Political Chamelon by demandside

PolitBehav DOI10.1007/s11109-016-9335-y

PoliticalChameleons:AnExplorationofConformity inPoliticalDiscussions

TaylorN.Carlson1 • JaimeE.Settle2

SpringerScience+BusinessMediaNewYork2016

Abstract Individualsdonotalwaysexpresstheirprivatepoliticalopinionsinfront ofotherswhodisagree.Neitherpoliticalscientistsnorpsychologistshavebeenable toﬁrmlyestablishwhythisbehavioroccurs.Previousresearchhasexplored,at length,socialinﬂuenceonpoliticalattitudesandpersuasion.However,theconcept ofconformitydoesnotinvolveattitudechangeorpersuasion;itmoreaccurately involvesself-censoringtomatchasociallydesirablenorm.Inanefforttoimprove ourunderstandingofthisbehavior,weconducttwoexperimentstoinvestigate perceptionsandbehavioralresponsestocontentiouspoliticalinteractions.Study1 askedparticipantstopredicthowahypotheticalcharacterwouldrespondtoa varietyofpoliticalinteractionsamongcoworkers.InStudy2,participantsdiscussed politicalissueswithconfederateswhowerescriptedtodisagreewiththem.The studiesrevealthatindividualsareuncomfortablearoundpoliticalinteractionsin whichtheyholdanopinioncountertothegroup.Participantsbothexpecteda hypotheticalcharactertoconforminStudy1andactuallyconformedthemselvesin thelabsessioninStudy2.

Keywords Conformity Discussion Contention Politics Opinions

Electronicsupplementarymaterial Theonlineversionofthisarticle(doi:10.1007/s11109-016-9335-y) containssupplementarymaterial,whichisavailabletoauthorizedusers.

& TaylorN.Carlson tfeenstr@ucsd.edu

1 DepartmentofPoliticalScience,UniversityofCalifornia,SanDiego9500GilmanDr.,LaJolla, CA92093,USA

2 GovernmentDepartment,CollegeofWilliam&Mary,MortonHall10,100UkropWay, Williamsburg,VA23187,USA

Introduction

TheAmericanpoliticalenvironmentisbothsocial(Berelsonetal. 1954;Campbell etal. 1960;Giusefﬁetal. 2013;KatzandLazarsfeld 1955;Lazarsfeldetal. 1968; Mondak 2010;Putnam 2001;Settleetal. 2011;Sinclair 2012;Zuckerman 2005) andpolarized(Abramowitz 2010;AbramowitzandSaunders 2008;Bafumiand Shapiro 2009;HaidtandHetherington 2012;HetheringtonandWeiler 2009; Iyengaretal. 2012;IyengarandWestwood 2015;Mason 2013, 2015).Asthe politicaldivisionsinAmericansocietybecomeincreasinglysalientandentangledin oursociallives,decisionsaboutwhatpoliticalinformationtodisclosetoothers becomesevenmoredelicate,assharingevennon-politicalinformationmayreveal likelypoliticalpreferences.Forexample,indicatingapreferenceforimportedbeer likeGuinnessorHeinekenoverdomesticbeerlikeBudweiserorMillercansignal liberalpoliticalideology(Khanetal. 2013).Liberalsandconservativeshavebeen showntodifferintheirdating(Klofstadetal. 2012),art(Wilson 1973),food,pet, andmoviepreferences(Haidt 2014;HaidtandWilson 2014).Sharingone’s opinionsmaybeespeciallyriskybecauseoftheincreasinglyhostileattitudes individualshavetowardout-partymembers(Iyengaretal. 2012;Iyengarand Westwood 2015;Mason 2015)andincreasedlevelsofpartisanbiasandanger (Abramowitz 2006, 2010;AbramowitzandSaunders 1998, 2005;Levendusky 2009;Mason 2013).Individualsmustchoosebetweensharingtheirtruepolitical opinions,amodiﬁedversionoftheiropinions(Cialdinietal. 1973;Hayes 2007; Hayesetal. 2005),ornothingatall(Noelle-Neumann 1993)astheynavigatewhat theyarewillingtorevealinthiscontentioussocio-politicalenvironment.

Thebalanceofopinionsinagivensocio-politicalinteractionwillinfluencean individual’scalculusaboutwhichpoliticalinformationtodisclose.Whenaperson interactswithotherlike-mindedindividuals,thisdecisionisnotnearlyasdifficult, becausethereislittleriskofsocialrepercussionsforsharingapoliticalopinionwith whicheveryoneagrees.However,individualswhofindthemselvesinapolitical opinionminorityfaceadifferentcost-benefitanalysis.Individualsinsuchaposition mightanticipateadversesocialconsequencesfordisagreeingwiththegroup (Huckfeldtetal. 2004;Mutz 2006)thataltertheircalculusonwhichpolitical informationtodisclose.Inanefforttoavoiddiscomfortandsocialostracization, theseindividualscouldmodifythepoliticalopinionstheyexpresstomatchthoseof thegroup.Thisstudyaimstounderstandwhetherindividualsbehavelikepolitical chameleons,temporarilyabandoningtheirtruepoliticalopinionsto conform to otherswhodisagree.

Despiteextantresearchonconformityfromapsychologicalperspectivedating backtothe1950s,thereislimitedevidenceof political conformity.Thelimited workonpoliticalconformitygenerallyconceptualizesconformityasupdated politicalpreferencesthathaveactuallychangedtomatchsocialnormswithina person’ssocialnetworks(Huckfeldtetal. 2004;HuckfeldtandSprague 1995;Mutz 2006;Sinclair 2012)orgroupnormsmorebroadly(LevitanandVerhulst 2015; Mutz 1998;Suhay 2015).Inotherwords,thesestudiesconsiderconformityasa changeinpoliticalattitudesinresponsetosocialinﬂuence.Weconceptualize

conformitydifferently.Politicalconformitydoesnotinvolveactualattitudechange, butinsteadinvolvesleadingotherstobelieveyousharetheirpoliticalviews,evenif youtrulydonot.Itisnotaresultofupdatingtruepoliticalbeliefsbasedonlearning newinformationfrompeers.Politicalconformityisabehavioralresponsetothe socialdiscomfortandstressstemmingfrompoliticaldisagreement.(Huckfeldtetal. 2004)arguethatcitizensareimmunetopressurestoconformiftheirpolitical preferencesaresociallyinvisible.However,theincreasingoverlapbetweenpolitical andsocialpreferences—andtheadventofsocialmedia—makeitmoredifﬁcultto keepone’spoliticalpreferencescompletelyshieldedfrompublicview.

Blendingsocialpsychologicalprincipleswithpoliticalscienceconcepts,we theorizethatface-to-facepoliticalinteractionsareuncomfortableandthisdiscomfortinﬂuencesthewayindividualsengagewithoneanotheraboutpolitics. Regardlessoftheirinterestinit,peopleexperiencepoliticsthroughoffhand observationsandinteractionswithpeopleintheirdailylives.Thesegenerally informal,face-to-faceexchangesarehardtocaptureandquantify,andthereis limitedevidenceabouthowpeopleperceiveandbehaveinresponsetothesecasual politicalinteractions.Althoughmostpeople,mostofthetime,arenotpayingmuch attentiontopolitics(Converse 1964),theincreasinglypolarizedandsocialnatureof politicscanforcepeopleintopoliticaldiscussionswheretheymuststrategically decidewhichpoliticalpreferencestodisclose.Motivatedbyadesiretoavoidthe socialconsequencesofpoliticaldisagreement,wearguethatindividualswill temporarily conformtoagroup’spoliticalopinion.Individualswhosensethatthey areinapoliticalopinionminoritywillallowotherstobelievethattheyallagree. Inanefforttoimproveourunderstandingofpoliticalconformity,weconducttwo experimentstoinvestigateanticipatedandobservedbehavioralresponsesto contentiouspoliticalinteractions.Study1askedparticipantstopredicthowa hypotheticalcharacterwouldrespondtoavarietyofpoliticalinteractionsamong coworkers.InStudy2,participantsdiscussedpoliticalissueswithconfederateswho werescriptedtodisagreewiththem.Thestudiesrevealthatindividualsare uncomfortablearoundpoliticalinteractionsinwhichtheyholdanopinioncounterto thegroup.Importantly,participantsbothexpectedahypotheticalcharacterto conforminStudy1andactuallyconformedinthelabsessioninStudy2.1

TheoreticalBackgroundandMotivation

Wemotivateourexplorationbyexplainingwhywemightexpectindividualsto conformtoothersinpoliticaldiscussions.Weﬁrstdevelopourargumentaboutthe inherentlysocialnatureofpolitics,andwhychangesintheperceivedcontentiousnessofpoliticshaveheightenedthesalienceofthesocialdimensionofpolitical behavior.Wethenconnecttheoriesandempiricalevidencefromsocialpsychology topoliticalsciencetodescribethetheoreticalfoundationsofpoliticalconformity.

1 Replicationdataandcodearepubliclyavailableon PoliticalBehavior’s Dataversepage https:// dataverse.harvard.edu/dataverse/polbehavior.

ASocial,Stressful,PoliticalWorld

Lasswell(1936)famouslywrotethat‘‘Politicsiswhogetswhat,when,andhow,’’ anobservationthatacknowledgesboththesocialandcontentiousaspectsofpolitics, andtheinterconnectionsbetweenthem.Electoralpoliticshighlightstheunderlying politicalconﬂictinsociety,makingcontentionmoresalientandvisible,andthus servingtoreinforcetocitizensthepervasivenessofdisagreementinthepolitical system.Asigniﬁcantportionofthiscontentionisgeneratedbydisagreementin policypreferencesbetweenelites.However,contentioninpoliticsextendsbeyond policydebates,andmuchofthewaypeopleexperiencethecontentionofpoliticsis throughparticipatinginsocialinteractions,orwatchingothersdoso.We conceptualizepoliticsasfundamentallyaprocessofpotentiallycontentioussocial interaction;peoplemayvoteintheballotboxalone,butlittleelseisdoneinasocial vacuum.

Theseanticipatedorexperiencedsociallycontentiouspoliticalinteractions inﬂuenceourperceptionsofpolitics,aswellasourbehavior.Forinstance,Hibbing andTheiss-Morse(2002)reportthatrespondentsinanationallyrepresentative samplewhofeeluneasyanduncomfortablearoundpoliticalargumentsaremore likelytosupport‘‘stealthdemocracy,’’aformofgovernmentwheredemocratic proceduresexistbutarenotvisibleandwhereordinarypeopledonotneedtoget involvedwithpolitics.Similarly,(Noelle-Neumann 1993)proposedthe‘‘spiralof silence’’theory,whichsuggeststhatsomepeoplewillsilencetheiropinionsinfront ofothersasawaytodisengagefromthesourceofcontention.Furtherresearch suggeststhatthepeoplemostsusceptibletooptingoutofpoliticalinteractionshave personalitytraitsthatmakepoliticaldiscussionstressfulanduncomfortable(Gerber etal. 2012;Hibbingetal. 2010;UlbigandFunk 1999).Theseﬁndingshighlight thatsomeindividualsareuncomfortablearoundpoliticalargumentsandthat disengagementfrompublicexchangeofopinionispreferable.

Thesefindingsaside,politicalscientistshavenotdevotedsignificantenergy tryingtounderstandwhichaspectsofpoliticalinteractionsmightbemoststressinducing.Pilotdata2 aimedatprovidinguswithguidancethattheliteraturecould notgiverevealevidencethatmanypotentiallystressfuldimensionsofthepolitical spherecausesomepeopleanxiety.Evenexposuretotheexistenceofconflictwas anxietyinducing:81.5%ofrespondentsreportthatreadingapollshowingthatthe oppositioncandidatewaswinningwouldcausethemtobeanxious,aswould knowingthattheywereapartisanminorityintheircommunity(67.1%)oramong theirfriends(64.4%).Beyondawarenessofcontention,largeproportionsofthe samplereportedthatobservingmanifestationsofthiscontentionwouldcausethem tobeanxious,situationssuchasseeingaprotestinthearea(65.3%),watchinga televisedpoliticaldebate(63.7%),orreadingapostonFacebookthatdisagreed withtheirviews(54%).Thepossibilityof engaging incontentioussituations appearstobesimilarlyanxiety-inducing,as76.3%ofthesamplereportedanxiety

2 Indevelopingourexperiments,wepilottestedavarietyofpoliticalstressorstimuli.Weasked individualsinaparticipantpool(n ¼ 280)atalarge,westernpublicuniversitytoidentifywhichofa seriesofpoliticalsituationswouldcausethemtobeanxious.Seethe‘‘Appendix’’sectionforfulldetails ofthepilotstudies.

atthethoughtofhavingadisagreeableconversationwithfriendsorneighbors. Interestingly,approximatelyathirdofthesamplereportedthatevenagreeablefaceto-faceoronlinepoliticalinteractionswouldcausethemtobeanxious.

Basedontheresultsthatinterpersonalinteraction—contentiousornot—hadthe potentialtobeasignificantsourceofstressinthepoliticalsphere,inthecontextofa surveyexperimentforarelatedprojectconductedonMechanicalTurk,wealso gatheredopen-endedresponsesfrom440surveyrespondentsaboutthefacetsof politicsthatarestressfulforindividuals.Ourcodingschemerevealedthatresponses fellintooneofthreemajorcategories:stressorsrelatedtotheoverarchingprocessof politics(suchasthebehaviorofcandidatesorofficials,orthewaythemediacovers politics);stressorsrelatedtothecontentofpolicy;orstressorsrelatedto participation.Ofthe1320messageswecoded,22%relatedtoparticipation;of thosecomments,42.8%identifiedfactorsrelatedtoindividualparticipation(such astryingtounderstandpoliticsordeterminingone’sbeliefs)while57.2%relatedto interpersonalinteractions.

Thesepilotstudiesweredesignedtoprovideageneralintuitionabouthow individualsrespondedtopoliticallycontentiousstimuliandtoprovidesomeface validitytothegeneralnotionthatengagingincontentioussituationscanbeanxietyinducing.Thedatacomefromconveniencesamplesofcollegestudentsand MechanicalTurkworkers,whichmeansthattheymightnotbereﬂectiveofthe generalpopulation.However,theexperimentalresultswepresentinthispaperrely onstudentsamples,sothestudentsamplepilotdataarequiteinformativeforour particularsample.

Why Political Conformity?

