Aging, Government Finances and International Capital Flows: Projections for Japan. November 2002

Transcription

1 Aging, Governmen Finances and Inernaional Capial Flows: Projecions for Japan November 2002 Rober Dekle Deparmen of Economics USC Los Angeles, CA 90089

2 2 I. Inroducion. Over he nex several decades, Japan s populaion will be ageing rapidly. In 1955, only 5.5 percen of he populaion was 65 years or older; by 1998, 16.2 percen were elderly. Projecions imply large increases in he elderly in he nex 20 years; by 2015, 25 percen of he populaion will be 65 or above. The main reason for his aging is he fall in he oal feriliy rae (birhs per family). The oal feriliy rae was more han 4 before 1949, declining sharply o 2.1 in I has begun o fall again since 1974 and he curren level of 1.4 was reached in There is sill lile sign ha his has sabilized or reurned o a higher level. In his paper, we revisi he impac of demographic change on Japanese saving and invesmen, and governmen budge deficis. There is widespread belief ha rapid aging will lead o major shifs in he Japanese saving and invesmen balance, and severely worsen Japan s fiscal siuaion. Using he laes governmen demographic projecions, we show ha he aging of he populaion underway will seadily lower Japan s oal saving rae from 30 percen of GDP oday o 19 percen of GDP in Japan s oal invesmen rae will decline from 28 percen of GDP oday o abou 22 percen of GDP in Given he more rapid decline in oal saving, Japan s curren accoun will seadily narrow from is curren level, and urn o defici around We also show ha he aging of he populaion will worsen governmen finances, as healhcare and social securiy spending soar. Unless governmen fiscal balances improve from he curren minus 7 percen of GDP o almos plus 5 percen of GDP over he nex decade or so, he curren governmen deb is no susainable. In he paper, we forecas fuure governmen spending from projeced demographics. Given he forecased governmen spending, large ax

3 3 increases will become necessary for he curren governmen deb o be susainable. In fac, we show ha axes as a percenage of GDP will need o be raised from he curren 28 percen o almos 50 percen by Mos of he earlier lieraure projecing he impac of demographic change on he Japanese saving-invesmen balance and governmen deficis daes back almos a decade. The daa are correspondingly almos a decade old. The earlier lieraure assumed fuure populaion growh raes ha are consan, and he economic projecions relied on ad-hoc behavioral assumpions. We allow fuure populaion growh raes and suppor raios (he raio of he labor force o he populaion) o change every 5 years, and our projecions are grounded in wellacceped microeconomic foundaions. Our paper is organized as follows. In Secion II, we briefly review he lieraure on, and pas rends in, Japanese privae and governmen saving raes and privae and public invesmen raes. In Secion III, we review he deerioraing Japanese governmen fiscal posiion in he 1990s. In Secion IV, we summarize he demographic changes undergoing in Japan, and presen he Japanese governmen s laes demographic projecions. In Secion V, we simulae he impac of demographic change on he fuure Japanese saving and invesmen raes, governmen deficis, and he curren accoun. In our simulaions, we adop he sandard small-counry, open capial marke markes, Ramsey opimal growh model. Specifically, we closely follow Culer, Poerba, Sheiner, and Summers (1990) modificaions o he Ramsey model, in examining he impac of changing demographics on savings and governmen deficis. Secion VI concludes. II. Pos-War Japanese Saving and Invesmen.

4 4 I is well-known ha he pos-war Japanese economy is characerized by very high saving and invesmen raes. In fac, Japan s saving raes are among he highes in he world only Ialy, Singapore, and Taiwan have higher saving raes. However, hese high Japanese saving and invesmen raes are primarily a pos-war phenomenon in fac a pos 1955 phenomenon. If he period of he Korean War is excluded, Japan s saving rae did no make i ino he double digis unil 1955, a full 10 years ino he pos-war period. Thus, we can immediaely rejec he view ha Japan s high saving rae is he resul of culural facors such as naional characer or Confucian and Buddhis eachings, because alhough culural facors were sronger in he prewar period, he saving rae was lower. The rends and flucuaions in Japanese saving and invesmen closely mirror he rends and flucuaions in Japanese GDP (Figure 1). For boh saving and invesmen raes, here is clear posiive associaion wih he growh in GDP, especially unil The broad rends in pos-war Japanese privae and governmen saving raes, invesmen raes, and he ne expor surplus GDP raios, are depiced in Table The privae saving rae rose seadily beween 1955 and he mid-1970s, peaking (firs) in Subsequenly, he rae fell unil he early 1990s, when i rose (again) o reach is pos-war 1 We depic gross saving and invesmen. Gross saving includes he depreciaion includes he depreciaion on capial. In his paper, we use gross, insead of ne saving because he laer requires daa on depreciaion. There is enormous conroversy regarding he proper measuremen of he capial depreciaion rae in Japan, and he use of gross savings allows us o sidesep his conroversy (Dekle and Summers, 1991; Hayashi, 1991; Horioka, 1995). 2 The privae secor includes households, privae unincorporaed non-financial enerprises, and corporaions. Corporae saving is small in Japan, and if households pierce he corporae veil, corporae saving can be considered par of household saving. The governmen secor includes he cenral, prefecural, and local governmens. Governmen saving excludes governmen invesmen.

