ReproducibleandReplicable:AnEmpiricalAssessmentoftheSocialConstructionofPoliticallyRelevantTargetGroups

RebeccaJ.KreitzerAssistantProfessorofPublicPolicyUniversityofNorthCarolina-ChapelHillCB#3435,AbernethyHallChapelHill,NC27599rkreit@email.unc.eduCandisWattsSmithAssistantProfessorofPublicPolicyUniversityofNorthCarolina-ChapelHillCB#3435,AbernethyHallChapelHill,NC27599cwsmith@unc.eduAcknowledgementsWethankHelenIngramandAnneSchneideraswellasFrankBaumgartnerfortheirfeedbackandinsightsonthisareaofourresearchagenda.WealsohavegreatappreciationforLorinBrucknerofUNC’sDavisLibrary’sResearchHubforherdatavisualizationexpertiseandassistance.

1

Nearlyaquartercenturyago,SchneiderandIngram(1993)introducedatheoryof

socialconstructionoftargetpopulationsintheAmericanPoliticalScienceReview.

Thisarticlehasbeencitedover1,700times,asthetheoryhasinspireddiscussion

andsparkeddebate.Thistheoryproposesthatwearelikelytoseesystematically

biasedpolicypatternsbecausepolicymakersarepressuredtorewardpositively

constructedgroups,especiallythosewithpoliticalpower,andalsobecause

policymakersareincentivizedtodeveloponerouspoliciesfornegatively

constructedgroups,beingespeciallyharshonthosegroupsthatalsohavelittle

power.Thistheoryhasbeenpivotalforpolicyscholarsbecauseitincorporates

value-ladencomponentsofthepolicymakingprocessinadditiontorationaland

instrumentalcomponentsofdesign(SchneiderandSidney2009).

Thisconcepthasprovidedasturdytheoreticalfoundationforthosewhoseek

toexplainnotonlytheshapeofpolicydesignbutalsothefeedbackandfeed-

forwardeffectsofpolicy,thoughithasitscritics.Sabatier(1999),forexample,

arguedthatthisframeworkis“largelynonfalsifiable”(11),butSchneiderand

Ingramaswellofaslewofscholarshaveprovidedempiricalevidenceforthetheory

aswellashavehighlightedthepredictivepoweroftheframework.Whilethe

scholarshaverebuttedcritiquesaswellasstrengthenedthetheoryovertime,there

aretwomattersthathaveyettoberesolvedinatidyway.

First,theauthorsdevelopfour“idealtype”targetgroups:advantaged,

contenders,dependents,anddeviants.Since1993,scholarshaveprovidedexamples

oftheseidealtypesthoughlargelyrelyingoncasestudies.Whilecasestudies

providerigorousanalysis,takentogether,westillfindthattherehasnotbeena

2

systematic,cross-casevalidationofthecategorizationofvarioustargetgroups.

Second,theauthorsnotethatsocialconstructionsareoftensubjecttocontention,

therebyhighlightingthecomplexityintroducedbythesocialconstructionoftarget

groups,butultimatelythetheoryimpliesthatthereisconsensusaroundthesocial

constructionofmanytargetpopulations.Indeed,SchneiderandIngramnote,“The

actualsocialconstructionoftargetgroups,aswellashowwidelysharedthe

constructionsare,aremattersforempiricalanalysis.Socialconstructionsare

measurable,empirical,phenomena”that“haveboundariesthatareempirically

verifiableandexistwithinobjectiveconditions”(1993,335).Toourknowledge,

therehasnotbeenasystematic,orstandardized,categorizationofpoliticallysalient

targetgroupsbasedonSchneiderandIngram’sfouridealtypes,norhastherebeen

anempiricalassessmentofwhetherortheextenttowhichconsensusaroundthe

socialconstructionsofthesevarioustargetgroupsexists.

Ourgoalforthisletteristoassesstheunderlyingassumptionsofthetheory.

Specifically,weask,canwepinpointthelocationoftargetpopulationsonSchneider

andIngram’stwo-by-twomatrixinasystematic,standardizedway?Ifso,arethere

widelysharedsocialconstructionsofsalienttargetgroups?Towhatdegreedohigh

levelsofconsensusemergearoundpoliticallyrelevantgroups?Whatarethe

theoreticalandpoliticalimplicationsforalackofconsensus,shouldsuchan

outcomebeuncovered?Werevisitthistheorytoofferanovelperspective,anddoso

byleveragingadvancesintechnologyandmethodologicalstrategies.

