distantwww.henhenlu.com

ExplainingWhySoManyPeopleDoNotSaveAnnamariaLusardi*CRRWP2001-05September2001CenterforRetirementResearchatBostonCollege550FultonHall140CommonwealthAve.
ChestnutHill,MA02467Tel:617-552-1762Fax:617-552-1750http://www.
bc.
edu/crr*AnnamariaLusardiisanAssociateProfessorofEconomicsatDartmouthCollege.
Theresearchreportedhereinwasperformed,inpart,pursuanttoagrantfromtheU.
S.
SocialSecurityAdministration(SSA)totheCenterforRetirementResearchatBostonCollege(CRR)fundedaspartoftheRetirementResearchConsortium.
ThisgrantwasawardedthroughtheCRR'sStevenH.
SandellGrantProgramforJuniorScholarsinRetirementResearch.
TheopinionsandconclusionsaresolelythoseoftheauthorandshouldnotbeconstruedasrepresentingtheopinionsorpolicyoftheSSAoranyagencyoftheFederalGovernmentoroftheCRR.
FinancialsupportfortheconstructionofthevariablesandthefinalsampleandpreliminaryworkonthisprojectwasalsoobtainedfromTheBrookingsInstitution,theNationalInstituteonAging(grantn.
R01-AG13893),andtheNationalScienceFoundation(grantn.
SES9905453).
TheauthorwouldliketothankRobAlessie,PattyAnderson,DougBernheim,NormanBradburn,GregDuncan,JeffDominitz,SvenFeldman,MikeHurd,DavidLaibson,GeorgeLoewenstein,DavidLoughran,ErzoLuttmer,BobMichael,SusanMayer,SendhilMullainathan,MatthewRabin,JonSkinner,NoriyukiTakayama,DickThaler,SteveVenti,MarkWarshawsky,participantsattheThirdAnnualConferenceoftheRetirementResearchConsortium,Washington,D.
C.
,andparticipantsatseveralotherconferencesanduniversities'seminarsforsuggestionsandcomments.
Anyerrorsaretheauthor'sresponsibility.
2001,byAnnamariaLusardi.
Allrightsreserved.
Shortsectionsoftext,nottoexceedtwoparagraphs,maybequotedwithoutexplicitpermissionprovidedthatfullcredit,includingnotice,isgiventothesource.
AbouttheSandellGrantProgramThispaperreceivedfundingfromtheStevenH.
SandellGrantProgramforJuniorScholarsinRetirementResearch.
Establishedin1999,theSandellprogram'spurposeistopromoteresearchonretirementissuesbyjuniorscholarsinawidevarietyofdisciplines,includingactuarialscience,demography,economics,finance,gerontology,politicalscience,psychology,publicadministration,publicpolicy,sociology,socialwork,andstatistics.
TheprogramisfundedthroughagrantfromtheSocialSecurityAdministration(SSA).
Eachgrantawardedisupto$25,000.
Inadditiontosubmittingapaper,successfulapplicantsalsopresenttheirresultstoSSAinWashington,DC.
FormoreinformationontheSandellprogram,pleasevisitourwebsiteathttp://www.
bc.
edu/bc_org/avp/csom/executive/crr/sandellguidelines.
shtml,sende-mailtocrr@bc.
eduorcallElizabethLidstoneat(617)552-1677.
AbouttheCenterforRetirementResearchTheCenterforRetirementResearchatBostonCollege,partofaconsortiumthatincludesaparallelcenterattheUniversityofMichigan,wasestablishedin1998througha5-year$5.
25milliongrantfromtheSocialSecurityAdministration.
ThegoalsoftheCenteraretopromoteresearchonretirementissues,totransmitnewfindingstothepolicycommunityandthepublic,tohelptrainnewscholars,andtobroadenaccesstovaluabledatasources.
Throughtheseinitiatives,theCenterhopestoforgeastronglinkbetweentheacademicandpolicycommunitiesaroundanissueofcriticalimportancetothenation'sfuture.
CenterforRetirementResearchatBostonCollege550FultonHall140CommonwealthAve.
ChestnutHill,MA02467phone:617-552-1762fax:617-552-1750e-mail:crr@bc.
eduhttp://www.
bc.
edu/crrAffiliatedInstitutions:MassachusettsInstituteofTechnologySyracuseUniversityTheBrookingsInstitutionNationalAcademyofSocialInsuranceUrbanInstitute2AbstractTherearevastdifferencesinwealthholdings,evenamonghouseholdsinsimilaragegroups.
Inaddition,alargepercentageofU.
S.
householdsarriveclosetoretirementwithlittleornowealth.
Whilemanyexplanationscanbefoundtorationalizethesefacts,approximatelythirtypercentofhouseholdswhoseheadisclosetoretirementhavedonelittleornoplanningforretirement.
Planningisshapedbytheexperienceofotherindividuals:individualslearntoplanforretirementfromoldersiblings.
Theyalsolearnfromtheexperienceofoldparents.
Inparticular,unpleasantevents,suchasfinancialdifficultiesandhealthshocksattheendoflife,provideincentivestowardplanning.
Inaddition,planningaffectswealthlevelsaswellasportfoliochoice.
Individualswhoplanaremorelikelytoholdlargeamountsofwealthandtoinvesttheirwealthholdingsinhighreturnassets,suchasstocks.
Thus,planningplaysanimportantroleinexplainingthesavingbehaviorofmanyhouseholds.
JELclassification:D91,E21,C21.
Keywords:Planning;siblings;wealthandportfoliochoice.
AnnamariaLusardiDartmouthCollegeDepartmentofEconomics6106RockefellerHall,Room327Hanover,NH03755-3514Phone:(603)646-2099Fax:(603)646-2122E-mail:Annamaria.
Lusardi@Dartmouth.
edu31.
IntroductionTheintertemporaloptimizationmodel,suchasthelife-cycle-permanentincomemodel,hasbeenthebasisfortheworkonsavings.
Thebasicimplicationofthemodelisthatagentsplanconsumptionandsavingconsideringtheirlife-timeresources.
Thus,agentsareforward-looking,theyanticipatethedeclineinincomeatretirementandsavetooffsetthechangeinresourcesinthefuture.
Therehavebeenseveralextensionsandmodificationstothisbasicmodel.
Forexample,aprecautionarymotiveandabequestmotivehavebeenincorporatedintotheintertemporalscheme.
1Itisstillratherdifficult,however,togenerateanaccuratematchofthedistributionofwealthintheU.
S.
.
Whatisparticularlydifficulttorationalizeisthehugedifferencesinwealthholdingsamonghouseholdsofsimilarcharacteristicsandeconomicstatus.
Manystudieshavedocumentedthathouseholdswithsimilarpatternsofincomeendupaccumulatingvastlydifferentamountsofwealth.
2Weknowrelativelylittleaboutwhatiscausingthehugedifferencesinwealthholdingsthatweobserveacrosshouseholdsandwhysomehouseholdssavesolittle.
Oneaspectofbehaviorthatthebasicmodelofsavingandintertemporaloptimizationhasnotmuchconsideredisthatplanningforretirementisacomplextask,andmanyindividualsmayperformitveryimperfectly,uptotheextremeofpostponingthedecisionuntilitistoolate.
Justtomentionafewofthedifficultiesinvolvedintheprocess,theinformationrequiredformakingdecisionsisextensive,andtherulesconcerningSocialSecurityandpensionsareratherelaborate.
ThereislittleevidencethathouseholdsarewellinformedabouttheirSocialSecurityandpensionbenefits.
3Verylittleworkhasbeendoneininvestigatinghowhouseholdsmakesavingplansandhowtheycollectalltherelevantinformationtomakesavingdecisions.
Inparticular,almostallmodelsassumethattherearenoplanningcostsand,forexample,nodifferencesinhowindividualsaccess(andevaluate)informationandhowtheyovercomeallthedifficultiesof1SeethereviewoftheworkonconsumptionandsavingsinDeaton(1992)andBrowningandLusardi(1996).
2SeeVentiandWise(1997)forathoroughexaminationofthedispersionofwealtharoundthetimeofretirement.
3SeeGustmanandSteinmeier(2000).
4devisingsavingplans.
However,thesecostsmayexistanddifferencesindevisingsavingplansandcarryingthoseplansoutcanbepowerfuldeterminantsofbothwealthholdingsandportfoliochoice.
Thistopicisofimportancenotonlytoadvanceourknowledgeofsaving,butalsotoinformthecurrentdebateontheeffectivenessofsavingincentives,ontheeffectsofprivatizingSocialSecurity,andontheconsequencesofdifferentpensionschemes(suchasthecurrentshiftamongemployersfromadefinedbenefitpensionschemetoadefinedcontributionscheme).
Inthispaper,IusedatafromtheHealthandRetirementStudy(HRS)toexaminesavingsinolderfamilies.
Ishowthatapproximatelythirtypercentofhouseholdswhoseheadisclosetoretirementhavedonelittleornoplanningforretirement.
Ialsoshowthatplanningisshapedbytheexperienceofotherindividuals:individualslearntoplanforretirementfromoldersiblings.
Theyalsolearnfromtheexperienceofoldparents.
Inparticular,unpleasantevents,suchasfinancialdifficultiesandhealthshocksattheendoflife,inducepeopletoplan.
Householdswhoseheaddoesnotplanhavesubstantiallylowerwealthholdingsthanhouseholdswhoseheadhasmadesomeretirementplans.
Theyalsoholddifferentportfoliosand,mostspecifically,theyarelesslikelytoholdstocks.
Thosewhodonotplanarealsomorelikelytofacedifficultiesinretirement.
Thepaperisorganizedasfollows:inSection2,Idescribesomeempiricalfindingsaboutsavingsandconsumption.
InSections3and4,Idescribetheempiricalevidenceonsavingsandportfoliochoiceandtheeffectsofplanningforretirement.
InSection5,Iexaminetheconsequencesofnotplanningforretirement.
InSection6,Idiscussmyfindings.
InSection7,Iprovidesomebriefconclusions.
2.
SomeFactsTheempiricalevidenceonsavingspresentssomechallengestoexistingmodelsofsavingbehavior.
BrowningandLusardi(1996)presentareviewoftheworkinthisliteratureandshowthatthereareseveralunresolvedissuesaboutsavings.
Manyworksemphasizethatthereishugeheterogeneityinhouseholdsavingbehavior,andmuchmorethancanbejustifiedintraditionalmodelsofsaving.
Forexample,DiamondandHausman(1984),Poterba,VentiandWise(1994),VentiandWise(1997,1998),andLusardi(1999)allshowthattherearevast5disparitiesinwealthholdingsandthatthosedisparitiespersistevenwhenlookingamonghouseholdsofsimilarageandeconomicstatus.
Notonlydowealthholdingsvarywidelyacrosshouseholds,butalsomanyfamiliesreportlowsavingsevenclosetoretirement.
DiamondandHausman(1984)wereamongthefirsttopointoutthatalargefractionofhouseholdsarriveatretirementwithaslittleas$1,500(in1966dollars).
Otherstudies,suchasHubbard,SkinnerandZeldes(1995)andBernheimandScholz(1993)notethatwealthholdingsareparticularlylowforhouseholdswhoseheadhasloweducation.
Asmentionedbefore,inthispaperIusedatafromtheHRS,asampleofU.
S.
householdswhoserespondentswerebornbetween1931and1941,toshedlightonsavingsofhouseholdswhoseheadisonlyafewyearsawayfromretirement.
Thissurveyprovidesdetailedinformationonwealthandtheretirementprocess,withafocusonhealth,labormarkets,andeconomicandpsycho-socialfactors,andprovidestheresearcherwithanunusuallyrichsetofinformationwithwhichtoanalyzehouseholdbehavior.
4Below,Ireportsomesimpledescriptivestatisticsabouthouseholdwealthholdings.
InTable1,IconsiderthesampleofhouseholdsinthefirstwaveoftheHRS,excludethosewhererespondentsarepartiallyorfullyretired,andthosewithrespondentsyoungerthan50orolderthan61.
Ireporttwomeasuresofwealth:liquidandtotalnetworth.
Thefirstmeasureofwealthisdefinedasthesumofcheckingandsavingaccounts,bonds,stocks,andotherassets,minusshort-termdebt.
