Winners Losers and Election Context

Embed Size (px)

Citation preview

  • 8/3/2019 Winners Losers and Election Context

    1/15

    University of Utah

    Winners, Losers, and Election Context: Voter Responses to the 2000 Presidential ElectionAuthor(s): Stephen C. Craig, Michael D. Martinez, Jason Gainous, James G. KaneSource: Political Research Quarterly, Vol. 59, No. 4 (Dec., 2006), pp. 579-592Published by: Sage Publications, Inc. on behalf of the University of UtahStable URL: http://www.jstor.org/stable/4148060

    Accessed: 13/10/2008 00:05

    Your use of the JSTOR archive indicates your acceptance of JSTOR's Terms and Conditions of Use, available at

    http://www.jstor.org/page/info/about/policies/terms.jsp. JSTOR's Terms and Conditions of Use provides, in part, that unless

    you have obtained prior permission, you may not download an entire issue of a journal or multiple copies of articles, and you

    may use content in the JSTOR archive only for your personal, non-commercial use.

    Please contact the publisher regarding any further use of this work. Publisher contact information may be obtained at

    http://www.jstor.org/action/showPublisher?publisherCode=sage.

    Each copy of any part of a JSTOR transmission must contain the same copyright notice that appears on the screen or printed

    page of such transmission.

    JSTOR is a not-for-profit organization founded in 1995 to build trusted digital archives for scholarship. We work with the

    scholarly community to preserve their work and the materials they rely upon, and to build a common research platform that

    promotes the discovery and use of these resources. For more information about JSTOR, please contact [email protected].

    University of Utah and Sage Publications, Inc. are collaborating with JSTOR to digitize, preserve and extend

    access to Political Research Quarterly.

    http://www.jstor.org

    http://www.jstor.org/stable/4148060?origin=JSTOR-pdfhttp://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/action/showPublisher?publisherCode=sagehttp://www.jstor.org/action/showPublisher?publisherCode=sagehttp://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/stable/4148060?origin=JSTOR-pdf
  • 8/3/2019 Winners Losers and Election Context

    2/15

    Winners, Losers, and Election Context:Voter Responses to the 2000 Presidential ElectionSTEPHENC. CRAIGANDMICHAELD. MARTINEZ, NIVERSITYFFLORIDAJASON GAINOUS, UNIVERSITY F LOUISVILLEJAMESG. KANE,FLORIDAOTERElectionsare sometimesseen as legitimizingnstitutions,promoting ystem-level upportamongcitizensbyallowing hemto haveinputinto the politicalprocess.However,priorresearch as foundthat this is less trueamong supportersof losingcandidates,who often exhibit owerlevels of political rustand satisfactionwithdemocracyWe analyzeNESsurveydatafrom 1964 to 2004, as well as surveysfrom Floridaand the nationfollowing he controversial residential lection of 2000, and findthat(1) losers exhibit owerlevels of polit-icaltrust,satisfactionwith democracy,onfidence hatgovernments responsive o citizens,andin early2001were essinclined o extend egitimacyo thenewlyelectedpresident;2) losersalso aremore ikelyto endorse"rationalizations"s explanationsof the electionoutcome,to be less satisfiedwith the choice of candidatesofferedn theelection,and to perceive he electoralprocessas unfair;and(3) voter nterpretationsf theelec-tionmediate he relationships etweenwinning/losingon the one hand,andtrust,responsiveness, ndsatis-factionwith democracyon the other.Thesefindingssuggestthat the so-called egitimizing unction of elec-tions is far froma universalphenomenon.

    lections reat the coreofdemocraticolitics.At eastin principle if not always in fact, they provide citi-zens with a chance to expresstheirpolicy views andpriorities,to participatedirectly n the politicalprocess,andto hold elected leaders accountablefor theiractions (Korn-bergand Clarke1992; Katz1997); as a result,elections areseen as legitimizinginstitutions that protect the system bygeneratingpopularsupportandby helping to "confinemasspolitical action to routine, peaceful channels" (Ginsberg1982: 7). There s, however,an importantcatch to the argu-mentbeingmadehere:Accordingto Ginsberg,"[t]heformalopportunity to participatein elections serves to convincecitizens that thegovernments responsiveo theirneeds andwishes" 7) (emphasis added; also see Rahn, Brehm, andCarlson1999).The catch, of course, is that this probably does nothappen as often as one might hope. In particular, t seemsunlikely thatvoters who support the loser(s) in an electionwill be as quickas those backingthe winner(s)to agreethattheir voices have been heard.This should be especiallytruewhenever, for example, candidatesobscure the real issuesby opting for either generalplatitudesor negative attacks,voters are unhappy with the electoral choices available tothem, or doubts exist about whether the election itself wasconducted in a fair and honest fashion. Indeed, priorresearch indicates that winners and losers do not always

    respond with equal enthusiasm either to the election out-come, or to the institutions and processes through whichthat outcome was rendered. Across a varietyof settings inadvanced industrial democracies, supporters of winningcandidatestend to havehigherlevels of systemsupportthando those who support the losing candidates(GinsbergandWeissberg1978; Clarkeand Acock 1989; Nadeau and Blais1993; Anderson and Guillory1997; Norris 1999; Nadeau etal. 2000; Anderson and Tverdova 2001; Anderson andLoTempio2002; Andersonet al. 2005; but also see Rahn,Brehm,and Carlson1999).1Moreover,t seems thatpolitical nstitutionshelp to shapethe magnitude of differencesbetween winners' and losers'levels of support. In inclusivesystems,where governmentalpower is sharedthroughfederalism,coalitionformation,orseparation of powers, for example, losers apparentlytakesome solace fromthe fact that theirpartisan representativeshave real (though limited) influence in policymaking. Incontrast,exclusivesystems customarilyproducemuch widergaps in supportbetweenwinners and losers,as the latterareeffectivelyshut out of meaningfulgovernmental nfluence atleast until the next election occurs (Andersonet al. 2005:120-40; Andersonand Guillory1997).Little progresshas been made, however, in identifyingthe individual-level processes that account for these pat-terns. Demographicvariablesand politicalpredispositions,including partisanship, extremist ideology, and attitudestowardsalient issues (Andersonet al. 2005: 73-89; Nadeauand Blais 1993) have only moderate influences on the rela-tionships between election outcome and system support,NOTE: An earlierversionof this articlewas presentedat the 2004 AnnualMeetings of the American Political Science Association, andreceived the Best PaperAwardpresentedby the APSAOrganizedSection on State Politics and Policy.We appreciatehelpful com-ments from TomCarsey,RonWeber,Dan Smith, and anonymousreviewers. The analysis by Anderson and Guillory suggests that this relationshipwill vary across contexts, for example, depending upon the degree towhich a political system is majoritarianor consensual (Lijphart1984;also see Norris 1999; Anderson and Tverdova2001).oliticalResearchQuarterly, ol. 59, No. 4 (December 2006): pp. 579-592

    579

  • 8/3/2019 Winners Losers and Election Context

    3/15

    580 POLITICAL ESEARCH UARTERLYand thereforeare of limited utility in explaininghow win-ners and losers might come to think differentlyabout theelectoralprocess. In the presentstudy,we considerwhetherfeelingsof system support are shaped by citizens'attitudesand beliefs about threesalientaspectsof the electoralexpe-rience. Specifically,we propose that winners' support andlosers'despondencymightbe explained,in part,by their(1)satisfactionwith the choiceofcandidates resentedto voters,(2) interpretationsf the election outcome, and (3) assess-ments of proceduralairness.We seek to determinewhetherany or all of these attitudes have a direct and meaningfulimpact on people'sattitudesabout the politicalsystem andits leaders (including their feelings of trust, beliefs aboutgovernmentalresponsiveness,satisfactionwith democracy,and legitimacy accorded to the newly elected president);and also whether they mediate the relationshipbetweencandidatepreferenceand politicalsupport, thereby helpingto explain why winners and losers tend to respond differ-ently after the ballots have been counted and a victordeclared.We test these hypotheses using data from severalopinion surveys,both nationaland statewidein scope, con-ducted over a number of elections (with a particular ocuson the controversialpresidentialracein 2000). Ourfindingsaffirmthe oft-noted contrastbetween winners and losers,and suggest some of the thought processes that underlaythose differences.Although not all aspects of the electioncontext examined here carry equal weight in shaping citi-zens' reactionsto election outcomes, it is clear that certainelements do matter-and that the impact of elections onpoliticalsupportis not invariablya positive one.