5 5 peak in There is a voluminous lieraure ha seeks o explain he paern and level of Japanese pos-war privae saving. 3 The lieraure suggess ha he mos imporan reason for Japan s high privae saving rae is rapid economic growh. The permanen income/life cycle hypohesis can explain he posiive impac of income growh on he privae saving rae if income growh is faser han expeced. This hypohesis may have been valid unil he early 1970s. The surge in privae saving from he mid-1970s o he early 1980s is relaed o he wo oil crisis in he 1970s. The explanaion given is ha hese crisis added furher fuel o he already rampan inflaion and precipiaed a recession, which in urn raised uncerainy abou he fuure and increased he perceived need o save for precauionary purposes. The fall in privae saving from he mid-1980s o he early 1990s is because of robus consumpion, simulaed by rising sock and land prices. In conras, he mid- o lae-1990s rise in privae saving is relaed o he recessionary economy, increases in unemploymen, uncerainy, and pessimism, all raising precauionary saving. Horioka (1991, 1992) finds ha he level and growh of Japanese GDP explains abou 65 percen of he variaion in he privae saving rae. 4 The lieraure suggess ha he second mos imporan reason or Japan s high privae saving rae is he favorable age srucure of he populaion. Unil he early 1970s, he proporion of he aged (over 65) o he working-age populaion (20-64) he so-called dependency raio was low in Japan. According o he life-cycle hypohesis, an increase in he dependency 3 See Horioka (1990), Dekle (1993), and Hayashi (1998) for a caalogue of reasons for Japan s high privae saving. 4 Horioka s resuls, however, mus be inerpreed wih cauion, since he includes variables wih differen orders of inegraion, I(.), in he same esimaing equaion. His demographic variables are I(2), bu he level and growh of GDP are I(0), and I(1), respecively.

6 6 raio has a significan negaive effec on he privae saving rae. In addiion, mos oher models including hose wih dynasic households predic a negaive relaionship beween he dependency raio and he privae saving rae. Horioka (1991,1992) finds ha adding he dependency raio o he equaion already including he level and growh of GDP raises he proporion of privae saving explained from 65 percen o 75 percen. Moreover, he esimaes ha a 1-percenage poin increase in he dependency rae will cause he privae saving rae o decline by 1 percenage poin. These and similar esimaes sugges ha he 12-percenage-poin increase in he dependency rae beween 1975 and 1998 has depressed privae saving by abou 12 percenage poins annually. The governmen saving rae rose unil he mid-1960s, hen gradually fell o is hisorical low in Subsequenly, he rae rose (again) unil he early 1990s, when i sared o decline o (almos) is new low in The rend in Japanese governmen saving is also closely relaed o economic growh. Governmen saving surged unil he mid-1960s, as growh raes were consisenly above governmen projecions, leading o rising ax revenues. From he mid-1960s, however, he demand for governmen services increased, dampening he budge surpluses. The recessionary 1970s led o couner-cyclical measures and a furher drop in governmen saving. To hal he decline in governmen saving, he Japanese governmen in he early 1980s inroduced budge freezes and reformed he ax sysem. These measures and srong economic growh in he mid- o lae 1980s led o rising budge surpluses. However, as he economy slumped in he early 1990s, falling ax revenues and he need for expansionary fiscal policy again depressed governmen saving raes. The invesmen rae also rose seadily, peaking in Since hen, i has fallen slighly.

7 7 Compared o household and governmen saving raes, he invesmen rae has remained comparaively sable. The main deerminan of Japanese invesmen has again been economic growh. 5 As GDP growh surged in he 1950s and 1960s, invesmen was able o grow o ake advanage of newly available echnologies. Since he early 1970s, he invesmen rae dipped somewha, bu has remained a a high level. The surge in invesmen raes in he lae 1980s is relaed o he cheap financing available o firms, owing o rising sock and land prices. Alhough privae invesmen has dipped in he 1990s, rising governmen invesmen owing o expansionary public works projecs in he mid- o lae 1990s has kep overall invesmen raes high. Japanese ne expors were in persisen defici unil he early 1970s, reflecing srong invesmen demand, bu inadequae saving. However, by he mid-1980s, he surge in saving and decline in invesmen pushed Japanese ne expors (as a percenage of GDP) ino record erriory. Subsequenly, as a resul of srong domesic consumpion in he lae 1980s and srong governmen invesmen in he 1990s, he ne expor surpluses (as a percenage of GDP) declined. III. The Japanese Fiscal Posiion in he 1990s. Governmen saving declined and public invesmen rose in he 1990s (Table 1). These rends in governmen saving and invesmen in he 1990s were caused by he recession, and also by srucural changes. The recession and he decline in he rae of economic growh lowered ax revenues. Srucural changes worsening governmen saving include ax reforms ha lowered ax elasiciies and ax revenues, and he aging of he populaion, which raised social securiy and 5 Kiyoaki and Wes (1996) find ha Japanese privae plan and equipmen invesmen beween 1961 and 1994 can be well explained by he flexible acceleraor model, wih lagged oupu as he sole explanaory variable.

8 8 healhcare expendiures. The deerioraion of governmen finances led o sharp increases in ousanding governmen bonds, raising concerns abou fiscal susainabiliy, and calls for fiscal reform. Governmen saving in he 1990s. Tax revenues declined because of he recessionary environmen of he 1990s. In addiion, governmen consumpion increased. Owing o he low cyclical variabiliy of Japanese unemploymen and social welfare benefis, however, governmen consumpion increases during he recession were capped. Governmen saving can be divided ino he full-employmen and cyclical componens. We esimae ha during he period , Japan s full-employmen governmen saving was abou 2.6 percen, slighly higher han acual governmen saving of 2.0 percen, leaving he cyclical componen of governmen saving a -0.6 percen. 6 Thus, much of he decline in Japanese governmen saving in he lae 1990s was no because of auomaic sabilizers, bu because of srucural facors, such as ax reducions. This low cyclical variabiliy of governmen saving is corroboraed in a recen IMF sudy showing ha a one-percenage poin increase in he oupu gap ranslaes ino an increase of he cyclical defici by abou a hird of 1 percen of GDP, which is abou half of he defici response in oher OECD counries (Muhleisen, 2000). Governmen saving declined since he early o mid-1990s, wih ax reducions supporing 6 We esimae he full-employmen governmen saving by regressing governmen saving on he oupu gap and a consan. We inerpre he esimaed value of he consan; which is he governmen saving rae when he oupu gap is equal o zero--as full-employmen governmen saving.

