Does Population Aging Represent a Crisis for Rich Societies?

Transcription

1 First drft Does Popultion Aging Represent Crisis for Rich Societies? by Gry Burtless THE BROOKINGS INSTITUTION Jnury 2002 This pper ws prepred for session of the nnul meetings of the Americn Economic Assocition entitled Is There n Aging Crisis? Atlnt, Georgi, Jnury 6, The views expressed re solely those of the uthors nd should not be ttributed to the Brookings Institution.

2 Does Popultion Aging Represent Crisis for Rich Societies? A common view mong policymkers is tht popultion ging represents crisis for industrilized countries. Becuse pensions nd old ge helth insurnce re lrgely funded out of public budgets, increses in the percentge of the popultion pst retirement ge must eventully cuse steep increses in the frction of government budgets devoted to old ge consumption nd rise in the level of public spending. According to one view, either txes or government debt will be pushed to unsustinble levels unless public progrms re significntly reformed to curtil retirement benefits. This wy of viewing the problem of popultion ging is highly misleding becuse it trets ging within the nrrow perspective of government budgets. In the United Sttes, the perspective is often even more distorted, becuse mny nlysts focus solely on ging s impct on the federl budget. Yet popultion ging lso hs notble impcts on stte nd locl budgets nd n even bigger impct on privte household spending. Some of these impcts represent offsets to the effect of ging on federl government budgets. Popultion ging is the result of two developments, lower birth rtes nd longer life spns. Holding constnt the mortlity rte, lower fertility inevitbly rises the frction of the popultion pst ny given ge, including the retirement ge. However, it lso reduces the percentge of the popultion below given ge, such s ge 20. Both the old nd the young re dependent popultions tht derive most of their support from the current output of ctive workers. Rich societies finnce much of the consumption of the ged through public retirement progrms, while most (though not ll) of the consumption of the dependent young is finnced out of privte household budgets. Young lbor force entrnts lso require public nd privte investments to equip them to ern s much s verge members of the working-ge popultion. Lower fertility reduces spending requirements for the young, fct missed when observers focus nrrowly on the budgets of government old ge progrms. The pprent crisis connected with popultion ging is prtly n illusion stemming from nrrow focus on the spending needs of hndful of public progrms. If the full spn of privte s well s public burdens is tken into ccount, the increse in the dependency burden ppers much more modest

3 Longer verge life spns lso increse the percentge of the popultion pst given ge. If people vlue longer lives, this development unmbiguously mkes the popultion better off, regrdless of whether it increses the percentge of lifetime income tht must be devoted to consumption pst the ge of 65. If popultion ging represents genuine crisis for present or future genertions, it must be the cse tht those genertions will suffer loss in lifetime net income, possibly s result of excess net contributions to support younger or older genertions. The pper will ttempt to show the circumstnces under which this outcome could occur. The Burden of n Aging Popultion The popultion of the United Sttes, like tht of ll industrilized countries, is growing older. Between 1950 nd 2000 the percentge of the U.S. popultion pst ge 65 incresed by hlf, rising from 8.1 percent to 12.4 percent of the totl popultion (Tble 1). The Census Bureu predicts tht Americns who re t lest 65 yers old will constitute more thn one-fifth of the popultion by The incresing shre of the ged in the popultion will plce strins on the federl budget becuse the elderly finnce much of their consumption with trnsfers from the federl government. The cturies of the Socil Security nd Medicre progrms mke regulr forecsts of the long-term spending needs of the two progrms. Their mid-rnge forecsts published in 2001 suggest tht combined outlys on the Socil Security nd Medicre progrms will climb inexorbly over the next 50 yers, incresing from 6.4 percent of GDP in 2000 to 12.5 percent of GDP in Prt of the predicted increse in spending is ttributble to fctors other thn popultion ging. 1 Nonetheless, the rising shre of the ged in the popultion will inevitbly boost outlys on trnsfer progrms in which eligibility nd nnul spending re closely linked to n pplicnt s ge. Prtly offsetting higher spending needs for the dependent old is the shrinking need to furnish support to the dependent young. Some of this burden is finnced through public budgets, primrily s spending on the ntion s schools nd universities. Government spending on eduction represented 5.1 percent of U.S. GDP in 2000, bout one-hlf percentge point less thn spending in 1975, when the frction of the popultion between ges 5 nd 24 ws ner post-wr high. Little of the budgetry sving fter 1975 is reflected in federl outlys, becuse n overwhelming shre of public spending on eduction is finnced by stte nd locl governments. Public support for the dependent young is much less importnt thn support received through privte household budgets, however. Most consumption of the young is finnced out of the - 2 -