Becauseconformityhasbeendemonstratedinawidevarietyoftasks(Asch 1956; Crutchﬁeld 1955;Latane 1996;NowakandVallacher 2001)inmanycross-cultural populations(Barryetal. 1959;BondandSmith 1996;Hofstede 1980;Triandis 1990),weexpectthatindividualswillalsoconformtothegroup’sopinionona politicalissue.Basedonexistingpsychologyliterature,weknowthathuman conformityisexplainedbythreegoalscentraltohumanbehavior:accuracy, afﬁliation,andpositiveself-conceptmaintenance(CialdiniandGoldstein 2004). Becausethesegoalsarecentraltohumanbehavior,weexpectthesamegoalstoalso motivate political conformity.

Althoughthisstudydoesnotdirectlytestthesemechanisms,itisplausiblethat theyarestillatwork.Forexample,thehumanneedtobeaccuratemotivates conformityespeciallyamongthosewhoareuncertain,havelimitedinformation,or arepresentedwithambiguousanswerchoices(Cialdini 2001).Thewidespread researchonpoliticalattitudeformationandchange,particularlywithinthecontext ofsocialnetworks,demonstrateshowaccuracymightbeanimportantmotivatorof politicalconformity.Politicalscientistshavedemonstratedthatpoliticallyuncertain individualsturntoknowledgeablemembersoftheirpeergroupsorelitestohelp informtheirpoliticalopinions(Ahnetal. 2010, 2014, 2013;DruckmanandNelson 2003;LupiaandMcCubbins 1998;Ryan 2010, 2011),whichleadsustoexpectthat uncertainindividualsmightrelyoninformationalcuesfrompeergroupswhen

statinganopinion.Additionally,priorworkhasshownthattheopinionofa homogenoussocialnetworkcansignalthatanattitudeisvalid(Festinger 1950), whichincreasesanindividual’slevelofconﬁdenceinholdingthatattitude(Levitan andVisser 2009).However,thesestudiesinadditiontothebountifulresearchon socialnetworksandattitudechange(HuckfeldtandSprague 1995;Huckfeldtetal. 2004),focusonattitudeformationandattitudechange,whicharedistinctfrom conformity.

Thehumandesireforgroupaffiliationcouldalsomotivatepoliticalconformity. Socialpsychologyresearchindicatesthatconformityandotherformsofbehavioral mimicrycanbuildrapportandhelpdevelopsocialrelationships(Chartrandand Bargh 1999)andmaybeemployedtogainsocialapprovalfromothers(Cialdiniand Goldstein 2004).Becauseindividualswanttofeelasenseofbelongingand affiliation,itislikelythattheywouldwanttopresentthemselvesassimilartoothers withwhomtheyinteract—self-expressionofpoliticalviewsshouldbenoexception. AlthoughindividualistculturesliketheUnitedStatestendtopromoteuniqueness insteadofconformity(Cialdinietal. 1999;KimandMarkus 1999),thestressand discomfortsurroundingpoliticaldisagreementmightbemorepotentthanthedesire tobeunique.Thedesireforaffiliationmightpromotepoliticalconformityeven moreinthepolarizedAmericanpoliticalenvironmentthatemphasizesan‘‘usversus them’’mentality,especiallyifweconsiderpartisanshipasaformofsocialidentity (Greenetal. 2002).

Individualsalsoconformtogroupbeliefstoprotecttheirself-esteemandselfconcept.Individualscanenhance—oratleastmaintain—theirself-esteemby conformingtogroupstheyvalue(BrewerandRoccas 2001;CialdiniandGoldstein 2004;Pooletal. 1998).Individualsmightvaluecertainsocialgroupssuchas family,coworkers,orparticularfriendgroups,regardlessoftheirpolitical afﬁliations.But,ifanyofthesevaluedgroupsconsistentlyholddifferentpolitical opinionsthananindividual,heorshemightfeelpressuredtoconformtotheir opinionstomaintainhisorheridentitywiththatgroup.Whilesomeindividuals maintaintheirself-esteembypromotingasenseofuniqueness(BlantonandChristie 2003;KimandMarkus 1999),whichcouldleadthemto avoid conformity,others willexperiencegreaterself-esteemfromidentifyingwithagroup,leadingthemto conformpolitically.

Althoughthesocialpsychologicalprinciplesexplainingsocialconformityshould alsoexplainconformityinapoliticalcontext,thereisauniquetensioninthe politicalcontextthatdistinguishesitfromothercontextsexploredinsocial psychology.Individualslivingindemocraticpoliticalsystemsarenotonlyallowed toparticipate,theyareexpectedto.Individuals,particularlyinAmerica,areoften inﬂuencedbyappealstotheir‘‘civicduty,’’(Campbelletal. 1954;Downs 1957; GerberandGreen 2000;Gerberetal. 2008)andpartofperformingone’scivicduty involvesengaginginpoliticaldiscussionanddeliberation(Bennettetal. 2000; Dalton 2008;Dryzek 1994;Lasswell 1941).

ThestrongsenseofcivicdutycoupledwithindividualistAmericanculturethat emphasizesuniquenessmightpressureindividualstoexpresstheirtruepolitical opinions,regardlessofthesocialconsequencesfordoingso.Motivatedbytheir desiretoperformtheircivicduty,Americansmightfeelpressuredtocontributeto

politicaldiscussions.Thiscanbetroublingforindividualswhodisagreewithothers participatinginthediscussionbecausewhiletheymightfeelpressuretoparticipate, doingsocouldrevealdisagreementresultinginundesirablesocialconsequences. Individualsinpoliticalminoritiesarethereforeplacedinachallengingposition wheretheymustartfullybalancetheexpectationtoparticipateinpolitical discussionswiththedesiretomaintaintheirsocialrelationships.Temporarily conformingtothegroup’sopinioncouldbeaviablecompromisetoachieveboth simultaneously.

TheoreticalExpectations

Putsimply,weexpectthatindividualsinapoliticalminoritywillconformtothe majoritypoliticalopinion.Whendiscussingpoliticswithagroupofotherswho disagree,individualswillsuccumbtopressurestoconform,motivatedbytheir desiretoalleviatethediscomfortassociatedwithpoliticaldisagreementand contention.Weexpectindividualsinapoliticalminoritytoconformwithgreater frequencythanindividualsinmorebalancedorhomogeneousdiscussionnetworks. Thepressuretoconformisgreatlyreducedwhenminoritydissentersarenotalone (Asch 1956).Whenindividualsviewthemselvesasthe only onesholdinga particularviewpointinaconversation,theywillbemotivatedtotemporarily conformtothegroup’sopinion.3 Wearticulatespeciﬁchypothesesforeachstageof ourstudy,butwebroadlyexpectthatindividualsinapoliticalopinionminoritywill bemorelikelytoconform.

Method&Sample

Wetesttheseexpectationsintwoexperiments.Weﬁrstuseavignetteexperimentto investigatewhetherindividualswillexpecthypotheticalcharacterstoconformin twodifferent,potentiallycontentiouspoliticalinteractions(Study1).Wethentest whetherindividualsactuallyconformthroughanAsch-basedlabexperimentwhere wedirectlyobserveandmeasureconformingbehavior(Study2).

Thedataforthisstudywerecollectedaspartofthepoliticalsciencestudent subjectpoolatasmallpublicuniversityontheeastcoast.Inthefalliterationofthe survey,studentstookapretestincludingbasicdemographicquestions,political interest,politicalknowledge,andpoliticalissuepositions.Approximatelythreedays later,theycameintothelabandeitherparticipatedinStudy1(n = 201)orStudy2 (n = 70),butnotboth.Inthespring,Study1wasrepeated,butconductedatthe sametimeasthepretestquestionsandtheentirestudywastakenonline.Atotalof 432studentsparticipatedinStudy1and t testsrevealthattherewerenodifferences inanswersonthedependentvariablesbetweenthesetofparticipantsthattookthe

3 Wefurthermoredonotexpecteveryoneinapoliticalminoritytoconform.Weexpectthatthereare individualdifferencessuchaspartisanattachment,politicalinterest,conﬂictavoidance,andsocialanxiety thatcontributetoanindividual’ssusceptibilitytopoliticalconformity.Weexploretheseindividual differencesinfuturework.

studyinthelabinthefallandthesetthattookitonlineinthespring.4 Summary statisticsaboutthesampleareshowninTable 5 inthe‘‘Appendix’’section.

Study1 Design

Wechosetofirstemployavignetteexperimentbecauseitisdifficulttoobserve organicpoliticaldiscussionsintherealworldandvignetteexperimentsare especiallyusefulwhenreal-worldobservationsofthebehaviorsofinterestarenot practical,feasible,orethical(Caroetal. 2012).Theexperimentwasdeliveredusing Qualtricssoftwareandwasembeddedintoalargersurveythatcoveredthevariety oftopicsdescribedabove.Onthefirstscreenofthestudy,participantsreadthe vignette.Thewordingofthevignette,andthefactthatthecharacteridentifiedwith thesubject’spoliticalparty,wereintendedtohelptheparticipantputhimorherself intothepositionofthefictionalcharacter.Inaddition,wedevelopedthecontextof thevignettebasedonthepilotdatadescribedpreviouslyinwhichindividuals describedwhataboutpoliticstheyfoundstressful.Vignetteexperimentsaremost effectivewhenthevariablesinthevignettefittheparticipants(Weber 1992).Levy andDubinsky(1983)andSchoemaker(1993)suggestthatvignettescanbe constructedtobettermatchtheparticipantsiftheyarebasedondescriptionsofthe situationsresearchersaimtoemulateintheirvignettes.Assuch,webasedour vignettesonpilotdatainwhichindividualsdescribedcontentiouspolitical interactions.Furthermore,wechosetofocusonaworkplaceenvironmentbecause politicaldisagreementismorecommonamongweakties,suchascoworkers (Huckfeldtetal. 2004;Mutz 2006;MutzandMondak 1998).Weemployeda2 9 2 designandparticipantswererandomlyassignedtooneofthefourresulting vignettes.

Considerthefollowinghypotheticalsituation.Whileyouarereadingthis story,pleasetrytoimagineyourselfinSally’sposition.Asyoureadthrough thedescriptionofthissituation,pleaseimagineasconcretelyandvividlyas possiblewhatSallyisthinkingandfeeling,andhowherthoughtsandfeelings willshapeherbehavior:

Sallyisaregistered[samepartisanshipassubject]andhasalwaysvotedfor [samepartisanshipassubject]candidates.Basedonconversationsfrom previouselections,Sallyknowsthathercoworkers [CONTEXTUAL MANIPULATION].LastThursdaymorningattheofﬁcewhenshewentto thecommonroomtopourherselfsomecoffee,severalofhercoworkerswere standingaroundtalkingabouttheupcomingelection.Sallystartedtolisten, andrealizedthatthegroupwastalkingabout [PROVOCATION

4 ThedifferencesincovariatesshowninTable 5 relatedtopoliticalengagementarelikelyduetothefact thattherewasagubernatorialelectionduringthefallsemesterwhilethestudywasbeingﬁelded.

MANIPULATION]

.Allofasudden,onecoworkerturnstoSallyandsays, ‘‘Sally,ofcourseyou’revotingfor[oppositionpartycandidate],aren’tyou?’’

Thevignettescriptimpliesthatthecharacterhasaclearpreferencefora candidatethatcontradictsthepreferenceofthecoworkerwhoposesthequestionto her.Thetwomanipulationsweredesignedtoalterotherfeaturesoftheinteraction. Intheﬁrstmanipulation,thecontextualmanipulation,wedescribethecontextofthe conversationusingtwodifferentconﬁgurationsofopinioninthegroup.Inthe PartisanMinority condition,thecharacterwasdescribedasbeinginanopinion minority(thetext‘‘Sallyknowsthathercoworkersarealmostallregistered [oppositepartyofSally]’’wasinsertedintothevignette).Inthe BalancedParty condition,textstatingthat‘‘Sallyknowsthathercoworkersarefairlyevenly dividedwithequalnumbersofDemocratsandRepublicans’’wasinsertedinstead. Theprovocationmanipulationfeaturedthe SupportOpponent condition(inwhich Sally’scolleaguesdiscussedtheirsupportfortheoppositionpartycandidate’s policypositions)andthe OpposeFavoredCandidate condition(inwhichthe colleaguesdiscusshowmuchtheyopposethepolicypositionsofthecandidateSally supports.)

Afterthevignette,thesubjectansweredaseriesofquestionsoftheformat‘‘What isthelikelihoodthat...’’Foreachofthesepost-treatmentdependentvariables, respondentsansweredona1–5scale,with1labeledas‘‘veryunlikely’’and5 labeledas‘‘verylikely.’’Weoperationalizeourconstructofinterest,political conformity,usingthequestion‘‘WhatisthelikelihoodthatSallyexpresseshertrue opiniontothegroup?’’Thisdependentvariablewaswordedtomakeclearthatthe character’strueopinionwasnotinﬂuencedbythegroup’sopinion,andthus assessessolelytheextenttowhichasubjectexpectsthecharactertopublicly conform.

Wearealsointerestedinthreeadditionaldependentvariablesthatmayhelpto elucidatewhysubjectsanticipatethatthehypotheticalcharactermightconform.The ﬁrstassessestheanticipatedimmediateresponsetothesituation.Ourtheory suggeststhatoneofthemotivatingfactorsforconformityisthediscomfortof engaginginpoliticaldisagreement,andthuswemeasuredtheparticipant’sestimate ofthelevelofdiscomfortexperiencedbythecharacterwiththequestion‘‘Whatis thelikelihoodthatSallyfeelsuncomfortableansweringthisquestion?’’

Additionally,wealsosoughttoassesswhetherparticipantsthoughtthatthis potentiallyuncomfortablesituationmighthavesubsequentconsequencesforthe characterandthusthefinaltwodependentvariablesexaminedthepotentialsocial ramificationsofthesituation.Thereisincreasingevidencethatpartisanshiphas becomeasalientsocialidentity,andwewantedtoinvestigatewhetheradescription ofanuncomfortablepoliticalencountermightchangeparticipants’responsesabout thecharacter’ssocialbehavior.Althoughnotadirecttestofaffiliationmotives drivingthepotentialforconformity,demonstratingadifferenceinanticipatedfuture socialbehaviorwouldsuggestthatrespondentsaremakingthelinkbetweenthe exchangeofpoliticalviewpointsandthenatureofthecharacters’social relationships.Wemeasurethisusingtwoquestions:‘‘Whatisthelikelihoodthat

Sallyinviteshercoworkersoverfordinnerinthenextsixmonths?’’and‘‘Whatis thelikelihoodthatSallywantstolookforanewjobwithinthenextyear?’’

