Download Download (pdf)

Survey
yes no Was this document useful for you?
   Thank you for your participation!

* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project

Document related concepts

Social perception wikipedia , lookup

Social dilemma wikipedia , lookup

Group dynamics wikipedia , lookup

Social tuning wikipedia , lookup

Social exclusion wikipedia , lookup

Transcript
CESifo Conference Centre, Munich
Workshop 2015
Workshop on
Public Opinion and UIFPolitical Economy
ofEducation
9 May
Trade͈offs Between Social Investment and
Passive Transfers in the New Welfare State:
New Political Coalitions in European Welfare
States?
Marius R. Busemeyer
Trade‐offsbetweensocialinvestmentandpassivetransfersinthenew
welfarestate:NewpoliticalcoalitionsinEuropeanwelfarestates?
Thisversion:March2015.
MariusR.Busemeyer
ProfessorofPoliticalScience
UniversityofKonstanz,FachD79
D‐78457Konstanz
Fon+497531882860,Fax+497531884103
Email:Marius.Busemeyer@uni‐konstanz.de
Web:http://www.polver.uni‐konstanz.de/busemeyer/prof‐dr‐busemeyer/
Abstract:
Inthewakeoftheglobaleconomicandfiscalcrisis,welfarestatesarenowenteringanew
phaseofausterity.Atthesametime,newsocialrisksrelatedtosingleparenthoodorcare
fortheelderlyleadtonewdemandsandexpectationsvis‐à‐visthewelfarestate.Thispaper
engagesinananalysisofhowthesecompetingdemandscomeintoconflictwitheachother
by using new and original survey data on individual‐level attitudes and preferences
towardssocialinvestmentsandpassivesocialtransfersineightEuropeancountries.Ifind
strongevidenceforcitizens’dislikeoftrade‐offs.Whenconfrontedwiththerealityoftrade‐
offs between different parts of the welfare state, citizens are less likely to support
additionalsocialinvestmentineducation,inparticularwhenthiswouldhavetobefinanced
bycutbacksinsocialtransferssuchaspensions.Ialsofindevidencefortheimplicationof
the “new politics” school that in the era of retrenchment, distributive conflicts within
existing welfare states exhibit a different political dynamic compared to the large‐scale
conflict about the scale and size of the welfare state. In particular, self‐interest more
narrowlydefinedasmembershipinparticularwelfarestateconstituencygroupsbecomesa
more important determinant of preferences when analyzing distributive conflicts within
existingwelfarestatesratherthantheconflictaboutredistributioningeneral.
1
Introduction
Backinthe1990s,welfarestatescholarsnotedtheendoftheeraofwelfarestateexpansion
andthedawnofaneraofpermanentausterity(Pierson2001;Stephensetal.1999).
AccordingtoPierson’s(2001)influential“newpolitics”schoolofthought,thisshiftin
circumstanceswasaccompaniedbyatransformationofthepoliticaldynamicofwelfare
statepolicy‐making.Respondingtoausteritypressures,policy‐makersareforcedtoengage
inretrenchment,tryingtominimizetheassociatedpoliticalcostsbyengaginginstrategies
ofblameavoidanceorobfuscation(Weaver1986).Inthewakeoftheglobaleconomicand
fiscalcrisis,itseemsthatwelfarestatesarenowenteringanewphaseofausterity(Pierson
2011),whichischaracterizedbyevenstrongerfiscalpressurestoconsolidatepublic
financeswithimplicationsforthesustainabilityofwelfarestates,inparticularinEurope.At
thesametime,newsocialrisksrelatedtosingleparenthoodorcarefortheelderlyleadto
newdemandsandexpectationsvis‐à‐visthewelfarestate(Bonoli2013;Hemerijck2013).
Thebulkofexistingscholarshipinthisfieldanalyzesthemacro‐levelofpolicy‐making,
implyingasimplisticunderstandingofvotersinthesensethattheywouldbestrongly
opposedtoanykindofmovementawayfromthestatusquo(Pierson1994,2001).Thus,
accordingtothe“newpolitics”argument,ifretrenchmenthappensatall,policy‐makers
essentiallyneedtohidethetruecostsofadjustmentinordertoavoidelectoralfallout.
Thispaperprovidesacontrastingperspectivetothisargumentbyexploringthedynamics
ofpolicytrade‐offsonthelevelofindividualattitudesandpreferences.Existingscholarship
onwelfarestateattitudesand,morerecently,onthe“feedbackeffects”ofpoliciesand
institutionsonmassattitudesandpatternsofpoliticalparticipation(cf.forrecent
overviewsoftheliterature:Béland2010;Campbell2012;Kumlin/Stadelmann‐Steffen
2014;Mettler/Soss2004;Svallfors2012)canonlyprovidelimitedinsightsinthisrespect,
becauseavailableinternationalsurveysonpublicopiniondonotmodelthesetrade‐offsina
satisfactorymanner.Inthepresentpaper,Icanrelyonoriginaldatacollectedinsurveyof
publicopinionineightWesternEuropeancountriesin2014.Amongotherthings,this
surveycontainsanexperimentalcomponent,whichallowstoanalyzehowpublicsupport
forwelfarestatespendingchangesoncedifferentkindsoftrade‐offssuchastaxincreases,
higherlevelsofpublicdebtandcutbacksinotherpartsofthewelfarestatearetakeninto
2
account.Ifocusonsupportforpublicspendingoneducation.Thisisbecauseprevious
researchhasshownthatincreasingeducationspendingishugelypopularamongvoters
(Ansell2010:Chapter4;Busemeyer2012),butonlywhenpolicyandfiscaltrade‐offsare
notfullyacknowledged.Thismakesitagoodtestcaseforstudyingtheimpactoftrade‐offs
onpatternsofpublicopinionandcoalitionformation.Furthermore,expandinginvestments
inhumancapitalformationisatthecoreoftherisingpolicyparadigmofthe“social
investmentstate”(Bonoli2013;Hemerijck2013;Moreletal.2011;
Vandenbroucke/Vleminckx2011),whichisgainingtractionamongpolicy‐makersatthe
nationalandtheEuropeanlevel.Hence,thepaperalsoprovidesinsightswithregardtothe
politicalviabilityofthisnewpolicyapproachasapotentialwayoutofthepolicydilemmas
ofcrisis‐riddenEurope.
Tobrieflyforeshadowthemainfindings,Ifindstrongevidenceforcitizens’dislikeoftrade‐
offs.Whenconfrontedwiththerealityoftrade‐offsbetweendifferentpartsofthewelfare
state,citizensarelesslikelytosupportadditionalsocialinvestmentineducation,in
particularwhenthesewouldhavetobefinancedbycutbacksinsocialtransferssuchas
pensions.Ialsofindevidencefortheimplicationofthe“newpolitics”schoolthatintheera
ofretrenchment,distributiveconflictswithinexistingwelfarestateshaveadifferent
politicaldynamiccomparedtothelarge‐scaledistributiveconflictaboutthescaleandsize
ofthewelfarestate,whichhadbeenpredominantintheeraofwelfarestateexpansion.In
particular,membershipinparticularwelfarestateconstituencygroupsisamoresignificant
determinantofindividualpreferenceswhenredistributiveconflictswithinthewelfarestate
arefullyacknowledged.Incontrast,consideringredistributivepreferencesmorebroadly
definedlarge‐scalecleavagestructuresindicatedbysocio‐economicpositionandpartisan
ideologyaremoreimportant.
Inthenextsection,Ibrieflyintroducetherelevantliteratureanddevelopatheoretical
argumentaswellasasetoftestablehypotheses.Subsequently,Iintroducethedatasetused
inthispaperandengageinananalysisoftwodifferentsetsofquestionscontainedinthe
survey,aimingatmeasuringtrade‐offsbetweensocialinvestmentandsocialtransfers.
3
Literaturereview
Welfarestatescholarshaveknownforalongtimethatreformingsocialpoliciescarrieshigh
politicalrisks,inparticularinaneraofretrenchmentandausterity,whenpolicy‐makersare
nolongerinapositiontohandoutgenerousbenefits,butinfactoftenhavetocutback
popularsocialpolicyprograms(Pierson1994,2001).Pierson’sworkwashighlyinfluential
inframingthedebateaboutthelinkbetweenwelfarestatereforms,organizedinterestsand
masspublics,buttoacertainextent,itimpliedaverysimplisticunderstandingofthelink
betweenpolicy‐makersandwelfarestateconstituencies.
Pierson’soriginalaccountfocusedmoreonorganizedinterestgroupsandwelfarestate
constituenciesmobilizingagainstcutbacksof“their”benefits,buthisargumenthasalso
beenappliedtothestudyofwelfarestateattitudes.Here,aninfluentialstrandinthe
literaturefindsthatsupportfordifferentsocialpolicieslargelydependsonself‐interest(e.g.