Researchshowsthatcrowdsourcingthetaskofcategorizationtoalarge

numberofnon-expertsallowsscholarstogeneratereproducibleandreplicable

3

resultsthatmimicthoseofexperts(Benoitetal.2016).Bycrowdsourcingthetask

ofevaluatingthesocialconstructionofvarioustargetpopulations,weareableto

placeaplethoraoftargetgroupsonthetwo-dimensionsofimportancetoSchneider

andIngram’stheory:poweranddeservingness.Upuntilnow,scholarshave

individuallytriedtosortthismatteroutforthemselves,whichinhibitsresearchers

fromreplicatingnotonlydataanalysis,butalsothedatacollectionandgroup

categorizationprocessesofotherscholars.Secondly,weareabletodiscernthe

extenttowhichaconsensusemergesaroundthestereotypesofthesegroups.

Ourresultssettlesomeofthemajorpointsofcontentionaroundthistheory

aswellasprovidenewinsightsintosocialconstructiontheoryinaneramarkedby

politicalpolarization,scapegoating,anddegenerativepolitics.Weendby

commentingonhowourmethodologicalapproachandresultsservetoreinvigorate

discussionandopennewavenuesofresearchforpolicyscholars.

SocialConstructionTheory

Schneider,Ingram,andtheircolleagues(SchneiderandSidney2009,Schneider,

Ingram,andDeLeon2014)explainthatwecanbestunderstandtheconstraintsand

motivationsofpolicymakerstodesignpoliciesthatcreate,maintain,orameliorate

disparitiesbetweenandamonggroupsthroughatheoryofsocialconstructionof

targetpopulations.Theyhomeinontwocharacteristicsoftargetgroups:social

constructionandperceivedpoliticalpower.

AccordingtoSchneiderandIngram,“socialconstructionsarestereotypes

aboutparticulargroupsofpeoplethathavebeencreatedbypolitics,culture,

4

socialization,history,themedia,literature,religion,andthelike”(1993,335).Those

withpositivesocialreputationsareviewedasdeserving,intelligent,public-spirited,

hardworking,andthelike.Ontheotherendofthespectrum,therearegroupsthat

areimbuedwithnegativestereotypes,suchasundeserving,selfish,andlazy.Inthis

context,politicalpowerisbestunderstoodas“votes,wealth,andpropensityofthe

grouptomobilizeforaction”aswellasaccesstopoliticallyinfluentialindividualsor

institutions(SchneiderandIngram1993,335).

Itisattheintersectionofsocialconstructionandperceivedpoliticalpower

thatcreatesthefouridealtypes:advantaged,contenders,dependents,anddeviants.

Advantagedtargetpopulationsarethosethatareviewedasdeservingandpolitically

powerful.Accordingthetheory,policymakersarelikelytoprovidebeneficialpolicy

treatmenttotheadvantaged.Contendersarethosewhoarepoliticallypowerfulbut

havepoorreputations.Policymakersarelikelytoprovidesubrosa,discretely

hiddenbenefitstocontendersduetotheirpoliticalpower,butarewillingtoprovide

punitive(buthollow)policytothisgroupwhenpublicinterestishigh.Dependent

targetgroupsaresympatheticgroupsthatarepositivelyconstructedbuthavelittle

politicalpower.Policymakershavelittleincentivetoproduceeasilyaccessibleand

highlybeneficialpoliciesfordependents,sowhenbenefitsareallocated,theytend

tobesymbolicorcomewithstringsattached(e.g.paternalisticsocialwelfare

programs).Finally,deviantsarethosetargetpopulationsthatareclearlyassociated

withnegativestereotypesandhavelittlepoliticalpower.Policymakersgain

politicalcapitalfordevelopingpunitivepoliciesforgroupscategorizedasdeviant.

5

Putsimply,politicianslike(andarerewardedfor)doinggoodthingsforgood

people,andbadthingstobadpeople.

Crowdsourcing

SchneiderandIngramprovidedagooddealoflatitudetoscholarsmeasurethe

socialconstructionoftargetgroups,guidingthemonlywiththeinstructiontorely

on“texts,suchaslegislativehistories,statues,guidelines,speeches,media

coverage…interviewsorsurveysofpolicymakers,mediarepresentatives,members

ofthegeneralpublic,andpersonswithinthetargetgroupitself”(1993,335).Upon

reflection,wecanseehowcomplexand“messy”thisprocesscanbe.Forexample,

SchroedelandJordan(1998)explainthatintheirefforttodeterminewherevarious

groupsfitintothetargetpopulationtypology,theyhadtousethreemethods,but

stillfoundthatthereweregroupsthatweredifficulttocategorize.Forthosegroups,

suchasgaymen,theydecidedtocategorizegroupsby“comparingtheirattributes

relativetooneanother”(e.g.gaymenversusintravenousdrugusers)(113).