Study1Hypotheses

Ourtheoryaboutwhatdrivespoliticalconformityisbasedprimarilyonthe conﬁgurationofopinionswithinthegroup,andthusweexpectthelargesteffects betweenthetwoconditionsinthecontextualmanipulation.Basedontheﬁndingsof Asch(1956),weexpecttoobservemoreconformityinthePartisanMinority conditionovertheBalancedPartisancondition.Toourknowledge,buttherehave beennopreviousstudieslinkingthewayinwhichdisagreementisprovokedtoa person’sresponse.Theprovocationmanipulationwasdesignedtoinduceachange inthesocialcostofreplyingbyalteringtheimplicitlydemandedresponsefromthe hypotheticalcharacter.IntheSupportOpponentcondition,Sallycouldvoiceher supportforherpreferredcandidatewithoutdirectlycontradictingwhather colleagueshavesaid;conversely,intheOpposeFavoredCandidatecondition, expressingsupportforherpreferredcandidatewouldforceSallytodisagreedirectly withthecollectiveopinionthathasbeenexpressed.Thus,weanticipatethatthe OpposeFavoredCandidateconditionmayalsocausehigherreportedanticipated conformity,althoughweconsiderthishypothesistobeexploratoryintheabsenceof previousliteratureonwhichtobasethisexpectation.

Second,wealsoexpectthePartisanMinorityconditiontocreatehigherlevelsof discomfortandagreaterthreatofdownstreamsocialconsequences:participantsinthe PartisanMinorityconditionwillreportmoreanticipateddiscomfort,expectSallytobe less likelytoinvitehercoworkersoverfordinner,andexpecthertobe more likelyto lookforanewjobthanparticipantsintheBalancedPartycondition.Weexpectthatthe OpposeFavoredCandidateconditionmayinducehigherlevelsofreporteddiscomfort, butwedonotexpectthedifferencebetweentheprovocationmanipulationstobe strongenoughtoaffectthedownstreamsocialconsequencevariables.

Finally,weexpectthatsubjectswhoreporthigherlevelsofdiscomfortwillbe lesslikelytoreportthatthecharacterwillreporthertruefeelings.Itispossiblethat theeffectofthetreatmentsonreportedlikelihoodofconformityis mediated bya subject’sreportofanticipateddiscomfort,butonlyifthetreatmentaffectsthe reporteddiscomfortlevelanddiscomfortlevelaffectsreportedconformity.

Results

SummaryStatistics

Webeginwithdescriptivestatistics.Whilemanyparticipantsmarkedthemidpoint oftheconformityscale,therewassigniﬁcantvariationacrossthequestion‘‘Whatis thelikelihoodthatSallyexpresseshertrueopiniontothegroup?’’Onaverage acrossallconditions,33.5%(95%CI29.0–38.3)ofparticipantsreportedavalue of1or2,indicatingthattheythoughtitwasunlikelyorveryunlikelythatthe characterwouldstatehertrueopinionswhenaskedaboutherpoliticalbeliefs,as showninFig. 7binthe‘‘Appendix’’section.

Turningtothepotentialmediatingvariable,weexaminehowcomfortableparticipantsexpectedthecharactertobeinthesituation.Thevastmajorityofthe samplereportsthatthesehypotheticalscenarioswillmakethecharacteruncomfortable,asshowninFig. 7ainthe‘‘Appendix’’section.

Finally,thetwosocialconsequencevariablesarehighlyskewedintheopposite direction.OurparticipantsdidnotprojectthatSallyandhercolleagueswould socializeatdinneroutsideoftheofﬁce(over90%markedascoreof3orlower), butneitherdidtheythinkshewouldlookforanewjob(over95%markedascore of3orlower).

TreatmentEffects

Weﬁrstexaminewhetherthetreatmentconditionsaffectedparticipants’projection ofconformity.ThereissomesuggestiveevidencethatparticipantsinthePartisan Minority(PM)conditionwerelesslikelytoreportthatSallywouldreporthertrue opinionthanparticipantsintheBalancedParty(BP)condition(PM = 2.86, BP = 3.03, p \ .10).Thiseffectseemstobedrivenbyanincreaseintheproportion ofrespondentswhoreportonthelowendofthescale:anincreasedproportionof participantsinthePMconditionreportedthatSallywouldbeveryunlikelytoreport hertruefeelings(PM = 11.0%,BP = 5.4%, p \ .10),asshowninFig. 1.Weﬁnd noevidenceofatreatmenteffectwithintheprovocationmanipulation,norany supportforaninteractionbetweenthetwotreatments.

ThePartisanMinorityconditionevokedacorrectlysignedbutstatistically insignificantdifferenceintheanticipatedlevelofdiscomfortascomparedtothe BalancedPartycondition,asdidtheOpposeFavoredCandidateconditionoverthe SupportOpponentcondition.Wedonotfindanyevidenceinsupportofan interactioneffectofthetreatments.Mediationanalysisrequiresasignificant associationbetweentheindependentvariableandthetheorizedmediator(Baronand Kenny 1986).Intheabsenceofasignificantrelationshipbetweenthetreatmentand thetheorizedmediator,wedonotconductaformaltestformediation.

Thereisevidencethatthecontextualmanipulationaffectedrespondents’perceptionsofoneofsocialramiﬁcationsoftheinteraction:subjectsinthePartisanMinority conditionthinkSallyismorelikelytolookforanewjob(PM—2.24,BP—1.91, p \ .0001).Whilesubjectsthinkthisisanunlikelyoccurrence,anincreasedproportionof peopleinthePMconditionreportascoreof3orhigheronthe5pointscale(Partisan Minority—37.6%,BalancedPartisan—25.2%, p \.01).Therearenosigniﬁcant differencesbetweenthetwoconditionsoftheprovocationmanipulationforeitherof thesocialconsequencevariables,noranyinteractioneffectsbetweenthetreatments.

FurtherExploringtheConformityDecision

Thesuggestivebutweakrelationshipsbetweenthetreatmentconditionshypothesizedtobegeneratemorediscomfort(PartisanMinorityandOpposeFavored Candidate)andthereporteddiscomfortlevelsuggestthatasubject’sreportof anticipateddiscomfortislessaconsequenceofthetreatmentconditionheorshe receivedandmoreafactorofindividualcharacteristicsthatpredisposeanindividual

Likelihood that Sally feels uncomfortable answering this question?

Likelihood that Sally invites her coworkers over for dinner in the next six months?

Likelihood that Sally wants to look for a new job within the next year?

Mean Differences between Treatment Conditions

Likelihood that Sally expresses her true opinion to the group? -0.50-0.2500.250.50

Fig.1 Differenceinmeansbetweentreatmentconditionsontwodependentvariables.Participantsinthe PartisanMinorityconditionwereslightlymorelikelytothinkthatSallywouldconform

tobemoresensitivetotheconsequencesofdisagreement,acrosscontextsand provocations.Intheabsenceofasignificantrelationshipbetweenthetreatment conditionsandreporteddiscomfortlevel,movingoutsidetheexperimental frameworkandfocusingontherelationshipbetweenthediscomfortandconformity, wefindstrongevidencethatasubject’sreportofbeinguncomfortableis significantlyrelatedtotheanticipationofconformity.Wearelimitedinourability toassesswhichindividualdifferencesmayexplainreportedconformity,butweare abletodemonstratethattherelationshipbetweendiscomfortandconformitypersists whencontrollingfordemographicandpoliticallyrelevantvariables:thehigher someoneperceivesthecharacter’slevelofdiscomfort,themorelikelytheywereto reportthatshewouldnotsharehertruepoliticalopinions,asshowninTable 1. Finally,subjectswhoanticipatedthatthecharacterwouldconformwerealsomore likelytoreportthatthecharacterwouldlookforanewjobandmuchlesslikelyto reportthatshewouldinvitehercoworkersovertodinner(seeTable 6 inthe ‘‘Appendix’’section).Asweelaborateonintheconclusion,whilebeingina partisanminoritydoesappeartoinduceconformity,andfeelingsofdiscomfortare stronglyassociatedwithconformity,thenatureoftherelationshipsbetweenthese variablesismorecomplexthanastraightforwardmediation.

Table1 Explainingpredictionof‘‘expressingtrueopinion’’

Dependentvariable:likelihoodofexpressingtrueopinion

BasemodelDemographiccovariatesPoliticalcovariates (1)(2)(3)

Uncomfortable -0.687*** -0.696*** -0.722*** (0.102)(0.103)(0.107) Male -0.131 -0.173 (0.181)(0.189)

White0.0800.042 (0.198)(0.203) 2012engagement0.021 (0.064) Politicalknowledge -0.095 (0.125)

(0.107) Ideology -0.076 (0.058)

(0.159)

Intercept1|2 -5.32 -5.40 -5.51 (0.477)(0.515)(0.799)

Intercept2|3 -3.53 -3.59 -3.68 (0.438)(0.479)(0.776)

Intercept3|4 -1.92 -1.98 -2.05 (0.413)(0.454)(0.761)

Intercept4|50.200.150.25 (0.424)(0.461)(0.768)

Observations415412393

AIC1153.951148.241090.07

Residualdeviance1143.951134.241066.07

Tablesformattedwith stargazer Hlavac(2015) * p \ 0.1;** p \ 0.05;*** p \ 0.01

Study2 Design

Study2includedthreeparts:apretest,alabsession,andaposttest.Participants (n = 70)tookthepretestonlinethreedayspriortothelabsession.Thepretestwas embeddedinthepoliticalsciencemasstestingsurvey,sothemultitudeofquestions andthetimeseparatingthepretestandlabsessionshouldhavereducedparticipants’

abilitytobedeliberatelyinternallyconsistentbetweensurveys.Threedaysafterthe labsession,participantstookabriefposttestsurveyonline.

Pretest

ThepretestincludedfourteenquestionsadaptedfromtheAmericanNational ElectionStudiesaboutpoliticalissues,embeddedwithinalargesurvey.Participants wereaskedtoindicatetheextenttowhichtheyagreedwithavarietyofpolicies listedinTable 2.

LabSession

Uponinformingconsentinthelab,participantsenteredasmallconferenceroomto discusspoliticalissuesina‘‘focusgroup’’withtwoother‘‘participants,’’whowere actuallyconfederatesactingaspartofthestudy.Participantsweretoldthatthey wereparticipatinginafocusgroupaboutstudents’politicalopinionsoncampus. Participantsandconfederatestookturnssharingtheiropinionsonthefourteen politicalissueslistedinTable 2.Participantswererandomlyassignedtostatetheir responsesoutloudoneachissue before or after theconfederates,witheachissue presentedoneatatime.Thoserandomlyassignedtogivetheirresponses before the confederateswereinthe controlcondition,becausetheywouldbegivingtheir responsestoeachpoliticalquestionwithoutknowingtheopinionsofthe confederatesontheissueathand,thereforegivingtheparticipantslimited informationabouthowtoconformontheparticularissue.Thoserandomlyassigned togivetheirresponses last wereinthe treatmentcondition becausetheywouldonly givetheirresponseafterhearingthattheconfederatesdisagreedwiththemonan issue,givingthemapositionwithwhichtoconform.Participantssharedtheir opinionintherandomlyassignedorderforoneissueatatime.Asidefromtheorder inwhichparticipantswererandomlyassignedtogivetheirresponses,the procedureswerethesameacrossthetreatmentgroups.Inotherwords,both treatmentsinvolvedexposingparticipantstodifferingviewpointsfromtheirown, butvariedtheorderinwhichthatinformationwasdisclosed.Thetreatment, therefore,wasdeliberatelyverysubtleanddesignedtotestwhetherpeoplewould conformtoagroup’sopinionwhengiventheopportunitytodoso.

Allparticipantsinteractedwithconfederateswhodisagreedwiththemonmost issues.Basedontheirpretestresponses,theconfederatesweretoldtoplaythe ‘‘role’’ofeitherRepublicansorDemocratsandtofollowthecorrespondingscripts, asshowninTable 7 inthe‘‘Appendix’’section.Ifaparticipantidentifiedasa Democrat,theconfederateswerescriptedtobeRepublicans;ifaparticipant identifiedasaRepublican,theconfederateswerescriptedtobeDemocrats;ifa participantidentifiedasanIndependent,theconfederateswererandomlyassignedto beeitherRepublicansorDemocratsineachsession.Theconfederateswereblindto thepurposeofwhichscripttheyweretoldtouseandthepartyidentificationofthe participants.Eachsessionincludedtwoconfederates,balancedbyraceandgender, suchthateachlabsessionincludedonemaleconfederateandonefemale

Voting(Faux)Onascaleof1to10,with10being‘‘veryimportant’’and1being‘‘notimportant atall,’’howimportantdoyouthinkitistovoteinelections?

Egypt(Faux)SomepeoplethinkthattheUSshouldcontinuegivingmilitaryaidtoEgyptatthe samelevelasbeforethemilitarytakeover.Othersfeelthatweshouldsuspend allmilitaryaidtoEgypt.Othershaveopinionsinbetween.Onascaleof1to9, where1isthattheUSshouldcontinuegivingmilitaryaidtoEgypt,and9isthat weshouldsuspendallmilitaryaidtoEgypt,wherewouldyouplaceyourself?

EconomicsOverthepastyear,wouldyousaythattheeconomicpoliciesofthefederal governmenthavemadethenation’seconomybetter,worse,orhaven’tthey mademuchdifferenceeitherway?

EnergySomepeoplethinkthatthemostimportantpriorityforaddressingAmerica’s energysupplyshouldbeexpandingexplorationandproductionofoil,coal,and naturalgas.Doyouagreeordisagree?

Minimum

Doyoustronglyfavor,favor,oppose,orstronglyopposeanincreaseinthe minimumwagefrom$7.25to$9.00anhour?

Somepeoplebelievethatabortionshouldbepermittedonlyifthelifeandhealth ofthewomanisindanger.Doyouagreeordisagree?Doyoudosostrongly?

TaxesOnascaleof1to10,with10being‘‘verylikely’’and1being‘‘veryunlikely,’’ howlikelywouldyoubetovoteforacandidatewhosupportsraisingtaxeson thewealthyandloweringtaxesonthepoor?

limitsonpowerplantsinordertoaddressclimatechange?

IsolationismSomepeoplebelievethatthiscountrywouldbebetteroffifwejuststayedhome anddidnotconcernourselveswithproblemsinotherpartsoftheworld.Doyou agreeordisagree?

