CountingWomen’sBallots:FemaleVotersafterSuffrageintheU.S.
ChristinaWolbrechtAssociateProfessorofPoliticalScience
UniversityofNotreDameNotreDame,[email protected]
PaperpreparedforpresentationattheEuropeanConferenceonPoliticsandGender,heldinUppsala,Sweden,June11‐13,2015.Thispaperreportsonworkco‐authoredwithJ.KevinCorder(WesternMichiganUniversity,USA)andsettoappearinCountingWomen’sBallots:FemaleVotersfromSuffrageThroughtheNewDeal,undercontractwithCambridgeUniversityPress.WegratefullyacknowledgedthesupportoftheFacultyResearchProgramattheUniversityofNotreDame,theResearchDevelopmentAwardProgramatWesternMichiganUniversity,andtheNationalScienceFoundation(SES‐9905843and9905307).SomeofthedataemployedweremadeavailablebytheInter‐universityConsortiumforPoliticalandSocialResearch(ICPSR);theConsortiumbearsnoresponsibilityfortheanalysesorinterpretationspresentedhere.Theauthorsaresolelyresponsibleforthecontentofthispaper.Thisconferencepapershouldnotbecitedwithoutpermission.
1
CountingWomen’sBallots:FemaleVotersafterSuffrageintheU.S.
OnAugust18,1920,Tennesseebecamethethirty‐sixthstateintheuniontoratifythe19thAmendmenttotheUnitedStatesConstitution.Afteramorethan70yearbattle,womenthroughouttheU.S.securedtherighttovote.ThenationalenfranchisementofwomenrepresentedthelargestexpansionoftheelectorateinAmericanhistory,nearlydoublingthesizeofthevotingagepopulation.1Millionsofcitizenswhohadnevercastaballotbecameeligibletodoso.
Justfourpresidentialelectionslater,theU.S.wouldexperienceoneofthemost
dramaticelectoraltransformationsinitshistory.RepublicanshadbeenascendantsincetheendoftheCivilWar,butthecrisisoftheGreatDepression—andtheparties’divergentresponsestoit—transformedpartisanalignmentsintheU.S.andusheredinaneraofDemocraticdominanceformuchofthe20thcentury.The“NewDealrealignment”ofthe1930sestablishedanewDemocraticcoalitionwhichwouldpersistfordecades.
ThecausesandconsequencesNewDealrealignmenthavereceivedconsiderable
attention.Scholarshavehadfarlesstosayaboutthecontributionofwomentothisprocess.Yetwehavereasontoexpectwomenmayhaveplayedakeyroleintheenormousvotermobilizationduringthisperiod.AlargenumberofstillelectorallyinactivewomenwereamongthosemostavailableformobilizationintotheDemocraticpartyinthe1930s.Atthesametime,womenwhohadvotedinthe1920swere—asaresultoftheirrelativelyfeweropportunitiestohavereinforcedtheirpartisanshipbycastingballots—amongthosemostavailableforconversionfromonepartisanallegiancetotheother.Thispaperasks:Didwomenandmenrealignviathesameordifferentprocesses?WhatwasthecontributionofwomentoNewDealrealignment?
Previousscholarshasbeenhinderedintheconclusionstheycouldreachregarding
theelectoralbehaviorofthefirstfemalevotersbecausewepossessverylimitedusefuldataonhowwomenvotedaftersuffrage.Withrareexceptions,officialrecordsreportonlythetotalnumberofvotescastoverallandforeachcandidate.Whetherwomencastballots,forwhichcandidates,andwithwhatconsequencescannotbedirectlydeterminedfromofficialrecordsalone.Reliablepublicopinionpolls—themodernsolutiontothisproblem—arevirtuallynon‐existentduringthisperiod.Earlyresearchersattemptedtodrawconclusionsfromtheavailableaggregateelectionandcensusrecords,butsinceRobinson(1950)socialscientistshaveunderstoodthedangersofwhatisknownastheecologicalfallacy(seebelow)andgenerallyshiedawayfromsuchanalysis.
1Wesay“nearly”becauseelevenstatesallowedwomentovoteinthe1916presidentialelection.Ontheotherhand,restrictiveinterpretationsofregistrationrules(ratificationoccurredafterregistrationdeadlinesinanumberofstates)systematicallydeniedwomenaccesstotheballotinArkansas,Georgia,Mississippi,andSouthCarolinain1920,delayingtheirparticipationinpresidentialelectionsuntil1924(Gosnell1930).Togetherwithblackmen,manyblackwomencontinuedtoexperiencesystematicexclusionfromthefranchiseuntilthesecondhalfofthe20thcentury.Forthatreason,weassumethatourconclusionsaboutthevotingbehavioroffemalevoterslargelydescribewhitefemalevoters.
2
Thispaperreportsonresultsfromaprojectwhichseekstoovercomethesedataandmethodologicalchallengesandindoingso,deepenandimproveourunderstandingofanimportantperiodinAmericanelectoralhistoryandAmericanpoliticaldevelopment.Combininguniquehistoricelectiondataandrecentmethodologicalinnovations,weareabletoestimatetheturnoutandvotechoiceofnewfemalevotersinthefivepresidentialelectionsfollowingwomen’snationalenfranchisement(1920‐1936)foralargerandmorediversesetofplaces—tenAmericanstates—thanhaspreviouslybeenpossible.Inthispaper,weemploytheestimatesforthepresidentialelectionsof1932and1936toexaminethecontributionsofwomentotheprocessofNewDealrealignment.
ExpectationsforWomenandNewDealRealignment
Theelectoralrealignmentthatoccurredbetween1928and1936hasbeenthe
subjectofaremarkablevolumeofscholarlywork.AquestionofparticularinterestistheextenttowhichtheshiftfromRepublicantoDemocraticdominancewascharacterizedbythemobilizationofpreviouslyinactivecitizensortheconversionofthosealreadyactive.TheconversionstorypositsthatwidespreaddissatisfactionwiththeRepublicanresponsetotheeconomiccrisispersuadedpreviousRepublicanvoterstosupporttheDemocraticnominee,FranklinD.Roosevelt(FDR),in1932,andthenreinforcethatpreferencebyvotingforFDRinsubsequentelections,resultinginalong‐termshiftfromRepublicantoDemocraticloyalty(Brown1988).ThoseparticularlyhardhitbytheDepression—e.g.,farmersandurbanindustrialworkers—werebelievedthemostlikelycandidatesforconversion(Gourevitch1984;Sundquist1973).
