Is Anybody Listening Evidence That Voters Do Not Respond to European Parties Policy ...

13 Pages·2011·0.18 MB·English

Is Anybody Listening? Evidence That Voters Do Not
Respond to European Parties’ Policy Statements
During Elections
James Adams
University of California at Davis
Lawrence Ezrow
University of Essex
Zeynep Somer-Topcu
Vanderbilt University
AlthoughextensiveresearchanalyzesthefactorsthatmotivateEuropeanpartiestoshifttheirpolicypositions,thereislittle
cross-nationalresearchthatanalyzeshowvotersrespondtoparties’policyshifts.Wereportpooled,time-seriesanalysesof
electionsurveydatafromseveralEuropeanpolities,whichsuggestthatvotersdonotsystematicallyadjusttheirperceptions
ofparties’positionsinresponsetoshiftsinparties’policystatementsduringelectioncampaigns.Wealsofindnoevidence
thatvotersadjusttheirLeft-Rightpositionsortheirpartisanloyaltiesinresponsetoshiftsinparties’campaign-basedpolicy
statements.Bycontrast,wefindthatvotersdorespondtotheirsubjectiveperceptionsoftheparties’positions.Ourfindings
haveimportantimplicationsforpartypolicystrategiesandforpoliticalrepresentation.
R
esearch on political representation in Europe policy debates over Left-Right social welfare issues (see,
emphasizes the linkages between parties’ policy e.g.,Dalton1985).
positionsandtheirsupporters’policybeliefs.Ac- In an evolving political environment wherein both
cording to this responsible party model of political rep- partiesandvotersshifttheirpolicypositions,policycorre-
resentation, it is normatively desirable that parties’ pol- spondencebetweenpoliticalpartiesandtheirsupporters
icy programs—and governing parties’ policy outputs— can be maintained through some combination of party
match the views of the party’s supporters, a desidera- elitesrespondingtotheirsupportersandthesesupport-
tum that reflects Sartori’s observation that “citizens in ersrespondingtopartyelites,i.e.,elitesmaydynamically
Westerndemocraciesarerepresentedbyandthroughpar- adjusttheirpolicypositionsinresponsetoshiftsintheir
ties.Thisisinevitable”(1968,471,emphasisinoriginal). supporters’beliefs,aprocesswelabelpartyresponsiveness,
Over the past 30 years, dozens of studies have analyzed andpartysupportersmaydynamicallyadjusttheirbeliefs
the mass-elite policy linkages that the responsible party inresponsetoshiftsintheirpreferredparty’spolicypo-
model highlights (see, e.g., Dalton 1985; Iversen 1994; sitions,aprocesswelabelpartypersuasion.Alternatively,
Powell 1989).1 These studies typically report reasonably rank-and-filevotersmayswitchtheirpartisanloyaltiesin
close matches between parties’ positions and their sup- responsetoparties’policyshifts,i.e.,votersmayengagein
porters’ policy preferences, particularly with respect to policy-basedpartisanswitching.Thelattertwoprocesses
JamesAdamsisProfessorofPoliticalScience,UniversityofCaliforniaatDavis,Davis,CA95616([email protected]).LawrenceEzrow
isaReaderinEuropeanPolitics,UniversityofEssex,WivenhoePark,ColchesterCO43SQ,UnitedKingdom([email protected]).Zeynep
Somer-TopcuisAssistantProfessorofPoliticalScience,VanderbiltUniversity,Nashville,TN37235([email protected]).
Earlier versions of this article were presented at the 2008 “Politics of Change” Workshop at the Vrije Universiteit, Amsterdam, at the
2008annualmeetingoftheMidwestPoliticalScienceAssociation,andatthe2010“PoliticalPartiesandComparativePolicyAgendas’
Workshop”attheUniversityofManchester.WethankJohnBartle,ThomasBra¨uninger,GaryMarks,SlavaMikhaylov,RandySteven-
son,HeatherStoll,GuyWhitten,andthreeanonymousreviewersforvaluablecomments.Allthreeauthorscontributedequallytothe
article.TheauthorsacknowledgetheUKEconomicandSocialResearchCouncil(ESRC)forfundingsupport(RES000–22-2895,“Sub-
constituency Representation across Western Europe”). Data and supporting information for this project can be found on the website
http://privatewww.essex.ac.uk/∼ezrow/.
1Analternativerepresentationcriterionemphasizesthelinkbetweengovernmentpoliciesandthemedianvoterpositionintheelectorate
(see,e.g.,McDonaldandBudge2005;Powell2000).
AmericanJournalofPoliticalScience,Vol.55,No.2,April2011,Pp.370–382
(cid:3)C2011,MidwestPoliticalScienceAssociation DOI:10.1111/j.1540-5907.2010.00489.x
370
371
ISANYBODYLISTENING?
bothinvolvevotersrespondingtoparties,andwecollec- actualstatementsintheparties’policymanifestos,which
tivelylabeltheseasprocessesofpartisanadjustments. formthebasisfortheparties’electioncampaigns.
Weanalyzethedynamicsofvoters’responsestoshifts
inEuropeanparties’Left-Rightpositions.Weaskthefol-
lowingquestions:WhenpartiesshifttheirLeft-Rightpol-
ShouldWeExpectVoterstoReact
icy statements, as reflected in their election manifestos,
toParties’PolicyStatements?
do citizens update their perceptions of the parties’ Left-
TheoreticalArguments
Rightpositions?And,doweobservepartisanadjustments
inresponsetoparties’shiftingpolicystatements,i.e.,do
citizens respond to parties’ statements by shifting their Wefirstconsiderthequestion:shouldweactuallyexpect
own Left-Right positions (a persuasion process) and/or citizenstoupdatetheirperceptionsofparties’Left-Right
theirpartisanloyalties(partisanswitching)? positionsinresponsetopoliticalelites’policystatements?
