family transfers involving three generations
TRANSCRIPT
7/28/2019 Family Transfers Involving Three Generations
http://slidepdf.com/reader/full/family-transfers-involving-three-generations 1/30
Family Transfers Involving Three GenerationsAuthor(s): Luc Arrondel and André MassonReviewed work(s):Source: The Scandinavian Journal of Economics, Vol. 103, No. 3, Intergenerational Transfers,Taxes and the Distribution of Wealth (Sep., 2001), pp. 415-443Published by: Blackwell Publishing on behalf of The Scandinavian Journal of EconomicsStable URL: http://www.jstor.org/stable/3441081 .
Accessed: 28/03/2012 05:44
Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at .http://www.jstor.org/page/info/about/policies/terms.jsp
JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of
content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms
of scholarship. For more information about JSTOR, please contact [email protected].
Blackwell Publishing and The Scandinavian Journal of Economics are collaborating with JSTOR to digitize,
preserve and extend access to The Scandinavian Journal of Economics.
http://www.jstor.org
7/28/2019 Family Transfers Involving Three Generations
http://slidepdf.com/reader/full/family-transfers-involving-three-generations 2/30
Scand. J of Economics 103(3), 415-443, 2001
FamilyTransfersInvolvingThreeGenerations*
LucArrondelCNRS-DELTA,-75014-Paris, [email protected]
AndreMassonCNRS-DELTA, -75014-Paris, [email protected]
Abstract
Most models of familytransfers onsideronly two generations nd focus on two motives:altruismandexchange.Theyalso assumeperfectsubstitution etween nter-vivosdownwardtransfersand bequests.Based on Frenchevidence,we show that parent-to-childransfers
belongto threedistinctcategories investmentn child'seducation,inancial ssistance,wealth
transmission),nd advocatea three-generationramework. hus,transfer ehaviorof parentstowardheirchildren s stronglynfluenced ythebehavior f theirownparents.There s also
some evidenceof the Cox and Starkdemonstration ffect:parentshelp their own parents,expecting o receivecomparableupportrom heirchildren.Suchbehavior anbe regardedsindirect eciprocity:hebeneficiary oesnotgivebackto the initialgiverbutto a thirdpersonof another eneration.
Keywords: ntergenerationalransfers; ltruism; xchange;ndirect eciprocity; emonstrationeffect
JELclassification:D10; D31;D63; D64
I. Introduction
Thebulkof theliteraturen intergenerationalamily inkshasfocusedonthedeterminantsf parent-to-childransfers,akingonlyintoaccount hecharacteristicsf these wogenerations.wingo dataimitations,heearlymodelsdealtmostlywith(post-mortem)equests;morerecently,ew datasetshaveencouragedshiftof interestowardsnter-vivosransfers,hichhave the advantagef being "voluntary".n this lattercase, two main
*A previousversion of this paperwas presentedat the conferenceon "IntergenerationalTransfers, axes,andDistribution f Wealth",Uppsala,June1999,wherewe receivedhelpfulcomments romFrankStaffordand otherparticipants.This version has benefited rom thefruitfulremarksand suggestionsof TullioJapelli,Chris CarollandJoe Altonjiat the TMRmeeting "Savingsand Pensions", Venice-Padua, eptember1999. The commentsof twoanonymousefereeswereveryconstructive.
?The
editorsof
the Scandinavianournalof Economics2001. Publishedby BlackwellPublishers, 08 CowleyRoad,OxfordOX4 1JF,UK and 350 MainStreet,Malden,MA 02148, USA.
7/28/2019 Family Transfers Involving Three Generations
http://slidepdf.com/reader/full/family-transfers-involving-three-generations 3/30
416 L.Arrondeland A. Masson
competing motives have been emphasized: altruism, where transfers aremade in order to even resources or well-being and smooth consumption
between family members, cf. Becker (1991), and (self-interested) exchange,where parentsuse gifts or a promise of inheritance as payment for a child's
services-insurance, supportor "attention"duringtheir old age; cf Kotlikoffand Spivak (1981), Bemheim, Shleifer and Summers (1985) and Cox
(1987).To disentangle these two motives, most tests concern the (negative) effect
of the recipient's income, for given parents' resources, on the probability orthe amount of the transfer,where the transfer is "compensatory" between
parent and child or between siblings; see Altonji, Hayashi and Kotlikoff(1997). The distinction appearscrucial when assessing the effects of govern-ment transfer policies. Under altruism, public redistributionis neutralized,
crowding out privatetransfers to the same beneficiary;on the otherhand,the
impact of public redistribution on exchange-motivated bequests is more
ambiguous.1The motivation for this paper,which is still exploratory,comes from two
main sources of dissatisfaction with this basic framework.The first concerns the assumptionthat various inter-vivos transfers(made
at different stages of the life cycle), as well as inter-vivos transfers and
bequests are quasi-perfectsubstitutes:besides the choice of the right timing,2only the total present value of parent-to-childtransfers has to be considered.Even though it makes modeling and predicting far easier, this view appearsquite unrealistic, owing to the strong heterogeneity of inter-vivos transfers.Free disposal of a home or payment of college fees for a 20-year-oldindependent child may be a substitute for co-residence. But major (official,
declared) gifts, received some 20 or 30 years later, play the role of an
'See McGarry1999) or Arrondel,Massonand Pestieau 1997). Parental ltruism,however,will havedifferentdistributionalonsequenceswhenparents reliquidity onstrained,makingnobequest.Governmentntervention an thenbe beneficial: ccordingo BeckerandMurphy(1988), public expenditures n the young and the elderlyhave provedto be both Pareto-
improving ndmoreequitable, spartof a "socialcompact"betweengenerationswhich "triesto achieve for poorerand middle-level[constrained]amilieswhat richer familiestend toachievewithoutgovernmenthelp: namely,efficient levels of investment n children and
supporto elderlyparents";ee alsoEhrlich ndLui(1991).2Altruistic arentswill thus have a strong ncentive o helpliquidity-constrainedhildren,butwill otherwisehave a preference or delayingtransfersas long as possible, owing to the
uncertaintybout heir child'sfuture ncome and the desireto have the lastword,preventingtheirchild fromoverconsumingwhenyoung;see Altonjiet al. (1997). In the same vein,McGarry1999) argues hat altruistic equestscould be positivelycorrelatedwiththe child'scurrent ncome: f the child's ncome ncreases,mplying hatshe is less liquidity-constrained,parentswill accordinglymake lower inter-vivosransferso her,so that in the last periodoftheir ife theywill havemoreresources eft for theirownconsumption,ut also forbequests.
? The editors of the Scandinavian Journal of Economics 2001.
7/28/2019 Family Transfers Involving Three Generations
http://slidepdf.com/reader/full/family-transfers-involving-three-generations 4/30
Family transfers nvolvingthreegenerations 417
anticipatednheritanceand are more likely to be substitutes or bequests.These two types of transfershave differentdeterminants nd could corre-
spond o differentmotives.The second shortcoming oncernsthe narrow ocus of most models on
bilateral intergenerationalrelations-between parents and children-
althoughrecent studiesadvocateexpandinghe analysisof familybehaviorfromtwogenerationso three; ee Coxand Stark 1998).
French videnceonparent-to-childransfersupportshisshift of perspec-tive. Gifts andbequestsdo not appear o be compensatory etweenparentsandchildren; ee Arrondel t al. (1997). Moreover,n France, here s little
scopefor redistributionf wealth ransfers etweensiblings.Deviation romequal sharing s bothstrongly imitedandcostly,owingto legal obligationsandthedistortions reatedby an inheritanceaxbasedon bloodrelationship.(Forthe sameinheritance,he tax canbe nil or below 20 percent or a son,55 percent or a nephew,and 60 percent or a strangero thefamily!)
Ontheotherhand,strongandhighlysignificant etrospective ffects,both
qualitativeand quantitative, avebeen systematicallyound for downwardtransfersand transmissionpractices.Obtained rom various data sets bydifferent
authors,these effects
appearrobust to alternativeeconometric
specifications; ee Arrondelet al. (1997). Hence,what is left in bequests(anddeclaredgifts) to childrenappearscommensuratewith what has beenreceived from parents,i.e., a higher life propensityto bequeathout ofinheritancehanout of humanresources.Moreover, ransmissionpatternsand behavior end to be reproducedromone generationo the next:every-thing being equal, inheritance eceivedthrougha will increasesthe prob-abilityof makinga will, doneesaremorelikely to becomedonors,and soon.
As far as child-to-parenttime or financial) ransfersare concerned, hemost strikingevidence of retrospective ffects concerns co-residence(co-habitation)with old parents. Based on the French "Caisse Nationaled'AssuranceVieillesse"(CNAV) urvey see below),Wolff 1998)has foundthatagedrespondentswhose ownparentshadmoved n withthem n thepasthad a significantlyhigher probability f being partof a child'shousehold.Similar indingshave been obtainedby Cox and Stark 1998) on US data:"the incidenceof sharinghousingwith parentswas 27 percenthigherfor
respondentswhose grandparentsadmovedin when the respondentswerechildren".
