A.E. Clark
ParisSchoolofEconomics(PSE),CNRS,48,boulevardJourdan,75014Paris,France Received10April2013;accepted30April2013
Abstract
Healthandwell-beingaresociallydetermined.Oneofthewaysinwhichthiscomesaboutisviasocialcomparisonswithotherindividualsinthe samepersonal,geographicorsocialnetworks,withthecomparisonsreferringeithertoincomeorotheraspectsofeconomicandsociallife.The existenceofsuchcomparisoneffectswithrespecttoincomemayhelptoexplainthesocialgradientinhealth.
#2013PublishedbyElsevierMassonSAS.
Keywords:Well-being;Comparisons;Income;Unemployment;Divorce;Religion;Socialhealthgradient Re´sume´
Lasante´ etlebien-eˆtresontde´termine´ssocialement:unedesraisonsdecettede´terminationestl’existencedescomparaisonssocialesavecles autresindividusappartenantaumeˆmegroupefamilial,ge´ographiqueousocial,soitparrapportaurevenu,soitparrapporta` d’autresaspectsdela viee´conomiqueetsociale.L’hypothe`sepre´sente´edanscetarticleestquecephe´nome`nedecomparaisonpeutaidera` comprendrelegradientsocial desante´.
#2013Publie´ parElsevierMassonSAS.
Motscle´s: Bien-eˆtre;Comparaisons;Revenu;Choˆmage;Divorce;Religion;Gradientsocialdesante´
1. Introduction
There are a number of parallels between health and subjectivewell-being,seenthroughthelens ofsomeonewho carriesoutmicro-econometricresearch.Inparticular,bothare veryoften measured on an ordinal scale and are subjective.
Thereisanobvioussimilaritybetweentheclassicself-assessed health question (in the British Household Panel Survey:
‘‘Wouldyousaythatyourhealthhasonthewholebeen...?’’, withthe possibleresponses: Excellent,Good,Fair, Poor,and Very poor) and subjective well-being (‘‘How dissatisfied or satisfiedareyouwithyourlifeoverall’’,withsevenresponses onan ordered scale).Thatthis similaritydid not leadto the analysisofwell-beingbeingsoeasilyacceptedasthatofhealth isasubject potentiallyworthy ofreflection.
Anyteethingproblemsinvolvedintheempiricalanalysisof well-being inthesocial sciencesseemtohavedied downby now. Here, I will take life satisfaction and so on as useful summary measures of how well overall the individual is functioning. Some of the research providing validating evidenceforthishypothesisismentionedinSection4of[1].
It is actually rather reassuring that the type of society is correlatedwithbothourhealthandwell-being:surelythebig surprisewouldhavebeenweretheynottobeso.Whichisnotto saythatweunderstandperfectlywhyitisthatsomesocieties performdifferentlyfromothersinthisrespect.Outcomesmay differ due tosocieties’ different characteristics (their health-caresystems,environmentalqualityandsoon);orbecauseof thetypeofpeoplewholivethere(inthesensethatwhatother people domayaffect my healthor well-being).In this latter case,therearespilloversorexternalitiesbetweenindividuals, anditisthischannelthatIwillreflectonhere.Ithereforeask:
Doyou makeme sick,ordoyoumakemehappy?
www.sciencedirect.com
Revued’E´ pide´miologieetdeSante´ Publique61S(2013)S184–S188
E-mailaddress:Andrew.Clark@ens.fr.
0398-7620/$–seefrontmatter#2013PublishedbyElsevierMassonSAS.
http://dx.doi.org/10.1016/j.respe.2013.04.003
Suchspilloversareaveryinterestingareaofresearch.Itis howevereasytocometothewrong conclusions.Individuals’
health/well-beingmaybecorrelated,butthatdoesnotshowthat theyinfluenceeachother.As notedabove, individualswithin thesamesociety(orthesameregion,townorevenroad)very likely share some of the same health environment (prices, health supply, exposure to environmental and health risk factors).Equally,it maybethe casethatindividuals whoare similartoeachotherchoosetoliveinacertainarea(birdsofa featherflocktogether).
In both of the latter cases, the causal link is not a direct relationshipfrommyownbehaviourtoothers’behaviour.We wanttoknowwhetheranexogenouschangetooneindividual’s behaviouraffectsthehealthandwell-beingofothersinhispeer group.Thisisimportant toknowfor policy purposes.Will a policythataffectsoneindividual’sbehaviouralsospilloverto thebehaviourofanotherindividual(ifIamabletomakeAstop smoking, whichmightthen encourageBtostop,whichthen leadsCtostop,whichthen...)?Equally,fromasocialwelfare pointofview,isthereariskthatmakingAbetteroffmightalso makeBworseoff?