Busemeyeretal.2009;Hasenfeld/Rafferty1989;Iversen/Soskice2001;Kangas2003;
Lynch/Myrskylä2009;Rehm2009;Svallfors1997,2004),e.g.pensionersaremorelikelyto
supportgenerouspublicpensionsandtheunemployedwouldbeexpectedtodemandmore
spendingonunemploymentinsurance.Furthermore,agrowingnumberofscholars–
drawinginspirationfromahugeliteratureonpolicyresponsivenessandfeedback(Béland
2010;Campbell2012;Eriksonetal.2002;Kumlin/Stadelmann‐Steffen2014;Mettler/Soss
2004;Wlezien1995;Soroka/Wlezien2010;Stimsonetal.1995)–positthatthereisa
directlinkbetweenpublicopiniononthewelfarestateandpolicyoutput(Brooks/Manza
2006,2007;Rehm2012):Popularattitudesaretranslatedintoactualpolicyoutput,since
policy‐makerscareforpublicopinioneveninbetweenelectionsbecauseoffearofelectoral
punishmentinthefuture(Stimsonetal.1995).Becausewelfarestateretrenchmentis
thereforegenerallyunpopular(Boerietal.2001),thegeneralexpectationinthisliterature
isthatwelfarestates“persist”(Brooks/Manza2006),i.e.large‐scalecutbacksareunlikely
becausetheyarepoliticallyrisky.Reformsmayonlybepossiblewhenpolicy‐makers
engagein“blame‐avoidance”andobfuscationstrategies(Pierson1994;Weaver1986)and
essentiallyhidetheirtrueintensionsandactionsfromvoters.
4
Takentogether,theworkbyPierson,BrooksandManzaimplyarathersimplistic
understandingofthelinkbetweenwelfarestatepoliciesandpublicopinion.Once
established,welfarestatepoliciesandinstitutionscreatestrong“positivefeedbackeffects”
(Pierson1993,2000)inthesensethatpopularattitudesandpreferenceswillincreasingly
reflecttheprevailingstatusquoasactorsadapttheirpreferencesandstrategiessothat
particularinstitutionaldevelopmentpathsget“lockedin”.Furthermore,asAndrea
Campbell(2002)andSuzanneMettler(2005)haveshownintheirwork,welfarestate
policiesdonotonlyinfluencethecontentofpolicypreferences,butalsosupportthe
creationofpoliticalcapacities,enablingwelfarestatebeneficiariestomoreeffectively
mobilizeindefenseoftheircherishedprograms.Forsure,therearepersistentcross‐
nationaldifferenceswithregardtowelfarestateinstitutionsaswellaspreferences,andthe
Pierson‐Brooks‐Manzaframeworkishelpfulinexplainingthem.However,scholarshave
notedparticularanomaliesinrecentyears,whicharehardtoexplainwithinthis
framework.
Mostimportantly,welfarestatechangeismorecommonthanwouldbeexpected.Directly
contestingPierson’sfindings,scholarshavenotedasignificantdegreeofretrenchmentand
numerouscutbacksinwelfarestategenerosityinOECDcountries(Allan/Scruggs2004;
Korpi/Palme2003).However,itisnotthecasethatretrenchmentisthe“onlygamein
town”(vanKersbergenetal.2014).Infact,somewelfarestateshavesignificantlyexpanded
theircommitmentto“socialinvestment”policiessuchasearlychildhoodeducation,active
labormarketpoliciesandcarefortheelderly(Bonoli2013;Hemerijck2013;Moreletal.
2011;Vandenbroucke/Vleminckx2011),oftenatthesametimeastheywerecuttingback
themoretraditionaltransfer‐intensivepartsofthewelfarestate(seeFleckensteinetal.
2011forthecasesofBritainandGermany).Someclaimthatthisshiftsignalsaqualitative
“recalibration”ofwelfarestatescateringtonewsocialrisks(Hemerijck2013:104),butof
course,itmightbethecasethattheexpansionofthesocialinvestmentcomponentsof
welfarestatesamountsmerelyto“sugarcoating”lesspopularretrenchmentreforms,since
theformerarerelativelycheapcomparedtotraditionalsocialinsurancepolicies(Bonoli
2013;Häusermann2010).
5
Theoryandhypotheses
Thesebriefexamplesshowthattheprocessofwelfarestatechangeismorecomplexthan
oftenassumed.Welfarestatesdonotjust“persist”,insteadthereisretrenchmentandthere
issystemicchange,oftenhappeningatthesametime.Onthebackgroundofexisting
literatureonwelfarestatereformsandgiventhelimitationsofexistinginternational
surveysofpublicopinionwithregardtothemodelingoftrade‐offs,however,wedonot
knowwhetherthesechangesareresponsestogenuinechangesinpopulardemandor
whetherthereisadisconnectbetweenpeople’sattitudesandpolicychangesinthesense
thatlatterresultfromincreasinglybindingconstraintsonthepartofpolicy‐makers,not
citizens’demands.Existingresearchhasnotbeenabletoanswerthisquestionina
satisfactorymanner,becauseavailablesurveysofpublicopinioncontainfewquestionson
socialinvestmentpoliciesand,moreimportantly,theydonottakeintoaccountimportant
constraintsandtrade‐offsbetweendifferentpartsofthewelfarestate.
Afirstimportantquestionisthereforehowpublicsupportforexistingornewsocial
policieswouldchangeoncebudgetarytrade‐offsarefullyacknowledged.Ontheonehand,
itcouldbethecasethattheacknowledgementofbudgetarytrade‐offshasnoeffect
whatsoeveronpublicsupport,simplybecausecitizensdonotcareaboutbudgetary
limitationswhenvoicingdemandsvis‐à‐vispolicy‐makers.Thisisactuallytheimplication
ofPierson’s(2001)“newpolitics”argument,whenappliedinfullyconsistentmanner.
Votersareonly(orprimarily)interestedinmaintaining“their”welfarestateprograms,
hencetheirsupportshouldnotwaveroncetrade‐offsaretakenintoaccount.Ontheother
hand,however,citizenscouldactuallymediatetheirdemandswhentheybecomeawareof
trade‐offs.Whenremindedthatadditionalspendingwouldhavetobefinancedwithhigher
taxes,higherlevelsofpublicdebtorcutbacksinotherpartsofthebudget,citizenscould
becomelesssupportiveofthesespendingincreases.Thisisbecauseinfacttheyareableto
prioritizebetweendifferentpolicyfieldsratherthansimplydemandingmorespendingat
anycost.
Furthermore,if“positivefeedbackeffects”indeedoccur,citizensshouldbemostreluctant
tofinanceadditionalspendingwithcutbacksinotherpartsofthewelfarestate,because
6
existingsocialpoliciescreatestrongsupportamongtheirbeneficiariesandbeyond.If
forcedtochoose,citizensmightbemoreopentofinanceadditionalspendingwithhigher
taxesorpublicdebt,incasepublicdemandforthisspendingissufficientlyhigh.Thecaseof
educationspendingstudiedinthispaperisagoodtestcaseinordertoteaseoutthese
differences,becausepreviousresearchhasshownthatlargemajoritiessupportmorepublic
spendingoneducationinprinciple(Ansell2010:Chapter4;Busemeyer2012).Therefore,
thereissufficientleewayforreductionsinpublicsupportoncedifferentkindsof
constraintsarementionedintheframingofthesurveyquestion.
Therefore,thefirsthypothesistobetestedis:
Hypothesis1:Ingeneral,individual‐levelsupportforincreasing(education)spendingdrops
oncetrade‐offsbetweenspendingandtaxationaswellastrade‐offsacrossdifferentpartsof
thewelfarestatearetakenintoaccount.Iexpectthedropinsupporttobestrongerinthe
lattercasesincetheserefertocutbacksinhighlyvisiblesocialpolicies.
ThesecondhypothesisIwanttotestinthispaperismorecomplexandrelatesto
differencesinthedeterminantsofspendingsupport,dependingonwhichkindsof
constraintsareenforced.Theultimategoalhereistofindoutwhetherthesocio‐political
coalitionsthatsupportspendingincreaseschangeoncetrade‐offsareacknowledged.A
plausibleexpectationisthatgeneralconflictsaboutthesizeand/orgenerosityofthe
welfarestatearedeterminedbyotherfactorsthanconflictsaboutthedistributionof
resourceswithinexistingwelfarestates.
Thebulkofthepertinentliteratureonwelfarestateattitudesandredistributive
preferenceshasbeenconcernedwithstudyingthefirstaspect,i.e.thedeterminantsand
variationofindividualpreferencesalongthedimensionof“moreorless”redistribution
and/orsocialspending(e.g.Alesina/Angeletos2005;Corneo/Grüner2002;Cusacketal.
2006;Fong2001;Kenworthy/McCall2008;Rehm2009).1Theliteraturehasidentifieda
1Ofcourse,somestudiesalsoexplorethevariationofattitudesacrosspolicyfields(e.g.
