nber -The Effect of Medicaid Expansion on the Take-up of Disability Benefits by Race and Ethnicity-此為英文文檔

NBERWORKINGPAPERSERIES

THEEFFECTOFMEDICAIDEXPANSIONONTHETAKE-UPOFDISABILITYBENEFITSBYRACEANDETHNICITY

BeckyStaiger

MadelineS.Helfer

JessicaVanParys

WorkingPaper31557

/papers/w31557

NATIONALBUREAUOFECONOMICRESEARCH

1050MassachusettsAvenue

Cambridge,MA02138

August2023

WethankNicoleMaestasandLaraShore-Shepardforhelpfulcommentsandsuggestions.TheresearchreportedhereinwasperformedpursuanttograntRDR18000003fromtheUSSocialSecurityAdministration(SSA)fundedaspartoftheRetirementandDisabilityResearchConsortium.TheviewsexpressedhereinarethoseoftheauthorsanddonotnecessarilyreflecttheviewsoftheSSA,anyagencyoftheFederalGovernment,ortheNationalBureauof

EconomicResearch.

NBERworkingpapersarecirculatedfordiscussionandcommentpurposes.Theyhavenotbeenpeer-reviewedorbeensubjecttothereviewbytheNBERBoardofDirectorsthataccompaniesofficialNBERpublications.

?2023byBeckyStaiger,MadelineS.Helfer,andJessicaVanParys.Allrightsreserved.Shortsectionsoftext,nottoexceedtwoparagraphs,maybequotedwithoutexplicitpermissionprovidedthatfullcredit,including?notice,isgiventothesource.

TheEffectofMedicaidExpansionontheTake-upofDisabilityBenefitsbyRaceandEthnicityBeckyStaiger,MadelineS.Helfer,andJessicaVanParys

NBERWorkingPaperNo.31557

August2023

JELNo.I13,I14,J15,J22

ABSTRACT

Publicdisabilityprogramsprovidefinancialsupportto12millionworking-ageindividualsperyear,thoughnotalleligibleindividualstakeuptheseprograms.MixedevidenceexistsregardingtheimpactofMedicaideligibilityexpansiononprogramtake-up,andevenlessisknownabouttherelationshipbetweenMedicaidexpansionandracialandethnicdisparitiesintake-up.Using2009-2020CurrentPopulationSurvey(CPS)data,wecomparechangesinSupplementalSecurityIncome(SSI)andSocialSecurityDisabilityInsurance(SSDI)take-upamongpeoplewithdisabilitieslivinginMedicaidexpansionstates,comparedtopeoplewithdisabilitieslivinginnon-expansionstates,beforeandafterMedicaidexpansion.Wefurtherexploreheterogeneitybyrace/ethnicity.WefindthatMedicaidexpansionreducedSSItake-upamongWhiteandHispanicrespondentsby10%and21%,respectively,andincreasedSSDItake-upamongWhiteandBlackrespondentsby9%and11%,respectively.WefurtherfindthatMedicaidexpansionreducedtheprobabilitythatdisabledrespondentshademployer-sponsoredhealthinsurancebyapproximately8%,aneffectprimarilyobservedamongBlackandother-racerespondents,suggestingthatexpansionreducedjob-lockamongtheSSDI-eligible,contributingtotheobservedincreaseinSSDItake-up.

BeckyStaiger

JessicaVanParys

DivisionofHealthPolicy

DepartmentofEconomics

andManagement

HunterCollege

SchoolofPublicHealth

695ParkAvenue,HW1534

UniversityofCalifornia,

NewYork,NY10065

Berkeley

andNBER

2121BerkeleyWay#5424

Berkeley,CA94720

jessica.vanparys@

bstaiger@

MadelineS.Helfer

1050MassachusettsAvenue

Cambridge,MA02138

mshelfer@

Adataappendixisavailableat/data-appendix/w31557

2

1Introduction

Approximately7.5%oftheworking-agepopulationintheUnitedStateshasaself-reporteddisability(

BureauofLaborStatistics

,

2022

).TheSocialSecurityAdministration(SSA)administerstwo?nancialassistanceprogramsforpeoplewithdisabilities.SupplementalSecurityIncome(SSI)actsasasafetynetof“l(fā)astresort”tolow-incomepeoplewithdisabilitieswhohavelimitedemploymenthistory,whileSocialSecurityDisabilityInsurance(SSDI)providesbene?tstopeoplewithdisabilitieswhohavesigni?cantpriorworkhistory(atleast?veofthelasttenyears)(

SocialSecurityAdministration

,

2022b

).Thesepaymentsrepresentasigni?cantsafetynetagainstextremepovertyforthispopulation.