    WINNING, LOSING, AND ELECTION CONTEXT

    Perhaps t is true thatby effectively "co-opting" itizens,electionscan "helpto increasepopularsupport forpoliticalleadersand for the regimeitself' (Ginsberg1982: 7). At thelevel of individual behavior, it appears that orientationssuch as trust,efficacy,and satisfactionwith democracytendto be at least somewhat more positive among those whovote or otherwiseparticipate n the political process (Gins-berg and Weissberg1978; Finkel 1985, 1987; Nadeau andBlais 1993; Rahn, Brehm, and Carlson 1999; Joslyn andCigler2001; also see Ciglerand Getter1977).2 Yetelectioncampaigns are, in many instances, extremely competitiveand hard-foughtcontests that producewinners, losers, andhardfeelingson both sides.One would expect hard feelings to be especially acuteamong those who end up on the short end of the stick;accordingly,nsofar as the generationof politicalsupport isconcerned,it is the

    losers'reactions[that]areabsolutelycrucial.Winnersarelikely to be overwhelmingly satisfied with a processthroughwhich the partyor candidatethey voted forgetselected. Losers'support is less obvious. That support

    requiresthe recognitionof the legitimacyof a procedurethat has producedan outcome deemed to be undesirable(Nadeauand Blais1993: 553).Just as the hard test of political tolerance is "thewilling-ness to permit the expression of those ideas and interests[that]one opposes"(Sullivan,Piereson,and Marcus 1979:784), a crucial aspect of legitimacyhas to do with losers'acceptanceof the election outcome as valid, and theirwill-ingness to consent to the winners'rightfulauthority o insti-gate policies to which losers may be strongly opposed. In abroader sense, the persistence of democratic institutionswould seem to depend on the acknowledgementby winnersand losers alike that the people are capableof casting theirballots in a wise and thoughtfulmanner, i.e., that the deci-sions renderedat the polls are not frivolous, and that theidea of democracyitself is not a sham. We believe that thelikelihoodof voters (especiallylosers) reachingthatconclu-sion depends in parton the meaningthey attachto the elec-tion outcome, their satisfactionwith the choice alternatives,and theirbelief that the election was conducted fairly.

    Interpretationsf the ElectionPrior research has found that an election can produce

    multiple interpretationshatvary by sourceand evolve overtime. Kingdon(1966: 23-34), forexample,discovereda con-gratulation-rationalizationffect among winning and losingcandidates:The former were much more likely than thelatter to believe that voters were at least reasonablywellinformed,and that they usually made the right decision atthe polls for the rightreasons(especiallyin termsof votingfor the better candidaterather han the party abel).Winnerswere thus inclinedto congratulatehe electorate-and, indi-rectly, themselves-for its good judgment, while losersrationalizedheirdefeatby assumingthatvotershad failed tomeet theirresponsibilitiesas democraticcitizens.Candidatesarenot the only ones who interpret he meaningof elections,of course. The media, political activists,and other opinionleaderssiftthroughthe cacophonyof campaigns n search ofthe realmessagethat(theybelieve)the electorate ntendedtoconvey, throwingout a good deal of wheat with the chaffalong the way (Hershey1992; Thomas and Baas1996).In the present study,we will look for evidence of con-gratulation-rationalizationt the mass level; that is, we willexamine in some detail the meaning attached to electionoutcomes by voters themselves. Why, for example, wasGeorgeW Bush elected president of the United States in2000? Was it because people responded favorablyto hisconservativepolicy agenda, his personal character,and hispromise to end the partisanconflict that had characterizedrelationsbetween Congressand the White House duringtheClintonera? Or did Bushbenefit more from the support ofspecial interests,the gullibilityof voterswho were deceivedinto supporting someone who was more ideologicallyextremethan he claimed to be, and the failure of state andlocal officials to ensure an honest vote tallyon election day?Following Kingdon (1966), we anticipate that there areFor a differentperspective, ee ClarkeandAcock(1989) andFreie 1997).

  • 8/3/2019 Winners Losers and Election Context

    4/15

    WINNERS, LOSERS,AND ELECTIONCONTEXT 581

    systematicdifferences n how winners and losers construethe meaning of an election, and that those interpretationswill affect the legitimacyaccordedby voters to the victori-ous candidate,as well as their feelings of trust in govern-ment and in the political process.ChoiceSatisfaction

    It is also our expectationthat politicalsupportwill varyaccordingto the degreeof satisfaction hat people feel withthe choice of candidates offered to them in a given cam-paign-and that such satisfaction, ike congratulation-ration-alization, s a mediatingfactorhelping to explain the diver-gent responses of winners and losers. The importance ofboth quantity and quality of options on satisfaction withchoice has been establishedoutside the politicalrealm,3butthere is little evidence to suggest that perceptionof choicehas much effect on such orientations as partisan strength(Craigand Martinez1989) or trust in government (Millerand Listhaug1990).Even so, and even in an era of increasing polarizationbetweenpartyelites (Abramowitzand Saunders1998; Het-herington 2001), we believe that choice matters to somevoters, though its salience may have less to do with issuesand ideology (e.g., those in the politicalcenterbeing unableto findexpressionfor their views at the ballot box; see Hib-bing and Smith 2004; or, conversely, those with moreextremepreferences eelingunhappywith partiesthat stressaccommodationand coalition-buildingoverideologicalclar-ity) than traditionalrepresentationalmodels envision. It alsomay have to do less with choice per se than with the beliefthat,in any given election,voters arepresentedwith at leastone candidatewho is deemed to be acceptable n terms ofpolicy direction,leadershipstyle, personalcharacter,and/orsome other factor or combination of factorsthought to beappropriateor the rightfuland effectiveexercise of politicalauthorityWhatever ts source,4we suspect that the level ofvoters' atisfactionwith the choicespresented o them on theballot will affect theirsupportforthe political regime.Procedural airness

    Finally, there is considerable evidence showing thatAmericanscare about decision-makingprocessesas well as

    outcomesand, specifically,hattheywill morereadilyaccept"unpalatable inding decisions"made by others if they per-ceive that those decisions were made in an appropriatemanner(HibbingandAlford2004: 62; also see HibbingandTheiss-Morse2002).5 Our approachhere is rooted in theconceptof proceduralustice, as elegantlydescribedby Lindand Tyler(1988). The idea is that people'sevaluationsofinstitutionsareshapednot only by the rewardsor outcomesthey receive,but also by theirperceptionsof the mannerinwhich those outcomes are allocated.Thus, a number of fac-tors might affect subjective proceduralfairness, includingprior expectations, beliefs about whether people wereafforded he opportunity o have theirsay, participants'evelof control, the perceivedneutralityof decisionmakers,andoverall assessmentsregardingthe fairness of the decision-making process. In short, procedural ustice theory positsthat citizens' evaluationsof an institution'sproceduresaffectthe level of trust and legitimacyaccorded to thatinstitution,independent of the substantive outcomes of the process.While much of the empiricalresearchthat emanates fromthis theory has centered on litigants'evaluationsof courtsand alternativedispute-resolutionmechanisms(see Lind andTyler1988: chap. 4-5), therealso is mountingevidence thatperceptions of fairness affect general support for politicalleaders and the politicalsystem as a whole (Tyler,Rasinski,and McGraw1985; Tyler,Casper,and Fisher 1989; Rasinskiand Tyler1988; Tyler1994; Mutz and Mondak1997).Accordingly,we propose that the levels of political trustand support expressedby winners and especiallylosers areshaped in part by theirgeneralbeliefs about the fairnessofthe electoral process. It is relatively easy, for example, forwinners to believe that the competition was fair, that theoutcome was an accuratereflection of public preferences,and that the institutionsheadedby those who have recentlybeen elected will govern wisely,well, and in the best inter-ests of all citizens. But what about the losers, especiallyonthose occasions when manyon the losing side conclude thatthe selection process was tainted and that, had the votesbeen counted accurately, their candidate would haveemergedvictorious?In sum, we hypothesizethe following:

    H1: Citizenswhosupportwinning andidates re more ikelythan thosewhosupportosers o have a senseof trust ngovernment,o believethat thesystem s responsiveopeoplelike themselves,o be satisfiedwith the demo-craticprocessas a whole,and to regardhe electionout-comeas a legitimate xpression f thepublic'swill.H2: The relationshipbetweenwinning/losing nd politicalevaluationss mediated ycitizens' eliefs boutwhetheror not the electionoutcome epresented true mandate.

