Japan's High Saving Rate in the 1960s through the 1980s

Transcription

1 Japan's High Saving Rae in he 960s hrough he 980s ABSTRACT An endogenous growh model wih boh he beques moive and he precauionary moive for saving is calibraed o he Japanese and he U.S. economy in order o quanify he conribuion of various facors ha accoun for huge differences in he saving and growh raes beween he wo counries during he Japan's high-growh period of he 960's hrough he 980's. Among ohers, saving incenives in Japanese ax sysem are found o be an imporan facor o Japan's high saving rae and fas growh rae. JEL Classificaion Number: E2, O40 Key words: Saving rae, Saving moive, Economic growh, Japan Deparmen of Economics, Sae Universiy of New York a Buffalo, 45 Fronczak Hall, Buffalo, NY Tel: (76) Fax: (76) jinkim@buffalo.edu.

2 I. Inroducion For several decades since he 960's, Japan's saving rae had been one of he highes among developed economies. Japan's average saving rae over he period was more han wice as high as ha of he Unied Saes (see Table ). ) Since he rade surplus equals he excess of savings over invesmen by he naional income accouns and he U.S. economy has had a large rade defici wih Japan for decades, a lo of aenion has been paid by economiss and policy makers o he difference in he saving rae beween he wo counries. Moreover, he period of high saving rae coincides wih a successful period of high income growh in Japan. Japan's huge savings are considered o provide funds o finance invesmen, which lead o fas accumulaion of capial and fas growh of produciviy. Possible explanaions for Japan's high saving rae in his period have been inensively discussed in he lieraure, and among he explanaions is he disincive ax sysem in Japan. The Japanese ax sysem promoes savings since mos incomes from ineres, dividend and beques are very lighly axed. Above all, axes on ineres and dividend income have wo imporan feaures in he Japanese ax sysem. Firs, hey can be axed separaely or aggregaely wih oher incomes a he axpayer's discreion, and high-income people can have ineres and dividend incomes axed separaely a a rae lower han ha on oher income. Second, some ypes of capial income are ax exemp, which is called Maruyu sysem. 2) Furhermore, here

3 is widespread evidence of abuse of his sysem such as opening muliple accouns using ficiious names. Capial gains from socks are no axed if hey are less han a cerain amoun and if he number of ransacions is no large. Beques ax in Japan is levied on beneficiaries or donee of beques while he U.S. esae ax is levied on donors. The burden of beques ax in Japan can be reduced since beques per person is small if he number of heirs is large. Moreover, land or real esae is appraised for ax purposes a a value lower han he marke value in Japan. This incenive encourages Japanese people o inves heavily in real esae, and in fac, hree quarers of axable beques in Japan is in real esae while only one quarer is in real esae in he Unied Saes. Several demographic characerisics in Japan are also considered o conribue o Japan's high saving rae. Firs, he Japanese have he longes life expecancy in he world. A long life span resuls in a long reiremen period for which he Japanese need more savings. Second, he populaion proporion of he aged was sill small in Japan before 990's, which implies here were relaively few old dissavers. Anoher explanaion offered in he lieraure is a culural facor; srong inergeneraional alruism among Japanese families. Horioka (984) shows from he aiudinal surveys ha saving for reiremen is no imporan for Japanese people as a saving moive. Using household survey daa, Hayashi (986) and Hayashi e al. (988) find ha boh he elderly who live independenly and hose who live wih heir grown children coninue o save and ha here are signs of significan wealh ransfer beween generaions. 3) The exisence of subsanial inergeneraional ransfers, however, can be he resul of several facors. Parens leave accidenal bequess o heir children if here is no perfec annuiy marke and hey die premaurely

4 (precauionary moive). 4) Second, selfish parens can use bequess as paymens for heir children's service o care for hem. 5) Finally, parens give heir wealh o children because hey are alruisic oward heir children as in Barro (974) and Becker (974) (beques moive). The firs wo cases can be well inegraed ino he life-cycle hypohesis as he exended life-cycle models where he implicaion of he life-cycle hypohesis would sill hold. The implicaion for he saving rae depends criically on wheher or no inergeneraional alruism exiss. While i is conroversial on which model is a beer fi for he Unied Saes, 6) he alruism model is more appropriae han he life-cycle model for he Japanese sociey. Conrary o he finding in Hurd (987) for he Unied Saes, Hayashi, Ando and Ferris (988) show ha old singles decumulae asses while he elderly wih children do no, which implies ha he beques moive is an imporan moive for saving. Hayashi (986) presens evidence ha suppors inergeneraional alruism: savings of he elderly have increased since he large expansion of Japan's social securiy in 973, which is inconsisen wih he predicion of he life-cycle model ha an unanicipaed increase in annuiies will be consumed. He also uses he Euler equaion