Others,however,haveemphasizedtheconsiderableevidenceofindividual‐level
stabilityofpartisanattachment,afindingdatingfromtheinfluentialearlyvotestudies(Berelson,Lazarsfeld,andMcPhee1954;Campbelletal.1960,1966;Lazarsfeld,Berelson,andGaudet1948)andapersistent,ifcontested,thesistoday(e.g.,Green,Palmquist,andSchickler2002).Ifpartisanshipisan“unmovedmover”(Johnston2006),thenDemocraticgainslikelycamelessfromparty‐switchingbyRepublicans,andmorefromthemobilizationofneworpreviouslyinactivecitizens,suchasthemassivenumbersofnewimmigrantsenteringtheU.S.intheprecedingdecades,respondingtotheappealoftheDemocraticparty(seeSalisburyandMacKuen1981).
Amidstdebatesovertheoryandmethod,scholarshavefoundsupportforboth
mobilization(Andersen1979;Campbelletal.1960;Campbell1985;Prindle1979;Wanat1979)andconversion(Burnham1970;EriksonandTedin1981;HawleyandSagarazu2012;LaddandHadley1975;Sundquist1983)asthemechanismsforNewDealrealignment.Differentelections,groups,andplacesmayhavebeencharacterizedbydifferentkindsofprocesses(e.g.,Andersen2014,Brown1988,Gamm1989;Nardulli1995).
Wehavereasontoexpectthatwomenwerebothmorelikelytoconvertfrom
previousRepublicansupportandmorelikelytobemobilizedintotheemergingDemocraticmajorityduringtheNewDealelections.Sincewomenhadbeenlesslikelythanmentoturnouttovoteintheelectionsofthe1920s,morewomenthanmenwereavailablefor
3
mobilization—thatis,thereweremorewomenwhowerenotalreadymembersoftheactiveelectorate.Indeed,intermsofsheernumbers,therewerelikelymorewomenavailableformobilizationthananyotherdemographicgroup.Atthesametime,womenwhohadvotedduringthe1920smayhavebeenmorelikelytoconvertsincetheirrelativelyshortexperienceatthepollsprovidedlessopportunitytoreinforcepartisanpreferences.Inotherwords,womenboastedlowerlevelsof“politicalimmunization”(McPheeandFerguson1962),theresistancetodisturbanceaccumulatedfromrepetitionandreiteration.
Yet,manyofthemostprominentrealignmentandpartisanshipscholarshavebeen
silentonthepotentialcontributionsoffemalevoters.Inhisinfluentialbook,CriticalElectionsandtheMainspringsofAmericanPolitics,Burnham(1970)notesthelikelydifferentialmobilizationofnativeandimmigrantwomenpriorto1928butdoesnotimplicatethemintherealignmentthatfollowed.Likewise,inSundquist’s(1983)importantstatementonpartyrealignment,DynamicsofthePartySystem,thechapteronNewDealrealignmentdoesnotmentionwomen.Kleppner(1982,89)identifies“immigrant‐stockvoters,theyoung,thosetowardthebottomoftheeconomicladder,theunemployed,reliefers,andcitizenswhohadchosentoabstaininthe1920s”asthoseresponsibleforincreasedturnoutduringtheNewDealperiod;whilemanywomenfellintothelastcategory,hedoesnotdiscusswomenspecificallywithregardstotheNewDeal.Similarly,Gosnell(1942,23),describingthesix‐foldincreaseinDemocraticvotersbetween1928and1936inPennsylvania,concludesthattheDemocratsattracted“ahugearmyofnewvoters—theyoungvotersandthosewhohadformerlybeennon‐votersthroughindifference.”Again,wemightexpectmanywomentofallintothatsecondcamp,buttheygounmentioned.Finally,andperhapsmostimportantlyintermsofestablishingtheconventionalwisdom,intheirclassicbook,TheAmericanVoter,Campbellandhiscolleagues(1960,153)implicate“theyouth,theeconomicallyunderprivileged,andtheminoritygroups,”butnotwomen,assourcesofthenewDemocraticmajority.
Afewscholarshaveatleastallowedforthepossibilityofaroleforfemalevotersin
theNewDealelections.DrawingexplicitlyonMcPheeandFerguson’s(1962)conceptofpoliticalimmunization,Andersen(1979)identifiestheenfranchisementofwomenasacentralcontributortothelargenumberofnon‐immunizedvotersinthe1920swhoweresubsequentlyavailableforDemocraticconversionandmobilizationduringtheNewDealperiod.Inotherwork,Burnham(1974,1015)pointstotheheavymobilizationofwomen,especially“ethnicwomen,”duringthe“1928‐1936realignmentsequence.”UsingdatafromBostonwards,Gamm(1989)findsthatwomenweremobilizedasDemocratslaterthanmen;increasesinDemocraticsupportbefore1934werearesultofsupportfrommen,butafterthattime,women’sgainssurpassedthoseofmen.
ThegeneralinattentiontowomenasafactorinNewDealrealignmentcanbe
attributedtoatleasttwofactors.First,theconventionalwisdomthatfemalevotershadfailedtomaterializeasauniquevotingblocortodiffermuchfrommenintheirpoliticalpreferenceshadbeenlargelyacceptedasconventionalwisdombythetimeKey(1955)hadcalledattentiontocriticalelectionsandtheNewDealperiodinparticularhadbecomethe
4
subjectofwidespreadscholarlyinterest.Asaresult,fewscholarslikelyviewedwomenaspotentialcontributorstotheprocessofelectoralchangeinthe1930s.
Second,thedearthofattentiontowomeninthepreviousliteratureonNewDeal
realignmenthighlightsthemethodologicalchallengestostudyinggenderdifferencesinelectoralbehaviorpriortotheadventofsurveyresearch.MuchoftheliteratureonNewDealrealignmentischaracterizedbymeticulousdataworkinwhichscholarscarefullyidentifiedcountiesandothergeopoliticaldivisionswithdemographicpopulationsthatallowedreasonable(inmostcases)inferencesabouttheturnoutandvotechoiceofparticulargroups(Andersen1979;Gamm1989;Key1955;Kleppner’s1982;Nardulli1995).Racial,ethnic,andimmigrantresidentialsegregationmakesisolatingsuchgroupsgeographicallyfarmoretenablethanidentifyingfemalevotersseparatefrommale.ThisempiricalchallengedoesnotjustifythefailuretosomuchasspeculateastothepossibledistinctivecontributionofwomentoNewDealrealignment,however.Rather,thislacunaeinthescholarlyliterature(withfew,importantexceptions)pointstoamoregeneralfailuretorecognizewomenashavingapotentiallyuniqueelectoralimpactduringthisperiod.
