384SocialScienceHistoryBullock,CharlesS.,III(1988) - PDF document

384SocialScienceHistoryBullock,CharlesS.,III(1988)Regionalrealignmentfromanoceholdingperspective.JournalofPolitics50:553…74.Burnham,WalterDean(1967)Partysystemsandthepoliticalprocess,inWilliamNisbetChambersandWalterDeanBurnham(eds.)TheAmericanPartySystems:StagesofPoliticalDevelopment.NewYork:OxfordUniversityPress:277…307.(1970)CriticalElectionsandtheMainspringsofAmericanPolitics.NewYork:Norton.(1991)Criticalrealignment:Deadoralive?inByronE.Shafer(ed.)TheEndofRealignment?InterpretingAmericanElectoralEras.Madison:UniversityofWis-consinPress:101…39.Burnham,WalterDean,JeromeM.Clubb,andWilliamH.Flanigan(1978)Partisanrealignment:Asystemicperspective,inJoelH.Silbey,AllanG.Bogue,andWil-liamH.Flanigan(eds.)TheHistoryofAmericanElectoralBehavior.Princeton,NJ:PrincetonUniversityPress:45…77.Campbell,Angus(1966)Aclassi“cationofthepresidentialelections,inAngusCamp-bell,PhilipE.Converse,WarrenE.Miller,andDonaldE.Stokes(eds.)ElectionsandthePoliticalOrder.NewYork:Wiley:63…77.Campbell,JamesE.(1997)ThePresidentialPulseofCongressionalElections.2nded.Lexington:UniversityPressofKentucky.(2000)TheAmericanCampaign:U.S.PresidentialCampaignsandtheNationalVote.CollegeStation:TexasA&MUniversityPress.Carmines,EdwardG.,andJamesA.Stimson(1989)IssueEvolution:RaceandtheTrans-formationofAmericanPolitics.Princeton,NJ:PrincetonUniversityPress.Chubb,JohnE.(1978)Systemsanalysisandpartisanrealignment.SocialScienceHis-tory2:144…71.Clubb,JeromeM.,WilliamH.Flanigan,andNancyH.Zingale(1980)PartisanRealign-ment:Voters,Parties,andGovernmentinAmericanHistory.BeverlyHills,CA:Sage.Converse,PhilipE.(1966)Theconceptofthenormalvote,inAngusCampbell,PhilipE.Converse,WarrenE.Miller,andDonaldE.Stokes(eds.)ElectionsandthePoliticalOrder.NewYork:Wiley:9…39.Ferejohn,JohnA.,andMorrisP.Fiorina(1985)Incumbencyandrealignmentincongres-sionalelections,inJohnE.ChubbandPaulE.Peterson(eds.)TheNewDirectioninAmericanPolitics.Washington,DC:BrookingsInstitution:91…115.Gans,DanielJ.(1985)PersistenceofpartysuccessinAmericanpresidentialelections.JournalofInterdisciplinaryHistory16:221…37.Glad,PaulW.(1964)McKinley,Bryan,andthePeople.NewYork:Lippincott.Hetherington,MarcJ.(2001)Resurgentmasspartisanship:Theroleofelitepolariza-tion.AmericanPoliticalScienceReview95:619…31.Jacobson,GaryC.(2000)Reversaloffortune:ThetransformationofU.S.Houseelec-tionsinthe1990s,inDavidW.Brady,JohnF.Cogan,andMorrisP.Fiorina(eds.)ContinuityandChangeinHouseElections.Stanford,CA:StanfordUniversityPress:10…38. PartySystemsandRealignments3837Mayhew(2002:47,51)alsosuggests1920asanalternativetothe1894…96realign-ment.Sundquist(1983:182)termsthe1920saminorrealignment.The1920elec-tionwasexaminedasthestartofthispro-Republicanrealignment(insteadof1896).Itfareswellinthepresidentialvoteanalysis(asinBartels1998:289).Republicanpresidentialcandidateswoninlandslidesin1920(WarrenG.Harding),1924(CalvinCoolidge),and1928(HerbertHoover).However,the1920startisnotstatisticallysigni“cantinthecongressionalanalysis(whetherornot1874isalsoincludedasarealignment).Withoutaplausiblereasonforadelayoftherealignmentatthepresi-dentiallevel,1896remainsthemostlikelystartofthisrealignmentasitremainstheonlyelectioninthisperiodwithasigni“cantrealignmenteectinbothpresidentialandcongressionalelections.8Arobustregressionofthecongressionalseatequationincludingsurgeanddeclineandpartysystemvariablesfor1874…92,1894…1928,1930…92,and1994…2004pro-ducedanadjustedof.68.Eachcoecientwassigni“cant(.01,one-tailed).Theestimatedeectswere0.30forsurgeanddecline,20.10forthe1874…92sys-tem,9.80forthe1894…1928system,23.66forthe1930…92system,and12.88forthe1994…2004system.Theconstant(thebaselineof1868…72)was35.00.9DwightD.Eisenhowers1952and1956presidentialelections,alongwithWoodrowWilsons1912and1916elections,areclassicdeviatingelections(Campbell1966:69…74).ControllingforEisenhowersdeviatingelections,the1968realignmentstart“tsthedatabetterthanthe1948alternative,analternativesuggestedbyMayhew(2002:146).WithavariableforEisenhowersdeviatingelections,theequationwitha1948realignment(a1948…2004partysystem)hasanadjustedof.38.Asimilarspeci-“cationwiththe1968onsethasanadjustedof.45.ReferencesAbramowitz,Alan,andKyleL.Saunders(1998)IdeologicalrealignmentintheU.S.electorate.JournalofPolitics60:634…52.Aldrich,JohnH.,andRichardG.Niemi(1996)ThesixthAmericanpartysystem:Elec-toralchange,1952…1992,inStephenC.Craig(ed.)BrokenContract?ChangingRelationshipsbetweenAmericansandTheirGovernment.Boulder,CO:Westview:Bartels,LarryM.(1998)Electoralcontinuityandchange,1868…1996.ElectoralStudies17:275…300.(2000)Partisanshipandvotingbehavior,1952…1996.AmericanJournalofPoliti-calScience44:35…50.Beck,PaulAllen(1999)ThechangingAmericanpartycoalitions,inJohnC.GreenandDanielM.Shea(eds.)TheStateoftheParties:TheChangingRoleofContemporaryAmericanParties.3rded.Lanham,MD:RowmanandLittle“eld:28…49.Black,Earl,andMerleBlack(2002)TheRiseofSouthernRepublicans.Cambridge,MA:HarvardUniversityPress. PartySystemsandRealignments381realignmentsoccurred,andthepossibilitythatthesuspectedrealignmentswereinsteaddeviatingelections.Theystandoutfromotherelections,includ-ingallofthealternativesraisedbyMayhew.Whywerentthesefourrealign-mentsclearinthestudiesthatMayhewexamined?Insomecases,deviatingelections,theinclusionofanaberrantin”uencepointelectionsuchas1912,orshort-term”uctuationsofsurgeanddeclinemayhavemaskedlong-termpartisanchange.Inothercases,thefactthatarealignmentwasnotobviousmayre”ectMayhewspointthatrealignmentsarenotbinaryevents.Indeed,thisanalysishasidenti“edtheonsetsofrealignmentsthatmayhavetakenseveralelectionstoemergeandseveralmoretocomplete.Thattherealign-mentprocessisnotinstantaneousandthatthismaycreatesomeuncertaintyastowhenexactlyarealignmentbeganshouldnotbeconstruedasevidencethatrealignmentshavenotoccurred.Theevidenceisthattheyhave.ThesoundnessofMayhewsoverallcritiqueoftherealignmentgenreisdiculttoassess.Realignmentsareonlyoneofmanycomponentsofafullappreciationofwhathastranspiredandwhy.Likeanysingleaspectofelec-toralhistory,theimportanceofrealignmentscanbeexaggeratedtotheexclu-sionofotherimportantqualities.Mayhewmayalsoberightaboutsomeorallofhisassessmentsoftherealignmentgenresclaims(2…14)regardingthecausesandconsequencesofrealignments„theseissuesarebeyondthescopeofthisstudy.Thatsaid,Mayhewiswronginconcludingthatrealignmentshavenotoccurredandthatthe1896realignment,inparticular,didnottakeplace.Thisanalysishasnotdelvedintothemanyrami“cationsthatrealign-mentsmayormaynothave,butitwouldseemobviousthatanunderstandingofU.S.electoralhistoryisseverelyimpededifitlacksanappreciationoftheveryreallong-termshiftsinthebalanceofpowerbetweenthemajorpartiesinpresidentialandcongressionalelections.1Foranexcellenthistoryofrealignmentresearch,seeRosenof2003.2Mayhew(2002:50…55)reliesheavilyonBartelss(1998:287…91)indirectassess-mentofcriticalelectionsusingstate-levelpresidentialreturns.Besidestheindi-rectionofthemethodandtheomissionofnonpresidentialdata,Bartelssanalysisraisesseveralconcerns:(1)itcountsstatesequallydespitepopulationandturnoutdierences;(2)realigningeectsarerestrictedtosingleelections;(3)subnationaleects(anelectionsstochasticerror)andpotentiallyshort-termeects(identi“edasnationalforces)arecountedincalculatingthelong-termimpactofanelection; PartySystemsandRealignments379Thecongressionalequationsfor1994intable3displayapatternsimilartothatofthe1930realignment.The1994startpointforthenewpartysys-temisthestrongestspeci“cation,butthefouralternativesleadinguptoit(1978,1982,1986,and1990)alsoindicateapartysystemstatisticallydierentfromtheNewDealera.Aretheseearlierelectionsrivalstartdatesto1994,ordotheysimplyre”ecttheoverlapofelectionsintheseriesandthelargedif-ferencebetweenpost-1994electionsandthosethatcamebefore?Transitionvariablesareexaminedtotestwhethertheseelectionswerepartofatransi-tionbetweenpartysystems.Thesearedummyvariablesforelectionsfrom1978to1992,from1982to1992,from1986to1992,andin1990and1992.Wheneachofthesetransitionvariablesistestedwiththe1994…2004partysystemvariable,noneremotelyapproachesstatisticalsigni“cance(.05).Itisalsoclearfromtable3thatarealignmenttookplaceatsomepointinthisperiod(thenonrealignmentspeci“cationfails),thattheNewDealsystemdidnotsurvive,andthat1994wasnotsimplyadeviatingelection.The1994electionwasabreakthroughforcongressionalRepublicans.TheydidespeciallywellintheSouth,wheretheyhadbeenstrongsincethe1960sinpresidentialvotingandwheretheyhadmadesomestridesatthecongressionallevelandbelow,asnotedabovewithrespecttostatelegisla-tivegains.IntheNewDealpartysystem,DemocratsheldanoverwhelmingmajorityofsouthernHouseseats.AlthoughRepublicansregularlywonaboutathirdoftheseseatsinelectionsfromthe1960sto1990,Democratscon-tinuedtodominatetheregion.Thischangedin1994.Since1994Republicanshaveheldamajorityofsouthernseats.CombinedwiththeirgainsintheMid-westandelsewhere,thepartythathadnotwonaHousemajoritysince1954andhadbeenunabletowinmorethan192seats,26shyofamajority,since1956hasnowwonsixconsecutivemajorities.Inshort,the1994midtermelec-tionbroughtthelong-anticipateddeepeningofthe1960srealignmentintocongressionalelections.ConclusionContrarytoMayhewsconclusionaboutthefoundationalempiricalclaimoftherealignmentgenre,thisreassessmentofelectoraldata“ndssolidevi-denceoffourrealignmentssincetheCivilWar.The“rstofthese,oftenunacknowledged,wasarealignmentin1874…76intheaftermathofRecon-struction.Afairlycompetitivepartysystem,withatilttotheDemocrats PartySystemsandRealignments377remotelyapproachesstatisticalsigni“cance.Coecientsassociatedwitheachofthetransitionvariableswereinthewrongdirection.Theapparenteectsoftheearlieralternativesinthe1930…32analysisaretheresultofthesealter-nativesincludingmanypost-1930elections.Finally,asexpected,arealign-mentde“nitelyoccurredinthisperiod.The1930and1932electionswerenotsimplydeviatingelections.IftheNewDealrealignmentisnottakenintoaccount,inessencesupposingasinglepartysystemfrom1894to1968inpresidentialvotingandto1994inHouseelections,theproportionofvarianceexplaineddropstozerointhepresidentialequationsandonly9percentinthecongressionalequations.TheStaggeredRealignmentof1968…94Howwelldothe1968and1994staggeredrealignmentyearsforthenewpartysystemholduptoalternatives?Atthepresidentiallevel,the1968startdateiswellsupported,butthe1972(+4)electionisalsoapossibility(seetable3).Althoughthetemporalsequencesuggeststhatthecaseisstrongerfor1968,thelargeoverlapinelectionsrequiresatesttodeterminewhether1968or1972wasinfactthestartandwhethereitheroftheseelectionswassimplyadeviatingelection.Table6presentstheanalysis.Equation1speci“es1968astherealigningelection.Equation2speci“es1972astherealignmentwith1968asadeviatingortransitionelection.Equation3speci“es1968astherealignmentwith1972asadeviatingelection.Thestrongestequationisclearlythethird:1968wastherealignmentonset,and1972wasapro-Republicandeviatingelection.The1968onsetregressionwith1972speci“edasadeviatingelectionismuchstrongerthanspecifying1972astherealign-mentonset(adjustedof.33versus.26intable3(+4)).Therealignmentbegunin1968increasedtheexpectedRepublicanpresidentialvoteby4.6percentagepoints,andthe1972electionprovidedanadditional10.6percent-agepointshort-termboosttotheRepublicanpresidentialvoteinthatyear.Thepresidentialvoteanalysisintable3alsoraisesthequestionwhetherthe1952electionorthe1956election(Š12andŠ16)mighthavebeentheonsetelectionfortherecentrealignment.Thesigni“canceofthe1952and1956speci“cations,however,mostprobablyisaconsequenceoftheirbeingdeviatingelections,assuggestedbythefactthatthefollowingtwoalternativepartysystemspeci“cations(Š8andŠ4)arenotsigni“cant.Aregressionincludingbothasingleelectiondummyfor1956asadeviatingelectionand PartySystemsandRealignments375Table5Thepossibilityofan1874…76electoralrealignment Dependentvariable:Democratictwo-partypercentages U.S.HouseseatsPresidentialvote Independentvariables()()()()()( ) …„„. „„. (.)