1

Nearly29%oftheapproximately9.9millionSSDIrecipientsand25%oftheapproximately7.6millionSSIrecipientsderive90-100%oftheirpersonalincomefromSSDIorSSIbene?tpayments(

MesselandTrenkamp

,

2022

).Andyet,thereisevidencethatnotallindividualswhoareeligibleforSSIorSSDIenroll,ultimatelyleavingmoneyonthetable(

ElderandPowers

,

2004

).

Incompletetake-upofpublicprogramsisabroaderpolicypuzzle(

Currie

,

2004

).Amongawideportfolioofexistingworkseekingtoanswerthisquestion,somestudieshavedemonstratedthattake-upinonepublicprogramcanhavespillovere?ectsontake-upinanother(

HamandShore-Sheppard

,

2005

;

Schmidtetal.

,

2020

).Forexample,priorresearchhasdemonstratedthatMedicaidexpansiona?ectsoverallSSIandSSDItake-up,thoughresultsaremixed(

BurnsandDague

,

2017

;

ChatterjiandLi

,

2016

;

Maestasetal.

,

2014

;

Schmidtetal.

,

2020

;

Sonietal.

,

2017

).Somestudieshavedocumentedsmallbutsigni?cantdecreasesinSSIenrollmentfollowingtheA?ordableCareAct’s(ACA)Medicaidexpansion(

BurnsandDague

,

2017

;

Sonietal.

,

2017

),suggestingthatforsomeSSIenrolleeswhowishtoremainintheworkforce(andareunabletodosowhilereceivingSSI),becomingeligibleforMedicaidoutsideoftheSSIpathwaymayhavebeenamoreattractiveoption.OtherstudiesshownorelationshipbetweenMedicaidexpansionandSSIapplicationsorbene?treceipt(

ChatterjiandLi

,

2016

;

Schmidt

etal.

,

2020

).Likewise,somestudiesshowthatMedicaidexpansionincreasedSSDItake-upasmoreindividualsbecameeligibleforMedicaidcoverage(

Maestasetal.

,

2014

),whileotherstudies?ndno

1SupplementalSecurityIncome(SSI)providesanaveragemonthlystipendofapproximately$623.SocialSecurityDis-abilityInsurance(SSDI)providesaveragemonthlycashbene?tsofapproximately$1,358(

SocialSecurityAdministration

,

2022a

).

3

relationship(

Schmidtetal.

,

2020

).

Furthermore,verylittleresearchexploresdi?erencesinSSIandSSDItake-upbyraceorethnicity.Thisgapintheliteratureissigni?cantgiventhehigherratesofself-reporteddisabilityamongBlackindividualsrelativetonon-HispanicWhites(

Goodmanetal.

,

2017

),andtheubiquityofracialinequitiesinotherhealthdomains,includinglifeexpectancy(

CaseandDeaton

,

2021

;

Cullenetal.

,

2012

;

Dwyer-

Lindgrenetal.

,

2022

),underlyingchronichealthconditions(

AlexanderandCurrie

,

2017

;

Boustanand

Margo

,

2014

;

Hickenetal.

,

2014

;

McGuireandMiranda

,

2008

;

Moreno?etal.

,

2007

),andhealthcareaccess(

Brownetal.

,

2016

;

Dickmanetal.

,

2022

;

Johnstonetal.

,

2021

;

Mahajanetal.

,

2021

;

Shietal.

,

2014

).Muchofthisgapisduetoalackofdata,astheSSAstoppedcollectingraceinformationin2002(

Martin

,

2016

).Despitethislimitation,researchershaveusedsurveydatatoprovideevidencethatracial/ethnicminoritiesmakeupadisproportionateshareoftheSSIpopulation,withBlackandAmer-icanIndianandAlaskaNative(AIAN)AmericanstwiceaslikelytoenrollinSSIasWhiteindividuals(

MusumeciandOrgera

,

2021

).