Thesecondmeasureisobtainedbyaddinghousingequity,otherrealestate,IRAsandKeoghs,businessequity,andvehiclestofinancialwealth.
Tolookmorecloselyatmajorcomponentsofwealth,Ialsoreporttheamountofwealthinretirementassets(IRAsandKeoghs)andhousingequity.
Thedatarefertotheyear1992,andallvaluesarein1992dollars.
SincetheHRSoversamplesblackandHispanichouseholdsaswellashouseholdsfromFlorida,Iusehouseholdweightstoobtainstatisticsrepresentativeofthepopulation.
4Thereareseveraladvantagesinusingthisdataset.
First,thereismuchinformationabouthouseholdswhichisnotpresentinotherU.
S.
datasets,aswillbeexplainedlater.
Second,sinceitconcentratesonaspecificcohort,thesamplesizeismuchbiggerthaninotherdatasetsthatcovertheentirepopulation.
Third,theimplementationofnewtechniquestoelicitinformationaboutwealthhasledtoratheraccuratewealthreports.
ForathoroughexaminationoftheHRS,thequalityofthedata,andcomparisonswithotherdatasets,seeJusterandSmith(1997)andSmith(1995).
Also,seethedataappendix.
6Thefirstimportantfeaturetonoteisthatthereisatremendousamountofheterogeneityinhouseholdwealthholdings,evenwhenlookingatanarrowagegroupinthepopulation.
Whilesomehouseholdshaveamassedlargeamountsofwealth,othershaveaccumulatedverylittle.
Thosedifferencespersistevenwhenaccountingforhouseholdcharacteristicssuchasincome,race,andeducation.
5ItisalsoapparentfromTable1thathousingisanimportantassetinmanyhouseholdportfolios,andmanyhavelittleinanythingelsethantheirhomeequity.
Itisanissue,however,whetherhouseholdsusehousingequitytosupporttheirconsumptionatretirement.
Aseriesofstudiesshowthatthereisalimiteddownsizingofhousingafterretirementandalimiteduseofcontractssuchasreversemortgages.
6Retirementassets,suchasIRAs,havebeenoneofthefastestgrowingcomponentsofhouseholdwealthinthe1980sand1990s.
However,ownershipandtheamountinvestedinthesetax-favoredassetsareheterogeneousacrossthesample.
7AsecondimportantfeaturetonoteinTable1istheproportionofhouseholdsthatarriveclosetoretirementwithlittleornowealth.
Aquarterofthehouseholdsinthesamplehavelessthan$30,000intotalnetworth.
Toputvaluesinperspective,thisamountwouldbuyanannuityofapproximately$2,200ayear.
Whiletotalnetworthisonlyapartialmeasureofaccumulation,sinceitdoesnotincludewealthinSocialSecurityandpensions,itishardtoborrowagainstretirementassets,anditisnotobvioushowhouseholdswithonly$850(thefirstdecileofthedistributionofnetworth)canoffsetpotentialshockstoincome,health,orfamilycircumstances,andhowquicklytheycanaccumulateastockofwealthintheremainingyearsuptoretirement.
Notonlyiswealthveryheterogeneous,butalsoportfoliosvarywidelyacrosshouseholds.
Retirementassets,suchasIRAsandKeoghs,areconcentratedamonghouseholdswhoseheadhasatleastahighschooleducation.
Onlyafractionofthepopulationholdstocks5SeeSmith(1995)andLusardi(1999).
6See,amongothers,VentiandWise(1990,1991),andSheinerandWeil(1992).
7Asubstantialportionoftotalnetworthisalsoaccountedforbybusinessequity.
Eventhoughthehouseholdsowning(oneormore)businessesaccountforonly15percentofthepopulationinthissample,theirwealthholdingsarelarge.
Theconditionalmeanandmedianare$75,000and$281,620respectively.
Again,itisnotclearthatbusinessequitycharacterizesaccumulationforretirement,sinceinthiscasetheretirementmotiveismixedwiththeenterprisemotive.
7andbonds,andthoseassetsarealsoheavilyconcentratedamonghouseholdswhoseheadhasahighschooleducation.
Mostimportantly,thelesseducatedrespondentsarenotonlylesslikelytoholdstocksandbonds,buttheyalsodonotholdbasicassets,suchassavingandcheckingaccounts.
Notethatasizableproportionofhouseholds,i.
e.
,15%,donotholdanyoftheconventionalfinancialassets(i.
e.
,checkingandsavingaccounts,certificatesofdepositsandTreasurybills,bonds,stocks,IRAsandKeoghs,andotherassets).
Infact,thecompositionofmanyportfoliosisrathernaive:asmanyas32%havealloftheirfinancialwealthinoneasset(theproportiongoesto47%ifweconsiderhouseholdswithzerooronlyoneasset).
Alargefractionofhouseholds(49%)donotholdanyoftheassetsthathavedeliveredrelativelyhighreturnsthroughouttheyears(notonlyrealestateandbusinesses,butalsobondsandstocks)andmorethan50%ofthefamiliesinthisagegroupdonotholdanystocksorIRAs.
AsimplewayofevaluatinghouseholdpatternsofaccumulationisprovidedinWarshawskyandAmeriks(2000).
TheyperformtheexperimentofimputingthecurrentwealthholdingsofU.
S.
households,asreportedintheSurveyofConsumerFinances(SCF),intooneofthemostpopularfinancialplanners:QuickenFinancialPlanner.
Accordingtothepredictionsoftheplanner,theyfindthatabouthalfofworkingmiddleclassAmericanhouseholdswillnothavefullyfundedretirements.
Somewillactuallyrunoutofresourcesveryshortlyafterretirement.
Thesefindingsareconcentratedonspecificgroupsofhouseholds,inparticularthosewithlittleeducationandtheonesclosetoretirement(age50ormoreasintheHRSsample).
Additionally,oneoftheproblemsthattheseauthorsemphasizeisthatmanyhouseholdshavelimitedresourcesuntillateintheirlife-cycleorstartsavingverylate,uptothepointwhereitisnotpossibletodomuchaccumulation.
SimilarfindingsarereportedbyMooreandMitchell(2000).
TheyusedatafromtheHRStodeterminehowmuchwealth(includingSocialSecurityandpensions)olderhouseholdshave,andhowmuchtheywouldneedtosaveiftheywishedtopreserveconsumptionlevelsafterretirement.
Theyconcludethatthemajorityofolderhouseholdswillnotbeabletomaintaincurrentlevelsofconsumptionintoretirementwithoutadditionalsaving.
Inparticular,themedianHRShouseholdwouldstillhavetosaveanadditional16%ofincometosmoothconsumption8afterretirement.
Theempiricalworksonconsumptionsuggestthatthesepredictionsmayturnouttobeaccurate.
Thereisamountingsetofevidencethatconsumptionfallssharplyatretirement,andmuchmorethancanberationalizedbyexplanationsconsistentwithtraditionalmodelsofsaving,and/orextensionsthattakenon-separabilitiesbetweenconsumptionandleisureintoaccount(suchas,forexample,thecostsofgoingtowork).
Inanearlystudy,Hamermesh(1984)reportsthatconsumptionbywhitemarriedcouplesearlyinretirementexceedsby14%theincomethattheirfinancial,pension,andSocialSecuritywealthcangenerate.
Thus,resourcesavailabletotheseretireesareinsufficienttoallowthemtosustainthelevelofrealconsumption.
Astheauthorshows,bothcross-sectionalandlongitudinaldatademonstratethathouseholdsrespondtotheinsufficiencyofresourcesbyreducingtheirconsumptionastheyage.
8Usingmorerecentdata,Bernheim,Skinner,andWeinberg(1997)againdocumentthatconsumptionexhibitsasharpdropatthetimeofretirementandanevenmorepronounceddeclinepost-retirement.
Mostimportantly,theyfindastrongnegativecorrelationbetweenretirementsavingsandthemagnitudeoftheconsumptiondrop;consumptionfallssharplyforthosewithlittlewealth.
Theyalsolookintothecompositionofconsumptionandfindlittleevidencethatrelativetastesforleisure,homeproduction,orwork-relatedexpensescanrationalizethelackofconsumptionsmoothing.
3.
1CostsofPlanningHowcanweexplainthehugeheterogeneityinsavingsthatweobserveempiricallyHouseholdsmaydiffersubstantiallyinthewaytheyimplementsavingplansbecausetheyfacedifferentcostsofplanninganddifferentwaysoflearning.
ThereissomeindirectevidenceontheseissuesbothintheSCFandintheHRS.
Inthe1995SCFrespondentsareaskedtoreportwhatsortofinformationtheyusetomakedecisionsaboutsavingandinvestment(seeTable2a).
TobeconsistentwiththeHRSsampleIuseintheempiricalwork,IexaminedalimitedagegroupintheSCF(theoneswhoare51-61yearsold).
Householdsrelyonseveralsourcesofinformation.
However,respondentsmostoften8Mariger(1987),HausmanandPaquette(1987),Banks,BlundellandTanner(1998)reportadditionalevidenceofadeclineinconsumptionatretirement.
9reportthattheyrelyonplannersorbrokers,readmagazinesandnewspapers,"callaround",andrelyonrelativesandfriends.
9Householdswhoseheadhashigheducation(morethanhighschooleducation)relymostoftenonplannersandbrokers,andalsomakeuseofaccountantsandlawyers.
Theyalsorelymoreoftenthanloweducationrespondentsonmagazinesandnewspapers.
Mostimportantly,bothhighandloweducationgroupsrelyonrelativesandfriendstomakedecisionsaboutsavingandinvestment.
Theeffortofsearchisnotonlyinfluencedbyhowhardthetaskisbutalsobyhowunpleasantitis.
First,obtainingandevaluatinginformationcanbeanunpleasanttaskforconsumerswithlittlefinancialliteracy.
Second,retirementisnotapleasanteventforeveryindividual.
Infact,somemayviewitasatimewhenoneisunproductive,orlonely,orunhealthy.
ThereissomeevidenceinsupportofthishypothesisintheHRS.
Table2breportstheproportionofrespondentsintheHRSsampleaccordingtohowtheyevaluatealistoffactsaboutretirement.
Notethatforasizablefractionofrespondentsretirementisnotnecessarilyaneventtolookfor.
Mostimportantly,thetableindicatesthathealthproblemsrepresentaworryformanyrespondents,andIwilltrytousethisinsighttoexplainhouseholdplanningbehavior,aswillbeexplainedinSections4.
1and4.
3.
3.
2EvidenceonPlanningAsimplewaytoevaluatehowhouseholdsobeythesimplepredictionsofthelifecyclemodelistolookdirectlyattheevidenceonplanning.
Littleworkhasbeendoneonthistopicsofar,butafewpapersoffersuggestions.
YakoboskiandDickemper(1997)examinedatafromthe1997RetirementConfidenceSurvey,whichcollectedinformationonAmericanworkers'sretirementplanningandsavingbehavior.
Theyreportthatalargeproportionofworkershavedonelittleornoplanningforretirement;only36%ofcurrentworkershavetriedtodeterminehowmuchtheyneedtosavetofundacomfortableretirement.
10Animportantfeature,however,9ThefiguresreportedinTable2aaretheproportionofrespondentswhohaveindicatedthespecificsourceofinformationlistedinthefirstcolumn(allfiguresareweightedtotakeaccountofoversamplingofrichhouseholdsintheSCF).
Sincemultiplesanswersarepossible,theproportionssumtomorethanone.
Theremainingcategoriesmentionedbyrespondentsrefertonotsavingorinvesting,notgettingadvice,usingothersourcessuchasinvestmentseminarsandclubs,materialfromwork,television,etc.
10Whileolderworkersaremorelikelytoreportaretirementsavinggoal,only44%ofpre-boomershavedonearetirementsavingneedscalculationascomparedwith32%oflateboomers.
SeeYakoboskiandDickemper10isthatmanyoftheworkerswhohavedonethecalculationcouldnotgiveafigurewhenasked.
Thus,accordingtothissurvey,asmanyas3/4ofworkershavelittleidearegardinghowmuchmoneytheyneedtoaccumulateforretirement.
Whenquestionedwhythecalculationwasnotattempted,manyreportedtheycouldnotsavemoreorretirementwastoodistanttoknowwhatwouldbeneeded.
Interestingly,asignificantproportionreportedtheycouldnotfindthetimeandalsothattheywereafraidoftheanswer.