Ultimately,theyclassifiedgaymenascontendersanddrugusersasdeviantsthough

onecouldimagineiftheyusedanothersetofcomparisons,theirclassificationmay

havecomeoutdifferently.1

Weshowherethatthetaskofuncoveringsocialconstructionsrootedin

“objectivereality”(SchneiderandIngram,335)canbefulfilledthrough

crowdsourcing.Today’stechnologyprovidesawidearrayoflow-costopportunities

1Ourresultsshowthatgaymenareontheborderbetweenbeingclassifiedasdependentsanddeviants,whereasopioidaddictsaremoreclearlyinthedeviantcategory.

6

tocrowdsource,ortheabilitytoemployalargegroupofindividualstodoaseriesof

smalltasksineffortstoproblem-solveonamassivescale(Benoitetal.2016).

Wearguethatthisareaofstudyactuallyrequiressomethingakinto

crowdsourcingbecausethetheoryhingesonwhatelectedofficialsthinktheir

constituents’perceptionsoftargetgroupsare.Crowdsourcingthecategorizationof

politicallyrelevantgroupsprovidesscholarswithempiricallyverifiable,systematic,

replicableresults,andallowsthemtodiscerntheextenttowhichaconsensus

emergesonthestereotypesofvariousgroups.GiventhefactthatSchneider,Ingram,

andtheircolleaguesplacemeaning-makingofgroupsandtheirreputationsatthe

centeroftheirtheory,notingthat“sharedunderstandingsamongpeoplegiveriseto

rules,norms,identities,concepts,andinstitutions”(SchneiderandSidney2009,106

,emphasisadded),wetakethattomeanthatsocialconstructionsarisefromthe

aggregationofpublicattitudes,whichcanbeuncoveredbyanalyzingpublic

discourse.Bycrowdsourcing,weareabletoaccuratelydeterminetheplacementof

73targetpopulationsalongthetwo-dimensionsoutlinedbythetheoryaswellasto

discernthelevelofagreementthatarisesaroundtheseconstructionsamong

membersofsociety.

MethodsandData

Consideringtheconstructionofgroupsisbasedlargelyonthepublicimageor

stereotypesofvariousgroups,weemployedthelaborof1,572workerstoappraise

thesocialconstructionof73targetpopulations.Werelyoncrowdsourcingthrough

Amazon’sMechanicalTurk(MTurk)toevaluatethedeservingnessandperceived

powerofseveralgroups.MTurkisamarketplacewhereindividualscanoptinto

7

performtasksbestdelegatedtohumans.ScholarsshowthatalthoughMTurk

workersarelessrepresentativethannationalprobabilitysamples,theyaremore

representativethansamplesrecruitedoncollegecampuses(Berinsky,Huber,and

Lenz2012,HuffandTingley2015).ResearchershavealsofoundMTurk

respondents’worktobereliable,especiallythosewhohavehighlyregarded

reputationsbasedonMTurkevaluationstandards(Peer,Vosgerau,andAcquisti

2014,Rouse2015).Oursamplereliesonthosewithatleasta99%accuracyrating,

andweincludedattentionchecks,whichresearchshowstoimprovethequalityof

thedata(Rouse2015).

Weposedtwotasks;eachconcernsoneaspectofthecentraltheory.First,we

explainedtotheworkers,“Somegroupsaremoreunited,easytomobilize,wealthy,

skilled,focusedontheirgoals,oraccustomedtovotingordirectlycontactingpublic

officials.”Weaskedthemtoevaluategroupsfrompowerless(0)topowerful(100)

basedontheseattributes.Second,wetaskedthemtoevaluatedeservingness,from0

(thoseviewedasgreedy,disrespectful,disloyal,immoral,disgusting)to100(groups

describedasgood,smart,hardworking,loyal,disciplined,generous).2Theworkers

weretaskedtoevaluateseventy-threetargetgroups,suchasethno-racialsubgroups

(e.g.Whitemen,youngBlackmen),professionalgroups(e.g.attorneys,teachers),

criminals(e.g.sexoffenders,welfarecheats),andmanyotherpoliticallysalient

groups(e.g.unions,illegalimmigrants).

