Perceived Threat and Authoritarianism Author(s): Stanley Feldman and Karen Stenner Reviewed work(s): Source: Political Psychology,

Vol. 18, No. 4 (Dec., 1997), pp. 741-770 Published by: International Society of Political Psychology Stable URL: http://www.jstor.org/stable/3792208 . Accessed: 04/01/2012 13:24
Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at . http://www.jstor.org/page/info/about/policies/terms.jsp JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms of scholarship. For more information about JSTOR, please contact support@jstor.org.

International Society of Political Psychology is collaborating with JSTOR to digitize, preserve and extend access to Political Psychology.

http://www.jstor.org

Political Psychology, Vol. 18, No. 4, 1997

PerceivedThreatand Authoritarianism
Stanley Feldman
State University of New York at Stony Brook

Karen Stenner
Duke University

that has looked at the role of There has been a long history of work on authoritarianism societal threat.Muchof the empiricalresearchin this traditionhas relied on aggregate data to examine the relationship between societal threat and authoritarian attitudes and data and a range ofperceived threatmeasures behaviors.Ouranalysis uses individual-level and to betterunderstand dynamicsof authoritarianism threat.Wealso move beyondthe the and hypothesize hypothesisof a direct relationshipbetween threat and authoritarianism, instead that the relationship involves interaction effects: societal threat activates authoritarian predispositions.As predicted,our analysisfinds no evidenceof a direct effect of societal threat but significant evidence of an interaction between authoritarian predispositionsand perceived threat. We consider the implicationsof these resultsfor our understandingof authoritarianism.
KEY WORDS: authoritarianism; societal threat;prejudice;punitiveness

It has long been hypothesizedthat various forms of threatwill contributeto authoritarianism. one of the first detailed discussions of the subject, Fromm In thatinsecurityis the majorfactorin the developmentof authoritari(1941) argued anism. Faced with an uncertainworld and lack of direction,people will "escape from freedom."One mechanismof escape is authoritarianism. Fromm'sanalyIn sis, this insecurity results from the rootlessness of the modem world and the
Earlierversions of this paperwere presentedat the 1994 AnnualMeeting of the International Society for Political Psychology and the 1994 AnnualMeeting of the AmericanPolitical Science Association. We would like to thankMarkFischle and Leonie Huddyfor their insightfulcomments.The dataused in this paper were collected by the National Election Studies and made available throughthe InterUniversity Consortiumfor Political and Social Research.Neither the NES nor the ICPSR bear any responsibilityfor the analyses or conclusions reportedin this paper. 741
0162-895X 0 1997 International Society of Political Psychology Publishedby Blackwell Publishers,350 Main Street, Malden, MA 02148, USA, and 108 Cowley Road, Oxford, OX4 1JF,UK.

742

Feldman and Stenner

the consequencesof capitalism.It is also possible to interpret workof Adornoet al. (1950) in termsof threatandinsecurity,althoughin this case the sourceof the threat is inconsistentchild-rearing practices,andthe consequencesresultfrompsychodynamic defense mechanisms. It is clear, however, that Adorno et al. were more concernedwith socializationandpsychodynamicprocessesthanwith the influence of social and economic threaton authoritarianism. A more direct connection is evident in the work of Rokeach (1960), who hypothesized that anxiety stemming from external threatis the underlyingcause of dogmatism and, therefore, intolerance. This theme reappearsin the work of Wilson (1973) and his colleagues who argued that "conservatism"-defined in this research in a mannervirtually indistinguishable from authoritarianism-is a response to generalized anxiety (fear of uncertainty) stemming from both external and intrapsychic causes. It is clear that there is a common tendency to charactertraits, implicate threatand anxiety in the development of authoritarian intolerance (Wilkinson, 1972), and even mass violence and genocide (Staub, 1989). Although the hypothesis is well-known, there is surprisinglylittle empirical evidence on the connection between threat, authoritarianism, and intolerance. Rokeach (1960) offered some indirectbut intriguingevidence on the relationship between perceivedthreatand the apparent level of dogmatismin religiousgroups. He also presentednonexperimental datashowinga correlation betweenanxietyand Wilson andhis colleagues (1973) offereda varietyof correlational data dogmatism. thatsuggesta relationship between anxietyand(social) conservatism. example, For they found that increasing conservatismis associated with fear of death, lower levels of stimulus-seeking,and dislike of complexity. The most concertedattemptto test this hypothesishas been the work of Sales (1972; 1973; Sales & Friend, 1973). The most compelling of his analyses used aggregate data to examine the relationship between environmentalthreat and In authoritarianism. one study, Sales (1972) examined the effect of the Great Depression on conversion rates among church denominations.He found that conversion rates from nonauthoritarian authoritarian to denominationsincreased duringperiods of presumedsocietal threat.In a second study, Sales (1973) found that a variety of archival indicators of authoritarianismwere higher during periods of societal stress (the Depression Era 1930s and the late 1960s) than in the lower stress years immediately preceding. Although his results were not perfectly consistent with the research hypothesis, Sales found that many of his indicators of authoritarianism-size of police budgets, power themes in comic books, prison sentences for sex offenders, purchases of attack dogs-were significantly higher in the periods of societal threatthan in the preceding lower threatyears. A recentstudyby Doty, Peterson,andWinter(1991) replicatedthe Sales study using 1978-1982 as the high-threat years and 1983-1987 as the low-threatperiod. In a useful variationon the Sales study,Doty et al. examinedthe transition from a

Perceived Threat and Authoritarianism

743

to hypothesis high-threat a low-threatperiod,thus guardingagainstthe alternative thatsocial indicatorsof authoritarianism simply tend to increaseover time. As was the case in the originalSales study,Doty et al. found thatmany,but not all, of their archivalindicatorsof authoritarianism registeredhighervalues in the hypothesized periodof societal threatthanin the low-threatperiod.In all, 13 of their20 measures showed significant change in the predicteddirectionacross the two periods. The most notable deviations from the expected pattern occurred for measures of "authoritarian aggression" (size of police budgets and support for the death penalty),whichmay simplybe a functionof growingpublicfearsaboutrisingcrime rates over the period studied. A study by Sales and Friend (1973b) turned from aggregated measures of authoritarianismto individual level effects. In two reportedexperiments, subjects were induced either to succeed or to fail on purported measures of "intelligence and ability." Although the effects were not very large, failure (as increased, while success decreased, the subjects' levels of authoritarianism measuredby F-scale items). And these effects were larger among subjects who attributed their performance on the tasks to internal causes. In one of the experiments, Sales and Friend also found that their manipulationof success and failure on the task led to differences in subjects' levels of obedience to an authorityfigure. A recent study by McFarland,Ageyev, and Hinton (1995) examined the relationshipbetween economic threat and authoritarianism using questionnaire datafrom samplesin the U.S. and Russia. McFarland al. constructedattitudinal et measuresto parallelat the individuallevel the social indicatorsof authoritarianism used by Sales and by Doty et al. in their aggregate analyses. Although these attitudinal measuresof the social indicatorswere only weakly intercorrelated, a as scale they showed substantial correlation with authoritarianism. Measures of economic threat-unemployment, low income, and subjective measures of personal economic distress, as well as concern regarding the national economy-were weakly correlated with the social indicator items in the U.S. sample, and more strongly correlated in the Russian sample. Perhapsthe most interestingfinding was thateconomic threatappearedto have a polarizing effect on authoritarianismscores in the U.S. sample. High levels of threat seemed to drive some individualstowardhigherauthoritarianism scores while pushingother toward lower authoritarianism.As a result of this increased variance, people the correlations among the authoritarianism items were significantly higher for those individuals reportingeconomic threatthan for those not so distressed. Despite the popularityof the threat-authoritarianism hypothesis,the Sales and Friendand the McFarlandet al. studies appearto be the only significantattempts to test the hypothesiswith individual-leveldata.As suggestive as the archivaldata may be, the lack of supportfor the hypothesisat the individuallevel leaves a major of In gap in our understanding authoritarianism. orderto addressthis deficiency, we must not only develop an appropriate researchdesign, but also consider the

744

Feldman Stenner and

that conceptualizationof authoritarianism would allow us to make sense of the observedrelationship.1 THREAT AND AUTHORITARIANISM The hypothesis entertainedin most of the literatureis that threator anxiety This produceshigherlevels of authoritarianism. is a simplifiedversionof Fromm's It is important,however, to make a distinctionbetween longoriginal argument. term societal threat and short-termthreateningevents. Fromm's argumentwas based on the structural factorsinherentin capitalismand modem society thatmay foster authoritarianism. a somewhatless grandscale, Wilson (1973) hypotheOn of sized thatgeneralizedanxiety is the centraldeterminant (authoritarian) conservatism. In these argumentsit is long-term, deeply felt threat and anxiety that contributeto the developmentof authoritarian characterstructures. On the other hand, the studies of Sales and of Doty, Peterson, and Winter examine the effects of more transitorythreaton authoritarianism. what does But, it mean if levels of authoritarianism-long presumedto be a personalitycharacteristic-increase underconditionsof short-term threatand then recede when that threat is removed? Only Sales and Friend (1973) seem to deal with this issue, conceding thatthe "notionthatcentralpersonalitytraits,such as authoritarianism, environment hardly is mightchangein responseto changesin the contemporaneous a commonplacein currentpersonalitytheorizing"(Sales & Friend, 1973, p. 163). Althoughtherearetheoristswho arguethatpersonalitytraitsarenot fixed andmay indeed exhibit significant short-termchange, there is anotherpossibility to consider:one which we believe providesa moresatisfactory accountof therelationship between transientthreatand manifestationsof authoritarianism. Perhapsit is not authoritarianism itself that increasesin the face of transientenvironmental threat, but ratherthe relationshipbetween authoritarianism its attitudinal behavand and ioral manifestationsor consequences, like intolerance.This implies that the typiresult from an interaction cally observed consequences of authoritarianism between authoritarian and threat. predispositions This hypothesisis consistentwith the aggregate-levelstudiesby Sales and by Doty, Peterson, and Winter. Since the researcherscould not directly measure authoritarian personalitytraits,the changes in behaviorsthatthey observedcould well be a productof the activation of existing authoritarian traitsby threat,rather thanan increasein the traitsthemselves.And while Sales andFriend(1973) found some effect of threat (failure) on a measure of authoritarian the "personality,"

1A recent study by Marcus et al. (1995) examined the effects of threat on intolerance,with threat operationalizedas both perceptionsof a specific threateninggroup, and of generalizedthreatfrom for of many groups. While their results are interestingand important our understanding intolerance, theirfocus is very differentfrom the one pursuedhere.