9 9 aggregae demand in he face of an unprecedened economic downurn. Paricularly in 1998, when he economy slipped ino recession, he governmen passed ax cu measures ha led o a subsanial decline in governmen saving in he following year. Marginal income and capial gains ax raes and healh insurance premia were cu, exempions for gif axes were raised, and ax deducions for home morgage holders were inroduced. The governmen also lowered corporae ax raes from 50 percen o 40 percen. Governmen saving can be broken down ino he social securiy surplus, he surplus in oher caegories, and healhcare expendiures (Figure 2). The social securiy surplus (benefis minus conribuions) fell from abou 2 percen of GDP in he early 1990s o abou 0.5 percen of GDP in 1999, owing o he recession (lowering conribuions) and increase in he elderly (raising benefis). Governmen healhcare expendiures rose from abou 3.6 percen of GDP in he early 1990s o abou 4.2 percen of GDP in he 1999, mainly owing o increase in he elderly, who use mos of he hospial services. However, he healhcare expendiure-gdp raio in Japan is smaller han in he U.S. (6.6 percen of GDP), or Germany (7.7 percen of GDP). The remaining caegory of governmen saving includes usual spending such as educaion, defense, and policing and firefighing. Saving in his caegory declined sharply from 9.5 percen of GDP o 4 percen of GDP, owing o he fall in (income and consumpion) ax revenues. Public Invesmen in he 1990s. Beween 1990 and 1999, he Japanese governmen passed 10 simulus packages, in an aemp o jump-sar he salling economy. The mos imporan componen of he governmen simulus packages were public works, which are included in public invesmen. However, as

10 10 shown in Table 1, he acual increases in public invesmen in he lae-1990s were raher moderae, compared o he prominen and headline grabbing role of public works in he simulus packages. There are wo reasons why acual public works fell shor of he levels announced in simulus packages. Firs, during he 1990s, he cenral governmen assigned roughly wo-hirds of he increased public works spending o local governmens (wihou providing a commensurae increase in funding). The capaciy, however, of local governmens o expand public invesmen was affeced by heir poor financial siuaion, and he coninued rise in public invesmen has increasingly been financed hrough local bond issues. The amoun of ousanding local governmen bonds increased from 12 percen of GDP in 1990 o 22 percen of GDP in Many local governmens surpassed he legally allowed hreshold of bonds-ousanding, and were pu under bond issuance resricions by he cenral governmen. Second, some of he public invesmen funds provided by he simulus packages remained unused, because of poor projec implemenaion. Ishii and Wada (1998) calculaed ha only percen of he simulus packages public works has ranslaed ino addiional demand during he mid- o lae 1990s. Governmen Deb and Liabiliies in he 1990s. The lae 1990s decline in governmen saving and rise in public invesmen led o sharp increases in governmen deb. Table 2 depics he fiscal balance-gdp raio, and several deb o GDP raios. The fiscal balance-gdp raio is lower han he difference beween he governmen saving-gdp raio and he public invesmen-gdp raio by abou 2 percen, mainly because of he inclusion of ne governmen land purchases in he fiscal balance. During he 1990s, he

11 11 governmen bough significan amouns of land from he privae secor o prop up land prices. The fiscal surplus declined coninuously in he 1990s, reaching abou minus 10 percen in Correspondingly, he raio of deb o GDP has risen sharply. By inernaional sandards, Japan s gross deb-gdp in 1999 was he highes among he G-7 counries Ialy s was 115 percen, and he U.S. s was 62 percen. Because of he parly funded naure of he Japanese pension sysem, as well as he governmen s major role in financial inermediaion, he Japanese governmen holds significan asses, keeping ne deb-gdp a a moderae level, and lower han in oher G-7 counries. However, since he asses of he social securiy sysem are more han offse by fuure pension obligaions, hey should be excluded when assessing Japan s deb siuaion. As a resul, Japan s ne deb excluding social securiy ne asses, a 85 percen, is significanly higher han in he U.S., 60 percen, and in Germany, 53 percen. The governmen s rue ne obligaions may be subsanially higher han he ne deb figures, because of unfunded liabiliies. There are hree main sources of unfunded liabiliies. The firs source are he fuure coss of governmen social securiy and healh schemes. Esimaes of fuure unfunded social securiy coss depend on demographic, economic growh, and ineres rae assumpions and range widely. In Japan, here are several social securiy schemes, bu he main scheme he Employees Pension Scheme derives one-hird of is (benefi) payous from governmen subsidies, and wo-hirds of is payous from payroll axes (conribuions). Given curren governmen subsidy and payroll ax raes, Chand and Jaeger (1996) esimae he presen (2000) value of Japan s unfunded social securiy liabiliies a 110 percen of GDP. Muhleisen (2000) esimae he presen value of ne unfunded liabiliies a 60 percen of GDP. Wih regards

12 12 o governmen healh benefis, on average, governmen subsidies cover abou 1/3 oal public healh insurance benefis (2 percen of GDP), wih he res covered by healh insurance conribuions and co-paymens. Given ha he elderly are exemp from healh insurance conribuions, and pay only small co-paymens, he fuure aging of he populaion is expeced o significanly raise he proporion of healh benefis covered by governmen subsidies. The second source of unfunded liabiliies are poenial losses on governmen asses. A porion of he governmen s asses represen sof loans ha may no be repaid. Many large public or join public-privae infrasrucure projecs financed from he Fiscal Invesmen and Loan Program (FILP) loans generae less revenue han budgeed, which may imply significan coningen liabiliies of he governmen. For example, much of he subsanial deb 3 percen of GDP--of he now privaized Japan Naional Railways is owed o FILP. Since mos of his deb will never be repaid, his deb will evenually have o be covered from he governmen budge. Oher public corporaions wih large accumulaed FILP deb include he Japan Highway (4 percen of GDP) and Housing and Urban Developmen Corporaions (2.5 percen of GDP). The hird source of unfunded liabiliies are he explici and implici governmen guaranees of privae secor lending. Explici guaranees are exended by FILP and oher governmen eniies o encourage lending by privae financial insiuions. Examples are guaranees of bank deposis by he Deposi Insurance Corporaion, and guaranees of lending by credi cooperaives o small- and medium-sized enerprises. Alhough hese guaranees do no enail fresh governmen lending, should he guaraneed loans no be repaid, he governmen mus cover he loans from he budge. The oal amoun of ousanding governmen-guaraneed bonds and loans amouned o abou 10 percen of GDP in Alhough hisorically, only abou 1