Busemeyeretal.2009),buteveninthatcase,thegoalislesstoidentifytrade‐offsbetween
7
numberof“usualsuspects”thathavebeenfoundtoaccountforvariationinsupportfor
redistribution,suchasincome,educationalbackground,age,genderorindividualpartisan
ideology.Morerecently,scholarshaveaddedfactorssuchasskillspecificity
(Iversen/Soskice2001),occupationalunemploymentrisk(Rehm2009),the“offshorability”
ofone’sjob(Walter2010),orperceptionsoffairnessanddeservingness(Alesina/Angeletos
2005;VanOorschot2006)–tonameafewexamples.
ThecoreargumentIwanttodevelopinthefollowingistosaythatconflictsaboutthe
overallsizeandgenerosityofthewelfarestateplayoutonadifferentpoliticaldimension
thanconflictsaboutthe(re‐)distributionofresourcesacrossdifferentpartsofthewelfare
state.Whereasconflictsabouttheoverallsizeofthewelfarestateandthegenerallevelof
redistributioninagivensocietyarerelatedtomacro‐levelcleavagesdefinedbyincome,
age,class,ideologyorsimilarfactors,materialself‐interestinthesenseofbelongingtoa
particularwelfarestateconstituencygroupmattersmorewhenitcomestodistributive
conflictswithinwelfarestates.Thus,givenacertaindegreeofwelfarestategenerosity,trade‐
offswithinthewelfarestatereflecttheclashofinterestsofdifferentconstituencies.
Inthepresentpaper,Ifocusondistributiveconflictsbetweenthebeneficiariesof“old”and
“new”socialpolicies(Vandenbroucke/Vleminckx2011),i.e.socialtransferprogramssuch
aspensionsontheonehandandnewkindsofsocialinvestmentpoliciessuchaseducation
andfamilypoliciesontheother.Ingeneral,thetrade‐offbetweensocialinvestmentsand
passivesocialtransfersisoftendepictedasonebetweenpoliciesthatcreatediffusebenefits
atsomepointinthefuturevs.socialpoliciesthathaveconcretebenefitsinthepresent(cf.
e.g.Streeck/Mertens2011).Forexample,today’sinvestmentsineducationandearly
childcarewillpayoffwhenyounggenerationswhoenjoythesebenefitstodayenterthe
labormarketatsomedistantpointinthefuture.Incontrast,pensionandunemployment
schemesentailimmediatebenefitsforthosewhoreceivethem.Cuttingbacksocialtransfers
–sotheargumentmightgo–isthereforedifficult,becausethediffusebenefitsofsocial
investmentsinthefutureareusuallynotsufficientlystrongtooutweighthematerialist
policies,buttoexplorehowthesupportforredistributionvariesacrossandwithinpolicy
fields.Inotherwords,theunderlyingdimensionremainsthesame.
8
interestsofpensionersandtheunemployed.Theelectoralconstituenciesandorganized
interestsassociatedwithsocialinvestmentarealsobelievedtobepoliticallylessinfluential
comparedtothedefendersofthetraditionalwelfarestate.Itisevenmoredifficultintimes
ofausterity,whenbudgetpressuresnecessitatereductionsinspending.Discretionarytypes
ofspendingsuchaspublicandsocialinvestmentsareeasiertocutbackthanentitlement‐
basedspending(Breunig/Busemeyer2012;Streeck/Mertens2011).
Thisperspective,however,neglectsthefactthatsocialinvestmentsoftenproduce
immediatebenefitsaswell–andoftenbeyondthegroupofthosewhoareimmediately
affected.Thecaseofearlychildhoodeducationisonesimpleexampleofthisprocess:The
expansionofchildcareprovisiondoesnotonlybenefitthechildrenthemselves,butitalso
increasestheopportunitiesforyoungparentstocombinefamilylifewithemployment.The
welfarestateconstituencygroupthatsupportstheexpansionofchildcareopportunities–
(young)parents–mighteasilybeaslargeassomedefendersoftraditionalsocialtransfers
(e.g.theunemployedorthedisabled).Therefore,thetrade‐offcouldbelessaboutfuture‐
orientedinvestmentsvs.consumption‐orientedsocialtransfers,butsimplyreflectthe
emergenceofnewredistributiveconflictsbetweenoldandnewwelfarestate
constituencies.
Thecentralhypothesisisthatincontrasttothegeneralpoliticalconflictabout
redistributionandthesizeofthewelfarestate,materialself‐interestmorenarrowlydefined
willplayalargerroleinredistributiveconflictswithinthewelfarestate.Thereasonisthat
whenitcomesto(re‐)distributionwithinthewelfarestate,servicesandtransfersarevery
visibleandconcreteforthosewhoimmediatelybenefitfromthem,whereastheissueof
redistributionandwelfarestategenerosityislocatedonamoreabstractlevelsothat
generalvalueandideologicalorientationsaswellastheindividual’sgeneralsocio‐
economicpositionwithregardtoincome,educationandagemattermore.
Hence,thesecondhypothesistobetestedis:
9
Hypothesis2:Conflictsabouttheoverallsize/generosityofthewelfarestateplayoutona
differentpoliticaldimensionthanconflictsaboutthe(re‐)distributionofresourcesacross
socialpolicysub‐fields.Intheformercase,ideology,normativeorientationsandthe
individuals’socio‐economicpositionintheclasshierarchyareexpectedtomattermore,
whereasinthelattercase,Iexpectmaterialself‐interestmorenarrowlydefinedtohavea
strongerimpact.
Operationalizingtrade‐offsattheindividuallevel:Introducinganewdataset
Methodologicallyspeaking,themeasurementoftrade‐offsinindividualpreferencesfor
welfarestatepoliciesischallenging.ExistingsurveyssuchastheEuropeanSocialSurvey
(ESS)ortheInternationalSocialSurveyProgramme(ISSP)oftencontainquestionsabout
individualpreferencesforpublicspendingorgovernmentinvolvementmoregenerally
defined.Unfortunately,thesequestionsusuallyneithertakeintoaccountthefactthat
individualsupportforpublicspendingshouldmatchindividualwillingnesstopaytaxes2
nordotheyexplicitlytakeintoaccounttrade‐offsacrosspolicyfields.Theseissuesare
crucial,however,inordertoproperlymeasuretherealindividual‐levelsupportforsocial
spendingacrossarangeofpolicies,sinceneglectingthesetrade‐offswillprobablyresultin
onoverestimationofsupportforspending.Thisisbecauserespondentsinpublicopinion
surveyshavebeenfoundtosupportbothlowertaxesaswellasagenerouswelfarestate
withnoconstraintsortrade‐offsenforced(Citrin1979;forarecentupdate:Giger/Nelson
2013).Also,ifnotforcedtochoose,theyarelikelytosupporthigherspendinginallpolicy
fields(seeGoerres/Prinzen2012foramoregeneraldiscussionoftheproblemsof
measurementinwelfarestatepreferences).
Forthepurposeofthispaper,Icanrelyonanoriginaldatasetthatmeasurestrade‐offsin
preferencesattheindividuallevelinanovelway,takingintoaccounttrade‐offsbetween
supportforspendingandfinancingontheonehandaswellastrade‐offsacrosspolicyfields
2TheISSPRoleofGovernmentmoduleisoneofthesurveysmostoftenusedtomeasure
individualspendingpreferences,anditpaysabitmoreattentiontothisissuethanother
surveys.Thewordingofquestionsaboutpublicspendingcontainsa“warning”to
respondentsthatlargespendingincreasesmightrequireataxincreasetopayforthis.This
is,however,aratherweakformofmodelingtrade‐offs.
10
ontheother.Thisoriginalsurveydatawascollectedintheproject“InvestinginEducation
inEurope”(INVEDUC).3Inthisproject,weconductedasurveyofpublicopinionon
educationandthewelfarestateineightEuropeancountries.Thecountriesareselectedas
representativesofdifferentwelfarestateregimesinNorthernandWesternEurope:
GermanyandFrancefortheconservativewelfarestateregime,theUKandIrelandforthe
liberalregime,DenmarkandSwedenforthesocialdemocraticvarietyaswellasItalyand
SpainasSouthernEuropeanwelfarestates.Thequestionnairewasdesignedbytheproject
teamattheUniversityofKonstanzincollaborationwithexpertsfromTNSInfratest
Sozialforschung–thesurveycompanyinchargeoftheactualfieldworkofthesurvey.The
originalquestionnairewaswritteninEnglishandthentranslatedintothedifferent
languagesinafour‐stepprocess,whichconsistedofatranslatabilitycheck,theactual
translation,alinguisticqualitycontrolandafinalqualitycontrolbytheprojectteam(for
detailsseeGensickeetal.2014).
Thetargetpopulationofthesurveyisthepopulationinagivencountryaged18andabove.