Infact,Cox and Stark nterprethis lastresultas a "demonstrationffect"on the part of forward-lookingparents (P), when they were the middlegenerationn familiesof threeoverlapping enerationswithtwo dependentgenerations.They helped their own parents(G) in the presence of theiryoungchildren K), in order o set an examplefor their children.The ideawas to inculcate n themthevaluesor feelings-guilt, obligation, iliallove,
O The editors of the Scandinavian Journal of Economics 2001.
7/28/2019 Family Transfers Involving Three Generations
http://slidepdf.com/reader/full/family-transfers-involving-three-generations 5/30
418 L.Arrondeland A. Masson
respectfor theirparents,etc.-that wouldlater secureold-agesupport ortheirparents,nthe formof care,attention r financial ransfers.
Such positive correlationsbetween past and presentsimilar transfers,whetherupwardor downward nes, can be given quitedifferent nterpreta-tions.An encompassing ndfruitful rameworks providedby a specialkindof (deferred)reciprocity hat anthropologistsall indirectreciprocity; eeMauss 1950).Thebeneficiary oesnotgivebackto the initialgiverbut to athirdperson theinitialgiverhavingbeenrewarded y anotherperson).Thisformof exchangerelies on a processof debt creationandpayment.Theheiris in debt to the familylineage,with an obligation o give backthat cannot
be fulfilled o thedeceased; heonly wayto redeem he debt s to payit backto one'schildren,as if bequestswere madein return or received nheritancefromone'sparents.Likewise,accordingo theCox andStarkdemonstrationeffect, parents,by helping their own parents, acquirea claim on theirchildren,as if theycouldprecommitheiroffspring o supporthemin theirold age (pay-as-you-go etirementystemsrest on the samelogic of recip-rocity).
SectionII is mainlytheoretical, oncerninghe heterogeneity f transfersand the
implicationsof a three
(or more) generationsramework.Three
kinds of parent-to-childinancial inter-vivos transfersare distinguishedaccording o the objectivepursued: ducation,assistance,andwealthtrans-mission, with specific determinants nd timing. Consequently,he appro-priatedivisionof the life cycle is intofour periods,accordingo a "KCPG"scheme: G representsold age (grandparent)nd K childhood(kid). Theintermediatepanis divided nto two periods:C correspondso the begin-ning of adult(child)life, when one leavesparents, tarts o work andmayhavekids,whereasP representsparents)matureagebeforeretirement,with
no more births.3Between hreeoverlapping dultgenerationsresp.agedC,P and G), numerous upwardor downward)nter-vivos ransfersmaytake
place, but thereappear o be only four broadcategoriesof "economicallyrelevant"ransfers, epictedn Figure1.
Dataand initial results arepresented n SectionIII. The FrenchINSEE1992wealthsurvey,with a nationally epresentativeampleof some 10,000households,shows that an individualof age P, around50 years old, canbelongto manydifferent amily generational ompositions.The most fre-
quent (45 percent), denoted (K)CPG, has three adult generations (thefourth,as a kid K, mayor maynot exist).The CNAV1992surveyprovidesdetailedinformationon intergenerationalransfers or around1,000 such
(K)CPG families, where each of the three adult generationshave beeninterviewed eparately.SectionIV reportsan econometricanalysisof the
3The "demonstration effect" instead requires a division of the life cycle into only three
periods, according to a KPG scheme.
? The editors of the Scandinavian Journal of Economics 2001.
7/28/2019 Family Transfers Involving Three Generations
http://slidepdf.com/reader/full/family-transfers-involving-three-generations 6/30
Family transfers nvolvingthreegenerations 419
Pastperiod Present period Nextperiod
(deceased) 0() G
XParent 3 |
Grandpren( (K) (4)
Fig.amily......flowsetwe three adult geeratios.........
Fig. 1. Family flows between threeadult generations
probability of the different transfers, mainly based on the two data sets
(amountsare not considered).The results confirmthe heterogeneityof inter-vivos downward transfers (education, assistance, wealth transmission), and
show the prevalence of retrospective and other effects which may be
interpreted as indirect reciprocities between three generations. Section V
concludes, emphasizingthe implications for public redistribution.
II. Theory: Heterogeneous Transfers and Indirect Reciprocities
Inter-vivos transfersmay consist of time transfersor financialtransfers.The
child is assumed to be an "adult" (C), meaning she must be above a
minimum age (18 or 22) and/or in an independent household. Moreover,"financial"transfers,especially from parentsto children,cover a large rangeof transactions:they include not only gifts of various assets, cash transfers
made once or regularly, alimony, but also in-kind transfers, such as the
paymentof rent, free disposal of a home, college fees, loans, cosignaturefora home mortgage,which have no obvious equivalentcash value.
?The
editors of the Scandinavian Journal of Economics 2001.
7/28/2019 Family Transfers Involving Three Generations
http://slidepdf.com/reader/full/family-transfers-involving-three-generations 7/30
420 L. ArrondelandA. Masson
Timing and Heterogeneity of Parent-to-child Inter vivos (Financial)Transfers
The heterogeneous nature of parent-to-childinter-vivos transfersleads us to
propose a division into three types according to the main objective pursued:education, assistance or transmission.
(i) Investments in the human capital of the child are received early in herlife cycle; these transfersadd to the child's presentand future income.
(ii) Financial assistance is often due to imperfect capital or insurancemarkets. Parentshelp liquidity-constrainedchildren by increasing their
present resources or extending their access to credit, especially for thepurchase of a home; they may offer a partial insurance against their
offspring'srisks of unemployment, divorce, etc.; they may also care for
their young grandchildren.These transfers,which are mainly received
by children still in the first partof their working life (C), primarilyaddto their consumption.
(iii) Inter-vivos transfers that are partof the wealth transmissionprocess are
often received laterin life, and take the form of stocks rather hanregular
flows. Often interpretedas an anticipated inheritance and as adding tothe child's wealth, they are therefore referredto as wealthgifts.
This division is clearly reminiscent of the Kotlikoff-Modigliani debate,relative to the share of "inheritedwealth" in total existing accumulation;see
Kessler and Masson (1989). Modigliani (1988) includes only "major" gifts
"adding to child's wealth" (case iii) in intergenerationaltransfers,whereas
Kotlikoff (1988) also counts as transfers "minor" gifts or financial help
"increasing child's consumption" (case ii), as well as college education fees
and other parents' spending (case i) as if investments in a child's educationand bequests were perfect substitutes; see Becker (1991). Although the
dividing line drawn by Modigliani between "minor" and "major" giftsseems somewhat artificial, most authors side with his position, advocatingthe limited degree of substitutability between education transfers and
financial assistance, on the one hand,and wealth gifts on the other.
Each type of transfer indeed leads to specific predictions. Since wealth
gifts are more likely to be substitutes for bequests, they should most often be
shared equally among siblings, especially in France; only the rich will beable to pay the "costs" of unequal division. Moreover,a specific motive for
wealth gifts should be to reduce estate taxation, implying that their prob-ability of occurrence increases with parents'wealth, or better taxable wealth.
The effect of a child's income is more ambiguous, depending on thedominantbequest motive.
Unequal sharing is likely to be more common for education transfers(depending on a child's ability) and financial assistance (depending on a
O The editors of the Scandinavian Journal of Economics 2001.
7/28/2019 Family Transfers Involving Three Generations
http://slidepdf.com/reader/full/family-transfers-involving-three-generations 8/30
Family transfers nvolvingthreegenerations 421
child's needs). Here only wealthy parents can, afford equal division between
grown-up children; see Hochguertel and Ohlsson (1999). The effect of the
child's income on education transfersis dubious, owing to income endogene-ity. But the probability of financial assistance should be compensatory,
regardlessof whether it is motivatedby altruism or exchange. In particular, t
should be negatively correlated to a child's current ncome; see Cox (1987).These predictions should be kept in mind when interpretingthe results
derived in US studies of inter-vivos transfers.Usually, the reference periodconsidered is really short-transfers given or received duringthe last year or
so-and the minimum threshold is quite low (US$100, 200 or 500). It is
therefore clear that such transfers concern mainly financialhelp or assistance(not major, declared gifts). The usual findings of (slightly) compensatoryinter-vivos transfers,more (often) benefiting less well-off children, should be
interpretedwith these qualifications.4Another key feature of heterogeneous inter-vivos transfersis their timing:
the grown-up child receives education transfers,or even financial assistance,much earlier in life than wealth gifts. However, even wealth gifts and
inheritance are most often received before retirement (at a mean age of 38
for gifts, and 45 for inheritancein France). Hence, the idea of dividing adult
working life into two stages, C and P: education transfers,and a large partof financial assistance, arereceived at age C, when you become independent,while wealth gifts are more likely to be received at age P, when old parentswant to speed up the transference of their property.The "representative" ife
cycle is thus of a KCPG type, divided into four periods with K denotingchildhood and G the retirementperiod. Families may have four overlapping
generationswith a middle generation of age P.5
Typology of Family Transfers between Three Adult Generations
We concentrate on inter-vivos, time and financial transfers in (K)CPGfamilies (with three or four generations), between three adult generations,
4Cox (1987), for instance,considersany paymentmadeby parentsduringJanuary-August1979 between"familyunits",whereanychild above 18 is a separate amilyunit(evenif shelives with her parents).Altonjiet al. (1996, 1997), Dunn(1997) and McGarry 1999) use
similardata;onlyCox andRank 1992)consider ransfers boveUS$200madeduring helastfiveyears.5How should the frontiersbetween these four stagesof equal lengthbe determinedn an
overlapping enerationsramework? he recent endency eemsto postpone hebeginningofadult ife (thebreakpointbetweenK andC) to the date of departureromtheparents'homeor, instead, o the age of 22, whichimpliesthatmostparent-to-childransfers ccurbetween
independent ouseholds;he FrenchCNAVsurveydoes precisely hat. Withthis convention,theKCPGdivisionof the life cyclebecomes:age K before22 yearsold,age C from22 to 43,ageP from44 to 65, age Gthereafter;ee Gokhale,Kotlikoff,SeftonandWeale 1999).