Onascaleof1to10,with10being‘‘verylikely’’and1being‘‘veryunlikely,’’ howlikelywouldyoubetovoteforacandidatewhosupportsthePatient ProtectionandAffordableCareAct,alsoknownasObamacare?

SomepeopleareafraidthegovernmentinWashingtonisgettingtoopowerfulfor thegoodofthecountryandtheindividualperson.Othersfeelthatthe governmentinWashingtonisnotgettingtoostrong.Whatisyourfeeling,do youthinkthegovernmentisgettingtoopowerfulordoyouthinkthe governmentisnotgettingtoostrong?

Onascaleof1to10,with10being‘‘verylikely’’and1being‘‘veryunlikely,’’ howlikelywouldyoubetovoteforacandidatewhosupportsdeferencetothe statesongaymarriage?

Onascaleof1to10,with10being‘‘verlikely’’and1being‘‘veryunlikely,’’ howlikelywouldyoubetovoteforacandidatewhosupportscuttingspending onprogramslikeMedicaidandMedicare?

Questionsaredisplayedinoneoftheordersinwhichthequestionswereaskedinthelabsession.The fauxquestionsweredeliberatelyplaced,butallotherswereinitiallyrandomized.Participantswerethen randomlyassignedtoeitherthequestionorderingshownabove,oradifferentquestionorderingcreated withthesameprocedure:Voting(faux),Egypt(faux),AffordableCareAct,Energy,Economics,Abortion (faux),Isolationism,GunControl(faux),MedicaidFunding,Taxes,Emissions,MarriageEquality, MinimumWage,GovernmentPower

confederate,oneofwhomwaswhiteandoneofwhomwasofaracialminority group.

Eachofthefourteenissueswaspresentedoneatatimeonascreenthatchanged tothenextissueautomaticallyafteroneminute.Participantsandconfederateswere instructedtostatetheiropiniononthequestiononthescreenanddiscussitifthey wanted.Becausethequestionspresentedonthescreenwerethesameasthepretest, participantsweresometimesaskedtoreportanumbertoindicatewheretheir opinionfellonascale.Tomaintainasmuchcontrolaspossible,confederateswere trainedtoavoiddiscussionandnottoprovidenewinformationifaskedby participants.Oneconfederatewas‘‘randomlyselected’’tobetherecorderforthe sessionandwasgivenaresponseformwiththediscussionquestionsandaspotto recordeachresponsefromtheparticipantandconfederates.Theresponseform, availableintheonlineseethe‘‘Appendix’’section,includedthesamequestion wordingandresponseoptionsasthepretest,withaspacetocircletheresponse givenbytheparticipantandeachconfederate.Intherareeventthatparticipants gavearesponsethatdidnotmatchtheoptionsontheresponseform,therecorder askedtheparticipanttoclarifyhisorherresponseintermsofthequestionresponse options.Discussionswiththeconfederatesafterthestudyindicatethatthiswasnota concernbecauseparticipantsgavetheirresponseswithintheframeworkofthe questionformat.

Inordertomakethesituationmorerealisticandconsistentwithprevious conformityresearch,theﬁrsttwoquestionswerestructuredslightlydifferently,with lesssocialpressureforconformity.Theconfederatesgaveneutralanswerstothe ﬁrsttwo‘‘faux’’questions,givingtheparticipantsnosignalstotheirpolitical leanings.Beginningonthethirdquestion,theconfederatesfollowedascript designedtodisagreewiththeparticipantsbasedonpretestresults.Overall,thelab sessionincludedten‘‘critical’’questionsonwhichtheconfederatesdisagreedwith theparticipantaccordingtothescript,andfour‘‘faux’’questionsdesignedtomake thestudymorerealistic,withconfederatesdisagreeingwitheachother,agreeing withtheparticipant,orprovidinganeutralresponse,asshowninTable 2 inthe textandTable 7 inthe‘‘Appendix’’section.Participantswererandomlyassignedto oneoftwoquestionorderings.Weheldthefauxquestionsinthesamepositionsand thenrandomizedtheorderofthetencriticalquestions.Participantswererandomly assignedtooneoftwoinitiallyrandomizedquestionorders,asexplainedinTable 2 inthe‘‘Appendix’’section.Aftercompletingallfourteenquestions,participants werethankedfortheirtimeandinstructedtoawaitafollowupsurveyinthecoming days.

Posttest

Threedaysafterthelabsession,participantswereemailedaposttestsurvey.The surveyincludedthesamefourteenquestionsthattheyansweredinthelargepretest surveyandinthelabsession.Thesequestionswereagainburiedwithinalarger survey,thoughnotaslargeasthepretest.Theprimarypurposeoftheposttestwasto examinethedistinctionbetweenpersuasionorattitudechangeandconformity.If participantsgavethesameresponsesonthepretestandposttest,butgaveadifferent

responseinthelabsession,thenwehavestrongevidencethatindividualswere indeedconforminginthelab.However,ifindividualsgavethesameresponseinthe labsessionandontheposttest,butthisresponsedifferedfromthepretest,thenthis couldbeevidenceofattitudechange.

Study2Hypotheses

Wehypothesizedthatparticipantsinthetreatmentconditionwouldconformata higherfrequencyandtoagreaterdegreethanparticipantsinthecontrolcondition. Basedonextantﬁndingsonconformityinsocialpsychology,weexpected participantstoconformtoagroup’spoliticalopinionwhentheyhadheardthe confederatesstateopinionswithwhichtheydisagreed.Inthecontrolcondition, participantswouldnotknowthepoliticalopinionsoftheconfederatesbeforestating theiropinions,sotheywouldhavelimitedinformationwithwhichtoconform.Itis possiblethatparticipantscouldintuitthattheconfederatesgenerallydisagreedwith theparticipantoverthecourseofthestudy,whichmeansthatwemightobserve somepreemptive‘‘conformity’’inthecontrolcondition.However,weexpecttosee agreaterfrequencyofconformityinthetreatmentcondition,whenparticipantsare certainofthegroup’sopinionspriortostatingtheirownopinion,comparedtothe controlconditionwheretheycanonlysurmisethegroup’sopinionsovertimeinthe study.BecauseStudy2isdesignedtobeadistinct,behavioraltestofwhether individualsactuallyconforminagroupsetting,asopposedtoreportingtheir expectationsofahypotheticalcharacter’sbehavior,theprimaryhypothesisisabout thedifferencebetweenthetreatmentandcontrolgroupsinthenumberoftimes participantsconformed.

Results

Atotalof70studentsparticipatedinthisstudy,butsevenwereremovedfromthe analysisbecauseoftreatmentadministrationerrorssuchasmissingconfederates, confederatesusingthewrongscript,andparticipantsknowingtheconfederates personally.Theremaining63participantswereincludedinmostanalyses.5 As showninTable 8 inthe‘‘Appendix’’section,thetreatmentgroupswerebalancedon mostkeycharacteristics.Participantsinthecontrolgroupweremarginallymore conservativethanparticipantsinthetreatmentgroup(p\ 10),andtherewere marginallymoreRepublicansthanDemocrats(p\ 10)inthecontrolgroup. Participantsinthetreatmentgroupalsoreportedpayingmarginallymoreattention topoliticsthanthoseinthecontrolgroup(p\:10).Therewerethirteenparticipants

5 Asdetailedlater,wemeasureconformityintwoways:potentialandpureconformity.Pureconformity requiresdatafromtheposttest;17participantsdidnotcompletetheposttest,sotheyarenotincludedin thepureconformityanalyses.Resultsforpotentialconformityholdwithandwithoutthese17 participants,butforstatisticalpowerpurposes,weincludethemintheanalysesforpotentialconformity. AsshowninTable 10 inthe‘‘Appendix’’section,participantswhodidnotcompletetheposttestdidnot meaningfullydifferfromthosewhodidcompletetheposttest,atleastbasedontheobservabledatawe haveavailable.

whoseposttestguessesofthetruepurposeofthestudywerefairlyaccurate.All resultsholdincludingandexcludingthoseparticipants.

SummaryStatistics

Ourprimarydependentvariableinthisanalysisisthenumberoftimesparticipants conformedacrossthetencriticalissuesduringthesession.Wemeasureconformity intwoways.First, potentialconformity meansthatinthelab,aparticipantgavean answerthatdifferedfromhisorherpretestresponse,movedinthedirectionofthe confederates,andcrossedthemidpointonthescale,suchthatthelabresponse actuallycounteredthepretestresponse.Forexample,ifonthepretestaparticipant indicatedthatheorshestronglyagreedwithsomething,butinthelabonlysaidthat heorsheagreed,thatwould not becodedaspotentialconformity.Ifthatparticipant saidthatheorshedisagreedorstronglydisagreedinthelab,thatwouldbe consideredpotentialconformity.Wecallthispotentialconformitybecausethe observedattitudeshiftinthelabhasthepotentialtobeconformity,butitcouldalso begenuineattitudechange.Second, pureconformity includestherequirementsof potentialconformity,inadditiontorequiringparticipantstogivethesameresponse onthepretestandtheposttest.Pureconformitythusﬁrmlydemonstratesaltering one’sopinion only inthepresenceofotherswhodisagree,whereaspotential conformityallowsforsomeﬂexibilityontheprivatepretestandposttestmeasures.

Notethatbothofourmeasuresofconformityrequiremovementacrossa midpointinthescale,amuchstricterrequirementthanpreviousstudiesexploring thepublicexpressionofopinions(LevitanandVerhulst 2015).Wedothisinorder todifferentiatetheconceptofconformityfromotherfactorsthatcouldinduce movementonaresponsescaleforanissuepositionbetweenapretestandalab session.Onquestionsutilizingaresponsescalewithmorethanfivepoints,some movementislikelytobeexpectedsimplybecauseofthelackofdistinctionina subject’smindonthescalepoints,forexamplea‘‘5’’andand‘‘6’’onaseven-point scale.Wecannotsaywithcertaintythatthismovementwouldrepresentconformity andisnotsimplyaformofresponseinstability.Bylimitingthemeasurementofour constructtoopinionsthatactually‘‘flipsides,’’wecanbemoreconfidentthat subjectsarepubliclyexpressinganopinionthatismeaningfullydifferentfromthe opiniontheyexpressedprivatelyonthepretest.

Figure 2 showsthedistributionofthefrequencyofconformityinbothpotential (Fig. 2a)andpure(Fig. 2b)measuresofconformityforallparticipantsandeach treatmentcondition.Thethickerlinesonthebottomoftheﬁguresshowtheraw distributionofparticipants,showingthatmostparticipantsconformedonceortwice andnooneconformedmorethanfourtimes.Moreimportantly,thethinlineson Fig. 2 showthecumulativedistribution.Theselinesillustratethatalthough participantsdidnotconformoften,88.9%ofparticipantsconformedonatleastone questionbypotentialconformitymeasures(94.1%inthetreatmentgroupand 82.8%inthecontrolgroup),and58.7%ofparticipantsconformedatleastonceby pureconformitymeasures(65.2%inthetreatmentgroupand52.2%inthecontrol group).ThesedistributionsarestrikinglysimilartothosefoundintheAsch(1956) experiments.

Frequency of Potential Conformity

Participants

Frequency of Pure Conformity

Total Number of Times Participant Conformed

(a) PotentialConformity

Total Number of Times Participant Conformed

(b) PureConformity

Fig.2 The thicklines slopingdownwardinthesefiguresreflectthedistributionofthefrequencyof conformityforallparticipantsandeachtreatmentgroup.The thinlines slopingupwardreflectthe cumulative distributionofthefrequencyofconformityforallparticipantsandeachtreatmentgroup. a showsthedistributionsusingthemeasureofpotentialconformityand b showsthedistributionsusing themeasureofpureconformity

Asamanipulationcheck,weuseda t testtoinvestigatedifferencesintheaverage frequencyofopinionchangebetweenthepretestandlabsessionbetweentheten criticalquestionsandthefourfauxquestions.Therewassignificantlymoreaverage changeinreportedopinioninthecriticalquestionsthanthefauxquestions (p\:001).Thismeansthattherewassignificantlymorechangeinreportedopinion onquestionsinwhichbothconfederatesintentionallydisagreedwiththeparticipant. Therewerenosignificantdifferencesbetweenthetworandomizedquestionorders, norweretheresignificantdifferencesbasedontheconfederateswithwhomthe participantsinteracted.

TreatmentMainEffects

Wehypothesizedthatparticipantswhostatedtheirresponseslast(treatment)would conformathigherlevelsthanparticipantswhostatedtheirresponsesfirst(control). Wetestthishypothesisusingstandard t testsandrandomizationinference. Randomizationinferenceisgainingtractioninpoliticalscience,especiallyin experimentalwork(Blattman 2015;Crabtreeetal. 2015;GerberandGreen 2012; Young 2016).First,usinga t testtoexaminetheeffectofthetreatmenton conformity,wefoundthatparticipantsinthetreatmentconditionconformed significantlymorefrequentlythanparticipantsinthecontrolconditionforpotential conformity(p\:01),asshowninFig. 3.Participantsinthetreatmentcondition conformedmorefrequentlythaninthecontrolconditionbypureconformity standardsaswell,butthisdifferenceisnotstatisticallysignificantbystandard thresholds(p ¼ :105).6 Itispossiblethatwearestatisticallyunderpoweredtodetect

6 Previousversionsofthismanuscriptreportedthepureconformityresultsasbeingsigniﬁcantatthe.05 level,butuponpreparingthereplicationdataandcodeinaccordancewith PoliticalBehavior’sdata

Fig.3 Differenceinthenumberoftimesparticipantsconformedbythepotential(a)andpure (b)conformityparametersineachtreatmentgroup. Lines represent95%conﬁdenceintervals

asignificantdifferencebetweenthetreatmentgroupsbypureconformitystandards. Becausepureconformityismeasuredbasedonposttestresults,onlythose participantswhocompletedtheposttestcanbeincludedintheanalysis,which reducesoursampleto46participantsforthepureconformitytests.AsFig. 3 illustrates,participantsconformedinbothconditions,butthefrequencyof conformitywassignificantlyhigherinthetreatmentgroupforpotentialconformity. Althoughitispossiblethatparticipantsinthecontrolconditionwereabletoguess thegroup’sopinionoverthecourseofthestudy,wefindthatparticipantswereno morelikelytoconformatthebeginningofthestudythanattheend,makingthis lesslikely.