Thesurveyusedcomputer‐assistedtelephoneinterviewing(CATI).Randomdigitdialing
(RDD)wasusedinallcountriesexceptSweden(whereacomprehensiveaddressregister
coveringmorethan90percentofthepopulationexists)inordertodealwiththeproblem
thatasizableandincreasingshareofthepopulationdoesnothavealandlineanymore,but
usesmobilephonesonly.Thedatasetcontainstwosetsofweightsthatareappliedinthe
descriptivestatisticsbelow:first,adesignweightappliedinthecaseoflandlinesonlyto
takeintoaccountdifferentselectionprobabilitiesofindividualsdependingonthenumber
oflandlinesperhouseholdandthenumberofpotentialintervieweesinagivenhousehold;
second,aselectivityweightthatcorrectsfordifferencesbetweenthesampleandthetarget
populationbyreferringtowell‐knownstratificationcharacteristicsofthelatter(using
stratificationvariablessuchasage,gender,educationlevels,occupationalstatus,regions
andemploymentstatus).Inthemultivariateregressionanalyses,Irefrainfromapplying
weightsbecausetheinclusionofindependentvariablescontrolsfordifferencesinthe
3Thisprojectisfinancedwitha“StartingGrant”fromtheEuropeanResearchCouncil,Grant
No.311769.
11
dependentvariablerelatedtothesecharacteristicsandavoidsrelyingonthe“blackbox”of
weighting.
ApretestofthequestionnairewasconductedinalleightcountriesinFebruary/March
2014.Theactualfieldworkforthesurveytookplacebetweenmid‐AprilandendofMay
2014.Theaveragelengthofaninterviewwas25minutes.Thenetnumberofinterviews
differedslightlybetweencountriesinordertotakeintoaccountdifferencesinthesizeof
thetargetpopulationinthatparticularcountry.TableA1intheappendixdisplaysthe
numberofinterviewspercountryaswelltheresponserate.Theaverageresponseratewas
roughly30percentwithahighof36percentinDenmarkandalowof20percentinIreland.
Interviewersweretrainedextensivelybeforethepretestaswellasthemainphaseofthe
fieldwork.MembersoftheprojectteamatKonstanzwereinvolvedinthebriefingof
interviewersinordertomakesurethatinterviewersalsounderstoodthecoregoalsofthe
project.Inordertoreducethedistancebetweentheinterviewerandtherespondent,the
fieldworkintheindividualcountrieswasconductedbynationalfieldofficesofTNS
InfratestSozialforschung,butcentrallycoordinatedandmonitored(Gensickeetal.2014:
13‐14).Theinterviewerswerenativespeakers.
Thedeterminantsoftrade‐offsinthewelfarestate(I)
Inthefollowing,Iemploytwodifferentsetsofquestionsthatwereincludedinthesurveyto
measuretrade‐offsinwelfarestateattitudes.Thefirstisasetofquestionsthattriedto
measuretheindividuals’willingnesstoinvestinapopularpolicysuchaseducationby
manipulatingthealternativeoptionsthatindividualscouldchoosefrom.Tothatextent,the
totalsamplewassplitintofourgroups(one“control”groupandthreedifferenttreatment
frames),4andassignmenttooneofthesegroupswasrandom,ofcourse.Theexactwording
ofthefourquestionswere:
Split1(“Control”):“Thegovernmentshouldincreasespendingoneducation.”
4Ofcourse,thesearenottreatmentsinthestrictexperimentalsense,becausethe
“treatment”ismeasuredtogetherwiththedependentvariable(supportforeducation
spending).
12
Split2(“Treatment1”):“Thegovernmentshouldincreasespendingoneducation,evenif
thatimplieshighertaxes.”
Split3(“Treatment2”):“Thegovernmentshouldincreasespendingoneducation,evenif
thatimpliescuttingbackspendinginotherareassuchaspensions.”
Split4(“Treatment3”):“Thegovernmentshouldincreasespendingoneducation,evenif
thatimpliesahigherpublicdebt.”
Followingthesestatements,respondentswereaskedwhethertheywouldstronglyagree,
agree,neitheragreenordisagree,disagreeorstronglydisagreewiththestatement.Table
A2intheappendixdisplaysmeansandstandarderrorsforthemostimportantvariablesin
theindividualgroupsaswellastheoverallnumberofvalidobservations,showingthatthe
randomassignmentofindividualstogroupsdidworkforthemostpart.TableA3contains
informationaboutthenumberofobservationsinthedifferentgroupsacrosscountries.
Figure1isasimplegraphicalrepresentationofthemeans(and95%confidenceintervals)
ofthecontrolgroupandthethreedifferenttreatmentswithouttakingintoaccount
differencesbetweencountries.Table1containstheresultsofaseriesofunpairedt‐testsof
differencesinmeansbetweenthecontrolandtreatmentgroupsinordertoconfirmthat
differencesbetweengroupsareallstatisticallysignificant.Responsesarecapturedona5‐
pointLikertscalefrom1(stronglydisagree)to5(stronglyagree).
Thefigureandthet‐testsclearlyshowthatsupportformoreinvestmentsineducationis
conditionalonthemenuofalternativechoicesthatrespondentsareconfrontedwith,which
isclearconfirmationofmyfirsthypothesis.Notsurprisingly,supportformoreeducation
spendingbythegovernmentishighestinthegroupwithoutanyconstraintsorqualifiers
added.Inthisgroup,meansupportformoreeducationspendingisat3.77pointsonthe5‐
pointscalementionedabove.Thisisequivalenttoashareof73.36percentofrespondents
expressingsupportformoreorevenmuchmorepublicspendingoneducation.Support
dropssignificantlyto3.13,whenrespondentsareremindedthatmoreeducationspending
13
mightimplyhighertaxes(equivalenttoashareof53.84percentsupportingmore(ormuch
more)spending).Whenrespondentsareconfrontedwiththefactthathighereducation
spendingwouldnotbefinancedwithhighertaxes,butwithhigherlevelsofpublicdebt,
supportdropsfurtherto2.95(ashareof42.10percent).Thisisacriticalthreshold,because
itrepresentsthewatershedbetweenindividualslargelysupportingincreasingeducation
spendingvs.themopposingspendinghikes.Mosttellingly,however,supportformore
educationspendingdropsevenfurtherto2.64whenrespondentsareconfrontedwiththe
possibilitythatincreasingeducationspendingcouldgoalongwithcutbacksinotherparts
ofwelfarestate,inparticularpensions.Thehugemajoritysupportingeducationspendingin
thefirstgroupisnowreducedarelativeminorityof27.19percentsupportingmoreor
muchmorespendingoneducation.
Thefactthatrespondentsbecomelesslikelytosupportspendingincreaseswhentheyare
remindedthattheyhavetopayforitintheformofhighertaxesorhigherlevelsofpublic
debtisnotsurprising.Whatissurprisingisthatreducingspendinginotherpartsofthe
welfarestate,inparticularpensions,isevenmoreunpopularthanrisingtaxesorincreasing
publicdebt.Themagnitudeofdifferencesinmeansbetweengroupsisstrikingasisthefact
thatallthesedifferencesarehighlystatisticallysignificant.Alsosomewhatunexpectedly,
respondentsdonotpreferincreasingpublicdebttorisingtaxes.Ifindividualswouldonly
caretomaximizeshort‐termbenefitsandminimizeshort‐termcosts,wewouldexpectthem
tobemoreopposedtotaxhikes,becausethesecreateimmediateshort‐termcosts.
However,individualsseemtobeawareofthepotentialnegativesideeffectsofhighlevelsof
publicindebtednessevenwhendebtisusedtofinancepublicinvestmentssuchas
education.Onecaveatthatneedstobeaddedhere,however,isthatindividualsinEuropean
countriescouldbemoresensitivetohighlevelsofpublicdebtbecauseoftheongoing
economicandfiscalcrisisinmanyEuropeancountries.
14
Figure1:Trade‐offsinthewelfarestate:meansandCIsofcontrolandtreatmentgroups.
15
Table1:Resultsofunpairedt‐testsofdifferencesinmeansbetweencontrolandtreatment
groups.
t‐Test
Difference
t‐Value(H0=no
p‐Value
difference)
Cvs.T1
0.5526487
17.2721
0.0000
Cvs.T2
1.184971
37.5701
0.0000
Cvs.T3
0.8132021
25.6048
0.0000
T1vs.T2
0.6323221
18.4245
0.0000
T2vs.T3
‐0.3717687
‐10.8912
0.0000
T1vs.T3
0.2605534
7.5556
0.0000
InthefollowingmultivariateregressionanalysesIwanttodeterminewhethersupportfor
educationspendingisinfluencedbydifferentfactors,dependingonwhichkindoftrade‐off
questionisasked.Thisexerciseaddressesthesecondhypothesisstatedabove,namelythat
materialself‐interestmorenarrowlydefinedwillbecomemoreimportantrelativeto
broadersocietalcleavagesoncetrade‐offsarefullyacknowledged.