    Forexample, Bottiand Iyengar 2004) discovered that the act of choos-ing from among appealing flavors of yogurt increased subjects'overallsatisfactionwith the selection, but that the act of choosing from a set ofunappealingflavors (sage, chili powder, tarragon, celery seeds) led todecreasedsatisfaction.With regard o quantity,hereis growingevidencethat while people often think that they would prefermore choices, toomany alternatives n a choice set may actually produce greater anxietyand, hence, less overallsatisfaction,e.g., see Schwartz(2004).4 As a purelypracticalmatter,our measure of choice satisfactioncapturesonly the degreeof satisfaction-not the reasoningbehind it. We includeit in our analysis n the belief that (1) some people in any election cast aballot for what they regardas the lesserof two evils; (2) losers are morelikely than winners to do so; and (3) when thathappens, it will tend toreduce the level of supportforgovernmentand the electoralprocess.

    What constitutes"appropriate" ay vary from one situation to the next,but within an electoralsettingwe takeit to mean that, forexample, therewas an open and honest exchange of views during the campaign,that allsides had an opportunity to be heard, that debate centered more onpolicy than on personality,and that an accuratevote count was taken onelection day (cf. Katz 1997).

  • 8/3/2019 Winners Losers and Election Context

    5/15

    582 POLITICAL ESEARCH UARTERLY

    H3: The relationshipbetweenwinning/losing nd politicalevaluationss mediatedbycitizens' atisfactionwith thechoice orselection) fcandidates fferedn the election.H4: Therelationshipetweenwinning/losingndpolitical val-uations s mediated yperceptionsfelectoralairness.While interpretations,choice satisfaction,and perceivedfairness are probablycorrelated,we expect that the originsof these orientationsare somewhat differentand, as a result,that each has different implications for system support.First, to the extent that interpretations ffectsupport, opin-ion leaders who help to shape those interpretations(Thomas and Baas 1996) may play an important role ineither enhancing or mitigatingthe legitimizingfunction ofelectionsas institutions.In an era of sharplypolarizedelites

    fightingfor swing voters in close elections (Fiorina 2005),the prospect that interpretationsof election outcomes bylosing candidatesand their followerswill contribute to highlevels of system support appearsbleak-though the evolu-tion of consensual accounts in the media (Hershey 1992)may sometimesprovidea push in that direction.Second, inextraordinary ircumstances,perceptionsof fairnessmightalso be impaired by partisan conflict among elites, butshould be affectedas well by (1) the public'sunderstandingof how the electoral process works and (2) the degree towhich citizens believe that every vote has been countedfairly with the medialikelyto have an impacthere as well).Finally,althoughsatisfactionwith electoral choices maywaxand wane in response to changes in candidaterecruitmentpatterns,we suspect that this variablewill be less suscepti-ble thaninterpretations f the election to elitemanipulation.

    DATA AND MEASURESWe use a number of datasets to test these propositions:NationalElectionStudy surveys from 1964-2004 allow usto examine, over a period of 40 years, the effects of presi-dential candidate preference on political support6; thatreview is supplemented with evidence from a national,cross-sectionaltelephone survey conducted from May 17

    throughJune 1, 2001, by the Florida Voterpolling organi-zation.7Our core analysesare based on a statewide, cross-sectional telephone poll (also done by Florida Voter from

    March24-April5, 2001) thatcontains he mostcompleteset of measuresof citizens'attitudes oward he electoralprocess.8Finally,a 1998 surveyof registered oters inFlorida rovides nopportunityo testpartof ourargumentwith data roma gubernatoriallection.9

    The principal dependentvariablesexaminedin ouranalyses reas follows seeAppendixA forcompleteques-tionwordingsn the 2001 Florida urvey'l):Political rust:1) how often hosewho rungovernmentanbe trusted o do theright hing; 2) is governments runby a fewbiginterests r forbenefitofallthepeople?ResponsivenessfGovernment:1) how much attention ov-ernmentpaysto whatpeople hink; 2) how muchelec-tions makegovernment ayattention.SatisfactionithDemocracy:espondent'satisfactionwithhow democracy orks n the UnitedStates.

    6 See www.umich.edu/-nes. The 1964-2002 data are taken from the NESCumulative File 1948-2002, while 2004 data are fromthe 2000-2004NES PanelStudy and the 2004 NES. Neither the NES nor its principalinvestigatorsbearany responsibility or our analysesor interpretations.

    7 Thesamplefor our nationalsurvey(N = 1000) was obtainedby random-digit dialing,with up to eleven callbackson all working numbers andinitial refusalsbeing attemptedbeforean alternatenumberwas selected.Upon reaching a working residentialnumber, interviewers asked tospeakto "theyoungestmale residentof your household, 18 years of ageor older, and a U. S. citizen, who is now at home." If no male was athome, interviewersasked to speak "with the oldest female, 18 years orolder,and a U. S. citizen, who is now at home." The overallresponseratewas 53 percent(AAPORStandardDefinitions).Ouranalysis s based on704 respondentswho indicated thatthey had voted for eitherGeorgeW.Bush or Al Gorein the 2000 presidentialelection.

    8 Respondentswere chosen randomly from a list of registeredvoters inFlorida, with up to four callbacksattempted on all working numbersand initial refusals. This yielded a total of 604 interviews, and ouranalyses are based on responses from 535 voters who indicated thatthey had voted for either GeorgeW Bush or Al Gore. Therewere nosubstantivedifferences n the results of the analyses n Tables4-6 usinga replicationthat included both voters and respondentswho failed tovote but had a preferencefor one of those candidates(total n = 559).Oursampling frame for this survey (a list of registeredvoters)does nottechnically fit the AAPOR 2004 standards for random-digit dialing,household, or mail surveys. Afterremoving bad numbers, non-regis-trants,and others who fell outside the targetpopulationfrom our cal-culations, the overall response rate was 45.5 percent, which is lowerthan we might havehoped. Basedon our experience and conversationswith various commercial pollsters, however, we believe that Floridatends to have a somewhathigherrefusalrate thanmostother states. Themarginof error for the surveyis plus or minus fourpercentagepoints.9 This statewide,cross-sectionalsurvey was conducted from November10-22, 1998, by Florida Voter. Respondentswere chosen randomlyfrom a list of registeredvoters;only those whose names were chosenfrom the list were actually interviewed, and up to four callbacks wereattemptedon all workingnumbers and initialrefusals. This procedureyielded 613 respondents,and a marginof errorof plus or minus fourpercentagepoints. Analysesare based on 502 actualvoters forgovernor.Results of the multivariateanalysis n Table8 aresimilarusing a datasetbased on responsesfrom 570 voters and non-voting supporters.Addi-tional information about any of the Florida Voter surveys can beobtained by contactingthe GraduateProgram n PoliticalCampaigningin the Political Science Departmentat the Universityof Florida.Formultivariateanalysesusing the two statewideFloridaVotersur-veys and 2004 NES, we addressed potential bias in missing data byusing the multiple imputationusing chainedequations(MICE)routine.MICEand similaralgorithmsoperate by replacingmissing values witha random draw from a distributionestimated from a maximum likeli-hood function basedon other variables n the dataset.(Kinget al. 2001provide a general discussion of multiple imputation, and the MICEpackageis explained in detail by Van Buuren and Oudshoorn 1999.)We drew five imputedvalues for each missingvaluewhich, when com-bined with the observednon-missingdata,form five replicatedatasets.Resultspresentedhere arepooled estimatesbased on separateanalysesof those datasets.