Results

2Weoutlinetheexactwordingofthetasksaswellaslistthe73groups,theirpointestimates,andstandarddeviationsintheappendix.

8

Figure1providesanillustrationoftheplacementofthesegroupsonthetwo-

dimensionalmatrixproposedbySchneiderandIngram(1993).Weaveragethe

scoresofthe1,572MTurkworkersgavetoeachofthegroupsondeservingnessand

power.Whereagroupfallsonthehorizontalaxisrelatestotheirlevelsof

deservingness,wherethosewhoarehighindeservingnessareontheleft.The

verticalaxisisrelatedtoperceivedlevelsofpoliticalpower.Thoseatthetopare

perceivedasverypowerful,whereasthoseatthebottomareviewedaspowerless.3

ThoughsomehaveinterpretedSchneiderandIngramtosuggestthatgroupsare

categorizedinadichotomousway—asdeservingorundeserving,andpolitically

powerfulorweak(SchroedelandJordan1998)—itshouldbemadeclearthat

groupsarearrangedonaspectrumacrossthesetwodimensions.

Atfirstglance,mostofthegroupsarewhereonemightintuitthemtobe.For

example,criminalsofallsortsareperceivedasdeviants.Meanwhile,childrenare

classifiedasdependents,anddoctorsareintheadvantagedgroup.However,afew

groupsarenotablydifferentinplacementthantheyareinSchneideretal.’s(2014)

figure.Forinstance,Schneideretal.categorizesmallbusinessesandtaxpayersas

advantaged,butourresultsshowthatareclearlyperceivedasdependentsbyour

sample.Youngblackmensimilarlyareinthedependentcategoryinoursample

insteadofthedeviantcategory,where“youngminoritymales”isplacedinthe

comparablefigure.Inotherinstances,groupslikeuninsured,DREAMers,andillegal

aliensaremoreclearlyinonecategorybasedonpower,whereastheyareplacedin

themiddleofthepowerspectrumintheSchneideretal.chapter.Itisalso

3Theorientationoftheaxesmimicsthatofthematrixinthe1993APSRarticle.

9

interestingtonotethatthereareveryfewgroupsthatsocietyviewsasboth

politicallypowerfulanddeserving.

Figure1illustrateswherethesegroupsarelocatedonthetwodimensions,

butacloseranalysisofthepointestimatesrevealthatthereisagreatdealof

variationontheextenttowhichsocietyagreesonlevelsofdeservingnessand

power.Evenwithaverylargesample,therearestillsomegroupswithagreatdeal

ofcontentionastowheretheyshouldbecategorized.Wedepictthisresultin

Figures2and3.Thesizeofthebubblerepresentsthemagnitudeofthestandard

deviationsindeservingnessandpower,respectively.Essentially,oureffortto

empiricallyestimatethepositionofvariousgroupselucidatestheunderlying

contentionandvariationthatexists.

[Figure2andFigure3]

Figure4presentsthisinformationinanotherway.Here,weprovidea

representativeseriesofscatterplotsthatallowustoseewhereeachofthe1,572

MTurkworkersevaluatedgroups’levelsofdeservingness(horizontal-axis)and

power(vertical-axis).Altogether,therearethreepatternsthatarise,ofwhichwe

provideafewexamplesinFigure4.

[Figure4abouthere.]

Tobegin,therearesomegroupsthathaveagreatdealofconsensusonboth

dimensions.Groupslikechildren,sexoffenders,andbigbanksofferthreeexamples

ofgroupsthatareeasilycategorized,inthiscase,asdependents,deviants,

contenders,respectively.Inthecaseofchildren,forinstance,mostpeopleview

themashighlydeservingbutperceivethemashavinglowlevelsofpower.Thisis

10

representedbythefactthatthedotsinthescatterplotareclumpedinthelower-left

handcorner.Weseesomethingsimilarforsexoffendersandbigbanks.

Ontheotherendofthecertaintyspectrumaregroupslike

“environmentalists.”Thescatterplotrevealsagreatdealofempiricalnoisearound

theevaluationofthisgroup.Noconsensusarisesoneitherdimensionofconcern.

Otherlessextremeexamplesofthispatternarisesinhowpeopleevaluatepoliceand

alsotransgenderpeople.Generallyspeaking,wefoundthispatternformostidentity

groups(e.g.AfricanAmericans,Latinos,gaymen).

