Impact of income inequality on health from middle and high income countries in 1991 2010

Chapter1:Introduction

ProblemStatement

Int h e recentyears,t h e researcheso n healthandi t s surroundingr e l a t i o n s h i p s hasbee no n t h e rise.E x p l a n a t o r y factorsaffecth e a l t h ast h e w h o l e p o p u l a t i o n i s p o i n t o f interestofmanyauthors.Theoutcomesofstudiesareamongmostcontroversies,notonlyi n th econclusions,butalso inthe discussions andcriticismoflimitations regardingthemeth odologies,data,underlying channelsof mechanisms.

Health is defined by the World Health Organization (WHO) as "a state of complete physical, mental, and social well-being and not merely the absence of disease or infirmity." In modern societies, health concerns are paramount, and advancements in technology and healthcare have significantly improved health outcomes, particularly in life expectancy and infant mortality rates While these metrics do not fully reflect the quality of life in terms of income, numerous studies have explored the connection between life quality and income-based measurements, establishing important correlations.

Health,atindividualorpopulationlevelhasexposedsomedegreesofrelationshipst o i n e q u a l i t y accordingtoRodger( 1 9 7 9 ) , Preston(1975),andDeaton( 2 0 0 1 ) Besidest h a t , therei s r e c e n t l y increase in studiesr e g a r d i n g health andp o p u l a t i o n healthandi t s n e x u s withincome,andespecially,inequality(Gravelleetal.,2002;Torre&Myrskylọ ,2014).Eventhoughthemeasurementofinequalityisitselfhardlyintuitive(Lynchetal.,2004), many economiststriedtoquantify itthroughnumbersofmetrics.Therefore,therelationship betweenincome inequalityandthehealtharebecomingimportant.

Inotheraspect,theassociationbetweeneconomicgrowthintermsofincomed i s t r i b u t i o n andqualityoflifemetricsareongoingtopicineconomicstudies.Thequalityo f lifecanonl yberaisedif growthandstandardofliving gotogether.Among determinantso f a h i g h l y developedsociety,h e a l t h andeducationarek e y opponents.Apartfromeducationattainment,whichisaprovenfactorinteractingwithwealthd i s t r i b u t i o n , healthataggregatele velsuchaslifeexpectancyandinfantmortalityratehas

10 exposedsomedegreesofconnectionstoincomeinequality accordingtoRodger(1979),Pre ston(1975),Deaton(2001).

Alternatively,t h e r e a r e empiricalworksofresearcheso n t h e c o n n e c t i o n betweenh u m a n capitalandeconomicgrowth,intermsofincomelevel.Asresults,thereisr e c e n t l y inc reasei n studiesregardinghealthandp o p u l a t i o n healthanditsn e x u s w i t h income,andmo reextending,incomeinequality(Gravelleetal.,2002;Torre&Myrskylọ,2014).

Equallyimportant,themutualeffectofhealthandincomeinequalityisasourceofdeb ateinmanypapers.Inonehand,somepapershavebeenindicatedthatthepartofthei n c o m e in equalityhypothesis.Ontheother side,theeffectsofhealthoutcomesonincomeconceptionandv i c e v e r s a haveb e e n i n v e s t i g a t e d f o r l o n g t i m e (Leighetal.,2009).However,theconnections betweenthreeco ncepts:economicinequality,healthprogressandtheirinteractionswithincomedrivingmechan ismarenoteasilyestablishedorobservewithsolidevidences.Theconsistentresultsofresearc hesofthisinterestarestille x c e p t i o n a l l y unconvincingbecauseofconflictingconclusio ns.

ResearchObjectives

Duet o t h e r i s i n g h e a l t h concernsi n welfare,e s p e c i a l l y w h e n i t comest o c hildm o r t a l i t y reductionandprolonghumanlongevity,manystudieshasbeenacceleratingt heknowledgeandconnectionsofhealthpoliciesintermsofincomedistributioninstrumentssu chasGiniorRobinHoodIndexes.ExploringthepatternofGinicoefficientlinkingtol i f e e xpectancyandIMR,withcontrolofsomeinsightfulfactorssuchaslevelofincome,healthspen ding,etc… arethemainpurposesofthisincomeinequalityonhealthindexesstudy,andarelargelytocon tributetoliteratures.

Thel o n g e v i t y andq u a l i t y o f l i f e a r e essentialt o moderns o c i e t i e s , b u t l a c k i n g o f u n d e r s t a n d i n g o f h o w i n c o m e i n e q u a l i t y c o u l d impacthealth,lackingconvi ncedevidences,particularlycombinedwithcontroversiesinunderlyingpatternsofpathwaysine videncesi n groupso f c ou nt ri es , m a k i n g perspectivesbecomedistorted.Therefore,t h e pro posed objectivesof thisresearchareto:

2 Estimatet h e effecto f G D P perheadi n conjunctiono f i n c o m e i n e q u a l i t y onhealthw i t h considerationo f differentiatinghigh,upperm i d d l e andlowermiddlei n c o m e countr ies.

Becausethemain objectiveofthisresearchis tore- examinetheeffectsofincome i n e q u a l i t y o n healtho u t c o m e , w i t h t h e attemptt o re vealu n d e r l y i n g m e c h a n i s m s w i t h evidencefrompatternsofdevelopedandlesserde velopedcountriesofdivergentlevels.Hence,theresearchwill trytoanswersomequestions:

1) Betteri nc om e differences( l o w e r inequality)leadt o betterl if e e x p e c t a n c y andr educinginfantmortalityrateataggregationlevel?

Furthermore,incomepercapita(GDPperhead)hasbeenlongtimeconsideredtheincenti veforpositivechangeinhealthperspectives;therefore,thisstudywillalsoe x a m i n e the secondresearchquestionon:

2) Whetherhigherincomepercapitaincreaseshealthindicatorsataggregationlevelo r in o t h e r hands,doesi n c o m e ass oc ia te w i t h differencesi n h e a l t h i n differentincomelevels

Researchmethodsandexpectedoutcome

Them a i n approacht o t h i s s t u d y is t o u s e p a n e l dataof4 8 countrieso f highandmidd lei n c o m e overt h e periodo f 2 0 years( 1 9 9 1 -

2 0 1 0 ) t o drawt h e r e s u l t s w i t h t h e attentiontounobservedheterogeneitybyusingfixe dandrandomeffectsaswellassomeeconometricm e t h o d s t o overcomet h e c o n f o u n d i n g a ndo t h e r issuesi n m o d e l s A l l t h e resultsaretobeexaminedinmannertoensurethereareno biasesaffectingtheinterpretationsandconcludingstatements.Data iscollectedfromvariousmacrosources.

Outcomesofresearchisprojectedtosupplementtherecentliteraturesandexpectedt o fulf illt h e understandingso f i n c o m e i n e q u a l i t y – h e a l t h r e l a t i o n Incomei n e q u a l i t y s h o u l d b e o n e o f m o s t importantm e d i t a t i o n s f o r p o p u l a t i o n h e a l t h ; i n t h i s casel i f e e x p e c t a n c y atbirth andinfa ntmortality rate,along withincome perhead.Inthathope,a n y actionsbygovernmentst hatadjustincomeinequalityordistributionandincomelevelhave directorindirectimpacts onhealthoftheirownpeople.

ThesisStructure

Thisstudyisstructuredtofeaturetheliteratureandframeworkoftheoryinf o l l o w i n g nextsection.Subsequently,theeconometricmodelswillbepresentedwithdatadescriptionsas wellasestimationstrategies.Finally,resultofestimationanddiscussionsareto beshownon two lastsectionsalongsideconclusions.

Chapter2:Literaturereview

TheoreticalBackground

Preston(1975)was leadingininvestigatethe impact ofpatternofincometohealthacrosscountries.Thes t r i k i n g resultin hismilestonepaperrevealedt h e relationshipbetweenp e r c a p i t a n a t i o n a l i n c o m e andl i f e e x p e c t a n c y atb i r t h f o r differentp e r i o d o f t i m e s Thisrelationshipis,however,wasa tdiminishingreturntoincome.Anotherconclusionwasthatiftheincomeinequalitywastober educed,thelifeexpectancycouldb e extendedf o r specificcountry,ceterisp a r i b u s.Theref ore,t h e n e g a t i v e r e l a t i o n s h i p betweenincomeinequalityandhealthhasbeensuggeste d.

TheFigure2 1 belows h o w s t h e m i l e s t o n e o f Preston(1975)w o r k s andremark ablyillustratestherelationshipofincomepercapitainformofrealGDPperheadandlifeexpecta ncyatbirthbyusingdataoffull48countriesovertheperiodof20years.Eventhoughthenoiseiss pottedaroundthedatastructure,theeffectofdiminishingratei s wellobservedwhenrealGDP percapitaraiseslifeexpectancyspecificallyattherangeo f 65-

45,000U.S.dollarsincrease.Thisalsocouldb e keyp r e d i c t i o n f o r p o l i c y recommendati onsast h e concentratedeffecti s v i s u a l l y demonstrated.

L ife e xp e ct a n cy a tb ir th (y e a rs ) 5 5 6 0 7 58 0 6 5 7 0

Moreover,t h e l i n k betweenincomest o h e a l t h , w i t h n o accountf o r i n e q u a l i t y directeffects,usuallyreferredtoas“absoluteincomehypothesis”,isreviewedbyDeaton(20 01).Thismeansthatincomedirectlyaffectshealth,nomatterhowtherelativeincomecompare dtoothers.Onthecontrast,“relativeincomehypothesis”drawstheoutcomeofhealthfromt heincomeinequality.Moreprecisely,therelativeincomehypothesisevolvest o theincomein equality hypothesis,whichproposesthedirecteffectfrominequality tohealth.T h e sizeabl eresearcheso n i n c o m e i n e q u a l i t y hypothesishaveb e e n d o n e u s i n g crosscountrydatal evel(Childs, 2013).

Theargumentso f P r e s t o n (1975)haved r a w n t h e importantconclusiona n d l a i d fo undationsforlargenumberofresearchesovertime, extendingtohealthmeasuresandi n c o m e aswellasincomeinequalityr e l a t i o n s h i p (Beckfield,2004).Lotsofstudiesshoweds omewhatquantitativeeffectseventhought h e debateh o w i n c o m e i n e q u a l i t y affectsp o p u l a t i o n h e a l t h continues.Nevertheless,t h e s e studiesarem o s t l y w it h caveatsand thereareneverabsoluteconclusions.

Theassociationfromi n c o m e i n e q u a l i t y t o h e a l t h hasbeenmodeledi n atleastse veralpapersformanyincomelevelcountriesandmanyregions(Wilkinson&Pickett,2006). Whiletheinsightsfromthesespecific researchesareveryhelpfulinshaping thepicture ofchangesinincome inrelationtohealth(oftenmeasuredbymortalityandlifeexpectancy), theconclusionsarenotcompellingenoughtodefinitelyconfirmtherobustconnectionbe tweenincome inequalityandhealthindications.

Significant efforts have been made to understand the complex relationship between inequality and health, particularly in the context of income development Notable literature, such as works by Leigh et al (2009) and Gravelle et al (2002), builds on foundational research by Deaton (2001), Preston (1975), Wilkinson (2002), and Rodgers (1979) These studies explore various theories, including the absolute income hypothesis, relative income hypothesis, and income inequality hypothesis, to explain the dynamics of this relationship (Wagstaff & Van Doorslaer, 2000).

Furthermore,E l l i s o n (2002),i n criticalthinkingp a p e r , hasbeenskepticalt h e p o s s i b l e “statisticalartefact”betweenaveragep o p u l a t i o n healthandincomeinequality,w hichwasderivedfromt h e curvilineara s s o c i a t i o n fromt h e i n d i v i d u a l l e v e l s H e alsog aves o m e explanationsf o r t h i s curvilineareffectandp r o p o s e d t h i s wasu n d e r l y i n g mechanismforrelative incomehypothesis.

Att h e c o r e o f u n d e r l y i n g mechanismo f i n c o m e i n e q u a l i t y o n h e a l t h o u t c o m e , socialinequalityisthemostinfluencingaspecthypothesizedbymany.Initially,in comei n e q u a l i t y is relatedt o s h r i n k i n g s o c i a l cohesiono r socialcapitalandi n t u r n increasesm o r t a l i t y (Kawachietal.,1997),whilesocialinequality ispersegroundf ormeasuringt h e meandeviationo f pairso f incomesi n wholep o p u l a t i o n i n Gini,accord ingt o Sen(1997).Byexploringthepathwaysofpsychologicalandphysiologicaleffectsonhea lth,W i l k i n s o n (2002)hasbeendrawntheconnectionsbetweenthesocialcohesionand

Social networks have significant health benefits, primarily due to their psychological effects on cognition, which can influence various social classes and lead to poorer health outcomes Wilkinson (1992) reviewed empirical studies using cross-sectional data, finding a relationship between societal inequality, characterized by the Gini coefficient, and health outcomes The social epidemiological transition supports the relative income hypothesis, as noted by Herzer and Nunnenkamp (2015) and Wilkinson (2002) Additionally, societal circumstances, particularly social inequality leading to economic disparities, are key dynamics that affect health through various pathways.

Astheoriesdescribeda b o v e , t h e diagramo f w h i c h mechanismcouldb e b r i e f l y de monstratedasfollowing:

Thesocialandeconomicaldeterminantsofrelationshipbetweenhealthandi n c o m e inequalitycouldbeattributedbyseveralfeaturesasinliteraturesofmanyresearchers.Thefirstis income percapita.Theincomepercapita asproxyofGDPperheadiscentralmediatorforso cio-economicsituationsthathasbeenconstantlyconnectedw i t h t h e positiveprogressofhealth.DevelopingofrealGDPpercapitaimplies improvingl i v i n g standardwiththegreatimpactonlifeexpectancyandinfantmortalityrate throughm a n y channels(Chenetal.,2014).

Secondly,Healthspendingpercapita,whichisaidedbywealthofeachindividualandhas directimpacto n healthprogressi n society.S u c h effecti s f o u n d i n literatureswrittenbyvanD eurzenetal.(2014),in whichpattern ofsp e n d i n g o nhealthi mpli cit ly headsto hospitals in majorcitiesinLowandMiddleIncomeCountries.