2. Evidence ofsocialinteractionsinwell-being
Giventheimportanceofthequestion,howcanweshowthat people are inter-related in this way? A number of different approaches are possible: considering the ways in which individuals are observedtobehave; asking peopleto express preferences over different hypothetical situations[2]; experi-mental[3];andneurological[4].
Increasingly,overrecentyearstheseapproacheshavebeen complementedwiththedirectanalysisofsubjectivewell-being data.Mostnon-economistswronglythinkthateconomistsare only interested in money.Let me, however, conform to that stereotype and start by considering social interactions with respecttoincome.Well-beingmeasuresareconsideredhereas proxiesforwhateconomistscallutility.Assuch,wecanlookat therelationshipbetweensubjectivewell-beingandnotonlythe individual’sincomebutalsotheincomeofthereferencegroup.
Asthebroadideaisthatindividualscomparetoeachother,we expect that reference-group income will attract a negative estimatedcoefficient inawell-beingregression.
Myfirstquestionisthereforenot only‘‘Doesyour money buyyouhappiness?’’,butalso‘‘Doesyourmoneymakeother people unhappy?’’ The empirical results here do seem to suggestthatmoneydoesmakemehappy,atleastaslongasI haveitandyoudonot.Inotherwords,individualsdocompare theirincomestoeachother,andthiscomparisonindeedaffects theirsubjectivewell-being.
Inordertocarryoutthiskindofempiricalanalysis,weneed both left- and right-hand side variables. I think that we are reasonably well-served with respect to the former by life satisfaction or otherpsychological functioning variables.But whataboutthelatter?Weincludeascontrolvariablesallofthe standard demographic controls, and own income. This just leavesreference-groupincome.Inordertohaveameasureof this,we needtoknowtowhomindividuals compare.
Veryrarely,surveysaskaboutsuchcomparisonsdirectly.In Wave3oftheEuropeanSocialSurvey(ESS)1in[5],thetwo most important reference groups for income were work colleagues(36%)andfriends(15%).TheJapaneserespondents in [6] alsocite workcolleagues andfriends as the two most importantreferencegroups,butwiththeirrelativeimportance inverted(42%and20%,respectively).Mostly,however,itisthe researcher himself who imposes some particular reference group, such as people who are similar to me (same age, education,job,region,etc.),myneighboursormyfamily,when carryingoutresearch.
SomeofmyearlierworkusedtheBritishHouseholdPanel Study (BHPS)2 to look at the relationship between job satisfaction and labour income. Job satisfaction rises with own income, but falls with others’ income. The empirical resultsin[7–9] showthatthisholdsfor:
theincomeof ‘‘peoplelikeyou’’;
partner’sincome;
theincomeof otheradultsinthesamehousehold;
theincomethatyouyourselfearnedinthesamejoboneyear ago.
These early findings have been replicated using different well-beingmeasures,suchaslifesatisfaction,andavarietyof referencegroups:asurveyofsomeofthesefindingsisprovided in[2].Thegeneralconclusionisthatthewell-beingproduced by income is dependent on the social context. There is an obvious link between this finding and the Easterlin paradox [10], whereby rising Gross DomesticProduct (GDP) inrich countries does not improveaverage happiness in the general population ofthesecountries.
Thereisnowgrowingagreementthatincomeisonlyoneof theelementsofa‘‘good’’life.Inthisrespect,we canwonder whether income issomehow specialinbeingaffected bythe social context, or whether this is rather a common finding acrossmanyaspectsofeconomicandsociallife,includingthe well-being derivedfromagoodstateof health.
Wecanstartthisexercisebylookingatthelabourmarket, and considering something that is undoubtedly bad: unem-ployment. We do indeed find that the unemployed report systematicallylowerlevelsofwell-beingthandotheemployed, in both cross-section and panel data. With respect to social interactions,weexpectthemirrorimageoftheresultfoundfor income: others’ unemployment should lead an unemployed individualtofeellessbad. Andthisiswhatthe dataactually shows.TheBHPSevidencein[11]revealsthatthewell-being gapbetweenemployeesandtheunemployed(withwell-being measured by the General Health Questionnaire GHQ-12) is significantly lower in regions with greater unemployment.
1TheESSisageneral-purposeacademicsurvey.Itisbiennialandhastodate coveredover30Europeancountries.Thefirstroundwasfieldedin2002/2003 andthefifthin2010/2011.Aswellasacoremodule,eachwaveincludestwoor more rotating modules. The data is freely available for download. More informationisavailableat:http://www.europeansocialsurvey.org/.
2Along-runningpaneldatasetforGreatBritainandsubsequentlytheUK.
Thefirstwavewascollectedin1991.Seehttps://www.iser.essex.ac.uk/bhps.