? The editors of the Scandinavian Journal of Economics 2001.
7/28/2019 Family Transfers Involving Three Generations
http://slidepdf.com/reader/full/family-transfers-involving-three-generations 9/30
422 L. Arrondeland A. Masson
generally living in independenthouseholds: adult children (C), their parents(middle generation P), and their retired grandparents (G). Such transfers
have the advantage of being both (more) free and observable, especiallywhen compared to transfers benefiting young kids who live with their
parents. The problem concerns their potential number, which is far too
large: to reduce it, only significant and economically relevant transfers
should be selected. For that purpose, we rely on some empirical figures
concerning the observed relative frequencyor the quantitative importanceof
various transfers. These figures are based on the CNAV 1992 survey on
French(K)CPG families presentedbelow.
Consider first financial transfers.We can safely disregardthose given bythe young generation. Less than 2.7 percent of adult children report having
given (generally limited) financial help to their parents of the middle
generation and even fewer parents report having received a transfer. This
means that only one-fourth of the proportion of middle generation house-holds help their own parents(10.8 percent). Furthermore,only 0.3 percent ofchildren help(ed) their grandparents.We hence consider only one type of
upwardtransfers,from the middle to the old generation.
Contrarilyto Bemheim et al.
(1985)and
others,in our
empirical analysiswe do not include among transfers family relations ("attention") such as
contacts and visits, especially between the middle and the old generation
(unless the latter has health or invalid problems); in fact, they often represent
reciprocal exchange, going both ways simultaneously. More specific time
transfers (including co-residence) may present similar pitfalls, albeit to a
smaller degree: some appearmore like simultaneous exchange of services;others seem quite anecdotal (gardening, etc.). Moreover, decisions concern-
ing time transfers areheavily constrained,depending stronglyon the distance
between family members. We therefore consider "economically relevant",
only those time transfers-of significant diffusion and importance-whichbenefit (physically or economically) dependent or needy individuals: old
(unhealthy) parents, pre-adult children with no personal resources, young
grandchildren,etc. These restrictive qualifications allow us to ignore time
transfers received by the middle generation. Moreover, among possible
upward time transfers, only parent-to-grandparentsupport is taken into
consideration (child-to-grandparent support is negligible).6 Finally, the se-
6The CNAVsurveyand the PSID surveycontain similar information bouttime transfersreceivedorgivenover the lastyearssee Altonjiet al. (1996).A quaterof middle-generationhouseholds ive supporto theirelderlyparentsorparents-in-law),utonly5.7 percentof the
young generationdeclare hattheyhelp theirgrandparents.Moreover,he help given by themiddlegenerations much moreimportant:whenthe spouseis not the "principal" elper,the latter s a child nmore han90 percentof thecases,but a grandchildn less than2 percentof thecases(theremaining percentbeingfriends,neighbors, rotherrelatives).
? The editors of the Scandinavian Journal of Economics 2001.
7/28/2019 Family Transfers Involving Three Generations
http://slidepdf.com/reader/full/family-transfers-involving-three-generations 10/30
Family transfers nvolvingthreegenerations 423
lected downwardtime transfers concern only care or education of pre-adultchildren or grandchildren.7
Let us now introduce the following notations: the currentmiddle genera-tion (age P) is denoted 2, its parents belong to generation 1, its children to
generation 3, and so on; so that ego, the "reference respondent" of the
middle generation, is now in position P2, her parents in position G1, and her
children in position C3; while one period ago, ego was in position C2, as
indicated in Figure 1. The preceding discussion leads us to retain only fourbroad categories of economically relevant inter-vivos transfersbetween the
three adult generations,with the following content:
(i) Type a (ascending) includes upwardtransfers from P2 (middle) to G1
(old): either financial or, more often, time transfers when old parentsare dependentor needy;
(ii) Type h (help) concerns downward transfers from P2 (middle) to C3
(young): either financial, care of grandchildren (in position K4), or
final outcome of child's education (now in position C3);
(iii) Type g (gifts) contains downward,mostly financial, transfers from G1
(old) to P2 (middle);
(iv) Type s is generation skipping, downward and mostly financial trans-fers, from G1to C3.
These transfersappearin the central rectangle of Figure 1. Where necessary,we add type b, expected post-mortem bequests, received by generation 2 at
the beginning of her retirement.
The usual literature, using a bi-generational framework, independently
analyzes transfers between the old (G1) and middle (P2) generations, and
those between the middle (P2) and the young (C3), while overlookinggeneration skipping. This is especially the case with bequest models, which
always give the leadership to the older generation. The crucial point,however, is that many alternative approaches are possible, which mayconsider the relations between the three adult generations at the same timeand/or give the leading role to younger generations. A sound empiricalprocedure is therefore to begin by trying to eliminate some of these many a
priori views, before considering more elaborate tests of specific models oftransfers. For instance, views of generalized risk sharing,credit expansion or
7Sincewe limitourselves o transfers etweenadultgenerations,we assume,by convention,that caregivento a pre-adult randchild age K), is receivedby the child(age C). Moreover,educationof pre-adult hildrens not directly aken ntoaccount;only its finaloutcome-thecurrent evel of educationof the now adultchild-is partof the tranferreceivedfrom herparentsbelonging o the middlegeneration.
? The editors of the Scandinavian Journal of Economics 2001.
7/28/2019 Family Transfers Involving Three Generations
http://slidepdf.com/reader/full/family-transfers-involving-three-generations 11/30
424 L.ArrondelandA. Masson
redistribution etween the threegenerationsdo not fit Frenchdata,whichcannot explain the prevalence of retrospective effects, or the asymmetryof
financial ransfers,.e., upward ransfers re far less significant handown-ward ransfers espiteeconomicgrowth.
"Indirect Reciprocities" between Generations
Preliminarynvestigation f Frenchdata ndicates hatthe view of transfersas partof indirectreciprocitiesbetweenthreegenerationss a constructive
pointof departure;ee Arrondeland Masson(1999a).Basedon some kindof intergenerationalcontract",hese formsof generalized xchangeallowus to encompassvarious ypesof retrospectivend demonstrationffects,aswell as mechanisms f preferenceormation nd nculcation f familyvaluesor socialnorms; ee Becker 1996).
Indirectreciprocitiesare based on a simple principleof reciprocal ri-lateralexchange; ee Mauss(1950). Supposegenerationa gives somethingto generationb. Generation thenacquiresa claim on the familialgroup,that will be-or has
alreadybeen-paidback in a
comparableorm bya
thirdgeneration;ikewise,generationb becomesin debt to the family,andwill be "obliged" o payit back,that s to replicatehe samekindof transferto a fourthgeneration.This process of debt creation and paymentmayengender ndefinitechains of mutuallyadvantageous ooperationbetween
generations;Masson (1999). Indirectreciprocitiesmay concern upwardor downwardransfers;moreover,hey maybe "backward-looking",hencurrent ransfers re influencedby past similar ransfers negeneration go,or "forward-looking",hen current ransfers re madein expectation hat
thesametypeof transferswill be reproducedn onegeneration'sime. Let usillustrate.
Forward-lookingndirectreciprocitiesentail an "initial obligationto
give". The Cox and Starkdemonstrationffect is an upward orm.Supportof own parentshas a positive externality:t increasesthe probabilityhatchildrenwill later do the same. Such reciprocitiesare motivatedby self-survival.Theproblemof enforcementoncerns hepossibility,n onewayor
another,to precommitfuture generationsin a long-termcontract,for
instanceby giving them "appropriateeliefs" as in the Hammond 1975)pensiongame.Likewise,parent-to-childransfers an be motivatedby twodifferentforms of altruism.Parentsmay only care about their child's
consumption: heir horizon extends over two generations.But forward-
looking parents,who care abouttheirchild'sutility,are endowedwith aninfinite dynastic horizon, since they expect their childrento exhibit a
comparabledegreeof altruismand to adoptsimilarbequeathingbehaviortowards heirownchildren,and so on. Thiscanbe interpreteds avariant f
? The editors of the Scandinavian Journal of Economics 2001.
7/28/2019 Family Transfers Involving Three Generations
http://slidepdf.com/reader/full/family-transfers-involving-three-generations 12/30
Family transfers nvolvingthreegenerations 425
downwardand forward-lookingndirectreciprocities,which appearmoti-vatedby lineagesurvival.
Backward-lookingeciprocities ntail a "finalobligation o give back".Thus, "retrospective" equests,as in Arrondelet al. (1997), emphasizingthe influenceof family background r history,maybe seen as a downwardvariantof such reciprocities, orrespondingo the following implicitrule:"Do untoyourchildrenas you would have likedyour parents o have doneuntoyou".8
We considervarious ndirectreciprocities etween,especially, hreeadult
generations,with reference o thecurrentmiddlegenerationin positionP2).