    10 Each of these measures was rescaledso that scores rangedfrom0 to 1,with the lattersignifyingmore positive feelings.Therewere slightvari-ations in questionwordings across the NES and FloridaVotersurveys.The codebook for the latter s in AppendixD. Sourcecode for our dataanalysis s in AppendixE.

  • 8/3/2019 Winners Losers and Election Context

    6/15

    WINNERS,LOSERS,AND ELECTIONCONTEXT 583BushLegitimacy:hetherrespondentsees GeorgeW Bush aslegitimate presidentof the United States.The central independent (dummy) variable is operational-ized in terms of a respondent's upportfor the winning can-didate; those who supported Bush received a score of 1,those who supportedGore a score of 0.In the 2001 statesurvey,congratulation-rationalizationsmeasuredby respondents'assessment of how well severalstatements (see Appendix A) applied to the election out-come. High scores on this variableindicate an acceptancethat the electorate'schoice conveyed a genuine preferencefor the winner in one or more key areas:a belief that voterschose Bush because of his commitmentto smallergovern-ment and taxes,his desire to improvethe workingrelation-ship between White House and CapitolHill, his abilitytohandlethe nation'simportantproblems,and the expectationthathe would be honest and not embarrass he officeof thepresidency as had his predecessor. Individuals with lowscoreswere more likely to rejectthose explanations,and tobelieveinsteadthat the Republicannominee won because ofhis backing from special interests, voters' lack of intelli-gence, their failureto reward the Democratsproperlyfor astrongeconomic performanceduringthe Clintonyears,andBush'sabilityto fool people into thinkinghe was less right-wing than he actuallywas. The index, which is rescaledtorangefrom0 to 1, has a Cronbach'saoof 0.79 and a meancorrelationof .31.We gauge choice satisfaction with a single item askingrespondentsto evaluate the choice of candidatesin 2000 incomparison with past presidential elections. Finally, ourmeasureof proceduralairness s similar to an item asked inNES(and the ComparativeStudyof ElectoralSystems)sur-veys since 1996, though we asked respondents how fair"electionsn the United States" re; n contrast,the NES ver-sions refer to "the ast election in the United States" 1996)or "thepresidentialelection we'vejust had"(2004). Each ofthese variables s rescaled fromzero to one.

    RESULTS

    Our firsthypothesis states that citizens who favorwin-ners aremore likely than those who prefer osers to expresspositive orientations toward the government, its leaders,and the political system in general.As shown in Table 1,NES post-presidential-electionsurveys from 1964-2004providemixed support for this proposition, a finding con-sistent with that reported by Anderson and his colleagues(2005: 66-67). Individualswho voted forJohnson in 1964,Nixon in 1972, Reagan n 1984, Bush "41" n 1988, Clintonin 1996, and Bush "43" n 2004 were indeed significantlymore trusting and more likely to describe the politicalsystem as responsive than were those who voted for theirprincipal challengers. Similarly,Bush "43" voters in 2004weremore satisfiedwith democracythan were Kerryvoters.However,a firstglance at results from the remainingelec-tions in the time series casts doubt on the generalizability f

    the overall argument.Supportersof successful challengersto incumbentpresidentsor vice-presidents(Nixon in 1968,Carter n 1976, Reagan n 1980, Clintonin 1992, and Bush"43"in 2000) were generallyno more trusting nor morelikely to perceivea high degreeof responsivenessthanwerethose who had voted for the statusquo."1While Anderson et al. (2005: 67) speculated that thepeculiaritiesof divided governmentand relatively frequentchanges in control of the executive might account for thelack of a clearpatternacrossU.S. elections, it is also possi-ble that attitudes about the political system are in fluxduringthe post-electionperiodbefore a new administrationtakes office (which is preciselywhen NES interviewers aretalkingwith respondents).Duringthatperiod, some peoplemay base their answers to the trust and responsivenessquestions on assessments of the previous four years, whileothers look ahead. If this is the case, H, would lead us toexpect sharp differencesin the levels of political supportbetween the new president'ssupporters and those whofavoredhis opponent only after the formerhas taken office.That is precisely the patternevident in Table2, whichshows scores forpoliticaltrust,responsiveness,and satisfac-tion with democracyin NES midterm surveys from 1966-2002. Two years into each president'sterm, citizens whorecalledvoting for the winner consistentlyreportedhigherlevels of trust and perceivedresponsivenessthan did thosewho had backed the defeated candidate. In almost everymidtermfrom 1966-2002, respondentswho recalledvotingfor the incumbent president exhibited consistently higherlevels of political supportthan those who had supportedthedefeatedcandidate; he sole exceptionis 1998, when Clintonand Dole supportershad equal levels of responsiveness.12The changeswe observe from presidential-election ears tomidtermyearsarelargelyconsistent with those reportedbyAndersonet al. (2005: 82-83), and implythat a causal arrowruns from candidatepreference o politicalsupport,notwith-standingthe possibilityof a recursiverelationship n presi-dential electionyears (Hetherington1998).While the 2000 NES Studyshows that between electionnight and the new president's nauguration,Bush support-ers were no more likely than Goresupportersto be satisfiedwith the government'strustworthiness,responsiveness, orthe way democracyworks, our own surveysin 2001 suggestthat differences between supportersof winning and losingcandidates become apparentvery earlyin a new president'sterm. Results from the statewide Florida survey in lateMarch/earlyApril2001 (top portion of Table3), as well as

    1Not surprisingly, upportersof the more successful third-partycandi-dates (Wallace in 1968, Anderson in 1980, and Perot in 1992 and1996) had notably ower scores on trust andresponsiveness hanmajor-partyvoters (see Hetherington1999). Perotvoters in 1992-96 alsowereless satisfied with democracy than their Democratic or Republicancounterparts.12 In our analysisof NES midtermsurveys,we rely on respondents'recallof their vote forpresidenttwo yearsearlier.The basic findingsfor2002reported n Table2 are the sameusing either the 2002 recall measure orthe 2000 vote reportforpanel respondents.

  • 8/3/2019 Winners Losers and Election Context

    7/15

    584 POLITICAL ESEARCH UARTERLY- TABLE1

    CANDIDATE HOICEAND POLITICALVALUATIONSPRESIDENTIALLECTION EARS)PoliticalTrust(2 item scale)Year Loser Winner Difference t sig (t)

    1964 0.45 0.65 0.20 9.30 0.001968 0.54 0.45 -0.08 -3.49 0.001972 0.25 0.44 0.20 11.11 0.001976 0.30 0.16 -0.14 -8.60 0.001980 0.24 0.16 -0.08 -4.02 0.001984 0.24 0.40 0.16 9.04 0.001988 0.21 0.37 0.16 8.87 0.001992 0.21 0.18 -0.04 -2.32 0.021996 0.19 0.28 0.09 4.46 0.002000 0.32 0.31 -0.01 -0.37 0.712004 0.22 0.47 0.25 11.37 0.00Responsiveness2 item scale)Year Loser Winner Difference t sig (t)

    1964 0.60 0.73 0.13 6.87 0.001968 0.66 0.66 0.00 -0.21 0.831972 0.56 0.67 0.11 8.10 0.001976 0.58 0.57 -0.01 -1.17 0.241980 0.53 0.54 0.01 0.77 0.441984 0.52 0.56 0.04 2.72 0.011988 0.49 0.60 0.11 7.09 0.001992 0.56 0.59 0.03 1.95 0.051996 0.55 0.62 0.06 3.53 0.002000 0.62 0.59 -0.03 -1.66 0.102004 0.57 0.68 0.11 6.49 0.00Satisfactionwith Democracy