A.E.Clark/Revued’E´ pide´miologieetdeSante´ Publique61S(2013)S184–S188 S185
Equally,withinthe household, the unemployedreport higher levelsofwell-beingiftheirpartnerisalsounemployed,andin generalastherearemoreunemployedadultsinthehousehold.
Inthissense,unemployment hurtslessthemorethereisofit around.
Now let’sconsider asecondtypeof market: the marriage market.Marriageistypicallyfoundtobeassociatedwithhigher levelsofwell-beingincross-sectiondata,althoughtheresearch in[12]notesthatthelion’sshareofthiseffectseemstocome frompeoplewhoarealreadyhappierbeingmorelikelytomarry ratherthanthecausaleffectitselfofmarriageonwell-being.In BHPSdata,thedivorcedreportlowerGHQscoresthandothe married.However,thewell-beingofthedivorcedriseswiththe percentage of individuals in the same region who are also divorced[13].Justasforunemployment,divorcehurtslessin high-divorceregions:divorceisestimatedtohavenoeffecton well-being ina region with a25% divorcerate. It might be counteredthat thisisa ‘‘plenty morefish in the sea’’effect, ratherthanreflectingsocialcomparisons.Onewayoftackling thisistoestimatethewell-beingofmenandwomenseparately, andconsideraseparateeffectofthemaleandfemaleregional divorcerates foreachsex.Theresultsshow thattheregional divorceratesarenotsignificantlycorrelatedwiththewell-being offemaledivorcees.Formen,themaleregionaldivorcerateis significantly positively correlated with well-being (with the coefficient on the female divorcerate being positive but not significant). Divorced men like having other divorced men around.
Unemploymentanddivorceareeasytomeasure.Inowturn tosomethingthatisrathermorenebulous,thesocialresourceof socialcapital.Inthe EuropeanSocial Survey,aproxyofthis notion can be approached by the frequency with which the individualmeetsfriends.Inalifesatisfactionregression,those whomeettheirfriendsmoreoftenreportsignificantly higher satisfactionscores(althoughwecanofcourseworryaboutthe directionofcausalityhere).Stickingwiththeideaofcomparing to others in the same region, we can calculate the regional frequency (at the NUTS2 level3) of meeting friends. This attractsanegativeandsignificant estimatedcoefficient inthe lifesatisfactionregression,andsomewhatmoresoforwomen than for men. This result is consistent with the idea of comparisontoothersaroundyou.
Recent work on European health data has uncovered a numberoffindingsthatareconsistentwithsimilarcomparison effectstothosealready mentionedregarding otheraspects of well-being.In particular,in[14], theindividual’sown health problems are estimated to have a smaller effect on the individual’sownwell-beingwhentheseproblemsaresharedby othersinthesame household.Two papers haveconsidereda potentialrelativerelationshipbetweenweightandwell-being.
First,[15]usesdatafromanumberofcountriesacrossEurope toshowthat,atagivenlevelofownBodyMassIndex(BMI), individualsreportfeelinglessoverweightinregionswherethe
average BMI is higher. Along the same general lines, [16]
appeals to data on partners in the German Socio-Economic Panel(SOEP)4toshowthatcoupleswherebothareobesehave similarlifesatisfactionlevelstocoupleswhereneitherisobese.
The evidence briefly describedso far is consistentwith a troublesharedbeingatroublehalved.Thismaybethoughtto reflectsolidaritywithinagroup,butitdoesalsomeanthatyour badnewsmakesmehappy,whichsoundsratherlesspleasant.
There is,however, onearea wherethis relationshipdoes not seemtohold:religion.
Theempiricalevidencepresentedabovehassuggestedthat theunemployedarehappierwhentheyliveclosetootherswho areunemployed,andthatthedivorcedortheindividualswith badhealthliketobewithotherdivorceesorotherpeoplewith healthproblems.Withrespecttoreligion,dopeopleofacertain religionthenalsoprefertolivewithotherswhoarelikethem?
WecanusedataattheNUTS2regionallevelfromtheESSto investigate [17]. We estimate life satisfaction separately for four main religious groups, as shown in Table 1: Roman Catholics,Protestants,otherreligions,andnoreligion.Thetop leftpanelofTable1showsthatthelifesatisfactionofCatholics increaseswiththe percentageof Catholicslivinginthe same region,butfallswiththepercentageprofessingnoreligion.The top right panelcarries out the same exercisefor Protestants:
theirsatisfactionriseswithboththepercentageofCatholicsand thepercentageofProtestants,butalsofallswiththepercentage with no religion. So far the results do indeed suggest that individuals like living with those who are like them. The surprise comes in the bottom right panel, where the life satisfactionofthenon-religiousisestimated.Thisisfoundtobe increasedwiththepercentageofCatholicsorProtestants,butto fallwiththepercentagewithnoreligion.Sothenon-religious arejustlikethereligious:bothdislikelivinginnon-religious areas.Wecarryoutanumberoftestsin[17]totrytounderstand whylivinginreligiousareas leadstohigher satisfaction, but concludethatnoneofsocialcapital,crimeandtrustcanexplain thiscorrelation.Whereasitoftenseemstobethecasethatthe effect of otherson myown health or well-being depends on whetherIaminsomewaysimilartothem,herethereligious seemtoincreasethewell-beingofall,whetherreligiousornot.
