Do childhood infections affect labour market outcomes in adulthood and, if so, how?

(1)

Short communication

Do childhood infections affect labour market outcomes in adulthood and, if so, how?

Jutta Viinikainen

^a,

*, Alex Bryson

^b

, Petri Böckerman

^c

, Marko Elovainio

^d

,

Nina Hutri-Kähönen

^e

, Markus Juonala

^f

, Terho Lehtimäki

^g

, Katja Pahkala

^h

, Suvi Rovio

ⁱ

, Laura Pulkki-Råback

^j

, Olli Raitakari

^k

, Jaakko Pehkonen

^l

aJyväskyläUniversitySchoolofBusinessandEconomics,P.O.Box35,FI-40014,Jyväskylä,Finland

bUniversityCollegeLondon,NIESR,London,UnitedKingdomandIZA,Bonn,Germany

cJyväskyläUniversitySchoolofBusinessandEconomics,UniversityofJyväskylä,Jyväskylä,Finland;LabourInstituteforEconomicResearch,Helsinki,Finland andIZA,Bonn,Germany

dDepartmentofPsychologyandLogopedics,FacultyofMedicine,UniversityofHelsinki,Helsinki,FinlandandNationalInstituteforHealthandWelfare, Helsinki,Finland

eDepartmentofPaediatrics,TampereUniversityHospital,Tampere,FinlandandFacultyofMedicineandHealthTechnology,TampereUniversity,Tampere, Finland

fDepartmentofMedicine,UniversityofTurku,Turku,Finland,DivisionofMedicine,TurkuUniversityHospital,Turku,Finland,MurdochChildren'sResearch Institute,Parkville,Victoria,Australia

gDepartmentofClinicalChemistry,FimlabLaboratoriesandFacultyofMedicineandHealthTechnology,FinnishCardiovascularResearchCenter,Tampere University,Tampere,Finland

hResearchCentreforAppliedandPreventiveCardiovascularMedicine,UniversityofTurku,Turku,Finland,Sports&ExerciseMedicineUnit,Departmentof PhysicalActivityandHealth,PaavoNurmiCentre,Turku,Finland

iResearchCenterofAppliedandPreventiveCardiovascularMedicine,UniversityofTurku,Turku,Finland

jDepartmentofPsychologyandLogopedics,FacultyofMedicine,UniversityofHelsinki,Helsinki,Finland

kResearchCentreofAppliedandPreventiveCardiovascularMedicine,UniversityofTurkuandDepartmentofClinicalPhysiologyandNuclearMedicine,Turku UniversityHospital,Turku,Finland

lJyväskyläUniversitySchoolofBusinessandEconomics,UniversityofJyväskylä,Jyväskylä,Finland

ARTICLE INFO Articlehistory:

Received21August2019

Receivedinrevisedform24January2020 Accepted29January2020

Availableonline30January2020 JELclassiﬁcation:

I1 I2 J01 J24 J3 Keywords:

Childhoodhealth

Infection-relatedhospitalization Education

Earnings Mediation

ABSTRACT

Aburgeoningbodyofliteraturesuggeststhatpoorchildhoodhealthleadstoadversehealthoutcomes, lowereducational attainment and weakerlabour marketoutcomes inadulthood. We focus onan importantbutunder-researchedtopic,whichistheroleplayedbyinfection-relatedhospitalization(IRH) inchildhoodanditslinkstolabourmarketoutcomeslaterinlife.Theparticipantsaged24–30yearsin 2001N=1706weredrawnfromtheYoungFinnsStudy,whichincludescomprehensiveregistrydataon IRHsinchildhoodatages0–18years.Thesedataarelinkedtolongitudinalregistryinformationonlabour marketoutcomes(2001–2012)andparentalbackground(1980).Theestimationswereperformedusing ordinaryleastsquares(OLS).TheresultsshowedthathavinganadditionalIRHisassociatedwithlower logearnings(b=-0.110,95%conﬁdenceinterval(CI):0.193;0.026),feweryearsofbeingemployed(b= 0.018,95%CI:0.031;0.005),ahigherprobabilityofreceivinganysocialincometransfers(b=0.012, 95%CI:0.002;0.026)andlargersocialincometransfers,conditionalonreceivingany(b=0.085,95%CI:

0.025;0.145).IRHsarenegativelylinkedtohumancapitalaccumulation,whichexplainsaconsiderable partoftheobservedassociationsbetweenIRHsandlabourmarketoutcomes.Wedidnotﬁndsupportfor thehypothesisthatadulthealthmediatesthelink.

* Correspondingauthorat:JyväskyläUniversitySchoolofBusinessandEconomics,POBox35,FI-40014UniversityofJyväskylä,Jyväskylä,Finland.

Phone:+350-40-5767804;Fax:+35814617194.