ExistingresearchalsosuggeststhatBlack,Hispanic,andAIANrecipientsreporthigherlevelsofpovertyandsmallerbene?tpaymentsthannon-minorityrecipientsinSSIandSSDI,onaverage(

Hendley

andBilimoria

,

1999

;

Martin

,

2007

;

MartinandMurphy

,

2014

;

Smith-Kaprosyetal.

,

2012

;

Tamborini

etal.

,

2011

).Thoughrace/ethnicityisunlikelytobeacausaldeterminantofanyofthesedisparities,itcouldbeaproxyforsharedexperiencesandconditionsthatresultinsystematicallydistinctpatternsoftake-upandbene?treceipt(

HendleyandBilimoria

,

1999

;

Martin

,

2007

;

Smith-Kaprosyetal.

,

2012

).Afteradjustingforindividualcharacteristicssuchasincome,education,andgeography,racialdisparitiesintake-upmaypersist,andcouldre?ectabroadercontextofstructuralbiasandexclusionaryactionstowardsnon-whiteindividuals.Policiesthatexpandaccesstoothersafetynetprograms—suchasMedicaidexpansion—mayexacerbateormitigateracialandethnicdisparitiesindisabilityprogramreceipt(

Creedonetal.

,

2022

).

PrevailinghypothesesregardinghowMedicaidexpansiona?ectstake-upofSSIorSSDIamongeligi-bleindividualsoftenconsiderthefollowingtrade-o?sthatindividualsface.ToqualifyforSSIorSSDI,individualsmustmeetcertaindisabilitycriteriaandhaveearningsbelowasubstantialgainfulactivity(SGA)threshold.ExpandingincomethresholdsforMedicaideligibilityhastwodi?erenthypothetical

4

implicationsforSSIandSSDIparticipation.Forindividualswithdisabilitieswhoarenototherwiseeligibleforpublichealthinsuranceprograms,SSIreceipttriggersautomaticenrollmentintoMedicaidinmanystates,andthuscanactasapathwaytohealthinsurance(

SocialSecurityAdministration

,

2022b

).Therefore,MedicaidexpansioncouldtheoreticallyreduceSSItake-upamongpeoplewhocouldqualifyforMedicaidoutsideoftheSSIprogramandwhowouldpotentiallyprefertoremainintheworkforce.

Incontrast,SSDI-eligibleindividualsfaceadi?erenttrade-o?:forgohealthinsurancethroughtheiremployer(whichisoftenlinkedtofull-timeemployment)inordertomeettheSGAthresholdandbe(potentially)uninsuredforthetwo-yearperiodbetweenqualifyingforSSDIandobtainingMedicarecoverage;orcontinueworkingtomaintainemployersponsoredhealthinsurance,anexampleof“joblock”(

Maestasetal.

,

2014

).Inthesecases,MedicaidexpansioncouldincreaseSSDItake-upasaportionoftheeligiblepopulationwillqualifyforhealthinsuranceviaMedicaidduringtheirwaitingperiodforMedicare.Withrespecttothe“joblock”channel,theevidenceonhowMedicaidexpansiona?ectsemploymentismixed.

Halletal.

(

2017

)?ndthatMedicaidexpansionincreasedworkforceparticipationamongindividualswithdisabilitiesusingdatafromtheHealthReformMonitoringSurvey(

Halletal.

,

2017

),while

Ne’emanetal.

(

2022

)useCPSdataand?ndthatMedicaidexpansionhadnoe?ectontheemploymentstatusofindividualswithdisabilities.

Ourstudyhastwogoals.First,wequantifyracial/ethnicdisparitiesinSSI/SSDItake-uprates.Sec-ond,wedeterminewhetherexpandingaccesstoMedicaidcana?ectracial/ethnicdisparitiesinSSI/SSDItake-uprates.Weseekto?llthesegapsbyusingtheCurrentPopulationSurvey(CPS)from2009-2020,whichispurportedtoaccuratelyidentify84%(asopposedtotheAmericanCommunitySurveys’66%)oftheworking-ageSSIandSSDIrecipientsinthesurvey(

Burkhauseretal.