    Year Loser Winner Difference t sig (t)1996 0.71 0.72 0.02 1.10 0.272000 0.73 0.72 0.00 -0.13 0.902004 0.55 0.77 0.22 12.73 0.00Note: Data on trust and responsivenessfrom 1964 to 2000 are fromthe NES CumulativeFile. Data on satisfactionwith democracyare fromrespective NEStime-seriesstudies. Data from 2004 are from the 2004 NES. Entries ndicate the scores (rescaledfrom 0 to 1) forvoters for the winning major-party andi-dateand for the losing major-party andidate.The t-test reportsthe significanceof the differencebetween the supportersof the winning major-party andi-date and losing major-party andidate.Pairwisedeletion of missing data.

    fromournationalurvey romMay/June001 (bottompor-tion of Table3), are consistentwith the NES midtermresults. nbothinstances, oterswho preferredGeorgeW.Bushforpresidentweresignificantlymoretrustful,morelikelyto believe hatgovernments responsive,moresatis-fied with the way democracys working n the UnitedStates, nd more ikely o acceptBushaslegitimateven nlightof theongoing ontroversyver he Florida ote count.Overall,hen,H is handsomelyonfirmed.Data rom heNESand our Florida urvey presentednthetopportion f Table ) alsoshowthatwinners nd osersofferdifferent ssessmentsf the electoral rocess. n 1996,when the NESbeganasking espondentso assess he fair-ness of themostrecentnationallection, majorityf votersgave he mostpositive atingpossibleon a five-pointcale;asexpected, owever,hosewhosupportedPresidentClin-

    ton'sreelectiongave a slightlystrongerendorsementof theprocess than did supporters of his defeated challenger.During and shortlyafter the legal dispute surroundingthe2000 presidentialelection, NES respondents of all stripeswere noticeablyless sanguine-and yet, once again in linewith our hypothesis, Bush supporters were significantlymore likely than Goresupportersto describethe election ashaving been fair.13Perceptionsof proceduralfairnesswerehigherin 2004, but the differencesbetweenBushand Kerryvoters were actuallymore pronounced than those betweenBush and Gore supportersin 2000. Finally, n the Floridasurveyconductedshortlyafterthe new presidenttook office

    13Resultsover timeare similardespite he evolutionn NESquestionwordingand ormat) otedearlier.

  • 8/3/2019 Winners Losers and Election Context

    8/15

    WINNERS, LOSERS,AND ELECTIONCONTEXT 585= TABLE

    PRESIDENTIALANDIDATE HOICEAND POLITICALVALUATIONSMIDTERMELECTION EARS)Political Trust(2 item scale)Year Loser Winner Difference t sig (t)

    1966 0.50 0.67 0.17 7.01 0.001970 0.48 0.53 0.05 2.43 0.021982 0.36 0.41 0.05 2.38 0.021990 0.28 0.36 0.07 4.92 0.001994 0.30 0.35 0.05 2.98 0.001998 0.37 0.42 0.05 2.34 0.022002 0.35 0.47 0.12 4.76 0.00Responsiveness(2 item scale)Year Loser Winner Difference t sig (t)

    1970 0.61 0.67 0.06 2.43 0.021998 0.59 0.59 0.00 0.13 0.902002 0.73 0.78 0.04 2.01 0.04Satisfactionwith DemocracyYear Loser Winner Difference t sig (t)

    2002 0.67 0.77 0.10 5.34 0.00Note: Data on trust and responsivenessfrom 1966 to 1998 are from the NES Cumulative File. Data from2002 are from the 2002 NES. Entries ndicate thescores (rescaledfrom 0 to 1) for voters for the winning major-party andidateand for the losing major-party andidate. The t-test reports the significance ofthe differencebetween the supportersof the winning major-party andidate and losing major-party andidate.Pairwisedeletion of missingdata.

    in early 2001, Bush supporters were substantiallymorelikelythan Goresupporters o endorse the generalstatementthat"electionsn the United States"are conducted fairly.Floridianswho backed the winner also were more posi-tive in their evaluations of other aspects of the electoralprocess. Bush supporters,for example, were more satisfiedthan Goresupporterswith the choice alternativespresentedto voters in 2000, relative to previous elections. Further,Goresupporters ended to believe that the presidentialelec-tion had turned on such factors as the influence of specialinterests and the shortcomings of ordinary voters (notgiving Democratsenough credit for a strong economy, let-ting themselves be misled about Bush's deological leanings,and their failure o understandthe issues), and less inclinedto believe that Bush won because of his conservativepoli-cies, perceived competence, honesty, and pledge to workmore effectivelywith Congress.Each of these differences sin the expected direction, of considerablemagnitude, andstatistically significant (see the second and third sets ofentriesin Table4, ft. 14).

    Hypotheses2, 3, and 4 suggest, of course, not only thatsupportersof winning and losing candidatestend to viewthe outcome in differentways, but that interpretationsofwhat the election means, choice satisfaction,and percep-tions of electoral fairness serve to mediate the relationshipbetween winning/ losing and political evaluationssuch astrust, governmentalresponsiveness(which is conceptuallyvery close to the concept of external efficacy,see Craig,Niemi, and Silver 1990), satisfaction with democracy,andlegitimacyThis argumentrequiresevidence that the medi-ating variables are related to orientations toward govern-

    ment and the electoral process, and that the relationshipbetween winning/losing and system-level orientations isobviated,or at least significantlyweakened, in multivariatemodels that include the mediatingvariables. Entriesin thecorrelationmatrix shown in Table5 indicatesupport forthefirstcondition;that is, people who attributedBush'swin tothe voters' wisdom rather than their ignorance, who wererelativelysatisfiedwith the choice of candidates in 2000,and who believed that the election was fairlydecided gen-erallyexhibitedhigher levels of support on all four dimen-sions that we measured.Our test of the second condition is presentedin Table6,where we estimate 12 ordered-logit regressionsthat showthe effects of election-attitudevariableson political-supportvariables,controllingfor candidatepreference.These sepa-rateestimates revealwhich, if any,of the formermediatetherelationship between candidate preference and the latter(dependent) variables (since the mediating effects of anyparticularvariablemight have been masked by the multi-collinearity n a multivariatemodel that included all threeelection-attitudevariablesand candidatepreference,as sug-gested by the correlations n Table5). In the firstcolumn ofequationsin Table6, estimating political trust, we see thatcongratulation-rationalizationlaysa mediatingrole in thatit is a significant predictorof trust and, moreover,its pres-ence changesthe effectof candidatepreferenceon trust:themagnitudeof the lattercoefficient is weak, and not signifi-cantlydiscernible fromzero. The two other equationsshowthat election fairnessand choice satisfactionalso are signifi-cantpredictorsof trust,thoughneither obviates the effect ofcandidatepreference.

  • 8/3/2019 Winners Losers and Election Context

    9/15

    586 POLITICAL ESEARCH UARTERLY- TABLE3

    PRESIDENTIALANDIDATE HOICEAND POLITICALVALUATIONSN 2001FloridaSurvey(March-April2001)Gore Bush Difference t sig (t)

    PoliticalTrust 0.28 0.49 0.21 9.05 0.00Responsiveness 0.48 0.63 0.15 5.54 0.00Satisfactionwith Democracy 0.70 0.77 0.08 3.36 0.00Bush Legitimacy 0.50 0.97 0.47 18.86 0.00

    NationalSurvey (May-June2001)Gore Bush Difference t sig (t)PoliticalTrust 0.36 0.47 0.11 5.73 0.00Responsiveness 0.52 0.63 0.11 3.59 0.00Satisfactionwith Democracy 0.69 0.80 0.10 4.82 0.00Bush Legitimacy 0.48 0.97 0.50 20.93 0.00Note: Data are from the 2001 FloridaVotersurvey of registeredvoters in Florida,and fromits nationalsurveyin 2001. Entries ndicate the scores (rescaledfrom 0 to 1) of voters for the winning major-party andidate and for the losing major-party andidate. The t-test reports the significanceof the differencebetween supportersof the winning major-party andidate and losing major-party andidate. Pairwise deletion of missing data.