4 erned incomes of prents nd other older reltives. If working ge dults do not support s mny children, their erned incomes cn obviously sustin higher level of consumption for themselves. One wy to mesure the burden imposed by young nd old dependents is to clculte the rtio of young nd old to the number of working ge dults. The Socil Security Actury hs performed this clcultion under the ssumption tht the working ge popultion is between 20 nd 64 yers old. The Actury s clcultions imply tht the pek dependency rte fter World Wr II ws ttined in 1965, when there were 95 Americns under ge 20 nd pst ge 64 for every 100 dults between 20 nd 64 yers old. The 1965 dependency rtio ws substntilly higher thn the rtio predicted for 2050, when the dependency rtio is expected to rech 80. On the other hnd, the totl dependency rtio in 2050 will be substntilly higher thn tody s rte, which is 49. The youth dependency rte hs fllen shrply since 1965 minly becuse the fertility rte hs declined. In the future, working ge dults will support substntilly more dependents thn they do tody, but fewer thn they did in The crisis of supporting lrge future dependent popultion will evidently involve smller burden thn ws borne by working ge dults in the 1960s. The Actury s mesure of the dependency burden implicitly ssumes the cost of supporting n ged person is equl to tht of supporting person under ge 20. If the ged hve greter consumption needs thn children, the future increse in dependency burdens is understted by the Actury s clcultions. Suppose n verge child consumes α times the verge dult s consumption, where 0 < α < 1. If ggregte output is Y, then the verge consumption of n dult is (1) C Y 19 α P = 89 = 0 + P = 20 where P is the popultion in ge group nd people ged 20 through 89 re ssumed to be dults. The stndrd dependency mesure implicitly ssumes tht ll working-ge dults re contributing to the support of the dependent young nd old, wheres in fct lbor force prticiption rtes re considerbly below 100 percent, even mong dults who re in the middle - 3 -

5 of their potentil work creers (see Tble 1). If the frction of the popultion tht works t ech ge,, is λ nd the longest lived person dies on his 90 th birthdy, then output is (2) Y = W 89 λ = 0 P, where W is the economy-wide verge wge. 2 Suppose tht ll dults shre eqully in consumption, while ll children receive identicl consumption llotments equl to α times the dult shre. This distribution scheme requires ech worker to give up prt of his output to support dependents, tht is, nonworking members of the dult nd child popultions. The portion of the wge tht ech worker must scrifice, τ, is one mesure of the burden tht ged nd youth dependents impose on ctive workers: W C 0 (3) τ = = Totl 19 W α ( ) + ( ) α λ P P P 20 λ P. The dependency burden cn be further divided into the portions tht re due to the young ( = 0,20), nonged dults ( = 20,64) nd the old ( = 65,89): (4) (4b) (4c) τ τ τ Young = Nonged dults Elderly = α = α P ( ) α 89 α ( ) 1 P + ( ) 1 P λ P + P 89 P 20 + λ P λ P ; P ; nd The intuition behind this mesure of the dependency burden is strightforwrd. To the extent tht the lbor force prticiption rte of children flls below α or the lbor force prticiption rte of dults flls below 100 percent, members of the ge group will require greter support from the working popultion. The dependency burden rises long with the percentge of the - 4 -

6 popultion in ge groups tht hve low lbor force prticiption rtes. It flls when these groups shrink s percentge of the totl popultion. Cutler et l. (1990) estimte tht the eductionl spending nd consumption requirements of child represent 72 percent of the consumption requirements of n dult, implying tht α = Using this estimte of α nd informtion on the ge structure nd lbor force prticiption rtes in selected yers, it is possible to construct estimtes of the dependency burden imposed by nonworkers. Tble 2 shows estimtes of the U.S. dependency burden nd its components in selected yers between 1950 nd Columns 1 through 3 disply estimtes of the dependency burden under the ssumption tht ge-specific lbor force prticiption rtes remined unchnged t their 1950 levels (see right-hnd columns in Tble 1). Entries in these columns show, s expected, tht the youth dependency burden rose between 1950 nd 1965 nd then fell shrply. The burden imposed by elderly dependents rose over the full period, but by less thn the youth dependency rte declined. Consequently, the overll dependency burden ws smller in 2000 thn it ws in either 1950 or Columns 4 through 6 show the trend in dependency burdens under the ssumption tht the ge-specific lbor force prticiption rte observed in 2000 will remin in effect throughout the period. The entries in these columns show steep drop in the youth dependency burden between 1965 nd 2000 nd continued though more modest decline through Becuse the old-ge dependency burden will climb much fster fter 2000 thn the youth dependency burden will fll, the overll dependency burden will rise lmost 6 percentge points between 2000 nd In spite of this rise, however, the overll dependency burden in 2050 will be only slightly higher thn the one borne by Americn workers in The trends just described re bsed on the ssumption tht prticiption rtes remin unchnged. In fct, prticiption rtes hve vried significntly over the pst hlf century. Femle prticiption rtes incresed fter 1950, nd mle prticiption rtes fell, especilly t older ges. The trend in femle prticiption hs produced the bigger impct on overll ctivity rtes. Age-specific prticiption rtes incresed for Americns between ges 15 nd 59 nd declined for people pst ge 60. To see the impct of chnging ctivity rtes on dependency burdens, compre columns 3 nd 5 in Tble 2. If 1950 ctivity rtes hd remined unchnged, the overll dependency burden would hve been virtully the sme in 2000 s it ws in Becuse lbor force prticiption rtes incresed fter 1950, the ctul dependency rte fell more - 5 -