Second,weuserandomizationinferenceteststoexaminethetreatmenteffecton potentialandpureconformity.Ineffect,weran10,000simulationsofour experiment,usingtheobservedvaluesoftheconformitymeasures,butrandomly reassigningthetreatmentcondition,andcalculatingthedifferenceofmeans betweentheshufﬂedtreatmentgroups.Wethencomparedtheseresultstothe observedmeandifferences,usingthetruetreatmentgroupassignments.The ﬁguresthuscompareourresultstothedistributionof10,000simulateddifferenceof meanstests.AsshowninFig. 4,theobserveddifferenceofmeansbetweenthetrue treatmentandcontrolgroupsisdistinctfromthevastmajorityoftherandomly generateddifferencesofmeansfromrandomlygeneratedtreatmentgroups.

Speciﬁcally,only0.10%ofthepermuteddifferencesweregreaterthanthe observeddifferenceforpotentialconformity,aswere3.02%ofthepermuted differencesforthepureconformitymeasure.Bothoftheseresultsprovideevidence thatindividualsinthetreatmentgroupconformed,bypotentialandpurestandards, morefrequentlythanthoseinthecontrolgroup.

Footnote6continued availabilityandreplicationpolicy,wediscoveredacodingerror.Theﬁndingspresentedinthepaper reﬂecttheresultsbasedonthecorrectedcode.

Potential Conformity by Condition

Conformity by Condition

Difference of Means (Treatment-Control)

(a) PotentialConformity

Histogram of Permutation Test Results

Difference of Means (Treatment-Control)

(b) PureConformity

Fig.4 Distributionof10,000permuteddifferencesofmeans.The verticalredline indicatestheobserved differencebetweenthetreatmentandcontrolgroupsforpotential(a)andpure(b)conformity,usingthe trueconditionlabels.Resultsindicatethat0.10%ofthepermuteddifferencesofmeansforpotential conformityaregreaterthantheobserveddifference,asare3.02%ofthepermuteddifferencesofmeans forthemeasureofpureconformity(Colorﬁgureonline)

Whiletheseresultsarecompellinggiventhatparticipantswererandomly assignedtothetreatmentconditions,wepushtheresultsfurtherbytestingthe effectsofthetreatmentonconformityinaregressionframework.Asshownin Tables 3 and 4,controllingforahostofcharacteristicsthatmightinfluence conformity,thequestionorderingtheparticipantreceived,andwhetherthe participantsurmisedthetruepurposeofthestudy,thetreatment still significantly affectedthefrequencywithwhichparticipantsconformedbypotentialconformity measures.Thetreatmenteffectsarenotstatisticallysignificantforpureconformity, butthiscouldbeaconsequenceofoursmallsamplesizeandlimitedstatistical power.Thetreatmentdoesnotappeartobeconditionalonanyofthedemographics wehaveincludedinourmodels,howeveroursmallsamplesizelimitsourabilityto fullyexplorethis.Thepatternofresultsshowninthisordinaryleastsquares frameworkisconsistentwithpoissonmodelsavailableinthe‘‘Appendix’’section.A descriptionofthecontrolvariablemeasuresisalsoavailableinthe‘‘Appendix’’ section.

ExplainingtheConformityDecision

BasedontheresultsfromStudy1indicatingthatparticipantsthoughtthata hypotheticalcharacterwouldbeuncomfortableinthepoliticallycontentious discussion,weincludedsomeself-reportmeasuresonthepost-testofStudy2to assesstheemotionalexperienceindividualshavewhileactuallyengagingin politicallycontentiousdiscussions.Weaskedparticipantstoreﬂectontheemotions theyexperiencedduringthelabsession.AsshowninFig. 5,veryfewparticipants

Histogram of Permutation Test Results

Conformity (10,000 Simulations)

Table3 Study2regressionmodels:potentialconformity

Dependentvariable

Potentialconformity

BasemodelStudycontrolsDemographic controls Political controls

Condition0.856***0.822***0.873***0.680*** (0.276)(0.268)(0.269)(0.310)

Knewpurpose -0.650* -0.677** -0.818** (0.331)(0.336)(0.368)

OrderA0.3850.475*0.411 (0.268)(0.272)(0.303)

Female0.2180.145 (0.274)(0.309)

White -0.210 -.145 (0.311)(0.350) 2012engagement0.061 (0.098)

Politicalknowledge -0.077 (0.173)

Politicalinterest -0.051 (0.278)

Partisanattachment0.026 (0.024)

Ideology 0.021 (0.098)

Constant1.379***1.324***1.312***1.203 (0.203)(0.262)(0.380)(1.055) Observations63636261 R2

Tablesformattedwith stargazer Hlavac(2015) * p \ 0.1;** p \ 0.05;*** p \ 0.01

reportedfeelinghappyorexcitedandthemostdominantemotionsweresurprise, frustration,anxiety,andconfusion.Veryfewpeoplereportedfeelingscaredor angry,butthepointstillremainsthatnegativeemotionsweremuchmoreprevalent thanpositiveemotionswhileinteractingwithpeoplewhodisagree.Thistrendis consistentacrossbothtreatmentgroups,whichsuggeststhatengagingwithpeople

Table4 Study2Regressionmodels:pureconformity

Dependentvariable

Pureconformity

BasemodelStudycontrolsDemographic controls Politicalcontrols

Condition0.3910.3710.379 -0.069 (0.236)(0.239)(0.239)(0.235)

Knewpurpose -0.330 -0.389 -0.396 (0.273)(0.274)(0.251)

OrderA -0.099 -0.143 -0.270 (0.240)(0.243)(0.218)

Female0.2670.027 (0.243)(0.229) White0.3350.556** (0.283)(0.267)

(0.069) Political knowledge -0.132 (0.129) Politicalinterest -0.140 (0.192) Partisan attachment 0.037** (0.016)

Ideology -0.226*** (0.068)

Constant0.652***0.800***0.4171.514** (0.167)(0.225)(0.328)(0.720) Observations46464645

R2 0.0590.0920.1460.472

AdjustedR2 0.0370.0270.0390.316

ResidualSE0.800(df=44)0.804(df=42)0.799(df=40)0.674(df=34)

Fstatistic2.750(df=1;44)1.418(df=3;42)1.368(df=5;40)3.036***(df=10;34)

Tablesformattedwith stargazer Hlavac(2015) * p \ 0.1;** p \ 0.05;*** p \ 0.01

withwhomonedisagreespoliticallyisagenerallymorenegativeexperience, regardlessofpressurestoconform.Ofcourse,theseemotionscouldbeconnectedto theexperienceofsimplyparticipatinginastudyordiscussingpoliticsgenerally, althoughwenotethatoursubjectswerepredominantlypoliticalsciencemajorswho aregenerallyveryinterestedinpolitics.

Proportion of Participants

Wealsoinvestigatedparticipants’previousexperienceswithpressuretohold particularpoliticalopinionsthroughself-reportmethods.Oftheparticipantswho answeredthequestion,63%reportedhavingfeltpressuredtoholdaparticular politicalopinionintheirdailylives.Therewasnotasigniﬁcantdifferenceinthe numberofstudentsreportingpriorfeelingsofpoliticalpressurebetweenthetwo treatmentgroups.Ofthoseparticipantsindicatingfeelingthispressure,the overwhelmingmajorityidentiﬁedfriends,family,andclassmatesasthesourceof thepressure.AsshowninFig. 6,31%oftheseparticipantsreportedfeeling pressurefromtheirteachersand13.8%reportedfeelingpressurefromtheirbosses. SurprisedAnxiousFrustratedConfusedHappyExcitedAngryScared Emotional

Fig.6 Oftheparticipantswhoreportedfeelingpressuretoholdaparticularpoliticalopinion,thisshows theproportionofparticipantsreportingfeelingthispressurefromeachsource

Fig.5 Proportionofparticipantsreportingexperiencingeachemotionduringthelabsession

Thesedescriptivestatisticsprovideevidencefortheprevalenceofthepressureto haveaparticularpoliticalopinionandfromwherethatpressurestems.

Conclusion

Democracyinherentlydependsupondiversepoliticalopinions.Huckfeldtetal.(2004) boldlyarguethat‘‘politicaldisagreementandheterogeneityconstitutethelifebloodof democraticpolitics’’(24).Thisnecessaryconditionfordemocracyisdiminishedina societywherepeoplefacepsychologicalbarrierstoparticipatingandexchangingideas freely.Ourtwostudiesshowthatpeoplebothexpectotherstohidetheirtruepolitical opinionsandactuallydosothemselves:whileapproximately33%ofrespondents expectacharactertoconforminavignette,approximatelytwo-thirdsofrespondents conformtheiropinionsinactualdiscussions.Wenotethatthesituationsinwhichthe hypotheticalcharacterorparticipantfoundhimorherselfwereneitherthreateningnor excessivelycontentious.Thetreatmentsinourexperimentssimplycreatedasituation inwhichtheparticipanthadreasontoexpectthatothershelddifferentpolitical viewpoints.Anypressureparticipantsfelttoadheretoamajorityopinionwerederived internally,notfromanyexplicitrepercussionsforfailingtoagree.

Weacknowledgethattherearesomelimitationswiththisstudy.Oursamplesizes weresmall,whichcouldmeanthatsomeofouranalysesareunderpowered.Our samplealsoconsistedofcollegestudents,whichisapopulationparticularly susceptibletoconformity(Sears 1986).However,oursamplewasdrawnfrom politicalsciencecourses,meaningthattheparticipantsprobablyhaveagreater interestinpoliticsandprobablyhavemorepoliticalinteractionsthanaverage Americans,whichshouldmakeithardertoﬁndaneffect.

OurmeasurementofconformityinStudy2wasparticularlyconservative.Simply movementinthedirectionofthe confederatesisnotenoughtobeconsideredconformityby eitherofourmeasures.Participantshadtoeffectivelycrossthemidpointonagivenresponse scalefortheirbehaviortobeconsideredconformity.Notonlydoesthisoperationalization fullycapturetheconceptofconformity—suchthatindividualsallowotherstoassume agreementwhenthereisnochangeinopinion—butitalsoreducesnoiseinthemeasure;itis unlikelythatparticipantswouldundergotrueattitudechangeoverthecourseofthestudy becausetheyweresurveyedonlythreedaysbeforeandthreedaysafterthelabsession.Our measuresofconformityaremoreconservativethantheonlyotherlabexperimentalevidence ofpoliticalconformity(LevitanandVerhulst 2015),whichonlyrequiresparticipantsto moveinthedirectionoftheconfederates.Ourmeasureofconformityexcludessimple changesinthemagnitudeofagreementordisagreement,asachangefrom‘‘stronglyagree’’ to‘‘agree’’isnotameaningfulexpressionofconformity.Werequireindividualstoshiftfrom someformofagreementtosomeformofdisagreement(orviceversa),whichweargueisa cleaner—andmoreconservative—measureofconformity.

Thedistinctionbetweenpotentialandpureconformityisalsoimportant.Forboth measures,wefindthatparticipantsinthetreatmentcondition,whogavetheirresponses last,conformedmorefrequentlythanparticipantsinthecontrolconditionwhogavetheir responsesfirst.However,thisdifferenceisonlystatisticallysignificantbystandard thresholdsforpotentialconformity.Wesuspectthatthelackofstatisticalsignificance

forpureconformityisafunctionofthesubtletyofourtreatment,theconservativenature ofthemeasurement,andoursmallsamplesize.Therandomizationinferenceresults bolsteroursuggestionthatindividualsinthetreatmentconditionconformedmore frequentlythanthoseinthecontrolconditionforbothpotentialandpureconformity. Furthermore,individualdifferencesarelikelyinﬂuentialintheunderlyingpropensityto conform.Potentialconformitydidnotrequireparticipantstogivethesameresponseson thepretestandposttest.Thismeansthatthetreatmentcouldhaveactuallyledtoattitude changeinathreedayfollowup.LevitanandVerhulst(2015)ﬁndevidenceforpersistent attitudechangeafterasimilarlabexperience,sothiscouldbeaplausibleexplanation.It alsocouldbetheresultofresponsebiasasparticipantsmighthaveremembered conforminginthelabanddecidedtogivethesameresponseasthelabsession.

Bothstudiesemployedtreatmentgroupswhereeachconditionhadthepossibilityof creatingdiscomfort:intheﬁrststudy,wecomparedtwotypesofenvironmentsthatare thoughttobestressful—partisancompetitionversuspartisanminoritystatus—andin thesecondstudy,thesubjectwasinanopinionminorityinbothconditions,justmore obviouslysointhetreatmentcondition.Thus,itwasintentionallydifﬁculttodetect differencesbetweentreatmentgroups.ThepressureinStudy2wasmuchlessexplicit thanthatofStudy1,butthefactthattherewereresultssuggestingthatconformityisan expectedandobservedoutcomeinbothconditionsinbothstudiesstrengthensour argumentthatconformityisprevalent.Furthermore,thereisevidenceinStudy1that individualleveldifferenceslikelymatterasmuchasthecontextinexplainingthe extenttowhichapersonconforms;whilethePartisanMinorityconditiondidincrease thereportedexpectationofconformity,theextenttowhichasubjectanticipatedthe charactertobeuncomfortable,regardlessofthecontextortheprovocation,was stronglyrelatedtoanticipatedconformity.Explorationofthecharacteristicsthatmake individualsuncomfortableinpoliticalinteractions,especiallywhentheyareinan opinionminorityposition,willbeafruitfulstepforwardinfutureresearch.

Ourﬁndingsarenotatoddswithotherworksuggestingthatpoliticalconformity andhomogeneityare not prevalent(Huckfeldtetal. 2004).Recallthatthisbodyof researchconceptualizesconformityasactualattitudechange.We,incontrast, considerconformitytobeadefensemechanismwherebyindividuals publicly expresspoliticalviewsthatdifferfromtheirprivatebeliefs,givingothersthe impressionthateveryoneisinagreement.Ourresultshavenobearingonthedebate overtheextentofprivateorpublicpoliticaldisagreementinsocialnetworks.But, ourtheoryactuallydependsonindividualsbeingexposedtopoliticaldisagreement, whichsituatesourstudynicelywithinthepoliticaldiscussionnetworkliterature.

