

ica

lC

on

serv

ati

ve

an

swe

rsco

mp

are

dto

rep

rese

nta

tiv

esa

mp

lea

ve

rag

e:

Lib

Mo

dC

on

Lib

Mo

dC

on

td

fd

td

fd

td

fd

td

fd

td

fd

td

fd

Rel

evan

ce:

Har

m2.

4639

1.36

0.25

21.

0851

5.34

20.

092

1.45

488.

152

0.13

217

.92

680.

002

1.37

28.

8055

0.21

20.

752

6.56

560.

312

0.55

Fair

nes

s6.

6340

1.81

0.66

0.07

509.

870.

011.

3149

5.63

0.12

217

.88

699.

292

1.35

28.

4853

3.44

20.

732

3.78

538.

532

0.33

Ing

rou

p2

7.85

402.

022

0.78

23.

2850

9.00

20.

292

3.97

493.

612

0.36

0.33

639.

170.

032

0.43

571.

932

0.04

22.

5857

0.16

20.

22

Au

tho

rity

27.

3046

1.20

20.

682

4.64

506.

782

0.41

29.

2648

8.12

20.

848.

0263

8.68

0.63

4.96

579.

520.

416.

5458

9.20

0.54

Pu

rity

210

.03

394.

042

1.01

24.

4050

0.56

20.

392

9.09

495.

012

0.82

6.73

619.

980.

541.

8357

6.95

0.15

1.90

584.

480.

16

Jud

gm

ents

:H

arm

8.37

435.

830.

803.

3851

2.00

0.30

2.95

513.

680.

262

7.08

778.

472

0.51

21.

7252

3.85

20.

152

3.12

510.

302

0.28

Fair

nes

s6.

6444

0.90

0.63

0.48

491.

210.

042

0.31

503.

192

0.03

29.

2482

7.57

20.

642

3.21

502.

352

0.29

23.

5646

8.95

20.

33

Au

tho

rity

212

.91

523.

462

1.13

27.

0149

4.97

20.

632

10.0

649

8.34

20.

904.

6980

4.21

0.33

2.34

473.

330.

223.

8552

3.87

0.34

Pu

rity

212

.84

462.

812

1.19

24.

8451

2.59

20.

432

7.61

515.

712

0.67

2.78

797.

420.

202

1.58

475.

992

0.14

1.50

534.

260.

13

Ty

pic

al

Lib

era

la

nsw

ers

com

pa

red

tore

pre

sen

tati

ve

sam

ple

ex

tre

me

:T

yp

ica

lC

on

serv

ati

ve

an

swe

rsco

mp

are

dto

rep

rese

nta

tiv

esa

mp

lee

xtr

em

e:

Lib

Mo

dC

on

Lib

Mo

dC

on

td

fd

td

fd

td

fd

td

fd

td

fd

td

fd

Rel

evan

ce:

Har

m2

2.17

30.6

42

0.78

22.

7951

.19

20.

782

2.85

55.9

52

0.76

27.

9140

.32

22.

492

4.55

47.8

22

1.32

23.

6846

.42

21.

08

Fair

nes

s0.

5928

.32

0.22

21.

4540

.00

20.

462

1.02

38.8

02

0.33

27.

7840

.17

22.

462

4.33

47.4

72

1.26

22.

3146

.83

20.

67

Ing

rou

p2

2.96

26.4

12

1.15

21.

3329

.51

20.

492

1.58

30.7

82

0.57

20.

8542

.26

20.

262

1.12

50.2

62

0.32

22.

1849

.95

20.

62

Au

tho

rity

20.

4626

.68

20.

180.

0830

.42

0.03

21.

7630

.62

20.

641.

8941

.58

0.59

1.01

48.0

80.

291.

7946

.24

0.53

Pu

rity

22.

2426

.27

20.

882

0.44

28.5

32

0.16

22.

4828

.84

20.

920.

7741

.46

0.24

21.

0048

.96

20.

292

1.05

47.3

42

0.31

Jud

gm

ents

:H

arm

1.87

26.8

90.