Buildingontheinsightsofthelargeliteratureonwelfarestateattitudescitedabove,the
modelcontainsvariablesrelatedtoself‐interestaswellasideologicalandnormative
orientations.Thelattermayatleastbepartlyendogenous,i.e.individuals’socialpolicy
preferenceswillalsocontributetoshapingtheirpartisanworldviewsandnormative
dispositions.However,theproblemofendogeneitycouldbelesssevereifoneassumesthat
ideologyandnormativeorientationsarefactorsthatarepartofdeepbeliefsystemsand
thusnotpronetoshort‐termchangestothesameextentassocialpolicypreferencesmight
be.
Morespecifically,thefollowingindependentvariablesareincludedinthemodels:
Educationalbackgroundmeasurestheindividual’seducationalexperiencesonafive‐point
scalefrombasiceducationtouniversity‐leveleducation.Previousresearchhasshownthat
highlyeducatedindividualsarealsomorelikelytosupportthefurtherexpansionof
16
educationalopportunities,eventhoughthismightseemcounterintuitivefromanarrow
rational‐choiceperspective(Busemeyer2012:230).Individual(net)incomeisgivenin
purchasing‐poweradjustedUS‐Dollars,re‐categorizedintofiveincomequintiles.In
contrasttoeducation,theassociationbetweenincomeandsupportforeducationspending
ismoreambivalent,sinceeducationalinvestmentsarelessredistributivethanothersocial
policies(Ansell2008,2010).Hence,Idonotexpectastrongassociationinthiscase.
Expectationsrelatedtodemographicvariablesaremorestraightforward:Retiredpersons
shouldbelesssupportiveofmoreeducationspending,whereasrespondentsstillin
education,womenandindividualswithsmallkidsathomeshouldexhibithigherlevelsof
support.
Inadditiontovariablesrelatedtodemographicsandthesocio‐economicposition,Iinclude
variablescapturingideologicalorientations.Left‐rightorientationcapturesindividuals’
self‐placementona10‐point‐scalefrom1(left)to10(right).Finally,generalpreferences
forincreasingthesizeofstatecouldbetakenasanindicatorofaprevalenceofanormative
orientationontheroleofthestate.Thisvariablemeasuresindividualresponsestothis
question:“Ashareofthenationalgovernment’sannualbudgetisspentonsocialbenefits
andsocialservices.Pleasetellme,accordingtoyou:Shouldthegovernmentspendmuch
more,more,thesame,spendlessormuchlessonsocialbenefitsandsocialservices?”Again,
responsesarerecordedona5‐pointLikertscalewithhighervaluesindicatingstronger
supportforstateinvolvement.
Withregardtomethods,Iapplyverysimpletechniques.Firstofall,Idichotomizethe
dependentvariablesinordertofacilitateinterpretationofthecoefficientestimates.Inall
fourcases,supportformoreormuchmorespendingiscodedas“1”,whereassupportfor
thesameorlessspendingiscodedas“0”.Second,thestructureofthedatasetishierarchical
(individualsnestedwithincountries).However,thenumberoflevel‐2units(countries)is
notsufficientinordertobeabletorunfully‐fledgedmulti‐levelmodelsortomodelthe
effectofmacro‐levelcharacteristicsonmicro‐levelvariablesdirectly.Instead,Itreatthe
multi‐levelstructureofthedatasetsimplyasafactorthatneedstobetakenintoaccountin
theselectionofthepropermodelspecification.Ithereforeemployasimplelogitmodelwith
17
countrydummiesandstandarderrorsclusteredbycountry(seetable2).Tomakethe
findingsmoreaccessible,averagemarginaleffectsareplottedinfigure2.
Takentogether,thepatternthatemergesfromtheregressionanalysesisthefollowing:
Firstofall,thereisadistinctionbetweenquestionsabouttheoveralllevelofpublic
spendingoneducationanditsfinancingviahighertaxesorpublicdebtontheonehandand
thequestionofredistributingresourcesfromotherpartsofthewelfarestate(inparticular
pensions)toeducationontheother.Inthefirstsetofquestions,educationalbackground,
left‐rightorientationandgeneralspendingpreferencesarestrongdeterminantsofsupport
foreducationspending:Highlyeducated,left‐leaningandstate‐lovingindividualsaremore
likelytosupporteducationspending,whereasindividualswithlittleeducation,arightist
ideologicalorientationandagreaterskepticismwithregardtotheroleofthestatetendto
bemoreopposed.Asexpectedandinlinewithpreviousresearch,individualincomedoes
notmattermuch,nordoesbeingfemaleorretired.
Inthequestionaboutredistributingresourceswithinthewelfarestate,however,these
variableslooseanyexplanatorypower.Instead,variablesrelatedtoself‐interestmore
narrowlydefinedturnouttobestatisticallysignificantnow:Theretiredare–forobvious
reasons–stronglyopposedtotakingawayresourcesfrompensionstoincreasespending
oneducation.Usingmodel3oftable2,thepredictedchangeintheprobabilityforaretired
persontosupporteducationspendingcomparedtoanon‐retiredpersonisareductionof
about6percentagepoints.5Viceversa,individualswithsmallkidsathomearemorelikely
tosupporttakingawayresourcesfrompensionerstofundeducationalinvestments.
Simulatingasimilarchangeasaboveyieldsanincreaseinsupportforeducationspending
forindividualswithchildrenathomeofabout10percentagepoints.Beingineducation
(student,vocationaltrainingorinschool)isalsoassociatedwithahighersupportfor
educationspending,whichisnotsurprising.Similartotheothervariablescapturingnarrow
self‐interest,theaveragemarginaleffectofbeingineducationbecomeslargeroncetrade‐
offsandconstraintsareacknowledged.Comparedtothevariableindicatingthepresenceof
5Genderissettomaleandthesmallkidsdummyvariableissetto“0”.Allothervariables
areattheirmean.
18
smallkidsathome,however,theeffectisstrongestinthecaseoftrade‐offsbetween
spendingandhigherlevelsofpublicdebt,i.e.youngpersonsineducationaremorelikelyto
supporteducationspendingevenifthatentailshigherlevelsofpublicdebt.Atfirstsight,
thisseemscounter‐intuitivebecauseitcouldbeexpectedthatyoungindividualshavea
stronginterestinkeepingpublicdebtlevelsdown,sincetheyhavetopayfortheselaterin
theirlives.However,aplausiblealternativeexplanationisthatyoungindividualshavea
shortertimehorizonandarethereforelessconcernedaboutincreasinglevelsofpublic
debt.Ingeneral,confidenceintervalsforthisgroupofindividualsarelargerbecausethey
representarelativelysmallshareofthesample.
Finally,itisnoteworthytoemphasizeagainthatneithergeneralsupporttowardsstate
spendingnorindividualpartisanideology–twovariablesthatareusuallyhighlycorrelated
withindividualspendingpreferences–donothaveanystatisticallysignificanteffecton
supportformoreeducationspendingoncetrade‐offsbetweendifferentpartsofthewelfare
stateareseriouslytakenintoaccount.Thiseffectisstrongestinthecaseofgeneral
spendingpreferences,whereaveragemarginaleffectsareindistinguishablefromzerofor
thoseindividualswhowereconfrontedwiththeconstraintsofpensionscuts,whereasthere
arestrongpositiveeffectsintheotherthreetreatmentgroups.
Exploringchangesinthemagnitudeofpredictedeffectsacrossthedifferentmodel
specificationsingreaterdetail,figure3containssimulatedchangesinthepredicted
probabilityofsupportingmoreeducationspendingthatresultfromachangeinthe
respectiveindependentvariablefromavalueonehalfofastandarddeviationbelowthe
meantoonehalfofastandarddeviationabovethemean.6Aninterestingobservationisthat
changesinthepredictedprobabilityresultingfromamanipulationoftheindependent
variablesarestrongestinthecaseofthegroupwithnoconstraintsenforced.Thisis
particularlythecasewithregardtogeneralsupportforstatespending,butalsowithregard
toindividualeducationalbackground.Onlyindividualpartisanideologyseemstohavea
strongereffectinthefourthgroup(increasingpublicdebt),perhapsindicatingthatleft‐
6Othercontrolvariablesaresetatthemeanvalue,thedummyvariables(retired,female,
student/ineducationandsmallkidsathome)aresetatzero.
19
leaningindividualsaremorewillingtoaccepthigherlevelsofpublicdebtinorderto
financesocialinvestments.Themostimportanttake‐awayfromFigure3isagainthe
observationthatbroadsocio‐economiccleavagesaccountforsignificantlylessofthe
variationinthedependentvariable,whenindividualsareconfrontedwiththepossibilityof
spendingcutsinpensionsinordertofinanceadditionaleducationspending.
20
Figure2:Determinantsofsupportforeducationspendingacrossdifferentmodels.
21
Table2:Determinantsofsupportforeducationspending,logitmodelswithclustered
standarderrorsandcountrydummies.