,

2014

).

Weuseadi?erence-in-di?erencesstrategytoestimatethecausale?ectofMedicaidexpansiononoverallSSIandSSDItake-up,aswellastake-upbyrace.We?ndthatSSItake-updecreasesforWhiteandHispanicrespondentswithdisabilities,whileSSDItake-upincreasesforWhiteandBlackrespondentswithdisabilitiesfollowingMedicaidexpansion.WeexplorethejoblockchannelbyshowingthatrespondentswithdisabilitiesarelesslikelytohavehealthinsurancethroughtheiremployerfollowingMedicaidexpansion,anindirectmeasureoflabormarketparticipationamongthelikelySSDI-eligiblerespondents.

5

Therestofthepaperproceedsasfollows.Section

2

discussesthedata.Section

3

introducesourempiricalstrategy.WepresentresultsinSection

4

,andconcludeinSection

5

.

2Data

TheCurrentPopulationSurvey(CPS)istheprimarydatasourceforlaborforcestatisticsintheUS,surveying60,000to90,000householdsannuallyvia?eldandtelephoneinterviews.Thesurveysamplesnon-institutionalizedindividualswhoare15yearsorolder,anditdoesnotoversampleminoritypopulations(incontrasttotheAmericanCommunitySurvey(ACS),whichisoftenusedinanalysesofdisabilityprogramtake-up).Eachobservationthusrepresentsahouseholdmemberofthesurveyedresidenceaged15orolder.WeusetheAnnualSocialandEconomicSupplement(ASEC)samplefrom2009to2020,whichincludes2,314,863respondent-years.SincetheASECasksaboutSSI,SSDI,andMedicaidenrollmentretrospectively(referringtothepreviouscalendaryear),welaggedsurveyyearsby1,andthusourstudyperiodis2008-2019.Weadditionallyrestrictedoursampletoworking-agerespondentsbetweentheagesof18and64yearsanddroppedrespondentswithahouseholdincomeabovethe90thpercentileofthesample.

Amongotherrespondentcharacteristics,theASECreportsraceandethnicity,householdincome,employment,earnings,SSIandSSDIparticipation,Medicaidenrollment,insurancestatus,andself-reporteddisability.Whiletherearemultiplequestionsaimedatassessingwhetherarespondenthasadisability,weseektoidentifyrespondentswhomostclosely?tthede?nitionofhavingadisabilityusedbytheSSAasqualifyingforSSIorSSDI.Speci?cally,theSSAde?nesqualifyingindividualsasthoseunable“toengageinanysubstantialgainfulactivitybecauseofamedicallydeterminablephysicalormentaldisability(ies)thatiseitherexpectedtoresultindeathorhaslastedorisexpectedtolastforacontinuousperiodofatleast12months”(

SocialSecurityAdministration

,

2023

).

Priorresearchsuggeststhatpreviousstrategiesusedtoidentifypeoplewithdisabilitieswhomaypotentiallybeeligiblefordisabilityprograms(suchasSSIandSSDI)mayhavebeeninsu?cient(

Burkhauseretal.

,

2014

).Forexample,onestudydemonstratesthatonly63.3%ofSSDIandSSIrecipientswerecorrectlyidenti?edintheCPSdatawhenusingacommon“six-questionsequence”ap-proach.Speci?cally,thisapproachcharacterizesindividualsashavingadisabilityiftheyrespondin

6

thea?rmativetoanyofsixquestions(6Q)intheCPSassessinghearing,vision,cognitive,ambulatory,self-care,andmobilitydi?culties(

forDiseaseControlandPrevention

,

2019

).However,combiningthissix-questionsequenceapproachwithanotherquestionregardingtherespondent’sabilitytowork(suchthattherespondentwillbecharacterizedashavingoneormoredisabilitywhenansweringa?rmativelytoanyofthesesevenquestions)hasbeenproposedasmoreappropriate.

2

Usingthisseven-questionsequencede?nition,wecharacterize158,078(11.38%)ofallworking-ageCPSrespondentsashavingoneormoredisability.Notably,thisisverysimilartothe11.6%identi?edinthepaperthatproposesthismethodology(

Burkhauseretal.