Morerecently,vanBaaletal.(2013)h a s c a r e f u l l y observedt h e h e a l t h caree x p e n d i t u r e andt h e l i f e e x p e c t a n c y l e a d i n g t o conclusiontherei s p o s i t i v e l y o b v i o u s impacto f healthcares p e n d i n g o n l i f e expectancy,p a r t i c u l a r l y i n Westerncountries.However,themarginaleffe ctaswellasunderlying mechanismofcausalrelationshipiss t i l l i n d o u b t

Grosssecondaryschoolenrolmentisgrossratiotopopulationofagegroupregardlessth eagethatcorrespondstolevelofsecondaryeducationlevel.Itisthebasici n p u t f o r econ omicgrowthgivent h e h u m a n development.S e c o n d a r y s c h o o l i n g attainmentappe aredu n d o u b t e d l y i n vastn u m b e r o f literaturei n relationt o economicgrowthmodelsa ndasbasicdeterminantsasaverageschoolingi n generalishighlycorrelatedwithlifeexpectanc yandeconomicoutput(Bloometal.,2004).Feinsteinetal.

(2006)hasconsolidatedthesubstantialdirect- effectofeducationonheathinlotofpapersw i t h complexmechanismofchannelsthateducati on’simpact.Logically,educationneedst o beincludedtomodelsforanexplanatoryperformanc eofmodelspecification(Groot&vandenBrink,2006).

Detailso f conceptualf r a m e w o r k w it h i t s determinantsw i l l b e discussedi n datade scriptionswhenmodelis to beconstructedin nextchapter.

EmpiricalStudiesFindings

Incomprehensivereviewwhetheri n c o m e i n e q u a l i t y i s a m a j o r determinanto f i n c o m e i n e q u a l i t y relationt o p o p u l a t i o n health,specifiedbylife e x p e c t a n c y and infantm o r t a l i t y rate,w i t h i n c o m e levelasm e d i a t o r , relevancieso f co nc ep ts havebe enpresentedbys o m e s i g n i f i c a n t researchesandstudies.T h e firme s t a b l i s h m e n t o f l iteraturesisessentialforbuildingtheconnectionsofconcepts.

Probablythemostfundamentalstudyforthetheoryandempiricalproofregardingrelati onof inequalityandhealthhasbeenconductedbyRodger(1979).FollowingPreston(1975)’sfindings, andusing individualtoaggregatehealthapproach,he

Thisfunctioncouldbedemonstratedforabsoluteincomehypothesis.Giventhe non linearo f functionalform,t h e meano f l i f e e x p e c t a n c y neededt o b e c o r r e l a t e d w i t h i n c o m e d i s t r i b u t i o n , andthereforemodeledbyGinicoefficient.

Initially,thepreciseformofincome- lifeexpectancyfunctionhasbeendefinedbyvariouscontrolvariablesjoiningthetwoconcep ts.Ontheempiricalside,Rodger(1979)hasexhibitedtheexcellentresultinwhichlowerequalit yleadedtohighermortality.Thisstatementheldtruethroughoutmanyspecifications,andalsofor lessdevelopedcountries,w i t h somewhatreducedsignificancelevel.Secondly,theinfantm ortalityratewass i g n i f i c a n t l y affectedbyincome i ne qu al i ty inr i c h countries.

T h e o ve ra ll importanceo ft h i s researchhasencouragedmanyevidence- basedworkswithinspiredtheory.

Inf r e q u e n t l y citeda r t i c l e t o connecti n c o m e d i s t r i b u t i o n andl i f e e x p e c t a n c y atb i r t h foralotofcountries inOECDandEurope,Wilkinson (1992)foundtherel ativelys t r o n g evidencefor relationbetweenincomechangesover timewith life expectancyusingcrosssectiondata.Theresultdrawnfrompaperwithimportantindicationtha tthegrossnationalproductperh e a d wasn o longera f a c t o r i n whichl i f e e x p e c t a n c y r e l a t e t o , atleastfordevelopedcountries.Ontheotherhand,relativeincomeratherthanabsoluteon ei s maincauseexplaining themechanismof incomedistribution on lifeexpectancy.

Nevertheless,withevidencefromacrosscountryandwithintheUnitedStatesofA merican,MellorandMilyo(2001)hascriticizedW i l k i n s o n (1992)anddemonstratedm o r e controversialresult, whenincome inequalitysometime raisedhealth outcome,but s o m e t i m e inoppositeway.Thispaperishighlightedwiththelongertimeofanalysis,theaccou ntofeducationf a c t o r o r t a k i n g differencet o investigatet h e m i d d l e mechanism.Moreo ver,thecasualrelationshipbetweenindividualhealthandinequalitywasnots t r o n g l y robu st.Finally,theincomeinequalityhypothesisassociatedwithsomepreviousworkswasskept ical.

(2002),inattempttoreplicatethesignificantworkofRodger( 1 9 7 9 ) , hast e s t e d relativei n c o m e h y p o t h e s i s andrelationsbetweenpercapitaincome,p o p u l a t i o n h e a l t h , andi n c o m e i n e q u a l i t y u s i n g n e w d a t a , newapproachesandmethodologies.Intheend,theyfoundn osignificantconnectionformentionedconceptsf o r eitherdevelopedordevelopingcoun triesascitedinDeaton(2001)’spaper,whichhesystematicallyr e v i e w e d manystudiesonli nksbetweeninequality,developmentand health.Questionshavebeenraisedforaggregatedataandmethodologiestroubles Asaresult ,relativeincomehypothesis hasbeenchallenged.

2 1 industrializednationsfrom1975to2006,TorreandMyrskylọ(2014)hasstronglyconfirm edtheeffectsofincomeequalityonhealthoutcome,especiallywhendifferencedbyage.Thati mpliedthemoreequalincomeyieldedthelowermortalityatyoungageandchildren.T h e y alsof o u n d t h a t t h e genderhadimportantimpactt o t h e r e s u l t Finally,giventheliteraturetheyr eviewed,theycastedthedoubt onworkofGravelleetal.(2002).

Inaddition,s o m e s c h o l a r s hasbeent a k i n g s u b s t a n t i a l effortst o reviewm a n y studiesandresearchesc o n c e r n i n g t h e n e x u s o f i n c o m e i n e q u a l i t y andhe ath,suchasLynchetal.(2004),Leighetal.

(2009),Wilkinson&Pickett(2006)andDeaton(2001).T h e y allfoundthequitecontradict edconclusionsthroughoutthelargenumbersofstudiesreviewed.However,theincomeinequal ityisindeedhavingthepowertoinfluenthealth,atvariousdegrees.Additionally,asr e c e n t d atai s increasinglyb e t t e r overyears;w i t h sophisticatedanalysismethods, thechannelsof mechanismsofhealthandincomei n e q u a l i t y hasbeenexplicitlyrevealed.Domination ofevidencesseems to befavorforthes l i g h t l y negativeormixedcorrelationandbetweeninequalityandhealthatlar gerangeofsignificance.

However,a g a i n s t t h e hypothesiso f i n c o m e i n e q u a l i t y i s d a n g e r o u s f o r health;Beckfield( 2 0 0 4 ) h a s r e c e n t l y attemptedt o replicatet h e workst r y i n g t o overc omet h e l i m i t a t i o n s o f cross- sectionalstudiesbyt e s t i n g o n o v e r 1 0 0 countriesw i t h 6 9 2 observations.Thefindingsre peatedlydeniedsuchrelationshipwithfixedeffectsmodelst o captureunobservedheterog eneity,e v e n w i t h prosperouscountriesa n d 2 measureo f i n c o m e inequality(Giniands hareofincome).Controllingforvariablesalsorenderedthen u l l hypothesis.

Ingeneral,themeasureofincomeinequalityindicatorsisalsothedebateofmanyresear ches.Inordertocounterpartthisarguments,KawachiandKennedy(1997)hasusedcrosssectionst udy,testedarangeofsixincomeinequalityindicatorsincludingGini,on5 0 statesoftheU.S.toreconfirmthehighassociationbetweenthemortalityandincomeinequality,whichhas beendemonstratedin manyecologicalresearches.

InspiredbyMellorandMilyo(2001),Childs (2013)hascheckedthevalidity ofi n c o m e i n e q u a l i t y hypothesis,w i t h paneldataf o r U S t o overcomet h e weakm o d e l specifications.Althought h e datad i d n o t supportt h e hypothesisw i t h incorrectsignsdet ected,conclusiono f p o s s i b l e associationbetweeni n c o m e i n e q u a l i t y andhealthwasdra wnduetoreasonsofcomplicatedchannelsofimpact.

Itshouldalsobenotedthattheincomeinequalityhypothesisisnotalwaysholdingforp o p u l a t i o n level.WagstaffandVanDoorslaer( 2 0 0 0 ) hasb e e n d o u b t f u l aboutt h e r o b u s t o f suchaggregatelevel,andclaimedthatonlyabsoluteincomehypothesis could beheldtrueontheU.Spopulationdata,notrelative-incomehypothesisorincome- inequalityhypothesis.

(2004).Developinganddeveloped nationsareoftheinterestinresearchinorder toinvestig atet h e differenceofeffectsintwogroupsofcountriescausedbywealth.TorreandMyrskylọ(20 14)andM a c i n k o etal.

(2004)haved o n e t h e examinationso n m a n y developedandindustrialcountriesandrealizedt he negativeeffectsofincomeinequality onlife e x p e c t a n c y andi n f a n t m o r t a l i t y rate.Int h e c a s e o f Macinkoetal.

Inl e s s relevant,t h e r e i s alsostudywhich f o c u s e d o n anotherformo f m o r t a l i t y s uchaswomen’sexperienceofchildmortality(vanDeurzenetal.,2014).Findingswereconsiste ntforotherstudieswherehigherinequalityisassociatedwithhigherchildm o r t a l i t y exp eriencesinindividual’sdatafrom52lowandmiddleincomecountries.Thiss t u d y alsoimplie dthewealthofpoorshouldbeimprovedtobridgethegapbetweenrichandpoor.

Mostrecently,HerzerandNunnenkamp(2015)chosetousepanelco- integrationt o enhancetherobustnessofresearchwheninterestingeffectofincomeinequalit yonlifeexpectancy.Thedifferenceoffindingbetweentwogroupsofdevelopedanddevel opingcountrieswassignificant.Theincomeinequalitysurprisinglyraisedthelifeexpectancyin richcountriesandtheoppositeresultinpoorones.Theresultsexposedtostablethroughm a n y sensitivechecksbut with littlemagnitudes.

Thedefiniteconclusionwhetherh e a l t h c a n beaffectedbyi n c o m e i n e q u a l i t y c o n s e q u e n t l y remainedopen Inaddition,the notablerobust resultsweredominat edfordevelopedcountries.Thecasualmechanisminadversewaywasalsoreviewedwithsomep o s s i b l e grounds.Finally,lackingofqualitydataandunknowninstrumentsbehindhealthan dincomewerereasonsforincompleteevidences.

Chapter3:Dataand ModelSpecifications

EmpiricalModel

Themodelofincomeinequality– healthrelationshipshavebeeneconometricallysetupbasedonworksofPreston(1975),Rod gers(1979),andMellorandMilyo(2001).T h e macro- modelcouldbespecifiedsimplyastheeffectofincomedistributionmeasureandfunctionofse tof incomeandcontrollingfactors:(Rodgers,1979)

Them a j o r effectsbetweent h e t w o h e a l t h measurementsandkeyi n d i c a t o r o f equality-Ginic o e f f i c i e n t - a r e modeledi n regressionconnectionsi n regardso f affectingfactorsi n accordancewithliterat uresandframeworkofvariablesnexus.This studyconsiderssix(6)specifications:

(2)includestherealGDPpercapitaasinthefundamentalconceptsbyRogers(1979),andisco nsistentwiththeapproachesbyTorreandMyrskylọ(2014);HerzerandNunnenkamp(2015) ,whichisthemostcommonfactorpairedwithGiniforsocio- economicenvironmentalmeasure.Thesubsequentingredientsarereflectedhealthcarespen dingandeducationfollowingtheresearchesofMacinkoetal.

(2004),andspeciallyMellorandMilyo(2001),whofoundcontroversialconclusions.T h o s e arehealthexpenditurepercapitaandsecondaryschoolingenrolmentratioswhichar eincorporatedinto 2 basicmodelsforlife expectancyandinfant mortalityrate.

Themorerefinedof2modelsextendedbyusinginteractiontermstodistinguisht h e e ffecto f G i n i andGDPpercapitau n d e r c o n d i t i o n s o f effectbetweenhigherandlowermi ddleincomecountries.Thesemodels arepresentedin (3),(4),(5),and(6).

(1) LifeExp it =α+1Giniit+2GDPpcit+3Healthpcit+4SchoolEnrollit+ eit(IVs:Trad eOpen,InvestRatio)

(2) IMRit=α+1Giniit+2GDPpcit+3Healthpcit+4SchoolEnrollit+eit

(3) LifeExpit= α+1Giniit+2GDPpcit+3Healthpcit+4SchoolEnrollit+5(H iixGiniit)+eit

(4) LifeExpit=α+1Giniit+2GDPpcit+3Healthpcit+4SchoolEnrollit+5(Hiix GDPpcit)+eit

(5) IMR it = α+1Giniit+2GDPpcit+3Healthpcit+4SchoolEnrollit+5(HiixGini it)

(6) IMRit= α +1Giniit+2GDPpcit+3Healthpcit+4SchoolEnrollit+5(HiixG DPpcit)+eit

IMRisInfantmortalityrate(deathsofinfantsunder oneyearoldper1000live births)

Giniisratioindexexpressedaspercentagefrom0,representsperfectequality,and100,perfecti nequality.Details explained in datadescription.

GDPpci s perc a p i t a G D P i n PPPadjustedatconstant2 0 1 1 internationald o l l a r s Forpar ticularcaseofArgentina,whendatamissingwasencountered,constantU.S.dollaratyear2010 is choiceofreplacement(World Bank,2015).

SchoolEnrolli s grosss e c o n d a r y s c h o o l i n g e n r o l m e n t ratiot o p o p u l a t i o n o f ag egroup(whatsoevertheage) forboth sexes(percentagepoint%)

InvestRatioisgrosscapitalformation(grossdomesticinvestmentasratiotoGDP)whichconsist soffixedassetandnetchangesin levelofinventories(percentagepoint%)

Hiisdummy variable.Valueof1iscountrieshighoruppermiddle income.0islowerm i d d l e

DatasourcesandDescription

Table3.1summarizesthelistofvariablesusedinstudyaswellastheirexpectedsigns, unit,denotation,sourcesofdataretrieved.Consequently,detailsofeachvariable,h o w they collectedandexplanationsonwhichpathwaystheyaffectdependentvariableso r includedin modelsarepresentedfollowing.