720.

2630

.71

0.09

0.14

31.6

80.

052

3.00

43.6

92

0.91

21.

0351

.88

20.

292

1.71

51.3

62

0.48

Fair

nes

s1.

4327

.50

0.54

20.

5936

.67

20.

202

0.92

35.6

62

0.31

22.

5042

.79

20.

762

0.34

49.4

92

0.10

20.

5250

.72

20.

15

Au

tho

rity

22.

8127

.16

21.

082

1.07

31.5

92

0.38

22.

2431

.74

20.

801.

1342

.01

0.35

0.37

51.1

30.

101.

0246

.02

0.30

Pu

rity

22.

3826

.72

20.

920.

1430

.10

0.05

20.

9329

.83

20.

342

1.28

44.8

42

0.38

22.

8559

.59

20.

742

2.05

50.3

02

0.58

Not

e.To

pp

anel

com

par

esp

red

icte

dan

dac

tual

answ

ers

for

the

stu

dy

sam

ple

,an

db

ott

om

pan

elco

mp

ares

pre

dic

tio

ns

toac

tual

answ

ers

inth

en

atio

nal

ly-r

epre

sen

tati

ved

atas

et,u

sin

go

nly

item

sco

mm

on

tob

oth

dat

aset

s(n

oit

ems

inco

mm

on

for

Ing

rou

pju

dg

men

ts).

Lib

=Li

ber

alp

arti

cip

ants

,Mo

d=

Mo

der

ate

par

tici

pan

ts,C

on

=C

on

serv

ativ

ep

arti

cip

ants

.Fo

rmu

las

use

dto

calc

ula

tet,

df,a

nd

dw

ith

ou

tas

sum

ing

equ

alsa

mp

lesi

zes

or

vari

ance

sca

nb

efo

un

din

the

sup

ple

men

t.d

oi:1

0.13

71/j

ou

rnal

.po

ne.

0050

092.

t002



Discussion

Results indicate that people at all points on the politicalspectrum are at least intuitively aware of the actual differences inmoral concerns between liberals and conservatives: they correctlypredicted that liberals would care more than conservatives aboutthe two individualizing foundations and that conservatives wouldcare more than liberals about the three binding foundations. Theresults also confirm previous studies of partisan misperception [24]

by showing that, in general, people overestimate how dramaticallyliberals and conservatives differ. Remarkably, people even morallystereotype their own ingroup, with liberals overestimating liberals’strong individualizing concerns and underestimating liberals’ weakbinding concerns, and conservatives exaggerating conservatives’moral concerns in the opposite directions.

Our results go beyond previous studies, however, in finding andexplaining an otherwise puzzling result: liberals were the leastaccurate. We presented three competing hypotheses about

Figure 1. Comparisons of moral stereotypes to actual conservative-liberal differences in moral foundation endorsement.doi:10.1371/journal.pone.0050092.g001



Ta

ble

3.

Act

ual

con

serv

ativ

e-lib

eral

diff

eren

ces

com

par

edto

tho

sep

red

icte

db

ylib

eral

s,m

od

erat

es,

and

con

serv

ativ

es.

Act

ua

ld

iffe

ren

ces

(co

n-

lib

):P

red

icte

dd

iffe

ren

ces

(Ty

pic

al

con

-T

yp

ica

lli

b)

Sa

mp

lea

ve

rag

eS

am

ple

ex

tre

me

sE

ve

ryo

ne

Lib

Mo

dC

on

Dif

fere

nce

S.E

.D

iffe

ren

ceS

.E.

Dif

fere

nce

S.E

.D

iffe

ren

ceS

.E.

Dif

fere

nce

S.E

.D

iffe

ren

ceS

.E.

Rel

evan

ce:

Har

m2

0.57

0.07

20.

880.

202

1.31

0.04

21.

780.

052

0.97

0.09

20.

580.

09

Fair

nes

s2

0.67

0.06

20.

890.

202

1.57

0.04

22.

020.

052

1.14

0.09

20.