Thesurprisinganswerweprovidetoeachoftheabove More specifically, given that our empirical measure of
questions is no. We find no substantively or statistically the Left-Right tone of party policy statements is based
significant evidence that voters adjust their perceptions oncontentanalysesoftheirelectionmanifestos,4weask:
ofparties’Left-Rightpositionsinresponsetothepolicy shouldweexpectvoters’perceptionsofparties’ideologies
statementsinparties’electionmanifestos—aconclusion tovarywiththeLeft-Righttoneofthestatementsinthese
that is striking given that interviews with European po- manifestos? We believe this is an interesting question,
liticalelitesthatweconducted(discussedbelow)suggest because there are considerations that point in conflict-
thatpartiescampaignonthebasisofthesemanifestos.We ing directions. The two considerations that suggest that
alsofindnoevidenceofcitizens’partisanadjustmentsin voters will indeed react to parties’ election manifestos
response to parties’ policy statements, i.e., we find no areasfollows.First,partiestypicallypublishtheirman-
evidence that voters adjust their Left-Right positions or ifestos in the run-up to the national election campaign,
their partisan loyalties in response to these policy state- andpreviousstudiesfindthatvotersdisplayheightened
ments. This latter conclusion holds both for analyses of interest in (and awareness of) political information that
national election surveys from five European countries2 appearsaroundthetimeofanationalelection.Inpartic-
andforseparateanalysesofEurobarometerdatafrom12 ular,Andersen,Tilley,andHeath(2005)reportanalyses
countries over the period 1973–2002.3 Simply put, we of the British electorate over the period 1992–2002 that
findthatwhenpartiesshiftthestatementsintheirpolicy demonstratethatcitizens’informationabouttheparties’
programs—statements that form the basis for the par- policypositionspeakedaroundthetimesofthenational
ties’electioncampaigns,accordingtothepartyeliteswe elections held during this period (1992, 1997, 2002).5
interviewed—thereisnoevidencethatvotersrespondby Andersen,Tilley,andHeath’sfindingssupporttheargu-
adjustingtheirperceptionsoftheparties’Left-Rightpo- mentspresentedbyGelmanandKing(1993),thatcitizens
sitions, their own Left-Right positions, or their partisan becomemoreenlightenedduringthecourseofnational
loyalties. electioncampaignsduetotheintensityofpoliticalmedia
Bycontrast,wefindthatEuropeancitizensdoreactto coverageduringtheseperiods,andalsobecausecitizens
theirperceptions ofparties’Left-Rightpositions,i.e.,cit- areespeciallymotivatedtoseekoutpoliticalinformation
izensadjusttheirLeft-Rightpositionsandtheirpartisan during election campaigns. Thus, while only a fraction
loyaltiesinresponsetotheparties’policyimages.Overall, ofrank-and-filevoterspresumablyreadtheparties’elec-
ourfindingstherebysuggestthatLeft-Rightideologydoes tionmanifestos,wemightexpectcitizenstobeespecially
matter tovoters,andthattheyreacttoparties’perceived attentive to media coverage of these manifestos because
ideologicalshifts.But,becausevoterperceptionsdonot theyarepublishednearthetimeofnationalelections(see
tracktheparties’actualpolicystatements,thereisadis- Topf1994).
connectbetweenshiftsinelitepolicydiscourseandvoter
reactions. Voters react to what they perceive the parties
4Aswediscussbelow,theComparativeManifestoProject(CMP)
standfor,buttheseperceptionsdonotmatchupwiththe
codingsofpartypolicymanifestosprovidetheonlyavailableover-
timeestimatesoftheLeft-Righttoneofparties’policystatements,
2ThefivecountriesareBritain,Germany,theNetherlands,Sweden, whichgeneralizeacrosstheEuropeanpartysystemsandthetime
andNorway,thefiveEuropeanpolitiesforwhichnationalelection periodsincludedinourstudy.
surveydataareavailableoveralengthytimeperiod.
5We note that the national election survey data that we analyze
3The12countriesthatareincludedintheEurobarometersurveys belowwerecollectedshortlyafterthenationalelectioncampaign
are Britain, France, Germany, the Netherlands, Portugal, Spain, tookplace,sothatrespondents’politicalinformationlevelsshould
Denmark,Luxembourg,Greece,Ireland,Belgium,andItaly. peakaroundthetimeofthesesurveys.
372
JAMESADAMS,LAWRENCEEZROW,ANDZEYNEPSOMER-TOPCU
Second, although there is no extant study that an- than 100,000 copies of the manifesto that were
alyzes whether the policy statements in parties’ election distributed....Andofcoursethepartyactivists
manifestosmatchtheirstatementsinothercampaignfo- read it. The manifesto basically is an outline of
rums(i.e.,politicaladvertisements,candidates’speeches, the kinds of arguments that we do share so it
etc.), we conductedinterviews with 26 party elites from basicallyharmonizesthemessageoftheparty.9
Germany,theNetherlands,andAustriaduringthesum-
mer of 2008, in which these elites stated that their par- Whiletheconsiderationsdiscussedabovesuggestrea-
tiesmakedeterminedeffortstocampaignbasedontheir sons why citizens’ perceptions of party ideologies may
policymanifestos.6 Thisisimportantbecauseitsuggests trackthestatementsinparties’electionmanifestos,other
that parties’ manifesto-based policy statements roughly considerations cast doubt on parties’ abilities to shape
reflect the policies they communicate to the public via theirpolicyimagesviatheircampaign-basedpolicystate-
other avenues, such as campaign advertisements, party ments. First, there is extensive empirical research that
elites’campaignspeeches,andmediainterviews.Typical argues that large segments of the mass public are inat-
werethefollowingexchanges: tentive to and uninformed about political issues (e.g.,
Converse 1964; Zaller 1992), so that even if the elec-
Q: Does the election campaign usually follow torate is relatively more politically informed around the
whattheelectionmanifestosays? timeofnationalelections—asthestudiesdiscussedearlier
A:Itisveryimportanttohave,withintheparty, suggest—significantsegmentsoftheelectoratemaynev-
and also with the local members...to have a erthelessbeinattentivetocampaign-basedmediacover-
view and an awareness of issues which are im- ageofparties’policystatements.Second,aseparatestrand
portantforusandforourvoters.Andtheyhave of research suggests reasons why even those voters who
to be in everything we do. So, they have to be registertheparties’shiftingpolicystatementsmayresist
central points in our election programs and in updating their perceptions of parties’ positions: specifi-
ourmaterialsandinspeechesontelevisionand cally,thisresearchdocumentsassimilation/contrasteffects
thingslikethat.7 wherebyvotersaligntheircurrentperceptionsofthepar-
Q: And how much do you think the election ties’positionswiththeirpreexistingpoliticalloyalties(see,
manifestoandtheissuesintheelectionmanifesto e.g.,AchenandBartels2006;GerberandGreen1999).