DependentVariables,R egressors andInstrumentalVariab les(IV)

Thedatasetinstudy iscollectedfromWorld Bank,C h i l d Mortality Estimates,andW o r l d IncomeInequalityDatabasefromUnitedNationsUniversity–

Becauseofthelackofinconsistency andmissingindatabase availability,this studya pproachest h e targetoff i r s t l y achievingt h e G i n i ando t h e r keyv a r i a b l e s , i e , l i f e e x p e c t a n c y andinfantmortality rate.Thatmeanscountriesselectedinstudieswo uldbechosenandclassifiedbasedontheavailabilitya n d accessibilityofthoseindicators.E q u a l l y importantly,theperiodofresearchiscondensedfrom1991to2010(20years),inwhicht h e dataobservedi s atdensestandm o s t surveyed;f o r t h e f u l l p o t e n t i a l employment.Accor dingtothefullsetofGini,lifeexpectancy,andinfantmortalityrateselectedfromconcerned databasesin1991-

2010,totalof48countriesarefinalizedasseto f targetednationsforresearch.Forthereasonof thiscaveat,thepatternofcountriesiss l i g h t l y skewedw i t h t h e heavierfocuso n regio no f Europe,SouthernAmerica,andCentralA s i a (9 1 6 %),a s i n detailedi n TableA 1 6 i n A p p e n d i x T h e re st o f countriese v e n l y l i e o n o t h e r regions.T h i s approachi s t o b e m a k i n g s u r e importantvariablesinformationarecollectedatitsbestintermofquantities,b utin theexpenseofrepresentativecomprehensionforallareasintheworld;whichmakesthefi ndingofeffectf o r separatearenearlyimpossible dueto thebiasness.

Ginihasbeenalmostcertainlythemostcitedandreferredamongmanyindicatorsofi n c o m e distribution measureso r wealthinequality.G i n i is b as ic al ly measureofratioofare abetweent h e Lorenzcurveandt h e hypotheticall i n e o f perfecte q u a l i t y (4 5degreeline),an dareaundert h e l i n e o f perfectequality.G i n i i n d e x i s u s u a l l y expressedaspercenta gefrom 0,representsabsoluteequality,and 100,absoluteinequality(Gini, 1921).

C u m u la ti ve sh ar eo fi n co m ee ar n ed 1 00 %

Bys i m p l e calculation,G i n i canb e e x a c t l y estimatedw i t h Lorenzc u r v e L(x)

Ginii n d e x e s arem a i n t a i n e d andconstructedbyW o r l d IncomeInequalityDatab asefromUnitedNationsUniversity–

WIDERWIID,2015),TheversionofGinidatabase,WIID3.3usedin s t u d y wasrelease di n September2 0 1 5 , haslargen u m b e r o f updatesaswellasimprovementsinobservationsand documentations.

BecauseGiniconstructioninWIIDisacompositecalculatedonefromsourcesofsurve ysandworkso f i nc om e inequality,i t c o n t a i n s s o m e assumptionsofconceptsandmethodol ogies.

AselectionofcriteriahasbeenscreenedtoensuretheGinidatachosenhaving theaccuratereflectionofnatureofinequalityscaleintroducedintomodels.WIID3.3descri edconsistentlytheconceptualgroundtobuildupthedatabase.Thisresearch followst h e s e descriptionst o f i l t e r t h e m o s t conceptswhicharem o s t appropriateandfu lfillthevariationofincomeequality.Theapproachesusedcoveredthe basics ofwelfareconcepts.

Conceptbasedontheincomeorconsumptionanddefinitionsoftheincomeitselfbecom est h e m o s t i m p o r t a n t principlest o t a c k l e t o s u c c e s s f u l l y u t i l i z e t h e G i n i i n d e x Thereisstilldisputesbetweenthetwomajorconceptsused(UNU-

WIDER,2015).Traditionally,income- basedconceptswillbeusedtocollectalldataonGiniandthisi s favoredapproachinresearc h.Otherconceptualmeasuresofcriteriathathaverulingrolesi n determiningthevalueofGiniin dextobeusedthemodelsaredetailedasfollowingsinorderofpriority:

Thequalityofsurvey- baseddataWIIDchoosestoapply.QualityofGinidataisclassifiedi n t o f o u r ( 4 ) levels:Hig h,Average,Low,andN o t k n o w n Thisqualificationreflectswhetherthe surveycovers income/consumptionconcepts,theconceptsunderlyingobservationsknowledgeable,andsurve yqualityitself(UNU-WIDER,2015);finalqualityratingis of mixofthese criteria.

Thehigherq u a l i t y rating,t h e m o r e f i t andreliableG i n i datai s T h e p r i o r i t y ofc hoosingi s alwaysH i g h q u a l i t y first,t h e n s o goingd o w n o n w h e n nos u r v e y b e t t e r available.

Iftherearetwoormoresurveysofsameyearobservedwithsamequalityrating,t h e n someotherconceptsdefinedinWIIDmanualwillbeusedasdeterminantsofwhichG i n i v a l u e to beused.Theorderasbelow:

Thewelfareconcepts;representt h e definitionso f income/ expenditurewhichiscoreofourevaluations.Therearefour(4)sub- categoriesincludedIncomegross/Incomedisposable/Consumption/Other.Becauseincome- basedinequality estimatesisourapproacho f choice,whichG i n i datap o i n t s withIncom e-basedw e l f a r e s i s prioritized.Otherwise,consumption-basedonescan beselectedinstead.

Ifwelfareconceptss t i l l ateo f t h e same,U n i t ofAnalysisconceptsa r e usedt o furtherf ilter.Theyc a n eitherHouseholdorPerson,withthelattermeansneedsofdifferents i z e d house holdinfluentsandb e takeni n t o account,andv i c e versa(UNU-

WIDER,2015).Obviously,t h e U n i t o f A n a l y s i s specifiedasPersonischosenoverother s.Ifthis first criterion ismissing,Householdis toreplace.

Otherthantheseabove,underconditionofnouniqueGinivalueisstillfound,seto f o t h e r conceptsh a v e tob e checkedt o finalizet h e G i n i Setofc o n c e p t s aremutuals u p p o r t i n g n o t i n orderconsistso f P o p u l a t i o n coverage,A g e coverage,Areac o v e r a g e , Revis ionnote,andSourcesofSurvey(whichreflectsreliabilityofsurvey).Thesecriteriabaseonlogi cofbettercoveragein terms ofrelatedconcepts.

TheinfantmortalityrateisextractedfromChildMortalityEstimates(CME),adatabasepro vidingthechildmortalityestimatesastheresultofresearchoftheUNGroupf o r Child

Datasetusedisbasedonmortalityindexesforpopulation levelworldwideoverthe l o n g periods.Infantmortali ty rateis among them o s t importanthealthindicatorswhichuse di n measuremento f levelo f developmenta s wellaskeyindicatorf o r healthc a r e infrastructu re andadvanceofspecific nation.

Lifee x p e c t a n c y i s o t h e r k e y dependentvariablewhichmeditatedo n e o f t h e most studiedaspectsinhealthscience.Lifeexpectancyatbirth(years)is“averageyearsofonenewborn i s expectedt o l i v e i f he/shew e r e t o passthroughl i f e s u b j e c t t o a g e - s p e c i f i c m o r t a l i t y rateofagivenperiod”,accordingtoUN(ChartingtheProgressofPop ulations,2000).WorldBankprovidesthelifeexpectancyatbirthforneededcountrieswith someprojectionsof its own.

Theincome percapitaisvery importantindicator forl iv in g standard withthe grea timpactonlifeexpectancyandinfantmortality ratethroughmanychannels(Chenetal.,201 4).Furthermore,t h e i n c o m e d i s t r i b u t i o n andi n c o m e percapitaa r e alsot h e t w o factorsthatcanrendertheincomeinequalitymoreorlesssevere.Inthecausaldirection,

AghionandMu r t i n (2010)hasbeeni n d i c a t e d i n endogenousg r o w t h theory,l i f e e x p e c t a n c y in turnhadimpact on GDP percapandgrowthusingIVinstruments.

PercapitaIncomeisunderlyingconceptinmanyotherdeterminantsofgrowth- relatedissuesi n economices ti ma t es Consequently,t h e i n c o m e pe rc ap it a is e a s i l y ex posedt o multi- connectionswithothervariablesandcorrelateswithunobservedheterogeneity,thisproblemsis highlightedas endogeneityandwill bemediatedpossiblybyIV.

TheGDPper capita,P P P (PurchasingPowerParities)fromin c o n s t a n t 2011 internati onaldollarsisunitofmychoicetorepresentthepowerofeconomicaloutputperheadeachnationc ouldproduce.T h e s e figuresbasedo n convertedinternationaldollarsw i t h 2011baseyearus ingPPP.Baseyear2011alsocorrespondstoperiod1991-

2010ofdataset.InternationaldollarshasthesamepurchasingpoweroverGDPastheU.S.doll arhasintheUnitedStates,withaim“tomeasurehowmuchlocalcurrencyisneededtobuyasm u c h asdoest h e currencyi n U.S.dollar”.Afteradjustingf o r i n f l a t i o n , t h e y arei n constantint ernationallycomparabledollars(Deaton&Heston,2008,p.1).

Thehealthexpenseisaidedbywealthofeachindividualandhasdirectconsequenceso n hea lthprogressi n p a r t i c u l a r s o c i e t y Sucheffectsa r e observedi n q u i t e concernedlitera tureswrittenbyvanDeurzenetal.

(2014),inwhichpatternofexpenditureonhealthi m p l i c i t l y headstohospitalsinmajorci tiesinLowandMiddleIncomeCountries.Beingskepticalregardingthec a u s a l i t y ofec on o m ic in eq u al i t y andhealth,Leighetal.

(2009)reviewedandusedempiricaldatatoconfirmtheongoingdebateonwaysofeffectswithrel ationtohealthpublicexpenditure.Otherpronouncedreview,byWilkinsonandPickett(2006),e x p l a i n e d t h e difficultieso n estimatingi n e q u a l i t y effecto n health,citingl i m i t a t i o n s o f o t h e r papersw i t h d o u b t f u l controlvariablesi n c l u d i n g h e a l t h e x p e n d i t u r e a mongmanyothers.

(2013)hasc a r e f u l l y observedt h e healthcaree x p e n d i t u r e andt h e l i f e e x p e c t a n c y l e a d i n g t o conclusiontherei s p o s i t i v e l y o b v i o u s impacto f healthcares p e n d i n g o n l i f e expectancy,p a r t i c u l a r l y i n Westerncountries.However,themarginale ffectaswellasunderlying mechanismofcausalrelationshipiss t i l l i n d o u b t

Asl i m i t ofin cl ud in g variablet h a t i s i n middle ofin co me andhealthoutcome,i t issu ggestedthatthecorrelatesbetweentheexpendituresonhealthandeducationarepronet o b e quitehighasin (Groot&vandenBrink,2006).

Thisindicatori s measuredi n convertedinternationaldollaru s i n g 2 0 1 1 P P P ratet o match othervariablesunit.ThesourceisderivedfromWorldHealthOrganization(WHO).

Averageschoolingingeneralishighlycorrelatedwithlifeexpectancyandeconomico u t p u t (Bloometal.,2 0 0 4 ) Besidest h a t , t h e r e aresignificantevidenceso f impacto f schoo lingo r educationi n principleo n i n c o m e percapita.Feinsteinetal.

(2006)hasconsolidatedt h e s u b s t a n t i a l direct- effecto f educationo n heathi n l o t o f papers.Additionally,theyha ve observedt h e evidences o f complicatedm e c h a n i s m o f channelst h a t education’seffectonsurroundingenvironme nts of oneconnects.

Beyondabove,Kemptner(2011)hasprovidedt h e associationofyearsofschoolingo n spe cificillness.Logically,educationneeds tobeincludedto models foranexplanatoryperformanceofmodelspecification(Groot&vandenBrink,2006).

Secondaryschoolattainmenti s grosse n r o l l m e n t s ratiot o p o p u l a t i o n o f agegroup (whatsoevertheage)thatcorrespondstolevelofsecondaryeducation.Itisthebasicinputf o r eco nomicgrowthbecausei t layst h e fundamentalsf o r lifetimel e a r n i n g andh u m a n developme nt.Secondary schoolingattainmentappearednumerous times invastnumber o f literatur ein relation toeconomicgrowthmodelsandasbasicdeterminants.

U n i t e d NationsEducational,Scientific,a n d CulturalOrganization(UNESCO)InstituteforS tatistics.Dataentryisweightedaverageandmeasuredaspercentagepoint( W o r l d Bank,2 015).

ThesummationofimportsandexportsshareofGDPhaslongseentheindicationforfreed omofmarketeconomies.Thisshareistypicaloneofmeasurementoftradeopenness.Tradein tegrationhasb e e n l o n g t i m e i n s p i r i n g numerouss t u d i e s r e g a r d i n g relationw itheconomicgrowthandmorerecently,healthofsociety.Consequently, includingt h e trades h a r e i n m o d e l s i s s o m e w h a t relevantandt h e r e a r e conside rableliteraturess u p p o r t i n g asi n (Herzer,2 0 1 4 ; H e r z e r , 2015)’sresearches.However,aut horl i k e Stevensetal.

In 2013, it was demonstrated that trade openness is a significant determinant affecting GDP per capita, life expectancy, and infant mortality rates However, while trade openness is correlated with infant mortality rates and life expectancy, inconsistencies arise when considering its interactions with domestic credit growth, public infrastructure, and educational achievement Additionally, concerns have been raised about trade openness contributing to insecurity, income inequality, and pollution.

Tradeopenness is thesum of exportsandimportsofgoodsandservicesmeasuredasashareofgross domesticproductfromWorld

BankviaWorldBanknationalaccountsdata,andOECDNationalAccountsdatafiles.Dataentryi sweightedaverageandmeasuredaspercentagepoint (WorldBank,2015).

The gross domestic investment ratio to GDP is a critical indicator of economic performance, highlighting the importance of fixed asset investment in economic growth Despite extensive research on the impact of capital investment on income levels and well-being, the relationship between investment and health remains underexplored As a result, gross domestic fixed investment is a suitable candidate for use as an instrumental variable in analysis, which will be further elaborated in the analysis strategy section.

Thegrossdomesticfixedinvestment(ratiotoGDP;Grossfixedcapitalformation)ist h e s u m o f l a n d i m p r o v e m e n t s , p l a n t , machinery,equipmentpurchases,andinfrastructures i n c l u d i n g neta c q u i s i t i o n s o f v a l u a b l e s D a t a i s f r o m W o r l d Bankv i a W o r l d Banknationalaccountsdata,andOECDNationalAccountsdatafiles.Dataentryi s weightedaverageandmeasuredaspercentagepoint (WorldBank,2015).

Inordertosuccessfully implementInstrumentalVariable(IV)estimationtodealw i t h endogeneity,a p p r o a c h i n g explanationo f InstrumentalVariableo n effecto f endogen ousfactorsispresentedasbelow:

The relationship between income inequality and health outcomes is complex, influenced by various confounding factors related to income levels and their fluctuations According to Pritchett and Summers (1996), individual prosperity can significantly affect physical health conditions Wealth not only impacts personal well-being but also influences access to healthcare, which in turn affects the income distribution within society and overall health progress Additionally, there are factors that influence health outcomes, such as life expectancy and infant mortality rates, that are not fully captured in existing models but still exert a significant impact on these dependent variables.