Threatand Authoritarianism Perceived

745

were fairly small. Moreover, since even observed changes in authoritarianism and pencil measures of authoritarianism still indirect indicatorsof an are paper unobservedconstruct,it is possible thatthe observedincreasesin the scale scores reflect an increased salience of authoritarian predispositions-and thereby increasedmanifestationof authoritarianism-in the face of threat,ratherthanactual gathersmomentumas one considers changesin the underlyingtrait.This argument that the most common measures of authoritarianism often tap rather specific manifestationsof the trait,like prejudiceand intolerance,which seem to us more akinto dependentvariables,thanto indicatorsof the independent variable:authoritarianpredisposition. We could find only one individual-levelstudythat-although the authors were not concerned with our particulartopic-provided evidence of an interaction between authoritarianism threat.The studyis one in a series thatwas designed and to test a frameworkknown as terrormanagementtheory (Rosenblattet al., 1989; Greenberget al., 1990). This theory posits that awareness of human mortality for (Rosenblattet al., 1989, p. 681). In producesa "potential overwhelmingterror" order to deal with this, societies develop cultural worldviews to provide their memberswith a sense of meaning in life. Threatsto this culturalworldview-in the formof deviancefromsocietalnorms-must therefore eliminated,especially be if the threatof mortalitybecomes salient. To test terror management theory the investigators used a fairly simple manipulationto increase the salience of the subjects' mortality. Specifically, subjects in the experimental condition were asked to answer two open-ended questions:one questionrequiringthem to say what they thinkwill happento them when they die, and the second askingthem to reportthe emotions thatthe thought of their death arouses in them. In a series of studies, Rosenblatt et al. (1989; Greenberget al., 1990) found that the mortalitysalience manipulationincreased subjects' rejectionto outgroupmembers and the former's levels of punitiveness towardsocial deviants:reactionsthatare highly characteristic authoritarians. of In one particular of greatestinterestfor our purposes,Greenberget al. study, before the experimental (1990) measuredtheirsubjects'levels of authoritarianism The dependent variable was attitude toward a person who was manipulation. constructedto be either similaror dissimilarto the subject.The analysis found a and pronouncedinteractionamong mortalitysalience, authoritarianism, reactions to a dissimilartarget.High and low authoritarians differedin theirreactionsto the dissimilar target only under conditions of mortalitysalience. And the mortality salience manipulation influencedattitudestowardthe targetonly amongthose high in authoritarianism. Thus, subjectsin this study exhibitedthe greatestrejectionof a dissimilartargetperson when they were both high in authoritarianism faced and with the reality of theirmortality:clearly, a threateningsituation. Although these authors were not attemptingto investigate the relationship between threat and authoritarianism, their results can be interpretedas clear evidence for an associationbetween the two, and a particular kind of association

746

Feldman and Stenner

at that.In theirlatterstudy, we observe the effects of the interactionof threatand authoritarianism the manifestationof authoritarian on behaviors, ratherthan a direct effect of threat on levels of authoritarianism. This is consistent with the perspective we have adopted in this paper and suggests that the interaction hypothesis in worthy of furtherinvestigation.Moreover, since a wide variety of potentialthreatshave been discussedin the literature-from threatsto self-esteem, identity and meaning, to threats to economic security and the stability of the betterthe specific types of threatthat political world-we still need to understand to lead individualshigh in authoritarianism become more intolerantand punitive. Indeed, a majorlimitationof the aggregate-levelstudies is that they are unableto distinguish among different types of threats, which become confounded in the simple comparisonof variousyears or periods of time. betweenthreatand Shouldtherebe strongempiricalsupportfor an interaction we would have more thanjust evidence for an interestinglittle authoritarianism, of hypothesis.Whatwould it indicatefor ourbasic understanding authoritarianism itself if its manifestationsdependheavily on the presenceof environmental threat and insecurity?Although it might be possible to interpretthis result within the the Adornoet al. conceptualization, psychodynamicmodeldoes not seem to predict a centralrole for threatin the activation of authoritarianism. Altemeyer's (1988, 1996) social learning account can even less comfortably accommodate such an interaction effect. In fact, we suggest that empirical evidence of a central role for threat in the activation of authoritarianismwould provide the basis for a wholly different interpretationof the origins and dynamics of authoritarianism itself. STUDY DESIGN For this investigation,we utilize datafromthe 1992 NationalElection Studies pre-postelection survey. As with most secondarydata analyses, this study is not ideal for the purpose at hand. The NES data allow us to construct a range of theoreticallyinterestingmeasuresof threat,but are less satisfactoryin termsof the availability of good dependent variables. Unfortunately,the NES data do not directly measure intolerance or punitiveness: two presumed importantconseHowever, we can make up for the lack of ideal quences of authoritarianism. variables by considering a set of variables that include measures of dependent minority group attitudes, attitudes toward the use of force, and general social and political attitudes. Having multiple dependent variables across these categories enables us to determine if the phenomenon we are observing is in fact the syndrome typically identified as authoritarianism. our results turn out to If be consistent across this range of dependentvariables, it will provide substantial evidence that the phenomenon we are studying accords with previous descriptions of authoritarianism.

Perceived Threat and Authoritarianism

747

The interactionhypothesis requires us to distinguish (more clearly than is predispositionsand the observed manitypically the case) between authoritarian in festationsof authoritarianism attitudesandopinions.One of the majorproblems with F-scale-typemeasuresis thatthey include items thatare often uncomfortably that close to the consequencesof authoritarianism we are interestedin explaining. The measure of authoritarian predispositions we use in our analysis taps the to alternativechild-rearingvalues. The measurearrays respondents'subscription respondentson a dimensionthatrunsfrom a belief at one end thatchildrenshould be well-behaved,obedient,and respectfulof elders, to a view at the otherend that children should be independent,responsible,and curious (see Kohn, 1977, for a discussion of the developmentof measuresof child-rearing values). it measureof authoritarianism, has long been Althoughthis is not a traditional noted that this dimension of child-rearingvalues is strongly related to other One authoritarianism measuresandto presumedconsequencesof authoritarianism. of the originalF-scale items, and one thathas survivedthroughthe yearsto appear in Altemeyer's Right-WingAuthoritarianism scale, is "Obedienceand respectfor authorityare the most importantvirtues children should learn."From a lengthy empiricalstudyof intolerance,Martin(1964, p. 86) concludes:"Thereis probably no otherquestion on which tolerantsdiffer from intolerantsmore sharplythanon child-rearingpractices. Attitudes on this subject directly and efficiently reflect general ethnocentrism.How to 'bring up' or socialize children is a matter of profoundconsequences,involving basic humanvalues andobjectives."And,based on many years of studyingchild-rearingvalues, Kohn (1977, p. 201) argues that "a conformistorientationincludes an unwillingnessto permitothersto step out of narrowly defined limits of what is proper and acceptable. Thus, a conformist orientationimplies not only intoleranceof deviantpolitical belief, but also intolerance of any beliefs thoughtto be threatening the social order-religious beliefs, to ethnic and racialidentifications,even beliefs aboutproperdress and deportment." We believe that this dimension of child-rearingvalues provides an excellent, nonobtrusivemeasureof authoritarian useful for predispositions.It is particularly our purposes since it does not include questions tappingpresumedconsequences of authoritarianism-like prejudice and intolerance-that make up our set of dependentvariables. ANALYSIS Given the natureof some of ourdependentvariables,we excludedall minority respondentsfrom the NES sample. Further,given that the variables of interest spannedboth the pre- andpostelection sections of the survey,and were addressed in their entiretyonly by those respondentswho answeredthe "long-form" the of we furthersubsettedthe datato includeonly those respondentswho questionnaire, answered the long-form questionnairefor both sections of the survey. From an