ResearchDesign
ThemethodologicalchallengeforthisresearchisdescribedbyOgburnandGoltra
(1919,413):“women’sballotsarenotdistinguishedfromthoseofmenbutaredepositedinthesameballotbox.”Invirtuallyallcases,ballotsarenotdistinguishedbythesexofthevoterwhocastthem,andthusofficialrecordscannottellushowwomenandmenvoted.Publicopinionandexitpolls,themodernsolutiontothisproblem,areunavailableorunreliableduringthisera.
Someearlyscholarscorrelatedavailableinformationaboutthegendercomposition
ofthepopulationandaggregateelectionreturnstomakeinferencesabouttheelectoralbehaviorofwomenandmen(e.g.,OgburnandGoltra1919;RiceandWilley1924).Thisprocessofinferringindividual‐levelrelationships(suchasbetweensexandturnout)fromaggregate‐leveldata(suchastheproportionofthepopulationthatisfemaleandtheproportionofthepopulationthatturnsouttovote)isknownasecologicalinference.However,Robinson(1950)explainshowthesecorrelationsarecharacterizedbywhathetermstheecologicalfallacy.Apositiverelationshipbetween,forexample,theproportionofthepopulationthatisAfrican‐Americanandtheproportionthatturnsouttovote,doesnotnecessarilymeanthatAfrican‐Americansaremorelikelytoturnoutthanothercitizens.Indeed,suchapositivecorrelationwasoftenfoundincountiesintheAmericanSouthinthefirsthalfofthe20thcentury;ratherthanindicatinghighAfrican‐Americanturnout,thatcorrelationwasdrivenbythepropensityofwhitestoturnoutathigherrateswheretheAfrican‐Americanpopulationwaslarger(Key1949).
Ecologicalinferencetodeterminegenderdifferencesinbehaviorisparticularly
challenging.Eveninsmallgeographicunits,wedonotobserveextremelyhighconcentrationsofwomenormen.Thisdistinguishesourapplicationfromthosethatfocusonotherpolitically‐relevantdivisions,suchasrace,class,orimmigration,whereresidentialsegregationresultsinhighconcentrationsofvariousgroupsinparticulargeographicareas
5
(often,ofcourse,notbychoice),permittingthedirect(orneardirect)observationofbehaviorforthatgroupinsomeespeciallyinformativegeographicunits.Thatis,observationswithextremelylopsidedconcentrationsof,e.g.,whitecitizens,produceverynarrowboundsofthepossibleelectoralbehaviorofwhites.If90%ofthepopulationinageographicunitiswhite,thelogicallypossibleturnoutandvotechoiceratesofwhitecitizensinthatunitmustbeverysimilartooverallturnoutandvotechoiceinthatunit.Givenarelativelylimitedrangeforthevariationinpercentagefemaleandgiventhetypicalconcentrationofthatpercentagearound50percent,thelogicallypossiblecombinationsofmaleandfemaleturnoutrangeacrossaverywideinterval.
Recentresearch,however,haspioneerednew,morereliableapproachesto
ecologicalinference(seeespeciallyKing1997).WeemployanapproachdescribedinWakefield(2005)whichbuildsonalargebodyofrelatedworkinbiostatisticsandepidemiology(seeRichardsonandMonfort2000)todevelopacomputationallymanageableBayesianstrategyfor2x2tables.WeextendWakefield’sapproachintwoways.First,weapplyWakefield’sapproach—developedfor2x2tables—tothemorecomplex2x4problem,estimatingDemocraticvote,Republicanvote,otherpartyvote,andabstention(fourpossibleoutcomes)formenandwomen(twopopulationgroups).Second,weintroducetheuncontroversialassumptionthatmaleturnoutwillexceedfemaleturnoutineachgeographicunitduringthisperiod.Asweshowbelow,thisapproachpermitsustogeneratereliableestimatesoffemaleandmaleturnoutandvotechoiceintheseelections.(OurmethodandestimationstrategyaredescribedingreaterdetailinCorderandWolbrechtN.d.).
Wethenemploythismethodologytoestimatetheturnoutandvotechoiceof
womenandmeninasampleofAmericanstates.Dataconstraintslimitedustoasample,ratherthanthefullpopulation.Weconstructedoursampleofstateswithtwoobjectives:(1)toobtainasmanyobservationsaspossibleineachstate(preferablyinexcessof100),and(2)toproduceasmuchcross‐statevariationinpolitically‐relevantvariablesaspossible.Dependingonthedataavailable,weobservepopulationcharacteristicsandelectoraloutcomesforvarioussub‐stategeographicunits,includingcounties,wardsofmajorcities,andMinorCivilDivisions(MCDs),thetermtheU.S.Censususestodescribethepoliticalsub‐unitsofcounties(usuallytownshipsandvillages).County‐leveldataweremadeavailablebytheInter‐universityConsortiumforPoliticalandSocialResearch(ICPSR1992,1992;seeCorderandWolbrechtN.d.forfurtherinformationonthedata).Electionreturnsandcensus(population)dataformultipleobservationsineachstateandelectionproducessufficientvariationtopermitsuccessfulestimation.Variationacrossstatesinpolitically‐relevantfactorspermitsustoexamineanumberofhypothesesregardingcontextualeffects.OurresultingsampleoftenstatesisdescribedinTable1.
6
Table1Samplestatesandcharacteristics
DateofPresidential Electoral Restrictions Women’s CollegeVote Party onState Suffrage1 Region Share(1920) Competition2 Voting3 Connecticut 1920 Northeast 1.3 One‐partyRepublican HighIllinois 1913 Midwest 5.5 One‐partyRepublican MinimalIowa 1919 Midwest 2.5 One‐partyRepublican MinimalKansas 1912 Midwest 1.9 One‐partyRepublican MinimalKentucky 1920 Border 2.5 CompetitiveDemocratic MinimalMassachusetts 1920 Northeast 3.4 One‐partyRepublican HighMinnesota 1919 Midwest 2.3 One‐partyRepublican MinimalMissouri 1919 Midwest 3.4 CompetitiveRepublican MinimalOklahoma 1918 Border 1.9 One‐partyDemocratic MinimalVirginia 1920 SolidSouth 2.3 One‐partyDemocratic High1.Source:Keyssar(2000).2.Source:Burnham(1981)for1914‐30.Seetextforcategorydefinitions.3.Source:Blakey(1928),Key(1949),andKeyssar(2000).