IfwearetomakeproposedIVvalid,thecrucialassumptions(orconditions)needst o p r o p e r l y e x p l a i n e d a n d satisfiedi n relationt o n e x u s o f IVitselfa n d endogenousvari ables(relevance)inadditionwithitsexogeneity.IVistobeusedinempiricalpaper“Wealthieri sHealthier”byPritchettandSummers(1996),inwhichsetofIVvariablesi n c l u d i n g

T e r m s ofTradeshocks,InvestmentratiotoGDP,BlackMarketPremium,PriceLevelDistor tion,andSimilarCountriesGrowthareutilizedtofeasiblyaddresstheendogeneity.T h e resultw asd e f i n i t e l y u n q u e s t i o n a b l e t h a t i n c o m e i s l i k e l y t o i n c r e a s e healthstatus,but theeffectoninfantmortalityrateislargerthanfor lifeexpectancy.

This research addresses issues of omitted variables and endogeneity using an instrumental variable (IV) approach It employs domestic investment ratios and trade openness as IVs in models to resolve endogeneity concerns These variables also influence life expectancy and infant mortality rates The study flexibly utilizes two IVs in separate models or as variables, contingent on tests and related checks The effectiveness of these IVs is verified in advance through F-tests in the first two-stage least squares (2SLS) estimations and supported by existing literature.

Gross domestic investment, or gross capital formation, plays a crucial role in enhancing economic well-being and prosperity by increasing gross domestic product and creating jobs Investment efforts are likely to lead to improved economic performance and potentially higher per capita income Research by Barro (2003) highlights the significant influence of trade openness and domestic investment on per capita income, while also noting that their effects are moderated by various other factors Additionally, the relationship between fixed investment ratios and growth has been explored, revealing a two-way correlation that favors investment (Blomstrom & Lipsey, 1993) Domestic investment, including fixed assets and net changes in inventory levels, significantly contributes to individual wealth and overall economic growth Trade openness, measured as the ratio of imports and exports to GDP, also shows a positive correlation with economic growth, as confirmed by Harrison (1996), who found that the strength of this effect varies based on data specifications and the types of countries studied.

Inordertogetevidenceofrelevanceassumption,correlationestimatesandregressions willbe usedto determinethedegreeofrelevanceoftwo(2)IVcandidates:d o m e s t i c inv estmentratioandtradeopenness.F- testfollowsfromfirststageregressionin2SLSm a k e t h e relevancec o n c l u s i o n viable.Additi onally,t h e c o r r e l a t i o n coefficienti s enhancingforinclusionorremovalanIVcandidateinter mofrelevanceassumption.

Besidesthat,beingtheinstrumentsmediatingforendogenousissue,theIVneedsn o t tobeinerrorterms;inotherwords,theydonothaveanyunobservedcorrelationwithtargeteddep endentvariables.Itse x o g e n e i t y w i l l b e s a t i s f a c t o r y andf i n a l l y rendert h e m o d e l validity.Modelnow canbetreatedwithIVtoeliminatebias.Exogeneitydeterminationi s se nsitivet a s k andrequiresm u c h o f argument.S e c o n d l y , i t i s m o r e o f perceptionineconom icsenseinrelationofIVinwholemodels,thantestsorprocedures,asIVonlyindirectlyaffectsde pendentvariable.

Ino r d e r t o convincet h a t IVcandidatesa r e ap pr op ri at e, t h i s s e c t i o n s h o w s h o w t h e s e IVcandidates;t r a d e openness,investmentratio,i s goodandbestfit i n thiss t u d y c o n t e x t

Recent studies indicate a lack of convincing empirical evidence linking health outcomes to trade openness and investment ratios Research has explored the quantitative effects of these factors on health, but findings have not effectively convinced audiences of their connections For example, Guzman (2008) attempted to explain well-being through life expectancy and infant mortality rates influenced by trade openness and internal conditions using OLS and panel data; however, results remained uncertain across various specifications and strategies Additionally, Herzer (2015) and Alam et al employed cointegration time series analysis, including Granger causality and unit root tests, to further investigate these relationships.

(2015),hasbeenappliedt o f i n d effectso n populationhealthcausedbytradeopenness.Theyf oundquiteimpactoftradeonlifeexpectancybutjustintwoparticularcountries,U.S.

(in40years)andPakistan(in30years)incaseofAlametal.

Intermofinfantmortalityrate,therehasbeentheinvestigationonfreetradebyAfrica nGrowthandO p p o r t u n i t y Act( A G O A ) o n infantm o r t a l i t y r a t e i n Sub-

SaharanAfrica(Panda,2014).Thefindingisfavoredby7reductionsininfantmortalityrate(p er1000).Thelimitationsinthisstudyaretheunveiledmechanismchannelsandunbalancedeco nomicsectorsinwhichinfantdeathsdeclinesaffects.Arearestrictingcoverisotherobstac lewhichmightbeanissueforglobalrepresentatives.

Significantly,Herzer(2014)haslaidthemillstoneonfewempiricalevidencesofrelatio nshipbetweenh e a t h andt r a d e o p e n n e s s whenh e usedp a n e l cointegrationtechniquef o r

Over the past 45 years, research has confirmed the long-term positive effects of trade openness, measured as a percentage ratio to GDP in current prices, on health indicators such as life expectancy and infant mortality rates While this finding may pose challenges for instrumental variable (IV) candidates like trade openness, these barriers could be mitigated due to the limitations of panel time series data The strength of the effect varies in both directions and is notably significant for low-development countries This research encompasses a range of countries, from high to low income, including many affluent nations The authors also highlight the potential role of trade openness in their models, suggesting diminishing health returns as income rises Similarly, Stevens et al (2013) reached comparable conclusions regarding the impacts of trade openness on health outcomes with fewer controlling factors.

Asconsequences,w i t h l i t t l e empiricalresearchesf o r associationb e t w e e n tradeo pennessuntilrecentlyasinHerzer(2014),extendedforthegrossdomesticinvestment,t h o s e indicatorsares t i l l p l a u s i b l e candidatef o r IV,givent h e realconnectionsw i t h end ogenouspercapitaGDP.Finally,thosecandidateswillbecheckingforeffectdirectlyo n lifee xpectancyandinfantmortalityrateasfirststagebyregressionstobetterconfirmationof function ofIVsinrelationto exogeneitycondition.

Estimation Method

Panel data estimation is essential for analyzing the dynamics of income distribution and its effects on life expectancy and infant mortality rates This method effectively utilizes extensive data to reveal variations within or across different panels (Park, 2011) It enables researchers to assess the reliable relationship between wealth inequality patterns over time in a specific country or track the evolution of the Gini coefficient across 48 countries within a defined time frame Additionally, panel data can mitigate biases caused by omitted variables and measurement errors, while controlling for individual-specific, time-invariant, and unobserved heterogeneity (Sửderbom, 2011).

Theomittedvariableduetounobservedtimeinvariantiscapturedinthemodel.Paneldataest imatorallowst o c o n s i s t e n t l y estimatingt h e effecto f t h e o b s e r v e d e x p l a n a t o r y v ariables.T h e r e i s s o m e significantpaneld a t a m o d e l s whicht o b e consideredi n t h i s s t u d y asbelow:

𝜀 (�= 0) PooledOLSissimpleyetpowerfulmodelforstandardregressions.Ifthereisnocross sectionalortimespecificunobservedheterogeneityappearedinmodels,pooledOLScanb e efficientandconsistent inestimates.

Thetwotreatmentswhichfrequentlybeusinginpaneldataanalysisarefixedeffecta ndrandomeffect.Fixedeffecti s m o s t commonlyappliedm o d e l f o r paneldata.T h e advanta geo f f i x e effecti s t h a t i t canp r o d u c e u n b i a s e d andconsistente s t i m a t e s i n t h e e x p e n s e ofdegreesoffreedom(incaseestimatingFEmodelusingLeastSquaresDummyVariable(L SDV) strategy).

Inopposedt o FEe s t i m a t i o n model,randomeffects( R E ) m o d e l examinet h e errorvar iances p e c i f i c t o g r o u p s o r t i m e s o f paneld a t a underassumptiont h a t heterogeneity(ind ividualeffect)notcorrelatedwith anyofregressors.T h e uii scomponentofcomposite errorterm; hencetheerrorcomponentmodel isalternativeforREmodel.

(� + �) EstimatingR E m o d e l needsu s i n g GeneralizedLeastSquared(GLS)estimatoro r FeasibleGeneralizedLeastSquared(FGLS)dependingonwhetherornotthecovariance

∑ofindividuali is known.Inmostcases,∑ is notavailable.

Endogeneityi s a significanti s s u e i n dealingw i t h econometricm o d e l s Itcanb e c ausedbytheunintendedcorrelationsbetweenregressorsanderrorterm.Thereasonoft h i s p roblemi s t h a t t h e r e a r e c o n f o u n d i n g factorsb o t h relatet o dependentandindependentv ariablesb u t n o t includedi n t h e m o d e l s Accordingt o S t o c k andW a t s o n (2003),threeso urcesofendogeneity couldbearisenfromunobservedomittedvariable,

� simultaneouscausality,andmeasurementerrors.Consequenceofendogeneityisbiasedino u t c o m e andcan beunreliablyusedfor interpretation.

Inordertoaddressthis,themethodofInstrumentalVariable(IV)hasbeenutilized U s i n g t h e instrumentalvariableconceptualapproach,i e , thevariablewhichaffectsdepend entvariablebutdoesnotcorrelatetoerrortermandother explanatory variables,t h e unbias edandconsistentresultw o u l d b e producedw i t h Two-stageLeastSquares(2SLS) estimator.

The Fixed Effect (FE) model is consistent but not always efficient, which is why Hausman's specification test, introduced by Hausman in 1978, is utilized to determine whether to use Random Effects (RE) or FE models with panel datasets The null hypothesis of Hausman's test posits that both models are consistent, meaning that individual effects are not related to any regressors If the null hypothesis is not rejected, it indicates that while the FE model is consistent, it is not efficient; conversely, the RE model becomes inconsistent and biased, making the FE model the preferred choice Ultimately, the coefficients of both models should not differ significantly (Park, 2011; Sửderbom, 2011).

BreuschandP a g a n ’ s ( 1 9 8 0 ) LagrangeM u l t i p l i e r (LM)t e s t e x a m i n e s whetheri n d i v i d u a l (ortime)specificvariance-componentsarezero.Thatmeansundernull hypothesis,� 0 :𝜎 2= 0.Ift hi s i s rejected,r a n d o m effectm o d e l ( R E ) c an b e usedm o r e efficientlythanpooledOLStodealwithheterogeneity;otherwise,pooledOLSmodelis goodenough(Park,2011;Baltagi,2008).

0.F-testoutcomeisbasedonthegoodness-of-fitmeasures(SSEorR 2 ).Ifnullhypothesis isrejected,thereisconsiderableincreaseofgoodness-of- fitinFEmodel.Hence,pooledOLSmodelisvulnerablecomparedtoFE(Park,2011).

Overall,the feasibleregressionsthen canb e a pp li ed totwomodels ande x t e n s i v e l y controlledformultiple issues ofeconometrics.

(1)FtestforFEeffectisapplied,then(2)LMtestforREeffectandfinally(3),Hausmantestf orREandFEtocheckforsuitableeffect.InstrumentalVariable(IV)estimationcanbefurtherfo llowedtoachievethefinerresults.TableA.2ofprocedureasi n Appendix(Park,2011):

TwoissuesofanyOLS-basedmodelsare theheteroskedasticityandserialcorrelation.Heteroskedasticitypresentsforthedispersionorun equalofvarianceduetosomereasonssuchasoutliers,ormixing ofmeasurements.The downsideofheteroskedasticityi s misleadingconclusions asaresult of wrong Fortstatist icseventhoughheteroskedasticityitselfdoes notcauseloss of biasness(Gujarati,2009).

Serialcorrelation(orauto-correlation)isonkind ofissuearisenintimeseriesanalysiswhenerrortermscorrelatetothepastoneeachothers.As i ncaseofheteroskedasticity,serialcorrelationd o e s n o t affectu n b i a s n e s s o r c o n s i s t e n c y o f estimatorb u t s t i l l problematicinconclusiondrawingbecausethestandarderrorsarealtere daswellastheFt e s t i s n o t m o r e reliable(Gujarati,2009).

DescriptiveStatistics

Table 4.3 presents descriptive statistics for various variables, highlighting significant variations, particularly in GDP per capita and health spending per capita These disparities indicate substantial differences in income and expenditure among populations, which likely contribute to inequalities affecting life expectancy and infant mortality rates (IMR) The average life expectancy in high and middle-income countries stands at 73.6 years, while the infant mortality rate is alarmingly high, with nearly 83 deaths per 1,000 births Additionally, the average percentage of individuals attending secondary education and the trade share of GDP are notably high, at approximately 95% and 78%, respectively This suggests that human development and trade in a global context may have been improving in recent decades.

Variable (realterm) Observations Mean StdDev Min Max

*t h e percentagep o i n t s i n t h i s caseover1 0 0 % i n c l u d i n g o f over-ageda n d under- ageds t u d e n t s becauseofearlyorlateschoolentranceandgrade repetition.

Source:WorldBank, ChildMortality Estimates,WorldIncomeInequalityDatabase

6 0 8 0 5 5 6 5 7 0 7 5 In fa nt M or ta lit yR at e 0 2 0 4 0 6 0 8 0

Real GDP per capita (USD) lifeexpc imr

Subsequentlywithdatastatistics,figuresinAppendixdemonstratehowdifferentt h e incomeinequalityandincomepercapitaaffecthealth,aswellasothermainpredictors,dat afrom48countries,1991-

2010.Figure4.4showslifeexpectancyaswellasinfantm o r t a l i t y r a t e drawna g a i n s t GDPpercapita.Itrepresentst h e overviewofrelationshipofincome levelandtwo depende ntvariablesinwhole sample 4 8 countriesover20years.IncomelevelasGDPpercapitaiskeyc ontrolfactorforincomeinequalityhypothesiswhichconnectsincomedistributionandhealth Italsohelpsverifytheabsolutei n c o m e h y p o t h e s i s Trendi s v i s i b l e througho u t t i n e periodw h i c h GDPpercapitapositivelycorrelateswithlifeexpectancybutnegatively withinfantmortalityrate.However,theassociationsareprioriknownasnotlinearandhavingdec liningrateeffect.