stay the same for the CDA during the electoral Thereareadditionalreasonswhypoliticalpartiesmay
campaign?Isitthebasisforthecampaignordoes face difficulties in shaping their policy images via their
itchangealot? campaign-based policy statements. Parties’ new policy
A:No,no,itis[thebasisforthecampaign].We pronouncementsoftenprovokerivalpartyelitestopub-
havesomeessential,veryimportantpartsofthe liclyderidethefocalparty’snewinitiativesasbeing“op-
party program which are fundamental for the portunistic,”“pandering,”“insincere,”“unrealistic,”and
imageoftheCDA....So,theissuesintheparty so on, and these characterizations may confuse voters
programareimage-buildingissues.So,thereisa aboutthefocalparty’sactualpolicyintentions.Further-
lot in the manifestos and there are some things more, members of the news media may weigh in with
whicharefundamentalforourimage.8 commentary about whether the focal party’s new pol-
Q:So,howmuchdoyouthinktheelectionpro- icy initiatives represent a “tactical” change in strategy
gramisreflectedintheelectoralcampaign?Be- as opposed to a fundamental change in policy direc-
cause,nobodyinthepublicreallyreadstheelec- tion; whether the policy shift is permanent or whether
tion program, so do you think that the party it merely represents a temporary victory for one of the
closelyfollowstheelectionprogram? party’sfactions;andwhether,intheeventthepartyenters
A: It’s not true that nobody reads the man- thegovernment,theparty’sleaderswillhavetheresolveto
ifesto. We’ve had, sometimes we’ve had more actuallypushthroughthepolicyreformstheyhaveadvo-
catedwhileinopposition(see,e.g.,Stokes2001).Inaddi-
tion,duringnationalelectioncampaignsboththemedia
6These elite interviews encompassed party leaders, members of
andopposingpartiesmayemphasizeaspectsofthefocal
parliament, campaign directors, and directors of think tank or-
party’s policy program that differ from the planks that
ganizations.Aprotocoloftheseinterviewswasthatwekeepthe
subjects’identitiesconfidential. the party itself prefers to emphasize, i.e., parties cannot
7InterviewwithapoliticianfromtheDutchGreenParty(June26, unilaterallysetthe“termsofthedebate”fortheelection
2008). campaignviatheirpolicystatements.
8InterviewwithapoliticianfromtheDutchChristianDemocratic
Party(June26,2008). 9InterviewwithapoliticianfromtheGermanGreens(July3,2008).
373
ISANYBODYLISTENING?
TheLinkbetweenParties’Statements periods.12 In each of these national election studies, re-
andTheirPolicyImages:Empirical spondentswereaskedtolocateallofthemajorpartieson
anideologicalscaletypicallyrangingfrom1(extremeleft)
Analyses
to 10 (extreme right).13 We used these survey responses
tocomputetheperceivedpartypositionforeachpartyin
WerequirelongitudinalmeasurementsoftheLeft-Right
eachelection.
toneofpartypolicystatementsalongwithvoters’percep-
tionsoftheparties’positionsinordertoanalyzewhether
ModelSpecification
parties’policyimagesshiftalongwiththeirpolicystate-
ments. Our measures of the Left-Right tone of parties’
We specify a multivariate regression model in order to
policy statements are based on the Comparative Mani-
evaluate whether parties’ policy images shift in tandem
festo Project’s (CMP) Left-Right codings of the parties’
withtheCMPcodingsoftheLeft-Righttoneofthestate-
electionmanifestos(Budgeetal.2001).Thesemanifestos
mentsintheparties’electionmanifestos.Thedependent
aretypicallypublishedintherun-uptothenationalelec-
variable is the change in the party’s policy image at the
tion,and,asdiscussedabove,oureliteinterviewssuggest
current election compared to its image at the previous
thatpartiesusethesemanifestosasthebasisfortheirsub-
election, as measured by NES respondents’ mean Left-
sequentelectioncampaigns.Theproceduresusedtomap
Rightplacementsoftheparty.Welabelthisvariable[Party
parties’policypositionsfromtheirelectionprogramsare
j’s perceived Left-Right shift (t)], to denote the shift in
described in detail in several of the CMP-related publi-
thefocalpartyj’sperceivedpositionbetweenthecurrent
cations,sothatweonlybrieflyreviewtheprocesshere.10
election t and the previous election t−1. Our key inde-
The coders match up quasi-sentences in the policy pro-
pendentvariablesare[Partyj’sstatedLeft-Rightshift (t)],
gramwithacategoryofpolicy(e.g.,welfare,defense,law
which denotes the shift in the CMP’s Left-Right coding
and order, etc.) and take the percentages of each cate-
of party j’s election manifesto between the current and
gory as a measureof the party’s priorities. Based on the
thepreviouselection,and[Partyj’sstatedLeft-Rightshift
mixture of policy priorities, the authors develop an in-
(t−1)], which denotes the shift in the CMP’s coding of
dex that measures the overall ideology for the program
partyj’smanifestobetweenthepreviouselectionandthe
ofeachpartyineachelectionyear.Theideologicalscores
electionbeforethat.Weincludethislattervariabletocon-
range from −100 to +100, with higher scores denoting
trolforthepossibilitythatsignificanttimelagsintervene
amoreright-wingemphasis.WehaverescaledtheCMP
beforecitizensprocessparties’policystatements(see,e.g.,
partyscorestoa1–10Left-Rightscale,sothattherangeof
Adams and Somer-Topcu 2009; Erikson, MacKuen, and
thesescoresmatchesthatofthevoters’perceptionsofthe
Stimpson2002).Wealsoincludealaggedversionofthe
parties’Left-Rightpositions(describedbelow).Welabel
dependentvariableinordertocontrolforautocorrelation
theCMP-basedcodingsofpartymanifestosastheparty’s
issues.14
statedposition.11
Ourspecificationfortherelationshipbetweenvoters’
Our measures of parties’ policy images, i.e., of vot-
perceptions of the parties’ Left-Right positions and the
ers’ subjective perceptions of the parties’ Left-Right po-
Left-Right tone of the party’s election manifesto is as
sitions,arederivedfromNationalElectionStudy(NES)
follows:
datafromGermany,GreatBritain,theNetherlands,Nor-
Partyj’sperceivedLeft-Rightshift(t)
way, and Sweden—the five Western European democ-
racies for which NES data are available for protracted =b +b [Partyj’sstatedLeft-Rightshift(t)]
0 1
+b [Partyj’sstatedLeft-Rightshift(t−1)]
2
+b [Partyj’sperceivedLeft-Rightshift(t −1)]
3
10Foramorethoroughdescriptionofthecodingprocess,seeAp- (1)
pendix2inBudgeetal.(2001).
11Wenotethatthereisawide-ranging,oftenheateddebateover
thevalidityoftheCMPLeft-Rightcodingsofpartyprograms,with 12Thelistofelectionsincludedinthesurveysisgivenintheap-
particularemphasisonwhetherthesecodingscaptureparties’Left- pendix.