748

Feldman and Stenner

original sample size of 2487, 2004 respondentsmeet these criteria,and of those, 1564 are white non-Hispanics. What follows below is an outline of the dependentvariablesfor the analysis, and of the independent variables we anticipated would be implicated in the tendency to hold stereotypical,punitive, and ethnocentricattitudes.We indicate for how conceptswere operationalized inclusionin the model, andourexpectations for theirimpacton the dependentvariables. Dependent Variables As we have noted,a limitationof the NES datais thatthey do not includedirect measuresof constructslike intoleranceor punitivenessthataretypicallyexamined We of as manifestations authoritarianism. have thereforechosen a rangeof dependent variables for our analysis. The variables fall into three general categories: and toward attitudestowardminoritygroups,social andpoliticalattitudes, attitudes has on the use of force. The vast literature authoritarianism shown strongconnections between these categories of attitudesand authoritarian predispositions.We to expect increasingauthoritarianism be associatedwith negativeand stereotypical attitudestowardminoritygroups and to the willingness to use force, both domesAuthoritarianism should also be relatedto key social tically and internationally. a versusoutgroupattachments, desirefor orderover freedom, orientations; ingroup and patriotism.If our empiricalresults are consistentacross this set of dependent variables we will be able to relate with considerableconfidence our findings to on literature authoritarianism. Attitudes Toward Minority Groups Negative racial stereotyping.The NES study included a batteryof questions on perceptionsof the traitsof four groups:whites, blacks, Hispanics,and Asians. For each group,respondentswere first askedto indicateon a seven-pointscale the extent to which groupmemberswere typically violent or peaceful. They thenrated each group on a scale of hardworkingversus lazy, and finally intelligent versus unintelligent.For each of the three minoritygroups,difference scores were computed to indicatethe extent to which groupmemberswere perceivedto have more negative characteristicsthan whites on each of the three trait dimensions. This produceda total of nine indicators:three groupsby three traits.In orderto obtain a somewhat more normal distribution,we took the squareroot of each of these differencescores.We thenaveragedthe nine indicators producea single measure to of negative racial stereotyping.The measurehas an estimatedreliability(coefficient a) of .90. Tendencyto blame blacks. The second measureof attitudestowardminority groupsis anitem gaugingthe extentto whichrespondents thoughtthatthe relatively

Perceived Threatand Authoritarianism

749

impoverishedcircumstancesin which blacks find themselves were due simply to lack of effort on the partof the latter.Specifically, the questionasked respondents to indicate the extent to which they agreed or disagreedwith the statement,"It's really a matterof some people not trying hard enough; if blacks would only try harderthey could be just as well off as whites." Gay rights.Respondentswere asked, "Do you favor or oppose laws to protect Do homosexuals againstjob discrimination? you favor/opposesuch laws strongly or not strongly?" Social and Political Attitudes Groupattachments-ingroup versusoutgroup.One of the majorhypothesized characteristics authoritarians thatthey are inordinatelyattachedto the domiof is nant ingroupsin society and rejectthe outgroups.In fact, Duckitt (1989) sees this as one of the defining characteristics authoritarianism. NES questionnaire of The asked people to indicate whether they feel close to a wide range of social and political groups,specifically, to indicatethe people "whoaremost like you in their ideas and interests and feelings about things." We designated the middle class, whites, and business people as ingroups,and blacks, Hispanics,Asians, and poor people as outgroups.Fromthese responseswe constructedtwo separatemeasures, one of ingroupattachment one of outgroupattachment, and each scored0 to 1, with 1 indicatingthatrespondents close to all of the groupsin the set, and0 indicating felt that they felt close to none of the groups (with values between 0 and 1 indicating other intermediatelevels of group closeness). Since we sought a measure that indicates the extent to which people both feel close to ingroupsand do not feel close to (feel distantfrom) outgroups,we multipliedthe ingroupattachment score the reverse of the outgroupidentificationscore. This yields a final measure by rangingfrom 0 to 1, with a score of 1 indicatingthatthe respondentfeels close to all of the ingroups and none of the outgroups.A score of 0 would occur if the respondenteither felt close to none of the ingroups, or felt close to all of the outgroups. Order versusfreedom. People high in authoritarianism should place a high value on orderin society but not on individualfreedom.We constructeda measure of this propensityfrom a value-ranking measuredesignedby Inglehart(1977). The question reads, "For a nation, it is not always possible to obtain everythingone might wish. On page ten of the booklet severaldifferentgoals arelisted. If you had to choose amongthem, which one seems most desirableto you? Which one would be your second choice?... 1. maintainingorderin the nation;2. giving the people more say in importantpolitical decisions; 3. fighting rising prices; 4. protecting freedomof speech."We thencreatedtwo variablesrepresenting extentto which the the respondentvaluedorderandfreedom.Each is scored 1 if thatvalue was ranked most important, if rankedsecond most important,and0 if not chosen. As with 0.5 the groupcloseness measure,we are interestedin the extent to which people both

750

Feldman and Stenner

value orderand do not value freedom.We thereforeconstructedthe final measure as the productof the ordervariablemultipliedby the reversedfreedomvariable. are TrueAmerican.Authoritarians expected to be both patrioticand inclined to demand social conformity.We might reasonablyexpect them to be concerned aboutthe definitionof citizenshipandto define citizenshipin a mannerthatrequires normsand practices.Happily,one subscriptionto and conformitywith traditional therearecertainqualitiesthatmake of the NES questionsreads,"Somepeople say a persona trueAmerican.Otherssay thatthereisn't anythingthatmakesone person more Americanthan another.I'm going to read some of the things thathave been mentioned. For each of the following, tell me how importantyou think it is in making someone a true American-extremely important,very important,someor what important, not at all important... Believing in God." Attitudes Toward the Use of Force is Supportfor the death penalty. A central aspect of authoritarianism the to use extrememeasuresto punishthose who violate laws. In the U.S., willingness the most extreme use of legitimate force is the death penalty. The relevantNES questionasks whetherrespondents"favoror oppose the deathpenaltyfor persons convicted of murder"(and whetherthey favor/opposethe death penalty strongly or not strongly). Defense spending.The willingnessto use force can extendbeyondthe nation's borders,both to protect the countryand to punish wrongdoers.One indicatorof such willingness to use force is supportfor greaterspendingon defense. The NES question we employ is, "Some people believe that we should spend much less money for defense. Othersfeel thatdefense spendingshouldbe greatlyincreased. Where would you place yourself on this scale, or haven't you thoughtmuch about this?" The original measure ranges from 1 (defense spending should be greatly decreased)to 7 (spend much more money for defense). for Support the Gulf war. Respondentswere asked:"Do you thinkwe did the right thing in sending U.S. militaryforces to the PersianGulf or should we have stayed out?"Note that this is a retrospective judgmentasked in the fall of 1992. INDEPENDENT VARIABLES Child-Rearing Values The centralexplanatoryvariableis a measureof authoritarian values formed from responses to a series of questions concerning"child-rearing values." These in questionspresentedrespondents turnwith fourpairsof qualities-independence or respect for elders; obedience or self-reliance;curiosity or good manners;and being considerateor well-behaved-and asked them to indicate which of the two

Perceived Threat and Authoritarianism

751

was the most desirablequality for a child to have. The four separateitems were coded such that the "authoritarian" response to each pair of qualities (respectfor elders, obedience, good manners, being well-behaved) was scored 1; the "nonauthoritarian" response(independence,self-reliance,curiosity,being considwas scored0; and respondentswho offeredthatthey were uncertainof their erate) choice, or thought that neither or both qualities were important,were scored 0.5. The authoritarianismmeasure is then a simple additive scale constructed by summing these four variables. The estimated reliability of the four item scale is .66. on The influence of authoritarianism intolerance,punitiveness,and negative attitudestowardoutgroupsis expected to be contingentupon-specifically, magnified by-conditions of fear andthreat.Thus, we constructeda rangeof variables measuring conditions which might be threateningto the individual, including political, economic, and personal threats.Then for each of the threatconditions by represented these variables,we constructedan interactiontermfor inclusion in the model, crossing each of the threatvariableswith the measureof authoritarianhas ism. Since the previousliterature touchedupona wide rangeof potentialthreats than we to thatmay contribute authoritarianism, have triedto samplebroadlyrather precisely defining threatand measuringit narrowly.One goal of this analysisis to compare the impact of the various conditions that have been hypothesizedto be to threatening authoritarians. Political Threat Perceived ideological distancefrom majorpolitical actors. This variabletaps as here operationalized the one aspect of what might be termed"politicalthreat," fromeach of fourmajorpolitical sumof a citizen's perceived"ideologicaldistance" actors:the RepublicanandDemocraticparties,andthe RepublicanandDemocratic presidentialcandidates.These representthe majorobjects on a citizen's political horizon and the main alternativesamong which political choices are made. The variable was constructed by summing the four absolute distances between a conrespondent'sown ideological position on a seven-pointliberal-conservative tinuum (self-placement), and the perceived positions of each of the two major parties,and the two majorpresidentialcandidates(Bush and Clinton). Thus we are interestedin the ideological distancethatindividualsperceive to exist between themselves and all of the majoractorsin the political arena,rather than how ideologically distant they feel from the nearest political alternatives ideological threat (politicalpartiesandpresidentialcandidates).We operationalize is in this mannerbecause we anticipatethat what is threateningto authoritarians that there may exist a party and/or ideological diversity per se, notwithstanding mattersareproximateto theirown. Given candidatewhose positions on important the authoritarian's exaggeratedneed to protect many empiricalfindings regarding and consequent intolerance of unconventional, deviant the dominant culture,