(. )…„. „„.„(.)(. )…Š. „.„Š.(. )(. )(.)(. ) … /....(. )(.)(.)(. )(. )(.) /… Š.. Š.. . (.)(. )(. )(. )(.)(.)Democraticpresidential.votemargin(×inon(.)(. )(.)year,Šinmidterms)Constant... .. .   Adjusted.. ....Standarderrorofestimate. . .. . .Durbin-Watson... ... Notes:Theparenthetical“guresare-ratios.Anasteriskindicatesthat.05,one-tailed.Theyear1912isexcluded.Partysystemvariablesareidenti“edbytheirinitialyear,whichmaybeamidtermelection.Themostrecentpartysystemisidenti“edasstartinginpresidentialelectionsin1968andincongressionalelectionsin1994.thecongressionalanalysis(equation2)butnotinthepresidentialanalysis.Inthepresidentialanalysistheadjusteddropsto.15,andthecoecientfallswellshortofsigni“cance.Inthecongressionalanalysistheadjustedis.46,andtheeectissigni“cant.Reconstructionanditsdismantlingmayhavehadasubstantialeectoncongressionalelectionsinthisperiod.AsaresultofReconstructionintheSouth,thepercentageofDemocraticseatswaskeptlowinthe1868,1870,and1872elections(onlyabout30percentoftheseatsin1868and1872).Thischangeddramaticallyin1874,withDemo-cratswinning63percentoftheseatsandmaintainingatleast45percentofU.S.Houseseatsuntil1894.Ifan1874…76realignmentoccurred,itwasintheoppositedirectionofan1894…96realignment.The“rstfavoredDemocratsastheSouthemerged PartySystemsandRealignments373Table4The1894electionasarealigning,deviating,orrealigninganddeviatingelection Dependentvariable:Democratictwo-partypercentageofU.S.Houseseats Independentvariables()()() …Š. „Š .(. )(.) ….(. )(. )(. )… Š.Š. Š.(.)(. )(.)deviatingsingleelection„Š. (. )(. )Democraticpresidentialvotemargin.(×inonyears,Šinmidterms)(.)(. )(.)Constant...   Adjusted...Standarderrorofestimate. . .Durbin-Watson... Notes:Theparenthetical“guresare-ratios.Anasteriskindicatesthat.05,one-tailed.Theyear1912isexcluded.Equations1,2,and3specify1894asarealigning,deviating,andbothrealigninganddeviatingelection,respectively.=Š4.37,&#x-277;&#x.800;.05).Misclassifyingevenacoupleofelectionssigni“cantlyblursthedierencebetweenthetwoperiods.Thesigni“canteectofthe1894deviatingelectioninthecongressionalanalysis(the“nalregressioninthesecondcolumnoftable3)raisesthepossi-bilitythat1894wassimplyadeviatingelection.Thispossibilityisexaminedmorecloselyintable4.Theequationsconsider1894asarealigningelection(equation1),asadeviatingelection(equation2),andasbotharealigningandadeviatingelection(equation3).Theresultssuggestthatthepartysystemexperiencedbothashort-termjoltandalong-termshiftintheRepublicansfavorin1894„that1894wasbothadeviatingandarealigningelection.Thedeviatingelectionspeci“cationindicatesthatRepublicanswerefavored(a17percentgaininHouseseats),asonemightexpectinthewakeofthe1893PanicwhileaDemocratoccupiedtheWhiteHouse.However,theoverall“toftheequationisnotasstrongasthatidentifying1894astheonsetofarealignment.Equation3,includingboththe1894singleelectionvariable PartySystemsandRealignments371tionordeletionofanelectionortwothatdoesnotbelonginapartysystemshouldproduceameasurementerrorandaweakenedestimatedeect.Usingdummyvariablesforthetransitionperiod,wecanalsodeterminewhetherthelead-uptothesuspectedrealignmentonsetinvolvedsigni“cantpartisanchangeofitsown.Twoadditionalpossibilitiesforeachrealignmentarealsoexamined.One,aninthalternativehypothesis,isthepossibilitythattherewasmentonoraboutthedesignatedrealignmentelections.Inthecaseofthe“rstrealignment,thiswouldamounttoasinglepartysystemextendingfrom1868to1928.Atenthpossibilityexaminedisthatthesuspectedrealignmentelec-tionwasactuallyadeviatingelectionratherthanthebeginningofanewpartysystem.Forexample,Mayhew(2002:53)doubtsthat1896wasarealigningelectionbutnotesthatthereislittledoubtthatthe1896electionstandsoutforitsshort-termchange.Theimpactofadeviatingelectionisestimatedusingadummyvariablescoredoneforthatyearandzerootherwise.Theempiricalsupportforthesealternatives,asre”ectedintheequationsoverall“tandthemagnitudeofthepartysystemscoecients,isthencomparedtothoseforthecanonicalrealignment.Table3presentsthepartysystemscoecientandthepercentageofexplainedvarianceforthecanonicalspeci“cation,itseightbracketedalter-natives,thepossibilitythattherewasnorealignment,andthepossibilitythatthesuspectedrealignmentwasactuallyadeviatingelection.Thesekeystatis-ticsarepresentedforboththepresidentialandthecongressionalanalyses.Inessence,tenalternativehypothesesaretestedforeachcommonlyhypothe-sizedrealignment.The1894…96RealignmentHowdoesthe1894…96realignmentstanduptothealternatives?Althoughsomerivalelectionsarestrongcontenders,noneisclearlysuperiortothe1894…96start.Inthecongressionalelectionanalysis,the1894startdateisstrongerthantheothersrangingfrom1882through1906andhastheonlystatisticallysigni“canteect.Thisdoesnottestwhetherthe1894startdateissigni“cantlydierentfromtheothers,butitdoesindicatethatHouseelectionoutcomesweresigni“cantlymoreRepublicanfrom1894to1928(=Š7.62,.05)thantheywerefrom1868to1892.Wecannotmakethesameclaimaboutthedierencebetweenan1868…96systemandan1898…1928system 370SocialScienceHistorythanabinaryprocess,dummyvariablesusedherecanonlyapproximatewhenthe“rststepsofarealignmentwereevident.Tocomplicatemattersfurther,many“rststeps(orpresteps)mayoccurwithoutshiftingthenormalvoteatall.DemocraticgainsamongCatholics,animportantcomponentoftheNewDealpartysystem,probablybeganin1928,whenDemocratAlSmithwastrouncedbyRepublicanHerbertHoover.Similarly,Republicangainsamongwhitesouthernerstookabigstepforwardin1964despiteLyndonJohnsonslandslidedefeatofBarryGoldwater.Nevertheless,canwesaythatsomedesignationsoftheobservableonsetofrealignmentsarebetterthanothers?Howwelldoalternativepossiblerealignmentscomparetothecanoni-calrealignments?Tables1and2con“rmedthecanonicalrealignmentssince1868:fourpartysystemsseparatedbythreerealignments.Realignmentsapparentlyoccurredin1894…96,1930…32,and1968…94.Toassesstheaccuracyoftheserealigningdatesandavoidprivilegingthesecanonicalelections,eightalter-nativerealigningelections(eightalternativehypotheses)bracketingeachcanonicalrealignmentyearareexamined(e.g.,the“rstrealignmentexaminedinpresidentialvotingis1896).Theeightalternativestartdatesofarealign-mentareset4,8,12,and16yearsbeforeorafterthecanonicalelection(des-ignatedastime).Thealternativeonsetelectionsforthesuspected1896(realignmentinpresidentialvotingare1880(Š16),1884(Š12),1888(Š8),1892(Š4),1900(+4),1904(+8),and1908(+12).Thehighlyunusual1912election,havingbeenidenti“edasanin”uencepointinboththepresi-dentialandthecongressionalanalyses,isexcludedfromtheremainderoftheanalysis.Theenddatesforthispartysystemarekeptconstant,inthiscase1928.Withthecongressionalrealignmentstartingin1894,thealterna-tivesarealsotwoyearsearlier,rangingfrom1878to1910.Thealternativedatesofthe1930srealignmentinpresidentialvotingrangefrom1916(Š16)to1948(+16).Aswiththepriorrealignment,thecongressionalanalysisbeginstwoyearsearlierandrangesfrom1914(Š16)to1946(+16).Thethirdrealignmenthypothesizedtostartin1968inpresidentialvotingrangesfromalternativesof1952(Š16)to1984(+16).Forthecorresponding1994congressionalrealignment,thealternativesrangefrom1978(Š16)to2002+8).Theequationsintables1and2arereestimatedusingthealternativerealignmentyearsandusingtheprecedingpartysystemasthebaselinecate-gory.Sincethesealternativepartysystemshavealargenumberofelectionsincommon,weshouldseeasimilarityincoecients.However,theaddi- Table2RealignmentsindicatedbyU.S.Houseseats,1868…2004 Dependentvariable:Democratictwo-partypercentageofU.S.Houseseats In”uence-pointAllcongressionalelectionscongressionalelectionsexcluded Independentvariables()()()()()( )()() Democraticpresidentialvotemargin.(×inonyears,Šinmidterms)(. )(. )(. )(. )(. )(. )(. )(. ) …„. „(.)(.)(.)(.)(.)(. )…ŠŠ.Š(.)(.)(.)(.)(.)(.) ….(.)(.)(.)(.)(.)(.)… Š. .Š„Š. (.)(.)(.)(. )(.)(.)Constant.. ... ...        Adjusted..... . . . Standarderrorofestimate........Durbin-Watson..... . . .  