Rightpolicypositionsasopposedtoparties’issuepriorities(see,e.g., 13IncaseswheretheLeft-Rightscalewasnot1–10,werecalibrated
Jandaetal.1995;McDonaldandMendes2001).Whileitisoutside
respondents’partyplacementstoa1–10scale.
the scope of this article to review this debate, we note that the
CMPcodingshavebeenwidelyusedinresearchthatanalyzesthe 14Wenotethatwealsoestimatedtheparametersofthemodelwhile
relationshipbetweenparties’Left-Rightpositionsandsurvey-based omitting the lagged dependent variable, and these estimates—
measuresofvoters’ideologicalpositions(e.g.,Adams,Haupt,and whichwereportinsupplementarymaterialspostedonourweb-
Stoll2009;Ezrow2007),whichisoneofthekeyrelationshipsthat site(http://privatewww.essex.ac.uk/∼ezrow)—supportedsubstan-
weanalyzeinthisarticle. tiveconclusionsthatwereidenticaltothosewereportbelow.
374
JAMESADAMS,LAWRENCEEZROW,ANDZEYNEPSOMER-TOPCU
TABLE1 DescriptiveStatistics:Dependentand TABLE2 TheRelationshipbetweenChangesin
IndependentVariables theLeft-RightOrientationsofParty
ManifestosandShiftsinVoters’
MeanAbsolute PerceptionsofPartyPositions(N =97)
MeanValue Value
Country-
Partyjsupporters’ −0.04(0.44) 0.30(0.33)
Basic Specific
Left-Rightshift (t)
Model(1) Intercepts(2)
Partyjshift (t)– −0.03(0.42) 0.33(0.25)
voterperceptions Partyj’sstated .039(.089) .047(.089)
Partyjshift (t)– 0.01(0.58) 0.44(0.38) Left-Rightshift (t)
CMPcodings Partyj’sstated .100(.074) .103(.074)
Left-Rightshift(t–1)
Notes:Thenumbersinparenthesesarethestandarddeviationsof
Partyj’sperceived .046(.144) .032(.136)
thereportedvalues.Allofthevariablesarecalibratedalonga10-
pointscalerunningfrom1(extremeleft)to10(extremeright). Left-Rightshift(t–1)
Themeasuresofpartysupporters’Left-Rightshifts,andofvoters’ Britain .157(.344)
perceptionsofparties’Left-Rightshifts,aredrawnfromNational Germany .107(.288)
ElectionStudydatafromGermany,GreatBritain,theNetherlands,
Norway, and Sweden. The complete list of elections and parties Norway −.031(.132)
includedinthedatasetisgivenintheappendix. Sweden .068(.125)
Intercept −.017(.046) −.051(.119)
AdjustedR2 .02 .04
where
∗∗p≤.01;∗p≤.05,two-tailedtests.
Partyj’sperceivedLeft-Rightshift(t)=thechangeinparty Notes:Standarderrorsareinparentheses.Fortheseanalysesthe
j’sperceivedLeft-Rightpositionatthecurrentelectiont dependent variable was the change in voters’ perceptions of the
comparedwithitspositionatthepreviouselectiont−1, party’sLeft-Rightpositionatthecurrentelectioncomparedtothe
previous election, basedonrespondents’partyplacements from
based on NES respondents’ party placements. The vari-
National Election Study data from Germany, Great Britain, the
able[Partyj’sperceivedLeft-Rightshift (t−1)]issimilarly Netherlands,Norway,andSweden(seetheappendixforthecom-
defined. plete list of elections and parties included in the data set). The
independent variables are defined in the text. In column 2 the
Party j’s stated Left-Right shift (t) = the change in party baselinecountryistheNetherlands.
j’sLeft-Rightpositionatthecurrentelectiont compared
with its position at the previous election t−1, based on
concern for the model we estimate. However, there are
the CMP’s Left-Right codings of the policy statements
other methodological concerns to address. The lagged
intheparty’selectionmanifestos.Thevariable[Partyj’s
statedLeft-Rightshift (t−1)]issimilarlydefined. dependentvariableincludedinourspecificationhelpsto
addresstheconcernofseriallycorrelatederrors(Beckand
Katz 1995), and a Lagrange multiplier test fails to reject
thenullhypothesisofnoserialcorrelation.Anothercon-
Results
cern is that there may be unobserved, election-specific
Our analysis encompassed 97 stated Left-Right shifts factors that influence the positions of parties’ partisan
by parties and their supporters in five European party constituencies in a particular election. We address these
systems—Germany, Britain, the Netherlands, Norway, concernsthroughtheuseofrobuststandarderrorsclus-
and Sweden—for which NES data were available over teredbyelection(Rogers1993).
lengthy time periods. The complete set of parties and The parameter estimates for the alternative models
electionsincludedintheanalysesarereportedintheap- thatweinvestigatedarereportedinTable2.Column1in
pendix.Table1reportsthemeansandstandarddeviations thetablereportsparameterestimatesforamodelwhere
oftheobservedvaluesofthevariablesinourdataset. thedependentvariable,[Partyj’sperceivedLeft-Rightshift
Ouranalysesencompassed25parties,eachobserved (t)], represents the change in the mean Left-Right po-
over an average of six elections, and should thus be re- sition ascribed to the focal party along the 1–10 Left-
garded as time-series cross-sectional data. Estimating a Right scale by NES respondents at the current election,
simple regression on the pooled data can lead to erro- comparedtothemeanpositionthatallNESrespondents
neousconclusionsifthereareunobserveddifferencesbe- ascribedtothepartyatthepreviouselection.Totheex-
tweenpartisanconstituencies(Hsiao2003);fortunately, tent that voters’ perceptions of parties’ Left-Right posi-
tests for party-specific effects indicate that this is not a tionsrespondtotheactualpolicystatementstheparties
375
ISANYBODYLISTENING?