Hence,Figure1, drawnn anoverlappingenerationsramework,epresentsthree time periods:preceding forbackward-lookingeciprocities), urrent,andnext(for forward-lookingeciprocities).Past inter-vivosransfers, ivenone generationago, are denoted-a, *h,*gand*s;expectedtransfers,n one
generation's ime are, denoted a*, h*, g* and s*; finally,since type btransfers epresentnheritance xpectations or the middlegeneration, ype*b transfersaccordinglyrefer to alreadyreceived inheritance.Note that
Figure1 has to takefive generationsntoaccount, romthegrandparentsothe
grandchildrenf
ego (P2),belongingo the middle
generation.Thenotations or indirect eciprocitiesollowdirectly,using chronologicalorder:*aa concernsupwardandbackward-lookingeciprocities such as a
retrospective iew of the "demonstrationffect");*hh,*ggand .bb concerndownward ndbackward-lookingeciprocities"retrospective"equestsarea formof bb);aa* concernsupward ndforward-lookingeciprocitiessuchas the Cox and Starkdemonstrationffect); hh*, gg* and bb* representdownwardand forward-lookingndirect reciprocities(such as dynasticaltruism).
Intheempiricalwork,we focusmainlyon backward-lookingeciprocitiesfor inter-vivos ransfers.aa, *hh, gg) andonthe demonstrationffect(aa*);otherforward-lookingdownward)ndirectreciprocitiesappearmorediffi-cult to test.
III. Cross-section Data (France 1992) and Initial Results
Weuse two Frenchhouseholdsurveys:havingbeen conducted n the same
year,withmanyquestionsn commonabout ntergenerationalransfers,heyprovide complementary nformation. The INSEE representative urvey
8Alsoexplainedby limitedrationalityrimperfectnformation,uch"retrospective"equestsbecomesociallyoptimalunder teady-state rowth; ee BevanandStiglitz(1979).This is whytheyareoftencalled"goldenrule"bequests.
? The editors of the Scandinavian Journal of Economics 2001.
7/28/2019 Family Transfers Involving Three Generations
http://slidepdf.com/reader/full/family-transfers-involving-three-generations 13/30
426 L.ArrondelndA. Masson
allows us to assessthe empiricalrelevanceof the KCPGdecomposition fthe life cycle,by lookingattherelative requency f three-adult enerations
families (Table 1). The CNAVsample,which covers only such families,offersresultscomparableo INSEE data on thatsub-population;moreover,simplestatisticsreveal the heterogeneity f downward-typesh, g ands-financialransfers etween hethreegenerationsTable2).
The INSEE 1992 RepresentativeData Set
TheINSEEwealthsurvey,basedon a sampleof around10,000 households,
is representativef the entireFrenchpopulation. t containsdetailed nfor-mationon eachhousehold'sdemographicndsocioeconomiccharacteristicsand sourcesof income(aswell as indications f thepasteconomic rajectoryof eachspouse).It alsoprovidesdetaileddataon the ownership nd amountof real and financial assets. Finally,a specific part of the questionnaireconcernsdynastic andrecall) nformation:arents'andchildren's haracter-istics (age, education,occupationalgroup), intergenerationalransfersre-ceivedandbequeathedhelpingout,gifts or inheritance) t anydate before
the interview.The survey has two main drawbacks or our purpose,however.First,householdswere interviewed lmostexclusivelyaboutfinancialandparent-to-childtransfers:hereis little information n timetransfers, s well as onreversetransfers, romchild to parent.Second,we havedirectinformationon income for the respondentonly-the potentialmiddlegeneration-notforherparents ndchildren.9
TheDiversity of Family (Generational)DemographicCompositions
Howmanyindividualsof matureage (say,50 yearsold) belongto a three-adultgenerations amily, .e., haveliving parentsand adultchildren?Stated
differently,how many P individualscomprisethe middlegenerationof a
(K)CPG family branch,with or withouta grandchild?10 s it is rather
complex, his latterdemographicompositions rarelyconsidered y econo-mists. Cox and Stark(1998), for instance,who were amongthe first toconsider hree
overlapping enerationswithinthe
family,use a
simplerKPG
representationwith only two adult generations.INSEE data can tell us
9Moreover,nlythe current evelsof householdncome andwealthareavailable, ot thoseatthe timeof pasttransfers.l?Data imitationprecludesotherchoices for the generation f reference uchas individualswithlivingparents ndgrandparents.
() The editors of the Scandinavian Journal of Economics 2001.
7/28/2019 Family Transfers Involving Three Generations
http://slidepdf.com/reader/full/family-transfers-involving-three-generations 14/30
Family transfers nvolvingthreegenerations 427
whether, from a purely demographic point of view, this "new" (K)CPGfamily composition is really dominant.11
Let us consider an individual in period P of her life cycle. The probabilitythat she has both a parent alive and a grown-up child increases regularlybefore 50 years old and decreases after 55, with a peak in the 50-55 span ofaround 48 percent. Thus, in any cohort of mature age (P), a majority ofindividuals do not belong to a (K)CPG triad. In particular,the rate is only44.2 percent in the 49-53 age bracket (the range selected in the CNAV
sample, presentedbelow).12What are the relative frequencies of family compositions for an individual
in the 49-53 age interval, referredto as ego? We omit the rare cases wherethis individual still has a grandparent alive, or where an intermediate
generation is missing (grandchild with no surviving child). As indicated in
Table 1, a family branch to which ego (position P) belongs may then have
any of the following eight compositions: (1) KCPG: four generations; (2)CPG: as 1, but ego has no grandchildren n the branchconsidered; (3) KCP:as 1, with ego an orphan;(4) CP: as 2, with ego an orphan; (5) KPG: as 2,but ego's child is still a "pre-adult";(6) KP: ego an orphan,with a pre-adultchild; (7) PG: ego has no descendants but one living parent at least; (8) P:
ego has neither living ascendants nor descendants.13Table 1 shows the results: 35 percent (24.0 + 11.0) of the individuals in
the 49-53-year-old cohort have lost both parents,31.8 percent (20.8 + 11.0)have no adult child, 10.1 percent (7.4 + 2.7) have no children at all. As
indicated, 44.2 percent belong to at least one (K)CPG family branch(KCPGor CPG). This three-adult generations composition (albeit a minority) is
"There s one difficulty,however: hree-adultenerationsamilies,referred o as triads,may
not have a (K)CPGcomposition. n the INSEEsurvey,9,507 householdswereinterviewed,correspondingo 16,580individualswho were the head of the householdor his/herspouse.Among these 16,580 individuals,15.0 percent (2,645 individuals)make up the middlegenerationn such triads: heyhaveat least one parent till alive andone "adult"child,are
independent r more than 22 years old. A proportionof these individuals,however,willalreadybe retired, elonging o a familycompositionwithtworetired enerations.Fortunatelyenough,evenif retirement eginsafter60 yearsold (still the legalretirement ge in France),the compositionwithtwo retiredgenerations epresentsess than a quaterof the triads:77.5percentof the individualswhocomprise hemiddlegenerationn triadsare less than60 yearsold.
12Three-adultenerationsfamilies with a middle generation n the 49-53 age bracketrepresentess than 28 percentof total (K)CPG families,when a 46-56 age span alreadycontains wo-thirds r so of these families.Thus, he narrow gebracket doptedn theCNAVsurvey,chosen(by familysociologists) n order o limit cohorteffects, leads in return o asamplewhich s far essrepresentativef theentire K)CPGpopulation.13Egomaybelongto several amily"branches"f different omposition, wing o thespacingof herchildren,withsome of age C, othersof age K. If ego is not anorphan,hepossibilitiesare KPGand(K)CPG;if ego is an orphan, hey are KP and(K)CP,where the parenthesisaroundK mean hatanegowithanadultchildmayormaynot havepre-adult randchildren.
? The editors of the Scandinavian Journal of Economics 2001.
7/28/2019 Family Transfers Involving Three Generations
http://slidepdf.com/reader/full/family-transfers-involving-three-generations 15/30
428 L. Arrondeland A. Masson
Table 1. Relative frequencies (%) of family generational compositions forindividuals of age P between 49 and 53
Composition % %
Onethree-adultenerations ranch:KCPG atleastone) 18.5CPG 25.7Total(K)CPG 44.2
Parent(s), o adultchild:KPG no KCPG) 13.4PG 7.4
Total(K)PG 20.8No parents, dultchild:
KCP(atleastone) 10.8CP(noJ) 13.1Total(K)CP 24.0
No parents, o adultchild:KP(atleastone) 8.3P (only) 2.7Total(K)P 11.0
Totalcohortaged49-53 100.0 100.0
KPGand(K)CPG (adultandpre-adult hildren) 20.8Total KPG 34.2
KPand(K)CP(adultandpre-adulthildren) 12.0TotalKP 20.3
Source: NSEE"Actifsinanciers992" urvey.
clearly the mostfrequent in the 49-53 age bracket; moreover, 18.5 percent
live in four overlapping generations families. By comparison, only 13.4percent of the cohort belong to a KPG family branch,while having no adult
child; however, 34.2 percent to be members of a KPG branch(a majority of
them also having an adult child)-a significant number for the test of the
pure demonstration effect describedby Cox and Stark(1996, 1998).In any case, these figures contradict the hypothesis of a representative
family for the analysis of intergenerationaltransfers. Moreover, the multi-
plicity of generational compositions constitutes another factor of inequality
in family support, besides the level of income or wealth, the geographicaldistance between family members, the strength of intra-generationallinks,or the heterogeneity of parents' preferences. Some individuals are thus
"lucky" to be early orphans, thereby not having to spend considerable time
and money, energy and stress, in order to support their old and dependentparents. This multi-dimensional heterogeneity should be taken into accountwhen designing new family policies or evaluating the consequences of a
withdrawalof the welfare state.