E-mailaddresses:jutta.viinikainen@jyu.fi(J.Viinikainen),a.bryson@ucl.ac.uk(A.Bryson),petri.bockerman@labour.fi(P.Böckerman),marko.elovainio@helsinki.fi (M.Elovainio),nina.hutri-kahonen@tuni.fi(N.Hutri-Kähönen),mataju@utu.fi(M.Juonala),terho.lehtimaki@tuni.fi(T.Lehtimäki),katpah@utu.fi(K.Pahkala),suvrov@utu.fi (S.Rovio),laura.pulkki-raback@helsinki.fi(L.Pulkki-Råback),olli.raitakari@utu.fi(O.Raitakari),jaakko.k.pehkonen@jyu.fi(J.Pehkonen).

http://dx.doi.org/10.1016/j.ehb.2020.100857

ContentslistsavailableatScienceDirect

Economics and Human Biology

j o u r n al h o m e p a g e : w w w . el s e v i e r . c o m / l o c a t e / e h b

(2)

1.Introduction

Childhood health haslasting effects on socioeconomic outcomes.Research hasestablishedthe importance of both initial healthendowmentandhealtheventsinchildhoodforsubsequent socioeconomicoutcomes.Bothstrandsoftheliteratureconclude that early health conditions are related toworse adult health, lowereducationalattainment,andlowerearnings(Douglasetal., 2018;Prinzetal.,2018).

Health disparitieshave early origins. Bothtwin studies and genome-wideassociationstudies(GWAS)showthatvariationin individualhealthispartlyexplainedbygeneticmakeup.GWASs have identified numerous single-nucleotide polymorphisms (SNPs)thatareassociatedwithhealth-relatedoutcomessuchas bodymassindex(BMI)(Lockeetal.,2015)andahigherincidence oftype2diabetes(Anderssonetal.,2013).Ameta-analysisbased ontwinstudiesalsoconcludedthatthereisasignificantgenetic component tohealth-related traits (Polderman et al., 2015). In addition to genetic inheritance, the in utero environment has consequences with regard to initial health endowment. It is challengingtodistinguishbetweentheeffectsofgeneticinheri- tanceandthoseofprenatalconditions,buttheresultsbasedon monozygotic twins with similar genetic makeup suggest that differencesininuteroconditionscausedifferencesinbirthweight, which is often used as a proxy for initial health endowment (Behrman and Rosenzweig, 2004). Further evidence for the importance of the effect of in utero conditions on subsequent healthcomesfromstudiesexploitingexogenousshockstofoetal health, such as the 1918 influenza pandemic (Almond, 2006;

Almond and Mazumder, 2005; Beach et al., 2018), the Dutch hunger winter (Roseboom et al., 2011), exposure to Ramadan (KuntoandMandemakers,2019;Majid,2015;Schoepsetal.,2018) andasuddenreductioninpollutionduetoplantclosuresduring pregnancy(ChayandGreenstone,2003;Parkeretal.,2008).

Studiesalsoshowthatchildhoodhealth mayhavelong-term effects onsocioeconomic outcomes in adulthood.For example, lowerself-reportedretrospective measures of childhood health andspeciﬁcchildhoodhealthconditionssuchasType1diabetes havebeenassociatedwithlowereducationalattainmentandlower earnings(Caseetal.,2005;Haasetal.,2011;Perssonetal.,2016;

Smith,2009).Towhatextent thesechildhood conditionsreﬂect initialhealthendowment,exogenousshockstochildhoodhealth or childhood environment are unclear because identiﬁcation strategieshavetypicallybeenunabletodistinguishamongthese competingexplanations.

Therearemanypotentialmechanismslinkingchildhoodhealth andsocioeconomicstatus inadulthood.First,worsehealthmay affecthumancapitalaccumulation,e.g.,duetoschoolabsences,or becausediseasemanagementorcomplicationsincreasethetime neededforhumancapitalaccumulation(i.e.,themarginalcostof educationalinvestmentsincreases). Earlyhealth conditionsmay alsoaffecttheabilitytolearn.Previousresearchhasdocumented, for example, that exposure to Ramadan fasting in utero has negativeeffectsoncognitivedevelopmentandschoolsuccesslater inlife(Majid,2015).Increaseduncertaintyaboutfuturehealthand labourmarketperformancemayalsoweakeneconomicincentives to invest in higher education (Lundborg et al., 2014). Second, childhoodhealthmayaffecthealthinadulthood,therebyaffecting occupationalchoices(Butlandetal.,2011),productivityandhours worked (Pelkowskiand Berger,2004; Vijan et al., 2004).Early infection-relatedhospitalizations(IRHs),forexample,havebeen linkedtometabolicoutcomes(Burgneretal.,2015a)andadverse cardiovascularphenotypes inadulthood(Burgneretal.,2015b), which mayaffectlabour market outcomes(e.g.,Cawley, 2015).

Third,those who have poorhealth may also be discriminated againstinthelabourmarket(Rooth,2011).

ThisstudyexaminesthelinksbetweenIRHsatages0–18years andlabourmarketoutcomesinadulthood.Ahospitalizationthatled to atleastoneovernightstaywas definedasinfection-relatedif either a primary or a secondary International Classification of Diseasescodeindicated aninfectiondiagnosis(Liu etal.,2016). Using rich, longitudinal,population-based data,we examinedto what extentlinksbetweenIRHsandadultlabourmarketoutcomesare mediatedthroughyearsofeducationandadulthealthandwhether any associations remain intact after adjusting for initial health endowment (the formal mediation analysis is explained in Section 2.6). A large body of literature has emphasized the importance of prenatal health endowment for socioeconomic outcomesinadulthood.Inthisstudy,wefocusonchildhoodand adolescentIRHs,whichhavegarneredmuchlessinterestintheprior literature.WefocusonIRHsforthreereasons.First,infectionsarea majorcauseofhospitalizationsamongchildren,thusconstitutinga significantpublic healthproblem (Gotoetal.,2016). Second,focusing onpost-prenatalhealthisimportantbecauseevenifprenatalhealth endowment affects variability in health, it is unlikely to be completelydeterministic.Tothatextent,itisworthinvestigating childhoodhealthinitsownright.Third,usinginformationonIRHs thatcoveragesfrom0to18,wecanexaminetherelativeimportance of early (ages 0–5) and later (ages 6–18) IRHs. This analysis adds to the earlier literature, which has stressed the importance of early childhoodhealth (Almond and Currie,2011)but has alsodocu- mentedthecrucialroleofadolescenthealthinadultlabourmarket outcomes(Lundborgetal.,2014).