752

Feldman and Stenner

to attitudesandbehavior,we expect authoritarians be fearfulof the mere existence in the political arenaof beliefs that are removedfrom theirown. Moreover,given the tendency of people to perceive that the public as a whole shares their own opinions (Ross, Greene, & House, 1977), partiesand candidateswho are far from the respondent'sideological position are likely to be seen by that individual as divergingfrom the position of most people in the U.S. Negativityof response to presidential candidates.Anotherpotentiallyimportant aspect of "politicalthreat"is the extent to which individualsreactnegatively to both of the majorpresidentialcandidates.We createda variablemeasuringthis constructfromresponsesto two groupsof questionsgaugingpositive andnegative emotions arousedby the candidatesand respondents'ratings of the candidates' possession of a variety of traits. The first group of questions involved eight items dealing with positive and negative emotions arousedby Bush and Clinton. These questions asked respondents to report,in turn,whethereach of the presidentialcandidates-"because of the kind of person he is, or because of somethinghe has done"-had ever made themfeel angry,hopeful, afraid,or proud.Foreach candidatewe formeda variable thatcountsthe numberof reported negativeemotionsminusthe numberof reported positive emotions. These two variableswere then multipliedtogetherso thathigh scores indicate a preponderanceof negative emotions toward both candidates. Thus, a respondentwould achieve a low score on this componentof the overall measureif he or she had largely positive emotions abouteither candidate. The second componentto our measureof negativity of response to political leadersinvolved individuals'ratingsof the candidates'possession of each of nine traits. ConsideringBush and Clinton in turn,respondentswere asked to indicate how well a numberof words or phrasesdescribedthe candidatein question.Since these posed a seriesof positive traits-intelligent, compassionate,moral,inspiring, strong, caring, knowledgeable,honest, and "gets the job done"-our measureof negativity first reversed,and then summedthe traitratingsfor each of the candidates. We then multipliedthe Bush and Clintontraitscores togetherso that,again, low scores would be obtainedon this componentif the respondentsaw either of the two candidatesas having largely positive traits. Our overall measureof negativity of response to the presidentialcandidates then consisted of an equally weighted summationof the negativity of emotions arousedby the two candidates,andthe negativityof theirtraitratings.It is important to recognize that, by the way in which we have constructedthis measure,high negativity scores cannot be producedby intense dislike of only one candidate,as might occur if the respondentwas a very strongpartisan.High negativity scores will only obtain if the respondenthas strongnegative feelings towardboth presidentialcandidates. Negativityof response to political parties. Justas people may be disturbed by the perception that both presidential candidates have predominantlynegative so characteristics, too may negative perceptionsof both political partiesbe threat-

Perceived Threat Authoritarianism and

753

ening. Unfortunately,we do not have availablethe same kind of emotion and trait items for partiesthatwere availablefor the candidates,and most resortinsteadto the "feeling thermometer"scores for the Democratic and Republican parties. Respondentsratedeach of the majorpartieson a scale rangingfroma very negative score of 0 to a highly positive score of 100. Since we are interestedin the degree of negativityfelt towardboth parties,we reversedthis scoringand then multiplied the two ratings. As before, high scores on this overall measure indicate strong negative reactionsto both political parties. Economic Threat Beyond the concept of political threat,it is clear that theremay be conditions of economic threator insecuritythataffect the extentto whichpeople areintolerant and/or punitive. To investigate the threatpotential of economic conditions, we included variables measuringeach respondent'slevel of threatin three respects that we anticipatemight affect one's sense of security and well-being: fear of and unemployment perceptionsthatthe economic situationof one's household,and of the nation as a whole, had worsened. We operationalized fear of unemployment first forming dummyvariables by indicatingwhetherthe respondent,and his or her spouse (if one existed and was in the workforce) were "very worried"about losing theirjobs, or (if unemployed) aboutnot being able to findjobs in the nearfuture.These variableswere then each weighted by the extent to which the householdwas dependentupon thatperson's of income-multiplying the dummyvariableby the proportion the family's income earnedby the individualin question-and summed.The formulafor constructing the variablethusvariedaccordingto the maritalandemploymentcircumstances of each respondent.High scores indicate a fear of unemploymentcombined with a high level of family dependenceupon the income of the personin question. Next, we formed a scale measure of negativity of retrospectivepersonal economic evaluation by summing responses to two items gauging respondents' perceptionsof the extent to which theirhouseholdeconomic situationshad gotten betteror worse over the last year.The two questionsare,"Weareinterestedin how people are getting along financiallythese days. Would you say thatyou (and your family living here)arebetteroff or worse off financiallythanyou were a yearago?" and "Do you thinkthatover the last year (your/yourfamily's) income has gone up more than the cost of living, has it fallen behind, or has it stayed abouteven with the cost of living?" Since we are interestedin perceptionsthat the respondent's economic situationhas gottenworse, we dichotomizedresponsesto bothquestions into "worse"versus "same"or "better." Ourmeasurethen combines responsesto both questions. Similarly, we constructed a scale measure of negativity of retrospective national economic evaluation by summing responses to three items gauging respondents'perceptionsof the extentto which the nationaleconomic situationhad

754

and Feldman Stenner

gotten betteror worse: (1) "How aboutthe economy. Would you say thatover the past year the nation's economy has gotten better, stayed the same, or gotten worse?"; (2) "Comparedto four years ago, would you say that the nation's economy has gottenbetter,stayedaboutthe same, or gottenworse?";and(3) "How aboutAmerica's ability to compete in the world economy? Over the past year has thatgotten better,stayedthe same, or gotten worse?"As was the case for personal economic evaluations,we dichotomizedresponsesfor each questioninto "worse" versus "same"or "better," formed a single measureby combining responses and to the threequestions. Fear of War Finally,we soughtto operationalizean important conceptraisedin the discussion of terrormanagementtheory.The researchwe reviewed raisedthe possibility that authoritarianism may be a worldview and modus operandiformed in partto protect the individual from the terrifying recognition of his or her inevitable does involve a fear of dying, then the threatof death mortality.If authoritarianism on ought to be a powerful magnifierof the effects of authoritarianism hostile and intolerantattitudes.Even if the frameworkof terrormanagementtheory is not accepted-and here we are entirely agnostic-the threatof death may be suffiWith this in ciently anxiety-provokingto disturbthose high in authoritarianism. mind, and given the destructivepotentialof nuclearwarfare,we included a final threatvariable which measuredthe extent to which respondentsexhibitfear of nuclear war. This was based on a question that asked respondentsto reporthow afraidthey were thata nuclearwar would occur in the nearfuture. Control Variables There are a numberof plausible rival influences on punitive and intolerant attitudes that need to be included in our analysis. We constructed variables measuringthe most importantconcepts among these: moral conservatism,religifundamentalism, osity, Christian political conservatism,years of education,political knowledge, and partyidentification.Finally, we includedvariablesmeasuring family income (in thousands of dollars), age (in years), and gender (dummy variableindicatingmale). Details of the construction these variablesareprovided of in the Appendix. Since the variablesin this analysis mostly do not have any naturalmetric, all of the variables-with the exception of family income, age, and years of education-were rescoredto range from a minimumof 0 to a maximumof 1 to aid in of interpretation the metricregressioncoefficients. Thus, a unit change in each of these variablesinvolves movement from the lowest possible score to the highest possible score. To furtherassist in computingand interpreting effects, given the

Perceived Threat and Authoritarianism

755

many interactionterms in the model, each variable was centered on zero by the subtracting variable'smean. For each of the nine dependentvariables,we estimateda regressionequation least squares.The specificationof these regressionsincludedall of the via ordinary independent variables described above as well as interaction terms between authoritarianism each of the threatvariables.By rescoring the authoritarian and values scale and the threatmeasuresto have means of zero, priorto child-rearing constructionof the interactionterms, we have simplified the interpretation the of coefficients. The coefficient on authoritarian can thus be interpredispositions pretedas the effect (partialslope coefficient) of this scale on the dependentvariable when all of the threatvariablesare at the same means. This will be almostexactly the same coefficient for authoritarian values thatwould be estimated child-rearing in a simple linear,additiveregressionmodel with the sampleindependent variables for the interactionterms). The coefficients for the interactionterms then (except indicatethe extentto which the effectof authoritarianism the dependentvariable on varies as a functionof each indicatorof threat.In all cases, the threatvariablesare coded so thatwe expect positive interactioncoefficients.2 With nine regressionequations,we obviously need to be cautiousin drawing conclusions about the impact of a particular independentvariableor interaction. Withthis manycoefficients,we could get statisticallysignificanteffects by chance. One strategy for dealing with this would be to combine these nine dependent variables into a smaller numberof composite measuresto reduce the numberof estimatedcoefficients.We have notchosen this strategybecausewe areparticularly interestedin discerningthe effects of authoritarian and predispositions threatacross a wide range of dependentvariables.Instead,we will look for consistent effects acrossthe nine regressionequations.If a variableor interaction reallyhas no effect on authoritarian attitudes,we should observe coefficients randomly distributed around zero across the nine equations. Consistentlypositive interactioneffects across the equationswould be extremelyunlikely to occur by chance. RESULTS Before we presentthe regressionresults,it is enlighteningto considerfirstthe betweenauthoritarian and simplerelationship predispositions the variousmeasures of threat.Accordingto the simple threat-authoritarianism hypothesis-which we earliercalled into question-environmental threatshould lead to increasesin the observedlevels of authoritarianism. Consistentwith the perspectivedeveloped in
2 Since we were including several interactionterms in the regression equations, we first estimated

term at a time, in orderto guardagainstthe chance separateequations,includingonly one interaction that we were missing some significant effects due to possible multicollinearity.These additional regressionestimates simply confirmedthe resultswe presentin the single equations.