Ourtenstatesamplepermitsustoobservethebehaviorandimpactoffemalevotersoverafarbroaderandmorediverserangeoftimeandplacethanpreviousresearch.However,wedonotclaimthatthesampleisrepresentativeofthebroaderU.S.electorate.Oursamplewasconstrained,sometimessystematically,bythelackofelectionandcensusreturnsforasufficientnumberofsub‐stateobservationsinmanystates;suchdataareunavailablethroughthewestinthisperiod,forexample.Asaresult,oursampledoesnotfullyreflecteachregionofthecountryandbothdiffersfromandissimilartothepopulationasawholeonanumberofdimensions.(ForacomparisonofoursampleandthebroaderAmericanelectorate,seeCorderandWolbrechtN.d.).
WeemployMCMCsimulationtoproduceestimatesoffemaleandmaleturnoutfor
eachofoursamplestatesinthefivepresidentialelectionsbetween1920and1936(seeCorderandWolbrechtN.d.).Whileourestimatessatisfyanumberofdiagnostictests,thebestindicatoroftheaccuracyoftheestimator,inourview,istherecoveryoftheknown(true)Illinoisoutcomesin1916and1920.ThestateofIllinoisenfranchisedwomenforasubsetofofficesin1913,andthuswomenreceiveddifferentballotsthandidmeninthepresidentialelectionsof1916and1920(seeGoldstein1984).Thiswasnotunusualinstatesthatenfranchisedwomenpriortothe19thAmendment;whatwasunusualisthatIllinoisalsoreportedtheoutcomes—numbersofvotescastandvotechoice—separatelyformenandwomen.TheIllinoisdatathusprovideusauniqueandvaluableopportunitytovalidateourestimates.
Figure1reportstheactualandestimatedquantitiesforIllinoisin1916and1920.
Thetablerevealsboththepromiseoftheapproachandafewhazards.Despitethesignificantchallengesofecologicalinference,particularlyforsexdifferences,estimatesofturnoutandpartisanvoteshareareremarkablyclosetotheobservedreturns.Indeed,the
7
errororuncertaintyassociatedwiththeestimateiscomparabletothemarginoferrorinthesortoflargepublicopinionsurveythatformsthebasisformostcontemporaryelectionresearch(+/‐3%for1920Republicanvoteshare,forinstance).Withtheexceptionofturnout,eachoftheobservedquantitiesisinthe90%Bayesiancredibleinterval—therangeofuncertaintyassociatedwiththeestimatedparameters.2Inbothelections,estimatesofoverallturnoutareveryclosedtotheobserved,butbiasedslightlydownwardforfemaleturnoutandslightlyupwardformaleturnout.Thissmallbias,whileproblematic,doesnotpreventusfromreachinghighlyaccurateconclusionsaboutchangesinturnoutandvotechoiceacrossthetwoIllinoiselections.
Figure1
Observedandestimatedvotechoiceandturnout,Illinois,1916and1920
WomenandNewDealRealignment
Wenowturntowhatourestimatesofturnoutandvotechoicecantellusabouttheroleofwomeninthetransformativepresidentialelectionsof1932and1936.Mobilization 2The90%Bayesiancredibleintervalistherangeofestimatedvaluesin90%ofthesimulations.TheBayesianapproachpermitsustobefairlyspecificabouttheconfidencewehaveinparticularconclusionsthatwereach.Inthetextweclaimtobeconfidentofafindingofdifference(e.g.,femaleRepublicanvoteshareexceededmaleRepublicanvoteshare)if90percentormoreofthesimulationssupporttheconclusion.Wemaynoteadifferenceinwhichasmallerproportionofsimulationssupporttheconclusionsbutweacknowledgelessconfidenceinsuchresults.ThechoiceofathresholdofthistypeintheBayesiancontextisnotoriouslydifficult(seeRafterty1996),sowedisclosespecificprobabilitiesassociatedwithimportantclaims.
8
oftheelectoratein1932and1936isexceptional,particularlygiventhelong‐termtrendstowarddecliningelectoralparticipationattheturnofthecentury(seeBurnham1965).Acrossoursampleasawhole,men’sturnoutaverages75%in1932and78%in1936.Women’sturnoutaverages48%in1932(a27pointturnoutgendergap)andby1936,morethanhalfofthewomeninoursamplestates(54%)turnouttovote,a24pointturnoutgendergap.Thestate‐levelestimates,reportedinFigure2,3highlightthisremarkablemobilization.Inourtensamplestates,morethan70%ofthemalevotingagepopulationturnsouttovoteineverystatesaveone(Virginia)in1932and1936.Femaleturnoutisalsohigh:above50%inhalfofthestatesinoursamplein1932andin6of10statesin1936.Indeed,weestimatefemaleturnouttoexceed60%infourMidwesternstatesin1936,alevelofmobilizationonpartwithhighturnoutelectionsintheU.S.today.
Figure2
Turnoutofwomenandmen,1932and1936
1932
3Becauseregionandpartycontextarecentraltoouranalyseselsewhere,weorganizestatesinthefiguresbythosecharacteristics:Westartwithourone‐partyDemocraticSouthernstate(Virginia)onthefarleft,thenreportonourtwoBorderstates,withtheone‐partyDemocraticstate(Oklahoma)followedbythecompetitive(Democratic‐leaning)state(Kentucky).WethenmoveontoourMidwesternstates,startingwiththeonecompetitive(Republican‐leaning)state(Missouri)andthenthefourone‐partyRepublicanstates(Kansas,Iowa,Minnesota,andIllinois).Finally,tothefarrightwepresentourtwoone‐partyRepublicanstatesintheNortheast(ConnecticutandMassachusetts).Bothregionandourpartycontextmeasureareindicated,alongwithstatelabel,onthex‐axis.
9
1936
Figure3showsthat,acrossoursampleasawhole,Democratswerehugely
successfulin1932and1936—particularlyinlightofRepublicandominanceinthe1920s.Intheelectionsleadingupto1932,weobservedapersistentRepublicanadvantageamongwomeninone‐partyRepublicanstatesintheMidwest.ThisdifferenceislargelyresponsibleforthesmallRepublicanadvantageamongwomeninoursampleasawholeinthefirstthreepresidentialelections,1920through1928.Theelectoraldisruptionsof1932and1936erodethisadvantage—andperhapsevenreverseitslightly—withtheendresultthatwomen’sandmen’ssupportforDemocraticnomineesettlesataboutthesamelevelby1936.Specifically,Democraticsupportinboth1932and1936averages57%amongmencomparedto60%amongwomeninourtensamplestates.