2.Dataandmethods

Thelongitudinaldatawereobtainedfromthreesources:1)the YoungFinnsStudy(YFS)/theFinnishHospitalDischargeRegister (FHDR); 2) the Finnish Longitudinal Employer-Employee Data (FLEED)ofStatisticsFinland;and3)theLongitudinalPopulation Census(LPC)ofStatisticsFinland.

The YFS is anon-going epidemiological study examiningrisk factorsforcardiovasculardiseasesinadulthood.Thestudystartedin 1980whensubjectsinsixagecohortsbornin1962,1965,1968,1971, 1974,and1977wererandomlychosenfromﬁveuniversityhospital regionsinFinlandtoproduceanationallyrepresentativesampleof Finnishchildrenandadolescents.Theoriginalsamplesizewas3596 (Raitakari et al., 2008). The YFS contains information on IRHs at ages 0– 18,whichoriginatefromtheFHDR.TheFHDRisanationwidedatabase withcomprehensiverecordsforpatientsdischargedfromhospitals from1969onwards.Thedataincludeinformationonadmissionand dischargedaysanddiagnoses.Thecompletenessandaccuracyofthe dataareadequate,accordingtovalidationstudies(Sund,2012).Inthis study,wefocusedonthethreeyoungestagecohorts,forwhomwe haveaccesstoinformationoninfectionsfortheentireageperiodfrom 0to18years.Thesamplesizeinthebaselinemodelis1706.In1980, thesechildrenwerebetween3and9yearsold.

Toobtaininformationonparticipants’educationalattainment andlabourmarketoutcomes,theYFSwaslinkedtotheFLEED,and information onparentalbackground(education andearningsin 1980) was drawn from the LPC data. The matching between datasetswasbasedonuniquepersonalidentiﬁers,thusavoiding problemsrelatedtoerrorsinrecordlinkages(RidderandMofﬁtt, 2007).Theuseofadministrativedataonincomeandeducational attainmentalso eliminatestherisk of systematicmeasurement errorrelatedtoself-reportedmeasures.

AllparticipantsoftheYFSprovidedwritteninformedconsent, andthestudywasapprovedbytherelevantinstitutional ethics committees. Parents or guardians provided written informed consentonbehalfoftheunder-agedchildrenenrolledinthestudy.

TheﬁnalcombinedYFS-FLEED-LPCdatahavebeenapprovedfor researchpurposesbyStatisticsFinlandundertheethicsguidelines

(3)

oftheinstitution,whichcomplywithallnationalstandards.Aﬂow chartofthestudysubjectsisdocumentedintheSupplementary Appendix(Fig.A.1).

2.1.Outcomevariables

Weusefouroutcomevariables,whichweremeasuredoverthe period2001–2012:1)thelogarithm ofaveragewageandsalary earnings; 2) share of years employed; 3) indicator of having received social income transfers (1 = yes; 0 = no);and 4) the logarithmofsocialincometransfersconditionalonreceivingany.

The outcome variables were drawn from Statistics Finland’s registrydataFLEED.

2.2.Explanatoryvariables

TheexplanatoryvariableofinterestwasthenumberofIRHsat ages 0–18 years. A hospitalization that included at least one overnightstaywasdefinedasinfection-relatedifeitheraprimary or a secondary International Classification of Diseases code indicatedan infectiondiagnosis(Liu et al.,2016).Alldiagnoses were established by certified health care professionals. In robustnesschecks, wealsoused anindicatorofhaving atleast oneinfectionbetweenages0and18asanexplanatoryvariable.

2.3.Potentialmediators

Theﬁrst potentialmediator was yearsof education in 2001.

Informationonthehighestcompleteddegreewasdrawnfromthe FLEEDandwasconvertedintoyearsofeducationusingtheofﬁcial estimatesofStatisticsFinlandregardingthenumberofyearsneeded toobtainadegree.Asasecondpotentialmediator,weusedtheadult healthindex.Theindexconsistsofsevenobjectiveindicatorsofthe surveyrespondents’medicalstate(i.e.,biomarkers;BMI,waist-hip ratio (WHR),triglycerides, HDL and LDL cholesterol levels, and systolic and diastolic blood pressure), which were based on measurements and blood tests conducted in 2001 by medical professionalsatlocalhealthcentres.Anindicatorvariableforeach biomarker was constructed to identify those individualswhose biomarkervaluesexceededthenationallyrecommendedlevels(1= high risk; 0 = low risk; the cut-off values are documented in SupplementaryAppendixTableA.1).Theseindicatorvariableswere summed to create a comprehensive health measure index.