990.

09

Ing

rou

p0.

120.

070.

060.

210.

580.

040.

780.

050.

460.

080.

240.

08

Au

tho

rity

0.43

0.06

0.77

0.16

1.05

0.04

1.09

0.05

0.90

0.08

1.11

0.08

Pu

rity

0.69

0.07

0.99

0.22

1.32

0.04

1.39

0.05

0.95

0.09

1.53

0.09

Jud

gm

ents

:H

arm

20.

550.

072

0.72

0.15

21.

000.

042

1.25

0.06

20.

650.

092

0.74

0.09

Fair

nes

s2

0.70

0.06

21.

100.

142

1.24

0.04

21.

630.

052

0.80

0.08

20.

750.

08

Ing

rou

p0.

470.

060.

790.

151.

150.

041.

410.

050.

930.

070.

760.

07

Au

tho

rity

1.20

0.06

1.89

0.13

2.04

0.04

2.22

0.05

1.59

0.10

2.04

0.09

Pu

rity

1.41

0.07

1.89

0.17

2.12

0.05

2.47

0.06

1.47

0.10

1.97

0.09

Act

ua

ld

iffe

ren

ces

(co

n-

lib

):P

red

icte

dd

iffe

ren

ces

(Ty

pic

al

con

-T

yp

ica

lli

b)

Sa

mp

lea

ve

rag

eS

am

ple

ex

tre

me

sE

ve

ryo

ne

Lib

Mo

dC

on

Dif

fere

nce

S.E

.D

iffe

ren

ceS

.E.

Dif

fere

nce

S.E

.D

iffe

ren

ceS

.E.

Dif

fere

nce

S.E

.D

iffe

ren

ceS

.E.

Rel

evan

ce:

Har

m2

0.34

0.09

20.

740.

222

1.45

0.05

21.

890.

062

1.08

0.10

20.

810.

10

Fair

nes

s2

0.35

0.08

20.

570.

232

1.60

0.05

22.

140.

062

1.14

0.10

20.

820.

10

Ing

rou

p0.

250.

110.

470.

340.

760.

050.

980.

070.

570.

110.

450.

12

Au

tho

rity

0.35

0.08

1.06

0.28

1.48

0.04

1.47

0.06

1.22

0.10

1.78

0.09

Pu

rity

0.63

0.10

1.37

0.33

1.79

0.05

2.01

0.06

1.28

0.10

1.80

0.10

Jud

gm

ents

:H

arm

20.

450.

092

0.66

0.29

21.

420.

052

1.75

0.07

20.

960.

112

1.07

0.11

Fair

nes

s2

0.32

0.08

20.

720.

242

1.11

0.05

21.

500.

062

0.66

0.10

20.

630.

10

Au

tho

rity

0.60

0.08

1.15

0.28

1.87

0.05

2.00

0.06

1.51

0.10

1.92

0.10

Pu

rity

0.82

0.09

1.91

0.32

1.92

0.05

2.26

0.07

1.24

0.12

1.82

0.11

Not

e.To

pp

anel

sho

ws

pre

dic

ted

and

actu

alan

swer

sfo

rth

est

ud

ysa

mp

le,a

nd

bo

tto

mp

anel

sho

ws

the

sam

efo

rth

eco

mp

aris

on

toth

en

atio

nal

ly-r

epre

sen

tati

ved

atas

et,u

sin

go

nly

item

sco

mm

on

tob

oth

dat

aset

s(n

oit

ems

inco

mm

on

for

Ing

rou

pju

dg

men

ts).

S.E.

=st

and

ard

erro

ro

fth

ed

iffer

ence

bet

wee

nth

em

ean

s,b

ased

on

po

ole

dst

and

ard

dev

iati

on

san

dn

ot

assu

min

geq

ual

size

so

rva

rian

ces

inth

etw

og

rou

ps.

Lib

=Li

ber

alp

red

icto

rs,

Mo

d=

Mo

der

ate

pre

dic

tors

,C

on

=C

on

serv

ativ

ep

red

icto

rs.

do

i:10.

