Ordinary Voting Behavior in the Extraordinary Elec

OrdinaryVotingBehaviorintheExtraordinaryElectionofAdolfHitler1GaryKing2,OriRosen3,MartinTanner4,andAlexanderF.Wagner5April2,20041OurthanksgotoColinFlint,J¨urgenFalter,andJoshTuckerfordataandothermaterials;toEricDick-son,JamieHolmes,KiraPetersen,andAlisonPostformanydiscussionsandsuperbresearchassistance;toJimAlt,WilliamBrustein,ColinFlint,MarkKishlansky,CharlieMaier,andJohnO’Loughlinforhelpfulcomments;andtoNIA/NIH(grantP01AG17625-01)andNSF(grantsSES-0112072andIIS-9874747)forresearchsupport.2DavidFlorenceProfessorofGovernment,DepartmentofGovernment,HarvardUniversity(CenterforBasicResearchintheSocialSciences,CambridgeMA02138;@Harvard.Edu).3AssistantProfessor,DepartmentofStatistics,UniversityofPittsburgh(2702CathedralofLearning,Pittsburgh,PA15260,(412)6249033,[e-mail]).4ProfessorofStatistics,DepartmentofStatistics,NorthwesternUniversity(2006SheridanRoad,Evanston,IL60208;[phone];[e-mail])5Ph.D.StudentinPoliticalEconomyandGovernment,HarvardUniversity([e-mail]).AbstractHowdidfreeandfairdemocraticelectionsleadtotheextraordinaryanti-democraticNaziPartywinningcontroloftheWeimarRepublic?Theprofoundimplicationsofthisquestionhaveled“WhovotedforHitler?”tobethemoststudiedquestioninthehistoryofvotingbehaviorresearch.Yet,despitetheoverwhelmingattention,scholarshavetreatedtheseelectionsasuniqueeventsandcomparisonwithotherelectionsasmostlyirrelevant.Indeed,theliteraturerarelybuildsoninsightsfromvotingbehaviorresearchinpoliticalscienceorfrommodernstatisticalmethods.Byaddingtheseapproaches,wefindthatsomeofthemostwidelyacceptedcurrenttheoriesofNazivotingdonotdistinguishtheWeimarelectionsfromalmostanyothers,inanycountry.Likesomeresearchers,weadoptarestrospectivevotingaccount,butwealsorecognizethatthevoterswhoweremosthurtbytheeconomicdepressionandhencemostlikelytoopposethegovernmentfallintotwoseparategroupswithdivergentinterests.Withthisadditionalfeatureandimprovedmeth-ods,weareabletoshowwhysomeofthedowntroddenturnedtotheNazisandothersturnedaway.TheconsequencesoftheelectionofHitlerwereextraordinary,butavailableevidencesuggeststhatthevotingbehaviorthatledtoitwasnot.1IntroductionThemostanalyzedelectionsinhistoryarethosewhichledtotheriseoftheNaziPartyinWeimarGermany.Indeed,understandingwhovotedforHitlerisespeciallytimelynow,giventheincreaseinradicalandextremistpartiesinmodernEuropeanelections.Yet,despiteallthiswork,theNazivotingliteraturerarelydrawsontheextensivevotingbehaviorliteratureinpoliticalscienceoronmodernstatisticalmethods.ThatdoesnotmaketheNazivotingliteraturewrong,andwedonotthinkitiswrong.Butitsuggeststhatlearningsomethingnewabouttheseextraordinaryeventsmaybepossible.Inthispaper,weofferaperspectiveontheseelectionsthatbuildsonthepoliticalsciencevotingbehaviorliteratureandcomparisonswithotherelectionssothatotherscholarsmaybeabletofollowup.Wealsointroduceintothisliteraturenewmethodsofecologicalinferencedesignedspecificallyforthisproblemthatgreatlyincreasetheamountofinformationwecanbringtobearonthesevenerablequestions.MuchofthecurrentliteratureontheWeimarRepublictreatstheelectionsasauniquehistoricalcase,andindeedtheconsequencesoftheelectionswereextraordinary.Incontrast,onceweaccountforaspectsoftheheterogenousinterestsofvotersintheseelectionsthatmayhavepreviouslybeenunderappreciated,wefindthatGermanvotersrespondedinafairlyordinaryway,consistentwiththeirinterestsandthusalsoconsistentwiththedominantpatternoffindingsaboutdemocraticelectionsinothercountriesandtimeperiods.Althoughthelimitednatureoftheevidenceavailablegeneratesconsiderableuncertaintiesinallanalysesincludingthosepresentedhere,ourresearchsuggestsanumberofnewconclusions.First,oneoftheleadingtheoriesintheliterature—thattheNaziswerea“catch-all”party,becausetheswingintheirfavoroccurredroughlyuniformlyacrossmanydifferentsocialgroups—missesthefactthatmostelectionsfromnumerouscountriesexaminedinthiswaydisplayessentiallythesametypeofuniformpartisanswing.Assuch,thistheory,whilenotwrong,doesnotdistinguishtheWeimarelectionsfrommostothersandsocannotbeusedtohelpexplainwhathappenedhere.Weinsteadfindthatanintention-basedretrospectivevotingaccountofthisseriesofelectionshelpstoorganizealargecollectionofotherwiseconfusingfacts.Inparticular,wefindthatthegroupsmosthurtbythedisastrouseconomicdepressiondidnothavehomogeneousinterests,ashadbeenpreviouslyassumed,andasaresultdidnotbehavethesame.ThosewhowereunemployedorathighriskofbecomingunemployedgavedisproportionatesupporttotheCommunists,forgoodreason,whereasthosewhowerehurtbytheeconomybutwereatlittleriskofunemployment(suchasself-employedshopkeepersandprofessionals,domesticemployees,andhelpingfamily1members)constitutedthegroupsthatgavethemostdisproportionatesupporttotheNazis.WebegininSection2withadescriptionofthehistoricalNazivotingliteraturefromtheper-spectiveofthepoliticalsciencevotingbehaviorliterature.Section3summarizesourmainsubstan-tiveargumentsanddiscusseshowourcon