    As with politicaltrust,perceptionsof responsivenessarehigheramongFloridianswho endorseda mandateinterpre-tation of the election,were satisfied with the choice of can-didates,and believed that the election was fairlyrun;how-ever,results n the second column of Table6 show thatonlycongratulation-rationalizationmediated the relationshipbetween candidate preferenceand responsiveness. In themultivariatemodel with congratulation-rationalization,hecoefficientfor winning/losingis very small and not signifi-cant at conventional evels. Once again,however,the equa-tions with choice satisfaction and perceptions of fairnessshow that while these variableshave significanteffects oftheir own, neither significantlymediates the relationshipbetween candidatepreferenceand perceivedresponsiveness.The findingsfor satisfactionwith democracyareslightlydifferent. On its face, this question seems to be tapping

    broad regime-levelorientationsrather than attitudes aboutspecific governmentalleadersand institutions, though it ispositively correlated with both winning/losing and each ofthe three indicators of election context (also see Canache,Mondak, and Seligson 2001). Consistent with what wehave observed thus far,the coefficient for congratulation-rationalization is positive and significant, controlling forcandidatepreference,and the latter's ffect fades when con-trollingforcongratulation-rationalization.n this case, per-ceptions of fairnessalso have a mediatingeffect: When wecontrol for beliefs about whether the election was con-ducted fairly,Bush supportersand Goresupportersappearequally satisfied with American democracy. As before,choice satisfaction is a significant predictorof satisfactionwith democracy,but it does not alter the latter'srelation-ship with candidatepreference.- TABLE 4

    EVALUATIONSF ELECTION ROCESSES YCANDIDATE HOICEElection FairnessData Source Loser Winner Difference t sig (t)

    1996 NES(V961460) 0.78 0.85 0.07 4.14 0.002000 NES(V001291) 0.51 0.69 0.18 8.40 0.002004 NES(V045042) 0.65 0.91 0.26 14.62 0.002001 FloridaSurvey 0.63 0.85 0.21 9.22 0.00

    Satisfactionwith ChoiceData Source Loser Winner Difference t sig (t)2001 FloridaSurvey 0.59 0.75 0.16 7.03 0.00

    Congratulation-RationalizationData Source Loser Winner Difference t sig (t)2001 FloridaSurvey 0.37 0.75 0.38 27.19 0.00Note: Entries ndicate the scores (rescaledfrom0 to 1) of voters for the winning major-party andidate and for the losing major-party andidate. The t-testreportsthe significanceof the differencebetween supportersof the winning major-party andidate and losing major-party andidate.Pairwisedeletion ofmissingdata.

  • 8/3/2019 Winners Losers and Election Context

    10/15

    WINNERS, LOSERS,ANDELECTIONCONTEXT 587= TABLE

    CORRELATIONSF POLITICALUPPORTANDATTITUDES OWARD LECTIONSatisfaction SatisfactionPolitical with Bush with Congratulation- ElectionTrust Responsiveness Democracy Legitimacy Choice Rationalization Fairness

    Political Trust 1.00 0.36 0.37 0.40 0.34 0.44 0.36Responsiveness 0.36 1.00 0.23 0.28 0.24 0.29 0.27Satisfactionwith Democracy 0.37 0.23 1.00 0.24 0.18 0.20 0.38Bush Legitimacy 0.40 0.28 0.24 1.00 0.19 0.63 0.46Satisfactionwith Choice 0.34 0.24 0.18 0.19 1.00 0.33 0.21Congratulation-Rationalization0.44 0.29 0.20 0.63 0.33 1.00 0.43ElectionFairness 0.36 0.27 0.38 0.46 0.21 0.43 1.00Note: Dataare fromthe 2001 FloridaVotersurvey of registeredvoters. Tableentries are correlationcoefficients (missingdataimputed). High scores indicategreater political trust, perception of responsiveness,satisfactionwith democracy,perception of Bush legitimacy,satisfaction with the choice of candidates,greaterperceivedfairness,and a greaterdegreeof congratulation believingthat the election outcome reflecteda positivevote for the winner,his policies,andhis leadershipabilities) and less rationalization(believing that the election turned more on the supportof specialinterests and the inadequacyof voters).

    Perceptionsof President Bush'slegitimacy immediatelyfollowingthe disputed 2000 election follow a differentpat-tern. Bush supporterswere predictablymore inclined thanGore supportersto accordlegitimacyto the new presidentand, unlike what we saw previously, he coefficientfor win-ning/losing remains positive and significantin our multi-variate models. As for the impact of election context, bothcongratulation-rationalizationnd electoral fairness havesignificant ndependenteffects n the multivariatemodels aswell; that is, legitimacy was highest among people whothought the election was fairlyrun and believed that theoutcome reflected voters' conscious choices rather thantheirinabilityto understandor appreciate mportant ssues.Attitudesabout the election do not, however, mediate thepreference-legitimacyrelationshipas they did with trust,responsiveness, and satisfactionwith democracy; in fact,choice satisfaction appears to have a trivial independenteffect on Bush legitimacy.The distinctivefindingsin our analysisof the legitimacyvariable are understandable n that this measure probablycaptures respondents' attitudes toward the new presidentmorethan their evaluationsof the office of the presidencyorthe government n general.Oursurvey questions relating otrust, responsiveness,and satisfactionwith democracyaskpeople to think about their regime-the government andthe way democracyworks-while the item on presidentiallegitimacyrefers to a particular ncumbent who had onlyrecentlytaken office followinga protractedand controver-sial post-election legal battle. Attitudes about the electoralprocess (specifically, ongratulation-rationalizationnd per-ceptions of electoral fairness)may have influenced assess-ments of this incumbent'slegitimacy,but they did not sig-nificantly undermine the basic direct effect of candidatepreference winning or losing).In contrast,the positivecorrelationbetween Bushprefer-ence and other political support variables disappears inmultivariatemodels that include congratulation-rationaliza-

    tion, providing support for H2. Those who preferredGorerationalizedthe election outcome more than did Bush sup-porters, of course, but individuals who acknowledged thevoters'competenceor discounted Gore'sdifficulties n com-municating his message displayed higher levels of trust,responsiveness,and satisfaction with Americandemocracy,regardlessof which candidate they preferred,than mightotherwise have been the case.On the otherhand, the limited effects of election fairnessareespeciallynotableconsideringthe context of our survey;that is, one might expect that if concerns about fairnesswould ever have effects on political support, it would beamongFloridiansshortlyafterthe 2000 presidentialcontro-versy In fact, support for H3 was weak, as perceptions offairness were relatedto political support and mediated therelationshipbetween candidatepreferenceand satisfactionwith democracy-but did not do so with respectto trust orresponsiveness.Finally,choice satisfactionwas a significantpredictor of trust, responsiveness, and satisfaction withdemocracy, but also did not mediate the relationshipbetween those variables and candidate preferenceas pre-dicted by H4. By and large, these findings suggest thatamongcitizenattitudes about the election process,it is con-gratulation-rationalizationhatplays the preeminentrole in

    14 SeeAppendixA forquestionwordings and responsecodes. In order toverify that the relationshipsbetween candidate choice and attitudesabout the election context in our 2001 statewidesurveyare not simplya function of partisanship,we estimated a multivariatelogit model ofeach variableas a function of winning/losing, partyidentification,andgeneral nterest n politics.As shown in AppendixB, TableBI, supportfor the winner is stronglyand positively correlatedwith congratulation-rationalization, choice satisfaction,and perceptions of fairness, evenwith the effects of other variables taken into account. Although parti-sanship is significantin each instance (with Republicans eaning moreheavily toward a mandate explanation, and scoring higher on bothchoice satisfactionand electoral fairness),the importance of winning/losing remains clear.