InformationonbiomarkerswasdrawnfromtheYFS.

2.4.Controlvariables

All models wereadjusted forindicators of thesexand birth cohort.Becausedifferencesinchildhoodhealthmayreﬂectdiffer- encesinfamilybackground(MilaniakandJaffee,2019),wealsoused thelogarithmoffamilyincomein1980andparentaleducationin 1980asadditionalcontrolvariables.Theindicatorvariableforhigh parentaleducationlevelequalledoneifatleastoneoftheparents hadobtainedsome universityeducation.Thesecontrol variables werebasedonStatisticsFinlandregistrydataFLEEDandLPC.

Previous studieshave emphasized therole played by initial health endowment in subsequent health outcomes (e.g.,Beach et al., 2018; Behrman and Rosenzweig, 2004; Majid,2015). To explorehowdifferencesininitialhealthendowmentaffectedour results,intherobustnesschecks,weadjustedthebaselinemodels forbirthweight(inkg),whichisacommonlyusedproxyforinitial healthendowment(BehrmanandRosenzweig,2004),andgenetic riskscoresforBMI(basedon32SNPs(Speliotesetal.,2010)),WHR (14SNPs(Heidetal.,2010)),triglycerides(41SNPs(Hernesniemi etal.,2015)),HDLcholesterol(38SNPs(Teslovichetal.,2010)),LDL cholesterol(58SNPs(Hernesniemietal.,2015))andbloodpressure

(29SNPs(Ehretetal.,2011)).Asanadditionalcontrolvariable,we also used a genetic risk score for years of education(74 SNPs (Okbayetal.,2016)).ThesedataweredrawnfromtheYFS.

2.5.Additionalvariables

Inthedescriptiveanalyses,wecomparedchildhoodhealthand educationaloutcomesstratiﬁedbyinfectionstatusinthefollowing dimensions: the number of physician-diagnosed childhood chronicconditions,participationinremedialeducation(in1980 and 1983)and schoolsuccess,indicated bygrade pointaverage (GPA)intheninthgrade.ThisinformationwasdrawnfromtheYFS.

2.6.Analysis

Two-samplet-testswereutilizedtocomparethecharacteristicsof those whoexperiencedatleastoneIRH withthose whodidnot experienceanyinfections.Theestimationsofadultoutcomeswere performed using ordinary least squares (OLS). To test potential pathwaysthatmaymediatetheassociationbetweenIRHandadult labourmarketoutcomes,weperformedaSobel(1982)mediation analysis.Thistestexamineswhethertheindirectassociationbetween theindependentvariable(X,i.e.,IRH)andthedependentvariables (Yk,wherekreferstoalabourmarketoutcome)throughpotential mediators(M_j,wherejreferstoyearsofeducationortheadult healthriskindex)issigniﬁcantlydifferentfromzero.Fig.1depicts themediationmodel,whichconsistsofthreeregressionequations

Yki¼

u

1þ

g

Xiþ

e

¹ⁱ ð1Þ

Mji¼

u

2þ

a

Xiþ

e

²ⁱ ð2Þ

Yki ¼ 

u

3þ

m

Xiþ

b

Mjiþ

e

³ⁱ ð3Þ whereireferstoanindividual,

g

referstothegrossassociation betweenIRHsandadultlabourmarketoutcomes(Fig.1,PanelA, pathc), αisthelinkbetweenIRHsand thepotentialmediators (Fig.1,PanelB,patha),βisthelinkbetweenthemediatorandthe labour marketoutcomes (Fig.1,Panel B, path b),and

m

is the remaining linkbetween IRHs and the labourmarket outcomes aftercontrollingforthemediator(pathc’).Themediatedlinkis calculatedastheproductofthecoefﬁcientsαandβ(i.e.,αβ),and thetotalproportionofthelinkthatismediatedthroughmediatorj isequaltoð

ab

_Þ=

g

.

3.Results

3.1.Descriptivestatistics

Twenty-ﬁvepercentofindividualshadexperiencedatleastone IRH (either primaryor a secondary infection diagnosis) during childhood,ofwhom61%hadoneinfection,23%hadtwoinfections and 16 % had at least three infections. Table 1 compares the observablecharacteristicsofparticipantswhohadexperiencedat least one IRH tothose who didnot experience any infections.

AmongthosewhohadexperiencedatleastoneIRH,theshareof menwashigher(p<0.01),theaverageagewaslower(p<0.01), andtheshareofchildrenwithhighlyeducatedparentswassmaller (p<0.05).Otherwise,theresultsofthebivariateanalyseswithout controls did not indicate the presence of differences in family background,initialhealthendowment(birthweight,GRSs)orthe numberofchronicconditionsinchildhood(ineachcase,p>0.10).

Table 2 presents labour market outcomes, educational out- comesandadulthealthcharacteristicsbasedonIRHstatus.The

(4)

bivariateanalysisrevealedthatthosewhoexperiencedIRHsduring childhoodweremorelikelytoparticipateinremedialeducation (p<0.01),hadlowerGPAattheninthgrade,hadfeweryearsof educationinadulthood(p<0.01)andwerelesslikelytocomplete higheducation(p<0.01).Participants withchildhood IRHsalso hadlowerearningsandalowershareofyearsemployed(p<0.01).