1371

/jo

urn

al.p

on

e.00

5009

2.t0

03



accuracy: 1) We found some support for the hypothesis thatmoderates would be most accurate, which they were in the case ofthe binding foundations. However, and most crucially, partisaninaccuracies were not mirror images of each other (in which casethe red and blue lines in Figure 2 would have opposite slopes). Onthe contrary, liberals and conservatives both tended to exaggeratetheir binding foundation differences by underestimating the typicalliberal and overestimating the typical conservative. 2) We found nosupport for the hypothesis that liberals would be most accurate;liberals were the least accurate about conservatives and aboutliberals. The largest inaccuracies were in liberals’ underestimationsof conservatives’ Harm and Fairness concerns, and liberals furtherexaggerated the political differences by overestimating their ownsuch concerns. 3) Finally, we found some support for thehypothesis that conservatives would be the most accurate, whichthey were in the case of the individualizing foundations. In linewith Moral Foundations Theory, liberals dramatically underesti-mated the Harm and Fairness concerns of conservatives. Thesefindings add to the literature on moral foundations by demon-strating a novel form of pragmatic validity [16] for the theory:

conceptualizing and measuring the moral stereotypes people haveof different social groups.

While we obtained a nationally-representative sample forcomparison of MFQ scores, it is important to note that thepredicted answers as typical liberals/conservatives all came from anon-representative Project Implicit sample. However, the partic-ipants in this study do ‘‘run the gamut’’ across the ideologicalspectrum, from very liberal to very conservative, and Figure 3demonstrates exaggeration across all 7 points on the politicalorientation item. Extreme liberals exaggerated the moral politicaldifferences the most, and moderate conservatives did so the least.Further, Nosek, Banaji, and Jost [34] showed evidence that strongconservatives at Project Implicit preferred conservative candidates,both implicitly and explicitly, as much as strong liberals preferredliberal candidates. Finally, across Project Implicit studies theliberal and conservative extremes show equivalent or near-equivalent extremity in implicit and explicit liking and identitywith partisan parties, politicians, and positions [35,36].

Nevertheless, we cannot completely rule out reference effects inthese predictions based on non-representative sampling. Inparticular, while the conservatives in this sample are indeed

Figure 2. Moral stereotypes about the typical liberal’s and typical conservative’s endorsement of the binding foundations(Ingroup, Authority, Purity), and individualizing foundations (Harm, Fairness).doi:10.1371/journal.pone.0050092.g002



Kevin Deutsch

Kevin Deutsch

Kevin Deutsch

conservative, they may also have different social experiences thana representative conservative. For example, conservatives who livein urban or predominately liberal enclaves might have greaterinsight into liberal beliefs than conservatives who live in rural orpredominately conservative enclaves. A useful follow-up investi-

gation would examine the effect of exposure to liberals andconservatives in one’s social context. If this is impactful, and if thepresent sample is systematically skewed in this regard, thenaccounting for social context may qualify the present conclusion ofconservatives having greater accuracy than liberals. It is also worth

Figure 3. Exaggeration of moral differences across political ideology.doi:10.1371/journal.pone.0050092.g003



noting that our single ideology item did not allow for participantsto indicate that they were libertarian, or that they were liberal onsocial issues but conservative on economic issues. Research onlibertarians has revealed a pattern of moral concerns unlike that ofliberals, moderates, or conservatives [37]; this finding, along withtheir coherent ideological identity separate from liberals andconservatives, makes libertarians a particularly interesting samplefor future studies using this paradigm. Do libertarians share themoral stereotypes about liberals and conservatives shown byparticipants in the current study? Do people hold consistent moralstereotypes about the ‘‘typical’’ libertarian, and are they accurate?Besides addressing these questions, future work should investigatedifferent possible antecedents of moral stereotyping, such asdifferential exposure to ideological caricatures in the media.