Thethirdpatternthatarisesincludesthosethathaveconsensusonone

dimensionbutlackconsensusontheother.Veterans,terrorists,“illegalaliens”are

examples.Thereissignificantagreementonthenotionthatterroristsarelowin

deservingness,butthereisnoconsensusastohowmuchpowerthistargetgroup

has.Thismakesintuitivesense.Somearelikelytobelievethatpeoplewhobecome

terroristsdosobecausetheyhavelittlepoliticalpower.Meanwhile,othersview

terroristsashavingagreatdealofpoliticalpower,asmeasuredbyinfluenceor

sophisticationoforganization(e.g.ISIL,KKK).Similarly,ourdatasuggeststhat

societyagrees“illegalaliens”haveverylittlepoliticalpower,butthereisgreat

disagreementonlevelsofdeservingness.Thisislikelytohavepoliticalimplications,

butinwhatdirection?Ifundocumentedimmigrantsareviewedasdeviants,we

shouldexpectpolicymakerstopunishthisgroupandtodosowithfewpolitical

consequences,butifsocietyviewsthemasdependents,wewouldexpectadifferent

setofpolicyoutcomes.Newadvancesinthetheorywouldprovidemoreguidance

forpredictingthefatesofambiguouslyconstructedorhighlycontestedgroups.

11

NewInsights

Thesocialconstructionoftargetpopulationstheoryhasbeencriticalinthepolicy

designscholarship.Atitscore,thetheoryisincrediblyhelpfulbecauseitprovidesa

greatdealofpredictiveinsight,especiallyasitrelatestohowandwhypolicymakers

exacerbateinequalitythroughthepolicyprocess.Ourresultsresolvethequestions

weraised,particularlyaboutthedegreeofcontentionaroundthesocial

constructionofgroups,butalsosparknewpointsofdiscussion.

Tobegin,weshowthatcrowdsourcingcanbeusedasahighlyreplicable,

accurate,andinexpensivewaytodeterminethecontemporarysocialconstructionof

targetpopulations.ThemethodssuggestedbySchneiderandIngram(1993)would

likelyleadtoagreatdealofinconsistencyinthecategorizingprocess,especiallyfor

highlycontentiousgroups.Throughcrowdsourcing,however,scholarswouldbe

abletocollectdataonjudgmentsofgroups’deservingnessandpowerinawaythat

isaccurate,reproducible,andreplicable.

SchneiderandIngram(1993)notedintheiroriginalarticlethatsomegroups

arelikelytobemorecontentiousthanothersbutimpliedthatweshouldlargely

expectconsensus.Ourresultsempiricallyassesstheclaim.Here,wecalculatedpoint

estimatesthatrepresenthowdeservingorpoliticallypowerfulsocietyviews

variousgroupsaswellasillustratetheextenttowhichconsensusarosearound

theseconstructions.Wefindthatwhilethereareseveralgroupswhosesocial

constructionsareclearlyagreedupon,thereareagreatmanythatareincredibly

contentious.

12

What’smore,whilecrowdsourcedworkersevaluatedgroupssimilarlyto

SchneiderandIngram’s(1993)hypotheticalplacementofgroups,manyareplaced

inawhollydifferentcategory.Itiswellknownthattheaggregationofanswersfrom

alargenumberoflaypeoplecanoftencomeclosertothe“truth”thantheestimates

ofafewexperts(LyonandPacuit2013).Thedifferencesmayresultfromthebiasof

experts.Additionally,ourresultsreflectthenotionthatsomegroupsbordertwo

categories.Makingpredictionsaboutoutcomesofthepolicyprocessforthese

groupswouldprovetobedifficult,asmovementfromonecategorytoanother

shouldresultinverydifferentoutputsofthepolicyprocess.Wealsofindthatthere

aregroupsthathaveconsensusononedimensionbutnottheother.

ThedifferencebetweenwhatSchneiderandIngramimplicitlyexpectedand

whatwefindspeaksvolumesbothaboutAmericans’worldviewaswellasto

questionsaboutwhogetswhat,when,how,andwhy.Theoretically,ourresults

suggestthatnotonlyshouldwethinkaboutthegroupsintermsofthetwo

dimensionsposedbythetheory,buttoalsothinkofathird:degreeofconsensus.

SchneiderandIngraminitiallydevelopedthistheorywhenthetwomajorAmerican

politicalpartieswereabletoagreeonagreatnumberofpolicymatters,butthis

thirddimensionislikelytobeparticularlyrelevantinaneraofpoliticalpolarization.