10
Figure3Democraticvoteshareofwomenandmen,1932and1936
1932
1936
11
Thefiguressofarhavegivenusageneralsenseoftheturnoutandvotechoiceofwomenandmenintheseelections.WhatissignificantabouttheseelectionsisthedramaticshiftfromaconsiderableRepublicanadvantageamongtheelectorateinelectionspriortothe1930stoaDemocraticadvantagebeginningin1932.Howdidthatcomeabout,andwhatroledidwomenplay?Wecanbestanswerthatquestionwithourestimatesbyexaminingthenumberofballotsgainedandlostbetweenelections.Aswehavediscussed,thecontributionsofmobilizationandconversiontoNewDealrealignmenthavebeenofkeeninteresttoscholars.Theprocessesappeartovaryacrossthesetwoelections—withlittleturnoutchangeandlargeDemocraticswings(conversion)in1932,andsignificantnewmobilizationandlittlechangeinpartisandistribution(mobilization)in1936—sowediscusstheoverallandpartisanmobilizationpatternsinthe1932and1936electionsinsequence.
TheElectionof1932
Figure4summarizesthetotalnumberoffemaleandmalevotersaddedtothe
electoratein1932,providinganindicationofthesizeofthemobilizationofmenandwomen.WiththeexceptionofVirginia,thesizeoftheactiveelectorateexpandedineverystate,althoughsometimesonlyveryslightly,over1928.Giventheextraordinarylevelofnewvotermobilizationin1928,bothmaintenanceandespeciallyexpansionofthesizeoftheactiveelectorateshouldbeconsideredanimpressivefeat.InahandfulofstatesintheMidwest(i.e.,Missouri,Kansas,andIllinois),weestimatethatnearlytwiceasmanynewfemalevotersareaddedtotherollsasmalevotersin1932,althoughonlyinIllinoisaremorethan90%ofthesimulationsconsistentwiththeconclusionthatfemalemobilizationoutpacedmalemobilizationin1932.Intheotherstates,thenumberofnewvotersiseithersimilarortherearejustslightlymorenewfemalevoters;wecanonlyconcludewithconfidencethatthesestatesaddedsimilarnumbersofnewmaleandfemalevoters.MassachusettsandConnecticut,bothofwhichexperiencedadramaticexpansionoftheelectoratein1928,standoutasstateswithalmostnonewvotersin1932;hereagain,thesimulationssuggestthisconclusionholdsforbothmaleandfemalevoters.Thus,withtheexceptionofIllinois,weestimatethatroughlyequalnumbersofnewmenandwomenenteredtheelectoratein1932.
12
Figure4Changeinnumberofvotescastbywomenandmen,1928to1932
DemocratsovercameaconsiderableRepublicanadvantagetobecomethemajority
partyin1932.Figure5displaysthechangeinthenumberofvotescastfortheDemocraticpresidentialnomineebetween1928and1932.OutsideoftheNortheast,thenumberofDemocraticvotesgainedisextraordinaryineverysamplestate.Innearlyallofthestates,menaccountforalargernumberofnewDemocraticvotersthandowomenbutinmostcases,thedifferencesaresmall.Illinoisistheexception:Nearly80%ofthesimulationsindicatethatwomenaccountedformorenewDemocraticvotersthandidmeninIllinoisin1932despitethefactthat,asineverystate,women’sturnoutlaggedthatofmen.Therearetwoothersstatesinwhichgenderdifferencesdoappear,bothintheSouthern/Borderregion:InOklahomaandVirginiamorethan90%ofthesimulationsareconsistentwiththeconclusionthatDemocratsgainedmorenewmalethanfemalevoters.ThesefindingsareconsistentwithevidencethatNewDealrealignmentunfoldedinuniquelylocalways,andthattheexperienceandcontributionsofwomenwereasmuchafunctionoftheirlocalcontextasthefactoftheirgender.
13
Figure5ChangeinnumberofDemocraticvotescastbywomenandmen,1928to1932
Notsurprisingly,thedeclineinRepublicanballotsineachstate—summarizedinFigure6—tellsthesamestory.ThedeclineinthelevelofsupportfortheRepublicancandidatewasroughlysimilaramongmenandwomeninmostsamplestates.InOklahomaandVirginia(whereDemocratsgainedmorementhanwomen),Republicanlossesamongmenweremuchlargerthanlossesamongwomen,aconclusionsupportedbymorethan90%ofthesimulations.Morethan70%ofthesimulationssupportthesameconclusioninKansasandMinnesota.Wethusarebeginningtoseesomeevidenceconsistentwithdifferingpatternsofmobilization(women)andconversion(men)asexplanationsforpartisanchangeamongmenandwomenin1932.
14
Figure6ChangeinnumberofRepublicanvotescastbywomenandmen,1928to1932
WerewomenandmenmobilizedasorconvertedtoDemocratsinthesamewayin
moststatesorbydistinctpaths?Ourestimatesdonotpermitustosaywithcertaintyhownewvoterscasttheirballots,orwhethervoterslostbyRepublicansin1932stayedhomeorcasttheirballotforDemocrats.Wecan,however,makereasonableinferences(withappropriatecaution)fromthepatternsobserved.Overall,ourgeneralconclusionisclear:WiththeexceptionoftheNortheasternstatesandIllinois,turnoutgainsaremodestornon‐existentin1932whiletheshareofvoteswonbyDemocratsincreasesdramatically,suggestingasignificantportionofthosenewDemocraticvoteslikelycamefromvoterswhocasttheirballotsfortheRepublicancandidatein1928;inotherwords,weconclude,asothershave,thatconversionwaslikelythedominantmechanismoverallin1932(cf.,Brown1988;HawleyandSagarzazu2012).
Theminimumnumberofconvertsrequiredtogeneratetheobservedlevelof
Democraticvotes4ineachstatein1932issummarizedinFigure7.ThenumberreportedinthefigureisthenumberofnewmaleorfemaleDemocraticvotersminusthenumberofnewmaleorfemalevoters.Thedifferencebetweenthesenumbersistheminimumnumberofconverts(previousRepublicanvoters)whomusthaveswitchedpartisanvotechoicetoaccountforobservedsupportfortheDemocraticcandidatein1932.SincethenumberofadditionalvotesforDemocratsissolargerelativetothenumberofnewvotersinevery
4ThefiguresummarizesdefectionsfromtheRepublicanPartyin1932,whichcouldresultinincreasingsupportfortheDemocraticcandidateorathirdpartycandidate.Thirdpartysupportisfairlylowinthesamplestatesin1932,about2.5%,sotheoverwhelmingnumberofRepublicanconvertssupportedtheDemocraticcandidate.