TheGinieffectonlife expectancyandinfantmortalityrateis shown inFigure4.5.G i n i representstheincomeinequalityinstudyandtherefore,essentialforpro jectingtheresultsofmodelregressions.Moreover,thepatternofdatapointsscatterswidelyb utthetrendingc o r e concentratesc l e a r l y w i t h o p p o s i t e directioni n lifee x p e c t a n c y an dinfantm o r t a l i t y rateversusGini.Lookingatlifeexpectancy,trendingassociationi snegativelylinearinnoticeablecloudofdata.Incontrast,infantmortalityrateispositivelycon nectedw i t h Gini,but thetrendonlypicksupataround Ginivalued40,andquitenotco ncentrated.ForGini20-

42 40,infantmortalityrateisnotverymuchimprovedovertime.T h i s phenomenono f scattere dd a t a i s evidenceo f uniformimprovemento f l i f e

6 0 8 0 5 5 6 5 7 0 7 5 In fa nt M or ta lit yR at e 0 2 0 4 0 6 0 8 0

Gini ratio % 50 60 70 lifeexpc imr expectancyversusGiniovertimeandreductionofinfantmortalityrateversusGini>40.T h e variationsarefairlylarge atsomecertaindegrees.

Speakingofhealthspending percapita,one among controlvariables, Figure4. 6shownbelowhasa goodsupplementf o r absolutei n c o m e hypothesis,w h i c h GDPp e r capi taimpactsonhealth.Healthindexesareabsolutelyinfluencedbythelevelofhealthcaree x p e n d i t u r e i n t e r m o f percapita.T h i s i s reasonablef o r t h e naturalconnectionsbetweent woconcepts.Thedirectinfluencearetobeexpectedbecausehealthcarespendingleadstoimpro vementinshorttermhealth,andinturnlongtermhealthcanbeaffectedbystablehealthcare spendingpolicies.T h e effectisatstrongestlevelatuptoover2 0 0 0 U S D t h e r a p i d l y s t e a d y e s p e c i a l l y t o infantm o r t a l i t y rate.K i n k i n g p o i n t i n trendcouldb e t r o u b l e s o m e whend o i n g regressionbecausei t cancreateunexpectedresults.

6 0 8 0 5 5 6 5 7 0 7 5 In fa nt M or ta lit yR at e 0 2 0 4 0 6 0 8 0

Health expenditure per capita (USD) lifeexpc imr

Anothervariablei s o f c o n t r o l i s s e c o n d a r y s c h o o l i n g e n r o l m e n t r a t i o , w h i c h i s f o u n d significantd e t e r m i n a n t i n literatureo f h e a l t h - e d u c a t i o n r e l a t i o n s h i p Figure4 7 s h o w n belowhasdrawnoverallpictureo f s e c o n d a r y schoolingandhealthmeasures.Thereisnosurprisepatternofschoolingsimilartowh atothervariableshavedone.From5 0 % oftotalsecondaryschooling,itdramaticallyraises lifeexpectancyanddropsinfantm o r t a l i t y rate.Thereisalsokinkpointobservedindatao finfantmortalityrate,itmights l i g h t l y affecttheregressionsignsin thetheory.

6 0 8 0 5 5 6 5 7 0 7 5 In fa nt M or ta lit yR at e 0 2 0 4 0 6 0 8 0

Secondary schooling enrolment rate % lifeexpc imr

Figure4 7 : L i f e expectancya n d infantm o r t a l i t y ratev e r s u s Secondaryschoolingenr olmentratio

TwotablesA.4andA 5 s h o w n i n A p p e n d i x summarizedt e s t s f o r t h e t w o ( 2 ) dependentvariables.TheHausmantestverifiedifthedifferenceincoefficientsnotsystematic. Criticalevaluationsfor classicallinearregressionmodelssuchasheteroskedasticityandseri alcorrelationarepresentedw i t h criticalstatisticsandrespectivep r o b a b i l i t y valuest o t e s t t h e hypotheses.T h e s p e c i f i c a t i o n errord u e measurementerrorcanbeovercomebyI Vinlateranalysis.Functionalformsandinvolvedcorevariablesisconsistentfromliteratures.

Allt e s t s w i t h Breusch-PaganLMTestandF - t e s t f o r c ho os in g betweenOLSo r R E andFEm o d e l h a v e beenfavoredt o t h e l a t t e r w i t h highsignificancelevel.Resultsrepresentsi n firstsectioni n TableA.4andA.5.A s conclusion,

Theresulto f Hausmant e s t indicatedt h a t t h e Randomeffect( R E ) i s f i t f o r l i f e e x p e c t a n c y modelat5%significancelevel(p- value=0.0623).Otherwise,theinfantm o r t a l i t y ratem o d e l i s m o r e f it w i t h Fixede ffect( F E ) at1 % significancel ev el ( T a b l e A.5).T h i s resulti s reasonablew i t h datai n whic ht h e G i n i oftens l o w l y changingo v e r t i m e andthevariationsnot easilycapturedbyFE.

The Wooldridge test for autocorrelation and the Wald test for heteroskedasticity have been utilized to identify econometric issues that may lead to incorrect statistical testing and conclusions All results meet the probability requirements for significance levels, confirming the presence of serial correlations and heteroskedasticity in the models To address these issues, robust standard errors and lagged values of variables will be employed as remedies using Stata commands.

In the analysis of data descriptions, the correlations among all variables are examined to investigate their mutual relationships, as detailed in Table A.6 of the Appendix A significant co-linearity is observed in the models, particularly between health spending per capita and GDP per capita, which shows a high correlation of 0.9 Additionally, potential multicollinearity exists at varying degrees (around 0.5) among pairs of explanatory variables, including the school enrollment rate in relation to the Gini index, GDP per capita, and health spending per capita It is logical to conclude that higher income per capita leads to increased healthcare expenditures (Lynch et al., 2004) Moreover, the Gini index and secondary school enrollment have complex effects on GDP growth rates (Bloom et al., 2004), which ultimately influence GDP per capita and healthcare spending.

Becausem u l t i c o l i n e a r i t y couldn o t producebiasness,andf o r t h e certainac ceptabledegreesofseverityoftwosignificantvariablesGDPpcandHealthpcinrelationw i t h o t h e r o n e s , t h i s s t u d y leavesc o l i n e a r i t y i s s u e t h e r e andd e l i b e r a t e l y n o t u s i n g anytreatmentsforit This is onesort ofmethodsrecommended.

MainvariableGi ni has s t a t i s t i c a l l y significantassociationwithtwo targetsvariab lesl i f e expectancyatb i r t h andIMRat5%significancelevel,-

0.54and0 6 7 respectively.Otherexplanatory variablesalsoexhibithighcorrelationsandc orrectsignsw i t h dependentones,e s p e c i a l l y f o r GDPp e r c a p i t a w i t h highestcoeffici ents0 7 5 and

Equallyimportant,twoIVhasdemonstratedamoderateperformancewheretheyaren ot m u c h i n r e l a t i o n t o dependentvariablescomparedt o o t h e r v ar ia bl es (ataround0.02-0.2),particularlylifeexpectancyatbirthwithremarkablylowsignificance,aswellm o d e r a t e l y lowtoIMR.Moreover,relevantcriteriainwhichIVsaffectendogenousGDPperca parefairlysolelysatisfiedforTradetoGDPratio(0.31).Asconclusion,IVsare mildlyappropriateremedyforourmodels.Hence,theywillbetestedfurtherbyregressionte chniquesfollowing.

The supplementary test results in Appendix A.7 assess the influence of proposed instrumental variables (IVs) on life expectancy (LE) and infant mortality rate (IMR) The primary aim is to evaluate the significance and effect of Trade Openness and Investment Ratio on these target variables Findings align with existing literature by Pritchett and Summers (1996), indicating that while both IVs show correct signs, they have minimal impact on the target variables, with only the Investment Ratio significantly affecting IMR at a 5% level This analysis, supported by correlation coefficients and indirect relationships with dependent variables, suggests that the two IVs can serve as safe instruments for life expectancy but not for IMR Further testing will be conducted in the first stage of the two-stage least squares (2SLS) method to confirm their strength in addressing endogeneity for life expectancy at birth For IMR, the fixed effects model with unobserved heterogeneity control, as per Hausman tests, is deemed more appropriate, and no suitable IVs are available for treatment.

ResultInterpretation

Forrobustnesspurpose,m o d e l s a r e testedt o seet h e differencesbetweent h e n OL Sregressionandfixedorrandomeffects,Table4.8and4.9showsthelimitations oft h e O LSandR E aswellasFE’ssignificantchangesi n outcomesi n termso f signandmagnitude.Inm o d e l s ( 1 ) and(2),t h e R E choiceproducedq u i t e differentmagnitudesfromOLSresults,event houghm a i n G i n i o n l y significantat1 0 1 % T h i s i s wheret h e advantageofREmodelisutil izedtocapturerealvariationofGiniversuslifeexpectancy.

Thesamepatternsi n outcomesf o r m o d e l s ( 3 ) and( 4 ) , wherealle x p l a n a t o r y variablesstatisticallysignificantbutdifferentmarginaleffects.However,allsignsareinex pectationexpectforhealthcarespendinginmodelFE(4),whichreflectsstrengthofFEi n accura telyestimatingcoefficients.Themodelsdon’tusenatural logarithmtransformationbecaus etherelationshipbetweenmultivariatemodelsdoesnotchangeduet o l o g transformationaccor dingtoM a c i n k o eta l

( 2 0 0 4 ) Moreover,t h e logstreatmentu s u a l l y makesvariablesinsignificant.This study,t herefore,useslevelsforallvariablesf o l l o w i n g theinequality- healthliteraturesofHerzerandNunnenkamp(2015),Macinkoetal.(2004) andLeighetal.(2009).

The study results are summarized in Tables 4.10 and 4.11, highlighting the application of life expectancy and infant mortality rate (IMR) in random effects (RE) and fixed effects (FE) specifications (1) and (4) Additionally, life expectancy is analyzed using instrumental variable (IV) trade openness and investment ratios to address endogenous issues in specifications (2) and (3) All variable signs align with prior expectations, except for health expenditure per capita in the FE model for IMR, which shows an inverse relationship Each model controls for GDP per capita, health spending per capita, and secondary schooling attainment rates, with Model 1 accounting for autocorrelations in residuals Stability is observed across all specifications, with most coefficients statistically significant at the 1% level Notably, the Gini coefficient, when negatively associated with life expectancy, is estimated to range from 0.036 to 0.038, indicating that a 10% increase in the Gini index corresponds to a reduction in life expectancy by approximately the same percentage.

0.37year,ceterisp a r i b u s.AscomparedtoTorreandMyrskylọ(2014),theyestimateddoublin g thatofGiniw o u l d resultindecrease1percentand1.3percentinlifeexpectancyfor menandwomen.Notably,significancel e v e l hasbeenincreasedforG i n i whenIVsperforma ncei s atworking.IntermofGDPlevelperhead,1000USdollarsboost(4.8%inmean)leadsto anaverage0 2 yearr a i s e i n l i f e expectancy,c e t e r i s p a r i b u s Similarlyf o r healthca reregarding directionbutwithmuchlesserquantitativeeffectwhenspendingmore 100U.S.dollarsforhealthcare,whichis6.9%moreinmean;givenallotherthingsequal,rende ro n l y 0.06yearreducedinaveragelifeexpectancy.Theimpactofsecondaryeducationiss u r p r i s i n g l y s l i g h t l y s t r o n g e r t h a n income;i m p r o v i n g 0 3 1 yearofl i f e e x p e c t a n c y i n averagewhileadditional10%in schoolingattained.

The Infant Mortality Rate (IMR) shows an inverse relationship with the Gini coefficient, indicating that a 10% increase in mean Gini can lead to an additional 1.6 infant deaths per 1,000 births In contrast, improvements in income and education, specifically a 10% rise in GDP per capita and secondary schooling ratios, are associated with reductions in infant mortality by 1.23 and 1.42 deaths per 1,000 live births, respectively Interestingly, while health expenditure is positively correlated with IMR, a 10% increase in spending results in 0.15 additional infant deaths per 1,000, highlighting a marginal yet statistically significant effect These contradictory findings emphasize the complex relationship between health spending and infant mortality outcomes, suggesting that healthcare systems may play a crucial role in influencing overall results, as supported by similar findings in Macinko et al.

(2013)findingswhenp u b l i c h e a l t h e x p e n d i t u r e i s positivelyassociatedw i t h IMRi f c ontrolf o r GDPpercapitai n multivariateanalysis.Moreover,Macinko’ss t u d y showedevid encesmorespendingonhealthnotalwayscorrelateswithbetterhealthi n those with highinc ome (OECDcountries).T h e lyingreasonwasblamed topossiblemulticolinearitybetweenex penditurepercapitavalueswithhighcorrelationswhichmightalsobetrueforthisstudy.Beside sthat,vanBaaletal.(2013),inlifeexpectancyandhealth spendingresearch,evenfoundsensitiveconclusionsregardingpositiverelationbetweenthem.

The influence of income inequality on health is significantly linked to social capital, which encompasses the social position and resources available to individuals, rather than just absolute income levels Societies with lower health outcomes often exhibit reduced social capital, leading to greater health disparities within communities (Kawachi et al., 1997; Wilkinson, 2002; Wagstaff & Van Doorslaer, 2000) Key factors contributing to this phenomenon include psychological impacts such as social stress and limited access to essential resources Additionally, GDP and per capita health spending play vital roles in health outcomes by enhancing access to quality healthcare, education, and nutrition, as well as promoting healthier lifestyle choices Notably, investment in education is essential for improving life expectancy and reducing infant mortality rates through the enhancement of knowledge and skills.

Inshort,theoverallperformanceofkeyfactorsseemtobesmallwhensensitivechang esm a d e , e.g.,s m a l l e r t h a n 0 5 yearw i t h l i f e e x p e c t a n c y increaseand1 6 infantdeat hsreductionwhenitmatterstoIMR.However,thereliabilityofestimatesishighduet o thefocus eddistributionof highandupper middleincome countries.Theseestimatesarecompara bletootherliteratureswhereonlymarginalimpactshavebeenfound(Leighetal.,2009).