ConsideringthehighlevelsofpoliticalpolarizationintheAmericanpolitical

landscape,wemightexpecttheretofindmultiplerealitiesbywhichDemocratsand

Republicans,orliberalandconservativeslive.Perhapstherearemultiple

constructionsoftargetgroupsthatshouldbemappedout.Drasticallydifferent

worldviewsoughttoleadtodifferentpolicyoutcomesasmajoritiesinlegislatures

13

shiftfromonepartytoanother.Bygatheringdataaboutrespondents’political

identificationanddemographicprofiles,onewouldbeabletogenerateanumberof

relevantandimportantpredictions.Thisadditionaldatawouldallowustouncover

theunderlyingpoliticaldeterminantsofconsensus,orlackthereof.Scholarswould

alsobeabletodelineatepredictionsabouthowpolicymakersofvarious

demographicgroupsmightviewtheworld;thatistosay,wemaygetatightergrasp

ofhowdescriptiveandsubstantiverepresentationarerelated.Gatheringthiskind

ofdatamightalsoallowustogainsystematicinsightonhowmembersoftarget

groupsviewthemselves,andtherebymakepredictionsaboutchancesforgroup

mobilization,perhapsevendevelopintuitiononthewhethermobilizationislikely

tohappenwithin(e.g.voting,lobbying)oroutside(e.g.protest,riots)ofAmerican

politicalinstitutions.Altogether,methodologicaladvancescanbeleveragedto

advancethecontributionsofthistried-and-truetheory.

Butourfindingsalsospeaktothepropositionthatnotallpolicyenhances

democracy.Forinstance,wefoundthattherewereveryfewgroupsthatcanbe

easilycategorizedasadvantaged,therearenogroupsinthemostupper-leftcorner,

andtherearemanymoregroupsinthecontenders’areaincomparisontothe

matrixdepictedinSchneideretal.’s(2014)mostrecentchart.Because“no

legislatorswanttoopenlydogoodthingsforshadypeople”(116),ourresultsmay

speaktoapublicmoodmarkedbycynicism,orportendanincreasinglylargegap

betweenharshrhetorictowardcontendersandthehollowregulationspresentedto

them.

14

Inaneramarkedbypolarization,whatdirectionshouldweexpect

policymakerstotakeongroupswhoseconstructionisunclearorambiguous?

Undocumentedimmigrantsmakeanexcellentcaseinpoint.Inthe1980s,theparties

wereabletoagreeonamnesty.AlthoughourresultsrevealthatAmericansview

undocumentedimmigrantsaspowerless,thereislittleagreementaroundtheextent

towhichtheyaredeserving.Itispossiblethatsocialconstructionsmaynowpivot

onpartisanlines,oritmaybethecasethatlackofconsensusorambiguityaround

theconstructionofagroupsuggeststhatthereisbothinterpartyandintraparty

contentionaroundhowtobestdealwiththesegroups.Indeed,groupsthatwefound

havingagreatdealofempiricalnoise(e.g.veterans,environmentalists)maybestbe

understoodaswedgegroups,potentiallyafifthcategory.Theseissuescanbe

resolvedempirically.

Beforeweclose,SchneiderandIngramsuggestedthat“contestedsocial

constructionsareinherentlyunstableandripeforpolicychangethatsubdividesthe

populationsintomoredeservingandlessdeservingcategories(2005,10),”and

furthermore,politicalentrepreneurscancapitalizebyscapegoatingnegatively

constructedgroups,therebysustainingorfacilitatingdegenerativepolitics,orthe

“exploitationofderogatorysocialconstructions,manipulationofsymbolsorlogic,

anddeceptivecommunicationthatmasksthetruepurposeofpolicy”(11).Our

resultsshow,unexpectedly,thattherearealargenumberofpoliticallyrelevant,

salienttargetgroupsthathaveagreatdealofcontentionandcontroversyaround

theirsocialconstruction.Animportantnextstepwouldbetoempiricallydiscernthe

consequencesofaseeminglygrowingnumberofcontestedsocialconstructionson

15

policy,particularlyasthoseimplicationsspeaktowhetherpolicyisincreasingly

likelytoresultindegenerativepolitics.

16

References

Benoit,Kenneth,DrewConway,BenjaminELauderdale,MichaelLaver,andSlava

Mikhaylov.2016."Crowd-sourcedtextanalysis:reproducibleandagile

productionofpoliticaldata."AmericanPoliticalScienceReviewno.110

(2):278-295.

Berinsky,AdamJ,GregoryAHuber,andGabrielSLenz.2012."Evaluatingonline

labormarketsforexperimentalresearch:Amazon.com'sMechanicalTurk."

PoliticalAnalysisno.20(3):351-368.