publishintheirpolicymanifestos,wewouldexpectpos- Voters’PartisanAdjustmentsto
itive and statistically significant coefficient estimates on Parties’PolicyStatements:Empirical
the [Party j’s stated Left-Right shift (t)] variable—which
Analyses
would indicate that voters adjust their perceptions con-
temporaneously with parties’ stated Left-Right shifts—
andpossiblyonthe[Partyj’sstatedLeft-Rightshift(t−1)] Theempiricalanalysesreportedabovecastseriousdoubts
on the proposition that European voters engage in sys-
variable,whichwoulddenotetimelagsbeforevotersup-
tematic partisan adjustments in response to parties’
date their perceptions of party positions. However, the
campaign-basedpolicystatements.Forifcitizensdonot
coefficientestimatesreportedincolumn1ofTable2are
evenupdatetheirperceptionsofparties’Left-Rightposi-
smallandstatisticallyinsignificant,andthustheydonot
tionswhenparties’policystatementsshift,thenitseems
support these expectations. These estimates imply that,
unlikelythatcitizensadjusttheirpartisanloyaltiesortheir
ceterisparibus,aone-unitshiftintheCMPcodingsofa
own Left-Right positions in response to parties’ policy
party’sLeft-Rightpositionatthecurrentelection(com-
statements. However, this logic is not necessarily sound
paredtothepreviouselection)isassociatedwithaLeft-
because parties may move citizens indirectly via their
Rightshiftinvoters’perceptionsoftheparty’sLeft-Right
policystatements:specifically,partiesmaycommunicate
position of less than one-twentieth of one unit, where
their Left-Right shifts to organizations and political in-
boththeCMPcodingsandvoters’partyplacementsare
terest groups (i.e., trade unions, business organizations,
calibratedalongthesame1–10scale.And,thecoefficient
etc.), and these organizations may in turn provide cues
estimateonthelaggedvariable,[Partyj’sstatedLeft-Right
shift (t−1)], is also near zero and statistically insignifi- thatpromptrank-and-filevoterstomakepartisanadjust-
ments.Suppose,forinstance,thattradeunionmembers
cant.Furthermore,whenwereestimatetheparametersof
holdgenerallyleft-wingpolicyviews,andthatthesemem-
themodelwhileincludingcountry-specificinterceptsto
bersalsotakepoliticalcuesfromtheirunion’sleadership.
controlforunobserved,country-specificfactorsthatin-
Thentotheextentthatunionleadersprovidemoreenthu-
fluencevoters’perceptions,theparameterestimates(re-
siastic endorsements to socialist, labor, and communist
ported in column 2 of Table 2) support the same sub-
stantive conclusions.15 Thus, we find no evidence that partieswhentheseparties’policystatementsshiftfarther
totheleft,wemightexpectthatcue-takingunionmem-
voterssystematicallyupdatetheirperceptionsofparties’
bers will more reliably support these parties when the
Left-Rightpositionsinresponsetotheactualpolicystate-
partiesadvocatemoreleftistpolicies—eveniftheunion
mentsthatpartiespresentintheirmanifestos.Giventhat
members themselves do not directly perceive the shift
previousresearchdocumentsthatcitizensacquiredispro-
in the focal party’s policy statements. The possibility of
portionate amounts of political information during na-
thistypeofcue-basedpartisanswitchingissupportedby
tionalelectioncampaigns,andourinterviewsthatsuggest
extensiveresearchdocumentingtheinformationalshort-
thatpartiescampaignbasedontheirmanifestos,thesere-
cutsthatcitizensemploytomakepoliticaldecisions(e.g.,
sultscastdoubtonparties’abilitiestoshapetheirpolicy
Sniderman,Brody,andTetlock1991).
imagesviatheircampaign-basedpolicystatements.
15Weconductedadditionalrobustnesschecksusingalternativever-
SearchingforEvidenceofVoters’Partisan
sionsofthedependentandindependentvariablesinequation(1).
Inoneversion,wecalibratedshiftsinparties’statedLeft-Rightpo- AdjustmentswithCross-SectionalSurvey
sitions(ascodedbytheCMP)againsttheshiftsofotherparties Data
inthepartysystem,i.e.wecreatedavariablelabeled[Partyj’srel-
ativestatedLeft-Rightshift (t)]thatwasdefinedasthedifference
Inanalysesbasedonpaneldata,wemightparseoutthe
between the focal party j’s stated Left-Right shift at the current
electionandthemeanstatedLeft-Rightshiftofallotherpartiesin extent to which voters’ partisan adjustments to parties’
thesystematthecurrentelection.Ourlogicforthisspecification stated policy positions reflect partisan switching as op-
wasthatwhenvotersassesswhetherafocalpartyhasshiftedtothe
posedtopartypersuasion.However,becausethedatawe
rightorleft,theymaycalibratetheparty’sshiftagainsttheshifts
ofotherpartiesinthesystem.Wealsoestimatedamodelwhere analyze are time-series cross-sectional, we instead assay
thedependentvariable,theparty’sperceivedLeft-Rightshift,was thesimplertaskofestimatingtheextenttowhicheither
calibratedagainsttheperceivedshiftsofotherpartiesintheparty
of these processes is taking place, i.e., we will search for
system,i.e.,wecreatedavariablelabeled[Partyj’srelativeperceived
anyevidenceofvoters’partisanadjustmentstoEuropean
Left-Right shift (t)]. Our parameter estimates for these alterna-
tive specifications—which we report in supplementary materials parties’ stated Left-Right shifts. Our logic is simple: to
postedonourwebsite(http://privatewww.essex.ac.uk/∼ezrow)— the extent that parties’ campaign-based policy commu-
supportedsubstantiveconclusionsthatwereidenticaltothosere-
nicationsmotivatepartisanadjustmentsintheelectorate
ported above. We thank an anonymous reviewer for suggesting
theseanalyses. duetopartisanswitchingand/orpartypersuasion(and,
376
JAMESADAMS,LAWRENCEEZROW,ANDZEYNEPSOMER-TOPCU
regardless of whether these effects are direct or occur variablebecauseofthepossibilityoflaggedpartisanad-
indirectly via cues from political organizations/interest justmentstoparties’Left-Rightshifts.Wealsoincludea
groups), we expect that when a party shifts its policy laggedversionofthedependentvariableinordertocon-
statements, the mean Left-Right position of the party’s trolforpossibleautocorrelationissues.18Welabelthisthe
supporters will shift in the same direction. This test is partisanadjustmentspecification—CMPcodings:
verysimple,andasdiscussedbelowitisincompleteinthat
Partyjsupporters’Left-Rightshift(t)
even if we find evidence that party shifts correlate with
=(cid:2) +(cid:2) [Partyj’sstatedLeft-Rightshift(t)]
shiftsbytheirsupporters,wecannotnecessarilyconclude 0 1
thattheseassociationsreflectacausalprocessofpartisan +(cid:2) [Partyj’sstatedLeft-Rightshift(t −1)]
2
adjustment.16 However, if we find that the positions of +(cid:2) [Partyjsupporters’Left-Rightshift(t −1)]
3
parties’ supporters do not shift in the same direction as
(2)
theirpreferredparties’policyshifts,thenwehavestrong wherePartyjsupporters’Left-Rightshift (t)=thechange
reasons to doubt that partisan adjustments of any kind
inthemeanLeft-Rightpositionofpartyj’ssupportersat
havetakenplace.Onthisbasisweproceed.
the current election t compared to the mean supporter
positionatthepreviouselection(t−1),basedonNESre-
ModelSpecification
spondents’self-placementsandtheirpartysupport.The
variable[Partyjsupporters’Left-Rightshift (t−1)]issim-
Ourlongitudinalmeasureofvoters’Left-Rightpositions
ilarlydefined.