15
state,itisextremelylikelythatthereweremanyconvertsfromtheRepublicantotheDemocraticparty.(ThisisanestimatebasedontheassumptionthatvirtuallyallnewvoterswereDemocraticsupporters.IfsomenewvoterssupportedtheRepublicans,thenumberofrequiredconvertswouldbehigher.IfsomeestablishedvotersshiftedfromtheDemocratstoorthirdpartiestotheRepublicanparty,thenumberofrequiredconvertsfrom1928Republicanswouldbehigher.Ontheotherhand,ifmanyRepublicanvotersfrom1928stayedhomeandthesevoterswerereplacedwithnewDemocraticvoters,thentheconversionnumberswouldbelower).Whiledifferencesinmanystatesaresmall,inmoststatesthereweremoremaleconverts(1928RepublicanvoterswhocastDemocraticorthird‐partyballotsin1932)thantherewerefemale.
Figure7
Minimum*numberofRepublicanconverts,1928to1932
*AssumesnonewvotersarecapturedbytheRepublicans
Overall,giventheestimatednumberofnewfemaleandmalevotersin1932inoursample(411,000newwomenand290,000newmen),andacceptingthecrudeassumptionthatnonewvotersareRepublicans,weestimatethat,acrossourtensamplestates,about656,000womenwhovotedRepublicanin1928votedforFDRin1932,andapproximately828,000menwhohadvotedRepublicanin1928castDemocraticballotsfouryearslater.Asaresult,conversionaccountsforabout62%ofthe1.06millionvotesgainedbyDemocratsamongwomenin1932inthesamplestates,andperhaps75%ofthe1.11millionvotesgainedbyDemocratsamongmen.
16
Thus,conversiongenerallyappearstoaccountformore,andmobilizationrelativelyless,ofthenewDemocraticvotescastbymenthanbywomen.Thisisnotparticularlysurprising;thelargernumberofmenalreadyvotingin1928impliesthatmoremenwereavailableforconversionand,lessso,mobilization.Whatmightbesurprisingisthattherelativelylargernumberofmaleconvertsoccursinspiteofthefactthatwomenaremorelikelytodefect.Thatis,womenwhovotedRepublicanin1928weremorelikelytocastDemocraticballots(defect)in1932thanweremenwhovotedRepublicanin1928.Theminimumnumberofconverts(fromabove)asapercentageoftheavailable1928Republicanelectorategivesasenseoftherateofdefection.Forexample,inMinnesota,weestimate89,000ofthe213,000womenwhocastRepublicanballotsin1928didnotdosoin1932,foradefectionrateof41%amongwomen.Incontrast,weestimate108,000ofthe347,000menwhocastRepublicanballotsin1928didnotdosoin1932,producingamaledefectionrateof31%.ConsistentwiththeestimatesfromMinnesota,women’srateofdefectionexceedsmen’sineverystate.Insomestates,likeOklahomaandIllinois,thegenderdifferencesindefectionratesarequitesmall.Inotherstates,likeMinnesotaandKentucky,thedefectionrateforwomenwasclearlyhigher.Sinceeachofthenumbersinthecalculationisestimatedwithuncertainty,theconfidenceintervalsforthesequantitiesarewide.Summingacrossallofthesamplestates,about80%ofthesimulationsareconsistentwiththeconclusionthatfemaledefectionratesexceededmaledefectionratesinoursamplestates.Inonestate,Minnesota,over85%ofthesimulationsareconsistentwiththeconclusionthatthepercentageofRepublicandefectorswashigheramongwomenthanmen.Therearenostateswhereevenmorethan50%ofthesimulationsareconsistentwithahigherminimumdefectionrateamongmen.
17
Figure8Minimum*percentageof1928Republicanballotsconverted(defectionrate),1928
to1932
*AssumesnonewvotersarecapturedbytheRepublicans
Thus,whilein1932mobilizationwasresponsibleforabiggerportionofnewDemocraticballotscastbywomen,thanitwasfornewDemocraticballotscastbymen,thepercentageofwomenwhochangedtheirvotechoicefrom1928to1932(defection)waslikelylargerthanthesimilarpercentageamongmen.Onceagain,thisfindinghighlightshowourconclusionsareshapedbythedataandmeasuresweexamine,andparticularlytheinteractionofturnoutandvotechoice.WhilemoreoftheDemocraticgainsamongmenwereattributabletoconversionthanmobilization,manyfewerwomenvotedin1928.Asaresult,evenwithfewerconvertsin1932overall,thelikelihoodthata1928voterchangedhervotefromRepublicantoDemocraticin1932waslikelymarginallyhigheramongwomenthanmen.Substantively,wenotethatthegreaterlikelihoodthatwomenwoulddefectin1932isconsistentwithclaimsthatwomen’slesserelectoralexperiencewouldmeanweakerpartisanties(e.g.,Converse1969,1976).
TheElectionof1936
In1936,themobilizationofnewvotersisclearlymoreextensivethanin1932and
inmoststatestherearemorewomenaddedtotheelectoratethanmen.Overall,acrossourtensamplestates,868,000morewomenparticipatein1936thanin1932,and570,000moremen.Figure9showsthatthispatterngenerallyholdsatthestatelevel;innearlyallofofoursamplestates,therearemorenewfemalevotersthanmale.Wecanbeconfidentof
18
thesedifferencesinIllinois,wheremorethan90%ofthesimulationssupporttheconclusionthatfemalemobilizationwaslarger,andsomewhatconfidentinConnecticutandKansas(wheremorethan80%ofthesimulationsshowmorenewfemalevotersthanmale.(Inanotherthreestates—Iowa,Missouri,andOklahoma—75%ormoreofthesimulationsarealsoconsistentwiththisconclusion).Thus,inmoststates,theexpectationthatthelargernumbersofinactivewomenavailableformobilizationwouldmeanwomenwereresponsibleformorenewballotsthanmenduringtheNewDealperiodisconfirmedinanumberofstatesin1936,whennewmobilizationwasconsiderable.
Figure9
Changeinnumberofvotescastbywomenandmen,1932to1936
WiththeexceptionofKentucky,Democratstendedtogainvotersin1936(seeFigure10).MoreofthenewDemocraticvotescomefromwomenthanmenineightofourtenstates,althoughagain,differencesaresmallinanumberofcases.Sincetheuncertaintyassociatedwiththeseestimatesarelarge,wecanonlybeconfidentaboutthesedifferencesinConnecticut;over90%ofthesimulationsindicatethatDemocratspickedupmorevotesfromwomenthanfrommen.Thelargemobilizationofvotersin1936generatedgainsfortheRepublicanpartyaswell(seeFigure11).Thedifferencesbetweenmenandwomenaresmall,however,soouroverallconclusionisthatin1936,inplaceswhereRepublicansgainedmalevotes,theygainedasimilarnumberoffemalevotes.5
5Massachusettsstandsoutasanexception,butthesimulationfailedtoconvergeforMassachusetts,1936,sowehavelessconfidenceinthatestimatesassociatedwiththatelection.