5 mehod o show ha his expansion of social securiy sysem does no have a posiive effec on he consumpion growh of older cohors. Dekle (990) suppors he applicabiliy of he alruism model o Japan by showing ha he wealh holding of independen elderly is significanly higher when hey have children and is posiively associaed wih he number of children. The objecive of his paper is o quaniaively analyze Japan's high saving rae over he period from he 960's hrough he 980's. The approach will involve finding ou how much of he saving rae difference beween he Unied Saes and Japan can be accouned for by each aforemenioned explanaion. Calibraing an endogenous growh model wih inergeneraional alruism and precauionary saving in addiion o oher facors such as ax incenives and populaion growh o boh Japanese and he U.S. economy, we use simulaions o find ou wha he saving and growh rae in Japan would be if he model's parameers ake he values of he U.S. economy. The res of he paper proceeds as follows. Secion II presens he heoreical framework and is seady sae equilibrium. Secion III shows he calibraion mehod and he parameer values esimaed for each counry. The simulaion resuls are presened in Secion IV, and Secion V concludes. II. Theoreical Framework The economy is populaed wih many dynasic families, and each family consiss of wo overlapping generaions; parens and children. Each agen lives he firs period for cerainy and survives o he second period wih he probabiliy ϕ. Parens care no only for heir own welfare bu also for heir children's. The uiliy of parens herefore depends on he welfare of all fuure descendans. In he firs period of life, each agen consumes and saves ou of oal income ha is he sum of labor income, gif (or beques) income and governmen ransfers. The budge consrain in he firs period is hus ( ) ( ) c + s = τn wl + τ g g + T, ()

6 where c is consumpion in he firs period, s is saving, w is he wage rae, l is exogenous labor supply, g is gif income, and T is governmen ransfers in he firs period. τ n and τ g are he ax raes on labor income and gif income, respecively. If alive in he second period, he agen consumes and leaves beques. Oherwise, all of he agen's income is bequeahed o children. Therefore, he second period budge consrain is c + q = R s + T (2) + ( ) g q n + = + + when parens survive in he second period, (3) = R s T oherwise. where c q T + + is consumpion in he second period, + is bequess, is governmen ransfers in he second period, and + g + is gifs o children. R + is he afer-ax real ineres rae. The populaion of a family grows a he consan rae (+n), and beques from parens, wheher inenional or accidenal, is assumed o be divided up evenly for all children as shown in (3). Since family members are alruisically linked, he head of a family akes ino accoun he oal presen value of family resources. Le W be he oal family wealh in period wih is law of moion as follows: W W + ( τ g) ( a) R + ( τ g) c ( a) ( n) c ( a) = + + (4) where a in a parenhesis represens he case when parens are alive in he second period. W ( g) ( d) W + = ( + nc ) ( d) τ R (5) + where d in a parenhesis represens he case when parens are dead in

7 he second period. If we define he uiliy of he family's head as he value funcion, () V, i should be a funcion of he oal family wealh, which is he only sae variable of he model: ( ) VW = max ϕ[ β uc ( ( a) ) + λ( + n) uc ( ( a) ) + λ( + n) VW ( ( a) ) ] + + ( ϕ) [ λ( + n) u( c ( d) ) + λ( + n) V( W + ( d) ) ] (6) where he firs and second erms in brackes are he expeced uiliy when parens are alive and dead, respecively. β is ime preference and λ is he alruism parameer which discouns children's uiliy. Since parens rea all heir children equally, he alruism parameer is muliplied by (+n). If he momenary uiliy funcion () u akes he consan relaive risk aversion form, we can show ha he value funcion akes he same form. σ Suppose VW ( ) = kw where k is a consan 7) and σ is he inverse of he ineremporal elasiciy of subsiuion. From he maximizaion problem in (6) wih he budge consrains (4) and (5), we can derive he consumpions as he funcions of family wealh W and k: ( ), ( ) and ( ) c a = JW c a = J W c d = JW where 2 3 σ [( )( ) ] ( ) ( σ ) σ ( σ ) σ β J2 = + n σ k τ g R + ( τ g) + ( + n) λ( τ g ), β J = J2 λ and ( τ g ) σ, σ [( )( ) ] ( ) ( σ ) σ ( σ ) σ σ τ ( ) J = + n k R + + n. Plugging hese ino (6), we ge one equaion from which k can be solved as a funcion of he ineres rae R and he parameers in he model: 3 g σ

8 βϕ ( ) σ ϕλ + n k = J + J σ σ σ σ ( nk ) ( ) R [ ( ) J ( nj ) g g 2] + ϕλ + τ τ + ( ϕλ ) ( ) σ 2 σ n + + σ J σ + ( ) ( + n ) k ( ) R σ ϕλ τ ( n ) 3 g + J σ [ ] 3 σ (7) Since he expeced family wealh in he nex period is he probabiliy-weighed sum of family wealh, we have ( ) ( ϕ) ( ) EW = ϕw a + W d ( ) [ ( ) ( ) ] ( )( ) ( ) [ ] = ϕ τ R τ J + n J W + ϕ τ R + n J W (8) g g 2 g 3 and in a seady sae of he model populaed wih a large number of families, he growh rae of family wealh EW + W should be equal o he labor income growh rae. Hence, EW W + ( n) = + g where g is he growh rae of per capia labor income. We assume ha here are wo firms in he economy: firm produces consumpion goods and firm 2 produces invesmen goods. The relaive price of invesmen goods o consumpion goods is p k. Le V be he presen value a = 0 of firm, which is he discouned sum of fuure ne cash flows. 8) Firm has he Cobb-Douglas echnology wih inpus of capial K and labor L. (9) V = AK L w L p K K + ρ = 0 s= o s α α {( τ p)( ) ( Ωc Ωd) k[ + ( δ) ] } ρs = ( τ p)( rb s + π) where µ s + µ s p es + ds + + π p + µ es s π. The ne cash flow is equal o he gross profi afer ax, less he cos of,