? The editors of the Scandinavian Journal of Economics 2001.
7/28/2019 Family Transfers Involving Three Generations
http://slidepdf.com/reader/full/family-transfers-involving-three-generations 16/30
Familytransfers nvolvingthreegenerations 429
The CNAV 1992 Survey on Three Adult Generation Families
The CNAV hreegeneration survey concerns, specifically, French families of
three-adultoverlapping generations, with a middle generation in the 49-53
age bracket (husband or wife). Three similar interviews were conducted perfamily, one for the middle generation, one for its parents,and one for a child
who lives in an independenthousehold or is more than 22 years old (the C
age). A total of 1,958 questionnaireswere gatheredfor the middle generation,but reliable data are available for only 1,217 couples in the old and middle
generations,andcomplete information(from the three different sources), was
only obtained for a thousand families-or rather amily branches.14
Each questionnaire contains detailed information on the household'sdemographic and socioeconomic characteristics, income, asset holdings,intergenerationaltransfers, etc. The questions were often adapted from theINSEE questionnaire performed the same year. The other advantage of the
CNAV survey is the large scope of intergenerational transfers covered:
downward and upward, financial and time transfers or services, visits and
contacts between parents and children, but also between grandparents,and
grandchildren,and detailed health status of each family member.15
Moreover, the CNAV survey is quite rich regarding dynastic data: inter-viewing the parents (G) of the middle generation provides information ontheir own parents(already deceased); likewise, interviewing the children (C)of the middle generation provides informationon theirown young children(ifthey exist). In short,the CNAV survey contains direct or indirect informationon the five generations representedin Figure 1-from the grandparents o the
grandchildrenof interviewed ego-as well as on all types of past andpresentinter-vivos andpost-mortemtransfers shown in the figure.16
14Thiscomplexselectionprocessraisesthe questionof the representativenessf the relevantthree-adultenerationsub-population. comparisonwith INSEEdatashowsthatthe CNAVsampleperformswell in thisrespect,althoughheamounts f assetsforownersareoften owerthan in the INSEE (typical wealth) survey.Moreover, he size of financialtransfers ssomewhat nderestimatedthis s onereasonwhywe consideronlyprobitanalysis).15The hree ndependent uestionnaireser familygive informationn incomeandwealth oreach adultgeneration, nd allow for cross-examinationf transfers ivenby one generationand receivedby another.Themost significantimetransfers, uchas supportof old parents,coverthe lastyears.A specificcomponent f timetransferss co-residence: 22-year-old rmore "adultchild"maybe still at home;on the otherhand,old parentsmaylive with theirchildrenof the middlegeneration.Thislatter ypeof co-residence oncerns16percentof theold parents n the CNAVsurvey,but,owingto a numerousprogeny,only 4 percentof theirchildrenof the middlegeneration.Moreover,or half of the children 2 percent),especiallyamong armers, o-residence s life-long(parents ndchildrenhavealways tayed ogether).16Typeb transfers,hatis expectedbequestor inheritance, reproxiedby middle-generationwealth.The mainshortcoming f the CNAVsurvey,however,s thatthe sampleconsistsonlyof three-adult enerationsamilies,making t difficult o perform ests based on the demo-graphicdiversity f families egowith orwithout hildren orinstance).
? The editors of the Scandinavian Journal of Economics 2001.
7/28/2019 Family Transfers Involving Three Generations
http://slidepdf.com/reader/full/family-transfers-involving-three-generations 17/30
430 L.Arrondel ndA. Masson
Diffusion of Financial DownwardTransfersbetween ThreeGenerations
To give some idea of the heterogeneityof financialdownward ransfers(typesh, g ands) betweenthe threeadultgenerations,we mayrefer to theitems distinguishedby the CNAVsurvey(the INSEE surveyhas similar
information)n the form of significant wealth)gifts on the one hand,andfinancialhelp on the other,divided nto fourcategories: 1) housing, hat is
paymentof rent or providinga separatehome free; (2) money, ncludingregularpaymentsor help for an important urchase; 3) loan, meaningamoney oan;(4) cosignature,or a mortgageandotherhelpforrepayment f
a mortgage.Thisalso revealswhethergrandparentsive as often,and in thesameway,asparentso theyounggeneration.
Recall that gifts, especially those received late in the life cycle, are
presumably artof the wealth transmission rocess;moreover,as we shallsee below,"housing"and"money"aremoreoften educationransfers, nd"loan"and"cosignature"reformsof financial ssistance.
Table2 simply presentssome descriptivestatisticsconcerning he fre-
quencyand compositionof the threetypes h, g and s of transfers,with
"loan" and "cosignature" roupedtogether.17 arent-to-childransfers-
Table2. Comparison f the natureof downwardinancialtransfers
Frequencyf different ransfersmong hebeneficiaries%)
Frequencymong Inter-thebeneficiary vivos Money Loanorgeneration%) gifts Housingapaymentscosignatureb
Middlegenerationo child 47.7 11.9 16.6 79.7 21.2P=t C
Oldgenerationo grandchild21.9 19.3 3.4 78.3 10.0G=* C
Oldgenerationo child 45.4 66.8 3.7 25.0 38.3
Source:CNAVurvey1992).Notes:aPaymentf rentor freehome
bFor omepurchasermortgage.
17Figure 1, drawn n a purely intergenerationalerspecive,does not representunions or
siblings.To makethe calculations eportedon Table2, however,we had to makespecificconventions: s the numberof individuals elated o ego of the middlegenerations far too
large,onlya limitedset of relevant amilybrancheswithinkinshipmustbe selected.In Table
2, the family unit taken into consideration"only" includesego and her spouse (middlegeneration),heir adultchild interviewedn the CNAVsample,as well as theirownparents(possiblyfour of them);moreover,requency ates arealwaysmeasuredn reference o the
beneficiary;ee Arrondel ndMasson 1999b).
( The editors of the Scandinavian Journal of Economics 2001.
7/28/2019 Family Transfers Involving Three Generations
http://slidepdf.com/reader/full/family-transfers-involving-three-generations 18/30
Family ransfersnvolvinghree enerations 431
whether received by the middle or the young generation-are twice as
frequent s grandparent-to-grandchildransfersgenerationkipping).More-
over,as expected,transfersa fromthe old to the middlegenerationmoreoftenconsistof inter-vivos ifts,whereas ransfersh fromthemiddleto theyounggeneration elativelymore often take the formof housingservices.Thecompositionof transfers ivento theyounggenerationby theirparentsandby theirgrandparentsre not thatdifferent,althoughparentsseem tospecializein housingservices-education transfers-whereasgrandparentsgiverelativelymoregifts.
IV. Probit Analysis of Transfers
We begin by presenting,based on INSEEdata,some preliminaryests ofthe typology introduced or parent-to-childinancialinter-vivos ransfers:education,financial assistance and wealth gifts (Tables3 and 4). Next,upward(financialor time) transfers, rom the middle to the old genera-tion, are studiedin more detail on the basis of CNAVdata,in line withthe Cox and Stark "demonstration ffect" (Table 5). We then consider
retrospectiveeffects for downwardfinancial transfers,again based onINSEE data (Tables3 and 4), while also referring o other Frenchdatasets and studies.
Parent-to-childFinancial Transfers:HeterogeneousMotives
The distinctionmadebetween education ransfers investmentn a child'shumancapital), inancialassistance o children proxyformissingcreditor
insurance), ndwealthgiftswhichareanticipated equests,mpliesthateachcategoryof inter-vivostransfershas specific determinants, ses specificformsof transmission, nd occurs more likely at a given stage of the lifecycle. Representative f the entire Frenchpopulation, he INSEE surveyallows for a loose testof thisternarydivision.
Table 3 focuses on educationtransfersand financialassistance. Thesample s restricted o non-retired arents,presumablyf age P, withat leastone grown-up above18) independenthild.It concernsonlyfinancialhelp,
excluding nter-vivos ifts, so thatthe thirdmotive-transmissionof wealthoranticipation f inheritance-canbe safelydisregarded.Inorder o disentangle ducationransfersromfinancialassistance,Table
3 compares ndependent robit estimatesfor the same four categoriesoffinancialhelp given by parents o adultchildrenas in the CNAVsurvey:"housing", "money", "loan" and "cosignature".The first explanatoryvariables,"transfers eceived",will be commentedon later.The relativedifferencesn theeffects of the othervariables re notonlystriking,butalso
C The editors of the Scandinavian Journal of Economics 2001.
7/28/2019 Family Transfers Involving Three Generations
http://slidepdf.com/reader/full/family-transfers-involving-three-generations 19/30
n Table3. Financial help given by (non-retired) parents to adult child (probit)DL0
Zs
Intercept
F Parents' characteristics
? Ownparent(s) alive
. Transfersreceived> Same form of transferreceivedb Other financial help, Gift received from own parents
Inheritancereceived from own parents? Otherinheritance or gift
Age (lOE-1)Age2 (O1E-2)WidowIncome (log.)Wealth 10OE-7)Wealth2(lOE-14)Social statusFarmer
Self-employed (small production unit)
Self-employed (big production unit)ProfessionalExecutiveMiddle managementWhite-collar worker
Qualifiedblue-collar workerOtherblue-collar worker(ref.)
Housinga Moneya
parameter t-value Parameter t-value
-2.632 -1.980 -3.770 -3.221
-0.172 -1.663 -0.104 -1.097
0.258 2.538 0.266 2.9540.016 0.201 0.080 1.0750.197 2.373 0.201 2.588
0.179 2.321 0.166 2.3580.328 2.703 0.022 0.184
0.043 1.044 0.017 0.5150.000 -0.693 0.000 -0.197
-0.197 -1.391 0.027 0.2220.085 1.575 0.261 4.6180.600 2.649 0.001 0.004
-0.162 -2.557 -0.028 -0.612
-0.141 -0.674 -0.103 -0.567-0.019 -0.093 -0.251 -1.417
0.231 0.695 0.160 0.514-0.297 -1.000 0.222 0.8180.002 0.008 -0.062 -0.310
-0.040 -0.188 0.211 1.1330.132 0.626 -0.160 -0.860
-0.111 -0.537 -0.065 -0.3700.000 0.000
7/28/2019 Family Transfers Involving Three Generations
http://slidepdf.com/reader/full/family-transfers-involving-three-generations 20/30
DiplomaNone(ref.) 0.000 0.000Primarychool 0.203 1.612 -0.068 -0.569Secondarychool 0.337 2.738 0.148 1.268Baccalaureate 0.324 1.971 0.151 0.962Graduate ndpostgraduatetudies 0.581 3.415 0.077 0.472Number f children -0.074 -1.981 -0.090 -2.642Number of adult children 0.116 2.498 0.063 1.481
Adultchildren'scharacteristics(mean)Girl(in %) -0.040 -0.434 -0.095 -1.133Age -0.029 -2.347 -0.016 -1.495Number fchildren -0.061 -1.024 -0.041 -0.877Diploma (in %)None(ref.) 0.000 0.000Primaryschool or secondary school 0.039 0.230 0.065 0.440Baccalaureate 0.268 1.450 0.225 1.367Graduate and postgraduate studies 0.457 2.517 0.297 1.831
Marital status (in %)r Single (ref.) 0.000 0.000CD Married -0.203 -1.891 -0.111 -1.148
Widow 0.862 1.281 0.548 0.9050 Divorced 0.004 0.010 0.427 1.460S Social status (in %)> No activity ref.) 0.000 0.000? Farmer 0.024 0.083 -0.123 -0.465i Self-employed 0.005 0.024 0.103 0.485? Executive -0.269 -1.545 -0.167 -1.021
White-collarworker -0.236 -2.075 -0.161 -1.531
|Blue-collarworker -0.165 -1.053 -0.139 -0.980
2 (40 df) 285.5 226.61 Numberof parents elping 420 574
? Numberof observations 1,813 1,813
a Source: INSEE "Actifs financiers1992"survey.Notes: aHousing:paymentof rent or providinga separatehome free; Money: regularpaymentor helpcosignature or a mortgageor otherhelp for repaymentof a mortgate.
7/28/2019 Family Transfers Involving Three Generations
http://slidepdf.com/reader/full/family-transfers-involving-three-generations 21/30
7/28/2019 Family Transfers Involving Three Generations
http://slidepdf.com/reader/full/family-transfers-involving-three-generations 22/30
Family transfers nvolvingthreegenerations 435
Table 4. (continued)
Probabilityf giving Probabilityf giving
parameter t-value Parameter t-value
Executive 0.311 1.546 0.241 1.184Middlemanagement 0.059 0.313 0.016 0.084White-collarworker -0.084 -0.547 -0.108 -0.580Qualified lue-collarworker 0.245 1.398 0.254 1.439Otherblue-collarworkerref.) 0.000 0.000DiplomaNone(ref.) 0.000 0.000
Primarychool 0.122 0.853 0.120 0.825Secondarychool -0.253 -1.851 -0.316 -2.257Baccalaureate 0.058 0.326 0.036 0.197Graduatendpostgraduatetudies -0.085 -0.449 -0.128 -0.670Number f children -0.149 -2.616Number fadultchildren 0.073 1.202 -0.009 -0.363
Adult children'scharacteristics mean)Girl (in %) -0.009 -0.105 0.013 0.143
Age 0.005 0.655 0.005 0.688Number f children 0.013 0.547 0.020 0.840Diploma in %)None(ref.) 0.000 0.000Primarychoolorsecondarychool 0.081 0.580 0.099 0.695Baccalaureate -0.028 -0.160 -0.021 -0.118Graduatendpostgraduatetudies 0.072 0.450 0.014 0.086Marital status (in %)Single(ref.)Married 0.038 0.346 0.045 0.409Widow -0.186 -0.426 -0.045 -0.101Divorced -0.095 -0.431 -0.034 -0.151Social status (in %)No activity ref.) 0.000 0.000Farmer 0.529 2.791 0.469 2.453Self-employed 0.513 2.875 0.509 2.828Executive 0.170 0.950 0.193 1.067White-collarworker 0.212 1.609 0.235 1.761Blue-collarworker 0.255 1.550 0.324 1.954
X2 (39df) 490.2 544.7Number f parents iving 403 403Numberof observations 3,668 3,668
Source: NSEE"Actifs inanciers 992"urvey.
Note:aThewealth ariables thesumofactualwealthplustheupdatedmountfthegifts.
component of wealth transmission. The first regression includes the level ofwealth of parents before the gift and their total number of children; thesecond regression introduces a dummy for taxable wealth, which in France
depends on the amount of wealth, marital status, and the total number ofchildren (it is an inheritancetax with an exemption threshold); see Arrondeland Laferrere(2001).
? The editors of the Scandinavian Journal of Economics 2001.
7/28/2019 Family Transfers Involving Three Generations
http://slidepdf.com/reader/full/family-transfers-involving-three-generations 23/30
436 L.Arrondeland A. Masson
The first regression reveals importantdifferences in the various forms offinancial help: education variables and parentalincome have no effect at all,
whereas wealth has a significant positive influence. The presence of grand-parent(s) now has a significantnegative impact, which may reflect their needfor costly support.Moreover,gifts are more often made by "self-employed"parentsto "self-employed" children("self-employed" including farmers and
professionals), in which case they constitute a privileged means of ensuringan efficient transmission of professional assets. Finally, the highly positiveeffect of being a widow indicates that considerable gifts are made as
anticipatedinheritance,after the death of the firstspouse.
In the second regression, the dummy for taxable wealth has a very highand significant effect, whereas the amount of wealth remains significant onlyat a 15 percent level. The probability of making a gift therefore appears
strongly influencedby tax considerations.19'20The previous set of regressions raises two main problems, however. The
first concerns the use of independent probit models. Arrondel and Wolff
(1998) show that multivariateprobit models for the four forms of financial
help, or for help and gifts, lead to comparablequalitativeresults. Moreover,the different forms of financial help appear significantly complementary;onthe other hand, gifts are rather substitutes for "loan". The second problemconcerns the potential omitted variable bias due to the absence of a directmeasure of the adult child's income (although proxied by education and
occupational group). CNAV data allow us to control for current income of
both the giver and the receiver. Since the same econometric specifications as
in Tables 3 and 4 lead to very close results, it is possible to estimate the bias:
introducingthe child's income in the regression does not qualitativelychangethe other coefficients. Moreover, transfersappearratheranti-compensatory:
for given parents'resources (income andwealth), the probabilityof a transferincreases (often) significantly with the current income of the interviewed
child.21
19On he otherhand,a dummyfor taxablewealthhas no effect on the different orms offinancialhelp (regressionsnot shown).Arrondeland Laferrere2001) have also analyzedtransfermotives declaredby givers:the most frequentare tax incentives and anticipateddivisionof anestate,alsoleading othehighestamounts iven.20naddition, haringpracticesvary considerablyor the different ormsof financialhelpand
gifts.Whengiftsareregisteredndmentioned n theestate,estatedateshow hat heyaresharedequally nmore han90 percentof thecases;unequal statesharing, ccurringmainly hroughgifts,is highlyconcentratedmong hetopwealth-holders. or inancial elp,the resultsare,onthecontrary, uite nagreement ithHochguertelndOhlsson's1999)conclusions n Americandata health ndRetirementurvey): qual harings theexception althoughhere ouldbesome
compensationver he ifecycle)andconcentrated, oreover,mongwealthy arents.21Replicatinghe Altonjiet al. (1997) study,Wolff(1998)also findsthatshiftingone unit incurrent ncomefrom heparents of the middlegeneration)o theirchildleadsto no reduction(indeed o a slight ncrease)n the transfer.
c The editors of the Scandinavian Journal of Economics 2001.