7 thn one-fifth, declining to less thn 44 percent. Almost ll of the decline ws due to shrp drop in the burden of supporting nonworking nonelderly dults. Becuse the prticiption rte of working-ge dults incresed, the dependency burden of supporting nonworking prime-ge dults shrnk by lmost hlf. The drop in prticiption rtes mong the ged slightly incresed the burden of supporting the dependent elderly, but this effect ws swmped by the impct of rising ctivity rtes mong nonelderly dults. If prticiption rtes remin unchnged between now nd 2050, the overll dependency burden in 2050 will remin well below levels observed in the 1950s nd 1960s. Bsed on trends of the pst couple of decdes, it seems likely U.S. prticiption rtes will continue to rise, lthough much more slowly thn they did between 1950 nd If prticiption rtes continue to inch up, the future rise in the dependency burden will be smller thn implied in Tble 2. Impct of Reduced Fertility nd Mortlity At ny given time, the popultion ge structure is complex product of pst fertility, mortlity, nd migrtion ptterns. As fertility nd mortlity rtes chnge, the ge structure will lso vry, ffecting dependency burdens. Most people who worry bout the future burden of old-ge dependency recognize it is the result of lower birth rtes nd longer life expectncy. I now consider the impct of reduced fertility nd lower mortlity on the dependency burden in the context of n initilly constnt popultion ge structure. Such structure would emerge fter sufficiently long period in which the ge-specific fertility nd mortlity rtes remin constnt. To mke the clcultions meningful, I begin by clculting the constnt ge structure tht would emerge if the fertility rte nd ge-specific mortlity rtes of 1950 hd remined unchnged for n indefinite period. Age-specific mortlity rtes for 1950 re displyed in the fourth column of Tble 1. The totl fertility rte between 1942 nd 1965 verged bout Hd the 1950 fertility nd mortlity rtes remined constnt over n indefinite period, the ultimte popultion growth rte would hve been slightly fster thn 1.1 percent yer. 3 Using lbor force prticiption rtes observed in 1950, we cn clculte the dependency burden ssocited with this constnt ge structure. The totl dependency burden ws 58 percent of workers wges, consisting of τ Young = 27 percent, τ Nonged dults = 24 percent, nd τ Aged dults = 7 percent. Chnges in fertility. Figure 1 shows how the dependency rte is ffected if fertility flls fter The solid line shows the trend in the totl dependency burden if the totl fertility rte - 6 -

8 drops from 3.15 to 2.06, which hs been the verge U.S. fertility rte since At mortlity rtes previling in 1950, totl fertility rte of 2.06 would eventully produce very slow declines in popultion. The broken line shows the dependency burden if the totl fertility rte drops to 1.3, the pproximte fertility rte of low-fertility-rte countries in the OECD. In the bsence of immigrtion, this reproduction rte will eventully generte lrge nd sustined declines in popultion. The ultimte rte of popultion decline is 1.4 percent per yer, rte of decline which would cut popultion size in hlf every 50 yers. Under either scenrio, the immedite impct of fertility decline is to reduce the dependency burden. Since the number of dependent young in the popultion flls shrply, the burden of supporting them must initilly decline. Eventully, lower fertility will be reflected in slower rtes of growth in the work force nd higher required levels of support for the dependent elderly, who will represent growing frction of shrinking popultion. Under the ssumptions of this exercise, however, the extr dependency burden to support the ged nd other nonworking dults will never be s lrge s the reduced dependency burden rising from smller number of dependent youths. The totl dependency burden is ultimtely lower when the fertility rte is 2.06 or 1.3 thn it is when the fertility rte is This fct will bring little comfort to worker who focuses solely on the tx she must py to support the retired elderly. The old-ge dependency burden increses from 7 percent in 1950 to n ultimte rte of 12 percent when the fertility rte flls from 3.15 to It rises to n ultimte rte of 18 percent when the fertility rte flls to 1.3. Nonetheless, the hevier burden of supporting lrger ged popultion is more thn offset by the lighter burden of supporting smller number of children. This bsic finding is obviously ffected by the ssumptions of the exercise. Of specil importnce is the ssumed vlue of α, the rtio of child to dult consumption requirements. If α were ssumed to be 0.38 insted of 0.72, the ultimte dependency burden ssocited with fertility rte of 2.06 would be identicl to the burden ssocited with fertility rte of With either fertility rte, the ultimte dependency burden represents 51 percent of worker s gross wge. 4 Thus, if children s consumption requirements re smll reltive to those of dults, substntil drop in the fertility rte is likely to result in long-term increse in dependency burdens. Even in this cse, however, the shrinking need to provide for the consumption needs of children will substntilly offset the effects of lower fertility rte on old-ge dependency burdens

9 Longer life spns. A second fctor boosting old-ge dependency is lengthening life spns. Age-specific mortlity rtes hve declined substntilly over the pst hlf century, incresing life expectncy t birth by lmost eight yers. The trend hs directly contributed to humn hppiness, regrdless of its effect on workers burden in supporting others or providing for their own old ge. Figure 2 shows the impct of lower mortlity rtes on the totl dependency burden. All of the clcultions re bsed on n initil popultion tht hs the stble ge structure implied by ge-specific mortlity nd fertility rtes observed in The solid line in the chrt shows how the dependency burden is ffected by stedy decline in ge-specific mortlity rtes, holding constnt the 1950 birth rte. To estimte the impct of longer life expectncy, I ssume the decline in ge-specific mortlity rtes mirrors the chnges ctully observed between 1950 nd 2000 (see columns 4 nd 5 in Tble 1). For exmple, the mortlity rte of persons between 65 nd 69 fell 38 percent between 1950 nd To clculte the nnul rte of decline in mortlity rtes of 65-to-69 yer-olds, I ssume this rte of decline persists over the full period from 1950 to 2150, with the mortlity rte of 65-to-69 yer-olds flling 38 percent every 50 yers. This rte of mortlity decline produces mjor boost in verge life spns. Age expectncy t birth in 1950 ws 68 yers. Under the ssumptions of this exercise, the verge life spn of person born in 2065 will be 84 yers, dding lmost one-qurter to the length of typicl life. 5 Perhps surprisingly, this extension in life spns dds significntly to the number of yers person cn be expected to work, even if the ge-specific lbor force prticiption rtes of 1950 remin unchnged. Bsed on 1950 ctivity rtes, person fcing the mortlity rtes in the 1950 life tble would be expected to work for totl 28 yers over n verge life spn. A person fcing 2000 mortlity rtes would be expected to work for totl of 30 yers, n increse in work yers of bout 8 percent. While this is less thn the proportionl increse in life spns implied by improvements in mortlity between 1950 nd 2000, it is lrge enough to offset much of the increse in old-ge dependency burdens. A common view of life spn improvements is tht they dd to the length of life fter retirement, incresing the size of the popultion which requires support. Mny deths occur well before norml retirement ge, however. To the extent tht these deths re postponed, the ffected individuls cn contribute for longer period to their own support nd the support of others. A child who dies before reching dulthood - 8 -