13 13 percen of governmen-guaraneed loans are never repaid, if he Japanese economy worsens, he percenage of unpaid loans could soar (Bayoumi, 1998). In addiion o he explicily guaraneed governmen loans and bonds, here are he implicily guaraneed governmen loans. Hisorically, he Japanese governmen has shown willingness o cover he irrecoverable problem loans of privae financial insiuions. For example, in 1998, he governmen auhorized 60 rillion yen (12 percen of GDP) in public funding o cover he irrecoverable loans of privae banks. 7 This willingness represens implici guaranees, and hese guaranees are coningen liabiliies of he governmen. In 2000, ousanding loans minus he capial and liquid asses of financial insiuions was abou 200 percen of GDP. If, as some bank analyss esimae, 10 percen of he loans are irrecoverable, hen he cos o he governmen of hese implici guaranees could be as high as 20 percen of GDP. Fiscal Susainabiliy and Inergeneraional Wealh Disribuion. The sharp increase in Japanese governmen deb in he 1990s has raised quesions abou he susainabiliy of his deb, and much policy work has been done in his area. Clearly, a curren Japanese governmen fiscal defici levels, he governmen deb will keep on growing. For given growh and ineres rae assumpions, he fiscal surplus exclusive of ne deb ineres paymens, or he primary fiscal surplus, necessary o sabilize he deb-gdp raio is: 7 The oal of public funds acually spen and included in governmen consumpion--in 2000 was abou 8 rillion yen (0.16 percen of GDP).

14 14 b = ( r gr)* d ( 1+ gr), where b is he primary surplus-gdp raio, r is he long-run real ineres rae, gr is he long-run real growh rae of GDP, and d is he deb-gdp raio. For example, assume ha r and gr are 0.06 and To sabilize he deb-gdp raio a he curren ne deb-gdp raio of 0.85, he governmen will have o run a primary fiscal surplus-gdp raio of almos 5 percen. Given he curren cyclically-adjused primary fiscal defici-gdp raio of abou 4 percen, o keep he deb- GDP raio a he curren level, he required increase in he primary balance would be 9 percen of GDP. This required adjusmen in he primary balance is somewha higher (because of our higher long-run real ineres rae assumpion) han, bu oherwise wihin he range (3-9 percen) repored in IMF (2000), and Jinno and Kaneko (2000). I would be very difficul for he governmen o achieve his adjusmen in he primary balance hrough fiscal reform in he near fuure. Thus, some analyss have argued ha he governmen may aemp o lower he real value of he deb hrough inflaion (Jinno and Kaneko, 2000; Miyao, 2001). Since Japanese governmen bonds pay a nominal coupon rae, inflaion will lower he real reurn on bonds, and he real ineres rae. From he equaion above, we can see ha he fall in he real ineres rae will lower he required adjusmen in he primary defici. In addiion o deb-susainabiliy, many analyss have focused on he disribuional effecs of governmen deb (Sakurai, 1998; Jinno and Kaneko, 2000; Miyao, 2001). The benefis of curren governmen spending largely fall on curren generaions, while he cos, in higher axes or inflaion, is borne by fuure generaions. Thus, deb financed governmen spending enails a

15 15 redisribuion of resources from fuure generaions o he curren generaion. These redisribuional effecs are paricularly high for social securiy and healhcare expendiures, where he benefis fall almos enirely on he elderly, and he coss are borne by he young. Governmen spending ha raises he fuure producive capaciy of he economy such as in educaion and physical infrasrucure will besow benefis, as well as coss, on he young, and he redisribuional effecs are smaller. Takayama and Kiamura (1999) idenified large inergeneraional imbalances in Japan, wih fuure generaions expeced o pay abou 3-4 imes more in ne axes and social securiy conribuions han he generaion currenly in reiremen. Of course, hese redisribuional effecs depend on Japanese households being non-ricardian. If he Japanese elderly raise heir bequess o compleely offse he coss o he young of he higher governmen deb, hen governmen deb has no redisribuive effecs. Recen Fiscal Reform Measures. To resrain increases in he deb-gdp raio, he governmen has proposed several fiscal reform measures in he 1990s. However, mos of he measures were posponed or abandoned, as he governmen sough o simulae demand, in ligh of he very weak domesic economy. Specifically, in 1997, he governmen enaced he Fiscal Srucural Reform Law. The goal of he 1997 Law was o eliminae fiscal deficis by The main insrumens in he 1997 Law were cus in governmen consumpion and invesmen, raher han ax increases. Public invesmen spending was o be cu by 7 percen in 1998, wih zero nominal growh unil 2001; and energy, educaion, and overseas developmen