19
Figure10ChangeinnumberofDemocraticvotescastbywomenandmen,1932to1936
Figure11ChangeinnumberofRepublicanvotescastbywomenandmen,1932to1936
20
Wehaveahardertimeinferringthemechanismsfortheallocationofpartisanvotesgainedin1936,comparedto1932,sincethenumberofnewentrantstotheelectoratewasquitelarge(over1.4million),butthegainsforDemocratsweresmaller(900,000),andRepublicansexperiencedgainsaswell.Asaresult,avarietyofpossiblemechanismscouldbeatworkbetween1932and1936:MobilizationofRepublicans,mobilizationofDemocrats,conversionfromDemocratictoRepublican,and/orconversionfromRepublicantoDemocrat.Itisnotpossibleforustosaywithmuchconfidencewhichmechanisms,andtowhatextent,explainelectoralchangein1936,orifdifferentmechanismscharacterizemenandwomen.Althoughtheyremainpossible,theestimates,intheaggregate,revealnostrikinggenderdifferencesinthesedynamics,exceptfortheevidenceofgreaterturnoutgainsamongwomencomparedtomen.
TheContributionofWomentoNewDealRealignment
Overall,wefindmenandwomencontributedinroughlyequalnumbers,butindifferentways,toNewDealrealignment.Inboth1932and1936wefindevidencethatthenumberofnewfemalevotersenteringthesystemexceededthenumberofnewmalevoters:Inoursample,120,000morewomenthanmenenteredtheelectoratein1932(almostexclusivelyattributabletoIllinois)and300,000morewomenthanmenenteredtheelectoratein1936.Aspreviousscholarshaveexpected(e.g.,Andersen1979),women’sunder‐mobilizationrelativetomentranslatedintomorenewfemalethanmalevotersduringtheNewDealrealignmentperiod.Atthesametime,wefindthatwomenintheelectoratepriortothe1930sweresomewhatmorelikelytodefectfromtheirpre‐NewDealpartisanship(i.e.,womenwhovotedRepublicanin1928weremorelikelytovoteDemocraticin1932thanweremenwhovotedRepublicanin1928),consistentwiththeexpectationthatpartisanshipwouldbelessstableamongfemalevoterswhohadbeenintheactiveelectorateforashorterperiod(e.g.,Andersen1979,Converse1969,1976).However,becausetherewerefewerwomenintheactiveelectorateoverall,mobilizationaccountedformorenewDemocraticvotesamongwomenthanitdidamongmen.
Despitethefactthatmorenewwomenthannewmenweremobilizedinthese
elections,thenumberofnewDemocraticvotesoverallthatcanbeattributedtowomenisnotsubstantiallydifferentthanthenumberofnewDemocraticvotesthatcanbeattributedtomen:Inourtensamplestates,weestimatethatjustoveronemillionfemaleDemocraticvoteswereadded(1.07million)in1932andalittlemorethanonemillionadditionalmaleDemocraticvotes(1.11million),plusabout550,000additionalfemaleDemocraticvotesandabout320,000additionalmaleDemocraticvotesin1936.Whilemorenewfemalethanmalevotersenteredtheactiveelectoratein1932and1936,morenewmaleDemocraticvotesweregeneratedbyconversion(menwerelesslikelytodefect,buttheirdefectionscamefromalargerbaseofestablishedvotersandthusgeneratedmoreDemocraticballots).Thus,thecombinedeffectsofmobilizationandconversionallowedtheDemocratstodrawnearlyequalnumbersofnewvotersfromtheranksofthemaleandfemaleelectoratein1932and1936:1.4millionadditionalvotesfrommenand1.6millionadditionalvotesfromwomen.
21
ConclusionsConventionalnarrativesofNewDealrealignmentoverlooktheimpactofnew
femalevotersonthefortunesoftheDemocrats.Bothmechanismsofrealignment—mobilizationandconversion—appeartohavecharacterizedwomen’scontributiontoNewDealrealignment:Duetolowerturnout,manywomenwereavailableformobilization,andthosewhohadvotedpreviously(mostlyforRepublicans)lackedthereinforcementofpartisanshipofferedbylong‐termpoliticalparticipation.Andindeed,manywomeneitherconvertedfromRepublicansupport(particularlyin1932)orwerenewly‐mobilizedbyDemocrats(especiallyin1936).
Interestingly,althoughtheshiftinsupporttotheDemocrats—measuredasthe
changeinDemocraticvoteshare—wassmalleramongmen(in1932),themaleelectoratewaslargersothenetresultisthatDemocraticgainscameinroughlyequalnumbersfrommenandwomen—about1.6millionadditionalwomenand1.4millionadditionalmenacrossbothelections.Inourtensamplestatesover4.5millionmenand3.3millionwomensupportedtheDemocraticcandidateRooseveltin1936.GiventhatDemocraticcandidateSmithreceived3millionvotesfrommenandonly1.7millionvotesfromwomeninthesamestatesin1928,thejointimpactofmobilizationandconversiononthefemaleelectorateacrossthe1932‐1936periodisclearlyverylarge.By1936,Democratshadattractedanimpressive95%morefemalevoters(from1.7millionin1928to3.3millionin1936)comparedtoabout50%moremalevoters(from3.0millionto4.5million).
AccountsofNewDealrealignmentthatignorethegradualandpersistentincreasein
women’sturnoutmissanimportantelementofmobilizationthataccountsforahugenumberofnewvotersinthe1930s.Insomestates,womenwereclearlywellincorporatedintoelectoralpoliticsbytheendoftherealignmentera—therateofturnoutforwomeninseveralsamplestateswasnearly65%by1936.Thislevelofmobilizationwouldbeconsideredremarkableinourcurrentera.
Womenshapedthesize,pattern,andoutcomeofNewDealrealignment.The
massivechangeintheelectoralfortunesoftheDemocraticpartyinthe1930swasdue,inroughlyequalmeasure,totheballotsofmenandwomen.Only16yearsaftertheextensionofsuffragetowomenandafterearlyaccountsdismissedwomen’ssuffrageasadisappointmentorfailure,womenwerealargeandincreasinglyimportantsourceofelectoralsupportfortheemergingDemocraticmajority.
22
References
Andersen,Kristi.2014.“ConstructingaNewMajority:TheDepression,theNewDeal,andtheDemocrats.”InTheCQGuidetoU.S.PoliticalParties,eds.MajorieR.Hershey,BarryC.Burden,andChristinaWolbrecht.Washington,D.C.:CongressionalQuarterlyPress.