  • 8/3/2019 Winners Losers and Election Context

    11/15

    588 POLITICAL ESEARCH UARTERLY- TABLE 6

    MULTIVARIATEODELSOF POLITICALUPPORTSatisfactionwith

    PoliticalTrust Responsiveness Democracy BushLegitimacyCoeff. std err sig. Coeff. std err sig. Coeff. std err sig. Coeff. std err sig.VotePreference 0.20 0.26 0.43 -0.01 0.25 0.98 -0.13 0.26 0.61 2.43 0.37 0.00

    Congratulation-Rationalization 3.27 0.56 0.00 2.33 0.54 0.00 1.69 0.56 0.00 4.56 0.76 0.00-2 log likelihood 2006.79 1643.70 1193.26 942.59PseudoR2 0.20 0.09 0.04 0.55VotePreference 1.08 0.18 0.00 0.57 0.17 0.00 -0.00 0.18 0.99 3.60 0.31 0.00ElectionFairness 2.15 0.33 0.00 1.49 0.31 0.00 2.58 0.33 0.00 2.47 0.39 0.00-2 log likelihood 2006.79 1643.70 1193.26 942.59PseudoR2 0.21 0.10 0.15 0.56VotePreference 1.14 0.18 0.00 0.66 0.17 0.00 0.35 0.17 0.05 3.83 0.31 0.00Satisfactionwith

    Choice 2.07 0.33 0.00 1.33 0.30 0.00 0.97 0.32 0.00 0.16 0.38 0.68-2 log likelihood 2006.79 1643.70 1193.26 942.59PseudoR2 0.21 0.09 0.04 0.50Note:Dataare from the 2001 FloridaVoter urveyof registeredvoters.Tableentries are ordered ogit coefficientsand associated standarderrors(missingdataimputed;threshold evels not shown). High scores indicate supportfor the winningcandidate(includingBushvoters and non-voterswho indicated that theypreferredBush),Republican dentification,greater nterest, greater atisfactionwith the choice of candidates,greaterperceived fairness,and a greaterdegreeof congratulation believing that the election outcome reflected a positivevote for the winner, his policies, and his leadership abilities) and less rationaliza-tion (believing that the electionturned more on the supportof specialinterestsand the inadequacyof voters).

    mediatingthe bivariaterelationshipbetween candidatesup-portand political supportobservedin Table3.15The typicalconcerns one might have about the general-izabilityof relationshipsobservedin any specific cross-sec-tionalsurveyare exacerbated n this case, given the extraor-dinary circumstancesin which the winner of the popularvote nationallyended up losing the electoralcollegeand thepresidencyitself aftera protracted egal dispute. However,datafrom both the 2004 NESand the survey done follow-

    ing the 1998 Florida gubernatorialelection provide addi-tional support for some of our key findings.In Table7, wepresentestimates of ordered-logit regressionsof our politi-cal support variables on candidate preferenceand percep-tions of fairness using data from the 2004 AmericanNationalElectionStudy On balance,the findingshere havemore similarities than differences with the statewide pat-terns evident in Table 6. As was true in Florida,all threenational-level regressions indicate that respondents whoregarded he election as fair exhibitedhigherlevels of polit-ical support, controlling for candidate preference.As wasalso seen for two of the threeregressions n the second rowof Table6, however,perceptionsof fairness did not obviatethe relationshipsbetweencandidatepreferenceand politicalsupport:Bushsupportershad higherlevels of trust and per-ceivedresponsivenesseven aftercontrollingfor theirgreatertendency to regardthe 2004 election as fair. These resultsoverall suggest that perceptions of fairness are related tosupport, but do not consistently mediate the relationshipbetween candidatepreferenceand broader attitudes aboutgovernmentand the politicalsystem as indicatedin H3.NeitherNES nor any other national omnibus surveys ofwhich we are awarehave askedrespondentsto assessexpla-nationsof electoral outcomes. Fortunately,he 1998 Floridadata enable us to examine relationshipsbetween candidatepreference, congratulation-rationalization, and politicalsupport in the context of a closely fought and contentiousgubernatorialelection which occurred two years beforerecountsof ballotswith hangingchadsput the state's lectoral

    " Wealso did some empiricaldiagnostics o assess the possibilityof recip-rocal effects in the models reported in Table6, using two-stage leastsquares(2SLS)on one imputed dataset. Our first-stageestimateswerelinear combinations of party dentification,politicalinterest,education,age, age squared, black, Hispanic, male, and response to a questionaboutwhether the respondent's ote was more for the lesser of the evilsor a positivevote for the preferred andidate.As we report n AppendixB, TableB2, the 2SLSresults show that congratulation-rationalizationhas significant,positive, and independenteffects on trust, responsive-ness, and satisfactionwith democracy,and the effect of vote preferenceon those variables either falls to zero (within confidence bounds) orchanges sign. Inotherwords,our basicfindingsare robust when we usea 2SLSmodel to controlforreciprocaleffectsof supporton congratula-tion-rationalization.Ideally, we would have preferred a two-stageordered ogit model with a correction or the uncertainty ntroducedbyusing an instrumentalvariable,but we are unawareof any routineinstandard tatisticalpackages includingR)thatmakes such a correction.However,results fromtwo-stageordered ogit models withoutcorrectedstandarderrors are substantively dentical to the 2SLSresults(and areavailableupon request).The robustness of our results across methodshelps allay concerns about violations of assumptionsin any particularestimation.

  • 8/3/2019 Winners Losers and Election Context

    12/15

    WINNERS,LOSERS,AND ELECTIONCONTEXT 589- TABLE7

    MULTIVARIATEODELSOF POLITICALUPPORTPoliticalTrust Responsiveness Satisfactionwith Democracy

    Coeff. std err sig. Coeff. std err sig. Coeff. std err sig.Vote Preference 1.10 0.15 0.00 0.38 0.15 0.01 1.26 0.17 0.00Election Fairness 1.57 0.30 0.00 1.90 0.27 0.00 2.20 0.29 0.00-2 log likelihood 2256.16 2239.52 1843.71PseudoR2 0.18 0.12 0.25Note: Data are from the 2004 NES.Tableentries are orderedlogit coefficients and associated standarderrors(missingdataareimputed;thresholdlevels notshown). High scores indicatesupportfor the winning candidate(includingBushvoters and non-voters who indicated that they preferredBush)and greaterperceivedfairness.

    system in the nationalspotlight. Our measure of congratu-lation-rationalizationn this survey reflects the particularthemes of the 1998 campaignand post-election analyses.Specifically, eople with low scoresbelieved thatRepublicanbusinessmanJebBushdefeatedDemocraticLieutenantGov-ernorBuddyMacKaybecause of his famousname, disarrayamongthe Democrats,help fromspecialinterests,and voterignorance;they tended to rejectthe notion that Bush'svic-tory was attributableeither to his conservativephilosophyor to voters' beliefs that he was better able than his oppo-nent to deal with the state'sproblems. Those who scoredhigh on this index felt the outcome hinged on Bush'scharisma, his reaching out to minorities, his post-1994movementtowardthe ideologicalcenter,and his opponent'sunrelenting negativism.16In the 1998 survey,those who had voted for the newgovernor-elect coredsignificantlyhigheron our measure ofpolitical trust (.392 to .321, t = 2.658, p < .01). However,they also scored significantly higher on congratulation-rationalizationvariable(.734 to .371, t = 19.87, p < .01)and, as we saw in the aftermath of the 2000 presidentialcontest, winning/losing in the 1998 gubernatorialelectionwas statisticallyunrelated to political trust when we con-trolledfor explanationsof the outcome (see Table8). Onceagain,then, H2 is supported.