Ifwecompareadulthealthbetweenthetwogroups,thosewho experiencedat least one IRHhad lowerHDL cholesterollevels

(p<0.1),higherBMIandWHR(p<0.01)andhighersystolicblood pressurein2001(p<0.1).

3.2.IRHandlabourmarketoutcomes

Table3showstheresultsfromfourmediationmodelswithfour differentoutcomevariables(logofaverageearnings,shareofyears employed,indicatorforobtaininganysocialincometransfersand Table1

DescriptivestatisticsbyIRHstatus.

All Mean(SD)

AtleastoneIRH Mean(SD)

NoIRHs Mean(SD)

Difference t-statistics N

NumberofIRHs(0–18years) 0.438 (1.075)

1.779 (1.522)

0 (0)

1.779 41.936*** 1706

Female(%) 0.505

(0.500)

0.440 (0.497)

0.526 (0.499)

0.086 3.066*** 1706

Agein2001 27.070

(2.469)

26.743 (2.474)

27.177 (2.459)

0.434 3.139*** 1706

Highlevelofparental education(%)(1980)

0.155 (0.362)

0.126 (0.332)

0.164 (0.370)

0.038 1.970** 1706

Parentalincome(1980) 12827.27

(7466.35)

12430.26 (7207.67)

12956.92 (7547.13)

526.66 1.255 1706

Initialendowment

Birthweight 3.516

(0.538)

3.492 (0.565)

3.524 (0.529)

0.032 1.033 1578

BodymassGRS 29.055

(3.327)

29.132 (3.295)

29.031 (3.339)

0.100 0.422 1090

Waist-hipratioGRS 15.170

(2.423)

15.095 (2.425)

15.193 (2.424)

0.097 0.562 1090

TriglycerideGRS 0.985

(0.094)

0.987 (0.099)

0.984 (0.092)

0.003 0.394 1090

HDLcholesterolGRS 44.698

(3.671)

44.900 (3.815)

44.636 (3.626)

0.264 1.007 1090

LDLcholesterolGRS 0.960

(0.078)

0.954 (0.082)

0.961 (0.076)

0.007 1.312 1090

BloodpressureGRS 30.376

(3.122)

30.378 (3.099)

30.376 (3.130)

0.003 0.011 1090

Childhoodhealth(0–18years)

Numberofchronicconditions 0.104 (0.382)

0.110 (0.412)

0.102 (0.372)

0.008 0.339 1706

Note:Tablereportsmeansandstandarddeviations(SD)inparentheses.Theunitofanalysisistheindividual.Differencesbetweengroupsweretestedusingatwo-samplet- test.Thethreecohortsstudied(aged24,27,and30in2001)weredrawnfromtheYFS.Theindicatorforhighparentaleducationequalledoneifatleastoneoftheparentshad obtainedsomeuniversityeducation(basedontheLPCdatafrom1980).IRHreferstoinfection-relatedhospitalizationandGRStothegeneticriskscore.Statisticalsigniﬁcance:

*p<0.1,**p<0.05,***p<0.01.

Fig.1.Mediationmodel.

(5)

Table2

DescriptivestatisticsandcomparisonoflabourmarketandeducationaloutcomesandadulthealthbyIRHstatus.

All Mean(SD)

AtleastoneIRH Mean(SD)

NoIRHs Mean(SD)

Difference t-statistics N Labourmarketoutcomes

Logofaverageearnings,2001–2012 9.397 (1.904)

9.158 (2.269)

9.475 (1.762)

0.317 2.618*** 1706

Shareofyearsemployed,2001–2012 0.788 (0.293)

0.747 (0.324)

0.801 (0.281)

0.054 3.075*** 1706

Indicatorforsocialincometransfers,2001-2012 0.889 (0.314)

0.898 (0.304)

0.886 (0.317)

0.011 0.632 1706

Logofaveragesocialincometransfers,2001–2012 7.056 (1.416)

7.109 (1.446)

7.039 (1.406)

0.071 0.839 1517

Educationaloutcomes

Education,years 12.800

(2.433)

12.424 (2.296)

12.922 (2.464)

0.498 3.659*** 1706

Higheducationlevel 0.192

(0.394)

0.148 (0.355)

0.206 (0.405)

0.058 2.826*** 1706

Indicatorofremedialeducation 0.253

(0.435)

0.328 (0.471)

0.243 (0.429)

0.086 2.140** 1070

Gradepointaverageinninthgrade 7.950 (0.936)

7.737 (0.987)

7.978 (0.926)

0.241 2.481** 906

Adulthealth

Bodymassindex2001 24.456

(4.335)

25.275 (4.928)

24.204 (4.106)

1.071 3.000*** 982

Waist-hipratioin2011 0.823

(0.077)

0.838 (0.076)

0.818 (0.076)

0.020 3.509*** 982

Triglyceridesin2001 1.255

(0.618)

1.307 (0.644)

1.240 (0.609)

0.067 1.409 982

HDLcholesterolin2001 1.284

(0.317)

1.253 (0.327)

1.294 (0.313)

0.040 1.682* 982

LDLcholesterolin2001 3.097

(0.775)

3.126 (0.782)

3.088 (0.774)

0.038 0.646 982

Diastolicbloodpressurein2001 71.249

(7.802)

71.626 (7.571)

71.133 (7.873)

0.492 0.839 982

Systolicbloodpressurein2001 120.930

(13.088)

122.342 (13.463)

120.495 (12.948)

1.847 1.878* 982

Healthriskindexin2001 1.435

(1.256)

1.606 (1.330)

1.382 (1.228)

0.224 2.375** 982

Note:Infectionsweremeasuredbetween0and18yearsofageapartfromtheresultsconcerningparticipationinremedialeducationandGPAinninthgradeinwhichcasethe numberofinfectionswasmeasuredbetweenages0and5.SeeTable1notesforadditionaldetails.