756

Feldman and Stenner

this paper,we found no such relationship.The largestcorrelationbetween any of the threatmeasures and authoritarian child-rearingvalues was -.14, for the perwas ceived ideologicaldistancemeasure.Moreover,thiscorrelation negative:more were associated with the perception of less ideological authoritarian responses discord (as well as with less negative perceptionsof the presidentialcandidates). values and The largestpositive correlationwas a mere .12, between authoritarian between authoritarian values and each of the fear of nuclearwar. The correlations from zero. three economic threatvariableswere all statisticallyindistinguishable We thus find no evidence that these sorts of threatsactually lead to significantly higherlevels of authoritarian predispositionsper se.3 The regressionestimates for the dependentvariables are shown in Tables I values and all of the threat through III. Since we have rescored authoritarian measuresas deviationsaroundtheirmeans, the coefficients for those variables(in the upperpanel of each of the tables) are virtuallythe same as would be obtained from estimatinglinear additiveregressions,withoutany interactioneffects. More values is the effect of that precisely, the estimated coefficient for authoritarian the dependentvariable when all of the threatmeasures are at their variable on sample means. Likewise, the coefficients for the threatvariablesgive theireffects on the dependentvariablewhen authoritarianism at its samplemean. Significant is interactioncoefficients indicatethatthe effect of authoritarianism thatdependon ent variablevarieswith level of threat(andconversely,thatthe impactof thatthreat conditionon the dependentvariablevaries with level of authoritarianism). Although the results are not perfectly consistent across the dependentvariables, a numberof the interactionterms are statisticallysignificantand substantermsare significantwith the wrong(negative) tively large.4Only threeinteraction sign, and only one of those is substantivelylarge. In contrast,twenty-one of the interactionterms are statisticallysignificantin the predicted(positive) direction. This is well beyond what could be expected by chance. The most consistentresult is the effect of the interaction of authoritarian predispositions and perceived ideological distance. This interactionis statisticallysignificantand substantively large in seven of the nine equations. Recall that this measure taps the overall ideological distance that respondentsperceive to exist between themselves, and each of the two majorpartiesandpresidential candidates.High scores indicatethat see all of these centralpoliticalobjectsas being distantfromtheirown respondents position.

3Nor, alternatively,that authoritarian predispositionsmake it more likely that people will perceive varioustypes of threat. 4 Since thereis no obviousreasonto expect anyof thecoefficientsforthe interaction termsto be negative, we use one-tailedtests of significance for these coefficients (p < .05) and two-tailedtests elsewhere. coefficient with the wrong(here,negative)sign shouldbe statistically Strictlyspeaking,no interaction significant using a one-tailed test. Nonetheless, we do indicate coefficients that would have been significantat p < .05 if we had used a two-tailedtest.

Perceived Threat and Authoritarianism of tudesTowardMinorityGroups Table I. Determinants Attii

757

Independentvariables values Authoritarian Perceived ideological distance Negativity of responseto presidentialcandidates Negativity towardpolitical parties Negativity of retrospectivenational economic evaluations Fear of unemployment Negativity of retrospectivepersonal economic evaluations Fearof nuclearwar * Authoritarianism Perceived ideological distance * Authoritarianism Negativity toward presidentialcandidates * Authoritarianism Negativity toward political parties * Authoritarianism Negativity of national economic evaluations * Authoritarianism Fearof unemployment * Authoritarianism Negativity of personal economic evaluations * Authoritarianism Fear of nuclearwar Christianfundamentalist Moralconservatism Political conservatism Partyidentification(Republican) Religiosity Gender(male) Age (10 years) Family income ($10,00s) Education(10 years) Political knowledge Constant AdjustedR2

Racial Stereotyping .11 (.019)* -.01 (.041) -.11 (.046)* -.14 (.040)* .05 (.016)* .04 (.023) -.01 (.012) .02 (.015) .39 (.136)* -.10 (.152) -.20 (.132) .06 (.049) -.04 (.074) -.10 (.039)# .04 (.047) .03 (.012)* .07 (.027)* .01 (.027) .00 (.018) -.04 (.020)* .01 (.010) .02 (.003)* .00 (.002) -.07 (.024)* -.14 (.026)* -.01 (.008) .20

Blacks Don't Try Hard .11 (.026)* -.18 (.057)* -.13 (.065)* .02 (.056) .03 (.022) .04 (.032) -.03 (.017) .02 (.021) .84 (.190)* .02 (.213) .37 (.184)* .08 (.069) .20 (.104)* -.03 (.054) .12 (.066)* .02 (.017) .22 (.038)* .07 (.038) .04 (.024) -.10 (.028)* .04 (.015)* .00 (.004) -.00 (.003) -.18 (.034)* -.16 (.037)* -.02 (.011) .21

Gay Rights .09 (.035)* .05 (.076) .09 (.087) .12 (.075) -.06 (.030) -.01 (.042) -.02 (.023) -.06 (.028)* -.09 (.254) .51 (.285)* -.24 (.246) .01 (.093) .02 (.139) .06 (.072) .14 (.088)* .08 (.022)* .43 (.051)* .21 (.050)* .13 (.033)* .05 (.038) .14 (.019)* .00 (.006) -.00 (.004) -.05 (.046) -.12 (.049)* -.09 (.015)* .24

Note: Entriesare unstandardized errors.* = p < .05 (one-tailed regressioncoefficients and standard test on interactioncoefficients, two-tailedtests for othercoefficients); # = p < .05 had a two-tailed test been appliedto the interactionterm instead. Source: 1992 National Election Study, non-Hispanicwhites completingboth pre- and postelection N questionnaires. = 1564.

758

Feldman and Stenner of Table II. Determinants Social and Political Attitudes Ingroup/ Outgroup .00 (.028) -.05 (.061) -.20 (.070)* -.06 (.061) -.01 (.024) -.04 (.034) -.02 (.018) -.03 (.023) .37 (.204)* -.29 (.229) -.07 (.198) .18 (.075)* -.10 (.112) -.11 (.058)# .14 (.071)* -.02 (.018) .05 (.041) .04 (.040) .09 (.026)* .01 (.030) .00 (.016) -.01 (.004) .02 (.003)* .19 (.037)* .08 (.040)* .00 (.012) .15 Ordervs. Freedom .02 (.040) -.09 (.086) -.36 (.098)* -.00 (.086) .02 (.034) -.01 (.048) .03 (.026) -.02 (.032) .17 (.289) .75 (.324)* .02 (.280) .21 (.105)* .04 (.158) -.09 (.082) -.06 (.101) -.03 (.025) .20 (.058)* .20 (.057)* .08 (.037)* -.00 (.043) -.07 (.022)* -.02 (.006)* .00 (.004) .03 (.052) .09 (.056) .05 (.017)* .07 True American .13 (.028)* -.22 (.061)* .01 (.070) -.09 (.061) .03 (.024) .06 (.034) .01 (.019) .09 (.023)* .38 (.205)* -.11 (.230) .42 (.199)* -.02 (.075) -.08 (.112) -.04 (.058) -.06 (.071) .10 (.018)* .23 (.041)* .17 (.041)* -.00 (.027) .38 (.031)* .04 (.016)* .01 (.004)* -.00 (.003) -.14 (.037)* -.16 (.040)* -.05 (.012)* .43

variables Independent values Authoritarian Perceivedideological distance Negativity of responseto presidentialcandidates Negativity towardpolitical parties Negativity of retrospectivenational economic evaluations Fearof unemployment Negativity of retrospectivepersonal economic evaluations Fearof nuclearwar * Authoritarianism Perceivedideological distance * Authoritarianism Negativity toward presidentialcandidates * Authoritarianism Negativity toward political parties * Authoritarianism Negativity of national economic evaluations * Authoritarianism Fearof unemployment * Authoritarianism Negativity of personal economic evaluations * Authoritarianism Fearof nuclearwar Christianfundamentalist Moralconservatism Political conservatism Partyidentification(Republican) Religiosity Gender(male) Age (10 years) Family income ($10,000s) Education(10 years) Political knowledge Constant AdjustedR2

Note: Entriesare unstandardized errors.* = p < .05 (one-tailed regressioncoefficients and standard test on interactioncoefficients, two-tailedtests for othercoefficients);# = p < .05 had a two-tailed test been appliedto the interactionterm instead. Source: 1992 NationalElection Study, non-Hispanicwhites completingboth pre- and postelection N questionnaires. = 1564.

Perceived Threat and Authoritarianism Table III. Determinants AttitudesTowardthe Use of Force of Death Penalty .16 (.031)* -.16 (.067)* -.14 (076) .11 (.066) .02 (.026) .08 (.037)* .01 (.020) -.00 (.025) .59 (.224)* .48 (.251)* -.15 (.217) .07 (.082) .02 (.122) .05 (.064) -.03 (.078) -.02 (.020) .18 (.045)* .16 (.044)* .08 (.029)* -.13 (.033)* .06 (.017)* -.01 (.005) .00 (.003) -.03 (.040) -.03 (.044) -.01 (.013) .11 Defense Spending .04 (.020)* .02 (.043) -.07 (.049) .00 (.042) -.04 (.017)* .00 (.024) .01 (.013) .05 (.016)* .34 (.143)* -.04 (.161) -.02 (.139) .10 (.052)* -.12 (.078) -.03 (.041) -.03 (.050) .02 (.013) .13 (.029)* .05 (.028) .08 (.018)* .03 (.021) .03 (.011)* -.00 (.003) .00 (.002) -.08 (.026)* -.06 (.028)* -.02 (.008)* .15

759

variables Independent Authoritarian values Perceivedideological distance Negativity of responseto presidentialcandidates Negativity towardpolitical parties Negativity of retrospectivenational economic evaluations Fear of unemployment Negativity of retrospectivepersonal economic evaluations Fearof nuclearwar * Authoritarianism Perceivedideological distance * Negativity toward Authoritarianism presidentialcandidates * Authoritarianism Negativity toward political parties * Authoritarianism Negativity of national economic evaluations * Authoritarianism Fearof unemployment * Authoritarianism Negativity of personal economic evaluations * Authoritarianism Fearof nuclearwar Christianfundamentalist Moral conservatism Political conservatism Partyidentification(Republican) Religiosity Gender(male) Age (10 years) Family income ($10,000s) Education(10 years) Political knowledge Constant AdjustedR2

Gulf War .02 (.034) -.14 (.075) -.68 (.085)* -.05 (.074) .02 (.030) -.02 (.042) -.01 (.023) -.01 (.028) .55 (.250)* .44 (.281)* .13 (.243) .17 (.091)* -.44 (.137)# -.01 (.071) -.09 (.087) -.01 (.022) .16 (.050)* .13 (.050)* .18 (.032)* -.01 (.037) .07 (.019)* -.03 (.006)* .00 (.004) -.04 (.045) .09 (.049) -.02 (.015) .16

Note: Entriesare unstandardized errors.* = p < .05 (one-tailed regressioncoefficients and standard test on interactioncoefficients, two-tailedtests for othercoefficients); # = p < .05 had a two-tailed test been appliedto the interactionterm instead. Source: 1992 NationalElection Study, non-Hispanicwhites completingboth pre- and postelection N questionnaires. = 1564.