10 contributes only trivilly towrd its own lifetime support. If its deth is postponed for thirty yers, its contribution towrd its own self support increses drmticlly. To determine the effects of fster mortlity improvements, I clculted the dependency burden when deth rtes decline twice s fst s the observed rte of improvement between 1950 nd Under this ssumption, the mortlity rte of 65-to-69 yer-olds flls 38 percent every 25 yers. The dependency burden under this lterntive ssumption is reflected by the top broken line in Figure 2. Under either ssumption regrding the decline in mortlity rtes, the totl dependency burden increses stedily but rther slowly fter If mortlity improves t the pce observed between 1950 nd 2000 nd lbor force prticiption rtes remin unchnged, the totl dependency burden will increse from 57.7 percent in 1950 to 58.4 percent in 2000, to 59.0 percent in 2050, nd to 59.8 percent in More rpid mortlity improvement would increse dependency burdens fster, but even if mortlity rtes fll t twice the rte experienced in the pst 50 yers the totl dependency burden will climb just 2.0 percentge points between 1950 nd The increse will be more rpid if children require less support thn ssumed here, but it will be much less rpid if lbor force prticiption rtes of dults between 20 nd 59 continue to rise. Combined impct. The lowest line in Figure 2 shows the combined impct of lower fertility rte nd reduced mortlity. It depicts the time pttern of dependency burdens if the totl fertility rte flls from 3.15 to 2.06 in 1950 nd ge-specific mortlity rtes begin to decline t the sme rte s ctully observed during the period. Under these ssumptions, the totl dependency burden flls shrply through 1965 nd then trends upwrd, returning to its 1950 level by 2015 nd grdully rising in cyclicl pttern over the following decdes. In 2050 the dependency burden clims one-hlf percentge point more of wges thn ws the cse in 1950, nd by 2150 the burden is 1.6 percentge points higher thn it ws in The perceived crisis in supporting future dependents my nonetheless look serious in By tht yer workers hve lredy benefited from shrp drop in the burden of providing for children but they still fce big future increses in the cost of supporting lrger ged popultion. Even smll increses in dependency burdens cn look dunting if the gross wge remins constnt or grows slowly. Figure 3 shows the trend in workers net wges fter subtrcting support contributions to the dependent popultion. (Net compenstion is mesured in logrithms - 9 -

11 to mke it esier to detect chnges in the rte of growth.) The lower broken line in the figure shows the net wge fter subtrctions of the dependency burden for ll nonworking members of the popultion, including children nd prime-ge dults s well s the elderly. The top broken line shows the net wge fter subtrcting the dependency burden solely for the dependent elderly. The light solid lines indicte the trend in net wges tht would result if there were no demogrphic shock in These lines simply show the underlying rte of growth in gross wge rtes, becuse dependency burdens do not chnge under stble ge structure. In performing the clcultions, I ssume gross wges increse 0.4 percent yer. This is less thn one-qurter the ctul rte of U.S. wge growth between 1950 nd Even with this very conservtive ssumption regrding nnul wge growth, net wges usully rise from yer to yer, though the rte of increse is sometimes extremely slow. The lower line in Figure 3 suggests tht the initil impct of the demogrphic shock is to llow workers to enjoy fster net wge growth. Net wges climb fster thn the gross wge between 1950 nd 1965, becuse ctive workers support shrinking number of child dependents. However, net wge growth slows drmticlly fter By 2040 workers obtin the sme net wge s they would hve received without demogrphic shock in In yers fter 2040, the net wge grows 0.07 percent per yer more slowly thn would hve been the cse in the bsence of the demogrphic shock. The trend in net wges looks considerbly different if workers ignore their gins from lower youth dependency rte nd focus only on the tx contribution required to support the retired elderly. After subtrcting higher contributions for the old-ge dependency burden, net wges do not grow t ll in the 45 yers fter In lter yers, the wge fter subtrcting the old-ge dependency tx grows 0.08 percent per yer more slowly thn would hve been the cse without the demogrphic shock. Genertionl Burdens The nlysis so fr hs focussed on the dependency burden borne by ctive workers t prticulr point in time. From this perspective, the burden of popultion ging cn be summrized by the implicit tx pid by ctive workers to support child nd dult dependents who do not work. This frmework does not fully cpture the influence of chnging popultion structure on successive genertions, for it fils to mesure the lifetime net benefits tht individuls derive from membership in reltively lrge or smll genertions. Individuls receive trnsfers when they re children nd, if they survive to join the work force, eventully provide -10 -