16 16 assisance were o be cu by 10 percen in 1998, wih annual reducions unil 2001 (Ishi, 2000, p. 149). However, wih he severe recession of 1997, fiscal consolidaion was pu on hold, and a wide-range of pump-priming measures were inroduced. In paricular, raher han declining, public invesmen for 1998 was increased by over 10 percen. Areas where he 1997 Law made progress were in healhcare and social securiy reform, which are imporan, given he aging of he populaion. In 1997, he conribuion rae and copaymens by paiens for he governmen healh insurance schemes were increased sharply (Ishi, 2000). In paricular, paiens aged 70 and above are required o pay a fixed proporion (10 percen) of heir medical coss. The governmen also capped prescripion drug prices, which are very high in Japan. In 2000, a pension reform bill based on he 1997 Law passed he legislaure. The bill conained provisions o cu lifeime pension benefis by abou 20 percen. Specifically, pension benefis for new reirees were cu by 5 percen; he age of pension eligibiliy will be gradually (from 2013) raised from 60 o 65; and pension benefis will be subjec o an earnings es. Analyss have esimaed ha he 2000 pension reforms will reduce governmen unfunded social securiy liabiliies from he curren 60 percen of GDP o 30 percen of GDP (IMF, 2000). Looking forward, he governmen is planning on implemening furher budge cus, once he economy fully recovers. Recenly, a poliical commimen has been made o cap governmen defici bond issues a 30 rillion (0.6 percen) of GDP in Alhough defici bonds reflec only a porion of oal governmen borrowing, his bond issuance ceiling should help lower fuure fiscal deficis. As sipulaed in he 1997 Law, public invesmen is due for furher cus. Criicism has been direced a he economic value of he public works projecs, as well as conracing

17 17 procedures. To address he efficiency issues, new cos-benefi guidelines for review of public works projecs were announced. Conracing procedures have also been reformed. Public works projecs in 2002 are scheduled o be cu by 10 percen, alhough wheher he cus will acually maerialize is unclear. Moreover, he governmen inends o change he form of public works from he radiional ype of consrucion projecs o broader social infrasrucure invesmen; for environmen and energy-relaed projecs, elecommunicaions neworks, scienific research, nursing homes, and he like. Wih regards o healhcare, conribuion raes and co-paymens, especially by he elderly, are planned o be increased furher. The governmen s saed goal is o resric he growh of medical coss of he elderly o no more han he rae of inflaion. The age of eligibiliy for special elderly medical care will evenually be raised from 70 o 75. Furher cus are also planned in social securiy; for example, here are suggesions ha average benefis should furher be reduced by abou 40 percen, o avoid large increases in fuure conribuion raes (Sakurai, 1998). IV. Aging and Suppor Raios. The economic consequences of populaion aging depend on he naure of underlying demographic change as well as on he relaionship beween he resource needs of individuals of differen ages. Figure 3 plos he Japanese governmen s projecions of he counry s populaion and he percenage of he oal populaion ha is elderly. 8 Japan s populaion is expeced o peak a close o 130 million in 2005, hen gradually decline o abou 100 million by abou The 8 The figures for were calculaed from daa presened in Japan s Saisical Yearbook. The figures from were calculaed from he medium projecions of he populaion by age group presened in he Minisry of Healh and Welfare (1998).

18 18 percenage of he populaion over he age of 65 has grown rapidly, especially since 1980, and now sands a abou 15 per cen. By 2020, ha percenage should approach 25 percen, and by 2050, 33 per cen. By 2030, he percenage of he very old (aged over 80) should exceed 10 per cen. These raes of populaion aging are much higher han in oher counries. For example, in he U.S., only abou 15 per cen of he populaion will be above he age of 65 by Declining feriliy is he principal source of he changing demographic paers (Takayama, 1998). In he years following he Second World War, he oal feriliy rae in Japan rose o abou 4 by However, feriliy declined sharply during he 1970s and 1980s. I was 2.1 per household in 1974, bu 1.4 per household by The oal feriliy rae is projeced o decline o abou 1.2 over he nex several decades. Moreover, Japan has allowed almos no immigrans, who, especially in English-speaking counries, have helped o keep he populaion young. These rends have imporan implicaions for he demographic srucure of he populaion over he nex half-cenury. The Suppor Raio. Demographic shifs affec he economy s consumpion opporuniies because hey change he relaive sizes of he self-supporing and dependen populaions. Following Culer, Poerba, Sheiner, and Summers (1990), we summarize hese changes by he suppor raio, denoed by α, which we define as he effecive labor force, LF, divided by he number of consumers, CON, α = LF / CON. The firs issue in measuring he suppor raio concerns he relaive consumpion needs of people a differen ages. We assume ha all people have idenical resource needs so ha:

19 19 CON = 99 N i, i= 1 where N i is he number of people of age i. The second issue concerns he effecive labor force. The firs measure, LF1, assumes ha all people aged are in he labor force, while individuals 19 and under or 65 and over are no: LF1 = 64 N i i= 20. This measure is used by he Japanese governmen in projecing he fuure labor force. The second measure, LF2, recognizes ha boh human capial and labor force paricipaion raes vary by age. We use daa on he average 1990 earnings of people of each age (measured in 5-year inervals) and sex ( W ij, where i is age, and j=m, male or F, female) along wih daa on age- and sex-specific labor-force paricipaion raes ( PR ). 9 ij 80 LF2 = ( W * PR * N + W * PR * N ) i= 15 im im im iw iw iw. This measure assumes ha earnings accuraely reflec a worker s human capial. If age- 9 The daa on earnings and labor-force paricipaion raes are from he Minisry of Labor (various years).

20 20 earnings profiles are hump-shaped, hen labor produciviy peaks in middle-age. Thus, his measure recognizes ha human capial of a sociey wih a high fracion of people in middle age is higher han ha of a sociey wih many older workers, whose earnings and labor-force paricipaion raes decline. The wo suppor raios using he wo measures of he labor force are repored in Figure 4. The wo suppor raios have very similar paerns, especially afer Using LF1, he suppor raio declines from 1 in 1990 o 0.80 in Using LF2, i declines o Beween 2005 and 2030, he second suppor raio declines more han he firs, owing o he fall in high-earning, prime age males. Given he similariy in he wo suppor raios, for he remainder of his paper, we focus on he governmen measure, LF1. V. Demographic Change, he Opimal Saving-Invesmen Balance, and Governmen Deficis. Here we simulae he impac of demographic change on fuure Japanese saving and invesmen raes, and governmen deficis, using he governmen s measure of he suppor raio, LF1. We adop he neoclassical framework and assume ha consumers maximize (lifeime) uiliy. Households base heir consumpion on boh curren and fuure income. Thus, consumpion can be deached from curren income; households adjus heir saving o keep consumpion growh consan ino he fuure. In our simulaions, we adop he sandard small-counry, open capial markes, Ramsey opimal-growh model (Barro and Sali-i-Marin, 1995, Ch. 3). Specifically, we closely follow Culer, Poerba, Sheiner, and Summers (1990) modificaions o he Ramsey model, in examining