and their partisan loyalties is derived from the national
Column 1 in Table 3, which reports the parameter
election studies described above, from Germany, Great
estimatesforthepartisanadjustmentspecification—CMP
Britain, the Netherlands, Norway, and Sweden. In these
codings(equation2),doesnotsupportthehypothesisthat
surveys the respondents were asked to place themselves
votersengageinpartisanadjustmentsinresponsetoshifts
onthe1–10Left-Rightscales,andtheywerealsoaskedto
inparties’campaign-basedpolicystatements.19Theesti-
indicate which party (if any) they supported. Our mea-
mateonthe[Partyj’sstatedLeft-Rightshift (t)]variable
sure of the Left-Right position of each party’s partisan
impliesthat,ceterisparibus,aone-unitshiftintheCMP
constituencyinthefocalelectionisthemeanLeft-Right
codings of a party’s Left-Right position at the current
self-placement of all respondents in that election who
reportedsupportfortheparty.17 election(comparedtothepreviouselection)isassociated
withaLeft-Rightshiftinthemeanpositionoftheparty’s
Wespecifyamultivariateregressionmodeltosearch
supportersofjust0.07units,alongthesame1–10scale.
for evidence of voters’ partisan adjustments to parties’
Thus,weobservenoevidenceofstatisticallyorsubstan-
Left-Right shifts. As we are interested in how the mean
tivelysignificantcontemporaneouspartisanadjustments
positions of the parties’ supporters shift in response to
by voters in response to shifts in the Left-Right tone of
parties’ stated Left-Right shifts, we specify a model in
parties’campaign-basedpolicystatements.Withrespect
which the dependent variable is the change in the party
to lagged effects, the coefficient estimate +.140 on the
supporters’meanLeft-Rightpositioninthecurrentelec-
[Partyj’sstatedLeft-Rightshift (t −1)]variableindicates
tioncomparedtopartysupporters’positionsintheprevi-
thatwhenpartiesshiftedtheirstatedLeft-Rightpositions
ouselection,asmeasuredbyNESrespondents’Left-Right
byoneunitatthepreviousinterelectionperiod,then,ce-
self-placementsandreportedpartysupport.Welabelthis
terisparibus,themeanpositionsoftheirsupportersshift
variable[Partyjsupporters’Left-Rightshift (t)].Ourkey
inthesamedirectionbylessthanone-seventhofoneunit
independentvariablesareagainthe[Partyj’sstatedLeft-
atthecurrent interelectionperiod.However,thislagged
Rightshift (t)]variable,whichdenotestheshiftinparty
estimateisnotstatisticallysignificant.20
j’s stated Left-Right position between the current and
the previous election, based on the CMP codings, and
thelaggedversionofthisvariable.Weincludethislatter 18Wealsoestimatedtheparametersofthemodelwhileomittingthe
laggedindependentvariable,andtheseestimatessupportedsub-
stantiveconclusionsthatwereidenticaltothosewereportbelow.
16Asdiscussedbelow,ifweobservethatpartysupportersshiftinthe
samedirectionastheirpreferredpartyduringthesametimeperiod, 19Weestimatedtheparametersofthismodelusingrobuststandard
thismayreflectpartisanadjustments,butalternativelyitmayreflect errorsclusteredbyparty(Rogers1993).
partiesshiftingtheirpositionsinresponsetotheirsupporters.
20Weconductedadditionalrobustnesschecksusingalternativever-
17The text of the survey questions relating to party support dif- sionsofthedependentandindependentvariablesinequation(2).
feredacrosscountries,withsomequestionsinvitingrespondents Specifically,weincorporatedthe[Partyj’srelativestatedLeft-Right
toidentifytheparty(ifany)they“feelcloseto,”othersasking,“Of shift(t)]variableandthe[Partyj’srelativeperceivedLeft-Rightshift
whichpartyareyouanadherent?”andsoon.Informationabout (t)]variablesdescribedabove(seenote15)inordertoaddressthe
thesesurveyquestionsisavailableontheEuropeanvotertrend-file possibilitythatwhenvotersassesswhetherafocalpartyhasshifted
athttp://www.gesis.org/en/research/EUROLAB/evoter/. totherightorleft,theycalibratetheparty’sshiftagainsttheshiftsof
377
ISANYBODYLISTENING?
TABLE3 RelationshipbetweenShiftsinPartySupporters’Left-RightPositionsandParties’
Left-RightShifts
Partisan
Adjustments CMPCodings+
Specification— Voter Economic Eurobarometer
CMPCodings(1) Perceptions(2) Conditions1(3) data2(4)
Partyj’sstated .070(.119) .037(.063) .027(.066) .027(.037)
Left-Rightshift (t)
Partyj’sstated .140(.141) .067(.088) .040(.109) −.002(.028)
Left-Rightshift (t –1)
Partyj’sperceived .635∗∗(.075) .645∗∗(.062)
Left-Rightshift (t)
Partyj’sperceived .326∗∗(.066) .374∗∗(.084)
Left-Rightshift (t –1)
Inflationrate −.021(.013)
change(t)
Unemployment −.010(.012)
ratechange(t)
Partyjsupporters’shift (t –1) −.056(.137) −.277(.152) −.332(.173) −.381∗∗(.079)
Intercept −.057(.046) −.046(.030) −.045(.044) −.038(.024)
Numberofcases 97 97 97 296
AdjustedR2 .033 .471 .496 .121
∗∗p≤.01;∗p≤.05,two-tailedtests.
Notes:Standarderrorsareinparentheses.Theanalysesreportedincolumn1arebasedonthespecificationgivenbyequation(2)inthe
text,wherethedependentvariableisthechangeinthemeanLeft-Rightself-placementsoftheparty’ssupportersintheyearofthecurrent
electioncomparedtothepreviouselection,ascomputedfromNESdatafromGermany,GreatBritain,theNetherlands,Norway,and
Sweden(seetheappendixforthecompletelistofelectionsandpartiesincludedintheanalysis).Theindependentvariablesaredescribed
inthetext.
1Forthismodelweestimatedcountry-specificintercepts,whichwereasfollows(standarderrorsinparentheses):Germany−.011(.010);
Britain−.175(.123);Norway.034(.071);Sweden.001(.094).TheNetherlandsisthebaselinecountry.Wealsoestimatedcountry-specific
intercepts(availableuponrequest)forthemodelsreportedincolumns3–4.
2Theanalysesreportedinthiscolumnarebasedonthespecificationgivenbyequation(2)inthetextandarecomputedfromEurobarometer
datafrom12Europeancountries.