12 for their own support nd contribute towrd the support of others. If n individul retires in old ge, he gin becomes dependent on support from others. Depending on the reltive size of the working-ge nd dependent popultions over the course of n individul s life, workers my be net tx pyers or net trnsfer recipients during their lifetimes. It is nturl to sk how vritions in fertility nd mortlity ffect the net trnsfers received by successive genertions. To nlyze this question, I hve clculted the rtio of birth cohort s lifetime consumption to its lifetime gross wges: Lifetime consumption α B = 0 (5) = 89 Lifetime wges 19 C + + λ W Fvored birth cohorts receive higher rtio of consumption to wges thn less fvored cohorts. I hve clculted this rtio for successive cohorts ffected by demogrphic shock. As before, I ssume the ntion begins with stble ge structure in which mortlity probbilities re the U.S. rtes previling in 1950 nd the totl fertility rte is Immeditely fter 1950 ge-specific mortlity begins to decline t the rte observed in the United Sttes between 1950 nd 2000 nd the totl fertility rte drops to 2.06, the verge U.S. fertility rte in the 1990s. Age-specific lbor force prticiption rtes sty unchnged before nd fter the demogrphic shock, remining t their 1950 levels (see Tble 1). The top pnel in Figure 4 shows the net trnsfers received by three brod ge groups before nd fter the demogrphic shock. Net trnsfers re clculted for youth (people ged 0-19), nonged dults (people ged 20-64), nd the elderly (people 65 nd older). In 1950, immeditely before the shock occurs, 22.0 percent of ggregte output is trnsferred to youth nd 4.5 percent is trnsferred to the elderly. Working-ge dults py net txes equl to 26.5 percent of ggregte output. As result of the drop in fertility, trnsfers to children decline while trnsfers to the elderly grdully increse. Note tht the initil impct of these two trends is to reduce the mount of ggregte output tht must be trnsferred from working-ge dults. The percentge of ggregte output tken from 20-to-64 yer-olds shrinks from 26.5 percent in 1950 to 23.4 percent in 2000 before rising to 27.1 percent by For people who re between 20 nd 64 yers old in 2000, the temporry reduction in dependency burdens cn improve the rtio of lifetime consumption to lifetime wges. = 0 p B+ 89 C = 20 p B+ p

13 To see this, consider the lifetime net trnsfers received by three birth cohorts (middle pnel of Figure 4). The oldest cohort is born in 1860 nd dies before the strt of the demogrphic shock. To compute the exct trnsfers received by ech cohort, I ssume tht the economy-wide gross wge remins constnt in every period. Members of the oldest cohort receive nnul net trnsfers equl to 21 percent of the gross wge when they re children. In their working-ge yers, they py nnul dependency txes equl 16 percent of the gross wge. 7 In old ge, they receive nnul net trnsfers mounting to 7.5 percent of the gross wge. The trnsfers they receive s children re much greter thn those they receive s old people, becuse reltively few of them survive long enough to collect substntil trnsfers in old ge. The lifetime consumption of the 1860 birth cohort is just 96 percent of its lifetime gross wges. When the popultion is growing rpidly nd is reltively short-lived, ech genertion contributes more to the support of young nd ged dependents over its life spn thn it receives in trnsfers during youth nd old ge. Reders who re fmilir with the Smuelson (1958) nd Aron (1966) expositions of the overlpping genertions model my be surprised to find tht members of growing popultion consume less over their lifetime thn they produce, even though they receive trnsfers from younger nd lrger genertions when they rech old ge. Smuelson nd Aron focus on the problem of finncing consumption of the ged. They do not consider the problem of supporting the dependent young. If the cost of supporting ech young person is sufficiently lrge, growing popultion cn impose youth dependency burdens on workers which re high in reltion to the gins workers obtin in old ge from being supported by lrge working-ge popultion. A criticl prmeter in this clcultion is α, the rtio of per cpit child to dult consumption. In the clcultions shown in Figure 4, α is equl to 72 percent. If it is hlf this lrge (α = 0.36), then under the other ssumptions of this exercise the genertion born in 1860 consumes lmost exctly wht it produces. If α = 0, ech genertion in stedily growing popultion consumes 6 percent more thn its lifetime gross wges. The sitution of the cohort born in 1910 is noticebly different from tht of the older cohort. Ech member of the cohort receives exctly s much net trnsfers in childhood s people born in 1860, becuse the ge structure nd dependency burden re ssumed to remin unchnged until the 1910 cohort reches dulthood. At ge 40, however, members of the 1910 cohort re ffected by the drop in fertility. Lower fertility reduces the number of children -12 -