Theotheranswerswerethattheprocessiseithertoocomplicatedortheydidnotknowhowtofindhelptodoit.
BenartziandThaler(1999)examinedataonretirementplanningofrecentlyhired(non-faculty)staffemployeesattheUniversityofSouthernCalifornia.
Theyreportthatmostoftherespondentsdidnotreadothermaterialthantheoneprovidedbytheirpensionfundcompanyanddidnotconsultwithanyoneotherthanfamilymembers.
Therearemanybooksonfinancialplanningandonsavingforretirement,anditisinstructivetoreadtheirsuggestions.
Manyexplainbasicconcepts,suchascompoundinterest,riskdiversification,etc.
Whilesourcesofadvice,toimplementthestrategiessuggestedinthesebooks,individualswouldstillneedtocollectadditionalinformationbeforemakinginvestmentdecisions.
11Itisalsousefultonotethecommon"mistakes"reportedinthesebooksconcerninghowpeoplemanagetheirpersonalfinances.
Forexample,inherlistofcommonmistakes,Glink(1999)placesattheverytopthetendencyofpeopleto"putitoff'tilltomorrow".
TheErnst&Young'sRetirementPlanningGuide(1997)liststhemajorobstaclestoasecureretirementandplacedprocrastinationinthirdplaceinorderofimportance,afterinflationandtaxes.
TheHRSprovidessomeinformationonindicatorsofplanning.
InsectionKon"RetirementPlans",respondentsareaskedtoreporthowmuchtheyhavethoughtaboutretirement.
Respondentscanonlychooseacrossfourdifferentanswerswhicharereportedat(1997)fordetail.
ThesefiguresareclosetomyfindingsintheHRS.
11ThereismuchinformationonretirementplanningontheInternet.
AsreportedbyB.
O'BrianinaWallStreetJournalarticleonFebruary7,2000,asimplesearchonthewebturnedup933siteswithretirementcalculators.
Theproblemofusingthatinformation,however,isthattherearemajordifferencesinthesuggestionsofferedbythemostpopularretirementplanners.
11thetopofTable3.
12Thefirstfeaturetonoticeisthatapproximatelyonethirdofrespondentsinthissamplehave"hardlythought"aboutretirement.
Thisisalargepercentage,inparticularconsideringtheageoftherespondents.
Manyrespondentsareonlyafewyearsawayfromretirementandtheeventisimminent.
Sincethewordingofthequestionunderconsiderationisrathergenericandcanleadtoseveralinterpretations,Ialsoreportthecharacteristicsofrespondentsacrossdifferentanswermodes(thefiguresarethefractionofhouseholdsineachgroup).
Consistentwiththefactthateducationandfinancialliteracycanbemoreconducivetoplanning(searchcostsarelower),respondentswhodonotthinkaboutretirementaremorelikelytohaveloweducation.
Notonlyistheireducationlow,butalsotheeducationoftheirfamilyoforigintendstobelow(fatherormotherdonothaveahighschooleducation).
Individualswhohavenotthoughtaboutretirementarealsolesslikelytobemarried.
Additionally,theyarelesslikelytohaveoldersiblings(olderthan62)thatcouldprovidesomeguidanceorexperienceonwhathappensafterretirement.
ThebottomthreerowsofthetablereporttheaveragescoresonthemeasuresofcognitiveabilitiesavailableintheHRS:1)abilitytothinkquickly(thescoregoesfrom1to5,where1meansexcellentand5poor);2)memory,whichmeasuresthenumbersofwordsonepersonisabletorecallintwosubsequenttrials(thetotalnumberofwordsis20,andthetotalscorethereforegoesfrom0to40);and3)analogy,whichmeasurestheabilitytoreporthowsomethingsarealike(thereare7questionstotaling2pointseachforatotalscoreof14).
Overallthepeoplewhohavenotthoughtaboutretirementreceivetheworstaveragescoreonallquestions.
Inparticular,thequestionaboutanalogyiswheretheygetascorethatissignificantlylowerthanthatofothergroupsofrespondents.
Thisisrelevant,sincethisisonecharacteristicthatcanbemostusefulasaskillforplanningforthefuture.
3.
3PlanningandHouseholdSavingsDohouseholdswhoseheaddoesnotplanaccumulatelesswealththanhouseholdswhoseheaddoessomeplanningInTable4,Ireportasimpleclassificationoftotalnetworth12Thesampleissmallerinthiscase(from5,292observationsitgoesto4,489observations)sincerespondentswhoreporttheywillneverretirecompletelyarenotaskedtoreporthowmuchtheyhavethoughtaboutretirement.
12acrosshowmuchrespondentshavethoughtaboutretirement.
Respondentswhohave"hardly"thoughtaboutretirementstandoutasaverydifferentgroupthanthosewhohavethought"alittle"or"alot"aboutretirement;theirmedianwealthholdingsarealmosthalfthesizeofthosewhohavethoughtaboutretirementandmanyhouseholdsinthisgroupreportnegativeorlittlesavings.
13Itisnoteasytointerpretthisevidencesincetherearemanyreasonswhyhouseholdshavelowwealthholdingsclosetoretirementandthevariablemeasuringplanningcouldsimplybeaproxyforeconomiccircumstanceorforpreferences.
Thereareseveraldifferentexplanationsforlowaccumulationthatcouldbeconsistentwiththelife-cyclemodeloranenlargedversionoftheintertemporaloptimizationmodelofsaving.
Forexample,householdsmayrelyonpensionsorSocialSecurity.
Theymayhavelittlesavingsbecausetheyhavelowlifetimeresourcesorhavebeenhitbymanyshocks.
Inaddition,theymayaccumulatelittlebecausetheydonotfacehighrisks(forexample,unemploymentorhealthrisks),ortheyhaveformalorinformalinsurance(throughanetworkoffamiliesandfriends)againstadverseevents.
Theymayalsoexpecttoreceiveinheritances,ortoenjoybigcapitalgainsontheirassets.
Inaddition,theymayexpectashortretirementperiod,dueperhapstoshortlongevity,ortheymaysimplybeimpatientanddiscountthefutureheavily.
Thislongbutstillpartiallistofexplanationsservestoemphasizethatitisveryhardtointerprettheevidenceonthelackofsavings,and,inparticular,thatoneneedsmuchdataaboutindividualcircumstancestoaddressthisissue.
Inthefollowingsection,IdescribetherichnessofinformationprovidedbytheHRSandthemanydeterminantsofsavingsIcanaccountforwhenanalyzinghouseholdbehavior.
3.
4TheDeterminantsofHouseholdSavingsAcriticismoftenraisedintheempiricalworkonsavingisthatresearchersuseaveryrestrictiveversionofthelife-cyclemodelanditbecomesperhapstooeasytofindevidencethat13Therearealsoanon-trivialnumberofrespondentsinthesamplewhoplantoneverretirecompletely.
Thisgroupisratherheterogeneousintermsofwealthholdings;somerespondentsreportahighamountofwealthandothersverylittlewealth.
Approximately30percentofthisgroupisaccountedforbytheself-employed,thusmanywillbeexcludedinthefinalsamplebecausemanyofthequestionsaboutretirementandfutureexpectationsarenotaskedoftheserespondents.
13doesnotsupportthepredictionsofthemodel.
OneoftheadvantagesofusingtheHRSisthatitdoesprovidealotofinformationonindividualrespondents.
Thisallowstheresearchertoexaminemanyofthereasonsforhouseholdbehaviortowardssavings.
Below,Ibrieflyexaminefourimportantsetsofinformationthatcanhelpingaininginsightsintohouseholdsavingbehaviorandinexplainingthedifferencesinpatternsofaccumulation:1)PensionandSocialSecuritywealth:UsingtheHRS,itispossibletolinktotheSocialSecurityrecordsofrespondentsandusethatinformationtocalculateSocialSecuritywealth.
14However,noteveryhouseholdhasgivenauthorizationtoaccesstheirSocialSecurityrecords,andIhaveusedimputedSocialSecuritywealthdataforthosehouseholds.
15Itisalsopossibletoconstructpensionwealthfromtheself-reportedpensioninformation.
16Thus,Icanrelyonacompletemeasureofhouseholdaccumulationwhenexaminingsavingbehavior.
2)Pasteconomiccircumstances:TheHRSprovidesinformationonpasteconomiccircumstances,suchaspastshocks.
Respondentsareaskedwhethertheyhavebeenunemployedinthepast.
Inaddition,theyareaskedtoreportwhethertheyfacedanyepisodesthatmadeitdifficulttomeetfinancialneeds.
Therearealsopositiveshocksinthepastthataffecthouseholdwealth,suchasreceivinginheritances,moneyfrominsurancesettlements,ormoneyfromrelativesandfriends.
Thesepositiveandnegativeshockscanbeanotherimportantexplanationforthewidedifferencesinwealthholdingsthatweobserveempirically.
3)Expectationsaboutthefuture:Inadditiontothepast,itisimportanttohaveinformationaboutfutureresources.
IntheHRS,respondentsareaskedtoreporttheprobabilitiesthathomepriceswillincreasemorethantheincreaseinthegeneralpricelevel,andthatSocialSecuritywillbecomelessgenerousinthefuture.
Respondentsarealsoaskedto14SpecialauthorizationisneededtoaccessSocialSecurityrecords.
FordetailontheconstructionofSocialSecuritywealth,seeMitchell,OlsonandSteinmeier(2000).
15IthankAlGustmanandTomSteinmeierforprovidingtheimputedSocialSecuritywealthdata.
FormoreinformationonpensionandSocialSecuritywealthintheHRS,seeGustmanandSteinmeier(1999).
16Foradetailedexplanationoftheconstructionofthepensiondata,seeVentiandWise(1997).
14reporttheprobabilityoflivingupto75and85.
17Inaddition,respondentsreporttheprobabilitytheywillhavetogivemajorfinancialhelptofamilymembersinthenext10years.
Mostimportantly,respondentsareaskedabouttheprobabilityoflosingtheirjobnextyear.
Iusethisvariabletoconstructameasureofincomevariation.
18Ifhouseholdshaveaprecautionarysavingmotive,theycarenotonlyaboutthedeclineinincomeatretirement,butalsoaboutrisk,whichcanbemeasuredempiricallybythevarianceofearnings.
Precautionarymotivescanbeimportantandareanotherpotentialexplanationforthelackofalongplanninghorizonandthepresenceofwhatcanresembleahighdiscountrateforthefuture.
4)Preferences:Another,notyetwellexploreddimensionalongwhichhouseholdscandiffer,ispreferences.
Whileitisveryhardtomeasureindividualpreferences,itisalsothecasethatparameters,suchasthecoefficientofriskaversionortherateoftimepreference,playapivotalroleinmanymodelsofintertemporaloptimization.
ThereisawaytoinferthisinformationintheHRS,andthereforetoaccountforvariationinpreferenceswhenexplaininghouseholdwealthholdings.
Inparticular,IusetheanalysisprovidedinBarsky,Kimball,Juster,andShapiro(1997)onwillingnesstotakegamblestoconstructproxiesforthecoefficientofriskaversion.
Ialsousedataonsmoking,drinking,caringaboutone'shealth,andexercisingregularlytoproxyfortherateoftimepreference.
Demographicvariablesthatcouldberelatedtotherateoftimepreference,suchaseducation,race,andcountryoforigin,arealsoincludedintheempiricalestimation.
Intheempiricalworkreportedbelow,Iexaminehouseholdbehaviorbyconsideringseveralregressionsofhouseholdsavings(andassetsownership)onthisextensivesetofvariables.
Thiscanbeseenasareducedformequationofasavingmodel,butalsoasawaytoassesshowwellwecanexplaindifferencesacrosshouseholdsbyusingthisrichsetofinformationabouthouseholds.
Inaddition,andmostimportantly,Icanexaminewhether,afteraccountingformanyofthereasonsthatcanexplainsavings,lackofplanningplaysanyroleinexplainingthedifferencesacrosshouseholds.
Sincelackofplanningistheresultofchoice,I17AnexcellentexaminationofsubjectiveprobabilitiesintheHRSisprovidedinHurdandMcGarry(1995).
18See,also,Lusardi(1998).
15cannotsimplyrunOLSregressions,buthavetoinstrumentforthisvariable.
4.