Althoughparticipantsdidnotengageinafulldeliberationinthelabexperimentin Study2,norwasthereafulldiscussionpreviewedinthevignetteexperimentinStudy 1,theresultsofthesestudieshaveimplicationsforthedeliberativedemocracy literature.Theevidencepresentedinthispaperthatindividualsinpoliticalopinion minoritiesarelesscomfortableandlesslikelytoexpresstheirtruepoliticalopinionsto thegroupindicatesthatgroupdeliberationsmightnotfullyreﬂectthenatureofthe opinionsofthegroup.Karpowitzetal.(2012)andKarpowitzandMendelberg(2007) discusstheimportanceofrepresentationofminorityvoicesindeliberations,butthis doesnotfullycapturetheextenttowhichvoicesmightbesilencedduetopressuresto

conform.Futureworkshouldtrytointegratethesetheoriestogethertoexaminethe impactthatconformitypressureshaveondeliberation.

Wehaveidentifiedmanyavenuesforfutureresearchonpoliticalconformity.We believethatthisinitial,andconservative,evidencepositionsresearcherswelltomove forwardtobetterunderstandthemechanismsandimplicationsofpoliticalconformity. Wesuspectthatindividualdifferencesinpersonality,socialanxiety,andconflict avoidanceinfluencethepropensitytoconformandthesedifferencesshouldbe exploredinmoredetailandinsampleswithmorediversityonthesedimensionsthan thoseusedinthesestudies.Similarly,thereismuchworktobedoneexploring variationinconformityacrossissuetypes.Ourstudiesherearenotwell-suitedto explorethisbecausequestionshaddifferentresponsescales,makingitdifficultto compareacrossissues.Futureworkshouldconsidervariationinconformitybetween socialandeconomicissues,and‘‘easy’’and‘‘hard’’issues.Therearemany opportunitiestoextendthisworktobetterunderstandotherpoliticalphenomena.

Ourresultssuggestthatindividualsarelesscomfortableinthepoliticalminoritythan inanenvironmentofopiniondiversity.Whilethereisdisagreementabouttheextentto whichgeographyislinkedtopolarization,oneoftheconsequencesofsortingbasedon politicalopinionisthatwhilemorepeoplewillﬁndthemselvessurroundedbylike mindedothers,increasingnumbersofpeoplewillalsoﬁndthemselvesinsettingswhere theirviewpointsareintheminority.Thus,ourresultssuggestthatitispossiblethatone oftheconsequencesofthecontemporary,polarizedpoliticallandscapeisincreased politicalconformity,asopinionminoritiesconformtotheopinionsofthosearound them.Whiletheimplicationsofpoliticalconformityarespeculative,weprovidestrong evidencethatitisapotentialbehavioralconsequenceofbeinginapoliticalopinion minority.Weseektoexplorethemechanismsmotivatingthisbehaviorinfuturework andtousethisknowledgetotheorizewaystoreducepoliticalconformity.Ultimately, individualsaltertheirpubliclystatedpoliticalviewsinthepresenceofopposingothers, behavinglikepoliticalchameleons,temporarilyabandoningtheirtruepoliticalopinions toconformtootherswhodisagree.

Acknowledgments TheauthorsthanktheWilliam&MaryOmnibusProjectforfacilitatingparticipant recruitmentandtheSocialNetworksandPoliticalPsychology(SNaPP)Labforprovidingboththe infrastructureandresearchassistantteamthatmadethisstudypossible.Theauthorsarealsogratefulfor supportfromtheNationalScienceFoundation(grantSES-1423788),aswellastheCharlesCenterat William&MaryforprovidinghonorsfellowshipfundingforStudy2.Finally,theauthorsthankthe anonymousreviewerswhosehelpfulcommentsgreatlyimprovedthispaper.

CompliancewithEthicalStandards

EthicalApproval Allproceduresperformedinstudiesinvolvinghumanparticipantswereinaccordancewiththeethicalstandardsoftheinstitutionaland/ornationalresearchcommitteeandwiththe1964 Helsinkideclarationanditslateramendmentsorcomparableethicalstandards.Thisarticledoesnot containanystudieswithanimalsperformedbyanyoftheauthors.

InformedConsent Informedconsentwasobtainedfromallindividualparticipantsincludedinthestudy.

Appendix

SeeFig. 7,Tables 5, 6, 7, 8, 9,and 10

Fig.7 DistributionofStudy1dependentvariables. a showsthedistributionofresponsestothequestion ‘‘WhatisthelikelihoodthatSallyfeelsuncomfortableansweringthisquestion?’’ b showsthedistribution ofresponsestothequestion‘‘WhatisthelikelihoodthatSallyexpresseshertrueopiniontothegroup?’’

Table5 Distributionofkeydemographicvariables(Study1) Fall(lab)Spring(online)

Table6 Meanlevelsofdiscomfortandsocialconsequences,byreportedconformity(Study1)

NotconformConform p value

Reporteddiscomfort3.8624.4460.000 Lookfornewjob2.0142.1930.050 Invitecolleaguesfordinner2.6232.4290.025

P valueisfromadifferenceofmeanstest

Table7 Study2script

QuestionTreatmentControl

(R)participant(D)participant(R)participant(D)participant

Onascaleof1to10,with10being ‘‘veryimportant’’and1being‘‘not importantatall,’’howimportantdo youthinkitistovoteinelections?

Onascaleof1to9,where1isthat theUSshouldcontinuegiving militaryaidtoEgypt,and9isthat weshouldsuspendallmilitaryaid toEgypt,wherewouldyouplace yourself?

Overthepastyear,wouldyousay thattheeconomicpoliciesofthe federalgovernmenthavemadethe nation’seconomybetter,worse,or haven’ttheymademuchdifference eitherway?

Somepeoplethinkthatthemost importantpriorityforaddressing America’senergysupplyshouldbe expandingexplorationand productionofoil,coal,andnatural gas.Doyouagreeordisagree?

Doyoustronglyfavor,favor,oppose, orstronglyopposeanincreasein theminimumwagefrom$7.25to $9.00anhour?

Confederate1: 6 Confederate1: 6 Participant:Participant:

Confederate2: 8 Confederate2: 8 Confederate1: 6 Confederate1: 6

Participant:ParticipantConfederate2: 8 Confederate2: 8

Confederate1: 5

5 Participant:Participant: Confederate2: 6

6 Confederate1: 5 Confederate1: 5

Participant:Participant:Confederate2: 6 Confederate2: 6

Confederate1: Better Confederate1: Worse Participant:Participant:

Confederate2: Better Confederate2: Worse Confederate1: Better Confederate1: Worse

Participant:ParticipantConfederate2: Better Confederate2: Worse

Confederate1: Disagree

Confederate2: Disagree

Confederate1: Agree Participant:Participant:

Confederate2: Agree

Confederate1: Disagree Confederate1: Agree

Participant:Participant:Confederate2: Disagree Confederate2: Agree

Confederate1: Strongly favor

Confederate2: Favor

Somepeoplebelievethatabortion shouldbepermittedonlyifthelife andhealthofthewomanisin danger.Doyouagreeordisagree? Doyoudosostrongly?

Confederate1: Strongly oppose Participant:Participant:

Confederate2: Oppose Confederate1: Strongly favor Confederate1: Strongly oppose

Participant:Participant:Confederate2: Favor Confederate2: Oppose

Confederate1: Disagree

Confederate2: Agree

Confederate1: Disagree Participant:Participant:

Confederate2: Agree

Confederate1: Disagree Confederate1: Disagree

Participant:Participant:Confederate2: Agree Confederate2: Agree

continued

QuestionTreatmentControl

(R)participant(D)participant(R)participant(D)participant

Onascaleof1to10,with10being ‘‘verylikely’’and1being‘‘very unlikely,’’howlikelywouldyoube tovoteforacandidatewho supportsraisingtaxesonthe wealthyandloweringtaxesonthe poor?

Doyoufavororopposemaking privategunsalesandsalesatgun showssubjecttobackground checks?

Doyoustronglyfavor,favor,oppose, orstronglyopposesettingstricter emissionlimitsonpowerplantsin ordertoaddressclimatechange?

Confederate1: 7 Confederate1: 3 Participant:Participant:

Confederate2: 8

Confederate2: 2 Confederate1: 7 Confederate1: 3

Participant:Participant:Confederate2: 8 Confederate2: 2

Confederate1: Favor Confederate1: Favor Participant:Participant:

Confederate2: Favor Confederate2: Favor Confederate1: Favor Confederate1: Favor

Participant:Participant:Confederate2: Favor Confederate2: Favor

Confederate1: Favor Confederate1: Oppose Participant:Participant:

Confederate2: Strongly favor Confederate2: Strongly oppose Confederate1: Favor Confederate1: Oppose

Participant:Participant:Confederate2: Strongly favor Confederate2: Strongly oppose

Somepeoplebelievethatthiscountry wouldbebetteroffifwejust stayedhomeanddidnotconcern ourselveswithproblemsinother partsoftheworld.Doyouagreeor disagree?

Onascaleof1to10,with10being ‘‘verylikely’’and1being‘‘very unlikely,’’howlikelywouldyoube tovoteforacandidatewho supportsthePatientProtectionand AffordableCareAct,alsoknown asObamacare?

Whatisyourfeeling,doyouthink thegovernmentisgettingtoo powerfulordoyouthinkthe governmentisnotgettingtoo strong?

Confederate1: Disagree Confederate1: Agree Participant:Participant:

Confederate2: Disagree Confederate2: Agree Confederate1: Disagree Confederate1: Agree

Participant:Participant:Confederate2: Disagree Confederate2: Agree

Confederate1: 8 Confederate1: 2 Participant:Participant:

Confederate2: 7 Confederate2: 3 Confederate1: 8 Confederate1: 2

Participant:Participant:Confederate2: 7 Confederate2: 3

Confederate1: Notgetting toostrong Confederate1: Gettingtoo powerful Participant:Participant:

Confederate2: Notgetting toostrong Confederate2: Gettingtoo powerful Confederate1: Notgetting toostrong Confederate1: Gettingtoo powerful

Participant:Participant:Confederate2: Notgetting toostrong Confederate2: Gettingtoo powerful

Table7 continued

QuestionTreatmentControl

(R)participant(D)participant(R)participant(D)participant

Onascaleof1to10,with10being ‘‘verylikely’’and1being‘‘very unlikely,’’howlikelywouldyoube tovoteforacandidatewho supportsdeferencetothestateson gaymarriage?

Onascaleof1to10,with10being ‘‘verlikely’’and1being‘‘very unlikely,’’howlikelywouldyoube tovoteforacandidatewho supportscuttingspendingon programslikeMedicaidand Medicare?

Confederate1: 3 Confederate1: 7 Participant:Participant:

Confederate2: 2 Confederate2: 8 Confederate1: 3 Confederate1: 7

Participant:Participant:Confederate2: 2 Confederate2: 8

Confederate1: 2 Confederate1: 8 Participant:Participant: Confederate2: 3

7 Confederate1: 2 Confederate1: 8

Participant:Participant:Confederate2: 3 Confederate2: 7

Thistableshowsthescriptusedbytheconfederates.Thetreatmentconditioncolumnsshowtheparticipantrespondinglasttoeachquestionandthecontrolconditioncolumnsshowtheparticipantgoing ﬁrst.Withineachtreatmentcondition,youcanseetheresponsestheconfederatesgave,dependingonthe participant’spartisanship

Table8 Study2balance table—allparticipants

Democrat,Female,andWhite reﬂectproportionsineach treatmentcondition.The remainingvariablesreﬂectthe meansineachtreatment condition.Tableincludes participantswhodidnot completetheposttest

Table9 Study2balance tableforthosewhocompleted posttest

Democrat,Female,andWhite reﬂectproportionsineach treatmentcondition.The remainingvariablesreﬂectthe meansineachtreatment condition.Tableonlyincludes participantswhocompletedthe posttest. ControlTreatment p value

Table10 Study2Summary Statistics

WholeSample

Tablesformattedwith stargazer Hlavac(2015)

Completedposttest

Didnotcompleteposttest

PilotStudy1:StressfulDimensionsofthePoliticalSphere

Inthefallof2010,aspartofasetofstudiesrunonasampleof280undergraduates atalargepublicuniversityintheWest,subjectswereaskedabouttheiranticipated emotionalresponsetoasetof13diversestimuliconsistingofavarietyofpolitical situationsthatpeoplecouldencounterintheirpoliticalenvironment,especiallythe environmentofacompetitiveorsalientelection.Thegoalofthispilotstudywasto characterizethepoliticalenvironment,differentiatingwhataspectsoftheroutine encountersapersonhasarelikelytoprovokeemotion,andwhetherdifferent emotionsareprovokedbydifferentscenarios.Respondentswerepresentedwith theseinstructions:

Howdoyoufeelaboutpolitics?Placea0inthecorrespondingcellinthe tablebelowifthepoliticalsituationdoesnotelicitthestatedemotion.Ifthe situationdoeselicitthatemotion,placeanumberinthecellthatcorresponds tothestrengthofyouremotionalreaction,from1(weak)to5(strong).A politicalsituationmayevokemorethanoneemotion.

Thefollowingstimuliwereplacedinatablewithfourothercolumnslabeled ‘‘Anxious’’,‘‘Angry’’,‘‘Enthusiastic’’,and‘‘Don’tKnow.’’

• Livinginacommunitywheremostofyourneighborsafﬁliatewithapolitical partyyoudon’tsupport

• Seeingbumperstickersoryardsignsinyourneighborhoodforcandidatesor partiesyoudon’tsupport

• Talkingwithyourneighborsorfriendsaboutpoliticswhenyouagreeonmost things

• Talkingwithyourneighborsorfriendsaboutpoliticswhenyoudisagreeonmost things

• Beingtheonlypersoninyourgroupoffriendswhosupportsacandidate,aparty, orapoliticalissue

• Readingapollpredictingtheopposition’scandidateislikelytowinan importantrace

• SeeingpoliticalprotestsinsomeothercitydepictedonTV

• Seeinglivepoliticalprotestsinyourarea

• Watchingapoliticaldebateontelevision

• Receivingapoliticalemailforwardwithwhichyoudisagree

• Receivingapoliticalemailforwardwithwhichyouagree

• Readingafriend’spostinyourFacebooknewsfeedthatexpressespolitical viewswithwhichyoudisagree

• Readingafriend’spostinyourFacebooknewsfeedthatexpressespolitical viewswithwhichyouagree

Theorderofthestimuliwasrandomizedacrossrespondents(Fig. 8).