The ideological ‘‘culture war’’ in the U.S. is, in part, an honestdisagreement about ends (moral values that each side wants toadvance), as well as an honest disagreement about means (laws andpolicies) to advance those ends. But our findings suggest that thereis an additional process at work: partisans on each side exaggeratethe degree to which the other side pursues moral ends that aredifferent from their own. Much of this exaggeration comes fromeach side underestimating the degree to which the other sideshares its own values. But some of it comes, unexpectedly, fromoverestimating the degree to which ‘‘typical’’ members of one’sown side endorse its values. Studies of ingroup stereotypes tend toshow that they are more accurate and less exaggerated thanstereotypes about an outgroup [38], especially for higher-statusgroups like Whites [39]. However, the current study found thatmoral stereotypes about an ideological group can be just asexaggerated when held by ingroup members as by outgroupmembers, and sometimes even more so. We suspect that this ispartially due to the fact that one can imagine members of one’sown ideological group more extreme than oneself; people could infact be motivated to differentiate themselves from their ideologicalgroup, imagining ‘‘typical’’ group members to be more extreme intheir moral profile (it would be interesting in future work to obtainmeasures of how ‘‘typical’’ participants rate themselves to be –perhaps everyone likes to see themselves as atypical when it comesto politics). But this may also be a unique feature of moralstereotypes, in that people are motivated to exaggerate the moralvalues of their group in ways that are in line with those samevalues.

The asymmetrical pattern found in moral stereotypes about theindividualizing foundations fits remarkably well with recent workon ideological opponent and own-group misperceptions. Examin-ing co-perceptions of conflicting groups such as pro-life/pro-choice and hawks/doves, Chambers and Melnyk [40] found thatpartisans saw their adversaries as motivated by an opposition totheir own core values, rather than being motivated by promotionof the adversaries’ values. This is consistent with the moralstereotypes that liberals appear to have of conservatives: liberalssee conservatives as being motivated by an opposition to liberals’core values of compassion and fairness, as well as being motivatedby their own (non-moral) values of ingroup loyalty, respect forauthorities and traditions, and spiritual purity (they may beparticularly likely to focus on issues in which these values come

into conflict). This misperception is asymmetrical: conservativesdid underestimate liberal moral concerns with the bindingfoundations, but they were no more likely to underestimate thanliberals themselves.

It is striking that instead of basic partisan outgroup derogation,in which both sides predict that the other is less moral in general,we found foundation-specific moral stereotypes about liberals andconservatives—and these moral stereotypes were largely shared byall. Participants across the political spectrum exaggerated liberalmoral disregard for Ingroup, Authority, and Purity, and conser-vative disregard for Harm and Fairness—that is, exaggerations ofthe patterns predicted by Moral Foundations Theory. Thissuggests that moral stereotypes might be unique in that they aremotivated (partisans want to cast the other side as immoral) andyet partisans share the same moral stereotypes about either side.Even more surprising, they share both of these moral stereotypeswith moderates, who are presumably not as motivated tostereotype either side. More research is needed to further delineatethe moral stereotypes of political partisans, for instance to see ifmoral stereotypes about members of political parties mirror thoseabout ideological groups, both in two-party political systems likethe U.S. and in multiparty systems like Italy. We also hope thatfuture studies can use Moral Foundations Theory’s finer resolutionof the moral domain to investigate specific moral stereotypes alongother social groupings, such as race, gender, social class, age, orweight.

Chambers and Melnyk [40] conclude: ‘‘Partisan group mem-bers suffer the misapprehension that their adversaries work toactively and willfully oppose their own sides’ interests rather thanpromoting the values that are central to their adversaries’doctrine…it is this perception that may spawn the feelings ofdistrust and animosity that partisans feel toward their rivals andmay ultimately fuel conflict between partisan groups’’ (p.1309). Inthis study, we focused on the moral values of ideologicalopponents, and their perceptions of the moral values of eitherside, in order to understand the moral ‘‘distrust and animosity’’endemic to the liberal-conservative culture war. We found thatthere are real moral differences between liberals and conservatives,but people across the political spectrum exaggerate the magnitudeof these differences and in so doing create opposing moralstereotypes that are shared by all. Calling attention to this uniqueform of stereotyping, and to the fact that liberal and conservativemoral values are less polarized than most people think, could beeffective ways of reducing the distrust and animosity of currentideological divisions.