EmpiricalEstimates4.
1HouseholdSavingsAsillustratedinTables3and4,respondentswhohave"hardlythought"aboutretirementrepresentalargegroupofthepopulationandtheystandoutfromtheothersintermsoftheircharacteristics.
Ihavethusdefinedasimpleplanningdummythattakesthevalueoneifrespondentshaveindicatedtheyhardlythoughtaboutretirement.
Thinkingaboutretirementisoneindicatorforretirementplanning,butthereismoreinformationintheHRSconcerningplanning.
Iusethisinformationtoconstructanindexthatmeasuresaseriesofactivitiestowardretirementplanning.
ThemethodIfollowinconstructingthisindexisasfollows:Iassignpointstorespondentsforhowmuchtheyhavethoughtaboutretirement(theoneswhohave"hardlythoughtaboutretirement"get1point,whiletheonesthathavethoughtalotaboutretirementget4points),andIaddpointsforeachadditionalplanningactivity.
Forexample,IaddapointifrespondentshaveaskedtheSocialSecurityadministrationtocalculatetheirretirementbenefitsandanotherpointiftheyhaveeverattendedaretirementseminar.
19IntheempiricalregressionsreportedbelowIusetwoindicatorsforretirementplanning:theindexIjustdescribedandthesimpledummyvariableforrespondentswhoreporttheyhavehardlythoughtaboutretirement.
Inthisway,Icanexaminedifferentproxiesforplanningandalsocheckupontherobustnessoftheempiricalresults.
Asmentionedbefore,Icannotsimplyregresssavings(orassetownership)onwhetherhouseholdshaveplannedforretirementsincebothvariablescanbetheresultofchoice.
Additionally,thevariablesmeasuringplanningarethemselvesproxiesandmaybeaffectedbymuchmeasurementerror,perhapsresultingfromtheambiguityininterpretingthequestionand/ormis-classificationinpickingtheappropriateanswer.
IaddresstheseproblemsbyperformingInstrumentalVariables(IV)estimation.
Asmentionedearlier,searchandpsychologicalcostsaffectplanning.
Iusetheageofsiblingstocapturethesecosts.
Onesimple(andperhapsinexpensive)wayoflearningistorely19Eventhoughitisarbitrarytoassignthesameweighttodifferentretirementactivities,itisveryusefultoconcentratealltheinformationonretirementplanningintoonevariable.
16onsiblings,inparticularoldersiblingsthatalreadyhaveexperiencewithretirementandsavings.
20InTable3,thereisevidencethatrespondentswitholdersiblingsaremorelikelytothinkaboutretirement.
ThereisalsoevidencefromtheSCFthatrelativesandfriendsareoftenconsultedtomakefinancialdecisions.
Sincetheorderofbirthmaycapturespecificpersonalityfeatures(somehavearguedthatafirst-bornislesslikelytorebel),andtherecanbeinteractionandlearningfromothermembersofthefamily,Ihavealsousedtheageofthesiblingsofthespouse.
Thiscouldalsoavoidtheproblemofgenetictransmissionofpreferences,atleastaslongaspreferencesofsiblingsandtheirspousesarenotperfectlycorrelated.
21TheinstrumentIuseisthedifferencebetweentheageoftheoldestsiblingandtheageoftherespondent(forrespondentswithoutsiblingsorwithonlyyoungersiblingsthisdifferenceissettozero).
Ialsousethedifferencesquaredtocapturepotentialnon-linearities.
Notethattheagedifferenceisnotsmall.
Forrespondentswitholdersiblings,thedifferenceissixyearsonaverage.
Additionally,manyhouseholdshavesiblingsthathavepassedtheretirementage(siblingsolderthan62).
Thus,thepotentialforlearningispresent.
Ialsousethenumberofoldersiblingstocapturethepotentialforinteractionandlearning.
Sincethenumberofsiblingscouldproxyforinheritances,Iaddamongthesetofcontrolsinthefirststageregressionavariableforwhetheratleastoneparentisstillalive.
InTable5(column1),Ireporttheresultsofregressingthesimpledummyfornotthinkingaboutretirementonthesevariables.
Theagedifferenceisagoodpredictorforplanning;respondentswitholdersiblingsarelesslikelytoreporttheyhavenotthoughtaboutretirement.
TheHRSreportsinformationonthefinancialsituationofsiblingsandwhetheritisbetterorworsethanthefinancialsituationoftherespondent.
Respondentsaremorelikelytoplanforretirementifthefinancialsituationofoldersiblingsisworsethantheircurrentfinancialsituation20IthankDavidLaibsonforprovidingmanysuggestionsontheinstrumentset.
21Onemightarguethatrespondentswitholdersiblingshavealsoolderparents.
Thereissome(suggestive)evidenceontheliteratureonsavingsthathouseholdswhowerebornbeforetheGreatDepressionhavedifferentpreferencesthanyoungergenerations.
Inparticular,theyaresupposedtobemorethriftyandalerttorisk.
Thosehouseholdscouldhavepassedthesetypesofpreferencesdowntotheirchildren.
AsdiscussedinBrowningandLusardi(1996),thisfindingisstillverycontroversial.
Additionally,giventherelativelynarrowagerangeofrespondents,itisdoubtfulthatdifferencesinpreferencesacrossadjacentgenerationsaresolarge.
Furthermore,giventhewayIhavedefinedit,respondentscouldhaveoldersiblingssimplybecausethesiblingsofthespouseareolder.
17(seeTable5,column2).
Thus,oldersiblingscannotonlyprovideinformationaboutretirementbutonecouldalsolearnfromthemistakesofothers.
Inaddition,witnessingthefinancialdifficultiesofsiblingsmayperhapsdiminishthereluctancetoundertakeanunpleasanttask.
Thesevariablesremainsignificantevenwhenaccountingforthelargesetofcontrolsinthefirststageregression.
Onemightarguethathouseholdssavetohelpouttheiroldersiblings(particularlytheonesinbadfinancialconditions).
However,inthefirststage,Iaccountfortheprobabilitythathouseholdswillhavetogivefinancialhelptofamilymembersinthefuture.
Thus,thesevariablesarenotlikelytoproxyforthisadditionalmotiveforsaving.
Topursuethisargumentfurther,apartfromtheexperienceofsiblings,agentscanlearnfromtheirownpastexperiences.
ThereisinformationintheHRSaboutwhether"inthelast20yearstherewereanyreallylargeexpensesoreventsthathavemadeitverydifficulttomeetyourfinancialgoals.
"Respondentswhoexperiencedthosepastshocksaremuchmorelikelytoplanforretirement(Table5,column3).
Pasteventscanprovidevaluableinformationand,inparticular,canprovideincentivesforputtingeffortinanunpleasanttask.
Note,however,thatIcannotusepastshocksasinstruments.
Pasteventscanhaveadirecteffectonwealth.
Householdsthatexperiencedpastshockscouldhavesufferedasubstantialdrainofresourcesandmaythereforedisplaylowerwealthholdingsthanhouseholdswhodidnotexperiencethoseshocks.
Thus,pastshocksarealsousedascontrolsinthefirststageregressions.
22Asmentionedbefore,therearecostsassociatedwiththepainofdealingwithanunpleasantevent.
AsreportedinTable2b,healthistheconcerncitedmostoftenbyrespondents.
Onewaytocapturethesecostsistouseinformationonparents.
Respondentsmaybeafraidofbeingunhealthyafterretirementiftheirparentshavebeenunhealthy.
Notonlycanillnessesbegeneticallytransmitted,butalsothehealthproblemsofaparentmaybeverypainfulexperiences(astheabovepastshocks).
IntheHRS,Ihaveinformationaboutwhetherthemother/fatheroftherespondent(andspouse)hadanillnesslasting3monthsormoreduring22Itisusefultonotethatintheregressionsofhouseholdsavingsonthelargesetofvariablesmentionedbefore,theempiricalestimatesforpastshocksbecomemorenegativewhenaccountingforwhetherhouseholdshavethoughtaboutretirement.
18thelastyearshe/hewasalive,andwhethertheyeverlivedinnursinghomes.
Thehealthexperienceofparentswhileprovidingworriesandmakingperhapsplanningmoreunpleasantcanalsoprovideinformationonthecostsonemayincurafterretirement.
Thus,thesignofthesevariablesisambiguousex-ante.
Theempiricalevidenceshowsthatrespondentswhoseparentssufferedhealthproblemsbeforedyingorlivedinnursinghomesaremorelikelytothinkaboutretirement.
Asinthecaseofsiblingsandpastshocks,painfuleventsmayinduceindividualstoputeffortintoplanningforretirement.
SinceinthefirststageregressionIaccountforthecurrenthealthstatusoftherespondentandthesubjectiveprobabilityoflivingupto75,thesevariablesarenotlikelytoaccountforindividuallongevity.
TheextensiveinformationaboutsiblingsandparentsprovidedintheHRShaspredictivepowerinexplainingthelackofthinkingaboutretirement(asmeasuredbythedummyfortheoneswhohave"hardlythought"aboutretirement)andplansforretirement(asmeasuredbytheindexvariable),evenafteraccountingforalltheothereconomicvariableslistedbefore.
Thepredictivepoweroftheinstrumentsisnotveryhigh(theR2isonly1%),buttheF-testontheexcludedinstrumentsrejectsstronglythehypothesisthattheinstrumentsarenotjointlysignificant(theF-valueisalwaysbiggerthan2,andsometimesmuchbigger,takingthevalueof6inlargeinstrumentsets).
23Toincreasethepredictivepower,Ihaveusedadditionalinformationaboutsiblings,suchaswhethersiblingsarefemale,haveahouseandwhethertheywork.
Additionaldemographicinformationaboutsiblings(whethertheyaremarried,havechildren,etc.
)hasbeenusedtoperformchecksandsensitivityanalysisusingdifferentsetsofinstruments.
Theinstrumentsappeartobevalid;thetestofover-identifyingrestrictionsisnotrejected(seelastrowofTable6).
Eventhoughthistestmayhavelowpower,itreportsadditionalevidenceinfavorofthechosenspecificationandthesetofinstruments.
Iperformtheempiricalestimationusingtwomeasuresofsavings:totalnetworthandfinancialnetworth.
InthelatterIexcludehousingequity(andotherrealestates),sinceitisnotclearthattheinvestmentinhousingistosupportconsumptionatretirement.
Ialsoexcludevehicles,andbusinessequity.
InTable6,Ireporttheresultsofregressingtotalandfinancialnet23SeeStaigerandStock(1997)andBound,JaegerandBaker(1995).
19worthoverpermanentincome24onalargesetofexplanatoryvariablesandtheindicatorforlackofretirementplanning(thedummyfornotthinkingaboutretirementisusedincolumns2and3,theindexforretirementplanningisusedincolumns4and5).
25Amongthesetofexplanatoryvariables,Iconsidernotonlyageandagesquaredtocapturethehump-shapedprofileofwealthholdings,butalsosomesimpledemographics,suchasthetotalnumberofchildren,andthenumberofchildrenstilllivingathome,gender,race,countryofbirth,maritalstatus,andeducationthatcanaccountforheterogeneityintastes.
Ialsoincludedummiesforregionsofresidence.
26Permanentincomeisincludedamongtheregressorstoaccountfornon-homotheticpreferences.
Ialsoaccountforhealthstatus,pastshocks,preferences(riskaversionandimpatience)andexpectationsaboutthefutureusingthevariablesthatIexplainedintheprevioussection.
Additionally,IaccountforbothSocialSecurityandpensionwealth(thesevariables,astheleft-handsidevariable,aredividedbypermanentincome).
AsemphasizedinBrowningandLusardi(1996),thereareothermotivestosaveapartfromprovidingforretirement.
Householdsmaysavetoleaveabequesttofuturegenerations,andIaccountforthismotivebyusinginformationontheintentionsofleavingbequeststoheirs.
Additionally,Iconsideraprecautionarysavingmotiveandproxyforitusingthesubjectivevarianceofearningsrisk.
Ialsoaccountforthefactthathouseholdsaccumulatelittlebecausetheycanrelyonhelpfromrelativesandfriendsincasetheyrunintoseverefinancialdifficultiesinthefuture.
Asmentionedbefore,Iconsiderthepossibilityofreceivingbequestsinthefuturebyusingadummyforwhetheratleastoneparentisalive,andIalsoaccountfortheprobabilityofgivingfinancialhelptofamilymembersinthefuture.