21 21 he impac of changing demographics on savings and governmen deficis. 10 Wih he model, we can examine how a sociey can adjus is saving, invesmen, and governmen deficis in response o changes in demographic variables. We simulae he model using plausible parameer values; and he projeced fuure pahs of he suppor raio, and he growh in he populaion and he labor force. (i) Skech of he Simulaion Resuls. As consumers seek o smooh consumpion over ime, consumpion per capia grows a a consan rae. However, as he suppor raio falls, GDP per capia grows a a slower rae han consumpion per capia, which raises he consumpion-gdp raio. Tha is, as he number of workers relaive o populaion falls, here are relaively less people susaining oupu (GDP), while consumers remain relaively numerous, raising he consumpion-gdp raio. The rise in he consumpion-gdp raio lowers he privae saving rae. The privae saving rae is projeced o decline from abou 28 percen oday o abou 15 percen by 2020, and abou 12 percen in To reduce disorions, he governmen seeks o mainain a per capia ax level ha grows 10 The Ramsey model assumes ha households are dynasic hey care abou heir children s and grandchildren s welfare (uiliy) as much as heir own. Of course, an imporan implicaion of dynasic households is ha Ricardian equivalence holds; governmen deb does no affec he inergeneraional disribuion of wealh. There is a large lieraure esing wheher he dynasic model is applicable for Japan(for a review, see Horioka, 2001). The dynasic model can be conrased wih he life-cycle model, in which households do no care abou heir children. Thus, in he life-cycle model, households bring down heir wealh (dissave) in old age. On he whole, he empirical ess suppor he dynasic model, and rejec he life-cycle model. The Japanese elderly, on average, leave large bequess o heir children, and his appears o be moivaed by alruism owards he nex generaion.

22 22 a he rae of per capia consumpion growh, implying a rising ax-gdp raio (given slower growh in GDP per capia). The ax-gdp raio is projeced o rise sharply from 28 percen oday o abou 45 percen in 2020, o reach almos 50 percen in Alhough aging raises social securiy and healhcare spending, increasing he governmen spending-gdp raio, he rise in he governmen spending-gdp raio is lower han he increase in he ax-gdp raio. Thus, he governmen saving rae gradually rises. The governmen saving rae rises from abou 2 percen of GDP oday o abou 10 percen in However, he decline in he privae saving rae is larger han he rise in he governmen saving rae, leading o a fall in he oal saving rae, from 30 percen oday o abou 24 percen in 2020, and abou 20 percen in The rising governmen saving rae evenually offses oday s ousanding governmen deb-gdp raio; fuure governmen spending, and public invesmen. Consequenly, afer iniially increasing, he governmen deb-gdp raio sars o decline in abou As he labor force declines (in absolue number), he need o equip workers wih capial equipmen decreases, and boh privae and public invesmen raes fall, resuling in a decline in oal invesmen. The privae invesmen rae falls from 20 percen oday o abou 16 percen in 2040; he public invesmen rae falls from 8 percen oday o abou 6 percen in The fall in oal saving is sharper han he fall in oal invesmen, resuling in a decline in he curren accoun-gdp raio from 2 percen of GDP oday o -1 percen of GDP in 2020, and evenually o -3 percen of GDP in Thus, afer iniially increasing, Japan s ne foreign asses-gdp raio sars o decline around 2020 and approaches 0 by (ii) Behavior of Firms.

23 We begin wih he producion funcion of a represenaive firm ha uses boh privae and public capial as inpus: 23 y = k$ γ γ λ m$ α 1 e h (1) where y is gross oupu per populaion (capia), k $ is he privae capial sock per effecive populaion, $m is he public capial sock per effecive populaion, and h is he consan rae of labor augmening echnical progress. We assume consan reurns o scale in privae and public capial, so ha ( 1 λ) = 2γ. In he above producion funcion, public capial is essenial for he produciviy of privae capial ie., public capial is no waseful. This goes agains convenional wisdom regarding he wasefulness of recen public invesmen in Japan. In our producion funcion, we are mosly concerned wih he produciviy of public capial over he long run (over decades), and public invesmen was cerainly producive in Japan in he pas (1960s and 70s), and may be producive again in he fuure. Noe ha when $k, $m and he suppor raios are consan, oupu per capia grows a a consan rae h. When he suppor raio is falling, however, oupu per capia grows a a slower rae han h. The supply of privae capial available o he firm depends on he global capial marke; he marginal produc of capial mus equal raes, and δ is he rae of depreciaion. We have: r + δ, where r is he gross inernaional real ineres 1 v 1 v k$ = ( r + δ)( avα λ ) m$, (2)

24 24 and hus privae invesmen per capia is: i$ = k$& + ( n + h+δ) k $. (3) where n is he populaion growh rae Thus, he pahs of privae capial and privae invesmen are solely deermined by he real ineres rae, he raes of growh of he labor force and populaion, echnical progress, and he pah of public capial. The governmen adjuss he level of public capial by changing he public invesmen rae, $ j : $ j = m & $ + m$ ( n + h+ δ) (4) (iii) Behavior of Consumers. The consumpion rae is deermined from forward-looking household behavior. Assume ha households wish o maximize heir lifeime uiliy, U, given by: U = 1 θ c n ee ( θ) 0 1 ρ d (5) where c is consumpion per capia, 1/θ is he ineremporal elasiciy of subsiuion, and ρ is he pure rae of ime preference. The budge consrain for households (in per-capia erms) is:

25 25 a& = α w + ( r n) a τ q 2 τ 2 (6) where a is oal asses per capia, which is comprised of privae capial, governmen bonds, and foreign asses, which are perfec subsiues in inernaional porfolios; w are wages; and τ is he lump-sum ax imposed on each person each period by he governmen. This lump-sum ax qτ 2 also imposes a deadweigh welfare loss of per person. 2 I can be shown (see Appendix) ha consumpion per capia always grows a h. Thus, while consumpion per capia grows a h, when he suppor raio is declining, oupu per capia ends o grow a less han h (see ii). The consumpion rae, c y is rising, lowering he privae saving rae. (iv) The Governmen Budge Consrain. Each period, he governmen issues governmen bonds of, &b o cover shorfalls in ax revenues: b& = ( r n) b τ + g + j (7)

26 26 where b is governmen bonds ousanding per capia. The increase in governmen bonds per capia is higher, he larger is he primary fiscal defici, which is he difference beween ax revenues per capia, and he sum of governmen consumpion g and public invesmen j per capia. As in Culer, e. al. (1990), we assume ha g is deermined by age-specific paerns of governmen consumpion. 11 Governmens spend differen amouns on people of differen ages. Spending on educaion benefi primarily children, while he elderly are he primary beneficiaries of healhcare and social securiy. Thus, even wihou changes in he srucure of governmen programs, demographic shifs can affec he level of g. We calculae per capia age-specific governmen spending paerns for Japan, focusing on he hree larges social expendiures: social securiy, healhcare, and educaion. For social securiy, we divide average social securiy expendiures in by he populaion over age 60. For healhcare, we allocae average healhcare spending in o differen ages, using he age-specific expendiure paerns repored in Ishi (2000). For educaion, we divide oal educaion spending in 2000 by he populaion beween ages 5 and 20. Demographic shifs can significanly aler governmen spending. Table 3 shows he projecions of oal governmen spending in 1995 yen and as a share of projeced GDP. We assume ha age-specific per capia expendiure paerns remain a he same real level beween 11 We also assume ha g eiher yields no uiliy o households, or ha governmen benefis do no affec he household s opimal choice of privae consumpion.

27 and Consisen wih curren Japanese governmen objecives (Ishi, 2000), we are no allowing any real increases in age-specific healhcare and social securiy spending. Tha is, if he average 67 year old receives 190 housand yen in governmen healhcare in 2000, an average 67 year receives he same inflaion adjused amoun in Oher governmen spending, mainly defense, policing, and adminisraion, are assumed o always equal he average raio o GDP of 5.6 percen. In our projecions, governmen spending rises from 25 percen of GDP in 2000 o 28 percen in 2015, and 33 percen in While educaion spending is projeced o decline, healhcare, and especially social securiy spending, are projeced o increase sharply, as he populaion ages. In paricular, in 2035, he populaion over 65 increases significanly (Figure 3), leading o sharp increases in social securiy and healhcare spending. I can be shown (see Appendix) ha he governmen will choose o levy a per capia lumpsum ax of ha grows a he rae of consumpion per capia growh, h. The governmen mus τ hen saisfy he following ineremporal budge consrain: τ 0 0 h e R d = b + ( g + j ) R d 0 0 (8) where b 0 is he governmen deb ousanding per person oday, and R is a discoun facor. This budge consrain says ha he presen value of ax revenues mus equal he presen value of 12 For social securiy, however, we assume ha he age of eligibiliy increases from 60 o 65 in 2015 (in accordance wih curren laws); alhough we assume ha per recipien benefis remain he same.

28 28 governmen consumpion plus public invesmen. If governmen ax revenues are insufficien o cover governmen spending oday, hen in he fuure, ax revenues mus exceed governmen spending for he governmen o saisfy is ineremporal budge consrain. (v) Projecions of Opimal Privae and Governmen Saving, Privae and Public Invesmen. As in Clarida (1993), we assume ha he governmen maximizes lifeime household uiliy (5), wih respec o c and $ j subjec o he consrains. We simulae he model using plausible parameer values, projeced fuure suppor raios (LF1), and fuure raes of populaion and labor force growh. In he simulaions, we allow suppor raios and raes of populaion and labor force growh o change every five years. Deails of he simulaion are given in he Appendix. For comparabiliy wih acual Naional Accouns Daa, we express our simulaions in erms of raios o GDP. We calibrae our model so ha he saring year (2000) corresponds o he average of he acual daa beween (he daa in Tables 1 and 2). For he iniial governmen deb-gdp raio, we use he raio of ne deb-gdp, inclusive of he social securiy ne asses (=45 percen of GDP). We accoun for ne fuure social securiy unfunded liabiliies by explicily incorporaing fuure social securiy benefis and conribuions ino our model. Of course, as alluded o earlier, here are oher unfunded and coningen liabiliies. Our saring year deb-gdp raio should be viewed as he lower bound. There is one excepion o his saring year calibraion exercise. Beween , he oal axes (including social securiy conribuions) colleced by he governmen averaged abou 27 percen of GDP. This ax rae was found o be simply oo low o be consisen wih our

29 29 model s ax smoohing and he saisfacion of he governmen budge consrain, (4). This is anoher way of saying ha unless he governmen sars running primary fiscal surpluses, governmen deb will no be susainable. Thus, we depar somewha from ax smoohing, and allow axes per capia o gradually increase from he year 2000 rae of 27 percen. 13 The governmen s ineremporal budge consrain is sill saisfied, which means ha fuure ax raes mus be higher han when axes are perfecly smoohed. Table 4 presens our projecions. Privae saving raes decline a few percenage poins unil 2010, and hen declines rapidly from 2010 o This paern is a resul of shifs in he suppor raio and increases in ax raes, which reduces disposable income. Alhough consumpion per capia always grows a a consan rae of h (=1.2 percen), as he suppor raio falls, oupu per capia growh is lower. Essenially, consumers are seeking o smooh heir consumpion when income is growing very slowly by lowering heir saving raes. Under ax smoohing, axes per capia increase a a consan rae, while oupu per capia grows a a slower rae; hus he ax-gdp raio rises over ime. However, he acual ax rae in he saring year (average, ) a 28 percen of GDP, is lower han wha is necessiaed by ax smoohing (33 percen) and he saisfacion of he governmen s ineremporal budge consrain. Tha is, unless curren ax raes are increased, he governmen will no be able o saisfy is ineremporal budge consrain. We allow axes per capia o increase more rapidly beween 2000 o 2015, and hen smooh increases in axes per capia from 2015 onwards. The sharp increases in ax raes beween 2000 and 2015 also conribues o he decline in privae saving raes, by 13 Culer, e. al. (1990) show ha deadweigh losses arising from deparures from ax smoohing are small.