Huff,Connor,andDustinTingley.2015."“Whoarethesepeople?”Evaluatingthe

demographiccharacteristicsandpoliticalpreferencesofMTurksurvey

respondents."Research&Politicsno.July-September:1-12.

Lyon,Aidan,andEricPacuit.2013."TheWisdomofcrowds:methodsofhuman

judgementaggregation."InHandbookofHumanComputation,599-614.

Springer.

Peer,Eyal,JoachimVosgerau,andAlessandroAcquisti.2014."Reputationasa

sufficientconditionfordataqualityonAmazonMechanicalTurk."Behavior

researchmethodsno.46(4):1023-1031.

Rouse,StevenV.2015."AreliabilityofMechanicalTurkdata."ComputersinHuman

Behaviorno.43:304-307.

Sabatier,PaulA.1999.TheoriesofthePolicyProcessBoulder,CO:WestviewPress.

Schneider,Anne,andHelenIngram.1993."Socialconstructionoftargetpopulations:

Implicationsforpoliticsandpolicy."AmericanPoliticalScienceReviewno.87

(2):334-347.

17

Schneider,AnneL,HelenIngram,andPeterDeLeon.2014."Democraticpolicy

design:Socialconstructionoftargetpopulations."InTheoriesofthepolicy

process,editedbyPaulASabatierandChristopherWeible,105-149.

Schneider,Anne,andMaraSidney.2009."Whatisnextforpolicydesignandsocial

constructiontheory?"PolicyStudiesJournalno.37(1):103-119.

Schroedel,JeanReith,andDanielRJordan.1998."Senatevotingandsocial

constructionoftargetpopulations:AstudyofAIDSpolicymaking,1987–

1992."JournalofHealthPolitics,PolicyandLawno.23(1):107-132.

Figure1:EstimatingthePowerandDeservingnessofSociallyConstructedGroups

Note:TheaxesarearrangedtomimicthematrixdevelopedbySchneiderandIngram(1993)

Figure2:EstimatingtheContentionofDeservingnessofSocialConstructions

Note:Thesizeofthebubblesrepresentthemagnitudeofthestandarddeviationindeservingnessforeachofthetargetgroupsbasedonn=1,572.

Figure3:EstimatingtheContentionofPowerofSocialConstructions

Note:Thesizeofthebubblesrepresentthemagnitudeofthestandarddeviationinpowerforeachofthetargetgroupsbasedonn=1,572.

Figure4:PatternsofConsensusamongIndividualRespondents

Note:SimilartoFigures1-3,deservingness(horizontalaxis)scalemovesfromhigh(left)tolow(right).Power(verticalaxis)shiftsfromhigh(top)tolow(bottom),asdelineatedinSchneiderandIngram(1993).

AppendixSurveyQuestionWordingforPowerandDeservingMeasuringPower:Somegroupsinsocietyhaverelativelymorepoliticalpowerandresourcesthanothers.Bypoliticalresourceswemeanthatsomegroupsaremoreunited,easytomobilize,wealthy,skilled,focusedontheirgoals,oraccustomedtovotingordirectlycontactingpublicofficials.Basedonwhatyouknowaboutthegroupslistedbelow,howpoliticallypowerfulwouldyousayeachofthesegroupsare,generallyspeaking.Here0meansthatmostpeopleinthatgroupareverypowerless.100meansthatmostpeopleinthatgroupareincrediblypowerful.MeasuringDeservingness:Somegroups,onaverage,areviewedaspeoplewhocontributetothegeneralwelfareofsocietyandworthy,andthusaredeservingofsympathy,pity,orhelp.Typically,wedescribemembersofthisgroupasgood,smart,hardworking,loyal,disciplined,generous,caringofothers,respectful,andcreative.Meanwhile,therearemanyothergroupsthatareviewedasaburdentothegeneralwelfareofsociety,andarebelievedtobeunderservingofsympathy,pity,orhelp.Typically,wedescribemembersofthisgroupasgreedy,disrespectful,disloyal,immoral,disgusting,dangerous,lazy,andexpectotherstocareforthem.Basedonwhatyouknowaboutthesegroups,howdeservingorunderservingwouldyousayeachofthesegroupsare,generallyspeaking.Here,0meansmostpeopleinthatgrouparecompletelyundeserving.100meansmostpeopleinthatgroupareverydeserving.

PointEstimatesandStandardDeviationsofAllGroups

GroupPowerRanking

(lowest=1)

PowerEstimate

PowerStd.