14 supported by ech wge erner, lessening the overll dependency burden. As result, the 1910 cohort pys lower net txes during its working-ge yers thn erlier cohorts. When it reches retirement ge the 1910 cohort receives more trnsfers thn older cohorts, in prt becuse it enjoys longer verge life spn. Over its lifetime, the 1910 birth cohort enjoys consumption equl to percent of its lifetime gross wges. This is substntilly higher rtio thn obtined by erlier cohorts. The youngest cohort, born in 2060, receives slightly more trnsfers in childhood thn older cohorts nd lso receives substntilly more trnsfers in old ge. Both chnges re due to improvements in longevity. Compred with erlier genertions, lrger percentge of the cohort born in 2060 survives through childhood nd much bigger frction survives to dvnced old ge. The combintion of lower youth dependency rte nd higher old-ge dependency rte modestly increses the net txes the cohort must py between ges 20 nd 64. On blnce, the consumption enjoyed by the cohort exceeds the gross wges it erns by 1.2 percent of its lifetime wges. Thus, when α = 0.72 the long-term effect of lower fertility nd grdul mortlity decline is to improve the consumption / wge rtio. The ssumption of zero wge growth is not very relistic. The bottom pnel in Figure 4 shows the impct of stedily rising wges on the clcultions. I ssume tht the rte of wge increse, ω, is 1.69 percent yer, the 50-yer U.S. verge since To clculte net trnsfers in meningful wy, I mesure ech yer s net trnsfer in reltion to the economy-wide rel wge rte, specificlly, the previling rel wge when the cohort reches dulthood (ge 20). When the wge rises throughout ech genertion s life spn, net nnul trnsfers received in old ge cn be substntilly lrger thn nnul trnsfers received s child, becuse they re finnced out of much bigger nnul output. Of course, rising wge lso mens tht the net trnsfers cohort must py during its working yers will be substntilly lrger thn the net trnsfers it receives in youth. Becuse genertions re reltively short-lived in decdes before the 1950 demogrphic shock, the beneficil impct of wge growth on trnsfers received in old ge is more thn offset by the dverse impct on rel trnsfer dependency burdens during cohort s prime working yers. When rel wges re climbing 1.69 percent yer, the genertion born in 1860 consumes 95 percent s much s it produces. Most of the difference between lifetime wges nd lifetime consumption is trnsferred to youthful dependents, who re very numerous when the totl fertility rte is The rtio of lifetime consumption to lifetime -13 -

15 wges becomes much more fvorble for cohorts which enter the work force fter the birthrte flls nd mortlity rtes decline. For exmple, the genertion born in 2060 receives lifetime income, including net trnsfers, tht mounts to 107 percent of its lifetime gross wges. This rtio is considerbly more fvorble thn tht obtined by the cohort born in These results re summrized in Figure 5, which shows the rtio of lifetime consumption to lifetime gross wges for successive birth cohorts over two century period under four different sets of ssumptions bout the long-term fertility rte nd wge growth. To clrify the exposition, I hve modified the erlier ssumption regrding the vlue of α. I now ssume tht α = 0.36, exctly hlf the level suggested by Cutler et l. (1990). 8 If () the totl fertility rte is 3.15, (b) mortlity rtes nd lbor force prticiption rtes remin stble nd equl to U.S. rtes in 1950, (c) wge growth, ω, is zero, nd (d) α = 0.36, then under the ssumption of stble popultion structure, ech genertion will enjoy precisely s much lifetime consumption s it erns in gross wges: The rtio of lifetime consumption to lifetime wges is The two lower lines in Figure 5 show the impct of demogrphic shock tht begins in In both cses the ge-specific mortlity rte is ssumed to begin flling fter 1950 ccording to the U.S. pttern observed over the period, nd this pttern of decline continues stedily from 1950 through In one cse I ssume tht the totl fertility rte drops to 2.06 fter 1950, pproximtely the fertility rte observed in the United Sttes since In the lterntive cse I ssume tht the totl fertility rte drops to 1.30, pproximtely the fertility rte in low-fertilityrte countries in the OECD. Not surprisingly, the bigger drop in fertility initilly produces bigger gins for the cohorts lredy in the work force when the birth rte flls. The cohort born in 1920 enjoys $ in lifetime consumption for every $ it erns in lifetime wges. However, the genertion born in 1965 receives slightly less in lifetime consumption tht it erns in lifetime wges, nd the rtio of consumption to wges grows less fvorble in every succeeding genertion. For the genertion born in 2060, lifetime consumption is just 94.2 percent of lifetime gross wges. When fertility is extremely low nd the verge life spn is lredy long (nd lbor force prticiption rtes remin stble), declining mortlity rtes cn reduce the consumption possibilities of successive cohorts. Even though nnul wge rtes re ssumed fixed in this scenrio, it is not obvious whether the genertion born in 2060 would consider itself worse off thn the genertion born in 1920, which obtins much more fvorble rtio of lifetime -14 -