As discussed above, the absence of strongly positive organizationsandinterestgroupsthatmonitorthesepol-
statistical associations between shifts in the Left-Right icystatements.Finally,wenotethattotheextentthatour
toneofparties’policystatementsandthemeanLeft-Right analysesdonotcontrolforthereciprocalcausalprocess
positions of their supporters provides strong evidence whereby parties react to voters’ policy shifts, this factor
that voters do not react to shifts in parties’ campaign- shouldinflatethepositivestatisticalassociationsbetween
based policy statements, to any substantively significant ourdependentandindependentvariables.Thus,thefact
degree:fortotheextentthatvotersshifteithertheirLeft- thatwedonotobservepositiveandstatisticallysignificant
Right positions (the party persuasion process) or their associations between party supporters’ Left-Right shifts
partysupport(thepartisanswitchingprocess)inresponse andshiftsinparties’policystatementsisstriking.
to parties’ Left-Right shifts, this should create positive
statistical associations between Left-Right shifts by par-
ties and shifts by their partisan constituencies. Further- DoCitizensRespondtoParties’Policy
more,thesepositivestatisticalassociationsshouldobtain Images?
regardless of whether voters respond directly to parties’
Ourfindingthatcitizensdonotsystematicallyadjusttheir
policystatements,orindirectlyviacuesfromthepolitical
Left-Rightpositionsortheirpartisanshipinresponseto
otherpartiesinthesystem.Ourparameterestimatesforthesealter- parties’ policy statements raises the question: to what
nativespecifications—whichwereportinsupplementarymaterials policy-based attributes of parties, if any, do voters re-
postedonourwebsite(http://privatewww.essex.ac.uk/∼ezrow)—
spond? An obvious candidate is parties’ policy images,
supportedsubstantiveconclusionsthatwereidenticaltothosewe
reporthere. i.e., citizens’ perceptions of the parties’ positions. Given
378
JAMESADAMS,LAWRENCEEZROW,ANDZEYNEPSOMER-TOPCU
ourfindingthatparties’policyimagesdonottracktheir mateonthe[Partyj’sstatedLeft-Rightshift (t)]variable,
campaign-basedpolicystatements,itseemsplausiblethat andonthelaggedversionofthisvariable,arebothnear
citizensreacttotheseimagesinlieuofreactingtothepar- zero and are statistically insignificant, which again sup-
ties’actualpolicystatements.Toevaluatethishypothesis, portstheconclusionthatparties’campaign-basedpolicy
weestimatedtheparametersofamodelthatwasidentical statementsexertlittleeffectonvoters.
toequation(2)exceptthatweincludedtwoadditionalin- Next, we estimated the parameters of a model that
dependentvariablesthatweintroducedabove,[Partyj’s was identical to the CMP Codings + Voter Perceptions
perceivedLeft-Rightshift (t)]and[Partyj’sperceivedLeft- specification,exceptthatwecontrolledforadditionalfac-
Rightshift(t−1)],whichrepresentthecurrentandlagged torsthatplausiblyinfluencecitizens’Left-Rightshifts.To
shiftsintheparties’policyimages,i.e.,theshiftsinvot- control for national economic conditions (which may
ers’perceptionsoftheparties’positions.Byincorporating motivatecitizenstoshifttheirLeft-Rightpolicyorienta-
theseperceptualvariablesintoamodelthatalsoincludes tions independently of the parties’ policy positions), we
the parties’ shifts as coded from their actual campaign includedthevariables[inflationratechange(t)],defined
statements,wecanevaluatetheextenttowhichcitizens’ as the change in the inflation rate since the last elec-
partisan adjustments reflect their reactions to their per- tion,and[unemploymentratechange (t)],definedasthe
ceptions of the parties’ positions, as opposed to voters’ changeintheunemploymentratesincethelastelection.
reactionstoobjectivemeasuresbasedontheactualstate- And,tocontrolforunmeasured,country-specificfactors
mentsintheparties’electionmanifestos.21 thatinfluencecitizens’Left-Rightpositions,weincluded
The parameter estimates for this CMP Codings + country-specificintercepts.Theparameterestimatesfor
Voter Perceptions specification, which we report in col- thismodel,whichwereportincolumn3ofTable3,again
umn 2 of Table 3, strongly support the hypothesis that supportthehypothesisthatcitizensreacttotheirpercep-
citizens react to their perceptions of the parties’ Left- tionsoftheparties’Left-Rightshifts,butnottoshiftsin
Rightshifts,butnottoshiftsintheLeft-Righttoneofthe theLeft-Righttoneoftheparties’actualcampaignstate-
parties’actualcampaignstatements.Theseestimatesre- ments.22
flectstrongandpositiverelationshipsbetweenLeft-Right
shiftsinthemeanpositionsofpartysupportersandshifts
invoters’perceptionsoftheparties’left-Rightpositions: RobustnessChecksBasedonanAlternative
thecoefficientestimate+0.635onthe[Partyj’sperceived DataSet:TheEurobarometerSurveys
Left-Rightshift(t)]variableindicatesthat,ceterisparibus,
While we have so far uncovered no evidence that voters
aone-unitshiftinaparty’sperceivedpositiononthe1–10
adjusttheirpartisanshiportheirLeft-Rightpositionsin
Left-Rightscaleisassociatedwitha0.635-unitshiftinthe
response to parties’ campaign-based policy statements,
mean position of the party’s partisan constituency on
our analyses encompass only five European electorates,
the same 1–10 scale, during the same interelection pe-
riod.And,theestimate+0.326onthe[Partyj’sperceived namelyBritain,Germany,theNetherlands,Sweden,and
Left-Rightshift (t−1)]variableindicatesthataone-unit Norway (the five European countries for which NES
data are available over long time periods). In order to
shift in a party’s perceived Left-Right position is asso-
ciated with an approximate 0.3-unit shift in the mean
22We estimated the parameters of two additional specifications
position of the party’s partisan constituency during the
in an effort to control for statistical issues that might affect our
subsequent interelection period. Both of these estimates substantiveconclusions.First,whilewefindnoevidenceofagen-
arestatisticallysignificant(p<.01),andtheysupportthe eral relationship between parties’ stated policy shifts and shifts
in their supporters’ positions, it seems possible that there are
hypothesis that voters adjust their Left-Right positions
“threshold effects,” whereby voters react to party policy shifts
and/or their party support in response to their percep-
if these shifts are sufficiently large. Second, we estimated a sec-
tionsofparties’Left-Rightshifts—particularlythelagged ond alternative version of our CMP Codings + Voter Percep-
relationshipsweidentify,whichcannotbeascribedtothe tionsspecification,inwhichshiftsinparties’perceivedpositions
weredefinedasthedifferenceintheparty’spositionasperceived
reciprocal causal process whereby parties react to their
by all respondents except for the party’s supporters in the fo-
supporters’Left-Rightshifts.Finally,theparameteresti- cal election. We estimated the parameters of this Voter Projec-
tion specification in order to evaluate the hypothesis that the
positive relationship that we estimate between parties’ perceived
21Wenotethatourmodeldoesnotsufferfrommulticollinearity Left-Right shifts and shifts in their supporters’ Left-Right self-
problemsbecausethecorrelationbetweenourmeasuresofparties’ placements is due to party supporters projecting their own po-
shifts based upon their campaign statements versus voters’ per- sitionsontotheirpreferredparty.Ouranalysesforthesealternative
ceptionsofparties’shiftsisverylow,i.e.,thecorrelationbetween specifications—whichwereportinsupplementarymaterialsposted
theobservedvaluesofthe[Partyj’sperceivedLeft-Rightshift (t)] onourwebsite(http://privatewww.essex.ac.uk/∼ezrow)—support
variableandthe[Partyj’sstatedLeft-Rightshift(t)]variableisonly substantive conclusions that are identical to those we reported
.05. above.