Fig.8 Proportionofparticipantsreportingthattheywouldexperienceanxiety(marked [3)oneachitem inthepilotstudy

PilotStudy2:FreeResponseAnswersaboutPoliticalStress

Toexploreourhypotheses,wetookadvantageofAmazon’sMechanicalTurk platform.MechanicalTurkisanonlineenvironmentwhereindividualscanhire otherstoaccomplishtasksinreturnformonetarycompensation(seeBerinskyetal. 2012 foramorecompletediscussion).Thesetaskscanbecompletedbyanyonewith accesstoMechanicalTurk,inotherwords,anyonewithacomputerandaninternet connection.SomepoliticalscientistshavevoicedconcernsaboutusingtheInternet populationforresearchgivencharacteristicsuniquetoitsmembers.Forexample, AnsolabehereandSchaffner(0000)arguethattheInternetpopulationissomewhat moreknowledgeablethantheoff-linepopulation.However,someevidencesuggests thismayresultfromrespondentssupplementingwhattheyknowbyusingGoogleor otherInternetsourcesBurnett(2012)(seealsofootnote23inBerinskyetal. 2012). Likewise,whiledisproportionatenumbersofgroupssuchasthedisabled,elderly,

poor,andminoritiesremainoff-line,increasingInternetpenetrationhasmadethis coveragebiascritiquelessconsequential(AnsolabehereandSchaffner 0000).

Recently,politicalscientistshavebegunusingMechanicalTurktorecruit subjectsforcomputer-basedexperiments.Berinskyetal.(2012)examinedthe validityofexperimentsusingtheMechanicalTurkplatform,findingthatitoften providesmorerepresentativesamplesthanthetypicalstudentandconvenience samplesdrawnforexperimentalresearch.Moreover,theydeterminedthatthreatsto validityincludingheterogeneoustreatmenteffects,subjectattentiveness,andthe prevalenceofhabitualsurveytakersofferonlyminorissuesinpractice.Perhaps mostconclusively,theyreplicatefindingsfromexistingexperimentalresearchinthe socialsciences.Thefindings(Berinskyetal. 2012)presentsuggestthatdrawing subjectsfromtheInternetpopulationprovidescomparableresultstotakingsubjects fromauniversity’sundergraduatepopulation.7

WegatheredourdatausingasurveyprogrammedinQualtrics.Alinktothe surveywasplacedintheMechanicalTurkenvironmentwiththetasktitle‘‘Survey ofPersonalBehaviorandPersonality.’’Itwasavailableinthetwoweekspreceding electiondayin2012,fromOctober20toNovember6.Thecompletionratewas 92%,andwehavecompleteresponsesfor1,834respondentsformostanalyses. Thesurveyincludedbatteriestoevaluatearespondent’ssocialanxietylevelusing theSIASscale,personality(usingtheTenItemPersonalityScale(TIPI)Gosling etal. 2003),andstandardsurveyquestionsfordemographics,politicalinterest, informationseeking,andpoliticalbehavior.Speciﬁcquestionwordingcanbefound intheappendix.Approximatelyhalfwaythroughthesurvey,aquestionwas includedtoverifythatsubjectswerereadingtheinstructionsandnotsimply answeringquestionsrandomly.

Thefreeresponseanswerswereembeddedintoanexperimentattheendofthe survey.Participantswererandomlyassignedtooneoffourgroups:acontrolthat skippedthetreatment,oneaskedtowriteaboutthreethingsinherdailylifethat causestress,oneaskedtowritethreethingsaboutpoliticsthatcausestress,andone askedtonamethreethingsthatbrightenlife.Participants’answerswerethen displayedonthescreenandthesubjectswereaskedtoconﬁrmtheirresponses. Althoughwedonotanalyzetheresultsoftheexperimentinthispaper,wedouse thefree-responseanswersgeneratedbysubjectsinthe‘‘politicalstress’’condition.8

Intotal,440respondentswereinthe‘‘politicalstress’’condition,generatinga totalof1320freeresponseanswers.Wecodedtheseresponsesduringthespringof 2013.Researchassistantsfamiliarwiththeprojectdesignedacodingschemeand trainedthreestudentscompletelyunfamiliarwiththeprojectintheactual implementationofthescheme.Responseswereﬁrstcodedintothreebroad categoriesofstressors:theprocessofpolitics,policyissues,andpolitical

7 Thedatausedinthispaperwereprimarilygatheredforthepurposesofconductingasurveyexperiment. WerecognizethechallengesofdrawinginferencesusingMechanicalTurkforsurvey-basedanalysis,and considertheseresultspreliminary.

8 Thewordingread:‘‘Peopleoftenﬁndthattherearemanythingsaboutpoliticsthatbringstresstotheir lives,suchasnegativecampaigning,contentiousdisagreementsbetweentheirfriendsorneighbors,orthe wordsoractionsofpoliticians.Beingasspeciﬁcaspossible,pleaselistuptothreethingsrelatingto politicsthataddstresstoyourlife.’’

participation.Over96%ofallanswerswerecodedintooneofthesethree categories,andtheratesofagreementbetweenthecodersexceeded80%atthe categorylevel.Responseswerethenfurthercodedintosub-categoriesandtopics. Intercoderagreementatthesub-categorylevelrangedfrom75–80%andagreement atthetopiclevelrangedfrom70–75%.Moredetailedinformationaboutthe validityofthecodingprocessisavailablefromtheauthorsuponrequest.The tablebelowshowstheresultsforthe21.97%ofresponsesthatwerecodedintothe ‘‘participation’’category(Table 11, 12, 13,and 14).

Percentofsubcategory

Sub-category:politicsandrespondent

42.76%ofparticipationcategory

Understandingpoliticsordeterminingbeliefs22.58

Stayingupdated4.84

Electionuncertainty23.39

Effectofpoliticsonrespondent14.52

Politicalpowerlessness16.13

Expressingpoliticalviews0.81

Votingandregistrationprocesses12.10

Other5.65

Sub-category:politicsandrespondent’ssocialnetwork

57.24%ofparticipationcategory

Opinionsexpressedbymembersofrespondent’ssocialnetwork43.37 Interpersonalinteractions55.42 Other1.20

Table12 Modelcontrolvariables

VariableQuestionwordingScale

FemaleWhatisyourgender?1=female,0= male

WhiteWhatisyourrace?(White,BlackorAfricanAmerican,AmericanIndian orAlaskanNative,Asian,NativeHawaiianorPaciﬁcIslander,Hispanic orLatino,Other)

2012 engagement

Pleaseindicatewhetheryoudidanyofthefollowingactivitiesduringthe 2012elections—Checkallthatapply:(a)Talkedtoanypeopletotryto getthemtovotefororagainstoneofthepartiesorcandidates,(b)went toanypoliticalmeetings,rallies,speeches,ordinnersinsupportofa particularcandidate,(c)woreacampaignbutton,putacampaignsticker onyourcar,orplacedasigninyourwindoworyard,(d)didanyother workforoneofthepartiesorcandidates,(e)contributedmoneytoan individualcandidaterunningforofﬁce,(f)contributedmoneytoa politicalparty,(g)contributedmoneytoanyothergroupthatsupported oropposedcandidates

1=white,0=not white

Sumofall activities(0–7)

Table11 Datafromthesecondpilotstudy

VariableQuestionwordingScale

Knowledge(a)Doyouhappentoknowhowmanytimes anindividualcanbeelectedPresidentofthe UnitedStatesunderthecurrentlaws?(b)Is theU.S.federaldeﬁcit—theamountby whichthegovernment’sspendingexceeds theamountofmoneyitcollects—now bigger,aboutthesame,orsmallerthanit wasduringmostofthe1990s?(c)Forhow manyyearsisaUnitedStatesSenator elected—thatis,howmanyyearsaretherein onefulltermofofﬁceforaU.S.Senator?

(d)WhatisMedicare?Aprogramrunbythe U.S.federalgovernmenttopayforold people’shealthcare,aprogramrunbystate governmentstoprovidehealthcaretopoor people,aprivatehealthinsuranceplansold toindividualsinall50states,aprivatenonproﬁtorganizationthatrunsfreehealth clinics(e)Onwhichofthefollowingdoes theU.S.federalgovernmentcurrentlyspend theleast?Foreignaid,Medicare,National defense,Socialsecurity

InterestSomepeopledon’tpaymuchattentionto politicalcampaigns.Howaboutyou?Would yousaythatyouwereverymuchinterested, somewhatinterestedornotveryinterestedin thepoliticalcampaignsin2012?

Partisan attachment

IdentiﬁcationwithaPsychologicalGroup scalefromGreene(2002).10items measuredona5pointscale

IdeologyWehearalotoftalkthesedaysaboutliberals andconservatives.Hereisaseven-point scaleonwhichthepoliticalviewsthat peoplemightholdarearrangedfrom extremelyliberaltoextremelyconservative. Wherewouldyouplaceyourselfonthis scale,orhaven’tyouthoughtmuchabout this?

Table13 Study2regressionmodels:potentialconformity

Dependentvariable

PotentialConformity

Sumofcorrectanswers(0–5)

1=notmuchinterested,2=somewhat interested,3=verymuchinterested

5(leastattached)—50(mostattached)

1=extremelyliberal,2=liberal,3=slightly liberal,4=moderate;middleoftheroad,5= slightlyconservative,6=conservative,7= extremelyconservative,8=haven’tthought aboutitmuch

BasemodelStudycontrolsDemographiccontrolsPoliticalcontrols

Condition0.483**0.471**0.499**0.361 (0.195)(0.196)(0.197)(0.225) Knewpurpose -

Table13 continued

Dependentvariable

PotentialConformity

BasemodelStudycontrolsDemographiccontrolsPoliticalcontrols

OrderA0.2260.2870.265 (0.189)(0.195)(0.213)

Female0.1290.091 (0.193)(0.210)

White -0.115 -0.078 (0.209)(0.227)

2012engagement0.027 (0.065)

Politicalknowledge -0.016 (0.111)

Politicalinterest -0.034 (0.181)

Partisanattachment0.020 (0.021)

Ideology0.014 (0.067)

Constant0.322**0.2740.248 -0.027 (0.158)(0.204)(0.282)(0.758) Observations63636261

Loglikelihood -95.456 -93.254 -91.168 -89.130 Akaikeinf.crit.194.912194.508194.337200.261

Tablesformattedwith stargazer Hlavac(2015) * p [ 0.1;** p \ 0.05;*** p \ 0.01

Table14 Study2regressionmodels:pureconformity

Dependentvariable

Pureconformity

BasemodelStudycontrolsDemographiccontrolsPoliticalcontrols

Condition0.4700.4420.438 -0.133 (0.329)(0.331)(0.333)(0.392)

Knewpurpose -0.440 -0.481 -0.497 (0.420)(0.425)(0.475)

OrderA -0.105 -0.153 -0.200 (0.322)(0.330)(0.347)

Female0.3030.158 (0.342)(0.358)

Table14 continued

Dependentvariable

Pureconformity

BasemodelStudycontrolsDemographiccontrolsPoliticalcontrols

White0.4310.665 (0.428)(0.478) 2012engagement0.008 (0.109)

Politicalknowledge -0.143 (0.190)

Politicalinterest -0.169 (0.317)

Partisanattachment0.062 (0.037)

Ideology -0.304** (0.133)

Constant -0.427* -0.261 -0.758 -0.035 (0.258)(0.318)(0.517)(1.417)

Observations46464645

Loglikelihood -52.708 -52.081 -51.161 -44.151 Akaikeinf.crit.109.416112.162114.323110.303

Tablesformattedwith stargazer Hlavac(2015)

* p \ 0.1;** p \ 0.05;*** p \ 0.01

References

Abramowitz,AlanI.(2006).Commentondisconnected:Thepoliticalclassversusthepeople.InP. S.Nivola&D.W.Brady(Eds.), Redandbluenation?CharacteristicsandcausesofAmerica’s polarizedpolitics (pp.72–84).Washington,DC:BrookingsInstitutionPress.

Abramowitz,A.I.(2010). Thedisappearingcenter:Engagedcitizens,polarization,&American democracy.NewHaven,CT:YaleUniversityPress.

Abramowitz,A.I.,&Saunders,K.(1998).IdeologicalrealignmentintheU.S.elections. Journalof Politics, 60(3),634–652.

Abramowitz,A.I.,&Saunders,K.(2005).Whycan’twealljustgetalong?Therealityofapolarized America. TheForum:AJournalofAppliedResearchinContemporaryPolitics, 3(2),1–22.

Abramowitz,A.I.,&Saunders,K.(2008).Ispolarizationamyth? JournalofPolitics, 70(2),542–555.

Ahn,T.K.,Huckfeldt,R.,Mayer,A.K.,&Ryan,J.B.(2013).Expertiseandbiasinpolitical communicationnetworks. AmericanJournalofPoliticalScience, 57(2),357–373.

Ahn,T.K.,Huckfeldt,R.,&Ryan,J.B.(2010).Communication,inﬂuence,andinformational asymmetriesamongvoters. PoliticalPsychology, 31(5),763–787.PresentedattheConferenceon SocialDilemmas,sponsoredbytheResearchGroupforExperimentalSocialScienceatFloridaState University.

Ahn,T.K.,Huckfeldt,R.,&Ryan,J.B.(2014). Experts,activists,anddemocraticpolitics:Are electoratesself-educating?Cambridgestudiesinpublicopinionandpoliticalpsychology Cambridge,MA:CambridgeUniversityPress.

Ansolabehere,S.,&Schaffner,B.Doessurveymodestillmatter? PoliticalAnalysis (Forthcoming).

Asch,S.E.(1956).Studiesofindependenceandconformity:Aminorityofoneagainstaunanimous majority. PsychologicalMonographs, 70(9),416.

Bafumi,J.,&Shapiro,R.Y.(2009).Anewpartisanvoter. JournalofPolitics, 71(1),1–24.

Baron,R.M.,&Kenny,D.A.(1986).Themoderator-mediatorvariabledistinctioninsocial psychologicalresearch:Conceptual,strategic,andstatisticalconsiderations. JournalofPersonality andSocialPsychology, 51(6),1173–1182.

Barry,H.,Child,L.,&Bacon,M.(1959).Relationofchildtrainingtosubsistenceeconomy. American Anthropology, 61,51–63.

Bennett,S.E.,Flickinger,R.S.,&Rhine,S.L.(2000).PoliticalTalkoverhere,overthere,overtime. BritishJournalofPoliticalScience, 30(1),99–119.

Berelson,B.R.,Lazarsfeld,P.F.,&McPhee,W.N.(1954). Voting:Astudyofopinionformationina presidentialcampaign.Chicago,IL:UniversityofChicagoPress.

Berinsky,A.J.,Huber,G.A.,&Lenz,G.S.(2012).Evaluatingonlinelabormarketsforexperimental research:Amazon.com’smechanicalturk. PoliticalAnalysis, 20(3),351–368.