Dependentvariable
(1)
(2)
(3)
(4)
Supportforeducationspending
No
Increase
Cut
Increase
constraints
taxes
pensions
publicdebt
Educational
0.183***
0.0631*
0.0547
0.0932**
background
(0.0418)
(0.0322)
(0.0656)
(0.0465)
Income
‐0.0157
0.0853*
0.0529
‐0.00933
(0.0517)
(0.0467)
(0.0355)
(0.0245)
Retired
‐0.0536
0.0704
‐0.356***
‐0.0107
(0.115)
(0.131)
(0.0756)
(0.108)
Student,inschoolor
0.433
0.514**
0.618***
0.666***
vocationaltraining
(0.293)
(0.250)
(0.162)
(0.226)
Female
0.0583
‐0.0371
‐0.0970
‐0.00473
(0.134)
(0.138)
(0.0938)
(0.0960)
Smallkidsathome
0.0966
0.252**
0.462***
0.0210
(dummy)
(0.128)
(0.122)
(0.112)
(0.169)
Left‐rightideological ‐0.0864***
‐0.0808**
‐0.0167
‐0.0912***
orientation
(0.0321)
(0.0320)
(0.0227)
(0.0290)
Generalspending
0.388***
0.266***
‐0.0602
0.269***
preferences
(0.0876)
(0.0607)
(0.0526)
(0.0711)
Constant
0.0206
‐0.705*
‐0.791**
‐1.116**
(0.447)
(0.395)
(0.375)
(0.492)
Countrydummies
Yes
Yes
Yes
Yes
PseudoR2
0.0782
0.0577
0.0236
0.0380
Observations
1,771
1,741
1,681
1,700
Robuststandarderrorsinparentheses
***p<0.01,**p<0.05,*p<0.1
22
Figure3:Comparingthemagnitudeofeffectsacrossmodels.
0,08
0,06
0,04
Education
0,02
Left‐Right
0
‐0,02
No
constraints
Taxincrease Pensioncuts Debtincrease
Statism
‐0,04
‐0,06
Note:Thefigurecontainschangesinpredictedprobabilityinthesupportforeducation
spendingforthedifferentgroups,simulatingachangeintherespectiveindependent
variablefrom½ofastandarddeviationbelowtothemeanto½ofastandarddeviation
abovethemean,holdingothercontinuousvariablesattheirmeansandsettingthedummy
variables(oldage,femaleandsmallkidsathome)tozero.
23
Thedeterminantsoftrade‐offsinthewelfarestate(II)
Thesecondsetofquestionsalludedtoabovearetwoquestionsthatforcerespondentsto
choosebetweensocialinvestmentsontheonehandandpassivesocialtransfersonthe
other.Inthiscase,thetotalsamplewassplitintotwogroups(notfourasabove),and
individualsselectedatrandomweregiveneithertheoneortheotherquestion.Thefirst
questionisaimedatmeasuringsupportforsocialinvestmentandreadsthus:
“Whatdoyouthinkaboutthefollowingstatement?Tobeabletofinancemorespendingon
educationandfamilies,thegovernmentshouldcutbackonoldagepensionsand
unemploymentbenefits.”
Incontrast,thesecondquestionissupposedtocapturesupportforthepassivewelfare
state:
“Whatdoyouthinkaboutthefollowingstatement?Tobeabletofinancemorespendingon
oldagepensionsandunemploymentbenefits,thegovernmentshouldcutbackspendingon
educationandfamilies.”
Again,respondents’answerswerecodedon5‐pointLikertscale(stronglyagree,agree,
neitheragreenordisagree,disagree,stronglydisagree).Forthemultivariateregression
analyses,thisvariableisthentransformedintoabinaryvariable,whichtakesthevalueof
“1”forindividualswhostronglyagreeoragreeand“0”fortheremainingcases.Thisisa
ratherconservativemeasureofsupportforsocialinvestmentandpassivetransfers,
becausetherearefewercasesinthesecategoriescomparedtothegeneralspending
questionusedabove.7
Whenlookingatthedescriptivestatistics,oneconspicuous,butunsurprisingfindingisthat
ingeneral,supportforspendingdeclinesprecipitouslyoncerespondentsareconfronted
withtrade‐offsacrosspolicyfields.Intheunconstrainedquestiononsupportforeducation
7Inpracticalterms,theresultsdonotdiffermuchnomatterwheretheexactcut‐offpointis
drawn.
24
spendingdiscussedabove,about73percentexpressedsupportformoreormuchmore
educationspending.Incontrast,only11.26percentdemandmoreormuchmorespending
on“educationandfamilies”ifthisadditionalspendingwouldbefinancedwithcutbacksin
passivesocialtransfers.Evenfewer(7.82percent)wouldliketoseemoreormuchmore
spendingonoldagepensionsandunemploymentbenefitsifthisistobefinancedby
cutbacksineducationspending(seetable3fordetails).Thelargestshareofrespondents
disagreeswithbothstatements,averysizableminorityevenstrongly.Theoverallmeanin
thesampleis2.11forthefirstquestionand2.06forthesecond.Thiscouldindicatethat
cuttingbackspendingonpensionsandunemploymentbenefitsforthebenefitofchildren
andfamiliesisslightlylessunpopular8relativetotheopposite,buttheoverallmessageis
clearandconfirmsthesimilarfindingoftheprevioussection:Individual‐levelsupportfor
spendingincreasesdropsprecipitouslyoncerespondentsareconfrontedwithacleartrade‐
offproblembetweendifferentsocialpolicies.Inotherwords,individualsdon’tlikebeing
forcedtochoose,independentofwhethertheyareaskedfortheirsupportforspendingon
socialinvestmentsorpassivesocialtransfers.
Infigure4andtheassociatedtable4,Ipresentfindingsfrommultivariateregression
analysesonthedeterminantsofsupportforsocialinvestmentandpassivesocialtransfers,
respectively.Ialsoincludeathirddependentvariablethatcapturesgeneralpreferencesfor
redistribution.Thisisdoneinordertoachieveasimilarcomparisonasabovebetweena
relativelyunconstrainedsettingontheonehandandamoreconstrainedsituationonthe
other,whererespondentsareconfrontedwithtrade‐offs.Thiswayitispossibletosee
whetherthecoalitionsthatsupportsocialinvestmentsandsocialtransfersaresimilarto
classicalredistributivecoalitions.Supportforredistributionismeasuredbyrespondents’
degreeofagreementwiththestatement:“Thegovernmentshouldreduceincome
differencesbetweentherichandthepoor.”Asabove,Iemployasimplelogitmodelwith
standarderrorsclusteredbycountryandcountrydummiesandIusethesamesetof
independentvariables.Again,resultsarepresentedintheformofplottedaveragemarginal
effects(Figure5)aswellasastandardregressiontable(Table4).
8Anunpairedttestconfirmsthatthedifferenceinmeansbetweenthesetwogroupsis
statisticallysignificantatthe.05level.
25
Ingeneral,thefindingspointinthesamedirectionastheonesabove:First,supporting
coalitionsforsocialinvestmentsdifferfromthoseofpassivesocialtransfersinimportant
ways.Increasingspendingonsocialinvestments,evenifitentailscuttingbackpensionsand
unemploymentbenefits,isstronglysupportedbyindividualswithsmallchildrenathomeas
wellasstudents.Thepredictedchangeintheprobabilityofsupportingsocialinvestmentis
anincreaseofabout5percentagepointsinbothcases.Incontrast,retiredpeoplearemore
likelytoopposesocialinvestmentsiftheyneedtobefinancedwithcutbacksinpension
spending.Here,thepredictedchangeinprobabilityisadecreaseofsimilarmagnitude(4
percentagepoints).Whenitcomestopassivesocialtransfers,however,individualswith
smallkidsathomearemorelikelytoopposespending(thepredictedchangeinprobability
is3percentagepoints).Ialsofindthathighlyeducatedindividualsaremoreopposedto
spendingonpassivesocialtransfers.Themagnitudeoftheeffectiscomparabletothatof
the“smallkids”dummy.
Second,therearealsosomeindicationsthatsupportforsocialinvestmentvs.passive
transfersisdeterminedbybelongingtoparticularconstituencygroupsratherthan
cleavagesbroadlydefined,whichisrevealedwhencomparingthefindingsfromthefirst
twomodelswiththelatteronefocusingongeneralredistributivepreferences.Inthecaseof
generalredistributivepreferences,variablessuchaseducation,income,left‐right
ideologicalorientationandgeneralspendingpreferenceshavemoreexplanatorypower
thaninthecaseofpreferencesforsocialinvestmentsandpassivetransfers,exceptforthe
factthatstudentsarealsolesslikelytosupportredistribution,whereasretiredpersonsare
morelikelytodoso.Whatisalsointerestingisthatgeneralsupportforsocialspendingis
positivelyassociatedwithsupportforredistribution(asonewouldexpected),but
negativelyassociatedwithsupportforsocialinvestments.
Takentogether,theseresultscanberegardedasastrongconfirmationofHypothesis2.