,

2014

).Inadditionalanalyses,weshowthatourresultsarerobusttousingonlythesix-questionsequencetoidentifytherelevantsample.

Respondentsreporttheirraceandethnicityasoneofsixcategories:White,Black,Asian,AmericanIndianorAlaskaNative(AIAN),Hispanic,orOther.WhilerespondentsintheCPScanreportmul-tipleracesandethnicities,wede?netheWhite,Black,Asian,AIAN,andHispanicraceidenti?ersforrespondentswhoidentifyastherespectiveracealone,andcapturerespondentswhoreportmultipleraceidentitiesintheOtherraceidenti?er.Allreportedracesarenon-Hispanicunlessotherwisespeci?ed.SSIreceiptisidenti?edforrespondentswhoreportSSIincomeinthepreviouscalendaryear,andwholisttheprimaryreasonforeligibilityasadisability.SSDIreceiptisidenti?edforrespondentswhoreportSocialSecurityincomeinthepreviouscalendaryearandlisttheprimaryreasonforeligibilityasadisability.Medicaidcoverageisidenti?edforrespondentswhoreportbeingcoveredbyMedicaidinthepreviouscalendaryear.Employer-basedhealthinsuranceisidenti?edforallrespondentswhoreportbeingthepolicyholderforemployer-sponsoredgrouphealthinsurancewithinthepastcalendaryear.Variablesforcurrentmaritalstatus,bachelor’seducation,homeownership,statepovertyrates,andresidenceinanon-metropolitanareaarealsoidenti?edfromthesample.

Ourstudycomparesdi?erentialchangesinSSIandSSDItake-upamongCPSrespondentsinstatesthatexpandedMedicaidin2014(“expansionstates”)tostatesthatneverexpandedMedicaid(“non-expansionstates”).Inoursample,wecharacterize26statesasexpandingin2014(Arkansas,California,

Colorado,Connecticut,Delaware,WashingtonD.C.,Hawaii,Illinois,Iowa,Kentucky,Maryland,Mas-

2Thisadditionalsurveyquestionisasfollows:“Atanytimeinthepreviousyear,didanyoneinthehouseholdhaveadisabilityorhealthproblemwhichpreventedthemfromworking,evenforashorttime,orwhichlimitedtheworktheycoulddo?”

7

sachusetts,Michigan,Minnesota,Nevada,NewHampshire,NewJersey,NewMexico,NewYork,NorthDakota,Ohio,Oregon,RhodeIsland,Vermont,Washington,WestVirginia)and17statesthatneverexpandedMedicaidduringourstudyperiod(Alabama,Florida,Georgia,Indiana,Kansas,Mississippi,Missouri,Nebraska,NorthCarolina,Oklahoma,SouthCarolina,SouthDakota,Tennessee,Texas,Utah,Wisconsin,Wyoming).WechecktherobustnessofourresultsagainsttheinclusionofthesixstatesthatexpandedMedicaidafter2014.Speci?cally,fourstatesexpandedin2015(Alaska,Arizona,In-diana,Pennsylvania),andtwostatesexpandedin2016(Louisiana,Montana).

3

Weprefertoexcludethesestatesfromthemainanalysistomaximizethepotentialpost-periodofourtreatmentgroup,andtohaveoneclearly-de?nedpreandpostperiod(relativeto2014)forbothexpansionandnever-expansionstates.

4

ThereareseveralkeyCPSdataqualityconcerns.First,whiletheCPSsurveysrespondentsfromall50USstatesandWashington,D.C.,largesamplingerrorsforsmallerstatesarepossible(

Martin

,

2016

).Inaddition,smallsamplesizesmayleadustoestimateresultswithlargestandarderrors.Second,somedataerrorshavebeenreportedforindividualsolderthan65(

Martin

,

2016

);however,wefocusexclusivelyonrespondentsaged18to64yearsold.Third,respondentsmayconfuseSSIandSSDIreceiptintheirresponse,orfailtoreportbene?ts,whichisalsoaconcernintheACS(

Martin

,

2016

).Fourth,CPSunder-reportsSSDIrecipientsandsigni?cantlyunder-reportsSSIrecipients,thoughthisissueexistsintheACSaswell,andtheACSdoesnotclarifywhatmakesrespondentseligibleforSSI/SSDI(

Martin

,

2016

).Wediscussunder-reportingimplicationsbelow.Despitetheselimitations,webelievethattheCPSisthebestavailabledatasetforthepurposesofourstudy.