    CONCLUSION

    Elections,even controversialones (Priceand Romantan2004), are often characterizedas legitimatinginstitutions,especially for citizens who supported the losing candidateand whose willingness to accept the final verdict is sup-posedly enhanced by the sense that their views have beengiven a fair and proper hearing during the course of thecampaign.Yet,consistent with prior research,our analysesuncoveredsome fairlysubstantialdifferencesbetween win-ners and losers. On the whole, the latter tend to be lesstrustful, ess certain of the responsivenessof governmentto

    popular concerns, less satisfied with the way democracyisworkingin the United States, and, at least in the 2000 pres-idential election, less inclined to extend legitimacy to thevictorious candidate.In this article,we have suggested that citizens' attitudesand beliefs about the election context help to explain thedifferential response of winners and losers. Our resultsindicate, for example, that even when controlling for theeffects of party dentification,supportersof GeorgeW Bushwere more likely than those who backed Al Gore to (1)believe that the election outcome represented a genuinemandate from voters, (2) express satisfaction with thechoice of candidatespresentedto voters in the presidentialrace, and (3) offer more generous assessments of whetherthe end result was achieved fairly More importantly,wefound that one of these attitudes, the interpretationof theelection, served as a link between candidatepreferenceandfeelings of political support. In other words, we read ourresults as evidence that Bush supportersexhibited higherlevels of support becauseof their understanding of whytheir candidatewas victorious.Perhapsit is not surprisingthat supportersof the win-ning candidatewould echo his or her campaignthemes andself-congratulatory xplanationsof th election outcome, northat those who supported the defeated opponent wouldfault the campaign, media coverage,voters' lack of intelli-gence, orjust the dumb luck of intrudingevents. Neverthe-less, Bennett(1980) reminded us that we should not mis-take the banalityof politicalaccounts for ineffectiveness.Nomatter how trite, elites' situationally appropriate explana-tions of political behavior can be powerful in helping toforge a bond between them and their supporters. Ourresults suggest that much the same is true with regardtovoters'explanationsof mass electoralbehavior:Simple,evensimplistic,understandingsof whya preferredcandidate ostcan depresscitizens'support for the political system.The problemis thatelections aresupposed to help legit-imize government primarily n the eyes of citizens who sup-portedthe losingcandidate.While electionsmaystillbe ableto serve that function if losers are persuaded that the out-come was a true reflection of voterpreferences, hat did nothappen very often among our respondents. If we are read-

    ing the datacorrectly, t seems likely that losing candidates16Our 1998 measure of congratulation-rationalization sed the same

    question wording and response format as our 2001 measure. SeeAppendixC.

  • 8/3/2019 Winners Losers and Election Context

    13/15

    590 POLITICALESEARCHUARTERLY=_TABLE 8

    MULTIVARIATEODELS F POLITICALUPPORTPoliticalTrust ExternalEfficacy

    Coeff. std err Sig. Coeff. std err sig.VotePreference -0.13 0.25 0.60 -0.03 0.22 0.89Congratulation-Rationalization 1.48 0.52 0.01 0.72 0.45 0.11-2 log likelihood 1450.30 2446.00Pseudo R2 0.03 0.01Note:Data re rom he 1998Florida oterurvey fregisteredoters.Table ntries reorderedogitcoefficientsndassociatedtandardrrorsmissing ataimputed;hresholdevelsnotshown).High cores ndicateupportor hewinning andidateincluding ush oters ndnon-voters ho ndicatedhat heypreferredush) nda greater egree fcongratulationbelievinghat he election utcome eflected positive ote for hewinner, ispolicies, ndhis lead-ershipabilities)nd ess rationalizationbelievinghat he election urnedmoreon thesupport f specialnterests nd the nadequacyfvoters).

    and other opinion leaderscontribute to the maintenanceofwidespread political cynicism in Americaby encouragingsome citizens to view elections (at least the ones they don'twin) in a very different ight than that imaginedby demo-cratictheorists.We maintain that the media'susual focus oncampaign strategy (see Fallows 1996; Cappella andJamieson 1997), and especially on the horserace,encour-agesvotersto view the competition througha similar ens-and therebymakes it more difficult for them to recognizethe existence of any fit between election outcomes andeither mass policy preferencesor assessments of candidatequalities.Likewise,losing candidates and their co-partisanapologistsare sometimes disinclined (Kingdon 1966), per-haps even more so todaythanin the past (Fiorina2005), toconcede that such fits exist and help to explain why elec-tions turn out the way they do. Rationalizationsmay pro-vide a psychological salve that helps losers to justify theexpenditureof time, effort,and money in their own minds,but we suspectthatthey do littleto instill public faith n theinstitutionsof government.It is, we would suggest, no accident that self-identifiedRepublicansand Democratsalike (as well as Independents)expressmorenegativeorientations owardgovernment,andtowardthe political process generally, han used to be thecase.'7Partof this shift maybe a function of polarizedelitesrepeatedlyalternatingbetween self-congratulatoryprocla-mationsof popularmandates and dismissiveinterpretationsof electoralsetbacks. If so, a little more humility on bothsides might go a long way to reconcilingthe two camps.

    APPENDICES

    The followingappendicesareavailableathttp://www.clas.ufl.edu/users/martinez/prq06/Appendix A: Question Wording and Response Codes, 2001Florida StateSurvey.Appendix B Table B : Explaining Attitudes about ElectoralContext.

    AppendixB:TableB2:Two-StageeastSquaresModels fPoliticalSupport.AppendixC:Congratulation-Rationalizationndexfor 1998 Guber-natorial urveyAppendixD: Codebookorthe March 001 FloridaVotersurvey.AppendixE:Source ode foranalyses.

    REFERENCES

    Abramowitz,Alan I., and Kyle L. Saunders. 1998. "IdeologicalRealignment in the U. S. Electorate."Journalof Politics 60(August):634-52.Alford,John R. 2001. "We'reAll in This Together:The Decline ofTrust in Government, 1958-1996." In John R. Hibbing andElizabethTheiss-Morse,eds., WhatIs ItAboutGovernmenthatAmericansDislike?,pp. 28-46. Cambridge, MA: CambridgeUniversityPress.American National Election Studies. 2005. The ANES Guide toPublicOpinionand ElectoralBehavior.Ann Arbor:Universityof Michigan, Center for Political Studies [producer and dis-tributor]. www.electionstudies.org).ccessed8 December2005.Anderson,Christopher.,AndreBlais,ShaunBowler, oddDono-van, and Ola Listhaug. 2005. Losers' Consent:ElectionsandDemocraticLegitimacy.Oxford:OxfordUniversityPress.Anderson,Christopher ., and ChristineA. Guillory.1997. "Politi-cal Institutions and Satisfaction with Democracy: A Cross-National Analysis of Consensus and MajoritarianSystems."AmericanPoliticalScienceReview91 (March):66-81.

    Anderson, ChristopherJ., and AndrewJ. LoTempio.2002. "Win-ning, Losing, and Political Trust n America."British ournalofPoliticalScience32 (April):335-51.Anderson,Christopher ., andYuliyaV Tverdova.2001. "Winners,Losers, and Attitudes about Government in ContemporaryDemocracies."International oliticalScienceReview22 (Octo-ber): 321-38.

    Bennett, W Lance. 1980. "The Paradoxof Public Discourse: AFrameworkor the Analysis f PoliticalAccounts."ournal fPolitics42 (August):792-817.Botti, Simona, and Sheena S. Iyengar.2004. "ThePsychologicalPleasurend Painof Choosing:WhenPeoplePreferChoosingat the Cost of SubsequentOutcomeSatisfaction."ournalofPer-sonalityand SocialPsychology7 (September):312-26.Canache,Damarys,JefferyK. Mondak, and MitchellA. Seligson.2001. "Measurement nd Meaningin Cross-NationalResearch

    7 SeeAlford2001:39-41),Craig1993:39-41),andAmerican ationalElection tudies2005).

  • 8/3/2019 Winners Losers and Election Context

    14/15

  • 8/3/2019 Winners Losers and Election Context

    15/15

    592 POLITICALRESEARCH QUARTERLY

    Tyler,Tom R., Kenneth A. Rasinki, and Kathleen M. McGraw.1985. "The Influence of PerceivedInjustice on the Endorse-ment of Political Leaders." ournalof AppliedSocialPsychology15 (8): 700-25.VanBuuren,Stef, and KarinOudshoorn. 1999. FlexibleMultivari-ateImputationyMICE.Leiden:TNO Preventieen Gezondheid.

    Received:anuary 0, 2006Acceptedor Publication:March 3, [email protected]@[email protected]@bellsouth.net