760

Feldman Stenner and

values vary with perceived ideological How does the effect of authoritarian for distance?Let us takeanexample.Considerthe case of support the deathpenalty, which is about averagefor the size of this interactioncoefficient. For someone at the 10th percentile of the ideological distance measure, the predicted effect of authoritarian values is just .08-very close to zero. On the other hand, someone at the 90th percentileon the ideological distancemeasurehas a predicted scoring coefficient for authoritarianism .26. Since all the variablesare coded to be of of unit range,a coefficient of .26 can be interpreted the fractionof the rangeof the as variable(support the deathpenalty)acrosswhich a respondent for would dependent be predicted to move if authoritarianism increased from its minimum to its maximumvalue (holding all else constant).It is clear both that authoritarianism has a substantialinfluence upon intolerantand punitive attitudes, and that the perceptionof ideological distancestronglymoderatesthe influenceof authoritarian values. Apart from the ideological distance measure, the interactionwith negative reactionsto the presidentialcandidatesis significantand correctlysigned in four of the nine equations,as is the interaction with negativeperceptionsof the national economic situation,while the interactionwith negative reactionsto the political parties is significant in two equations. (And there are no significant negative interactioncoefficients in any of these cases.) In sharpcontrast,the interactionof authoritarian predispositionsand personaleconomic conditionsis not statistically with significant(in the predicteddirection)in any of the equations.The interaction fearof unemployment does littlebetteras it is significant(in thepredicteddirection) only once. The interactionwith fear of nuclear war is statisticallysignificant in threeequations,but the coefficients are relativelysmall. Since the largest and most consistent interactioneffects emerged for the interaction of authoritarianism with the ideological distance measure, we ran additional analyses to assure ourselves that these significant effect were not spurious.It is possible thatscoreson the ideologicaldistancemeasurearecorrelated with education, political knowledge, or ideological extremity. The significant interactioneffects might then be a function of these other variables and have nothing to do with the perception of ideological diversity. We therefore reestimatedall of the equationsadding,one at a time, interactionsbetween authoritarianpredispositionsand education,political knowledge, and ideological extremeffect of ity. None of these additionssignificantlyalteredthe estimatedinteraction authoritarian and ideological distance. Thus, this finding appears predispositions to be robustto alternativespecifications. These resultsprovide supportfor our interaction version of the threat-authoritarianism andno support all for the simplehypothesisof a directeffect at hypothesis of short-termthreat on authoritarian values. The largest and most consistent interactioneffects are registeredfor variables that tap perceived societal threat. When people perceive that the political parties and presidentialcandidateshold ideological positions very differentfrom them, or have very negative reactionsto

Perceived Threatand Authoritarianism

761

the presidentialcandidates,or see a deteriorating nationaleconomy, the impactof is greatlymagnified.These resultssuggest, then, that authoritarian predispositions ratherthan threats to individuals' immediate well-being-for example, to their household economic situation-it is perceived threat at the societal level that arouses the ire of authoritarians amplifies the impact of authoritarian and predison intolerantand punitive attitudes. positions Authoritarian Predispositions and the Effects of Threat To this point we have discussed the regressionresultsin termsof the extent to which various threats moderate the effect of authoritarian predispositionson a of dependentvariables. However, interactioneffects can always be interrange pretedfromtwo perspectives.We can also examinethe extentto which the various threat measures affect the dependent variables as authoritarianism varies. One obvioushypothesisis thatincreasingperceptionsof threatshouldbe relatedto more ethnocentricand punitive attitudesfor those high in authoritarianism. This effect of threatmightthen go to zero as authoritarianism decreases.Thatis, we mightfind thatpeople high in authoritarianism sensitiveto andprovokedby theperception are of threat,while those low in authoritarianism predispositionsare relatively unaffected by increasingthreat. Before re-examiningthe interactioneffects from this alternativeperspective, it is useful first to look at the coefficients for the threatvariables.Recall that, by our coding, these coefficients yield the effect of each threat measure on the values are at the sample mean. These dependent variable when authoritarian estimatesarevery close to those thatwe wouldhave obtainedby estimatinga model withoutthe interaction terms.Lookingacrossthe dependentvariables,we note that it is clear that the coefficients for the threatmeasuresare typically close to zero. Wherethey are statisticallydifferentfromzero, the coefficients arenegative.Thus, if we had tested a simple model that predictsa direct,unconditionalrelationship between threatand authoritarian and (intolerant punitive)attitudes,we would have found no evidence at all in supportof thatprediction. we Turningnow to the interactionof threatand authoritarianism, find support for only half of the expectedpattern. do findthe predictedeffect of threatamong We those high in authoritarianism. Here, greaterperceptionsof threatare associated with more ethnocentricand punitive attitudes.Moreover,the resultsclearly show that it is only among those high in authoritarianism threathas this effect. As that we have just seen, even given average levels of authoritarianism, none of these threatvariableshas a significantpositive effect on any of the dependentvariables. we result.Here,as And, amongthose low in authoritarianism, observea surprising of threat increase, we actually observe more liberal (libertarian) perceptions responses on our dependentvariables.This patternemerges for every one of the significant interaction effects across the nine equations. The role of threat is thereforemore complex than we originally anticipated.High levels of perceived

762

Feldman and Stenner

threattend to polarize the responses of those high and low in authoritarianism. Those high in authoritarianism become more punitive and ethnocentricunder conditionsof threat,while those low in authoritarianism become even less punitive and ethnocentricunderthese conditions. We might be tempted to attributethese results to some methodological problemwere it not for two others studies that find this same polarizationeffect. The first study comes from the terrormanagementresearch that we discussed earlier(Greenberget al. 1990). In one of the terrormanagementstudies, subjects were given a modified version of the F-scale. Recall that the threatmanipulation here involved making the subjects' mortality salient. When the effects of this were examinedseparately those high andlow in authoritarianism, for manipulation resultsemerged.As predicted,amongthose high in authoritarianism the divergent threatwas associatedwith morenegative attitudestowarda targetperson mortality constructedto be highly dissimilarto the subject. However, among those low in authoritarianism, mortalitythreatwas associated(thoughmore weakly) with more attitudestowardthe dissimilartargetperson. positive The second study, one we also discussed earlier,is the comparativeanalysis of the effects of economic threat on authoritarianism the U.S. and Russia in & Hinton, 1995). Using a 16-itemshortformof Altemeyer's (McFarland, Ageyev, scale-and bearin mind thatwe considerthis more Right Wing Authoritarianism a measureof authoritarian attitudes(manifestations)than of authoritarian predisposition-these researchersfound that economic threatleads to a polarizationof scores on the authoritarianism measure.Comparedto those in low-threatconditions, there were higher proportionsof people registering both high and low authoritarianism among those experiencing threat. Interestingly,this result appeared in the U.S. sample but not in the Russian sample. Apparently,then, this polarizationeffect is quite real, emerging in three very different studies of the authoritarian attitudesin the U.S. The absenceof impactof threaton characteristic the polarization effect in the Russian data indicates that more cross-national researchis needed to determinethe generalizabilityof this finding. Perceived Threat and the Structure of Authoritarian Attitudes The nine dependentvariables we have examined in these regressions were chosen because we believe them to be characteristic manifestationsof authoritarianism.If this is so, shouldn'tthese variablesbe stronglyinterrelated? authoritarIf ian predisposition is a common factor underlying these variables, we might reasonablyexpect to find a patternof correlationsamong the variablesreflecting this fact. However, the resultsjust presentedsuggest that the picture may be far more complex than this. We have shown that certaintypes of threatpolarize the attitudesof those high and low in authoritarianism, to causing high authoritarians become more intolerantand punitive, while low authoritarians become less so. Conversely,underconditionsof low threat,the attitudesof those high and low in

Perceived Threatand Authoritarianism

763

authoritarianism are less distinguishable. Thus, since the correlation between authoritarian predisposition and authoritarian manifestations varies with levels of threat, the correlations among the attitudinal manifestations of authoritarianism should likewise depend upon threat. Since the ideological diversity measure produced the largest and most consistent interaction effects with authoritarianism in the earlier analysis, we examined the correlations among the nine dependent variables separately for those respondents above and below the mean in perceptions of ideological diversity. These correlations are shown in Table IV. Correlations above the main diagonal are for those respondents reporting greater than average perception of threat, and those below the diagonal are for those respondents experiencing the low-threat condition. Overall, the correlations among the dependent variables-our manifestations of authoritarianism-are substantially lower for those below the mean in perceptions of ideological diversity. The average correlation is only .07 for this group. The average correlation among the dependent variables then increases to .20 for those above the mean in perceptions of diversity. Or consider just the correlations among the dependent variables of Table I: our three measures of attitudes toward minority groups. Under conditions of low threat the three correlations average .18, which rises to .30 when threat is above the mean.
Table IV. Correlations Attitudes Among Authoritarian Stereotyping 1.00 .24 .15 .05 .00 .20 .08 .05 .01 Black Effort .42 1.00 .13 -.04 .03 .13 .15 .01 .05 Order/ True Death Defense Gay $ Rights Ingroup Freedom American Penalty .17 .05 .05 .36 .22 .24 .30 1.00 -.04 .06 .14 .02 .09 .06 .04 .11 1.00 .06 -.13 .09 -.02 .12 .11 .11 .15 1.00 .01 .07 .01 .07 .33 .33 .04 .12 1.00 -.01 .13 -.01 .29 .23 .13 .14 .22 1.00 .06 .15 .30 .35 .05 .18 .31 .27 1.00 .16 Gulf war .02 .13 .18 .19 .22 .15 .18 .26 1.00

Racial Stereotyping Blacks Don't Try Hard Gay Rights Ingroup/ Outgroup Ordervs. Freedom True American Death Penalty Defense Spending Gulf war

Note: Entriesare correlationcoefficients. Those correlationsabove the diagonalare for those respondentsabove the mean in perceptionsof ideological diversity;those below the diagonalare for those below the mean in perceptionsof ideological diversity. Source: 1992 NationalElection Study, non-Hispanicwhites completingboth pre- and postelection N questionnaires. = 1564.