16 consumption to wges. The cohort born in 1920 hs n expected life spn of 71.4 yers; the cohort born in 2060 cn nticipte living for 83.7 yers, gin of 12.3 yers or bout 17 percent. When the birthrte fter 1950 flls to 2.06 rther thn to 1.30, the long-term prospects of successive genertions re considerbly brighter. Genertions born between 1865 nd 1930 obtin less fvorble rtio of lifetime consumption nd wges thn would be the cse if the fertility rte dropped to 1.30, but genertions born fter 1930 obtin much more fvorble rtio. Even the genertion born in 2060 receives surplus of lifetime consumption over lifetime wges, implying tht the trnsfers it receives from other genertions re lrger thn those it pys to support other genertions. This outcome is possible becuse popultion growth is slightly positive, implying tht ech genertion cn be supported in old ge by slightly lrger popultion of ctive workers. 9 The top two lines in Figure 5 show the effects of fster wge growth on these clcultions. I ssume tht gross wges rise 1.69 percent yer, the verge rte of U.S. wge growth fter I consider two fertility scenrios, one in which the totl fertility rte drops to 1.3, nd the second in which it flls only to As ws the cse when ssumed wge growth ws zero, higher long-term fertility rte produces more fvorble long-term rtio of lifetime consumption to lifetime wges. However, even very low fertility rte still produces fvorble long-term rtio of lifetime consumption to lifetime wges. The genertion born in 2060 receives $ in lifetime consumption for every $ it erns in lifetime wges. This is virtully identicl to the rtio obtined by the genertion born in 1850, when life spns were much shorter nd popultion growth ws significnt nd positive. The long-term outlook is even more fvorble if the popultion growth rte drops to 2.06, s it hs in the United Sttes, rther thn to 1.3, s hs been the cse in Jpn nd much of Western Europe. Conclusions This nlysis suggests tht the extr burdens connected with popultion ging re considerbly smller thn commonly supposed. Under plusible ssumptions, the trend towrd n older popultion my ctully reduce the lifetime consumption scrifice tht ech genertion must mke in order to support other genertions. To be sure, popultion ging implies tht the tx rte needed to support the retired elderly must rise, fct which hs been highlighted in most populr discussion. But this extr burden will be t lest prtly offset by reduced need to provide support to the young, who will become less numerous reltive to the ctive workforce -15 -

17 thn hs been the cse in the pst. Even if dults provide only modest consumption support to the dependent young, the lrge drop in the youth dependency rtio implied by current birth rtes will offset much of the extr burden of supporting lrger retired popultion. The extr burden of n older popultion will be smller still if wge growth remins positive nd if lbor force prticiption rtes mong the working-ge popultion continue to rise. Three fctors my ccount for widespred pessimism concerning the support burdens implied by current demogrphic trends. First, s noted by Cutler et l. (1990), the rich countries hve lredy derived much of the consumption benefit to be gined from lower youth dependency rte, but they hve not yet experienced the full impct of higher old-ge dependency burdens. In terms of Figure 5, recent genertions of ctive workers nd retirees hve enjoyed the consumption dvntges ssocited with rising rtio of lifetime consumption to lifetime wges, but future workers will fce the inevitble unwinding of prt or ll of this dvntge. If future wge nd popultion growth rtes re low, future workers my hve to trnsfer more to the retired elderly thn they cn ever expect to receive themselves s trnsfers in old ge. Disregrding the considerble improvement in their own life spn compred with tht of erlier genertions, future genertions my consider themselves worse off thn erlier genertions tht hd the opportunity to consume more thn they produced during their lifetimes. Second, mny pessimists my believe the public nd privte trnsfers needed to support n ged dult re much greter thn those needed for dependent child. For exmple, Cutler et l. (1990) estimte tht the medicl nd non-medicl consumption needs of person older thn 64 re 1.27 times those of yer-old nd 1.76 times those of child. Even under this ssumption, however, the extr dependency burden of lrger elderly popultion will be substntilly offset by reltively smller popultion of dependent children. Moreover, the ultimte burden of old-ge support does not depend on the mount of support needed to mke n ged dult s well off s nonged dult. It depends on the ctul level of support provided to the ged. It my be the cse tht n ged dult requires 1.27 times s much income s nonged dult to stisfy the sme medicl nd non-medicl consumption needs. However, if society enforces distributionl rule tht provides n identicl consumption shre to every dult, whether ged or nonged, the retired elderly must ccept consumption llotment tht does not fully stisfy their consumption needs (ssuming their needs require them to receive 1.27 times the nonged dult llotment to ttin the sme level of hppiness). I see little evidence tht -16 -

18 OECD countries provide enough support to their retired elderly so tht the verge income of the elderly is 27 percent greter thn tht of nonged dults. The best evidence is tht the equivlent income of the elderly in rich societies is pproximtely equl to or less thn tht of the nonged (Bosworth nd Burtless, 1998). Finlly, workers view tx contributions to support the retired elderly in very different light thn they regrd most trnsfers to the dependent young. On the one hnd, most Americns regrd public trnsfers to the elderly more fvorbly thn public trnsfers to children nd nonged dults. However, they regrd public trnsfers in much less fvorble light thn they do their privte trnsfers to nonworking dependents within their own fmilies. Fmily bredwinners derive greter stisfction in providing for the consumption needs of their own spouses nd child dependents thn they do in pying txes to support the comfortble retirement of strngers. Even though bredwinners in rich countries hve enjoyed sizble consumption gins from lower fertility, they do not view these gins s n offset to the higher txes they must py to support the retired elderly. The perceived crisis of popultion ging my be nturl result of the different wys we support ged nd nonged dependents. This should not obscure the fct tht the true economic burden of n ging popultion is fr different nd less worrisome thn populrly supposed