Manyofthesevariablesplayaroleinexplaininghouseholdsavings.
Householdswhoseheadhasahigheducationhavehighersavings.
Marriedcoupleshavehighsavings,while24Permanentincomeisconstructedbyregressingtotalhouseholdincomeonasetofdemographics,firmcharacteristics,occupationandeducationdummiesandthosedummiesinteractedwithage,andsubjectiveexpectationsofincomechangesinthefuture.
25Seethedescriptionofthefinalsampleinthedataappendix.
26Forbrevity,theestimatesofthesedemographicvariablesarenotreported,butarebrieflydiscussedinthetext.
20childrenhaveadepressingeffectonwealth.
Householdswhoexperiencednegativeshocksinthepastenduphavinglowerwealth,whilereceivinginheritancesorothertransfersleadstohighersavings.
Householdswithabequestmotiveaccumulatemore,whilethosewhoareimpatientaccumulateless.
Aninterestingresultfromtheregressionisthathouseholdswhohavealargepensionaccumulatemoreratherthanlesswealth,27showingthathouseholdswhohavemuchinretirementassetsalsohavemoreinotherformsofaccumulation.
Themostimportantresultisthathouseholdswhodonotplanforretirementenduphavingmuchlowersavingsthanhouseholdswhohavethought(alittleoralot)aboutretirement.
Thedummyforlackofplanningisnegativeandstatisticallysignificant.
Theindexforplanningshowsasimilareffect.
Respondentswhohavetakenstepstowardretirementplanningaccumulatemorewealth.
Thus,planningcontinuestohaveaneffect,evenafteraccountingformanyofthevariablesthatcanexplainsavings.
Theestimatesindicateverylargeeffects.
Respondentsthatdonotplanhaveapproximately100%lesssavingsthanrespondentsthathavedonesomeplanning.
Eventhoughtheestimatesarenotverypreciseandmaysufferfromtheproblemofhavingweakinstruments,theyshowthatlackofplanningforretirementisanimportantdeterminantofthelowwealthholdingsofmanyAmericanhouseholds.
4.
2RobustnessandSensitivityChecksIhaveperformedseveralchecksontheempiricalspecificationandthechoseninstruments.
First,Ihaveconsidereddifferentmeasuresofaccumulation.
Ihaveexperimentedwithusingameasureoffinancialnetworththatincludesotherrealestate,andIhaveexaminedtheeffectsofincludingandexcludingbusinessequity.
ResultsdonotchangeandIalwaysobtainthatplanninghasasignificanteffectoneverychosenmeasureofaccumulation.
Second,Ihaveexperimentedwithdifferentsetsofinstruments.
Ihaveusedasinstrumentsonlytheagedifferencebetweentheoldestsiblingandtherespondentandsomecharacteristicsofthesiblings(gender,maritalstatus,whethertheyhavechildren,etc).
Onepotentialconcernofthepreviousinstrumentsetisthatparentsmaycompensatethechildreninworsefinancialconditionsbyleavinglargerbequests.
Whilethereislittleevidencethatbequests27SimilarresultsarereportedbyGustmanandSteinmeier(1999),whousetheHRSpensiondatafromthePensionProviderdataset.
21aresplitunequally,itisworthinvestigatingwhetherresultschangewhenthefinancialconditionofsiblingsisnotincludedinthesetofinstruments.
Ihavealsoexcludedtheinformationabouttheparents.
Onecouldarguethathealthshockscreateadrainofresourcesandcouldproxyforreducedinheritances.
ThereisnotenoughinformationintheHRStodisentanglewhetherparentswerecoveredbyinsuranceandwhethertheyhaveabequestmotive.
Additionally,itiswellknownthathealthandwealtharehighlycorrelated,andsomeofthiscorrelationmaysimplybeduetounobserveddifferencesintimepreference:thosewhoinvestinhealthalsosaveforthefuture.
Healthyparentsarealsotheonesmostlikelytoplaceahighvalueonthefutureandtosaveatanaboveaveragerate.
Thistendencytosavecouldhavebeenpasseddowntochildren,alongwithattitudesabouthealth.
Hence,parentalhealthmayaffectsavingsdirectly.
Whilemyinstruments(beinginahospitalornursinghomebeforedying)maycapturemorethangeneralhealth,theymaystillsufferfromtheabovementionedproblem.
Thus,Ihaveexperimentedwithdifferentsetsofinstrumentsthatincludeandexcludetheinformationaboutparents.
Ishouldnote,however,thattheover-identificationtestdoesnotindicatethattheinstrumentsusedinTable7areinvalid.
Thesetofinstrumentshaslesspredictivepowerthantheoneusedbefore.
However,estimatesdonotchangeconsiderably,eventhoughtheyarestatisticallysignificantonlyatthe10%level.
Ihavealsoincludedamongtheregressorsadditionalindicatorsforfamilyresources,suchastheeducationofthefamilyoforigin.
Resultsarenotaffected.
Additionally,Ihaveusedasanindexofretirementplanningtheanswerstohowmuchrespondentshavethoughtaboutretirement.
Thus,IdonotconsiderattendingretirementseminarsandaskingtheSStocalculateretirementbenefits.
Resultsdonotchangesubstantially.
284.
3HouseholdPortfolioChoicePortfoliochoicecanrevealagreatdealabouthouseholdbehavior,anditisherethatwemaybeabletodifferentiateclearlyamongdifferentmodelsofbehavior.
Itiswellknownthateventhoughreturnsonstockshaveoutperformedbonds,onlyarelativelysmallfractionofhouseholdsinvestinstocks.
Infact,animportantandunresolvedpuzzleiswhysofew28Forbrevity,estimatesarenotreportedbutareavailablefromtheauthoruponrequest.
22householdsholdstocks.
29Additionally,asmentionedbefore,manyhouseholdportfoliosseemratherunsophisticated.
Similarevidenceisprovidedwhenlookingathowindividualschoosetoinvesttheirretirementassetsindefinedcontributionplans.
BenartziandThaler(2001)findevidenceofverynaivediversificationstrategies.
Manyparticipantsindefinedcontributionsavingplanssimplydividetheircontributionsevenlyacrossthefundsofferedbytheplan.
If,asmentionedbefore,muchefforthastobeexertedtoobtaininformationaboutcomplexinvestmentassets,suchasstocks,agentsfacinghighcostswillbelesslikelytoinvestinthoseassets.
Thedependentvariableintheregressionsisnowadummyforwhetherhouseholdsholdstocks.
Asforthepreviousregressionsonhouseholdsavings,Ihaveconsideredalargesetofcontrolsthatcanproxyforbothhouseholdresourcesandpreferencesthatcanexplainstock-ownership.
Intheseregressions,ratherthanconsideringameasureoftotalpensionwealth,Idistinguishamongthosewhohavedefinedcontributions,definedbenefits,andothertypesofpensions.
Respondentswithdefinedcontributionscanusuallychoosehowtoinvesttheirpensionassets,andthismayalsoaffecttheallocationoftheirnon-pensionassets.
Sincemanytransactioncostscanbefixed,Ialsocontrolforothergrossfinancialassets(i.
e.
,totalfinancialassetsminustheamountinvestedinstocks),andfinancialassetssquared.
Asreportedbefore,financialwealthholdingsaresosmallformanyhouseholdsthatitisunlikelytheyinvesttheseamountsinstocks.
Asbefore,Iusetheagedifferencebetweenrespondentsandoldersiblingsanddummiesforthefinancialstatusofsiblingsandhealthstatusofparentstocapturethecostsofplanningandthepotentialsforlearning.
Unfortunately,thereisnoinformationintheHRSontheportfoliosofsiblingsorofparents.
ChitejiandStafford(1999)reportevidencethattheholdingsofstocksamongcertaingroupsofthepopulation,suchasAfricanAmericans,isstronglyinfluencedbywhethertheirfamilyoforiginheldstocks.
Iwillproxyforthepotentiallearningfromthefamilyoforiginbyusinginformationonwhetherthefinancialconditionofparentsisgoodorpoor.
Itisveryhardtofindhighpredictivepowerfortheinstrumentsgiventhelargeset29See,amongothers,HaliassosandBertaut(1995).
23ofcontrolsusedintheregression.
Toincreasethepredictivepower,inparticularfortheindexofretirementplanning,Iaddedadummyforwhetherrespondentsworkinasmallfirm.
Thelowerdegreeofinteractionandthedecreasedpossibilitiesofundertakingretirementactivities(suchasattendingretirementseminars)shouldincreasethecostsofplanning.
Theinstrumentsetseemsvalid.
Thetestofover-identifyingrestrictionsisnotrejected(seelastrowofTable7).
AparsimoniousspecificationforstockownershipisreportedinTable7.
Eventhoughnotreported,inadditiontothevariableslistedinTable7,Ihaveaccountedforage,gender,race,countryofbirth,maritalstatus,numberofchildren,education,andregionofresidence.
TheestimatesIobtainareconsistentwithotherworkonstock-ownership.
30Forexample,householdswithhigheducationandpermanentincomearemorelikelytoinvestinstocks.
Blacksaremuchlesslikelytoinvestinstocks.
Respondentsreportingexcellentorgoodhealtharemorelikelytoinvestinstocks.
Theoneswithdefinedcontributionpensionsaremorelikelytoinvestinstocks.
Grossfinancialassetsalsoaffectsstockownership.
Themostrelevantresult,however,isthatevenaftercontrollingformanyfactorsthatcanexplainstockownership,lackofplanningisastrongdeterminantofportfoliochoice.
Boththedummyforthelackofplanningandtheindexforretirementactivitiesindicatethathouseholdswhoplanforretirementaremorelikelytoholdstocksintheirportfoliosthanhouseholdswhodidlittleornoplanning.
TheseresultsareconsistentwiththeevidenceprovidedinMunnell,SundenandTaylor(2001)concerningparticipationandcontributionto401(k)plans.
Theyreportthatashortplanninghorizonisthemostimportantvariableinexplainingtheparticipationdecision.
Planninghorizonsalsoaffecttheamountsworkerscontributetotheplan;ashortplanninghorizonreducesthecontributionratebyroughly1percentagepoint.
Iplantoexpandmyresearchinthisdirectionandexaminewhetherandhowplanningcostsaffectholdingandinvestmentinretirementassets,suchasIRAsandKeoghplans.
5.
TheConsequencesofPlanning:EvidencefromRetireesThepreviousestimatesshowthat,evenaccountingformanydeterminantsofsavings,30SeeHaliassosandBertaut(1995)andHeatonandLucas(2000).
24individualsthatdonotplanendupaccumulatinglesswealththanindividualswhoplanforretirement.
Thiscouldbetheresultoflowsavingratesaswellasdifferentportfoliochoices.
DoesthisfacthaveconsequencesWhileplanningisfoundtoaffectprivatewealthandportfoliochoice,itcouldwellbethathouseholdwhodonotplanstillmanageacomfortableretirement,forexamplebecausetherearevariablesIhavenotcontrolledfor(helpfromchildren,othersourcesofsupport,etc.
),orbecauseitishardtomeasurepensionandSocialSecurityaccurately,orbecausethespecificationofpreferencesisnotaccurate.
31However,asmentionedbefore,thereismuchevidencethatconsumptionfallssharplyatretirement,andmuchmorethancanberationalizedbyexplanationsconsistentwithtraditionalmodelsofsaving.
Mostimportantly,Bernheim,Skinner,andWeinberg(1997)showthatthedropinconsumptionismuchsharperforthehouseholdswhicharriveatretirementwithlittlewealth.
Astheauthorsreport:"ourresultsappeartosuggestthatonaverageindividualswhoarriveatretirementwithfewresourcesexperiencea"surprise"—theytakestockoftheirfinancesonlytodiscoverthattheirresourcesareinsufficienttomaintaintheiraccustomedstandardsofliving,e.
g.
becausepensionincomeislessthanexpected,orbecausetheyrecognizethatsavingswillgolessfarthantheyhadhoped.
"ThereissomeinformationintheHRStoassesstheexperienceofhouseholdswhoserespondentisalready(partiallyorfully)retired.
Thoserespondentsarealsoaskedhowmuchtheyhadthoughtaboutretirement.