764

Feldman and Stenner

In addition to illustrating the polarizing effect of threat on authoritarian attitudes,these results raise serious concerns about the propertiesof F-scale-like authoritarianism measures.Since these measurestypically include a goodly proof items that are fairly concrete manifestationsof authoritarianism-like portion attitudes toward minority groups, conventional social behavior, and punitiveness-the observed internalconsistency of such scales may be in part a function of levels of threat. This may account for Altemeyer's (1988) finding that the intercorrelations scale (and among the items of his Right Wing Authoritarianism its estimatedreliability)variedconsiderablyover time. hence, On the other hand, our indicatorof authoritarian predispositions-the childvalues measure-was chosen in large partbecause it is a relatively lowrearing level, unobtrusivemeasure.As such, it shouldbe far less sensitive to the effects of threat.In fact, the mean correlationamong the four child-rearingitems increases only slightly between the low- and high-threatconditions:from .28 to .34. This nicely reinforcesour centralconclusion thatthreatdoes not have a directeffect on authoritarian predispositions,but ratherincreases the connections between those authoritarian attitudesandbehaviors. predispositionsanda varietyof characteristic DISCUSSION Although we cannot claim to have presenteda seamless patternof empirical interaction evidence, this analysis has found some intriguingand often substantial effects between various indicators of threat, authoritarian predispositions,and measuresof stereotyping,intolerance,andpunitiveness.Whatarethe implications of these results?Let us summarizethe majorfindingsof this analysisone at a time, and consider their significance for our understanding the impact of threaton of authoritarianism. so doing, we will also consider the broaderimplicationsof In these findings for our understanding the natureof authoritarianism of itself. We found no evidence to supportthe simple hypothesis that perceptionsof threatincrease levels of authoritarian predisposition.None of the correlationsof the threatmeasureswith authoritarianism largeandmanywere negative.Thus, was the aggregate-levelrelationshipsbetween societal threatand indicatorsof authoritarianismdo not result from a direct relationshipbetween threatand individuals' levels of the trait.This does not mean, however, thatwe can rejectthe hypotheses of researcherslike Frommand Wilson, who arguethatauthoritarianism shaped is by the long-term consequences of threat and insecurity. These hypotheses are beyond the scope of the presentstudy. As we anticipated,threatappearsto affect authoritarians increasing the by connection between their predispositions,and their political and social attitudes. This mechanismcan accountfor the aggregateresultsshowing effects of threaton authoritarian that behaviors, without having to adopt a view of authoritarianism substantial short-term in levels of the predisposition. From expects possibly change

Perceived Threatand Authoritarianism

765

the perspectiveadoptedhere, authoritarian predispositionsmay not change at all is not precluded),butthe manifestations authoritarianism-inof (althoughchange tolerance,prejudice,punitiveness,and the like-will be more pronouncedunder conditions of threat. Perhapsmore importantly,it is not just that threatmagnifies the effects of authoritarianism that the observed consequences of authoritarianism but depend the presence of threat. In the absence of threat, we find little heavily upon connection between authoritarian predispositions and the dependent variables. threat appears to be critical to the activation of authoritarianism. Thus, Any must,therefore,accountfor the central satisfactoryexplanationof authoritarianism role of threat.Althoughit might be possible to incorporate externalthreatinto the Adorno et al. (1950) explanationof authoritarianism, their focus was on the role of childhood socializationand internalconflict. It is not obvious how this account could incorporateour findings. Threat was certainly not a central part of their explanation.It is even less clear how Altemeyer's (1988, 1996) social learning accountcouldaccommodatea centralrolefor externalthreat.Altemeyerdoes argue thatperceptionsof a dangerousworld help to mediatethe effects of authoritarianism, but this is very differentfrom the interactiveeffect of threatevidenced in our analysis.Moreover,thereis nothingin Altemeyer'sexplanationto suggest why his should be activated only under conditions of acquired trait of authoritarianism threat. Whatkind of explanationswould farebetter?In partthe answerdependsupon the types of threatto which authoritarianism sensitive.Ouranalysisindicatesthat is The political threatsare especially salient to authoritarians. most substantialand consistent result was the pronouncedinteractioneffect of authoritarianism with perceived ideological diversity.The more ideological distanceauthoritarians perceive between themselves and the two partiesand the presidentialcandidates,the more prejudiced,intolerant,and punitivethey become. We also found substantial effects for the interactionof authoritarianism with negative assessments of the presidential candidates. When people consider the candidates to be bereft of positive qualities, arousingfeelings of anger and fear, the effect of authoritarian predispositionsis greatlyexacerbated.Shouldthe political threatsworkin tandem, the magnitudeof impact is very soberingindeed. For example, the influence that authoritarianism exerts on support for the death penalty is magnified by both of ideological diversityand fear and angertowardpolitical leaders. perceptions Economic threatsalso interactedwith authoritarianism several cases, but in note thatit is perceptionof a deteriorating nationaleconomy thatis consequential. In sharpcontrast,threatsto personal economic conditions-whether unemployment or more generalpersonaleconomic decline-appear to have no aggravating effect on authoritarians. Taken together, this pattern of results indicates that authoritarianism activatedand aggravatedby threatsto the political and social is order but not by threatsto personal well-being. We think this is a very critical is finding. Authoritarianism activatedwhen thereis a perceptionthatthe political

766

Feldman and Stenner

or social orderis threatened. This is at odds with most previousdiscussionsof the which have focused on directpersonalthreats. role of threatin authoritarianism, So where does this line of reasoninglead us? It is no doubtpremature argue to in favor of a particularconceptualizationof authoritarianism the basis of the on analysispresentedhere.But we can single out whatappearto be the most promising leads in previousdiscussions of authoritarianism. We earliernoted that threatis centralto Rokeach's (1960) theory of dogmatism. Rokeacharguedthatpeople develop closed-mindedbelief systems as a direct consequence of threat and anxiety. Although Rokeach did not explicitly consider the effects of transient threat on closed-mindedness, it is plausible that someone in possession of a rigid belief system that was generated to ward off threat would be most inclined to exhibit intolerance when faced with an anxiety-provoking situation. One of Rokeach'scentralclaims aboutbelief systemsis thatthey areorganized aroundthe distinctionbetweenbelief anddisbelief subsystems.Reactionsto others are based, therefore,upon degrees of similarity-dissimilarity. Thus, intoleranceis an outgrowth of the clash of beliefs. Although starting from a very different perspective,Duckitt (1989) has recently offered a comparableconceptualization of authoritarianism is based on intense groupidentification,and the resulting that "straintowardcohesion" (Duckitt, 1989, p. 63). Duckitt argues that groupmembershipandconformityto groupnormsprovidethe basis for self-esteem.The more intensean individual'saffective identificationwith theirgroup,the morededicated he or she is to its integrityand cohesion, resultingin conformityand conventionalism; respect for and submission to authority;and intoleranceand punitiveness towardthose who fail to conformto the ingroupnorms (Duckitt, 1989, p. 70). involves a heightenedattachment Accordingto Duckitt,then,authoritarianism to the ingroupand an associatedrejectionof the outgroup.This not only parallels Rokeach's distinctionbetween belief and disbelief systems, it likewise marksout a centralrole for threat.Duckittmakesreferenceto the extensive literature social in to psychology regardinghow externalthreatscan heightenattachment the ingroup and rejectionof outgroups.The parallelbetween Duckitt's focus on ingroup/outgroup distinctions and Rokeach's analysis of belief and disbelief systems may provide the basis for a more encompassingview of authoritarianism. Anotherperspectivethat likewise highlightsthe role of threatin dynamicsof authoritarianism focuses on the authoritarian's presumedneedfor orderor structure in the world. Such argumentshave been made in related,but somewhatdifferent ways, by Gabennesch(1972), Kohn (1977), and Forbes(1985). Kohn emphasizes orientationstowardconformityversus self-direction.His measureof child-rearing values-a directantecedentof the measuresused in this paper-taps the desirefor conformity.Threatis thus defined in termsof challenge to this conformity.Forbes are (1985) argues, likewise, that authoritarians conformists,who tend to yield to the opinionsof theiringroupin orderto wardoff the anxietyproducedby the threat of "isolationand loneliness"(Forbes, 1985, p. 160). The authoritarian's anxiety is