Acknowledgments

We thank Christian Smith and Steve Vaisey for making available theirnationally-representative moral foundations data, and we thank SelinKesebir for assistance with statistical formulas.

Author Contributions

Conceived and designed the experiments: JG BAN JH. Performed theexperiments: JG. Analyzed the data: JG BAN. Wrote the paper: JG BANJH.

References

1. Farah J (2003) Democrats are immoral. World Net Daily November 20.Available: http://www.wnd.com/news/article.asp?ARTICLE_ID = 35719. Ac-cessed 2008 Oct.

2. Feingold M (2004) Foreman’s wake-up call. The Village Voice January 21–27.Available: http://www.ontological.com/RF/rfproductionfiles/2004rufus/reviews/vvfeingold.pdf. Accessed 2008 Oct.

3. Leo J (2002) Ethics are made easy when anything goes. Townhall.com.Available: http://townhall.com/columnists/JohnLeo/2002/07/15/ethics_are_made_easy_when_anything_goes. Accessed 2008 Oct.

4. Krugman P (2007) The Conscience of a Liberal. New York, NY: Norton.5. Haidt J, Graham J (2007) When morality opposes justice: Conservatives have

moral intuitions that liberals may not recognize. Soc Justice Res 20: 98–116.



Kevin Deutsch

Kevin Deutsch

How do you think Libertarians would be stereotyped by the participants of the study?

6. Judd CM, Park B (2005) Group differences and stereotype accuracy. In DovidioJ, Glick P, Rudman LA, editors. On the Nature of Prejudice: Fifty Years afterAllport. Malden MA: Blackwell Publishing. pp. 123–138.

7. Jussim L, Harber KD, Crawford JT, Cain TR, Cohen F (2005) Social realitymakes the social mind: Self-fulfilling prophecy stereotypes bias and accuracy.Interact Stud 6: 85–102.

8. Ryan CS (2002) Stereotype accuracy. Eur Rev Soc Psychol 13: 75–109.9. Allport GW (1954) The nature of prejudice. Cambridge MA: Addison-Wesley.

10. McCauley C (1995) Are stereotypes exaggerated? A sampling of racial genderacademic occupational and political stereotypes. In Lee Y, Jussim LJ, McCauleyC, editors. Stereotype accuracy: Toward appreciating group differences.Washington, DC: American Psychological Association. pp. 215–243.

11. McCauley C, Stitt CL (1978) An individual and quantitative measure ofstereotypes. J Pers Soc Psychol 36: 929–940.

12. Nosek BA, Smyth FL, Hansen JJ, Devos T, Lindner NM, et al. (2007)Pervasiveness and correlates of implicit attitudes and stereotypes. Eur Rev SocPsychol 18: 36–88.

13. Tajfel H, Turner JC (1986) The social identity theory of intergroup behavior. InWorchel S, Austin WG, editors. Psychology of Intergroup Relations. Chicago:Nelson-Hall.

14. Greene S (2004) Social identity theory and party identification. Soc Sci Q 85:136–153.

15. Judd CM, Park B (1993) Definition and accuracy in the assessment of socialstereotypes. Psychol Rev 100: 109–128.

16. Graham J, Nosek BA, Haidt J, Iyer R, Koleva S, et al. (2011) Mapping the moraldomain. J Pers Soc Psychol 101: 366–385.

17. Graham J, Haidt J, Nosek BA (2009) Liberals and conservatives rely on differentsets of moral foundations. J Pers Soc Psychol 96: 1029–1046.

18. Cannon PR, Schnall S, White M (2011) Transgressions and expressions:Affective facial muscle activity predicts moral judgments. Soc Psychol Pers Sci 2:325–331.