Asforthesampleofnon-retiredrespondents,alargeproportionoftheserespondentshadnotthoughtaboutretirement(520outof1172observationsreportthattheyhadhardlythoughtaboutretirement).
Respondentsarealsoaskedtoratetheirretirementexperienceintwoindependentquestions.
Inonequestion,theyareaskedtoreporthowretirementhasturnedouttobeandinasecondquestiontheyareaskedtocompareretirementyearstotheyearsjustbeforetheyretire(Tables8a,b).
Asimpleclassificationacrossdifferentanswermodesshowsthatalargeproportionoftherespondentswhohavenotthoughtaboutretirement(43%)rateretirementasnotatall31SeeBrsch-SupanandStahl(1991)foranalternativecharacterizationofpreferences.
25satisfying.
Attheotherextreme,alargemajorityofrespondentswhohavethoughtalotaboutretirement(68%)rateretirementasverysatisfying.
Theotherindicatorreportssimilarfindings.
Morethan54%ofrespondentswhohavenotthoughtaboutretirementhaveratedtheretirementyearsnotasgoodastheyearsbeforeretirement,whileamuchsmallerfractionofrespondentshaveratedretirementnotasgoodiftheyhavethought(alittleoralot)aboutretirement.
Thisevidenceisonlysuggestive,butitisagainconsistentwiththeevidenceonthelowamountofaccumulationfornon-plannersprovidedintheprevioussections.
Thesehouseholdsmayindeedbelikelytoexperienceanegative"surprise"afterretirement.
6.
DiscussionHowcanwerationalizethefindingsreportedaboveAsuggestiontotakeawayfromtheempiricalanalysisisthatitistoosimplistictoassumethattherearenocostsofplanning.
Thosecostsexistandtheymayevenbesizableforsomehouseholds,forexample,thosewithlittlefinancialliteracy.
Inaddition,notonlysearchcostsshouldbeconsideredwhenmodelingsavingdecisions(forexample,thetimespentcollectinginformationor"callingaround"),butalsothepsychologicalcostsofdealingwithunpleasantevents.
Oneimportantquestionthenbecome:HowlargehavethesecoststobetopreventsomanyindividualstoplanforretirementAmodeltakenfromtheliteratureinpsychologyprovidessomechallengingpredictions.
Foragentsthatdisplayhyperbolicdiscounting,evensmallcoststhathavetobepaidimmediatelyleadtowideregionsofinaction:agentspostponeactionsthatimplyimmediatecosts.
Asreportedinmuchwork,researchonbothanimalandhumanbehaviorhasledpsychologiststopositthatdiscountratesarenotexponential,asassumedinmanytraditionalmodelsofsaving,butrathertheyarehyperbolic.
32Thesetimepreferencesareassociatedwithdecreasingimpatienceovertime.
Hyperbolicagentsdisplayarelativelyhighdegreeofpatiencewhenchoosingrewardstobeaccruedinthedistantfuture,butareveryimpatientiftherewardsaretobeobtainedinthenearbyfuture.
32SeethereviewofhyperbolicdiscountinginAinslie(1992).
26AusefulguidingframeworkistheonedescribedbyO'DonoghueandRabin(1999b).
33Asintheirmodels,Iconsiderforconveniencequasi-hyperbolicpreferencebutImodelexplicitlythecostsofplanningthatconsumersface.
Givenmyempiricalfindings,Iassumetherearecostsinsettingupasavingplanandthesecostsareoftwotypes:searchcostsandpsychologicalcosts.
Individualshavetoexertefforttocollectinformationaboutreturnsondifferentinvestmentopportunities.
Inaddition,individualshavetoexertefforttoovercomethepainofdealingwithanunpleasanttask.
Onecanthinkofthesecostsassimplyadditiveor,alternatively,assumethatpsychologicalcostsmakesearchcostsmoreburdensome.
Theeffortofsettingupasavingplanisassumedtoaffecttheinterestratethatindividualscanobtainontheirsavings(forexample,becauseinvestinginstocksgivesahigherexpectedreturn).
Theoptimizationmodelbecomesasfollows:[]∑++=+=itititTitttePCvePCvUδβ1subjecttothefollowingbudgetconstraint:()()tttttCYAerA++=+11whereTindicatesmaximumlife-time,Cisconsumption,Yisincome,Aissavingsandrtheinterestrate.
Planningincreasestheinterestrateobtainedonsavingsbutatthesametimeisaburdenanddecreasesutility(Pindicatesthelossinutilitywhichisafunctionoftheeffort"e"ofplanning).
Inthissetup,individualschoosetheoptimalamountofeffortbyequatingthecostsofplanningtothebenefitsofplanning.
Agentsthatfacerelativelylowcostswillexerthigheffortandobtainhighreturns.
Additionally,forsomepreferencespecifications,thesehighreturnsleadhouseholdstosavemore(i.
e.
,thesubstitutioneffectisgreaterthantheincomeeffect).
Iftheseconditionspersistover-time(costsalwaysremainhighforcertaingroupsofindividuals),individualswithlowplanningcostsendupaccumulatinghigheramountsofwealththanindividualswhohavethesamecharacteristicsbutfacehighercostsofplanning.
Itisimportanttohighlightthedifferencesinthismodelwithrespecttotraditionalmodelsofsaving.
Itisobviousthatplanningcostsinhibitbehaviorevenintraditionalmodels.
However,whatisdifferentherewithrespecttothosemodelsisthatevensmallcostscanhaveabigeffect33See,also,Akerlof(1991),Laibson(1997),andO'DonoghueandRabin(1999a,1999b).
27onbehavior,whilecostshavetobereallylarge,forexample,topreventalife-cycleconsumertokeepassetsinlowreturnsecurities.
Inthisalternativemodel,consumersdiscounttheimmediatefutureatahighrate,andthishighratecombinedwithacostthathastobepaidimmediatelygivesrisetomuchinaction.
Infact,contrarytothepredictionsoftraditionalmodels,someactionsmayneverbetaken;consumersmaycontinuetoprocrastinateindefinitelyandneverputanyeffortintoplanning.
Thismaybeparticularlyrelevantforbehaviorsuchassavingandportfoliochoice.
Thereareusuallynodeadlinesorspecificperiodswhendecisionshavetobemade.
Decisionscanbemadeeveryday,butsinceoneneedstoputtimeandeffortintoit,decisionscanbesubstantiallydelayedandperhapsnevertaken.
Thismodelhastwodesirablefeatures.
First,itcanrationalizewhyevensmallcostscancreatesubstantialdelaysinmakingdecisionsonimportantmatterssuchassavingforretirement.
Second,itexplainsanotherimportantempiricalfinding,i.
e.
,thefactthatpeoplethatdonotplanenduphavinglowwealthatretirement.
34Whilethisframeworkisverypromising,thereareperhapsother,andevensimpler,explanations.
Empiricalresultsfromthemacroeconomicliteratureonconsumptionhasledsomeresearcherstoassumethattherearetwotypesofagentsintheeconomy:afirsttypethatbehavesaccordingtothepredictionsoftheclassicallife-cycle-permanentincomemodel,andasecondtypecalled"ruleofthumb"consumers,thatsimplyconsumeaccordingtotheircurrentincome.
Empirically,thefractionof"ruleofthumb"consumersintheeconomyhasbeenestimatedtorangefrom20to50%.
35Onecouldassumethatthesetwogroupsarenotexogenous,butagentsswitchfromonegrouptotheotherdependingontheamountof"learning"acquiredaboutretirement.
Forexample,many(perhapsyoungconsumers)startoutas"ruleofthumb"consumers.
However,astheywitnessshockstooldersiblings(whogothroughretirement)ortoparents,theyswitchtotheothergroup.
Whilecrude,thisschemeisconsistentwiththefactthatwedonotseemuch34Ifsavingdecisionswerepurelyrandom,non-plannerscouldendupwithhigherratherthanlowerwealthholdingsthanplanners.
35SeeCampbellandMankiw(1990)fortheevidenceonmacrodataandLusardi(1996)fortheevidenceonmicrodata.
See,also,thereviewofthisworkinBrowningandLusardi(1996).
28savinguntilage50orso,whichisalsothetimewhenoldersiblingsstarttoretireandparentsaremorelikelytoexperiencehealthproblems.
367.
ConcludingRemarksInthispaper,Iexaminethewealthholdingsofhouseholdswhoseheadisclosetoretirement.
Contrarytothepredictionsofmanytraditionalmodelsofsaving,Ifindthatalargeshareofhouseholdshavenotthoughtaboutretirementormadeanyplansforretirement.
Lackofplanningresultsinlowwealthholdingsandinportfoliosthatarelesslikelytocontainhighreturnassets,suchasstocks.
Muchresearchisneededtodeterminethereasonswhyhouseholdsdonotplanforretirement,andwhethertheprovisionofinformation,forexample,onSocialSecurityandpensionbenefits,canplayaroleinaffectinghouseholddecisionmakingand,ultimately,thefinancialsecurityofmanyAmericanhouseholds.
36Thereare,ofcourse,alternativeexplanationsforthisfinding.
Animportantoneisrelatedtobuffer-stockmodelsofsaving.
Peopleareprudentbutalsoimpatientand,onaverage,theyholdlittleamountsofwealthuntillateinthelifecycle.
Thismodel,however,doesnotexplainwhysomanypeopledonotplan,evenwhenclosetoretirement,whenuncertaintyshouldberesolved.
Additionally,itdoesnotexplainwhyplanningisinfluencedbytheexperienceofsiblingsandparents.
29DataAppendixThedatausedinthispaperisfromthefirstwaveoftheHealthandRetirementStudy(HRS).
TheHRSisarepresentativesampleofindividualsbornintheyear1931-1941(approximately51-61atinterview),butblacks,Hispanics,andFloridianswereoversampled.
Theindividualdeemedmostknowledegeableaboutthefamily'sassets,debts,andretirementplanningwasaskedquestionsonhousing,wealth,andincome.
OnedistinctivefeatureoftheHRSistheattentionpaidtoexpectationsaboutfutureevents.
Economicmodelsofhouseholdbehavioralmostalwaysincludeasignificantroleforthesevariables,butnotmuchinformationisprovidedinmanycommonlyuseddatasets.
TheHRSincludesabatteryofquestionsaboutsubjectiveprobabilitiesconcerninghealth,longevity,homeprices,changesinSocialSecurity,andsomemacroeconomicvariablessuchasinflationandrecessions.
AsecondinnovationoftheHRSistheuseofbracketingorunfoldingtechniquestoreducethesizeofthemissingdataprobleminthemeasurementoffinancialvariables.
Itiswellknownthatmissingdatarepresentsamajorprobleminsurveymeasurementsofhouseholdwealth.
IntheHRSrespondentswhoreportedtheydidnotknoworrefusedtoprovideanestimateofthesizeofanetworthcomponentwereaskedtoreportthevalueinasetofbrackets.
Smith(1995)andJusterandSmith(1997)reportanevaluationofthesetechniquesandadetaileddescriptionoftheiradvantagesinimprovingtheaccuracyofinformationabouthouseholdwealth.
Toconstructthefinalsample,Ideletedtherespondentswhoarepartiallyorfullyretiredatthetimeoftheinterview.
Ialsodeletedtherespondentsthatdonotreportinformationonthevariablesusedintheempiricalestimation.
Theself-employedarenotaskedmanyofthequestionsaboutsubjectivefutureprobabilitiesandtheyaredeletedfromthesample.
Sincethedistributionoftheratiooftotalandfinancialnetworthtopermanentincomeissowide,Itrimthedistributionandexcludethetopandbottom1%.
Thenumberofobservationsinmyfinalsample3,265.
Thefollowingtablereportssimplestatisticsofthevariablesusedintheempiricalestimation.
30DescriptivestatisticsofthefinalsampleVariablesmean(std.
dev.
)Financialnetworth/permanentincome.
687(1.
211)Totalnetworth/permanentincome2.
411(2.
548)Stockownership.
261(.
439)Havenotthoughtaboutretirement.
237(.
425)Indexofretirementactivities2.
759(1.
638)Age54.
40(3.
857)#ofchildren3.
105(2.
020)#ofchildrenathome.
838(1.
032)Male.
501(.
500)White.
722(.
448)Black.
180(.
384)U.
S.
born.
918(.
274)Married.
658(.