Table3

Mediationbyeducationalattainment.

Outcomevariable(Y) (1) (2) (3) (4) (5)

Relationbetweennumberof IRHsandlabourmarket outcomes(pathc)

Relationbetween numberofIRHsandyears ofeducation(patha)

Relationbetweenyearsof educationandlabourmarket outcomes(pathb)

MediatedlinkbetweenIRHs andtheoutcomevariable throughyearsofeducation (Sobeltest)

Proportionoftotal connectionthatis mediated Model1:

Y=Logofaverage earnings,2001-2012 (n=1706)

0.110**

[0.193;0.026] 0.169***

[0.273;0.066]

0.210***

[0.173;0.247]

0.036*** 0.323

Model2:

Y=Shareofyears employed,2001-2012 (n=1706)

0.018***

[0.031;0.005]

0.169***

[0.273;0.066]

0.032***

[0.026;0.037]

0.005*** 0.294

Model3:

Y=Indicatorforsocial incometransfers, 2001-2012(n=1706)

0.012*

[0.002;0.026]

0.169***

[0.273;0.066]

0.008**

[0.015;0.002]

0.001** 0.115

Model4:

Y=Logofaverage socialincome transfers,2001-2012 (n=1517)

0.085***

[0.025;0.145]

0.165***

[0.271;0.059]

0.076***

[-0.104;0.047]

0.012*** 0.147

Note:Resultsarebasedonthelinearregression(OLS)method.Eachrowreferstooneofthefourestimatedmediationmodels.Theunitofanalysisistheindividual.Thepaths areexplainedinFig.1.Thethreecohortsunderstudy(aged24,27,and30in2001)weredrawnfromtheYFS.Allmodelsincludedcontrolsforthebirthyear,sex,parental educationlevel(1980),andparentalearnings(1980).The95%conﬁdenceintervalsfortheparameterestimatesarereportedinparenthesis.Statisticalsigniﬁcance:*p<0.1,**

p<0.05,***p<0.01.

(6)

logofaveragesocialincometransfers).Theresultspresentedin column1showtheoverall(gross)associationbetweenIRHsand eachoftheselabourmarketoutcomes(pathc,Fig.1).Ahigher numberofIRHswasassociatedwithlowerearnings,alowershare of years employed,a higher probabilityof receiving any social incometransfers(i.e.,theextensive margin ofadjustment) and highersocialincometransfersifreceivingany(i.e.,theintensive marginofadjustment).Havinganadditionalinfectioninchildhood was associated with 11.0 % lower earnings (b = 0.110, 95 % conﬁdence interval (CI): 0.193; 0.026; p<0.010), a 1.8 percentagepointlowershareofyearsemployed(b=0.018,95

%CI:0.031;0.005, p<0.006), a 1.2percentage point higher probabilityofreceivinganysocialincometransfers(b=0.012,95% CI:0.002;0.026,p<0.087)andan8.5%increaseinthesizeof socialincometransfersconditionalonreceivingany(b=0.085,95

%CI:0.025;0.145,p<0.006).

3.3.Potentialpathways

The resultsinTable3 alsoshowthemediationresultsusing educationasapotentialpathwaybywhichIRHsmayaffectadult labour market outcomes. Column 2 presents the relationship betweenIRHsandyearsofeducation(patha,Fig.1),whilecolumn 3presentstherelationshipbetweenyearsofeducationandlabour marketoutcomes(pathb,Fig.1).AhighernumberofIRHs was relatedtofeweryearsofeducation(Column2,p<0.003),whereas ahighernumberofyearsofeducationwasassociatedwithbetter labourmarketoutcomes(Column3,p<0.02).BasedontheSobel mediationtest(Column4),theindirectassociationbetweenIRHs and labour market outcomes mediated through education was statisticallysigniﬁcant(p<0.048),andtheproportionofthetotal associationmediatedwas32%intheearningsequation,29%inthe shareofyearsemployed,12%intheprobabilityofreceivingsocial incometransfersand15%intheamountofsocialincometransfers conditionalonreceivingany(Column5).

Table4showsthemediationresultsthatwereobtainedusing adulthealthasapotentialmediator.Basedontheresults(Column 3), adult health was a signiﬁcant predictor of labour market outcomes. A higher number of health risks in adulthood was related to lowerearnings (b = -0.139, 95 % CI: -0.208; -0.069,

p<0.000)andalowershareofyearsemployed(b=-0.016,95%CI:

-0.028;-0.003,p<0.014).Thepointestimatesalsosuggestthata higher number of health risks was associated with a higher probabilityofreceivingsocialincometransfers(b=0.009,95%CI:

-0.007;0.026,p<0.254)andlargertransfersreceived(b=0.057,95

%CI:-0.010;0.125,p<0.095).However,therelationshipbetween childhoodIRHsandthenumberofadulthealthrisks(Column2) wasweak:oneadditionalIRHwasrelatedtoan0.02increasein thenumberofhealth risks(p>0.6).Thus, IRHsdo notseemto predicttheadulthealthriskindex.Consequently,theproportionof thetotalassociationthatwas mediatedthroughthehealthrisk index was low (1–3 %), and according to the Sobel test, this mediationpathwaywasnon-signiﬁcant(p>0.6).