7/28/2019 Family Transfers Involving Three Generations
http://slidepdf.com/reader/full/family-transfers-involving-three-generations 24/30
Family ransfersnvolvinghree enerations 437
IndirectReciprocities: Child-to-parent Timeor Financial) Transfers
Table5, derived from the CNAVsurvey,concernsupward financialandtime) transfersgiven by ego's household(middlegeneration,49-53 yearsold) to ego's parents(type a transfers).Financial transferamounts arereported or the last five years.Time transfers epresent upportcurrentlygiven to parentspresumablyn need,owingto healthproblemsor old age.Healthstatus s divided nto threecategories: o problem, omeproblemandbad health.In addition,we knowif old parents-generation1-themselveshavehelpedtheir(now deceased)own parents-generation0-in the past,so we have information bout heexistenceof *aa ndirect eciprocity.
Table 5 refers to two bivariateprobit models for financial and timetransfers.Among the 1,216 households of the middle generationwithreliable nformation n the old generation, 92 gavetimetransfers, utonly88 madea financial ransfer.The resultsof the firstmodelarereproducedncolumns1 and 2. Thesecondmodel adds crossedvariables orchildrenandeducation,andadoptsa more restrictivedefinitionof financial ransfersseebelow);forsimplicity theresultsdo notchangefor timetransfers), nlytheresultsobtained orthese financialransfers rereproducedincolumn3).
Theprobability f timetransferscolumn1) is explainedby manycharac-teristicsof the old generation. t increaseswith parents'age (andincome),and is higherforwomen who live alone. As expected,bad health ncreasesthe probability f transfer; n the otherhand,distancefrom childrenhas astrongnegativeeffect,as well as thenumber f children.Ego is less likelytohelpherparentsf she hasmanysiblings.Theexistenceof professionalhelphas a positive effect, as if professionalhelp and family help were "com-plementary".22Moreover, or given characteristics f ego, the significant
negativeeffect of parents'bequeathablewealth (also found on a samplerestricted o two childrenormore) s evidenceagainst he strategicbequestsin Bernheimet al. (1985); see Perozek(1998) for similarfindings.Theretrospective ffect *aa is almostsignificantat a 10 percent evel: parentswho helpedtheirownparents end to receivemoresupportduring heir oldage.
Fewerchild'scharacteristicsavesignificant ffects on the probability ftimetransfers.nparticular,henumber f childrenathomehas no influencewhatsoever, espitethe cost involvedforparents-perhapsowingto a weakdemonstrationffect;see Coxand Stark 1996).More mportantly,owever,the probabilityof transfers higherfor daughters, ingle children,and forchildrenwho have some healthproblems.These resultsmaybe interpretedas indirect vidence n favorof the Cox andStarkdemonstrationffect:these
22Thisconclusion needs much furtherqualification, owever,since the specificconditionsrequiredo obtainprofessional elpare ikelyto createa selectivitybias.
C The editors of the Scandinavian Journal of Economics 2001.
7/28/2019 Family Transfers Involving Three Generations
http://slidepdf.com/reader/full/family-transfers-involving-three-generations 25/30
438 L. ArrondelandA. Masson
Table 5. Transfersgiven to elderlyparents (bivariateprobit)
(1) (2) (3)Time transfer Financial transfera Financial transferb
Asympt. Asympt. Asympt.Parametert-value Parametert-value Parametert-value
Intercept
Old generation'scharacteristics
Support f ownparents n the
pastAgeLiving as coupleWomanloneFreehomeprovidedbythe
familyProfessionalhelpSomehealthproblemsBad healthIncome log)Wealth1OE-6)
Number f children liveGift ochildrenDistanceromchildren
Middle generation'scharacteristics
Two arentsaliveParent(s)-in-lawliveAgeSex(female:1)Livingas couple
EducationNumber f childrenChildrenX education
-4.636 -2.312 -1.876 -0.525 -2.968 -0.792
0.144
0.0540.0100.3360.173
0.3090.1340.3110.179
-0.040
-0.080-0.014-0.191
1.577
6.257 -0.16 -1.331 -0.012 -1.2810.055 -0.408 -1.333 -0.426 0.1962.038 0.081 0.319 0.051 -0.9861.358
3.0531.261 -0.009 -0.061 0.012 0.0822.592 0.142 0.877 -0.073 -0.3821.759 -0.029 -0.183 -0.064 -0.386
-2.968 -0.033 -2.012 -0.029 -1.685
-3.841 -0.018 -0.679 -0.035 -1.049-0.140-6.934
0.578 2.393-0.101 -0.769
-0.050 -1.607 -0.029 -0.5760.445 4.874 -0.172 -1.316
-0.356 -2.376 -0.260 -1.184
-0.003 -0.070 0.016 0.8320.039 0.976
0.579 1.953-0.019 -0.130-0.022 -0.424-0.163 -1.146-0.350 -1.515
0.084 2.2690.367 2.495
-0.020 -2.156
Number f children thome -0.005 -0.086Number f childrennotat 0.007 0.470
homeSomehealthproblems 0.236 2.146Bad health -0.044 -0.278Income log) 0.068 0.679 0.342 2.189Wealth 10E-6) 0.002 0.389 0.014 1.648
Log-likelihoodDisturbance correlationNumber of children helpingNumber of observations
-851.20.069(0.771)
292 881,216
0.309 1.7630.013 1.367
-823.90.103 (1.090)
78
1,216
Source: CNAVsurvey1992.Notes: Time transfer:help givento parentsowingto healthproblemorold age.aFinancial transfersinclude inter-vivos gift, alimony, regular payments, money help, loan and
housinghelp.bAlimony s not include in financial ransfers.
c Theeditorsof the Scandinavianournalof Economics 001.
7/28/2019 Family Transfers Involving Three Generations
http://slidepdf.com/reader/full/family-transfers-involving-three-generations 26/30
Familytransfers nvolvingthreegenerations 439
helpers are themselves more likely to need supportin their old age, because
they have a longer life expectancy, are alone, or have already had some
health problems. But other explanations are possible: social norms mayaccount for the fact that daughtersdevote more of their time to parents (theresult holds even for working daughters); single persons may have more free
time and are more sensitive to parentalpressure; and helpers, who experi-ence painful physical and psychological stress are more likely to have some
health problems (reversecausation).23On the other hand, Table 5 shows that financial transfers(column 2) are
poorly explained by the usual variables;parents' wealth again has a negative
effect andchild's income a
positiveeffect
(a signof
upward altruism?).Note
that there is no significant effect of the total number of children (either
independent or at home). Moreover, the probability of financial transfer
increases when both parents are alive. There is no sign of substitution
between financial and time transfers made by ego's household to her parents;the correlation is weak and insignificant.24
In the second model, a more restricted definition of financial transfers is
introduced(column 3). In 10 cases, only alimony has been given, which maybenefit ego's parents-in-law as well as ego's parents. Dropping these 10
ambiguous cases leaves us with 78 helpers. The same specification leads to
close results: in particular (regression not shown), the coefficient for the
number of ego's children is positive, though small and imprecisely estimated
(0.061, t-statistics of 1.4). Cox and Stark(1998), who obtain a similar result,
argue that it is already favorableto their demonstrationeffect, since "havingmore children represents increased competition for ego's household re-
sources".25In order to take into account child costs, these authors suggestthat better educated egos will indeed "commit more resources to their
23Theremaybe a problemwith thedependent ariable elative o time transfershatshouldbe
givenonly to parentspotentiallyn need:say,havingsome healthproblemsor badhealth,or
beingabove80 yearsold. Thenumber f observationss then reduced o 915, withonly250
helpers,but estimates emain eryclose:onlythe *aa etrospeciveffectlosesany significance.Moreover,we addtwoexplanatoryariables,hepresenceof parents-in-lawndtheexistenceof time transfers iven to themby ego's household: he lattervariable s highly significant(coefficientof 1.591, t-statistics bove8), indicating trongcomplementarityf time transferstoparents ndparents-in-lawlackof instruments oes notallowus to correct orendogeneity
bias).24It is impossible o test the *aaindirectreciprocityor upward inancial ransfers:oo few
parentsdeclarehavingmadefinancial ransferso theirownparents.25Unlesshaving more children is correlatedwith downwardaltruism,which, in turn, ispositivelycorrelatedwiththepropensityo give to parents.Moregenerally, otethatthe Coxand Starkdemonstrationffectis notentirelyconvincing, inceit does notprevent ree-ridingbehavior etweengenerations,eading o timeinconsistencies: achgenerationeeksto receive
old-agesupportwithout irsthelping ts ownparents.Transmission f preferences r inculca-tionof social norms s certainly artof the answer.
? The editors of the Scandinavian Journal of Economics 2001.
7/28/2019 Family Transfers Involving Three Generations
http://slidepdf.com/reader/full/family-transfers-involving-three-generations 27/30
440 L. Arrondeland A. Masson
children'shuman capital, leaving less to give to parents". Hence the predic-tion that, holding income constant, ego's education "will interactnegatively
with children, reflecting budget constraints",while the coefficient of child-ren will be positive, reflecting the demonstration effect. This is preciselywhat the results of the last columns of Table 5 show, again in close
agreementwith those of Cox and Stark,and "in favor" of the demonstration
effect.26
Indirect Reciprocities: Parent-to-child Financial Transfers
French data systematically reveal the importance of retrospective effects:*hh, *gg and *bb. The pervasiveness of qualitative and quantitative retro-
spective effects for bequests (-bb) is documented in Arrondel et al. (1997):those who make wills are more likely to have received inheritancethrougha
will; there is a specific link between the amounts of wealth inherited and
bequeathed.