Blanton,H.,&Christie,C.(2003).Devianceregulation:Atheoryofidentityandaction. Reviewof GeneralPsychology, 7,115–149.

Blattman,C.(2015).WhyIworryexperimentalsocialscienceisheadedinthewrongdirection. http:// chrisblattman.com/2015/12/07/if-you-run-ﬁeld-experiments-this-might-be-paper-that-will-make-itharder-to-publish-your-work-in-a-few-years/

Bond,R.,&Smith,P.(1996).Cultureandconformity:Ameta-analysisofstudiesusingasch’s(1952b, 1956)linejudgmenttask. PsychologicalBulletin, 119,111–137. Brewer,M.B.,&Roccas,S.(2001).Individual,self,relationalself,collectiveself.PsychologyPress. ChapterIndividualValues,SocialIdentity,andOptimalDistinctiveness,pp.219–37. Burnett,C.(2012).Artiﬁcialintelligence:Comparingsurveyresponsesforonlineandofﬂinesamples.In APSA2012annualmeetingpaper. Campbell,A.,Gurin,G.,&Miller,W.E.(1954). Thevoterdecides.Peterson:Row. Campbell,A.,Converse,P.E.,Miller,W.E.,&Stokes,D.E.(1960). TheAmericanvoter.NewYork: Wiley.

Caro,F.G.,Ho,T.,McFadden,D.,Gottlieb,A.S.,Yee,Christine,Chan,Taizan,etal.(2012).Usingthe internettoadministermorerealisticvignetteexperiments. SocialScienceComputerReview, 30(2), 184–201.

Chartrand,T.L.,&Bargh,J.A.(1999).Thechameleoneffect:Theperception-behaviorlinkandsocial interaction. JournalofPersonalityandSocialPsychology, 76,893–910. Cialdini,R.B.,Wosinska,W.,Barrett,D.W.,Butner,J.,&Gornik-Durose,M.(1999).Compliancewith arequestintwocultures:Thedifferentialinﬂuenceofsocialproofandcommitment/consistencyon collectivistsandindividualists. PersonalityandSocialPsychologyBulletin, 25,1242–1253. Cialdini,R.B.(2001). Inﬂuence:Scienceandpractice (4thed.).Newton,MA:Allyn&Bacon.

Cialdini,R.B.,AlanLevy,C.,Herman,P.,&Evenbeck,S.(1973).Attitudinalpolitics:Thestrategyof moderation. JournalofPersonalityandSocialPsychology, 25(1),100–108. Cialdini,R.B.,&Goldstein,N.J.(2004).Socialinfluence:Complianceandconformity. AnnualReviews Psychology, 55,591–621. Converse,P.E.(1964). Thenatureofbeliefsystemsinmasspublics.Glencoe:FreePress. Crabtree,C.,Fariss,C.J.,&Kern,H.L.(2015).Truthreplacedbysilence:Afieldexperimentonprivate censorshipinRussia.AvailableatSSRN2708274. Crutchfield,R.S.(1955).Conformityandcharacter. TheAmericanPsychologist, 10(5),191–198. Dalton,R.J.(2008).Citizenshipnormsandtheexpansionofpoliticalparticipation. PoliticalStudies, 56, 76–98.

Downs,A.(1957).Aneconomictheoryofdemocracy. HarperandRow Druckman,J.,&Nelson,K.(2003).Framinganddeliberation:Howcitizens’conversationslimitelite inﬂuence. AmericanJournalofPoliticalScience, 47(4),729–745. Dryzek,J.S.(1994). Discursivedemocracy:Politics,policy,andpoliticalscience.Cambridge,MA: CambridgeUniversityPress. Festinger,L.(1950).Informalsocialcommunication. PsychologicalReview, 57(5),271–282. Gerber,A.S.,&Green,D.P.(2000).Theeffectsofcanvassing,telephonecalls,anddirectmailonvoter turnout:Aﬁeldexperiment. AmericanPoliticalScienceReview, 94(3),653–663. Gerber,A.S.,&Green,D.P.(2012). Fieldexperiments:Design,analysis,andinterpretation.NewYork, NY:WWNorton.

Gerber,A.S.,Green,D.P.,&Larimer,C.W.(2008).Socialpressureandvoterturnout:Evidencefroma large-scaleﬁeldexperiment. AmericanPoliticalScienceReview, 102(1),33–48.

Gerber,A.S.,Huber,G.A.,Doherty,D.,&Dowling,C.M.(2012).Disagreementandtheavoidanceof politicaldiscussion:Aggregaterelationshipsanddifferencesacrosspersonalitytraits. American JournalofPoliticalScience, 56(4),849–874.

Giusefﬁ,K.E.,Smith,K.B.,&Hibbing,J.R.(2013).Socialanxiousnessandpoliticalparticipation. Paperpresentedatthe2013AnnualMeetingoftheAmericanPoliticalScienceAssociation, Chicago,IL.

Gosling,S.D.,Rentfrow,P.J.,&Swann,W.B,Jr.(2003).Averybriefmeasureofthebig-ﬁve personalitydomains. JournalofResearchinPersonality, 37,504–528.

Green,D.P.,Palmquist,B.,&Schickler,E.(2002). Partisanheartsandminds.NewHaven,CT:Yale UniversityPress.

Greene,S.(2002).Thesocial-psychologicalmeasurementofpartisanship. PoliticalBehavior, 24(3), 171–197.

Haidt,J.(2014).Yourpersonalitymakesyourpolitics.TimeMagazine. http://science.time.com/2014/01/ 09/your-personality-makes-your-politics/.

Haidt,J.,&Wilson,C.(2014).CanTIMEpredictyourpolitics?Seehowyourpreferencesindogs, Internetbrowsers,and10otheritemspredictyourpartisanleanings.TIMEMagazine. http://time. com/510/can-time-predict-your-politics/

Haidt,J.,&Hetherington,M.J.(2012).Lookhowfarwe’vecomeapart.CampaignStops,September17. http://campaignstops.blogs.nytimes.com/2012/09/17/look-how-far-weve-come-apart/

Hayes,A.F.(2007).Exploringtheformsofself-censorship:Onthesprialofsilenceandtheuseof opinionexpressionavoidancestrategies. JournalofCommunication, 57,785–802.

Hayes,A.F.,Glynn,C.J.,&Shanahan,J.(2005).Willingnesstoself-censor:Aconstructand measurementtoolforpublicopinionresearch. InternationalJournalofPublicOpinionResearch, 17(3),298–323.

Hetherington,M.J.,&Weiler,J.D.(2009). Authoritarianismandpolarizationinamericanpolitics Cambridge,MA:CambridgeUniversityPress.

Hibbing,J.R.,&Theiss-Morse,E.(2002). Stealthdemocracy:Americans’beliefsabouthowgovernment shouldwork.Cambridge,MA:CambridgeUniversityPress.

Hibbing,J.R.,Ritchie,M.,&Anderson,M.R.(2010).Personalityandpoliticaldiscussion. Political Behavior, 33(4),601–624.

Hlavac,M.(2015).Stargazer:Well-formattedregressionandsummarystatisticstables.Rpackage version5.2 http://CRAN.R-project.org/package=stargazer.

Hofstede,G.(1980). Culturesconsequences:Internationaldifferencesinwork-relatedvalues.Beverly Hills,CA:Sage.

Huckfeldt,R.,Johnson,P.E.,&Sprague,J.(2004). Politicaldisagreement:Thesurvivalofdiverse opinionswithincommunicationnetworks.Cambridge,MA:CambridgeUniversityPress.

Huckfeldt,R.R.,&Sprague,J.(1995). Citizens,politics,andsocialcommunication:Informationand inﬂuenceinanelectioncampaign.Cambridge,MA:CambridgeUniversityPress.

Iyengar,S.,Sood,G.,&Lelkes,Y.(2012).Affect,notideology:Asocialidentityperspectiveon polarization. PublicOpinionQuarterly, 76(3),405–431.

Iyengar,S.,&Westwood,S.J.(2015).Fearandloathingacrosspartylines:Newevidenceongroup polarization. AmericanJournalofPoliticalScience, 59(3),690–707.

Karpowitz,C.F.,&Mendelberg,T.(2007).Groupsanddeliberation. SwissPoliticalScienceReview, 13(4),645–662.

Karpowitz,C.F.,Mendelberg,T.,&Shaker,L.(2012).Genderinequalityindeliberativeparticipation. AmericanPoliticalScienceReview, 106(3),533–547. Katz,E.,&Lazarsfeld,P.F.(1955). Personalinﬂuence:Thepartplayedbypeopleintheﬂowofmass communications.NewYork:FreePress. Khan,R.,Misra,K.,&Singh,V.(2013).Ideologyandbrandconsumption. PsychologicalScience, 24(3), 326–333.

Kim,H.S.,&Markus,H.R.(1999).Devianceoruniqueness,harmonyorconformity?Acultural analysis. JournalofPersonalityandSocialPsychology, 77,785–800.

Klofstad,C.A.,McDermott,R.,&Hatemi,P.K.(2012).Thedatingpreferencesofliberalsand conservatives. PoliticalBehavior,120 Lasswell,H.D.(1936). Politics:whogetswhat,when,how.PeterSmith. Lasswell,H.D.(1941). Democracythroughpublicopinion.GeorgeBantaPublishingCompany.

Latane,B.(1996).Dynamicsocialimpact:Thecreationofculturebycommunication. Journalof Communication, 46,13–25.

Lazarsfeld,P.F.,Berelson,B.,&Gaudet,H.(1968). Thepeople’schoice:Howthevotermakesuphis mindinapresidentialcampaign.NewYork:ColumbiaUniversityPress.

Levendusky,M.(2009). Thepartisansort:Howliberalsbecamedemocratsandconservativesbecame republicans.Chicago:UniversityofChicagoPress.

Levitan,L.C.,&Verhulst,B.(2015).Conformityingroups:Theeffectsofothersviewsonexpressed attitudesandattitudechange. PoliticalBehavior

Levitan,L.,&Visser,P.(2009).Socialnetworkcompositionandattitudestrength:Exploringthe dynamicswithinnewlyformedsocialnetworks. JournalofExperimentalSocialPsychology, 45, 1057–1067.

Levy,M.,&Dubinsky,A.J.(1983).Identifyingandaddressingretailsalespeople’sethicalproblems:A methodandapplication. JournalofRetailing, 59(1),46–66.

Lupia,A.,&McCubbins,M.D.(1998). Thedemocraticdilemma:Cancitizenslearnwhattheyneedto know?.Cambridge,MA:CambridgeUniversityPress.

Mason,L.(2013).Theriseofuncivilagreement:IssueversusbehavioralpolarizationintheAmerican electorate. AmericanBehavioralScientist, 57(1),140–159.

Mason,L.(2015).’Idisrespectfullyagree’:Thedifferentialeffectsofpartisansortingonsocialandissue polarization. AmericanJournalofPoliticalScience, 59(1),128–145.

Mondak,J.J.(2012). Personalityandthefoundationsofpoliticalbehavior.Cambridge,MA:Cambridge UniversityPress.

Mutz,D.C.(1998). Impersonalinﬂuence:Howperceptionsofmasscollectivesaffectpoliticalattitudes Cambridge,MA:CambridgeUniversityPress.

Mutz,D.C.(2006). Hearingtheotherside:Deliberativeversusparticipatorydemocracy.Cambridge, MA:CambridgeUniversityPress.

Mutz,D.C.,&Mondak,J.J.(1998).Theworkplaceasacontextforcross-cuttingpoliticaldiscourse. JournalofPolitics, 68(1),140–155.

Noelle-Neumann,E.(1993). Thespiralofsilence:Publicopinion-oursocialskin.Chicago:Universityof ChicagoPress.

Nowak,A.,&Vallacher,R.R.(2001).Societaltransition:Towardadynamicalmodelofsocialchange. ThePracticeofSocialInﬂuenceinMultipleCultures,151–71.

Pool,G.J.,Wod,W.,&Leck,K.(1998).Theself-esteemmotiveinsocialinﬂuence:Agreementwith valuedmajoritiesanddisagreementwithderogatedminorities. JournalofPersonalityandSocial Psychology, 75,967–975.

Putnam,R.D.(2001). Bowlingalone:Thecollapseandrevivalofamericancommunity.Touchstone BooksbySimonandSchuster.

Ryan,J.B.(2010).Theeffectsofnetworkexpertiseandbiasesonvotechoice. PoliticalCommunication, 27,44–58.

Ryan,J.B.(2011).Socialnetworksasashortcuttocorrectvoting. AmericanJournalofPoliticalScience, 55(4),752–765.

Schoemaker,P.J.H.(1993).MultipleScenariodevelopment:Itsconceptualandbehavioralfoundation. StrategicManagementJournal, 14(3),193–213.

Sears,D.O.(1986).Collegesophomoresinthelaboratory:Inﬂuencesofanarrowdatabaseonsocial psychology’sviewofhumannature. JournalofPersonalityandSocialPsychology, 51(3),515–530.

Settle,J.E.,Bond,R.,&Levitt,J.(2011).Thesocialoriginsofadultpoliticalbehavior. AmericanPolitics Research, 39(2),239–263.

Sinclair,B.(2012). Thesocialcitizen:Peernetworksandpoliticalbehavior.Chicago:Universityof ChicagoPress.

Suhay,E.(2015).Explaininggroupinﬂuence:Theroleofidentityandemotioninpoliticalconformity andpolarization. PoliticalBehavior, 37,221–251.

Triandis,H.C.(1990).Cross-culturalstudiesofindividualismandcollectivism.InJ.J.Berman(Ed.) Nebraskasymposiumonmotivation (Vol.37,pp.41–133).

Ulbig,S.G.,&Funk,C.(1999).Conﬂictavoidanceandpoliticalparticipation. PoliticalBehavior, 21(3), 265–282.

Weber,J.(1992).Scenariosinbusinessethicsresearch:Review,criticalassessment,andrecommendations. BusinessEthicsQuarterly, 2(2),137–160.

Wilson,G.D.(1973).Conservatismandartpreferences. JournalofPersonalityandSocialPsychology, 25(2),286–288.

Young,A.(2016).Channellingfisher:Randomizationtestsandthestatisticalinsignificanceofseemingly significantexperimentalresults.WorkingPaperasofFebruary2016. Zuckerman,A.S.(2005).Thesociallogicofpolitics:Personalnetworksascontextsforpolitical behavior.TempleUniversityPress.