19 ENDNOTES 1 Becuse of the incresed sophistiction, utiliztion, nd cost of medicl goods nd services, federl helth cre spending would be expected to increse fster thn ntionl income, even if the popultion ge structure remined unchnged. 2 If the wge rte vries by ge, λ my be re-defined to designte the product of group s prticiption rte nd the rtio of the group s wge reltive to the economy-wide wge. 3 To simplify the clcultions, I ssume the potentil life spn hs up 18 periods corresponding to the 5-yer intervls shown in Tble 1. The mximum potentil life spn is thus ssumed to be 90 yers. 4 If the fertility rte flls to 1.30 rther thn to 2.06, the ultimte dependency burden rises slightly compred with the dependency burden when fertility is 3.15 or If α = 0.38, the extr consumption needs of lrger elderly popultion cn outweigh the cost svings from reduced spending on children. Even in this cse the chnge in dependency burdens is smll, however. The ultimte dependency burden with totl fertility rtes of 3.15 or 2.06 is 51 percent of the gross wge. When the totl fertility rte flls to 1.3, the totl dependency burden ultimtely rises to 52 percent. 5 Becuse I use the simplifying ssumption tht life spn cn extend no longer thn 90 yers, my clcultions understte the increse in life expectncy tht will occur if mortlity rtes continued to decline for the next 150 yers. Some of the improvement in mortlity rtes will improve life expectncy t ge 90, nd this improvement is ignored in the clcultions. 6 Rel finl consumption per ctive worker in the United Sttes incresed 1.69 percent yer between 1950 nd Rel NNP per ctive worker incresed t the sme nnul rte. Even though the pce of rel wge growth slowed fter 1975, the nnul rte of increse in consumption per worker incresed more thn three times fster thn the 0.4 percent rte of growth ssumed in Figure 3. 7 Note tht this is substntilly less thn the dependency burden fced by working dult. Under the ssumptions of the model, working-ge dults who do not work receive trnsfers from the people who work. The net trnsfer shown in Figure 4 indictes the difference between trnsfers received by members of n ge group nd support contributions mde by working members of tht group. 8 If we ssume higher reltive consumption level mong children, the effect of lower totl fertility rte on the rtio of lifetime consumption to lifetime wges is even more fvorble to future genertions. 9 Although totl fertility rte of 2.06 is consistent with stble or slightly declining popultion under the mortlity rtes in effect in 2000, predicted future declines in the mortlity rte will reduce the number of children needed to ensure stble or slightly growing popultion. An incresing proportion of ech genertion will survive through the child-bering ges, so ech womn will not need to ber s mny children in order to mintin fixed popultion

20 REFERENCES Aron, Henry J. (1966) The Socil Insurnce Prdox, Cndin Journl of Economics 32 (August), Bosworth, Brry, nd Gry Burtless (1998) Popultion Aging nd Economic Performnce, in Brry Bosworth nd Gry Burtless, eds., Aging Societies: The Globl Dimension (Wshington: The Brookings Institution), Cutler, Dvid M., Jmes M. Poterb, Louise M. Sheiner, nd Lwrence H. Summers (1990) An Aging Society: Opportunity or Chllenge? Brookings Ppers on Economic Activity (Spring), Smuelson, Pul A. (1958) A Exct Consumption-Lon Model of Interest with or without the Socil Contrivnce of Money, Journl of Politicl Economy 66 (December),

21 Tble 1. U.S. Popultion Age Structure, Mortlity, nd Lbor Force Prticiption, 1950 nd 2000 Percent Age group Percent of U.S. popultion Mortlity rte / Activity rte b/ Difference Difference Difference / Percent of persons live t the first indicted ge who die within five yers fter ttining tht ge. b/ Persons who re employed or seeking work s percent of ll persons in ge group. Source: Author's tbultions of dt from U.S. Census Bureu, Bureu of Lbor Sttistics, nd Ntionl Vitl Sttistics Reports (Vol. 47, no. 13). Tble 2. U.S. Dependency Burden under Alterntive Assumptions Regrding Lbor Force Prticiption, Percent 1950 Activity Rte 2000 Activity Rte Dependency burden (1) (2) (3) (4) (5) (6) τ Young τ Nonged dults τ Aged dults τ All dependents Source: Author's clcultions s explined in text

22 Figure 1. Impct of Lower Fertility nd Lower Mortlity on Dependency Burden, Figure 1. Effect of Reduced Fertility 62% 62% Totl dependency burden (% of wge) 60% 58% 56% 54% 52% 60% 58% 56% 54% 52% 50% 50% TFR drops to 2.06 TFR drops to 1.3 Figure 2. Impct of Lower Mortlity Rte 62% 62% Totl dependency burden (% of wge) 60% 58% 56% 54% 52% Reduced mortlity combined with lower fertility rte 60% 58% 56% 54% 52% 50% 50% Mortlity decline Fster mortlity decline Both mortlity nd fertility decline -21 -

23 Figure 3. Trend in Net Wge fter Demogrphic Shock 5.5 Logrithm of fter-tx wge Wge fter old-ge tx Wge fter ll dependents tx

24 Figure 4. Age Pttern of Net Trnsfers to Dependents, Before nd After Demogrphic Shock 30 Net trnsfers to ech ge group s percent of ggregte output, by clendr yer % of totl output Yer % of gross wge when cohort ttins ge Annul per cpit trnsfers to members of birth cohort s percentge of gross wge when cohort becomes dult: Annul wge growth = 0% per yer Birth yer of cohort % of gross wge when cohort ttins ge Annul per cpit trnsfers to members of birth cohort s percentge of gross wge when cohort becomes dult: Annul wge growth = 1.69% per yer Birth yer of cohort

25 Lifetime Consumption s % of Lifetime Wges Figure 5. Rtio of Lifetime Consumption to Lifetime Wge in Successive Birth Cohorts (α=0.36) 115% TFR = % ω = 1.7% 105% TFR = 1.30 TFR = % ω = 0.0% 95% TFR = % Birth yer of cohort -24 -

26 -25 -