Withsomeexceptions,thegeneralpatternisthatvariablesrelatedtobroadsocietal
cleavagesaremorepowerfulinaccountingforvariationingeneralsupportfor
redistribution.Incontrast,whenrespondentsareforcedtomakechoicesbetweensocial
26
investmentsandpassivetransfers,variablesrelatedtoself‐interestmorenarrowlydefined
suchashavingsmallchildrenathome,beingineducationorretiredbecomemore
important.
Table3:Trade‐offsinspendingpreferences:socialinvestmentvs.passivesocialtransfers.
Morespendingoneducationand
Morespendingonoldage
familiesandcutbackonoldage
pensionsandunemployment
pensionsandunemployment
benefitsandcutbackspending
benefits
oneducationandfamilies
Percent
Cumulative
Percent
Cumulative
Stronglyagree
1.60
1.60
1.14
1.14
Agree
9.66
11.26
6.68
7.82
Neitheragree
15.29
26.55
15.12
22.94
Disagree
43.37
69.92
49.62
72.56
Strongly
28.44
98.36
25.20
97.76
100.00
2.24
100.00
nordisagree
disagree
Don’tknow/no 1.64
answer
Overallmean
2.11(0.986)
2.06(0.887)
(andstandard
deviation)
27
Figure5:Supportforsocialinvestments,passivetransfersandredistribution.
28
Table4:Determinantsofsupportforspendingonsocialinvestmentsandpassivesocial
transfersaswellasredistributionmoregenerallydefined.
(1)
(2)
(3)
Dependentvariable Supportfor Supportfor
General
spendingon spendingon supportfor
social
social
redistri‐
investments transfers
bution
Educational
0.0240
‐0.236***
‐0.0559**
background
(0.0344)
(0.0350)
(0.0255)
Income
‐0.0311
‐0.0981*
‐0.136***
(0.0241)
(0.0511)
(0.0193)
Retired
‐0.384***
0.168
0.287***
(0.138)
(0.140)
(0.0863)
Student,inschoolor 0.527***
‐0.103
‐0.531***
vocationaltraining
(0.196)
(0.418)
(0.157)
Female
‐0.217
‐0.0237
‐0.00111
(0.155)
(0.0882)
(0.0489)
Smallkidsathome
0.452***
‐0.568***
‐0.0437
(dummy)
(0.147)
(0.180)
(0.110)
Left‐rightideological
0.0209
0.0945**
‐0.197***
orientation
(0.0209)
(0.0376)
(0.0287)
Generalspending
‐0.381***
0.0751
0.458***
preferences
(0.0706)
(0.0741)
(0.0231)
Constant
‐0.999***
‐2.621***
1.000***
(0.256)
(0.310)
(0.218)
PseudoR2
0.0508
0.0594
0.1197
Observations
3,430
3,423
6,873
Robuststandarderrorsinparentheses
***p<0.01,**p<0.05,*p<0.1
29
Conclusions
Thispaperhasaddedanewperspectivetotheongoingdebateaboutwelfarestate
retrenchmentbyfocusingontrade‐offsbetweendifferentpartsofthewelfarestate,in
particularsocialinvestmentsandpassivesocialtransfers.Ifoundthatcitizensindeed
dislikebeingforcedtochoosebetweendifferentkindsofsocialpolicies.Although
investmentsineducationareverypopularinprinciple,supportforspendingdrops
precipitouslywhencitizensareremindedthatincreasingspendingimpliesraisingtaxes,
increasinglevelsofpublicdebtorcuttingbackpensions.Ifforcedtochoose,citizensare
evenmorelikelytoaccepthighertaxesorhigherlevelsofdebtratherthancuttingback
otherpartsofthewelfarestate.
Thesecondmajorfindingofthepaperisthatthepoliticalcleavagestructuresindistributive
conflictsthatplayoutwithinexistingwelfarestatesmightbeverydifferentfromthekindof
coalitionsthathavebeenidentifiedinthegeneralstruggleabouttheirexpansion.Existing
researchalmostexclusivelyfocusesonthelatterforthesimplereasonthatgenerally
availablesurveysofpublicopiniondonotincludequestionsthatwouldadequatelymeasure
trade‐offs.Therobustandconsistentfindingthatemergesfromtheanalysisofouroriginal
surveydataisthatdistributivestruggleswithinexistingwelfarestatesaremuchmore
determinedbyself‐interestmorenarrowlydefinedratherthanbroadsocietalcleavages
relatedtoideologicalorientations,norms,valuesorsocio‐economicposition.
Thisalsohasimplicationsfortheworldofpolicy‐making.Theviabilityofthe“social
investmentstate”asthenewdominatingparadigminsocialpolicy‐makingcouldbe
hamperedbythepoliticaloppositionofpowerfulconstituencygroups,ifexpandingsocial
investmentswouldgoalongwithcutbacksinotherpartsofthewelfarestate.Giventhe
limitedfiscalroomofmaneuverofmostEuropeangovernments,thisseemsunavoidablefor
themostpart,however,sinceincreasingtaxesorlevelsofpublicdebtisnolongerfeasible.
Inaway,therefore,welfarestatesare–again–caughtbetweenarockandahardplace,
becausetheadoptionofexistingwelfarestatestomeetnewdemandsrelatedtonewsocial
riskswillmostlikelygoalongwithpoliticalconflictsaboutthecuttingbackofexisting
schemes.The“recalibration”ofwelfarestatesalongthenewsocialinvestmentparadigmis
30
notimpossibleasHemerijck(2013)showsinhiscomprehensivebook.ButasIhaveargued
inthispaper,policy‐makersmayexpecttofaceuphillbattlesagainstficklepublicopinion.
31
References
Alesina,A.,&Angeletos,G.‐M.(2005).FairnessandRedistribution.AmericanEconomic
Review,95(4),960‐980.
Allan,J.P.,&Scruggs,L.(2004).PoliticalPartisanshipandWelfareStateReformin
AdvancedIndustrialSocieties.AmericanJournalofPoliticalScience,48,496‐512.
Ansell,B.W.(2008).UniversityChallenges:ExplainingInstitutionalChangeinHigher
Education.WorldPolitics,60(January2008),189‐230.
Ansell,B.W.(2010).FromtheBallottotheBlackboard:TheRedistributivePoliticalEconomy
ofEducation.Cambridge:CambridgeUniversityPress.
Béland,D.(2010).Reconsiderungpolicyfeedback:Howpoliciesaffectpolitics.
Administration&Society,42(5),568‐590.
Boeri,T.,Börsch‐Supan,A.,Tabellini,G.,Moene,K.O.,&Lockwood,B.(2001).WouldYou
LikeToShrinktheWelfareState?ASurveyofEuropeanCitizens.EconomicPolicy,
16(32),9‐50.
Bonoli,G.(2013).TheOriginsofActiveSocialPolicy:LabourMarketandChildcarePoliciesin
aComparativePerspective.Oxford,NewYork:OxfordUniversityPress.
Breunig,C.,&Busemeyer,M.R.(2012).FiscalAusterityandtheTrade‐offbetweenPublic
InvestmentandSocialSpending.JournalofEuropeanPublicPolicy,19(6),921‐938.
Brooks,C.,&Manza,J.(2006).WhyDoWelfareStatesPersist?JournalofPolitics,68(4),816‐
827.
Brooks,C.,&Manza,J.(2007).WhyWelfareStatesPersist:TheImportanceofPublicOpinion
inDemocracies.Chicago:UniversityofChicago.
Busemeyer,M.R.(2012).InequalityandthePoliticalEconomyofEducation:AnAnalysisof
IndividualPreferencesinOECDCountries.JournalofEuropeanSocialPolicy,22(3),
219‐240.
Busemeyer,M.R.,Goerres,A.,&Weschle,S.(2009).Attitudestowardsredistributive
spendinginaneraofdemographicageing:Therivalpressuresfromageandincome
in14OECDcountries.JournalofEuropeanSocialPolicy,19(3),195‐212.
Campbell,A.(2002).Self‐Interest,SocialSecurity,andtheDistinctiveParticipationPatterns
ofSeniorCitizens.AmericanPoliticalScienceReview,96(3),565‐574.
Campbell,A.L.(2012).Policymakesmasspolitics.AnnualReviewofPoliticalScience,15,
333‐351.
Citrin,J.(1979).Dopeoplewantsomethingfornothing:Publicopinionontaxesand
governmentspending.NationalTaxJournal,32(June1979),113‐129.
Corneo,G.,&Grüner,H.P.(2002).Individualpreferencesforpoliticalredistribution.Journal
ofPublicEconomics,83,83‐107.
Cusack,T.,Iversen,T.,&Rehm,P.(2006).RisksatWork:TheDemandandSupplySidesof
GovernmentRedistribution.OxfordReviewofEconomicPolicy,22(3),365‐389.
Erikson,R.S.,MacKuen,M.B.,&Stimson,J.A.(2002).Themacropolity.Cambridge,New
York:CambridgeUniversityPress.