379
ISANYBODYLISTENING?
evaluatewhetherourconclusionsgeneralizetootherEu- Insummary,wefindnoevidencethatcitizensreactto
ropeanelectorates,wereestimatedtheparametersofthe shiftsinEuropeanparties’campaign-basedpolicystate-
partisan adjustment specification (equation 2) using al- mentsbysignificantlyshiftingeithertheirownLeft-Right
ternativemeasuresofpartisanconstituencies’Left-Right positionsortheirpartisanloyalties.Thisconclusionholds
shifts that were derived from Eurobarometer surveys. regardlessofwhetherwemeasurepartysupporters’Left-
These surveys have been administered in the following Right shifts using national election study data from five
Western European democracies beginning in the early countries (Britain, Germany, the Netherlands, Sweden,
1970s:Britain,Italy,Denmark,France,Portugal,Greece, and Norway) or Eurobarometer data from 12 European
Spain, Luxembourg, the Netherlands, Belgium, Ireland, democracies. Simply put, we find no evidence that the
andGermany.23 Thesesurveyscontainthesameitemin mean Left-Right positions of party supporters move in
each year in each country, asking approximately 1,000 tandemwiththeLeft-Rightpolicystatementsthatparties
respondents per country to place themselves on a scale publishintheirelectionmanifestos.Bycontrast,wefind
runningfrom1(extremeleft)to10(extremeright),and a strong and positive relationship between citizens’ per-
alsotoindicatetheparty(ifany)thattheysupport.24We ceptions of parties’ Left-Right shifts, and the Left-Right
used these Eurobarometer items to construct the vari- shiftsoftheseparties’supporters.
ables [Party j supporters’ Left-Right shift (t)] and [Party
j supporters’ Left-Right shift (t−1)] that we used as our
dependentvariablesintheNES-basedanalysesdescribed
DiscussionandConclusion
above.(Wenotethatwecannotusethesedatatoconstruct
measuresofvoters’perceptionsofpartyshiftsbecausethe
Eurobarometersurveysdonotaskrespondentstolocate Ourfinding,thatvotersreactstronglytotheirperceptions
theparties’Left-Rightpositions.)Wethenreestimatedthe ofparties’Left-Rightshiftsbutnottoparties’actualshifts
parameters of equation (2) on all partisan constituency as coded by the Comparative Manifesto Project (CMP),
Left-Right shifts that are covered in the Eurobarometer raisestroublingquestionsaboutthenatureofmass-elite
surveys—296 party constituency shifts in 12 European policylinkages.InsituationswherepartiesshifttheLeft-
democracies,roughlytriplethenumberofpartyshiftsin Rightorientationsofthepolicystatementsthattheypub-
ourNES-basedanalyses.25 Column4inTable3presents lishintheirelectionmanifestos,wefindnoevidencethat
theparameterestimates,whichdisplaythesamepatterns voters respond by adjusting their own Left-Right posi-
astheestimatesbasedontheNESdata:thecoefficientes- tions,theirpartisanloyalties,oreventheirperceptionsof
timatesonthekeyindependentvariables,[Partyj’sstated theparties’Left-Rightpositions.Thesefindingsarestrik-
Left-Right shift (t)] and [Party j’s stated Left-Right shift ing given that previous research concludes that citizens
(t−1)], are almost exactly zero, so that there is no evi- displayheightenedlevelsofpoliticalattentionaroundthe
denceofstatisticallyorsubstantivelysignificantpartisan timeofelectioncampaigns,andthatoureliteinterviews
adjustmentsbyvotersinresponsetoshiftsinparties’pol- suggestthatpartiescampaignonthebasisoftheirmani-
icystatements.26 festos.
While our findings may appear surprising, we have
notedconsiderationsthatcastdoubtonparties’abilities
23WenotethatSpainandPortugalwerenotincludedintheEuro- toshapetheirpolicyimagesduringpoliticalcampaigns,
barometersurveysuntil1981and1986,respectively.
includingthefactthatmanycitizensareinattentivetopol-
24IneachEurobarometersurvey,respondentsareaskedtoindicate itics;thatrivalpartiesandmediaelitesmaychallengethe
their party support by answering the question: “If there were a
sincerityofthefocalparty’spolicypromises;thatcitizens
‘generalelection’tomorrow,whichpartywouldyousupport?”
tend to assimilate new political information into their
25Theseanalysescovered82partiescompetingin69elections.We
preexisting schemas; and that citizens must balance the
estimatedtheparametersofthismodelusingrobuststandarderrors
clusteredbyparty(Rogers1993),andaLagrangemultipliertestfails policystatementsthatpartiesissueduringelectioncam-
torejectthenullhypothesisofnoserialcorrelation. paignsagainstparties’statements(andbehavior)during
26Wereestimatedtheparametersofequation(2)onseveraldiffer- the precampaign period. It is this balance of conflicting
ent subsets of parties using the Eurobarometer data, in order to considerations that prompted us to undertake this re-
explorewhethervoters’partisanadjustmentstoparties’campaign-
search project. We have diligently searched for evidence
basedpolicystatementsaremediatedbythetypeofpartyorbythe
characteristicsofthepartysystem.Thesesupplementaryanalyses that citizens react to the policy statements that parties
included separate estimates on all governing parties, and on all
oppositionparties;onallleft-wingparties,andonallright-wing (2006,chapter5)identifyasrevolvingprimarilyaroundLeft-Right
parties;andontheBritish,Dutch,andItalianparties,i.e.,thethree economic issues. All of these analyses continued to support our
partysystemsintheEurobarometersurveysthatBenoitandLaver substantiveconclusions.