3.4.Robustnesschecks

Ourresultsimplythateducationexplainsaconsiderablepartof thelinkbetweenIRHsandlabourmarketoutcomes,whiletherole of adult health as a mediator is small and statistically non- significant.However,missinginformationintheadulthealthrisk indexreducedourestimationsample inTable 4(i.e.,mediation throughadulthealth),whichlimitsthecomparabilitybetweenthe results using the two different mediators. To explore whether varyingsamplesizehasimplicationsforourfindings,weestimated themodelsinTable3usingthesamesampleasinTable4.Basedon theresults(TableA.2),thesamplesizewasnotamajordriverofour findings, and education remaineda significant mediator in the reducedsample.

We alsoperformed several robustness checks toreduce the possibilitythatourresultsweredrivenbyconfoundersthataffect bothIRHsandlabourmarketoutcomes.First,weaugmentedthe baselinemodelswiththemeasuresofinitialhealthendowment i.e.,birthweightandGRSs(Table5).Thesecontrolvariableshad onlya modestimpactonthe pointestimates thatdescribe the connectionbetweenIRHsandlabourmarketoutcomes.Thus,the observedinitialhealthendowmentdoesnotseemtobeamajor driverofourresults.Thispatternisinaccordancewiththeﬁndings based on bivariate analyses that showed that there were no statisticallysigniﬁcantdifferencesininitialendowmentbetween thosewhohadatleastoneIRHandthosewhohadnone(Table1).

Table4

Mediationbyadulthealth.

Relationbetweennumberof IRHsandlabourmarket outcomes(pathc)

Relationbetween numberofIRHsand health(patha)

Relationbetweenhealth andlabourmarket outcomes(pathb)

MediatedlinkbetweenIRHs andtheoutcomevariable throughhealth (Sobeltest)

Proportionoftotal connectionthatis mediated

Model1:

Y=Logofaverage earnings,2001-2012 (n=982)

0.068*

[0.145;0.010]

0.016 [0.053;0.086]

0.139***

[0.208;0.069]

0.002 0.033

Model2:

Y=Shareofyears employed,2001-2012 (n=982)

0.019**

[0.033;0.006]

0.016 [0.053;0.086]

0.016**

[0.028;0.003] 0.000 0.013

Model3:

Y=Indicatorforsocial incometransfers,2001- 2012(n=982)

0.009 [0.009;0.027]

0.016 [0.053;0.086]

0.009 [0.007;0.026]

0.000 0.017

Model4:

Y=Logofaveragesocial incometransfers,2001- 2012(n=869)

0.068*

[0.004;0.141]

0.013 [0.058;0.085]

0.057*

[0.010;0.125]

0.001 0.011

Note:Resultsarebasedonlinearregression(OLS)method.Eachrowreferstooneofthefourestimatedmediationmodels.Theunitofanalysisistheindividual.Thepathsare explainedinFig.1.Thethreecohortsunderstudy(aged24,27,and30inyear2001)aredrawnfromtheYFS.Allmodelsincludedcontrolsforthebirthyear,sex,parental education(1980),andparentalearnings(1980).The95%conﬁdenceintervalsfortheparameterestimatesarereportedinparenthesis.Statisticalsigniﬁcance:*p<0.1,**

p<0.05,***p<0.01.

(7)

Second,theresultsremainedintactwhenthemediationmodel wasaugmentedwithageneticriskscoreforeducation(Supple- mentary Appendix Table A.3). Finally, we used Oster’s (2019) methodtoevaluatetherobustnessofourresultsregardingomitted variablebias. The results inTable 6 suggest that unobservable factorsmayconfoundthelinksbetweenIRHsandeducationyears, andbetweeneducationyearsandsocialincometransfer.However, theseresultsarehighlyconservativeassettingRmax=1impliesthe outcomevariablewouldbefullyexplainedbythetreatmentand fullsetofcontrols.Otherwise,theresultsbasedonOster’s(2019) methodsuggestthatthesignsofourestimatesindicatingthelinks between years of education and labour market outcomes are robusttosubstantialselectiononunobservables.

Theresultsbasedonthemodelwitheducationasamediator remained robust if we replaced the number of IRHs with an indicatorvariableforthosewhohadexperiencedatleastoneIRH inchildhood(SupplementaryAppendixTableA.4).Inaddition,we estimatedseparatemediationmodelsusingearly(ages0–5)and

later (ages 6–18) IRHs as the explanatory variable of interest (SupplementaryAppendixTablesA.5andA.6).Theseresultswere inaccordancewithourmainﬁndingspresentedinTable3.Thus, thelinksbetweenIRHsandlabourmarketoutcomesaresimilar during early childhood andlater childhood. We consideredthe possibilitythatthis resultreﬂectsthecorrelationbetweenearly and later IRHs.However, this didnotseem tobethecase:the correlationbetweenthenumberofearlyandlaterIRHswasweak (r=0.14),andthelinkbetweenlaterIRHs,educationandlabour marketoutcomes remainedintact ifthemodelwas augmented withearlyIRHs(SupplementaryAppendixTableA.7).