Andersen,Kristi.1979.TheCreationofaDemocraticMajority,1928‐1936.Chicago:TheUniversityofChicagoPress.
Berelson,BernardR.,PaulF.Lazarsfeld,andWilliamN.McPhee.1954.Voting:AStudyofOpinionFormationinaPresidentialElection.Chicago:UniversityofChicagoPress.
Blakey,GladysC.1928.AHandyDigestofElectionLaws.Washington,D.C.:LeagueofWomenVoters.
Brown,Courtney.1988.“MassDynamicsofU.S.PresidentialCompetitions,1928‐1936.”AmericanPoliticalScienceReview82(December):1153‐81.
Burnham,WalterDean.1970.CriticalElectionsandtheMainspringsofAmericanPolitics.NewYork:Norton.
Burnham,WalterDean.1974.“TheoryandVotingResearch:SomeReflectionsonConverse’s‘ChangeintheAmericanElectorate.’”AmericanPoliticalScienceReview68(September):1002‐23.
Burnham,WalterDean.1981.“TheSystemof1896:AnAnalysis.”InTheEvolutionofAmericanElectoralSystems,eds.PaulKleppner,WalterDeanBurnham,RonaldP.Formisano,SamuelP.Hays,RichardJensen,andWilliamG.Shade.WestportCT:GreenwoodPress.
Corder,J.KevinandChristinaWolbrecht.N.d.CountingWomen’sBallots:FemaleVoteChoicefromSuffrageThroughtheNewDeal.Manuscript.
Campbell,Angus,PhilipE.Converse,WarrenE.Miller,andDonaldE.Stokes.1960.TheAmericanVoter.NewYork:JohnWileyandSons,Inc.
Erikson,RobertS.andKentL.Tedin.1981.“The1928‐1936PartisanRealignment:TheCasefortheConversionHypothesis.”AmericanPoliticalScienceReview75:951‐62.
Gamm,Gerald.1989.TheMakingoftheNewDealDemocrats:VotingBehaviorandRealignmentinBoston,1920‐1940.Chicago:UniversityofChicagoPress.
Goldstein,JoelH.1984.TheEffectsoftheAdoptionofWomanSuffrage:SexDifferencesinVotingBehavior—Illinois,1914‐21.NewYork:Praeger.
Gourevitch,PeterAlexis.1984.“BreakingwithOrthodoxy:ThePoliticsofEconomicPolicyResponsestotheDepressionofthe1930s.”InternationalOrganization38(Winter):95‐129.
Gosnell,HaroldF.1942.GrassrootsPolitics:NationalVotingBehaviorofTypicalStates.Washington,D.C.:AmericanCouncilonPublicAffairs.
Gosnell,HaroldF.1930.WhyEuropeVotes.Chicago:UniversityofChicagoPress.
23
Green,DonaldP.,BradleyPalmquist,andEricShickler.2002.PartisanHeartsandMinds:PoliticalPartiesandtheSocialIdentityofVoters.NewHaven:YaleUniversityPress.
Hawley,GeorgeandInakiSagarazu.2012.“Wheredidthevotesgo?ReassessingAmericanpartyrealignmentsviavotetransferesbetweenmajorpartiesfrom1860to2008.”ElectoralStudies31:726‐39.
Johnston,Richard.2006.“PartyIdentification:UnmovedMoverorSumofPreferences?”AnnualReviewofPoliticalScience9:329‐51.
Key,V.O.,Jr.1949.SouthernPoliticsinStateandNation.NewYork:AlfredA.Knopf.
Key,V.O.,Jr.1955.“ATheoryofCriticalElections.”JournalofPolitics17(February):3‐18.
Keyssar,Alexander.2000.TheRighttoVote:TheContestedHistoryofDemocracyintheUnitedStates.NewYork:BasicBooks.
King,Gary.1997.ASolutiontotheEcologicalInferenceProblem:DeconstructingIndividualBehaviorfromAggregateData.Princeton,NJ:PrincetonUniversityPress.
Kleppner,Paul.1982.WhoVoted?TheDynamicsofElectoralTurnout,1840‐1940.NewYork:Praeger.
Ladd,EverettC.andCharlesD.Hadley.1975.TransformationoftheAmericanPartySystem:PoliticalCoalitionsfromtheNewDealtothe1970’s.NewYork:Norton.
Lazarsfeld,PaulR.,Berelson,BernardR.,andHazelGaudet.1948.ThePeople’sChoice.NewYork:ColumbiaUniversityPress.
McPhee,WilliamN.andJackFerguson.1962.”PoliticalImmunization.”InPublicOpinionandCongressionalElections,eds.WilliamN.McPheeandWilliamA.Glaser.NewYork:FreePressofGlencoe.
Nardulli,PeterF.1995.“TheConceptofCriticalRealignment,ElectoralBehavior,andPoliticalChange.”AmericanPoliticalScienceReview89(March):10‐22.
Ogburn,WilliamF.andInezGoltra.1919.“HowWomenVote:AStudyofAnElectioninPortland,Oregon.”PoliticalScienceQuarterly34:413‐33.
Prindle,DavidF.1979.“VoterTurnout,CriticalElections,andtheNewDealRealignment.”SocialScienceHistoryIII(Winter):144‐70.
Raftery,AdrianE.1995.“BayesianModelSelectioninSocialResearch.”SociologicalMethodology25:111‐163.
Rice,StuartA.andMalcolmM.Willey.1924.“AmericanWomen’sIneffectiveUseoftheVote.”CurrentHistory20(July):641‐47.
Richardson,S.andC.Monfort.2000.“EcologicalCorrelationStudies.”InSpatialEpidemiology:MethodsandApplications,eds.P.Elliott,J.Wakefield,N.G.Best,andD.J.Briggs.Oxford:OxfordUniversityPress.
Robinson,WilliamS.1950.“EcologicalCorrelationandtheBehaviorofIndividuals.”AmericanSociologicalReview15:351‐57.
24
Salisbury,RobertH.andMichaelMacKuen.1981.“OntheStudyofPartyRealignment.”JournalofPolitics43:523‐530.
Sundquist,JamesL.1983.DynamicsofthePartySystem:AlignmentandRealignmentintheUnitedStates.RevisedEdition.Washington,D.C.:TheBrookingsInstitution.
Wakefield,Jon.2004.“Ecologicalinferencefor2×2tables(withdiscussion).”JournaloftheRoyalStatisticalSociety167:385‐445.
Wanat,John.1979.“TheApplicationofNon‐Analytic,MostPossibleEstimationTechnique:TheRelativeImpactofMobilizationandConversionofVotesintheNewDeal.”PoliticalMethodology6(3):357‐74.