Retrospective effects for financial help are shown in Table 3 (top part)based on INSEE data and comparableconclusions are obtained with CNAV
data. Among the transfers received (mainly from own parents), the same
form of help always has a positive, significant and high coefficient, whereas
the effects of other transfers are less importantand not always significant.This is clear evidence of specific *hhindirect reciprocity or transmission of
behavior.
Retrospective effects for gifts are shown in Table 4 for INSEE data.
Among the transfers received (again, mainly from own parents), gifts are the
only variable with a strong and significant (positive) effect. This is in favor
of gg indirectreciprocity.27
Forward-lookingand downward indirect reciprocities, hh*, gg* or bb*,appear less obvious to test, since we lack specific predictions derived from
an explicit model. Unequal sharing gives some indication of such recipro-
cities, however. Estate data show that in the rare cases of unequal sharingof
French estates (8 percent), which occur mainly in rich or self-employed
26Weadd to this last specificationanotherexplanatory ariable, he existenceof financialtransfersgiven to ego's parents-in-law:he positivecoefficient(1.064, t-statisticsabove4)suggests,onceagain, complementarityetween-this time-financial transferso parents ndparents-in-law. ut the result suffersfrom an endogeneitybias (unfortunately, e cannotinstrumentheexistenceof transfersoparents-in-law).27VeryimilarresultsareotainedusingCNAVdata.Moreover,Wolff(1998)has consideredotherretrospectiveffects.He findsevidenceof .gs effects: he existenceof giftsorinheritancereceivedfrom own parents ncreases he probabilityhat the old (age G) generationmakestransferso theirgrandchildren.gs effectsare,however, maller han*ggeffects).On the other
hand, sh effectsarerejected: eceivinggiftsor inheritanceromgrandparentsas no influenceontheprobabilityor the middlegenerationo make ransferso theirchildren.
? The editors of the Scandinavian Journal of Economics 2001.
7/28/2019 Family Transfers Involving Three Generations
http://slidepdf.com/reader/full/family-transfers-involving-three-generations 28/30
Family transfers nvolvingthreegenerations 441
families, the only registeredvariablethat has explanatorypower for the
discrepancybetweenthe amountsreceivedby siblingsis demographic:he
advantaged hildrenare thosewho themselveshavechildren; ee Arrondelet al. (1997). These heirs will be able to perpetuate amily lineage orenterpriseand to replicate he same transmission rocessin favor of theirown offspring.As unequalsharingoccursmainlythroughgifts rather haninheritancein more than 80 percentof the cases),this result s evidenceofgg* thanbb*.
V. Conclusions
This paperaddresses ntergenerational, ainly inter-vivos ransferswithinFrench amilies.Itemphasizeshegreatheterogeneity f downwardinancialtransfers hat are made at distantdates over a lifetime:a four-periodifecycle, frequentlywith threeoverlapping dultgenerations, ppears o be amore realistic as well as a moreappropriateepresentation.ndeed, n linewith some otherrecentstudies,ourfindingsadvocateexpandinghedomainof analysisof the familialbehavior rom wo to three adult)generations.Wedid this by studyingindirectreciprocitiesbetweengenerations,where thebeneficiary f a transferdoes not give back to the initialgiverbutreplicatesthe samekind of transfer o a thirdpersonof anothergeneration.Borrowedfromanthropologists,his conceptmaybe linkedto the recentburgeoningliterature n endogenousastesandpreferenceormation;ee Becker 1996).It also blursthe somewhatunsatisfactory istinctionbetween altruisticandexchangemotivesfortransfers.
Frenchdatarepeatedly how the importanceof retrospective ffects onvariousdownwardparent-to-child)ransfers: or eachtransfer,parentsare
stronglyinfluencedby the corresponding ehavior of their own parents.Moreover,heempiricalanalysisof upwardchild-to-parent)ransfers howssome (less robust)evidenceof forward-lookingndirectreciprocities, uchas theCox and Stark 1996, 1998)"demonstrationffect":parentshelptheirown (old)parentsbecausetheyexpectto receivecomparableupportn oldage from heirchildren.
Our empiricalconclusionshave some bearingon the effectiveness ofintergenerational olicy redistribution. ndirectreciprocitiesintroduce a
(partial) inertia in family transferbehavior,especially with respect toBarro-Beckercompensatoryltruism. n the Cox andStark ramework,orinstance, ransfers o old parentswill notbe sensitive o short-termhangesin (retirement r health)policies:householdsof the middlegenerationwillonly reactto changes n the long runwhichconcern hemselves.Likewise,transferso children, tronglynfluenced y familybackground, ill notfullyrespond o policychanges suchas a rise in socialsecurityor other ransfersto theelderly)andmayonlyadaptwithlong lagsto a newenvironment.
? The editors of the Scandinavian Journal of Economics 2001.
7/28/2019 Family Transfers Involving Three Generations
http://slidepdf.com/reader/full/family-transfers-involving-three-generations 29/30
442 L. Arrondel and A. Masson
References
Altonji, J. G., Hayashi,F. andKotlikoff,L. J. (1996), The Effects of IncomeandWealth
on Time and Money Transfersbetween Parents and Children,NBERWorkingPaperno. 5522.
Altonji, J. G., Hayashi,F. and Kotlikoff,L. J. (1997), ParentalAltruism and Inter VivosTransfers:TheoryandEvidence,Journalof PoliticalEconomy105, 1125-1166.
Arrondel,L. and Laferrere,A. (2001), Taxationand Wealth Transmission n France,Journalof PublicEconomics79, 3-33.
Arrondel,L. and Masson,A. (1999a), Les transfertsentregenerations: 'Etat, le marcheet les familles,Futuribles247, 5-40.
Arrondel,L. and Masson, A. (1999b), Family TransfersInvolving Three Generations,Delta Doc9916, Paris.
Arrondel,L. andWolff,F C. (1998), Les transfertsntervivosen France: nvestissementshumains,aides financiereset transmissiondu patrimoine,Economie et Prevision 135,1-27.
Arrondel, L., Masson, A. and Pestieau, P. (1997), Bequest and Inheritance:EmpiricalIssues and French-US Comparison,in G. Erreygersand T. Vandevelde(eds.), IsInheritanceLegitimate?,Springer-Verlag,Heidelberg.
Barro,R. J. (1974), Are GovernmentBonds Net Wealth?,Journalof Political Economy82, 1095-1117.
Becker, G. S. (1991), A Treatise on the Family, EnlargedEdition, HarvardUniversityPress,Cambridge,MA.
Becker, G. S. (1996), Accountingfor Taste, HarvardUniversity Press, Cambridge,MA.
Becker,G. S. and Murphy,K. M. (1988), The Familyandthe State,Journalof Law andEconomics31, 1-18.
Bemheim,B. D., Shleifer,A. andSummers,L. H. (1985), The StrategicBequestMotive,Journalof PoliticalEconomy93, 1045-1076.
Bevan,D. L. and Stiglitz, J. E. (1979), Intergenerational ransfersandInequality,GreekEconomic Journal1, 8-26.
Cox, D. (1987), Motives for PrivateIncomeTransfers,Journalof Political Economy95,508-546.
Cox, D. and RankM. R. (1992), Inter Vivos Transfersand Intergenerational xchange,Reviewof Economicsand Statistics74, 305-314.
Cox, D. and Stark,0. (1996), Intergenerational ransfersand the DemonstrationEffect,Mimeo,BostonCollege and HarvardUniversity.
Cox, D. and Stark,0. (1998), FinancialTransfers o the Elderlyand the "Demonstration
Effect", Mimeo,BostonCollege and HarvardUniversity.Dunn, T. A. (1997), The Distributionof Intergenerationalncome TransfersAcross and
WithinFamilies,Mimeo, SyracuseUniversity.Ehrlich, I. and Lui, F T. (1991), IntergenerationalTrade, Longevity and Economic
Growth,Journal
ofPolitical
Economy99, 1029-1059.
Gokhale,J.,Kotlikoff,L. J., Sefton,J. andWeale,M. (1999), Simulating he Transmissionof WealthInequalityvia Bequests,Mimeo,BostonUniversity,MA.
Hammond,P. (1975), Charity:Altruismor CooperativeEgoism?, in E. S. Phelps (ed.),Altruism,Moralityand EconomicTheory,Russell SageFoundation,New York.
Hochguertel, S. and Ohlsson, H. (1999), Inter Vivos Gifts: Compensatoryor EqualSharing?,Mimeo,UppsalaUniversity.
Kessler,D. andMasson,A. (1989), Bequestand WealthAccumulation:Are Some Piecesof the Puzzle Missing?,Journalof EconomicPerspectives3, 141-152.
? The editors of the Scandinavian Journal of Economics 2001.
7/28/2019 Family Transfers Involving Three Generations
http://slidepdf.com/reader/full/family-transfers-involving-three-generations 30/30