Fleckenstein,T.,Saunders,A.M.,&Seeleib‐Kaiser,M.(2011).TheDualTransformationof
SocialProtectionandHumanCapital:ComparingBritainandGermany.Comparative
PoliticalStudies,44(12),1622‐1650.doi:Doi10.1177/0010414011407473
Fong,C.(2001).Socialpreferences,self‐interest,andthedemandforredistribution.Journal
ofPublicEconomics,82,225‐246.
Giger,N.,&Nelson,M.(2013).TheWelfareStateortheEconomy?Preferences,
32
Constituencies,andStrategiesforRetrenchment.EuropeanSociologicalReview,
29(5),1083‐1094.
Goerres,A.,&Prinzen,K.(2012).CanWeImprovetheMeasurementofAttitudesTowards
theWelfareState?AConstructiveCritiqueofSurveyInstrumentswithEvidence
fromFocusGroups.SocialIndicatorsResearch,105,515‐534.
Hasenfeld,Y.,&Rafferty,J.A.(1989).TheDeterminantsofPublicAttitudesTowardthe
WelfareState.SocialForces,67(4),1027‐1048.
Häusermann,S.(2010).ThePoliticsofWelfareStateReforminContinentalEurope:
ModernizationinHardTimes.Cambridge,NewYork:CambridgeUniversityPress.
Hemerijck,A.(2013).Changingwelfarestates.Oxford,NewYork:OxfordUniversityPress.
Iversen,T.,&Soskice,D.(2001).AnAssetTheoryofSocialPolicyPreferences.American
PoliticalScienceReview,95(4),875‐893.
Kangas,O.(2003).Thegrasshopperandtheants:popularopinionsofjustdistributionin
AustraliaandFinland.JournalofSocio‐Economics,31,721‐743.
Kenworthy,L.,&McCall,L.(2008).Inequality,PublicOpinionandRedistribution.Socio‐
EconomicReview,6,35‐68.
Korpi,W.,&Palme,J.(2003).Newpoliticsandclasspoliticsinthecontextofausterityand
globalization:Welfarestateregressin18countries,1975‐95.AmericanPolitical
ScienceReview,97(3),425‐446.
Kumlin,S.,&Stadelmann‐Steffen,I.(2014).Citizens,policyfeedback,andEuropeanwelfare
states.InS.Kumlin&I.Stadelmann‐Steffen(Eds.),Howwelfarestatesshapethe
democraticpublic:Policyfeedback,participation,voting,andattitudes(pp.3‐16).
Cheltenham,UK,Northampton,MA,USA:EdwardElgar.
Lynch,J.,&Myrskylä,M.(2009).AlwaystheThirdRail?PensionIncomeandPolicy
PreferencesinEuropeanDemocracies.ComparativePoliticalStudies,42(8),1068‐
1097.
Mettler,S.(2005).SoldierstoCitizens:TheG.I.BillandtheMakingoftheGreatestGeneration.
Oxford,NewYork:OxfordUniversityPress.
Mettler,S.,&Soss,J.(2004).TheConsequencesofPublicPolicyforDemocraticCitizenship:
BridgingPolicyStudiesandMassPolitics.PerspectivesonPolitics,2(1),55‐73.
Morel,N.,Palier,B.,&Palme,J.(Eds.).(2011).TowardsaSocialInvestmentState?Ideas,
PoliciesandChallenges.Bristol:PolicyPress.
Pierson,P.(1993).WhenEffectBecomesCause:PolicyFeedbackandPoliticalChange.
WorldPolitics,45(4),595‐628.
Pierson,P.(1994).DismantlingtheWelfareState?Reagan,Thatcher,andthePoliticsof
Retrenchment.Cambridge:CambridgeUniversityPress.
Pierson,P.(1996).TheNewPoliticsoftheWelfareState.WorldPolitics,48,143‐179.
Pierson,P.(2000).IncreasingReturns,PathDependence,andtheStudyofPolitics.
AmericanPoliticalScienceReview,94(2),251‐267.
Pierson,P.(2001).CopingwithPermanentAusterity:WelfareStateRestructuringin
AffluentDemocracies.InP.Pierson(Ed.),TheNewPoliticsoftheWelfareState(pp.
410‐456).Oxford:OxfordUniversityPress.
Pierson,P.(2011).Thewelfarestateovertheverylongrun.ZeSArbeitspapier,02/2011.
Rehm,P.(2009).RiskandRedistribution:AnIndividual‐LevelAnalysis.Comparative
PoliticalStudies,42(7),855‐881.
Rehm,P.(2012).SocialPolicybyPopularDemand.WorldPolitics,63(2),271‐299.
Soroka,S.N.,&Wlezien,C.(2010).DegreesofDemocracy:Politics,PublicOpinion,andPolicy.
33
Cambridge,NewYork:CambridgeUniversityPress.
Stephens,J.D.,Huber,E.,&Ray,L.(1999).TheWelfareStateinHardTimes.InH.Kitschelt,
P.Lange,G.Marks&J.D.Stephens(Eds.),ContinuityandChangeinContemporary
Capitalism(pp.164‐193).Cambridge,NewYork:CambridgeUniversityPress.
Stimson,J.A.,MacKuen,M.B.,&Erikson,R.S.(1995).Dynamicrepresentation.American
PoliticalScienceReview,89(3),543‐565.
Streeck,W.,&Mertens,D.(2011).FiscalAusterityandPublicInvestment:IsthePossiblethe
EnemyoftheNecessary?MPIfGDiscussionPaper,11/12.
Svallfors,S.(1997).WorldsofWelfareandAttitudestoRedistribution:AComparisonof
EightWesternNations.EuropeanSociologicalReview,13(3),283‐304.
Svallfors,S.(2004).Class,AttitudesandtheWelfareState:SwedeninComparative
Perspective.SocialPolicy&Administration,38(2),119‐138.
Svallfors,S.(2012).WelfareStatesandWelfareAttitudes.InS.Svallfors(Ed.),Contested
welfarestates:WelfareattitudesinEuropeandbeyond(pp.1‐24).Stanford:Stanford
UniversityPress.
VanKersbergen,K.,Vis,B.,&Hemerijck,A.(2014).TheGreatRecessionandwelfarestate
reform:Isretrenchmentreallytheonlygameleftintown?SocialPolicy&
Administration,48(7),883‐904.
VanOorschot,W.(2006).MakingthedifferenceinEurope:deservingnessperceptions
amongcitizensofEuropeanwelfarestates.JournalofEuropeanSocialPolicy,16(1),
23‐42.
Vandenbroucke,F.,&Vleminckx,K.(2011).Disappointingpovertytrends:Isthesocial
investmentstatetoblame?JournalofEuropeanSocialPolicy,21(5),450‐471.
Walter,S.(2010).GlobalizationandtheWelfareState.TestingtheMicrofoundationsofthe
CompensationHypothesis.InternationalStudiesQuarterly,54(2),403‐426.
Weaver,R.K.(1986).Thepoliticsofblameavoidance.JournalofPublicPolicy,6(4),371‐
398.
Wlezien,C.(1995).ThePublicasThermostat:DynamicsofPreferencesforSpending.
AmericanJournalofPoliticalScience,39(4),981‐1000.
34
Appendix
TableA1:Numberofinterviewspercountryandresponserate.
Country
Numberofinterviews
Responserate
Denmark
1,000
36%
France
1,003
28%
Germany
1,500
26%
Ireland
1,000
20%
Italy
1,002
28%
Spain
1,000
24%
Sweden
1,100
33%
UnitedKingdom
1,300
21%
Total
8,905
27%
35
TableA2:Meansandstandarddeviation(inparentheses)formostimportantconfounding
variablesacrossgroups.
“Noconstraints” “Taxincrease”
“Pensioncuts” “Debt
increase”
Max.number
2,237
2,238
2,148
2,178
3.42(1.52)
3.37(1.52)
3.37(1.53)
3.36(1.54)
3.03(1.39)
3.07(1.41)
3.09(1.44)
3.03(1.42)
Retired
0.32(0.47)
0.34(0.47)
0.35(0.48)
0.34(0.47)
Student/in
0.04(0.20)
0.04(0.20)
0.03(0.18)
0.03(0.18)
Female
0.55(0.50)
0.54(0.50)
0.53(0.50)
0.54(0.50)
Smallkidsat
0.14(0.35)
0.13(0.34)
0.14(0.35)
0.15(0.36)
Left‐rightid.
4.81(2.30)
4.76(2.34)
4.86(2.36)
4.86(2.31)
General
3.36(0.97)
3.41(0.95)
3.36(0.97)
3.36(0.94)
ofvalidobs.
Educational
level
Individual
income
(quintiles)
education
home
spendingpref.
36
TableA3:Distributionofnumberofobservationsacrosscountries.
Control
Treatment1
Treatment2
Treatment3
Denmark
266
225
231
253
France
263
231
255
251
Germany
358
377
378
377
Ireland
258
262
230
242
Italy
243
274
240
228
Spain
272
245
237
240
Sweden
264
318
261
245
UK
313
306
316
342
37