30 30 lowering disposable income. By 2040, ax raes need o increase o almos 50 percen of GDP, for he governmen o recoup is curren ousanding ne deb (45 percen), projeced fuure spending (Table 3), and projeced public invesmen (Table 4). Governmen saving raes rise from abou 1-2 percen of GDP in 2000 o abou 10 percen in 2020, owing o he increased ax receips. Governmen saving raes decline somewha in 2035, because he spike in he over 65 populaion (Figure 3). Privae and public invesmen raes gradually fall over ime, as he need o equip workers wih capial equipmen declines. Because of high governmen saving and falling public invesmen, he fiscal surplus (governmen saving minus public invesmen) urns posiive afer Consequenly, he governmen ne deb-gdp raio increases unil 2020, and falls hereafer. The decline in he ne deb-gdp raio is fairly rapid beween 2020 and The decrease in privae saving is sharper han he increase in governmen saving, resuling in a fall in oal saving. The oal saving rae declines from abou 30 percen in 2000 o 24 percen in 2015, 21 percen in 2030, and 19 percen in Toal invesmen declines from 28 percen in 2000 o 25 percen in 2015, 23 percen in 2030, and 22 percen in Thus, he decline in oal saving is sharper han he decline in oal invesmen, leading o declining curren accoun surpluses. Japan s curren accoun surplus is projeced o become negaive in 2015, and negaive from hen onwards. Consequenly, Japan s ne foreign asses, afer peaking relaive o GDP in 2015, will sar o decline, and will approach 0 by (vi) Comparison wih Earlier Projecions of he Japanese Saving-Invesmen Balance and he Governmen Budge.

31 31 Many papers have projeced he impac of demographic change on he Japanese savinginvesmen balance, alhough papers projecing fuure Japanese governmen budge balances are fewer. Despie he variey of mehodologies and modeling assumpions, mos earlier papers like our paper--projec declining saving and invesmen raes, wih saving raes declining faser han invesmen raes, leading o curren accoun deficis beween 2025 and On he whole, he earlier papers predic deerioraing governmen budge balances, unless here is drasic fiscal reform. The proposed fiscal reforms in he earlier papers range from ax increases o cus in social securiy benefis and in public invesmen. In our paper, we have focused on ax increases. Mos earlier papers projecing fuure saving and invesmen raes are based on ad hoc economeric specificaions. Horioka (1991, 1992), using reduced-form ime-series economerics, esimaes ha Japanese privae saving will become -15 per cen by Oishi and Yashiro (1997) develop a small (7 equaions) simulaneous equaion economeric model. They projec ha in 2025, he gross saving rae and he gross invesmen rae will reach 6 percen and 14 percen of GDP, respecively, leading o a ne expor defici-gdp raio of 8 percen. The Economic Planning Agency s (1997) large (270 equaion) simulaneous equaion model of he Japanese economy projecs ha by 2025, he gross saving and gross invesmen raes will be 23 percen and 25 percen, respecively. Their model projecs ha by 2050, he gross saving and gross invesmen raes will be 15 percen and 20 percen, respecively. Also using a large-scale macroeconomeric model, he Japan Cener of Economic Research (2000) esimaes ha in 2025, he privae saving rae will be 3.7 percen. Recenly, researchers have made projecions based on explici uiliy maximizing frameworks. Auerbach e. al. s (1989) overlapping-generaions model--based on he life-cycle

32 32 hypohesis where agens reduce heir wealh in old age--finds ha Japan s saving-invesmen balance will narrow unil 2030, and become negaive in Miles and Cerney s (2001) apply Auerbach e. al. s model o he closed economy, and projec ha Japan s gross saving rae will be 18 percen by 2020, and 14 percen by McKibbin and Nguyen (2001) develop a mulicounry model o simulae he economic effecs of Japan s aging populaion. Since in heir model, Japan is no longer small, changes in Japan s saving and invesmen behavior can impac inernaional real ineres raes, unlike in our paper. In addiion, he auhors adop he Blanchard (1985) saving funcion, where agens build up heir financial wealh in heir young years o mainain a arge level of consumpion in heir elderly years. McKibbin and Nguyen projec ha by 2040, he gross saving and he gross invesmen raes will be 20 and 22 percen, respecively, leading o a curren accoun defici-gdp raio of -2 percen. Wih regards o he effec of aging on Japanese governmen budge balances, mos earlier research porray falling ax revenues and rising pension benefis. Masson and Tryon (1990) find ha beween 2000 and 2025, he budge defici-gdp raio will deeriorae by 2.5 percenage poins, owing o increasing pension burdens. The Japan Cener of Economic Research (2000) esimaes ha he governmen saving-gdp raio will decline from 2 percen in 2000 o -2.4 percen by The Economic Planning Agency s (1997) simulaneous equaion model predics a -0.5 percen governmen saving rae in 2025, and a 0.3 percen governmen saving rae in To achieve his budge urnaround, ax raes (oal axes+social securiy conribuions)/gdp) are projeced o gradually increase from he curren 28 percen o 35 percen in 2025 and remain a ha level hereafer. Muhleisen (2000) uses a simulaneous economeric equaion model and projecs seadily improving governmen finances. The fiscal balance (governmen saving minus