Portfolio Allocation over the Life Cycle Evidence

PortfolioAllocationovertheLifeCycle:EvidencefromSwedishHouseholdData*BjörnAndersson
AbstractThispaperprovidesempiricalevidenceonlife-cyclepatternsintheassetallocationofSwedishhouseholds.DataonhouseholdportfolioallocationarecollectedfromtheHINKsur-veysfortheperiod1982-1992,andportfoliosharesofdifferentassetcategoriesareregressedonage,period,andcohortdummiesaswellassocio-economicanddemographicvariables.Thereareevidentdifferencesintheageprofilesforthedemandofdifferentassets.Thefrac-tionof“risky”financialassetsfollowsahump-shapedageprofile,asdoestheshareoftotalrealassets.Whiletheprobabilityofownershipof“safe”financialassetsincreasesoverlife,theweightintheportfoliohasaU-shapedagepattern.Thisisalsotrueforthefractionoftotalfinancialassets.Furthermore,therearedifferencesintheassetallocationofdifferentbirthcohorts;theportfolioweightofrealassetsisrelativelyhigherforthe“baby-boom”generation,whileyoungergenerationsaremorepronetoinvestin“risky”financialassets.Keywords:lifecycle;assetdemand;portfoliochoice;cohort;differentialmortality;SwedenJELclassification:C2,D91,E21,G11,J10*IwouldliketothankThomasLindh,StefanHochguertel,SaraLindbergandErikZetterströmforconstructivecomments.Naturally,theusualdisclaimerapplies.IamalsoindebtedtoSaraLindbergforproductivecooperationwiththedata.FundingfromtheNationalSocialInsuranceBoardisgratefullyacknowledged.
BjörnAndersson,DepartmentofEconomics,UppsalaUniversity,P.O.Box513,SE-75120,Uppsala,Swedentel:[phone],fax:[phone],e-mail:[e-mail]11IntroductionAwell-knownpredictionfromthelife-cyclehypothesisofsavingisthatthesizeofanindi-vidual’swealthisrelatedtohisorherage.Initsmoststylisedform(e.g.Modigliani,1986)themodelpredictsthattheaccumulationofwealthwilldisplaythefamiliarhumpshape(orratherapinnacleshape)overtheindividual’slifecycle,wherethewealthisbuiltupduringtheworkingyearsandconsumedduringretirement.However,neitherinthestripped-downver-sionofthelife-cyclemodel,norinrecentextensionsincluding,forexample,precautionaryandbuffer-stockmotivesforsaving,ismuchsaidabouttheallocationofwealthacrossdiffer-entassets.Mostdiscussionsofage-dependentassetdemandcanbefoundintherelatedlit-eratureonhouseholdportfoliochoice,butmainlyintheoreticalwork.Inempiricalinvestiga-tionstheissueisusuallylimitedtoaninclusionofageasacontrolvariableinregressorsets.Determiningpossiblelife-cyclepatternsforhouseholds’portfoliochoiceisimportantinthelightoftheongoingpopulationageingintheindustrialisedcountriesaswellasthe”baby-boom”generation’spotentialinfluenceonthemacroeconomy,somethingwhichrecentlyhasattractedattention.Studiesofthestabilityofsocial-securitysystemshavemainlyfocusedonthelevelofretirementsavingofthe“babyboomers”.Recentreformsofsocial-securitysys-temse.g.inSwedenallowabiggerroleforindividualandpublicpensionfunding,withmoreindividualresponsibilityfortheinvestmentdecisions,buttherehasbeensurprisinglylittleresearchonhowthesavingis,andmightbe,allocatedacrossassets.Furthermore,life-cyclerelatedshiftsintheportfolioweightsofrealandfinancialassetsmightinfluencethedetermi-nationofothermacroeconomicvariablessuchasrelativepricesonassets,orthewealthdistri-bution.Eventhoughthereseemstobeanagreementamongfinancialadvisorsoftheadvantageforindividualsofdifferentagestoholdportfolioswithdifferentcomposition–typicallyoflessriskynaturetheolderonegets–therehasbeennoagreementabouttherationaleforthisinthetheoreticalportfolio-choiceliterature(seeCanneretal.,1997).Inastatic“mean-variance”model(Markowitz,1952;Tobin,1958)life-cyclerelatedportfoliochoicecanonlybeanissueifdemographicsareallowedtoaffectunderlyingparameterssuchastheriskaversion,sincethismeasurewillgoverntherelativeamountofwealthallocatedtorisklessvs.riskyassets.Indynamicmodelsthescopeforlife-cycleconsiderationsiswidersincetheinvestmentopportu-nitysetcanbeallowedtochange,andthisinvalidatesthemyopicbehaviouroftheinvestorinthestatic“mean-variance”world(Merton,1971).2Thefactthatinvestmentopportunitysetsdoseemtovaryovertimehasstimulatedinterestin“long-horizon”portfolio-choicemodels1whereacentralfeatureisthattheopportunitytorebalancetheportfolioinlatertimeperiodswillaffecttheportfoliochoicetoday.Sinceriskyassetsprovideahedgeagainstadverseshiftsininvestmentopportunities,a“long-horizon”investorwillholdlargerproportionsinriskyassets,wherethehedgingcomponentwilldependontheinvestmenthorizon(Brennanetal.,1997).Relatedtothisistheliteratureonback-groundrisk,i.e.riskintroducedbystochasticstatevariables.2Forexample,Bodieetal.(1992)andViceira(1999)showthattheintroductionoflabourincomeintotheportfolio-choicemodel,andthustheintroductionofhumanaswellasfinancialcapital,willcauseinvestorstorebalancetheportfolioinordertomitigatetheexposuretoriskimplicitinthehumancapital.Whatthismeansintermsofre-allocationofwealthbetweenrisklessandriskyassetsovertimeultimatelywilldependonthenatureofthestochasticpropertiesoftheassetsandthestatevariables.However,thegeneralnotionfromthesemodelsisthattheamountoffinancialwealthinvestedinriskyassetsshouldfalloverlife,orperhapsfollowahump-shapedprofil