4.Discussion

Usinglinkedlongitudinalregistrydata,wefoundthatchildren who experienced IRHs in childhood had weakerlabour market outcomesintermsofearningsandemploymentthanchildrenwho had notexperiencedIRHs in childhood.Theyalsohad a higher Table5

IRHsinchildhoodandlabourmarketoutcomes.Baselineresultsandresultswithcontrolsforinitialhealthendowmentsbasedonthesamesamplesize.

Relationbetween numberofIRHsand labourmarket outcomes(pathc)

Relationbetweennumber ofIRHsandyearsof education(patha)

Relationbetweenyearsof educationandlabour marketoutcomes(pathb)

Mediatedlink betweenIRHsandthe outcomevariable throughyearsof education (Sobeltest)

Proportionof total connection thatis mediated

Model1:

Y=Logofaverageearnings,2001-2012 (n=1040)

0.035 [0.112;

0.042]

0.036 [0.113;

0.041]

0.204***

[0.342;

0.067]

0.208***

[0.345;

0.070]

0.138***

[0.105;

0.171]

0.137***

[0.104;

0.170]

0.028 0.028*** 0.802 0.790

Controlsforinitialhealthendowments No Yes No Yes No Yes No Yes No Yes

Model2:

Y=Shareofyearsemployed,2001-2012 (n=1040)

0.011 [0.024;

0.003]

0.011 [0.025;

0.003]

0.204***

[0.342;

-0.067]

0.208***

[0.345;

0.070]

0.020***

[0.014;

0.026]

0.020***

[0.014;

0.026]

0.004*** 0.004*** 0.388 0.365

Model3:

Y=Indicatorforsocialincometransfers, 2001-2012(n=1040)

0.012 [-0.006;

0.031]

0.012 [0.006;

0.031]

0.204***

[0.342;

0.067]

0.208***

[0.345;

0.070]

0.004 [0.012;

0.004]

0.004 [0.012;

0.004]

0.001 0.001 0.064 0.069

Model4:

Y=Logofaveragesocialincometransfers, 2001-2012(n=922)

0.077**

[0.003;

0.152]

0.080**

[0.006;

0.155]

0.199***

[0.341;

0.057

0.201***

[0.344;

-0.059]

0.047***

[0.080;

0.013]

0.045***

[0.079;

0.012]

0.009 0.009* 0.120 0.114

Note:Initialhealthendowmentsincludecontrolsforbirthweight,BMIGRS,WHRGRS,triglycerideGRS,HDLGRS,LDLGRSandbloodpressureGRS.Resultsarebasedonlinear regression(OLS)method.Eachrowreferstooneofthefourestimatedmodels.Theunitofanalysisistheindividual.ThepathsareexplainedinFig.1.Thethreecohortsunder study(aged24,27,and30inyear2001)aredrawnfromtheYFS.Allmodelsincludecontrolsforthebirthyear,sex,parentaleducation(1980),andparentalearnings(1980).

The95%conﬁdenceintervalsfortheparameterestimatesarereportedinparenthesis.Statisticalsigniﬁcance:*p<0.1,**p<0.05,***p<0.01.

Table6

Robustnesstoomittedvariablebias(Oster’smethod).

Yearsofeducation Logofaverageearnings Shareofemploymentyears Indicatorforsocialincome transfers

Logofaveragesocialincome transfers

(1) (2) (3) (4) (5) (6) (7) (8) (9) (10)

Treatment variable

dfor β=0given Rmax

Identiﬁedset givend=1and Rmax

dfor β=0given Rmax

Numberof IRHs

0.236 0.169,0.591 .. .. .. .. .. .. .. ..

Yearsof education

.. .. 0.575 0.210,4.935 0.607 0.032,0.742 0.083 0.008,2.335 0.429 0.369,-0.076

Rmax 1.0 1.0 1.0 1.0 1.0

Note:ResultsarecomputedusingOster’s(2013)Statamodulepsacalc.BaselinemodelsarebasedonOLSandtheycontrolforthebirthyear,sex,parentaleducation(1980),and parentalearnings(1980).Basedontheresults,theunobservableshouldbe0.236(Column1),0.575(Column3),0.607(Column5),0.083(Column7),0.429(Column9)timesas importantastheobservablesinordertoproduceazerotreatmenteffect(i.e.β=0).Alternatively,Oster’smethodcanbeusedtoestimateboundsforthetreatmenteffects assumingthatunobservablesareasimportantasobservables(d⁼¹⁾âs^suggested^byÂltonjiêtâl.^(2005).^Basedôn^these^results,^we^cannot^reject^the^hypothesis^that^the^links betweenIRHandyearsofeducation(Column2)aswellasbetweeneducationyearsandsocialincometransfersindicator(Column8)arezero.Otherwise,theresultsbasedon Oster’s(2019)methodsuggestthatthesignsofourestimatesarerobusttosubstantialselectiononunobservables(Columns4,6,10).