19. Graham J, Haidt J (2011) Sacred values and evil adversaries: A moralfoundations approach. In: Shaver P, Mikulincer M, editors. The SocialPsychology of Morality: Exploring the Causes of Good and Evil. New York:APA Books.

20. McAdams DP, Albaugh M, Farber E, Daniels J, Logan RL, et al. (2008) Familymetaphors and moral intuitions: How conservatives and liberals narrate theirlives. J Pers Soc Psychol 95: 978–990.

21. Van Leeuwen F, Park JH (2009) Perceptions of social dangers moral foundationsand political orientation. Pers Indiv Diff 47: 169–173.

22. Robinson RJ, Keltner D, Ward A, Ross L (1995) Actual versus assumeddifferences in construal: ‘‘Naive realism’’ in intergroup perception and conflict.J Pers Soc Psychol 68: 404–417.

23. Dawes RM, Singer D, Lemons F (1972) An experimental analysis of the contrasteffect and its implications for intergroup communication and the indirectassessment of attitude. J Pers Soc Psychol 21: 281–295.

24. Chambers JR, Baron RS, Inman ML (2006) Misperceptions in intergroupconflict: Disagreeing about what we disagree about. Psychol Sci 17: 38–45.

25. Cohen GL (2003) Party over policy: The dominating impact of group influenceon political beliefs. J Pers Soc Psychol 85: 808–822.

26. Greenberg J, Jonas E (2003) Psychological motives and political orientation—theleft the right and the rigid: Comment on Jost et al. (2003) Psychol Bull 129: 376–382.

27. Rokeach M (1956) Political and religious dogmatism: An alternative to theauthoritarian personality. Psychol Mon 70: 43.

28. Tetlock PE (1984). Cognitive style and political belief systems in the BritishHouse of Commons. Journal of Personality and Social Psychology 46: 365–375.

29. Jost JT, Glaser J, Kruglanski AW, Sulloway FJ (2003) Political conservatism asmotivated social cognition. Psychol Bull 129: 339–375.

30. Sibley CG, Duckitt J (2008) Personality and prejudice: A meta-analysis andtheoretical review. Pers Soc Psychol Rev 12: 248–279.

31. Carter JD, Hall JA, Carney DR, Rosip JC (2006) Individual differences in theacceptance of stereotyping. J Res Pers 40: 1103–1118.

32. Graham JW, Taylor BJ, Olchowski AE, Cumsille PE (2006) Planned missingdata designs in psychological research. Psychol Meth 11: 323–343.

33. Smith C, Vaisey S (2009) Charitable giving and moral foundations in anationally-representative sample. Unpublished data, University of NorthCarolina.

34. Nosek BA, Banaji MR, Jost JT (2009) The politics of intergroup attitudes. In JostJT, Kay AC, Thorisdottir H, editors. The Social and Psychological Bases ofIdeology and System Justification. Oxford UK: Oxford University Press. pp.480–506.

35. Lindner NM, Nosek BA (2009) Alienable speech: Ideological variations in theapplication of free-speech principles. Polit Psychol 30: 67–92.

36. Smith CT, Ratliff KA, Nosek BA (2012) Rapid assimilation: Automaticallyintegrating new information with existing beliefs. Soc Cog 30: 199–219.

37. Iyer R, Koleva S, Graham J, Ditto PH, Haidt J (2012) Understanding libertarianmorality: The psychological dispositions of self-identified libertarians. PLoSONE 7: e42366.

38. Linville PW, Fischer GW, Salovey P (1989) Perceived distributions of thecharacteristics of in-group and out-group members: Empirical evidence and acomputer simulation. J Pers Soc Psychol 57: 165–188.

39. Ryan CS (1996) Accuracy of Black and White College Students’ In-Group andOut-Group Stereotypes. Pers Soc Psychol Bull 22: 1114–1127.

40. Chambers JR, Melnyk D (2006) Why do I hate thee? Conflict misperceptionsand intergroup mistrust. Pers Soc Psychol Bull 32: 1295–1311.



graham -the moral stereotypes of liberals and conservatives

Documents