3EmpiricalStrategy

We?rstcompareunadjustedratesofSSIandSSDItake-upacrossnon-expansionandexpansionstates,beforeandafterMedicaidexpansionin2014.Theunadjustedcomparisonoftake-uprates

providesadescriptiveoverviewthathelpscontextualizeouradjustedresults.Wethenformalizethese

3Twostates—VirginiaandMaine—expandedMedicaidin2019.However,wedonotincludethesestatesinouranalysisgiventheirinsu?cientpost-expansionperiod.

4Notethatduetothedatalagdescribedabove,datafromtheexpansionyear(i.e.year0inouranalysis)isobtainedfromthe2015survey.

8

comparisonsinadi?erence-in-di?erencesdesign:wecomparechangesintheprobabilitythatrespondentswithdisabilitiesreportreceivingMedicaid,SSI,and/orSSDIinexpansionstatescomparedtonon-expansionstates,beforeandafterexpansion.Speci?cally,weestimatethee?ectofMedicaidexpansionontheprobabilitythatY=1forrespondentilivinginstatesinyeartusingthefollowingmodel:

P(Yist=1)=β1{Expansion}s*{Post}t+?Xit+6s+6t+"ist(1)

where{Expansion}sisanindicatorthatthestateexpandedMedicaid;{Post}tisanindicatorforthepost-period(de?nedasyearsgreaterthanorequalto2014);Xitisavectorofrespondentcharacteristics(raceindicators,maritalstatus,homeownershipstatus,bachelor’sdegree,andnon-metropolitanresidency,andstatepovertyrate)thatmaybecorrelatedwithtake-up;and6sand6tarestateandyear?xede?ects,respectively,tocontrolforanyseculartrendsintake-up."ististheerrorterm,andweclusterstandarderrorsatthestatelevel.Wedroptheyearofexpansionsinceitisa“transitional”year,whereastatemayonlyhaveexpandedMedicaidaccessforpartoftheyear.SinceYistisbinary,weestimateEquation

1

asalinearprobabilitymodel.

β1isthecoe?cientofinterest;itrepresentsanestimateofhowMedicaidexpansiona?ectstheaverageprobabilityofMedicaid/SSI/SSDItake-up,dependingonmodel(Yist).Next,weestimateheterogeneoustreatmente?ectsbyrace/ethnicity.Speci?cally,weinteract{Expansion}s*{Post}twithindicatorvariablesforallracecategories:{White}i,{Black}i,{Hispanic}i,{Asian}i,{AIAN}i,{Other}i.Then,thetripleinteractiontermon{Expansion}s*{Post}t*{Black}i,forexample,canbeinterpretedasthee?ectofMedicaidexpansionontake-upamongBlackrespondentslivinginexpansionstates,relativetoBlackrespondentslivinginnon-expansionstates(withanalogousinterpretationsforeachrace).

Tobenchmarkthee?ectofMedicaidexpansionrelativetoexistingracialdisparitiesintake-upprobabilities,wereportthecoe?cientsassociatedwitheachraceindicator(omittingWhiteasthecomparisongroup).Forexample,thecoe?cienton{Black}icanbeinterpretedastheaveragetake-uprateofBlackrespondentsrelativetoWhiterespondents,intheabsenceofMedicaidexpansion.Comparingthesecoe?cientstothecoe?cientsonthetripleinteractiontermallowsustoevaluatewhetherMedicaidexpansionclosedorexacerbatedanyexistinggapsintake-upbetweenWhiteand

9

non-Whiterespondents.