DeservingRanking

(lowest=1)

DeservingEstimate

DeservingStd.

abortionproviders 59 31.81 24.97 72 45.46 33.37africanamericans 47 35.87 23.99 47 61.51 29.24americanindians 25 22.55 21.56 42 67.26 28.92attorneys 40 67.08 23.33 50 37.77 29.33autoindustry 49 64.44 24.06 30 35.20 27.60bigbanks 17 85.22 19.60 25 23.24 27.04bigcorporations 11 86.52 18.18 31 25.01 27.63blacklivesmatter 65 34.30 25.49 73 45.94 34.49ceos 19 82.36 19.71 44 29.64 28.96children 13 12.56 19.10 28 76.60 27.52collegestudents 33 31.25 22.72 23 62.35 26.92congress 18 84.26 19.71 32 27.75 27.84criminals 15 15.44 19.22 5 17.70 23.69disabled 20 19.06 19.80 26 75.04 27.27doctors 52 61.19 24.32 48 61.32 29.30DREAMERs 43 25.81 23.50 62 51.23 31.34elderly 64 33.66 25.47 8 76.33 24.26environment 54 41.37 24.43 56 60.92 30.50ex-felons 4 11.87 15.83 34 31.22 27.97farmers 38 37.09 23.15 13 69.19 25.31feminists 57 37.93 24.75 65 50.17 31.71forprofcolleges 68 54.93 26.75 33 28.10 27.93gaymen 48 33.40 24.05 59 56.70 31.12gunmanuf 67 64.32 26.57 35 27.18 28.16hackers 71 43.69 28.73 11 20.59 24.59homeless 1 7.37 13.42 64 63.47 31.61homeowners 42 43.94 23.41 16 62.53 25.69illegalaliens 14 13.64 19.17 69 35.45 32.32insuranceco 34 74.63 22.78 19 25.02 26.55jobcreaters 58 61.17 24.79 39 62.36 28.76laborunions 61 57.16 25.10 57 48.01 30.59latino 36 31.96 22.92 46 57.94 29.17lesbians 51 31.41 24.10 63 56.54 31.37marijuanasmoker 24 24.38 21.46 58 38.91 30.84media 31 74.76 22.39 24 28.47 27.04mentallyhandicapped 7 11.32 16.89 37 73.78 28.31middleclass 21 44.14 20.88 7 67.44 24.18military 63 65.35 25.25 55 64.10 30.49millennials 41 38.40 23.39 40 52.65 28.86

mothers 62 41.77 25.13 9 74.56 24.42muslimmen 37 25.91 23.08 68 47.64 31.92muslims 32 25.35 22.67 70 49.48 32.45opiodaddict 3 10.53 15.26 61 33.42 31.30pharmaceuticalcompanies 27 80.24 21.62 21 24.82 26.79police 44 64.51 23.76 52 59.92 30.05pollutingindustries 70 61.07 27.70 3 12.05 19.76poorfamilies 6 14.14 16.67 27 69.42 27.49primarycarephysicians 56 54.27 24.61 38 61.03 28.38prisoners 2 8.65 14.51 20 25.19 26.68richpeople 12 84.30 18.75 51 32.53 29.79scientists 46 53.26 23.97 36 66.57 28.18sexoffender 5 10.79 16.61 2 9.55 18.80smallbusiness 28 39.43 22.10 17 68.02 25.69smokers 23 24.19 21.12 49 32.69 29.33soldiers 66 44.95 25.89 15 74.32 25.55students 26 28.03 21.62 18 64.33 26.55superpacs 72 72.52 29.22 10 19.67 24.53taxpayers 53 41.96 24.35 14 71.60 25.34teachers 50 40.35 24.09 12 74.45 24.65teaparty 69 45.65 26.82 43 29.70 28.94teenagers 10 16.64 18.07 41 55.05 28.90terrorists 73 30.62 29.55 1 5.84 15.27transgender 35 23.74 22.79 71 55.95 32.92unemployed 8 14.96 17.11 54 58.41 30.18uninsured 9 15.92 17.89 66 57.65 31.75vegans 39 27.40 23.19 60 47.08 31.27veterans 55 38.78 24.53 6 79.01 24.02wallstreetbrokers 30 76.55 22.24 29 25.25 27.55welfarecheats 22 17.29 20.90 4 12.27 21.59welfaremothers 16 16.83 19.55 67 58.28 31.77whitemen 60 68.36 24.97 45 53.94 29.06whitewomen 45 49.98 23.90 22 62.48 26.79youngblackmen 29 25.04 22.20 53 58.35 30.16