Itwouldbeofinteresttounderstandhowthisoperates:dothe religious make fewer demands on others, leading to better outcomes,ordotheyratherprovidesocialresources(eitherat theindividuallevelorvia religiousinstitutions)?
It is last of interest to consider a potential relationship betweenthekindofsocialcomparisonsdescribedaboveandthe socialgradientofhealth.Oneideaisthattheslopeofthissocial gradientbetweenincomeandhealthmightbesensitivetothe intensityofincomecomparisons:themoreindividualscompare theirincometoeachother,thegreateristhewell-beingreturnto havinghigherincome(asyoubothhavemoredollars,andmore dollarsthanothers).
3NUTS:nomenclaturedesunite´sterritorialesstatistiques/Nomenclatureof territorialunitsforstatistics.
4Along-runningpaneldataset forGermany,startingin1984: seehttp://
www.diw.de/en/soep.
Separateestimatesofthesocialgradientbycountry,using datafromthe1996WorldValuesSurvey(WVS)arefoundin [18]:inthispaper,Appendix2providesestimatedoddsratiosof the effect of high relative income on self-assessed health separatelyfor38WVScountries.Icanmatchthesetofigureson comparisonintensityfromtheESSasfoundinTable5of[5].
Unfortunately, onlynine countriesoverlapped betweenthese two (Bulgaria, Finland, Germany, Norway, Poland, Russia, Spain,SwedenandSwitzerland),whichdoesnotprovidemuch statisticalpower.Wesuspectthatgreatercomparisonintensity might yield a steeper health gradient. With only nine observations, the results of this analysis are unlikely to be contractually-binding.Butevenso,itisworthnotingthatthe twoareindeedpositivelylinkedintheaboveexercise,witha correlationcoefficientofaround0.4.Amoreseriousattemptto lookatthisquestioncouldnotonlyfindabettermatch,interms ofcountriesandyears, tothe ESScomparisonintensitydata, butcould also perhapsestimate socialgradients for different groups within a country (by sex and age, for example, or region),providingfarmoreobservations.
3. Conclusions
It is likely banal to say that health and well-being are correlated.Itisactuallymoredifficulttothinkwhat itwould meanweretheynottobe.Inthisshortnote,Ihavesuggested that the empirical analysis of health and well-being could perhapsbeboughtclosertogether(botharetypicallymeasured ordinallyandsubjectively).Inaddition,oneofthekeythemes intheEconomicsofHappinessliterature,thatofcomparisons,
mayhelpustobetterunderstandthesocialgradientinhealth.
Moreinteractionandcollaborationbetweenthedifferentfields maywellshortlytelluswhetherthishypothesisturnsouttobe true.
Disclosure ofinterest
The author declares that he has no conflicts of interest concerningthisarticle.
References
[1] ClarkAE,FrijtersP,ShieldsM.Relativeincome,happinessandutility:an explanation for the easterlin paradox and other puzzles. J Econ Lit 2008;46:95–144.
[2] AlpizarF,CarlssonF,Johansson-StenmanO.Howmuchdowecareabout absoluteversus relativeincome and consumption?JEconBehavOrg 2005;56:405–21.
[3] McBrideM.Money,happinessandaspirations:anexperimentalstudy.J EconBehavOrg2010;74:262–76.
[4] FließbachK,WeberB,TrautnerP,DohmenT,SundeU,ElgerC,etal.
Social comparison affectsreward-relatedbrain activity in thehuman ventralstriatum.Science2007;318:1305–8.
[5] ClarkAE,SenikC.Whocomparestowhom?Theanatomyofincome comparisonsinEurope. EconJ2010;120:573–94.
[6] ClarkAE,SenikC,YamadaK.ThejonesesinJapan:incomecomparisons andfinancialsatisfaction.Paris:PSE,Mimeo;2013.
[7] ClarkAE,OswaldAJ.Satisfactionandcomparisonincome.JPublicEcon 1996;61:359–81.
[8] ClarkAE.L’utilite´ est-ellerelative?Analysea` l’aidededonne´essurles
[8] ClarkAE.L’utilite´ est-ellerelative?Analysea` l’aidededonne´essurles