Table4.10:Effectsofincomeinequality usingfixed-effectsandrandom-effects

VARIABLES Lifeexpectancy p-value IMR p-value

Table4 1 1 : E f f e c t s o f incomei n e q u a l i t y u s i n g instrumentalv a r i a b l e s onl i f e expectancy

VARIABLES Lifeexpectancy p-value Lifeexpectancy p-value

The interaction effects of income inequality and income levels on life expectancy vary significantly between high, upper-middle, and lower-middle-income countries, as summarized in Tables 4.12 and 4.13 Notably, the Gini coefficient shows no impact on life expectancy at birth in wealthier nations compared to poorer ones Conversely, GDP levels significantly influence life expectancy in high and upper-middle-income countries, with advanced economies showing a minimal reduction in life expectancy of 0.15 years despite a 5% improvement in trade openness and a 0.11-year decline linked to investment ratios These subtle differences, though small, reinforce the diminishing effect of income on life expectancy, consistent with past studies by Preston (1975), Rodger (1979), and Wilkinson (1992) This aligns with the relative income hypothesis, which suggests that absolute income becomes less significant for wealthier individuals, as noted by Gravelle (2002) Furthermore, the health benefits of increasing income are typically more pronounced at lower income levels, reflecting a curvilinear relationship identified by Deaton (2001).

Othervariablesc a n b e m o r e o r l e s s comparablet o previousm o d e l s i n termo f qu antitative coefficientsandallproduceexactsignsasprioriexpectation.

Table4 1 2 : Regressionso n l i f e expectancyw i t h interactionso f Ginia n d GDPperhead -Trade Opennessinstrument

Table4 1 3 : Regressionso n l i f e expectancyw i t h interactionso f Ginia n d GDPperhead –InvestmentRatioinstrument

Fort h e IMRasd e p e n d e n t variable,G i n i h a s noo r l i t t l e impactwhent a k i n g differenceintoaccountacrossincomelevel(Table4.14).ThatcanbeinterpretingGini– asmeasuremento f i nc om ed is tr i bu t io n – hascoveredt h e p a t t e r n s atl e a s t f o r countrieswithG N I over1 , 0 2 6

U S dollarsi n sample;t h u s p r o v i n g t h e p l a u s i b i l i t y o f i n c o m e i n e q u a l i t y aswellasrelativeincomehypothesis(whichtakesrelativeaverageincomeofp o p u l a t i o n relatedt o individualhealthinsteado f i n c o m e i n e q u a l i t y ( W a g s t a f f & Van

Doorslaer,2 0 0 0)).T h i s f i n d i n g is consistentw i t h whatT o r r e andMyrskylọ(2014)hasf o u n d forincomeinequalityandinfantdeaths.

Conceivably,t h e p e r c a p i t a i n c o m e discriminateseffectbetweent h e l e s s a ndh i g h l y d e v e l o p e d c o u n t r i e s W h i l e f o r h i g h l y developedcountries,1 0 % i ncreasecorrespondstoreductionofonly0 7 3 infantdeaths,sameraiseinlesserdevelopednatio ns(i.e.,poorerones)savesmorethan3.31childrenunderoneyearoldper1,000liveb i r t h s , ot hersunchanged.Feasibleexplanationisarguablyt o b e lyingunderabsolutei n c o m e hypot hesiswhenotherthingsremain constant.

Thehealthspendingperheadisreexaminedtobestillinoppositesignbut,whentreate dindistinguishingbetweenhigh/ upperandlowermiddle income,nolongers t a t i s t i c a l l y significant.Tosummarize, interactionswithGiniratioandindexhaveclarifiedhowdiversegroupsofincomedrivet heeffectofincomelevelanddistributiono n c a p a b i l i t y o f reducingt h e infantd e a t h s andp r o l o n g l i f e ofpeople.G i n i asm a i n determinanto f healthi s n o actori n m o d e l s 1 , 2;asi n c o m e i n e q u a l i t y hypothesisi s somewhatsusceptibleduetoincomeleveld iscrimination.

M (GeneralizedMethodofMoments)asproofofdynamicmodelsinTableA.15inA p p e n d i x UsingsystemGMMestimatorbyArellanoandBond(1991),whichincludinglaggedIVsv ariablea n d seto f aslagso f endogenousvariablesi n m a t r i x , moreinformationw i l l b e exploi tedi n t o revealtransmissionmechanismo n vitali n d e x e s o f health.

GeneralizedMethodo f M o m e n t s (GMM)b e t t e r u t i l i z e s efficientestimates o f dynamicp a n e l databyu s i n g m u l t i p l e i n s t r u m e n t s Iti s ablet o d e a l w i t h uno bservedheterogeneity,heteroskedasticity,alsoallowedforautocorrelation.SystemGMMisc onsideredsystemofequations, wheredependentvariableisdynamic,dependingonitso w n pastvalues,whileexplanatoryvariablesarenottotallyexogenousandcanprobablyb e correlatedwitherrorinpastorcurrenttime.InStata,estimatorArellanoandBondforsystemGM

The models are robustly evaluated using Sargan and Hansen tests for over-identified restrictions, revealing that while they perform well, excessive instruments can weaken their reliability Utilizing five instruments, the one-step system GMM incorporates Trade Openness, Investment ratio, and lags to two levels of endogenous GDP per capita Overall, the performance of system GMM with level equations is modest, with only two significant coefficients at the 10% level Although the expected signs are maintained, GDP per capita does not show significant results, while health expenditure per capita reflects the correct sign despite its insignificance Structurally, system GMM with an IV style serves as an additional verification step for the validity of the two main models, confirming that the models hold up well despite vulnerabilities associated with the numerous instruments and methodologies used.

Chapter5:Conclusions

ConcludingRemarks

This study examines the hypothesis of income inequality and its impact on vital health indices across 48 high and middle-income countries The complex relationship between income distribution and health outcomes remains a topic of debate among researchers With the availability of long-term data and advancements in econometrics, the pathways linking Gini coefficients and other controls to health concerns can be explored more effectively Utilizing panel data analysis, the study thoroughly assesses the relationship between income inequality and public health outcomes, revealing that while Gini is a significant predictor of life expectancy, its negative effects are notable However, the impact on infant mortality rates was unexpectedly positive when accounting for per capita health expenditure Additionally, Gini's influence appears minimal in higher-income countries compared to GDP per capita Overall, the quantitative findings align with previous studies, reinforcing the complex interplay between income inequality and health.

Research indicates that GDP impacts infant mortality and life expectancy differently in poorer and richer countries The relationships between the Gini coefficient and life expectancy, as well as infant mortality rates, highlight significant findings in the literature Notably, an increase of 10% in the Gini coefficient is associated with a reduction of 0.37 years in life expectancy across the entire sample Additionally, a 10% rise in secondary schooling attainment correlates with an increase of 0.31 years in life expectancy The effect of the Gini coefficient remains consistent regardless of income group distinctions For GDP per capita, a 10% increase leads to an extension of life expectancy by 0.4 years In contrast, higher and upper-middle-income countries experience a positive effect of 0.26 years, while lower-middle-income countries see an average increase of 0.72 years Without separating income levels, a 10% change in income inequality influences infant mortality rates significantly.

GDPbyhead,oreducationenrolmentcouldshrinkorraiseinfantmortalityc o r r e s p o n d i n g l y by1 6 , 1.23,and1.42for every1 0 0 0 livebirths.Whenrestrictingattentionto in co me subgroups seriously,GDPpe rcapitai s up1 0 % holding down0 7 3 newborndeathsinhigherincomecountriesand3. 31onaverageinlowermiddlecountries.

Eventhoughfairlyconsistency,interestingoutcomeinsenseofcommonrationalealsof o u n d f o r healthc a r e s p e n d i n g variable.T h e signi s u n e x p e c t e d a n d magnitudei s m o d e r a t e l y small,10%morehealth care spending foreachheadonlyextend lifeto0.087-

0.1year,andessentially raisebyaround0.15childdeathsrate.T h i s effectisperceivedandr eflectedinvanBaal etal.

(2013)’srelatedpap er,whichevidencesareobviousforp o s i t i v e impactofhealthspend ingonlifeexpectancyineconomicallyhighincomecountries.However,t h e effecti s n o t clearin strengthd u e t o weakm a r g i n a l quantities.T h e possiblereasonistheoverallhealthcares pendingisnotabletocapturethenaturalvariationsofecologicaltransmissionofhealthoutco mes.Moreover,generalhealthspendingi s s u s t a i n a b l y biasedtowards o m e groupso f p a r t i c u l a r l y vulnerabilities,f o r instance,v e r y youngo r o l d aging.Finally,t h e underl yingmechanismo f healthcaree x p e n d i t u r e andlifeexpectancyaswellasmortalityisn otyetfullyunderstood(vanBaaletal.,2013).

Insummary,t h e s e p a t t e r n s o f findingsarei n circleo f s u p p o r t i n g a s s o c i a t i o n betweeni n c o m e i n e q u a l i t y andhealthliteratures.Therea r e m o r e s u p p o r t i v e evidencest h a t , eventhoughmanydifficultiesencounteredwhenm a k i n g j u d g m e n t o f economici n e q u a l i t y inconnectionwithphysiologicalcharacteristics,thepathways underlyingunequalsocietiesarebeingrevealed.Thedistributionofprosperitythroughthisst udyhasproveditselfasoneofthemajorexplanatoryfactorsforhealthataggregatelevel.Infact,g ivent h e moderater o b u s t n e s s , lowerG i n i hasprolongedl i f e expectancyandreducedne wbornsd e a t h s bysmallmarginsi n m i d d l e andhighi n c o m e countries(ranging0.37-

Thiss t u d y i s n o t abletou n d o u b t e d l y c o n f i r m t h e i n c o m e i n e q u a l i t y n e g a t i v e impacto n high/ upperm i d d l e versuslowerm i d d l e i n c o m e countries;hence,t h e conclusioni s m e r e l y t h a t i n c o m e i n e q u a l i t y h y p o t h e s i s h o l d s f o r upperm i d d l e i n c o m e countriesupw ard.However,thisiscontrasttosomeresearcheswhichonlyconnecti n c o m e inequalitytode velopingcountriessuchasWagstaffandVanDoorslaer(2000);or

Less healthy societies raise significant concerns regarding social cohesion and inequality, often leading to psychological effects such as social stress and limited access to resources (Kawachi et al., 1997; Wilkinson, 2002; Torre & Myrskylä, 2014; Macinko et al., 2004) These populations typically experience lower productivity and ongoing social challenges Government policies aimed at closing the income gap not only focus on income redistribution but are also politically motivated (Wagstaff & Van Doorslaer, 2000) Progressive income taxation serves as an effective tool for wealth distribution in societies and plays a crucial role in alleviating extreme poverty.

This study demonstrates a consistent positive relationship between GDP per capita and health outcomes, significantly increasing life expectancy at birth and reducing infant mortality rates across various specifications The impact of GDP per capita varies between lower-middle and high/upper-middle-income countries, with stronger effects observed in economically disadvantaged regions Despite some limitations in the GMM test regarding income levels, the findings reinforce the importance of economic development policies in lower-middle-income countries Overall, the absolute income hypothesis remains robust, highlighting the critical role of income in enhancing health metrics.

Educationisthemostlikelypowerfulpredictorforhealthbesidepercapitaincome.Itp r o v e d t o b e r o b u s t andconsistentinm a j o r specifications.Investmenti n s e c o n d a r y educationisamongmostvisiblepathwaystosustainablyaddressandimprovep o p u l a t i o n health.O n theo t h e r hand,healths p e n d i n g perc a p i t a i s l e s s influentwhenproduc einversesignsandminormagnitudewhilestillrobust.Thiscontroversialfindingcanbeex plainedandconsistentwithMacinkoetal.

(2004)whentheyfoundhealthpublicexpenditurespositivelyrelatedtoIMRwithcontrolforG DPpercapita.However,governmentswithpoor- centricpoliciescannotonlyredistributeincome,butinvestmoreo n healthservicesforpoorerc ommunities.

Byemployingrecent datasourcesandpanel analysis with attention toendogeneityissuesandothereconometricconcerns,thisempiricalstudytriestoaddresss omel i m i t a t i o n s o f previousones.Butevens o , t h e researchc a n n o t s o l v e t h e c a u s a l i t y o f relationships.Suchcausationistobeexploredbyimplementingp a n e l cointeg rationm e t h o d s infutureresearches.Secondly,theGiniiswellknowntocorrelatewithnu mbers

59 ofeconomicv a r i a b l e s ; therefore,t h e r e leavesfuturestudiesw i t h alternativemeasurement so f i n c o m e i n e q u a l i t y t o bei m p l e m e n t e d f o r c o m p a r i s o n Thirdly,classif icationo f geographicalareai s alsoi s s u e t o t h i s s t u d y whenc o u n t r i e s w i t h l o w i n c o m e countriesi g n o r e d andcoveredareai s d i s t o r t e d Lastly,u n d e r l y i n g mechanism andcomprehensiveaspectsofsocialinequalityshould bedeepeningtorefinethemodel o f incomedistributiononhealth.Furthermore,morestudiesneedtobecarriedouttofindwhich subgroupso f s o c i e t y b e n e f i t m o s t fromclosingt h e i n c o m e gap,especiallyi n fra gmentsofextremelyrich orpoorin diversecountries.

Tobeconclusion,incomeinequalityhypothesisremainsholdinthisresearchforlow er/ uppermiddleandhighincomepopulationacrossnations.Lowerinequalityisablet o i m p r o v i n g l i f e e x p e c t a n c y andr e d u c i n g i n f a n t deaths.O n o t h e r hand,evidencet h a t GDPpercapitahelpsustainabsolutei n c o m e hypothesisa r e reflectedint h i s researchfindings.H

Policyimplication

The findings on income inequality and per capita income highlight the need for central governments to implement policies that can enhance life expectancy and reduce infant mortality rates Understanding the underlying pathways of income distribution's impact on health can guide effective policy-making A less economically distorted society is better equipped to address chronic social issues, fostering a work ethic and creating a harmonious living environment Moreover, economically equal communities tend to experience lasting health benefits Therefore, governments are encouraged to adopt measures to redistribute income and gradually close the wealth gap, with progressive income taxation being a key instrument to alleviate extreme poverty and promote equity in society.

GDP per capita has consistently demonstrated a positive correlation between income and health outcomes, significantly increasing life expectancy at birth and reducing infant mortality rates across various contexts The impact of GDP per capita is notably stronger in lower-income countries compared to upper-middle-income nations, reinforcing the absolute income hypothesis These findings suggest that governments and relevant authorities should implement policies that promote economic development to enhance GDP per capita, particularly in lower-middle-income countries.

Educationisalsosuggestedasoneofthemajorinfluencerforhealthalongsidepercapitain come.Itproveditselftobeconsistentinspecificationsandbestlikelycandidatef o r inmacro policiesifcentralgovernmentswanttoefficientlyimpactthelongrunhealthi n d e x e s Spec ifically,thesecondaryeducationisoneofbasicinputtoeffectivelycontributetolifetime achievementsandpopulationhealthaccordingtothesestudyfindings.M o r e i n v e s t m e n t o n fundamentaleducationbygovernmentsi s l i k e l y t o h e l p accomplishtargetsof prolonginglifeexpectancyandreducinginfantmortalityrate.