Threatand Authoritarianism Perceived

767

apparentlyreduced by cleaving to the ingroup, uncritically submitting to the ingroupnorms, insisting on the compliance of others, avoiding "disruptivecontacts"with outgroupmembers,and exaggeratingtheirdifferences. Finally, Gabennescharguesthatindividualswith limited perspectiveson the world around them-whether due to insularity, limited socialization, lack of education or experience-are left prone to a particularworldview that he calls social realityas "encompassing "reification." Such individualsseem to understand a superordinate normativedimension, an externallocus where events are determined, where moral authorityresides, and to which men must adaptthemselves" (Gabennesch,1972, p. 862-863). Thus, social institutions,customs andnormsare understoodnot as humanproducts,fashionedby humansand thus changeableby them, but ratheras installedby God, nature,or the cosmos: the productof some Institutions culturearethusaccordedreverenceandunquesand externalauthority. and authority" tioningobedience,being "superordinate infusedwithtranscendental (Gabennesch,1972, p. 864). The connection between Gabennesch'snotion of reificationand authoritariIndividualswho come to reify the social anism as a worldviewis readilyapparent. world are therebyinclined towardmoraland culturalabsolutism;to rigid conventionalism and unquestioningconformity to external authority;to disavow selffor directionandto vilify thehuman.Most important ourpurposeshere,we imagine that those who uncriticallysubmit to traditionalmoral authoritiesand the established social orderare bound to judge harshly,and refuse to tolerate,individual deviance and diversityof public opinion. More generally, we can expect them to be threatenedand disturbedby any events that challenge the self-evident truthof establishedbeliefs and the integrityof the social order. Although all of the perspectives we have mentionedhere provide a way of integrating (with varying ease) the role of threat into a conceptualizationof there is one last resultfrom our analysis that is more difficult to authoritarianism, It deal with, andhence more theoreticallydiscriminating. is not simply thatpeople are aggravated societal threat.Those who are who are high in authoritarianism by low in authoritarianism sensitive also, but they become less intolerantand are punitive under conditions of threat. Unfortunately,it has been common in the as literatureto treatlow authoritarianism just the absence of the authoritarianism authoritarian (see Kohn, 1972). But the polarizationeffect we have identified trait here indicates that much more attentionneeds to be paid to both poles of this dimension. Of the conceptualizationsof authoritarianism we have discussed here, that only Duckitt (1989) explicitly considers both ends of the continuum.He defines as which authoritarianism "theindividualor group'sconceptionof the relationship or should exist, thatis, the appropriate normativerelationship,between the group and its individualmembers"(Duckitt, 1989, p. 71). Fromthis definitionhe is able to provide a descriptionof both ends of the authoritarianism dimension (Duckitt 1989, p. 71): "At one extreme would be the belief that the purelypersonalneeds,

768

Feldman and Stenner

as inclinations,and values of groupmembersshould be subordinated completely as possible to the cohesion of the group and its members. At the other extreme of would be the belief thatthe requirements groupcohesion shouldbe subordinated as completely as possible to the autonomyand self-regulationof the individual in thus defines "low"scores on authoritarianism member."This conceptualization relations.The activationof on individual-group terms of a libertarian perspective this libertarianperspective by threat-just as the authoritarian perspective is activatedby threat-would accountfor the polarizationeffects thatwe, andothers, as have observed.In addition,Duckitt's focus on authoritarianism a group-related leads naturallyto the predictionthat it is threatsto group integrity, phenomenon ratherthan status, cohesion and/or identity that will activate authoritarianism, threatsto personalwell-being. The perspectiveswe have discussedhere providea differentand enlightening Our analysis provides no way of clearly way of looking at authoritarianism. distinguishingamong these viewpoints, but it does markthem out as some of the of most fruitfulavenuesto pursueif we areto furtherourunderstanding the origins anddynamicsof authoritarianism. Each,given theirsharedconcernwith threatand anxiety, is capableof encompassingthe empiricalevidence regardingthe interaction of authoritarian predispositionsand threat.Duckitt's perspectiveappears,at this point, to most comfortablyaccommodateour findings regardingthe importance of societal ratherthan personal threat,and regardingthe activationof the libertarian position by threat.But it is far too soon to claim direct supportfor any one perspectivebased upon the evidence presentedhere. It is our hope that these findings regardingthe centralrole of societal threatin activatingauthoritarianism of formulation one of the most enduring will lead us to develop a more satisfactory conceptsin the social sciences, andto betteraccountfor some of the most important phenomenain social behavior. APPENDIX: ADDITIONAL NES VARIABLES The moral conservatismvariable was constructedby summing four items measuring(1) the extent to which respondentsbelieved thatthe "newerlifestyles" were contributingto the breakdownof our society; (2) the extent to which they believed that this country would have many fewer problems if there were more family ties; (3) the extent to which they disagreedthatthe emphasison traditional world was always changing and we should adjustour view of moral behaviorto those changes;and (4) the extent to which they disagreedwith being moretolerant of people who chose to live accordingto their own moral standards, even if they were very differentfrom theirown. Religiositywas measuredby a simple additivescale createdfromfive individual items tapping an array of religious attitudes and behaviors. These items measuredwhetherrespondentsconsideredreligion to be an important partof their

Perceived Threat and Authoritarianism

769

life; the extent to which they said their religion providedguidance in day-to-day living; and the frequency with which they prayed, read the Bible, and attended religious services. Christian fundamentalistbeliefs were measuredby a dummyvariableindicata belief thatthe Bible was the actualword of God, "tobe takenliterally,word ing for word." Political conservatism was measured by respondents' placement of their political views on a seven-point scale ranging from "extremelyliberal"to "extremelyconservative." Finally,party identificationwas indicatedby responsesto the standard question "generally speaking, do you usually think of yourself as a Republican, a Democrat, an Independent,or what?" along with its follow-up probes gauging strengthof partisanship(from professed partisans),or partisan"leaning"(from those who had deniedpartisanship). Thus the final scale rangedacrossfive points, from a low of "0" for strong Democraticpartisans,throughto a high of "1" for strongRepublicans.

REFERENCES
Adorno, T. W., Frenkel-Brunswick, Levinson, D. J., & Sanford,R. N. (1950). The authoritarian E., personality.New York:Harper& Row. San Altemeyer,B. (1988). Enemiesoffreedom: Understanding right-wingauthoritarianism. Francisco: Jossey-Bass. MA: Harvard. Altemeyer,B. (1996). The authoritarianspecter. Cambridge, of Alwin, D. F., & Krosnick,J. A. (1985). The measurement values in surveys:A comparisonof ratings and rankings.Public OpinionQuarterly,49, 535-553. in Doty, R. M., Peterson,B. E., & Winter,D. G. (1991). Threatand authoritarianism the United States, 1978-1987. Journal of Personalityand Social Psychology, 61, 629-640. and A Duckitt,J. (1989). Authoritarianism groupidentification: new view of an old construct.Political Psychology, 10, 63-84. Forbes, H. D. (1985). Nationalism, ethnocentrism,and personality. Chicago: University of Chicago Press. Fromm,E. (1941). Escapefromfreedom. New York:Holt, Rinehartand Winston. as Gabennesch,H. (1972). Authoritarianism world view. AmericanJournal of Sociology, 77, 857-875. J., Greenberg, et al. (1990). Evidencefor terrormanagement theoryII: The effects of mortalitysalience on reactionsto those who threatenor bolster the culturalworldview.Journal of Personalityand Social Psychology, 58, 308-318. Kohn, M. L. (1977). Class and conformity.Chicago:Universityof Chicago Press. scale: A balancedF-scale with left-wing reversals. Kohn,P. M. (1972). The authoritarianism-rebellion Sociometry,35, 176-189. Marcus,G. E., Sullivan,J. L., Theiss-Morse,E., & Wood, S. L. (1995). Withmalice towardsome. New York:CambridgeUniversityPress. Martin,J. G. (1964). The tolerantpersonality.Detroit,Wayne State UniversityPress.

770

Feldman and Stenner

in McFarland,S. G., Ageyev, V. S., & Hinton, K. (1995). Economic threatand authoritarianism the United States and Russia. Paperpresentedat the 1995 meeting of the International Society for Political Psychology. Rokeach,M. (1960). The open and closed mind.New York:Basic Books. Rosenblatt,A. et al. (1989). Evidence for terrormanagement theoryI: The effects of mortalitysalience on reactionsto those who violate or uphold culturalvalues. Journal of Personality and Social Psychology,57, 681-690. Ross, L., Greene, D., & House, P. (1977). The 'false consensus effect': An egocentric bias in social Social Psychology, 13, 279-301. processes.Journal of Experimental perceptionand attribution and Sales, S. M. (1972). Economic threat as a determinantof conversion rates in authoritarian nonauthoritarian churches.Journal of Personalityand Social Psychology,23, 420-428. Journalof PersonalityandSocial Psychology, Sales, S. M. (1973). Threatas a factorin authoritarianism. 28, 44-57. of Sales, S. M., & Friend,K. E. (1973). Success and failureas determinants level of authoritarianism. BehavioralScience, 18, 163-172. Staub,E. (1989). The roots of evil. Cambridge: CambridgeUniversityPress. Stenner, K. (1996). Societal Threatand Authoritarianism: Racism, Intolerance,and Punitivenessin America. Part I: Evidence from the General Social Survey 1972-1994. Paperpresentedat the AnnualMeeting of the AmericanPolitical Science Association, San Francisco. Wilkinson,R. (1972). The brokenrebel. New York:Harper& Row. Wilson, G. (1973). Thepsychology of conservatism.New York:Academic Press.