Thevalidityofourempiricalapproachinidentifyingthecausale?ectofMedicaidexpansionontake-upreliesonanassumptionthat,but-fortheexpansion,take-uptrendsinexpansionandnon-expansionstateswouldhavebeensimilar.Whilethisisnotdirectlytestable,wecantestforthepresenceofdi?erentialtrendsintake-uppriortotheexpansion.Wedothisusinganeventstudyinwhicheventtimeismeasuredasyearsfrom2014.Weestimatethesamemodelasinourmaindi?erence-in-di?erencesspeci?cation,butweinteractindicatorsforyearsrelativeto2014withanindicatorforwhetherthestateexpandedMedicaid,andwithindicatorsforrespondentrace/ethnicity.Becauseoftherelativelysmallsamplesize,weexpectthismodeltogeneratelesspreciselyestimatedcoe?cients.However,anadditionaladvantageofthisapproachisthatitallowsustoevaluatewhetherMedicaidexpansion’se?ectsonSSI/SSDItake-upchangeinmagnitudeorsigni?canceovertime.

4Results

Table

1

reportsdi?erencesinthecharacteristicsofrespondentswithdisabilitiesinoursamplebe-tweenexpansionandnon-expansionstatesinthepre-expansionperiod.Ofthe30,138respondentsreportinginnon-expansionstatespriortoexpansion,16.6%receivedSSI,26.8%receivedSSDI,and32.1%wereenrolledinMedicaid.Ofthe49,007respondentsinexpansionstatesinthepre-expansionperiod,18.3%,23.5%,37.7%receivedSSI,SSDI,andMedicaidpriortotheexpansion,respectively.Whilenon-expansionandexpansionstatesreportsimilaraverageageandsharemale,respondentsinnon-expansionstatesweremorelikelytobeBlack(7.7percentagepointdi?erence),andlesslikelytobeWhite,Asian,orHispanic(3.1,2.4,and2.6percentagepointdi?erence,respectively).Respondentslivinginnon-expansionstateswerealsomorelikelytoownahome(7.7percentagepointdi?erence)andmorelikelytobelivinginanon-metroarea(8.5percentagepointdi?erence).Therearesmallerdi?erencesinmarriagerates,bachelor’seducation,andpovertyrates.Theseimbalancesmotivateourdi?erence-in-di?erencesandeventstudyapproaches.

Figure

1

Panel(a)reportsdi?erencesintheprevalenceofdisabilitybyrace,whileFigure

1

Panel(b)reportsSSIandSSDIreceiptamongrespondentswithdisabilitiesbyrace.WeobservethehighestratesofdisabilityamongAIANrespondents(21%),followedbyBlackandOther(16.9%and16.6%,

10

respectively).Wecharacterize13%ofWhite,9%ofHispanic,and6%ofAsianrespondentsashavingadisability.

Inpanel(b),thebluebarsreportthepercentofworking-agerespondentswithdisabilitieswhoreceiveSSI.Theredbarsreportthepercentofworking-agerespondentswithdisabilitieswhoreceiveSSDI.BlackandAIANrespondentshadthehighestratesofSSIreceipt(26%and24%,respectively),whileWhiteandAsianrespondentshadthelowestratesofSSIreceipt(15%and13%,respectively).Black,White,andAIANracerespondentsreportedthehighestratesofSSDIreceipt(27%,25%,and22%respectively).TherateofSSIreceiptishigherthanSSDIreceiptforAIANandHispanicrespondents,althoughratesarerelativelycloseforBlackandAsianrespondentsaswell,re?ectingunderlyingdi?erencesinprogramenrollmentforpeoplewithdisabilitiesbyrace.

Figure

1

providesadditionalevidenceonchangesinSSIandSSDItake-upovertime.Speci?cally,Figure

1

plotssmoothedmeansofoverallunadjustedtake-upofthesetwoprogramsseparatelyinexpansionandnon-expansionstatesacrossourstudyperiod.

5

Panel(a)showsincreasingSSItake-upleadingupto2014forbothnon-expansionandexpansionstates,withgenerallylowertake-upratesinnon-expansionstatesversusexpansionstates.After2014,SSItake-upamongexpansionstatesbeginstodecline,whiletake-upinnon-expansionstatesbecomesrelativelystable.

Panel(b)showsincreasingSSDItake-upamongourstudypopulationforbothexpansionandnon-expansionstatesinmostyearspriorto2014,withgreatertake-upinnon-expan