Byc o n t r a s t , healths p e n d i n g p e r c a p i t a i s q u i t e c o n t r a d i c t o r y w i t h unpredictedsignsandsmallmagnitudewhilestillrobust.Thisfindinghassamepatternandconsi stentw i t h Macinkoetal.

(2004)whentheyfoundhealthpublicexpenditurespositivelyrelatedt o IMRwithcontrolfor GDPpercapita.However,becauselocalcommunitiesarelessorm o r e benefitedfromp u b l i c healthcarei n termo f healthfinance,g o v e r n m e n t s s h o u l d alwaysi m p r o v e t h e h e a l t h careaccessandi n v e s t o n p u b l i c healths e r v i c e s T h e m o s t importanti s t o relocate e f f i c i e n t l y thef u n d s , reducew ast in g andf i x i n g the shortagein h u m a n resourcesan dcapablehealthsystems.

Limitationsandfurtherresearches

Thisstudyhasrepeatedlyaddressednumberofcaveatswhenattemptingtoprovet h e relationshipbetweenincomeinequalityandhealthsuchasunobservedheterogeneityandu s i n g datao f r a t h e r homogenouscountrieso f n o t t o o l o w income.T h i s i s caseo f Mellorand Milyo(2001)whensmallnumberofcountriesofonespecificgroup.Measurementerroro f

G i n i andotherscouldb e problematicf o r t h e realeffectt o b e revealed.Furthermore,endog eneityhasbeentakencaretoavoidunbiasness.

However,thisstudystillposessome limitations arisenwhichcouldbeideasfori mprovementsinrefinedresearches.Thedirectionofassociationisparticularlyimportanti f po licieswanttobeefficientandlessrisky.Thepaneldataeffectsusedinstudycannot

61 effectivelysolvethecausalassociationofrelatedfactors.Nextstepofexploringtherootcauseo fcausationin modelis to usepanelcointegrationinresearches.

Besidesthat,errorsi n measurementofvariablesarealsosourceo f b i a s n e s s , e s p e c i a l l y w i t h i n c o m e d i s t r i b u t i o n G i n i i s connectedt o o t h e r v a r i a b l e s throug hconversiono f socialeconomicalsurveyss o i t s relationshipw i t h o t h e r indexescanb e corr elating.U s i n g o t h e r measurementso f i n c o m e d i s t r i b u t i o n willgiveu s consolidatedass ociationofincomeinequalityonhealth.Ontheotherhands,highqualitydatahasbeenemergingo vertimeofferingsupportformodelanalysis.

Thirdly,geographicaldiscriminationisotherconcerntothisstudywhencountriesw i t h lowincomecountriesignoredandcoveredareaisdistorted.Sexandagemixingindifferentg roupofpopulationinlargesetofdatasampleisaswellasanobstacletoshedlightontruthofi ncome inequality hypothesis.Subgroups willbecautiouslyinvestigateandwhendat ais readilyavailable,ageandsexofarefurtherfocusof newresearches.

Intheend,mechanismofincomeinequality–health is still much ofcontroversiesevenw i t h recents t u d i e s Itcanbeinfluencedbymanyfactorsd e e p l y l y i n g i n socialinequality,andi n h u m a n characteristicssuchasbiologicaldifferences,n a m e l y s exandage( T o r r e

& M y r s k y l ọ ,2014).Therefore,therei s anemergencyt o researchwhichsubgroupsofsoci etybenefitmostfromclosingthewealthgap,andfocusonthediversityo f groupsofdifferentsoci alandeconomicalstatuses.

(2015).Accountingforcontributiono f tradeopennessandforeigndirectinvestmentinlife expectancy:Thelong-runandshort-runanalysisinPakistan.SocialIndicatorsResearch,1- 16.

(1991).Sometestsofspecificationforpaneldata:MonteCarloevidenceandanapplicationt oemploymentequations.Thereviewofeconomic studies,58(2),277-297.

Baltagi,B.(2008).Econometricanalysis ofpaneldata JohnWiley&Sons.

Barro,R J., &Sala-i-Martin,X.(1995).Economicgrowth.NewYork.

(2004).T h e effecto f h e a l t h o n economicgrowth:aproductionfunctionapproach.Wo rlddevelopment,32(1),1-13.

Beckfield,J.(2004).Doesincomeinequalityharmhealth?Newcross- nationalevidence.Journalof Health andSocialBehavior,45(3),231-248.

(2014).Healthandwealth:S h o r t panelGrangercausalitytests f o r d e v e l o p i n g c o u n t r i e s TheJournalo f InternationalT r a d e & EconomicDevelopment,23(6),755-784.

(1sted.,p.1).NewYork.Retrievedfromhttp://www.un.org/esa/population/ pubsarchive/chart/9.pdf

(2016).Childmortality.org.Retrieved1 7 J u n e 2 0 1 6 , fromhttp://childmortality.org

(2001).Health,inequality,andeconomicdevelopment(No.w8318).Nationalbureauofec onomic research.

Deaton,A , & H e s t o n , A (2008).UnderstandingP P P s a n d PPP- basedn a t i o n a l accounts(No.w14499).NationalBureauofEconomicResearch.

(1998).C o n s i s t e n t covariancematrixestimationw i t h s p a t i a l l y dependentpan eldata.Reviewofeconomicsand statistics,80(4),549-560.

Ellison,G T (2002).Lettingthe G i n i o u t o f t he bo tt l e?

Challengesf a c i n g ther e l a t i v e i n c o m e hypothesis.SocialScience&Medicine,54(4),5 61-576.

(2006).4 W h a t aretheeffectsofeducationonhealth?.InMeasuringtheeffectsofeducatio nonhealthandcivicengagement:Proceedingsof theCopenhagensymposium.

(1921).MeasurementofInequalityofIncomes.TheEconomicJournal,31(121),1 2 4 – 126.http://doi.org/10.2307/2223319

(2002).Income,income inequality andhealth:whatcanwelearnfromaggregatedata?.So cialscience&medicine,54(4), 577-589.

Groot,W., &vandenBrink,H M.(2006).4 A.What doeseducationdo to our health?.Gujarati,D.N.(2009).Basiceconometrics.TataMcGraw-HillEducation.

Guzman,J.A.(2008).TradeOpennessand Well- being:DoComplementaryConditionsMatter?.ProQuest.

(2015).T h e R i s e i n LifeE x p e c t a n c y andEconomicGrowthin the20thCentury.TheEconomicJournal,125(584),838-852.

Harrison,A ( 1 9 9 6 ) Opennessandgrowth:A time-series,c r o s s - c o u n t r y analysisf o r developing countries.Journal ofdevelopmentEconomics,48(2),419-447.

Herzer,D.(2014).Thelong- runrelationshipbetweentradeandpopulationhealth:evidencefromfivedecades.InAnn ualConference2014(Hamburg):Evidence- basedEconomicP o l i c y (No.100441).Vereinf ü r Socialpolitik/

Herzer,D.(2015).Thelong- runeffectoftradeonlifeexpectancyintheUnitedStates:anempiricalnote.AppliedEconomi csLetters,22(5),416-420.

(2015).Incomei n e q u a l i t y andhealth:Evidencef r o m developedanddevelopingcount ries.Economics,9(4),1.

Hoechle,D ( 2 0 0 7 ) R o b u s t standarderrorsf o r panelregressionsw i t h c r o s s - s e c t i o n a l dependence.Stata Journal,7(3),281-312.

Kawachi,I.,&Kennedy,B.P.(1997).Therelationshipofincomeinequality tomortality:doesthechoiceofindicatormatter?.Socialscience&medicine,45(7), 1121- 1127.

( 1 9 9 7 ) Socialcapital,i n c o m e inequality,andmortality.Americanjournalofpubliche alth,87(9),1491-1 4 9 8

(1996).Incomedistributionandmortality:c r o s s sectionalecologicals t u d y oft h e R o b i n H o o d indexi n theU n i t e d States.Bmj,312(7037),1004-1007.

(1905).Methodsofmeasuringtheconcentrationofwealth.Publicationsoft h e Americans tatistical association,9(70),209-219.

A , & W o l f s o n , M.(2004).Isincomeinequalityadeterminantofpopulationhealth? Part1.A systematicreview.Milbank Quarterly,82(1),5-99.

(2004).Wageinequality,thehealthsystem,andinfantmortalityi n wealthyindustrializedc ountries,1970–1996.Socialscience&medicine,58(2), 279-292.

(2001).Reexaminingtheevidenceofanecologicalassociationbetweeni n c o m e i n e q u a l i t y andh e a l t h Journalo f HealthP o l i t i c s , P o l i c y andL a w,26(3),487-522.

Newey,W.K , & West,K D (1987).Hypothesis testing with efficient methodo f momentsestimation.InternationalEconomic Review,777-787.

Park,H.M.(2011).Practicalguidestopaneldatamodeling:astep-by- stepanalysisusingstata.PublicManagement& P o l i c y AnalysisP r o g r a m , Internatio nalUniversityo f Japan.

( 1 9 7 9 ) Incomeandi n e q u a l i t y asdeterminantso f mortality:aninternationalcross- sectionanalysis.Populationstudies,33(2),343-351.

Shkolnikov,V M , Andreev,E M , Zhang,Z.,Oeppen,J , & Vaupel,J.W

StataCorp.2011.Stata 12 BaseReferenceManual.CollegeStation, TX:

StataPress.StataCorp.2011.StataStatistica l Software:Release12.CollegeStation, TX:

Stevens,P.,Urbach,J.,&Wills,G.(2013).Healthytrade:Therelationshipbetweenopentrade andhealth.ForeignTradeReview,48(1),125-135.

Stock,J.H.,&Watson,M.W.(2008).Heteroskedasticity‐robuststandarderrorsforfixed effectspaneldataregression.Econometrica,76(1),155-174.

TheWo rl d Bank,W o r l d DevelopmentIndicators(2015).Retrievedfromhttp:// data.worldbank.org/indicator/

Torre,R.,&Myrskylọ,M. (2014).Incomeinequalityandpopulationhealth:Ananalysiso f paneldatafor21 developedcountries,1975–2006.Population studies,68(1),1-13.

UNU-WIDER,‘WorldIncomeInequalityDatabase(WIID3c)’,September2015,https:// www.wider.unu.edu/project/wiid-world-income-inequality-database' vanBaal,P.,Obulqasim,P.,Brouwer,W.,Nusselder,W.,&Mackenbach,J.

(2013).Theinfluenceof health carespending on lifeexpectancy(No.35). vanDeurzen,I.,vanOorschot,W.,&vanIngen,E.

(2014).Thelinkbetweeninequalityandp o p u l a t i o n healthinl o w andmiddleincome countries:p o l i c y mytho r socialreality?.PloSone,9(12),e115109.

(2000).Incomeinequality andhealth:whatdoestheliteraturetellus?.Annualreviewof publichealth,21(1),543-567.

Wilkinson,R G (1992).Incomed i s t r i b u t i o n andl i f e expectancy.Bmj,304(6820),1 6 5 -

Wilkinson,R G.(2002).Unhealthy societies:theafflictions ofinequality.Routledge.

(2006).Incomei n e q u a l i t y andp o p u l a t i o n health:a review andexplanationoftheevidence.Socialscience&medicine,62(7), 1768-1784.

Ho notrejected Ho notrejected PooledOLS

Horejected(Fixedeffect) Ho notrejected FEmodel

Ho rejected Horejected Hausmanr ejectedoth erwise test:FE

Variable lifeexpc IMR Gini GDPpc Healthpc School

Enrol ment InvestmentRatio TradeOpenness lifeexpc 1.00

Romania RussianFederation Australia CzechRepublic KyrgyzRepublic Serbia

Belgium ElSalvador Luxembourg Spain lifeexpc imrgini gdppc healthpc invest~e trades~e lifeexpc imr gini gdppc healthpc investment~e

R-sq: within =0 6 2 5 2 Obsper g ro up : min = 4 between= 0 6 2 2 0 avg= 14.2 overall= 0 6 2 5 6 max= 16

F(4,618) = 257.71 corr(u_i,X b ) =- 0 3 9 3 1 Prob>F = 0.0000 lifeexpc Coef Std.E r r t P>|t| [95%C o n f I n t e r v a l ] gini -.0403224 0097228 -4.15 0.000 -.0594162 -.0212287 gdppc 0001713 0000167 10.24 0.000 0001385 0002042 healthpc 0006826 0000944 7.23 0.000 0004973 0008679 schoolenrolshare 0331182 0040622 8.15 0.000 0251408 0410955

_cons 67.75399 5418913 125.03 0.000 66.68982 68.81816 sigma_u 2.8138057 sigma_e 77871075 rho 92885992 (fraction ofv a r i a n c e d u e to u_i) Fte st t h a t a l l u _ i = 0 : F(46,6 1 8 ) = 142.83 Prob>F=0.0000

Var sd = sqrt(Var) lifeexpc e u 18.33578

R-sq: within =0.6246 Obspergroup:min= 4 between=0.6247 avg= 14.2 overall=0.6289 max= 16

Waldchi2(4) = 1091.45 corr(u_i,X) =0(assumed) Prob>chi2 = 0.0000 lifeexpc Coef Std.Err z P>|z| [95%Conf.Interval] gini -.0382807 0095479 -4.01 0.000 -.0569941 -.0195672 gdppc 0001527 0000153 9.97 0.000 0001227 0001827 healthpc 0007534 0000909 8.28 0.000 0005752 0009316 schoolenrolshare 0330098 0040272 8.20 0.000 0251166 0409029

_cons 67.99819 6574098 103.43 0.000 66.70969 69.28669 sigma_u 2.6485063 sigma_e 77871075 rho 92043139 (fraction ofvariancedue to u_i)

BreuschandPaganLagrangianmultipliertestforrandomeffects lifeexpc[id,t]=Xb+u[id]+e[id,t]

Test: Var(u)=0 chibar2(01) = 3235.45Prob>chibar2=

V_B))S E gini -.0403224 -.0382807 -.0020418 0018361 gdppc 0001713 0001527 0000186 6.72e-06 healthpc 0006826 0007534 -.0000708 0000252 schoolenro~e 0331182 0330098 0001084 0005321 b= c o n s i s t e n t u n d e r H o a n d H a ; o b t a i n e d f r o m x t r e g B = i n c o n s i s t e n t u n d e r H a , e f f i c i e n t u n d e r H o ; o b t a i n e d f r o m x t r e g Test: Ho: differencei n c o e f f i c i e n t s n o t s y s t e m a t i c chi2(4)= ( b - B ) ' [ ( V _ b - V _ B ) ^ ( - 1 ) ] ( b - B )