Notes:Theparenthetical“guresare-ratios.Anasteriskindicatesthat.05,one-tailed.Thedierencebetweenthe1868…92and1894…1928systemsissigni“cant.02(=2.19,one-tailed)inthe“rstsetofregressions.Thedierencesbetweenthe1894…1928and1930…92systemsandthe1930…92and1994…2002systemsaresigni“cantat.01(one-tailed).Inthesecondsetofregressions,eachpartisaneraisstatisticallydistinctfromadjacenterasat.01(one-tailed).Arobustregressionanalysisindicatedthatthe1868,1872,1890,1910,1912,1914,1934,and1936electionswerein”uencepoints.Theseelectionswereexcludedfromequations5…8.Coecientsindicatingdierencesbetweenadjacentpartysystemsaresetinboldtype. 366SocialScienceHistoryhamexaminedchangesinthenational(usuallytwo-party)presidentialvotewithdummyvariablesovertenelections,with“veelectionsspeci“edaspre-realignmentelectionsandthenext“veasrealignment-eraelections.Possiblerealignmentswereexaminedbymovingawindowoftenelectionsoneelec-tionatatime.Thepresentanalysisdiersbyincludingcongressionalelec-tions,takingadirectregressionapproachratherthanexaminingresiduals,examiningpossiblerealignmentssimultaneously,anddetectingandexclud-ingin”uencepointsthatwoulddistortthe“ndings.TheCanonicalPartySystemsTables1and2presenttheresultsoftheanalysisusingthepresidentialvoteandHouseelectionindicatorsofpartystrength.The“rstsetoffourequa-tionsineachtableusesalloftheelectionsintheseries,andthesecondsetusesthosenotexcludedasin”uencepoints.Therobustregressionanalysissetasideonly1912(theWilson-Roosevelt-Taftelection)fromthepresidentialanalysisandexcludedeightelections(1868,1872,1890,1910,1912,1914,1934,and1936)fromthecongressionalanalysis.Theequationsetsareestimatedexcludingeachpartysystemvariableinsequencetoavoidperfectmulticol-linearity.Eachsetoffourequationsintables1and2isessentiallythesameequation,hencethesamesummarystatisticsandsymmetryofcoecientswithdierentbaselineeras.Eachversionisreportedtofacilitatecomparisonofthepartyeras.Theconstantineachequationisthemeanofthedepen-dentvariablefortheeraoftheexcludedvariable.Thisservesasthebaselineforcomparisons.ThecoecientsindicatethedierenceinthemeanDemo-craticPartystrengthbetweentheelectionsoftheperioddesignatedbythevariableandthebaselineera.Forexample,accordingtoequation1intable1,theexpectedDemocratictwo-partypresidentialvotefrom1868to1892was49.31percent,andthisdroppedby3.95percentagepointsintheerafrom1896to1928.Theimportantcomparisonsinvolvethecoecientsinboldtypenexttothediagonal,whichcompareadjacentpartysystems.Thesecoecientsindicatethemagnitudeofanychangeinexpectedpartystrengthfromonesystemtothenext.Adistinctpartysystemoughttobesigni“cantlydierentinitspartisanshipfromtheprecedingsystemandfromthefollowingsystem. PartySystemsandRealignments365re”ectthis,electionsfrom1932to1964arecountedaspartoftheNewDealpartysystemfortheanalysisofpresidentialvoting,andelectionsfrom1930to1992arecountedasNewDealsystemelectionsfortheanalysisofcon-gressionalelections.Thepost…NewDealrealignmentsystemiscodedoneforelectionssince1968inthepresidentialanalysisandsince1994inthecon-gressionalanalysis.Inadditiontothepartysystemvariables,thecongressionalanalysisin-cludesameasureoftheeectsoftheon-yearpresidentialsurgeandthemid-termdecline(A.Campbell1966;J.Campbell1997).Thevariableisthedier-encebetweentheDemocraticpresidentialcandidatesvotepercentageand50percent,takingapositivesigninonyearsandanegativesigninmidterms.Apartyshouldgainseatsintheonyearandloseseatsatthenextmidterminproportiontoitspresidentialvotemargin(J.Campbell1997).Controllingfortheseshort-termoscillationsisnecessarysothattheyarenotconfusedwithlong-termshifts.Thesesurgeanddeclineeectshavenonetpartisaneect.Withintheon-yearandmidtermcycle,themidtermdeclinecancelstheon-yearsurge.Theeectsofrealignmentonsetelectionsareestimatedstraightfor-wardlyusingmultipleregression,withappropriateprecautionsforserialcor-relation.Althoughtheanalysisspans136years,itispossiblethatafewelec-tionsmayhaveunduein”uence,producing“ndingsre”ectingthatelectionratherthanelectionsingeneral.Toavoiddistortionfromin”uencepoints,inadditiontoexaminingthefullsetofelections,bothoftheanalysesareexaminedwithaleastmediansquares(LMS)robustregressiontechnique(RousseeuwandLeroy1987).LMSeectivelyidenti“esandexcludesin”u-encepoints.Itremovesanysubjectivityindeterminingwhich,ifany,elec-tionsaresounusualthattheyshouldbesetaside.Also,toavoidprivilegingorpreselectingthehypothesizedrealignmentyearsorspecifyingsomearbi-traryspanforrealignments,theanalysisconsidersawiderangeofalterna-tiverealignmentonsetelectionsaswellasthepossibilitiesthatthesuspectedrealignmentsdidnotoccuratallorweredeviatingelections.Overthecourseoftheanalysis,withonlythreejusti“edexceptions,everypresidentialelec-tionisconsideredasapossiblerealignmentonset.Whilethismightappearahistorical,thepurposeistoavoidimposingonthedataanypreconceivednotionofwhentheonsetofarealignmentmayormaynothaveoccurred.ThisresearchisinseveralrespectsanextensionofWalterDeanBurn-hams(1970:13…18)analysis(seealsoLawrenceandFleisher1987).Burn- PartySystemsandRealignments363ti“edaspartofthenewpartysystem.Eachfollowedthestartofaneconomicdepression,in1893and1929,respectively,eventsthatmayhaveprecipitatedoracceleratedtheserealignments.Thethird-andfourth-partysystemvariablesdierinthepresidentialandcongressionalanalyses.Afteryearsofweatheringinternalpartydivi-sions,thesplinteringofthepartycoalitionin1948(theDixiecratandPro-gressivebolts),andthetwindeviatingelectionsoftheEisenhoweryears,theDemocraticPartysmajorityinpresidentialvotingdissolvedinthelate1960s(Burnham1991:115;Shafer1991:47;AldrichandNiemi1996:88;Beck1999:39).Althoughtherehadbeensigni“cantregionalshiftsincongressionalelec-tionssincethe1960s,includingRepublicangainsintheSouth,DemocraticdominationofHouseelectionssurviveduntiltheearly1990s.Inelectionsfrom1958to1992,Republicansneverwonmorethan192seats(44percent).TherecentrealignmentbeganwithRepublicaninroadsinpresidentialvotingintheformerlysolidDemocraticSouthinthe1950sand1960s.contrasttotheNewDealsystem,Republicanssince1968havehadagoodmeasureofsuccessinpresidentialelections,winningsevenoftencontests.Aggregateshiftsinboththedirectionandthestrengthofpartyidenti“cationfollowedinthemid-1980s(Norpoth1987;Petrocik1987:349;Miller1991;AbramowitzandSaunders1998;Bartels2000;Hetherington2001),thoughchangeamongwhitesouthernerswasevidentearlier(Ladd1985;Wol“n-ger1985;Stanley1988).Between1952and1980DemocratsoutnumberedRepublicansamongvotersbyanaverageof54percentto38percent,with8percentindependents(Campbell2000:211).In1984thegapnarrowedtoatwo-pointDemocraticadvantage(47percentto45percent),butDemocratsretainedcontroloftheHouse.Thenin1994Republicansbrokethroughincongressionalelections,gaining54seatsandaHousemajorityforthe“rsttimein40years.Itwas,asEverettCarllLadd(1995:22)putit,anotherstepinthevastpartisanrealignmentthathasbeenoccurring.AccordingtoAlanAbramowitzandKyleL.Saunders(1998:647),ThedramaticRepublicanvictoryinthe1994midtermelectionandthereelectionofaRepublicanCon-gressin1996re”ectedalong-termshiftinthepartyloyaltiesoftheU.S.electorate.HouseRepublicanshavecontinuedtoholdnarrowmajoritiesforsixconsecutiveelections.Priortotherealignmentappearinginpartyidenti“cationnumbersinthe1980sanddeepeningintoHouseelectionsinthe1990s,somedeclaredthepartysystemdealignedandtherealignmenthollow(Wattenberg1987), PartySystemsandRealignments361man1976;Gans1985),thisresearchreexaminestheevidencethatnationalpartisanrealignmentshavetakenplacesince1868.Speci“cally,wastherearealignmentin1896,1932,andthelate1960s?WithrespecttoMay-hewsassessmentsofthe15claimsoftherealignmentgenre,ishecorrectinrejectingtheideathatrealignmentonsetelectionsaredistinguishablefromnonrealigningelections(claim1,thegenresfoundationalempiricalclaim[2002:14])andthatthereexistedasystemof1896(claim15[ibid.:30])?TheconventionalviewisthatthepartysystemfromtheCivilWarto1892washighlycompetitive.In1896orthereabouts,intheaftermathoftheeconomicpanicof1893andwiththeDemocraticPartycontrolledbyitspietistic,populist,Bryan-ledwing,theelectoratebecamemoreRepublican(Key1955:12;Glad1964:199;Burnham1970;Kleppner1972;McCormick1986).ThisRepublicaneralasteduntiltheGreatDepressionof1929.Thein”uxofCatholicimmigrantvotersandtheeconomiccrisisproducedaneraofDemocraticPartydominance.Althoughitshowedsignsofsplinteringandweakeningasearlyas1948,thisNewDealpartysystemsurvivedintothe1960sinpresidentialvotingandthe1990sincongressionalvoting.Whilethereisnotyetaprevailingviewofthenatureofthepost…NewDealpartysystem,thepartiesarenowagainnearparity(Ladd1997).Beforereexaminingtheevidenceofwhethertheseorotherrealignmentstookplace,thede“nitionofrealignmentinusehereshouldbemadeclear.Forourpurposes,arealignmentisadurableandsubstantialshiftinthepartiesnationalelectoralbalanceofpower.Thisde“nitioncomportswellwithcom-monusageaswellaswiththede“nitionsusedinmanypreviousstudies,includingMayhews(seealsoKey1959:198;Campbell1966:74;Clubbetal.1980:22;Miller1991:568;Bartels1998).Italsoallowstheeectsandcausesofrealignmentstobeexaminedempirically.Realignmentsmayhavedierentcausesandmaytakeplaceindierentways,butdierencesintheirdevelop-mentshouldnotbeconfoundedwithwhetherarealignmentoccurred.Thisde“nitionalsorecognizesthatrealignmentsvaryinmagnitudeandduration.Itprovidesforrealignmentstobemoreorlessregionalized(Bullock1988)ortoinvolvechangesindierentpopulationorregionalsubgroups(inter-activechange),solongasthesechangesyieldanationalshiftinrelativepartystrength.Manysubnationalelectoralshiftsarepoliticallyimportanteveniftheyosetoneanotherandproducenonationalchange,butourconcernisexclusivelywithnationalshifts.Thelimitedpurposeofthisstudyistodeter-minewhetherandwhennationalpartisanrealignmentsoccurred(Mayhews 360SocialScienceHistoryandathird(JamesA.Gar“eldsin1880)waswonwithasmallerpluralitythanJohnF.Kennedysmarginin1960.RealignmentsQuestionedDespitethisrecordandtheextensivescholarshiponrealignments,theutilityoftherealignmentconcepthasbeencalledintoquestion.Onereasonforthisisitsunsettledde“nition.Realignmentshavebeende“nedasdurablechangesintheissuesthatpoliticallydividethenation(e.g.,Key1955;Sundquist1983)oringrouporregionalattachmentstotheparties(Key1955;Petrocik1981,1987).Theyhavealsobeende“nedasdurableshiftsinthebalanceofpowerbetweentheparties(Key1959)orsigni“cantchangesinthenormalvote(Converse1966).Withinthisperspective,somerequireachangeinwhichpartyholdsmajoritystatus(Burnhametal.1978:49;Gans1985)orshiftsinpartyidenti“cation(Chubb1978;Miller1991).Somede“nitionsbundlevariouspoliticalchanges,includingchangesinpartyvoteshares,theextentofintrapartycon”ict,third-partyactivity,issuepolarizationlevels,turnoutrates,andpublicpolicies(Burnham1970:6…10;1991:115…16).Thisconcep-tualmuddlehascausedsometoabandontheterm(CarminesandStimson1989;Shafer1991;Silbey1991).Therealignmentconcepthasalsobeencriticizedasbeingunrealisti-callydichotomous,requiringovernightchangeinasingleelectionratherthanbeginninginanelectionandextendingoversomeperiod(CarminesandStimson1989:20).Substantialaggregatepartisanchangetakestime.Thoughsecularrealignmentsexplicitlyprovideforgradualchange(Key1959)anddespiterepeatedclari“cationsthatrealignmentstakeplaceoveryears(Key1955:11;Kleppner1981:7;Sundquist1983:11…12),realignmentsarefre-quentlymischaracterizedassingle-electionevents.DavidMayhew(2002:6)hasmovedbeyondpreviouscriticstoclaimthattheconceptofrealignmentandanumberofpropositionsregardingthecausesandconsequencesofrealignmentshavebecomeanimpedimenttounder-standingAmericanelectoralhistory,thattheyconstituteafailedmodelofillumination.Withtheexceptionof1932,hedisputesthatanyrealign-mentshavetakenplace(ibid.:47,141).Heconcludesthatthereisinsucientevidencetosingleout1896asarealigningelectionandobservesthatnocerti“ableelectoralrealignmenthasoccurredsince1932(ibid.:35).InlightofMayhewsscorchingcritiqueandthatofseveralothers(Licht-