474)Divorced.
175(.
380)Widowed.
083(.
275)Separated.
035(.
184)Northeastregion.
188(.
391)Midwestregion.
249(.
433)Westregion.
152(.
359)Highschool.
383(.
486)Somecollege.
210(.
408)College.
119(.
324)Morethancollege.
099(.
299)Excellenthealth.
257(.
437)Verygoodhealth.
322(.
467)Goodhealth.
290(.
454)Pastunemployment.
365(.
481)Pastshocks.
319(.
466)Receivedinheritances.
180(.
384)Receivedmoneyfromrelatives.
073(.
260)Receivedmoneyfrominsurancesettlements.
053(.
224)Highriskaversion.
654(.
476)Moderateriskaversion.
123(.
329)Mediumriskaversion.
106(.
308)Permanentincome/100050.
410(20.
980)Probabilitytoliveto75.
660(.
278)ProbabilitythatSSwillbelessgenerous.
599(.
295)Probabilitythathousepriceswillgoup.
485(.
292)31Probtogivemajorfinancialhelptofamily.
417(.
312)Descriptivestatisticsofthefinalsample(cont.
)Smoker.
267(.
443)Heavydrinker.
049(.
216)Noregularexercise.
432(.
495)Talkstodoctorsabouthealth.
783(.
412)Bequest.
428(.
495)Parentsstillalive.
685(.
464)Varianceofincome1.
85(7.
155)Canrelyonhelpfromrelatives&friends.
421(.
494)Neverretirecompletely0.
099(.
299)Pensionwealth/permanentincome1.
702(2.
435)SocialSecuritywealth/permanentincome2.
662(2.
267)#ofobservations3,26532ReferencesAinslie,George.
1992.
Picoeconomics,NewYork:CambridgeUniversityPress.
Akerlof,George.
1991.
"ProcrastinationandObedience.
"AmericanEconomicReviewPapersandProceedings81:1-19.
Arnone,William,NormanBarker,RobertGarner,MartinNissenbaum,GlennPape,KennethRouse,andDavidVoss.
1997.
Ernst&Young'sRetirementPlanningGuide,NewYork:JohnWiley&SonsPress.
Banks,James,RichardBlundellandSarahTanner.
1998.
"IsThereaRetirement-SavingPuzzle"AmericanEconomicReview88:769-788.
Barsky,Robert,MilesKimball,ThomasJusterandMatthewShapiro.
1997.
"PreferenceParametersandBehavioralHeterogeneity:AnExperimentalApproachintheHealthandRetirementSurvey.
"QuarterlyJournalofEconomics62:537-579.
Benartzi,ShlomoandRichardThaler.
1999.
"RiskAversionorMyopiaChoicesinRepeatedGamblesandRetirementInvestments.
"ManagementScience45:364-381.
Benartzi,ShlomoandRichardThaler.
2001.
"NaiveDiversificationStrategiesinDefinedContributionSavingsPlans.
"AmericanEconomicReview91:79-98.
Bernheim,Douglas,JonathanSkinnerandStevenWeinberg.
1997.
"WhatAccountsfortheVariationinRetirementWealthAmongU.
S.
Households,"NBERWorkingPaperNo.
6227.
Bernheim,DouglasandJohnKarlScholz.
1993.
"PrivateSavingandPublicPolicy,"inJamesPoterba(ed.
),TaxPolicyandtheEconomy,MITPress:73-110.
Brsch-SupanAxel,andKonradStahl.
1991.
"LifeCycleSavingandConsumptionConstraints.
Theory,EmpiricalEvidence,andFiscalImplications.
"JournalofPopulationEconomics4:233-255.
Bound,John,DavidJaeger,andReginaBaker.
1995.
"ProblemswithInstrumentalVariablesEstimationWhentheCorrelationbetweentheInstrumentsandtheEndogeneousExplanatoryVariableisWeak.
"JournaloftheAmericanStatisticalAssociation90:443-450.
Browning,MartinandAnnamariaLusardi.
1996.
"HouseholdSaving:MicroTheoriesandMicroFacts.
"JournalofEconomicLiterature34:1797-1855.
Campbell,JohnandGregoryMankiw.
1990.
"PermanentIncome,CurrentIncomeandConsumption.
"JournalofBusinessandEconomicStatistics8:265-279.
Chiteji,NginaandFrankStafford.
1999.
"PortfolioChoicesofParentsandTheirChildrenasYoungAdults:AssetAccumulationbyAfricanAmericanFamilies.
"AmericanEconomicReviewPapersandProceedings,89:377-380.
Deaton,Angus.
1992.
UnderstandingConsumption,Oxford:OxfordUniversityPress.
Diamond,PeterandJerryHausman.
1984.
"IndividualRetirementandSavingBehavior.
"JournalofPublicEconomics23:81-114.
Glink,Ilyce.
1999.
"100QuestionsYouShouldAskAboutYourPersonalFinancesandtheAnswersYouNeedToHelpYouSave,Invest,andGrowYourMoney,"NewYork:RandomHouse.
Gustman,AlanandThomasSteinmeier.
1999.
"WhatPeopleDon'tKnowabouttheirPensionsandSocialSecurity.
"InWilliamGale,JohnShoven,andMarkWarshawsky,eds.
Public33PoliciesandPrivatePensions,Washington,D.
C.
:BrookingsInstitution,forthcoming.
Gustman,AlanandThomasSteinmeier.
1999b.
"EffectsofPensionsonSaving:AnalysiswithDatafromtheHealthandRetirementStudy.
"Carnegie-RochesterConferenceSeries,50:271-236.
Haliassos,MichaelandCarolBertaut.
1995.
"WhyDoSoFewHoldStocks"EconomicJournal105:1110-1129.
Hamermesh,Daniel.
1984.
"ConsumptionDuringRetirement:TheMissingLinkintheLifeCycle.
"ReviewofEconomicsandStatistics66:1-7.
HausmanJerryandLynnPaquette.
1987.
"InvoluntaryEarlyRetirementandConsumption,"inGaryBurtless(ed.
),"Work,HealthandIncomeAmongtheElderly,"WashingtonD.
C.
:BrookingsInstitution:151-181.
Heaton,JohnandDeborahLucas.
2000.
"PortfolioChoiceandAssetPrices:TheImportanceofEntrepreneurialRisk.
"JournalofFinance,55:1163-1197.
Hubbard,Glenn,JonathanSkinnerandStephenZeldes.
1995.
"PrecautionarySavingandSocialInsurance.
"JournalofPoliticalEconomy103:360-399.
Hurd,MichaelandKathleenMcGarry.
1995.
"EvaluationoftheSubjectiveProbabilitiesofSurvivalintheHealthandRetirementStudy.
"JournalofHumanResources30:S268-S292.
Juster,ThomasandJamesSmith,1997.
"ImprovingtheQualityofEconomicData:LessonsfromtheHRSandAHEAD.
"JournaloftheAmericanStatisticalAssociation92:1268-1278.
Laibson,David.
1997.
"GoldenEggsandHyperbolicDiscounting.
"QuarterlyJournalofEconomics112:443-478.
Lusardi,Annamaria.
1999.
"Information,Expectations,andSavingsforRetirement,"inHenryAaron(ed.
),BehavioralDimensionsofRetirementEconomics,Washington,D.
C.
:BrookingsInstitutionandRussellSageFoundation:81-115.
Lusardi,Annamaria.
1998.
"OntheImportanceofthePrecautionarySavingMotive.
"AmericanEconomicReviewPapersandProceedings88:449-453.
Lusardi,Annamaria.
1996.
"PermanentIncome,CurrentIncomeandConsumption:EvidencefromTwoPanelDataSets.
"JournalofBusinessandEconomicStatistics14:81-90.
Mariger,Randall.
1987.
"ALife-CycleConsumptionModelwithLiquidityConstraints:TheoryandEmpiricalResults.
"Econometrica55:533-557.
Mitchell,Olivia,JanOlsonandThomasSteinmeier.
2000.
"SocialSecurityEarningsandProjectedBenefits,"inO.
Mitchell,P.
HammondandA.
Rappaport(eds.
),ForecastingRetirementNeedsandRetirementWealth,Phildelphia:UniversityofPennsylvaniaPress:68-94.
Moore,JamesandOliviaMitchell.
2000.
"ProjectedRetirementWealthandSavingAdequacy,"inO.
Mitchell,P.
HammondandA.
Rappaport(eds.
),ForecastingRetirementNeedsandRetirementWealth,Philadelphia:UniversityofPennsylvaniaPress:327-359Munnell,Alicia,AnnikaSundenandCatherineTaylor.
2001.
"WhatDetermines401(k)ParticipationandContributions,"CenterforRetirementResearch,BostonCollege.
O'Brian,Bridget.
2000.
"CalculatingRetirementIt'sNoSimpleEquation,"WallStreetJournal:February7,2000.
34O'Donoghue,TedandMatthewRabin.
1999a.
"DoingItNoworLater.
"AmericanEconomicReview89:103-124.
O'Donoghue,TedandMatthewRabin.
1999b.
"ProcrastinationinPreparingforRetirement,"inHenryAaron(ed.
),BehavioralDimensionsofRetirementEconomics,WashingtonD.
C.
:BrookingsInstitutionandRussellSageFoundation:125-156.
Poterba,James,StevenVenti,andDavidWise.
1994.
"TargetedRetirementSavingandtheNetWorthofElderlyAmericans.
"AmericanEconomicReviewPapersandProceedings84:180-85.
Sheiner,LouiseandDavidWeil.
1992.
"TheHousingWealthoftheAged,"NBERWorkingPaperNo.
4115.
Smith,James.
1995.
"RacialandEthnicDifferencesinWealthintheHealthandRetirementStudy.
"JournalofHumanResources30:S159-S183.
Staiger,DouglasandJamesStock.
1997.
"InstrumentalVariablesRegressionswithWeakInstruments.
"Econometrica65:557-586.
VentiStevenandDavidWise.
1998.
"TheCauseofWealthDispersionatRetirement:ChoiceorChance"AmericanEconomicReviewPapersandProceedings88:185-191.
Venti,StevenandDavidWise.
1997.
"Choice,Chance,andWealthDispersionatRetirement,"mimeo,DartmouthCollege.
Venti,StevenandDavidWise.
1991.
"AgingandtheIncomeValueofHousingEquity.
"JournalofPublicEconomics44:371-395.
Venti,StevenandDavidWise.
1990.
"ButTheyDoNotWanttoReduceHousingEquity,"inD.
Wise(ed.
),IssuesintheEconomicsofAging,Chicago:UniversityofChicagoPress:13-29.
Warshawsky,MarkandJohnAmeriks.
2000.
"HowPreparedAreAmericansforRetirement,"inOliviaMitchell,P.
HammondandA.
Rappaport(eds.
),ForecastingRetirementNeedsandRetirementWealth,Phildelphia:UniversityofPennsylvaniaPress:33-67.
Yakoboski,PaulandJenniferDickemper.
1997.
"IncreasedSavingbutLittlePlanning:Resultsofthe1997RetirementConfidenceSurvey,"EBRIIssueBrief,191,November.
35Table1:ThedistributionofhouseholdwealthPercentileLiquidNetWorthIRAs&KeoghsHousingEquityTotalNetWorth5-6,0000002500027,980506,000042,00096,0007536,00015,00085,000222,20090110,00045,000150,000475,00095199,50075,000200,000785,000Mean(Std.
Dev.
)46,171(178,654)16,492(49,754)61,613(100,646)227,483(521,467)Note:Thistablereportsthedistributionoftotalnetworthanditscomponentsacrosshouseholdswhoseheadis50-61yearsoldandnotfullyorpartiallyretired.
Thetotalnumberofobservationsis5,292.
Allfiguresareweightedusingsurveyweights.
36Table2a:HowdoyoumakedecisionsaboutsavingandinvestmentSourcesTotalsampleLoweduc.
Higheduc.
Callaround0.
250.
220.
30Relatives/friends0.
210.
210.
21Financialplanners/brokers0.
280.
140.
45Accountants0.
070.
020.
14Lawyers0.
030.
020.
04Magazines/newspapers0.
270.
210.
35Materialinthemail0.
110.
080.
15N.
ofobservations798284514Note:Thistablereportsthefractionofhouseholdswhousethesourcesofinformationlistedinthefirstcolumn.