F(4,619) = 78.34 corr(u_i,X b ) = -0.5155 Prob>F = 0.0000 imr Coef Std.E r r t P>|t| [95%C o n f I n t e r v a l ] gini 1588347 0359751 4.42 0.000 0881867 2294827 gdppc -.0005857 0000622 -9.41 0.000 -.0007078 -.0004635 healthpc 0010187 0003509 2.90 0.004 0003295 0017078 schoolenrolshare -.1420931 0150997 -9.41 0.000 -.171746 -.1124403

7.6190296 2.895912 87376849 (fraction ofv a r i a n c e d u e to u_i) Fte st t h a t a l l u _ i = 0 : F(46,6 1 9 ) = 52.68 Prob>F=0.0000

Var sd = sqrt(Var) imr e u 106.0995

Waldc h i 2 ( 4 ) = 374.14 corr(u_i,X ) = 0( a s s u m e d ) Prob>c h i 2 = 0.0000 imr Coef Std.E r r z P>|z| [95%C o n f I n t e r v a l ] gini 1702108 0347011 4.91 0.000 1021979 2382237 gdppc -.0004425 0000511 -8.66 0.000 -.0005426 -.0003423 healthpc 0004814 0003261 1.48 0.140 -.0001578 0011206 schoolenrolshare -.1425683 014881 -9.58 0.000 -.1717346 -.1134021

BreuschandPaganLagrangianmultipliertestforrandomeffects imr[id,t]=Xb+u[id]+e[id,t]

Test: Var(u)=0 chibar2(01) = 1898.37Prob>chibar2=

Testsf o r t h e e r r o r c o m p o n e n t m o d e l : imr[id,t]= Xb +u [i d] +v [i d, t] v[id,t]= l a m b d a v [ i d , ( t - 1 ) ] + e [ i d , t ]

V_B))S.E gini 1588347 1702108 -.0113761 0094891 gdppc -.0005857 -.0004425 -.0001432 0000355 healthpc 0010187 0004814 0005373 0001296 schoolenro~e -.1420931 -.1425683 0004752 0025607 b=consistentunderHoandHa;obtainedfromxtregB =inconsis tentunderHa,efficientunderHo;obtainedfromxtreg

Random-effectsG L S r e g r e s s i o n Number of obs = 669

RobustS t d E r r z P>|z| [95%C o n f I n t e r v a l ] gini -.0382807 0233347 -1.64 0.101 -.0840159 0074546 gdppc 0001527 0000483 3.16 0.002 000058 0002474 healthpc 0007534 0002032 3.71 0.000 0003551 0011517 schoolenrolshare 0330098 0107689 3.07 0.002 0119031 0541164

_cons 67.99819 1.546485 43.97 0.000 64.96714 71.02925 sigma_u 2.6485063 sigma_e 77871075 rho 92043139 (fraction ofv a r i a n c e d u e to u_i)

RobustS t d E r r t P>|t| [95%C o n f I n t e r v a l ] gini 1588347 0941394 1.69 0.098 -.0306582 3483276 gdppc -.0005857 0001387 -4.22 0.000 -.0008649 -.0003065 healthpc 0010187 0003657 2.79 0.008 0002826 0017548 schoolenrolshare -.1420931 0436419 -3.26 0.002 -.2299398 -.0542465

RegressionwithDriscoll-Kraaystandard errors Number of obs = 670

Method:Fixed-effectsregression Number of groups = 47

Groupvariable(i):id F( 4, 46) = 124.01 maximumlag:2 Prob>F = 0.0000 withinR-squared = 0.3361 imr Coef

KraayStd.E rr t P>|t| [95%Conf.Interval] gini 1588347 0451758 3.52 0.001 0679006 2497688 gdppc -.0005857 0000293 -19.97 0.000 -.0006447 -.0005266 healthpc 0010187 0003188 3.20 0.003 0003769 0016604 schoolenrolshare -.1420931 0296809 -4.79 0.000 -.2018378 -.0823485

Fixed-effects(within)regression Numberofobs = 670

(Std.Err.adjustedfor47clustersinid) gdppc Coef

RobustSt d.Err t P>|t| [95%Conf.Interval] gini -24.38816 35.35491 -0.69 0.494 -95.5539 46.77758 healthpc 3.927062 4492532 8.74 0.000 3.022763 4.831362 schoolenrolshare 32.33735 18.42666 1.75 0.086 -4.753586 69.42829 investmentshare 147.3332 33.44948 4.40 0.000 80.00289 214.6635 tradeshare 59.47592 15.43631 3.85 0.000 28.40424 90.5476

_cons 6123.816 2423.913 2.53 0.015 1244.731 11002.9 sigma_u 11230.659 sigma_e 1582.6127 rho 98052851 (fraction ofvariancedue to u_i)

Waldchi2(6) = 117.97 corr(u_i,X) = 0(assumed) Prob>chi2 = 0.0000

(Std.Err.adjustedfor47clustersinid) lifeexpc Coef

RobustSt d.Err z P>|z| [95%Conf.Interval] gini -.0383525 0233636 -1.64 0.101 -.0841443 0074394 gdppc 0001384 0000426 3.25 0.001 000055 0002219 healthpc 0007828 0002109 3.71 0.000 0003696 0011961 schoolenrolshare 0328515 0107015 3.07 0.002 011877 0538261 investmentshare 0088031 019067 0.46 0.644 -.0285675 0461737 tradeshare 0030622 0090016 0.34 0.734 -.0145807 0207051 _cons 67.82052 1.541029 44.01 0.000 64.80016 70.84089 sigma_u 2.5086778 sigma_e 77864249 rho 91212953 (fraction ofvariancedue to u_i)

Fixed-effects(within) regression Numberofobs = 670

(Std.Err.adjustedfor47clustersinid) imr Coef

RobustSt d.Err t P>|t| [95%Conf.Interval] gini 1596326 0907058 1.76 0.085 -.0229487 342214 gdppc -.0003468 000142 -2.44 0.019 -.0006326 -.0000609 healthpc 0005134 0003772 1.36 0.180 -.0002459 0012728 schoolenrolshare -.1386419 043062 -3.22 0.002 -.2253212 -.0519627 investmentshare -.1366209 0545942 -2.50 0.016 -.2465133 -.0267285 tradeshare -.0459075 0324247 -1.42 0.164 -.1111751 0193601

_cons 34.1106 6.551349 5.21 0.000 20.92342 47.29778 sigma_u 6.7712854 sigma_e 2.8121308 rho 85289589 (fraction ofvariancedue to u_i)

G2SLS random-effectsI V regression Numbero f o b s = 669

R-sq: within = 0.6231 Obs per group: min = 4 between = 0.6157 avg = 14.2 overall = 0.6178 max = 16

Waldc h i 2 ( 4 ) = 1006.75 corr(u_i,X ) =0 ( a s s u m e d ) Prob> c h i 2 = 0.0000 lifeexpc Coef Std.E r r z P>|z| [95%C o n f I n t e r v a l ] gdppc 0002014 0000389 5.18 0.000 0001251 0002776 gini -.036168 0097694 -3.70 0.000 -.0553156 -.0170203 healthpc 0005209 0001958 2.66 0.008 0001371 0009047 schoolenrolshare 0303537 0045518 6.67 0.000 0214324 0392751

_cons 67.46121 7939808 84.97 0.000 65.90503 69.01738 sigma_u 3.0666311 sigma_e 78377151 rho 93868381 (fraction ofv a r i a n c e d u e to u_i) Instrumented: gdppc

G2SLS random-effectsI V regression Numbero f o b s = 669

R-sq: within = 0.6242 Obs per group: min = 4 between = 0.6177 avg = 14.2 overall = 0.6203 max = 16

Waldc h i 2 ( 4 ) = 1007.10 corr(u_i,X ) =0 ( a s s u m e d ) Prob> c h i 2 = 0.0000 lifeexpc Coef Std.E r r z P>|z| [95%C o n f I n t e r v a l ] gdppc 0001932 0000444 4.35 0.000 0001062 0002803 gini -.0382287 009621 -3.97 0.000 -.0570855 -.019372 healthpc 0005706 0002183 2.61 0.009 0001428 0009984 schoolenrolshare 0314538 0045775 6.87 0.000 0224821 0404256

_cons 67.53321 9497314 71.11 0.000 65.67178 69.39465 sigma_u 4.5367962 sigma_e 77948985 rho 97132606 (fraction ofv a r i a n c e d u e to u_i)

Source SS df MS Numbero f o b s = 669

Total 12248.3006 668 18.3357794 RootM S E = 2.5621 lifeexpc Coef Std.E r r t P>|t| [95%C o n f I n t e r v a l ] gini -.0607617 0145679 -4.17 0.000 -.0893664 -.032157 gdppc 0000513 0000141 3.63 0.000 0000235 0000791 healthpc 0012452 0001507 8.26 0.000 0009492 0015411 schoolenrolshare 0396939 0068733 5.78 0.000 026198 0531899

Source SS df MS Numbero f o b s = 670

Total 70980.5651 669 106.099499 RootM S E = 6.1939 imr Coef Std.E r r t P>|t| [95%C o n f I n t e r v a l ] gini 5142051 0351514 14.63 0.000 4451841 5832261 gdppc -.0001073 0000342 -3.14 0.002 -.0001744 -.0000402 healthpc -.0010086 0003643 -2.77 0.006 -.001724 -.0002932 schoolenrolshare -.1006145 0166159 -6.06 0.000 -.1332405 -.0679884

Waldchi2(7) = 169.17 corr(u_i,X) = 0(assumed) Prob>chi2 = 0.0000

(Std.Err.adjustedfor47clustersinid) lifeexpc Coef

RobustSt d.Err z P>|z| [95%Conf.Interval] gini -.0324822 0238124 -1.36 0.173 -.0791537 0141892 gdppc 0003326 000063 5.28 0.000 0002091 0004562 healthpc 0008217 0002197 3.74 0.000 0003911 0012524 schoolenrolshare 0236519 0098354 2.40 0.016 0043749 0429289 hi 1.373233 2.050601 0.67 0.503 -2.645871 5.392337 higdppc -.00021 000075 -2.80 0.005 -.000357 -.000063 higini 0114881 0425485 0.27 0.787 -.0719054 0948816 _cons 67.50106 1.23819 54.52 0.000 65.07425 69.92787 sigma_u 2.5823365 sigma_e 74843057 rho 92250954 (fraction ofvariancedue to u_i)

RobustS t d E r r t P>|t| [95%C o n f I n t e r v a l ] gini 1088636 0773407 1.41 0.166 -.0468152 2645424 gdppc -.0015878 000294 -5.40 0.000 -.0021795 -.000996 healthpc 0004255 0003462 1.23 0.225 -.0002714 0011223 schoolenrolshare -.0863804 0335111 -2.58 0.013 -.1538349 -.018926 hi 0 (omitted) higdppc 0012354 0003029 4.08 0.000 0006257 001845

Fixed-effects(within)regression Numberofobs = 670

(Std.Err.adjustedfor47clustersinid) imr Coef

RobustSt d.Err t P>|t| [95%Conf.Interval] gini 18973 1162311 1.63 0.109 -.0442311 4236912 gdppc -.000586 0001384 -4.23 0.000 -.0008645 -.0003074 healthpc 0010269 0003523 2.91 0.005 0003177 0017361 schoolenrolshare -.1394172 0441364 -3.16 0.003 -.2282592 -.0505752 hi 0 (omitted) higini -.1600483 1273683 -1.26 0.215 -.4164274 0963307

_cons 33.69399 6.008265 5.61 0.000 21.59998 45.788 sigma_u 8.3634263 sigma_e 2.8909924 rho 89326524 (fraction ofvariancedue to u_i)

G2SLS random-effectsIV regression Numberofobs = 669

R-sq: within = 0.6472 Obs per group: min = 4 between = 0.6062 avg = 14.2 overall = 0.6122 max = 16

Waldchi2(6) = 1092.74 corr(u_i,X) =0(assumed) Prob>chi2 = 0.0000 lifeexpc Coef Std.Err z P>|z| [95%Conf.Interval] gdppc 0004569 000098 4.66 0.000 0002648 0006489 gini -.0255298 0102326 -2.49 0.013 -.0455853 -.0054743 healthpc 0006567 0001371 4.79 0.000 0003881 0009254 schoolenrolshare 0172763 0065158 2.65 0.008 0045056 030047 hi 2.370363 3.88475 0.61 0.542 -5.243607 9.984333 higdppc -.0003062 0000836 -3.66 0.000 -.0004701 -.0001424 _cons 66.77758 2.914548 22.91 0.000 61.06517 72.48999 sigma_u 13.414869 sigma_e 75902815 rho 99680879 (fraction ofvariancedue to u_i) Instrumented: gdppc

G2SLS random-effectsIV regression Numberofobs = 669

R-sq: within = 0.6472 Obs per group: min = 4 between = 0.6338 avg = 14.2 overall = 0.6411 max = 16

Waldchi2(6) = 1092.95 corr(u_i,X) =0(assumed) Prob>chi2 = 0.0000 lifeexpc Coef Std.Err z P>|z| [95%Conf.Interval] gdppc 0002339 0000591 3.95 0.000 000118 0003498 gini -.0352637 0097259 -3.63 0.000 -.054326 -.0162014 healthpc 0009258 0001061 8.73 0.000 0007179 0011337 schoolenrolshare 0292089 0049781 5.87 0.000 019452 0389658 hi 1.019794 1.023294 1.00 0.319 -.9858264 3.025414 higdppc -.0001258 0000523 -2.40 0.016 -.0002283 -.0000232 _cons 67.93498 8524632 79.69 0.000 66.26419 69.60578 sigma_u sigma_e rho

2.852436 75856653 9339491 (fraction ofvariancedue to u_i) Instrumented: gdppc

G2SLS random-effectsIV regression Numberofobs = 669

R-sq: within = 0.6240 Obs per group: min = 4 between = 0.5892 avg = 14.2 overall = 0.5938 max = 16

Waldchi2(6) = 1013.63 corr(u_i,X) =0(assumed) Prob>chi2 = 0.0000 lifeexpc Coef Std.Err z P>|z| [95%Conf.Interval] gdppc 0001997 0000378 5.28 0.000 0001256 0002739 gini -.0424753 0107565 -3.95 0.000 -.0635577 -.0213929 healthpc 0005382 0001878 2.87 0.004 0001701 0009064 schoolenrolshare 0309435 0044326 6.98 0.000 0222558 0396312 hi -2.291822 1.332786 -1.72 0.086 -4.904035 3203915 higini 0203916 0240636 0.85 0.397 -.0267723 0675555 _cons 68.56537 8828464 77.66 0.000 66.83502 70.29571 sigma_u 3.1331971 sigma_e 78461682 rho 94099011 (fraction ofvariancedue to u_i) Instrumented: gdppc