Fractionsarereportedinthetotalsampleofolderrespondents(50-61yearsold)andacrosshighandloweducationgroups.
Thedataisfromthe1995SurveyofConsumerFinancesandallfiguresareweightedusingsurveyweights.
Table2b:SomeunpleasantfactsaboutretirementTheprospectofillness&disabilityNotdoinganythingproductiveBeingboredMissingpeopleatworkworryalot0.
230.
120.
090.
07worrysomewhat0.
290.
170.
160.
23worryalittle0.
260.
190.
170.
29worrynotatall0.
220.
520.
580.
41Note:Thistablereportsthefractionofhouseholdsaccordingtohowtheyratedtheunpleasantfactsaboutretirementlistedinthefirstrow.
ThedataisfromtheHRSandallfiguresareweightedusingsurveyweights.
37Table3:WhothinksaboutretirementHowmuchhaveyouthoughtaboutretirementALotSomeLittleHardlyatAllTotalSampleN.
ofobservations1,3311,0396811,4384,489CharacteristicsLessthanhighschool0.
200.
130.
220.
320.
22Highschool0.
380.
350.
370.
370.
37Morethanhighschool0.
420.
520.
400.
310.
41Familyhashigheducation0.
450.
530.
460.
400.
45Married0.
640.
680.
610.
530.
61N.
ofsiblingsolderthan620.
230.
280.
220.
190.
23Abilitytothinkquickly2.
292.
202.
252.
422.
30Memory12.
913.
812.
912.
613.
1Analogy6.
327.
006.
405.
806.
35Note:Thistablereportsthecharacteristicsofrespondentsacrossdifferentresponsestothequestion:"Howmuchhaveyouthoughtaboutretirement"Allfiguresareweightedusingsurveyweights.
38Table4:ThinkingaboutretirementandtotalnetworthPercentileHowmuchhaveyouthoughtaboutretirementWillneverretirecompletelyAlotSomeAlittleHardlyatall502,010-120-500-3,7002541,30050,50028,5008,80017,57550116,200128,00092,00060,00095,70075241,000266,800208,000147,000259,00090437,000474,500485,700346,500745,00095636,500752,0001,009,000613,3501,335,000Mean(Std.
Dev.
)224,252(504,987)239,298(422,639)245,304(638,957)165,367(448,924)289,960(630,551)N.
ofobs1,3311,0396811,438629Note:Thistablereportsthedistributionoftotalnetworthacrossdifferentresponsestothequestion:"Howmuchhaveyouthoughtaboutretirement"Inthelastcolumn,itreportsthedistributionofwealthforrespondentswhoplantoneverretirecompletely.
Allfiguresareweightedusingsurveyweights.
39Table5:Theeffectsofoldersiblingsandparentsonlackofplanning1234constant0.
2385(0.
0089)0.
2390(0.
0090)0.
2502(0.
0104)0.
3084(0.
0186)agedifferencebetweentheoldestsiblingandtherespondent-0.
0131(0.
0062)-0.
0108(0.
0064)-0.
0111(0.
0064)-0.
0122(0.
0064)agedifferencesquared0.
0004(0.
0003)0.
0004(0.
0003)0.
0004(0.
0003)0.
0005(0.
0003)numberofoldersiblings0.
0234(0.
0100)0.
0314(0.
0109)0.
0300(0.
0109)0.
0289(0.
0108)oldersiblingshaveworsefinancialsituationthanrespondent-0.
0398(0.
0191)-0.
0387(0.
0191)-0.
0383(0.
0190)oldersiblingshavebetterfinancialsituationthanrespondent-0.
0168(0.
0168)-0.
0143(0.
0168)-0.
0140(0.
0167)pastshocks-0.
0333(0.
0160)-0.
0325(0.
0160)parentslivedinnursinghomesbeforedying-0.
0624(0.
0174)parentshadanillnesslasting3monthormorebeforedying-0.
0507(0.
0194)R20.
0020.
0030.
0050.
012Note:Thistablesreportstheregressionsoflackofplanningforretirementonthevariableslistedinthefirstcolumn.
Thedependentvariableisadummyvariable,whichequalsonewhenrespondentsreporttheyhavehardlythoughtaboutretirement.
40Table6:ExplaininghouseholdsavingsLackofplanningdummyPlanningindexTotalNetWorthFinancialWealthTotalNetWorthFinancialWealthCoeff.
Std.
errCoeff.
Std.
errCoeff.
Std.
errCoeff.
Std.
errconstant1.
94076.
53972.
51683.
3177-2.
53676.
4063-0.
22313.
2651planning-2.
30441.
1005-1.
40860.
55830.
90180.
42410.
55840.
2161excellenthealth0.
56440.
18200.
18620.
09230.
62460.
19610.
22440.
0999verygoodhealth0.
15920.
16160.
08080.
08200.
25130.
17150.
13790.
0870goodhealth0.
16890.
16200.
09920.
08220.
24830.
17870.
14900.
0911permanentinc.
/1000-0.
02750.
0074-0.
01030.
0038-0.
03410.
0093-0.
01440.
0047pastunemployment-0.
10810.
10300.
00360.
0523-0.
15000.
1070-0.
02230.
0546pastshocks-0.
57830.
1113-0.
31490.
0565-0.
59060.
1159-0.
32320.
0591receivedinheritances0.
57080.
12780.
20610.
06480.
51320.
13860.
17010.
0706moneyfromrelatives0.
56730.
18400.
18120.
09340.
46260.
20490.
11560.
1044moneyfrominsurance1.
00650.
22160.
54590.
11241.
06100.
22890.
57960.
1166highriskaversion0.
04410.
1540-0.
07790.
07810.
05500.
1590-0.
07040.
0810mediumriskaversion0.
06870.
1881-0.
11260.
09540.
05840.
1934-0.
11920.
0985moderateriskaversion0.
11400.
2142-0.
05010.
10870.
17330.
2310-0.
01190.
1177varianceofincome0.
00320.
00620.
00300.
00320.
00310.
00640.
00290.
0033prob.
liveto750.
22010.
19100.
07490.
09690.
17050.
18740.
04540.
0955prob.
SSmoregener.
-0.
08460.
15940.
02210.
0809-0.
15830.
1707-0.
02380.
0870prob.
housepriceup-0.
08980.
1681-0.
10820.
0853-0.
11900.
1695-0.
12560.
0864prob.
givehelptofam.
0.
08660.
15340.
02200.
07780.
06690.
15830.
00960.
0807bequests0.
98790.
09800.
27720.
04970.
94370.
10630.
24950.
0542canrelyonhelp-0.
01080.
09630.
01040.
04890.
00210.
09790.
01810.
0499parentalive-0.
13510.
1090-0.
06090.
0553-0.
17720.
1137-0.
08700.
0580smoker-0.
29910.
1107-0.
09140.
0561-0.
30590.
1127-0.
09520.
0574heavydrinker-0.
26070.
2234-0.
05890.
1133-0.
34990.
2414-0.
11500.
1230noregularexercise-0.
13560.
1016-0.
04140.
0516-0.
09170.
1096-0.
01400.
0559talktodocabouthealth0.
15950.
12830.
07900.
06510.
10210.
14430.
04240.
0735pension/permincome0.
07290.
03050.
01170.
01550.
03230.
0463-0.
01380.
0236SSwealth/perminc.
0.
08030.
02340.
01180.
01190.
08020.
02400.
01180.
0122AdjustedR20.
07000.
01000.
02100.
0009pvalueofoverid.
test0.
2320.
9090.
2990.
953Note:Thistablereportsinstrumentalvariablesregressionsoftotalandfinancialnetworthoverpermanentincomeonthevariableslistedinthefirstcolumn.
Eventhoughnotreported,regressionsincludeseveraldemographicvariables.
Refertothetextforacompletelistofthevariablesandalistoftheinstrumentsusedintheestimation.
Twoindicatorsofplanningareusedintheestimation:adummyvariable,whichequalsoneifrespondentshavehardlythoughtaboutretirementandanindexforplanning.
41Table7:PlanningandstockownershipLackofplanningdummyPlanningindexCoeff.
Std.
errCoeff.
Std.
errCoeffStd.
errCoeff.
Std.
errconstant0.
45370.
23610.
37570.
23240.
06740.
13600.
02910.
1333planning-0.
45770.
1845-0.
41110.
19000.
08120.
04030.
07150.
0452excellenthealth0.
09060.
03090.
08860.
03070.
08460.
02950.
08280.
0300verygoodhealth0.
08610.
02690.
08540.
02640.
08370.
02550.
08320.
0252goodhealth0.
05290.
02680.
05210.
02650.
04790.
02530.
04740.
0253pastunemployment-0.
01570.
0174-0.
01470.
0169-0.
01750.
0166-0.
01770.
0161pastshocks-0.
02580.
0183-0.
02320.
0185-0.
01400.
0162-0.
01290.
0165inheritances0.
06660.
02110.
06860.
02080.
07270.
01990.
07420.
0196highriskaversion0.
00390.
02570.
00280.
0261-0.
02370.
0236-0.
02060.
0230mediumriskavers.
-0.
00540.
0312-0.
00660.
0307-0.
02560.
0309-0.
02390.
0302moderateriskavers.
0.
03340.
03560.
02690.
0361-0.
00470.
0307-0.
00630.
0304permanentinc/10000.
00270.
00120.
00260.
00120.
00230.
00130.
00230.
0013prob.
liveto750.
00950.
03060.
00770.
0302-0.
00780.
0278-0.
00740.
0276prob.
givefin.
help.
0.
00280.
00250.
00260.
00250.
00370.
00240.
00350.
0024bequests0.
03770.
01610.
03430.
01580.
03630.
01550.
03320.
0152varianceofincome0.
00170.
00110.
00140.
00110.
00150.
00100.
00130.
0010parentsarealive-0.
00190.
0181-0.
00140.
0178-0.
00810.
0174-0.
00650.
0172smoker-0.
02210.
0181-0.
01920.
0177-0.
01560.
0181-0.
01380.
0180heavydrinker-0.
00710.
0363-0.
00700.
0359-0.
00430.
0344-0.
00480.
0345noregularexercise-0.
00970.
0168-0.
01100.
0165-0.
01000.
0161-0.
01150.
0159talkdocabouthealth0.
01120.
02250.
01440.
02170.
02260.
02020.
02570.
0196grossfin.
wealth0.
00220.
00040.
00220.
00040.
00200.
00050.
00200.
0005fin.
wealthsq.
/1000-0.
00410.
0012-0.
00410.
0012-0.
00400.
0013-0.
00410.
0013def.
benef.
/perm.
inc-0.
00590.
0052-0.
00770.
0071def.
contr.
/perm.
inc.
0.
04560.
00950.
04440.
0095otherpens.
/perm.
inc0.
00430.
00960.
00870.
0088AdjustedR20.
03500.
07300.
13100.
1520pvalueofoveridtest0.
49600.
39400.
26000.
2010Note:Thistablereportsinstrumentalvariablesregressionsofstockownershiponthevariableslistedinthefirstcolumn.
Eventhoughnotreported,regressionsincludemanydemographicvariables.
Refertothetextforacompletelistofthevariablesandalistoftheinstrumentsusedintheestimation.
Twoindicatorsofplanningareusedintheestimation:adummyvariablewhichequalsoneifrespondentshavehardlythoughtaboutretirementandanindexforplanning.
42Table8a:RetirementandplanningHowmuchhaveyouthoughtaboutretirementHowhasyourretirementturnedouttobeAlotSomeAlittleHardlyatallVerysatisfying0.
680.
500.
350.
22Moderatelysatisfying0.
280.
410.
460.
35Notatallsatisfying0.
040.
090.
190.
43N.
ofobservations34321792520Table8b:RetirementandplanningHowmuchhaveyouthoughtaboutretirementHowisyourretirementcomparedtotheyearsjustbeforeyouretiredAlotSomeAlittleHardlyatallBetter0.
570.
440.
350.
18Aboutthesame0.
220.
310.
360.
24Notasgood0.
110.
150.
220.
54Retiredlessthan1yearago0.
100.
100.
070.
04N.
ofobservations34321792520Note:Thesetablesreportthefractionofrespondentsaccordingtohowtheyhaveratedretirementandhowmuchtheyhavethoughtaboutretirement.