9 he addiional capial purchased. τ p, Ωc and Ωd are he corporae ax rae, he invesmen ax credi and ax saving from depreciaion allowances. ρ s is he discoun facor or he cos of capial for firm ha has wo alernaive mehods of financing, borrowing or issuing equiies. ( rb + π) and ( d + pe pe + π) are he nominal reurns of bonds and equiies issued by firm where π is he inflaion rae, d denoes dividend yield, and pe pe denoes capial gain. The deb-equiy raio µ is assumed exogenous for he firm. From he maximizaion problem, we have p p α α k ( τ p) αak L = ( Ωc Ωd) pk ( + ρ ) ( δ) ( α ) α k (0) α AK L = w () I is assumed ha firm 2 has he linear echnology wih capial inpu only, which is an assumpion required for susained income growh in he economy. Le V 2 be he presen value of firm 2. V2 = p BK p K K + ρ = 0 s= o 2s ρ2s = ( τ p)( rb2s + π) where {( τ p) k 2 ( Ωc Ωd) k[ 2+ ( δ) 2] } µ 2s + µ 2s And he firs-order opimizaion condiion is p e2s + d2s + + π p + µ e2s 2s π., p p k ( τ ) ( ) p pkb = Ωc Ωd pk ( + ρ2 ) ( δ) k (2) Wih he assumpion ha all four kinds of asses in he economy (bonds and equiies of firm and 2) exis in posiive amouns, heir afer-ax raes of reurn should be equalized. 9) If we assume in addiion ha he dividend raios for boh firms are he same, we have r = r r, and b b2 b

10 p p = p p p p, and e e e2 e2 e e ( τ )( r + π) = ( τ ) d + ( τ )( p p + π) i b d c e e (3) where τ i, τ d and τ care he ax raes on ineres income, dividend income and capial gains, respecively. 0) Assuming ha he deb-equiy raios of firm and 2 are he same, he coss of capial in boh firms are equalized. From equaions (0), (), (2) and he definiion of he cos of capial, we have τ p Ω Ω c d B K K α α α α L L + ( ) K δ K L L µ pe =+ ( -τ p )( rb + π) + d + + π π + µ p + (4) µ e In he following analysis, d, p p and µ are assumed consan over ime, which guaranees wih equaion (3) ha he real ineres rae r b is consan over ime. Since he righ-hand side of (4) is consan and L e grows a he same rae as he family populaion, K grows a a consan rae in he equilibrium. In a seady sae of he model when K ( = K + K2 ) grows a a consan rae and he raio K K2 is consan, equaion (4) becomes τ p Ω Ω c d e ( α ) α µ pe B+ ( δ) g = + ( τ )( + π) π + µ + µ π p rb d p (5) where g is he growh rae of per capia income. Equaions (7),(9) and (5) form a se of non-linear simulaneous equaions. Since no closed-form soluions exis, he equilibrium values of k, R ( i)( rb ) e = + τ + π π, and

11 g should be derived hrough simulaions. III. Parameerizaion of he Model We calibrae he model's basic parameers using he acual Japanese and he U.S. daa and some consensus esimaes in he lieraure. Table 2 repors he benchmark values of he parameers for each counry. Mos Japanese ax raes and wo ax incenives are drawn from Shoven and Tachibanaki (988). These ax raes are marginal raes because he marginal rae is a relevan measure for he incenive effecs of axaion. The U.S. ax raes are marginal raes drawn from King and Fulleron (984). However, he gif (beques)income ax raes of boh counries are he average raes calculaed by he auhor based on he daa in Barhold and Io (992). The comparison of he U.S. and he Japanese ax sysem reveals ha saving incenives in he Japanese ax sysem are sronger because personal income from capial and inergeneraional ransfers are very lighly axed. ) Japanese firms have a considerable advanage from deb-financed invesmen since Japanese households can have ax exempion on mos of heir ineres income, and firms can deduc ineres expenses. This advanage is eviden in Table 2 as he average deb-o-equiy raio of Japanese firms is higher han ha of he U.S. firms. Dividend yield, which is he raio of dividend income o equiy value, is higher in Japan as well. Higher deb-o-equiy raio and dividend yield resul from several feaures of Japanese corporae and personal income axes which alleviae he problem of double axaion on dividends. For example, he corporae income ax rae is lower for earnings paid ou as dividends and dividends income can be axed separaely from oher income a lower raes.

12 The depreciaion raes for boh counries are calculaed from Shoven e al. (988) and King e al. (984) as he weighed average wih he weigh of he capial sock share. The inflaion and populaion growh raes are calculaed from Maddison (99). The consumer price indices are used for calculaing inflaion raes. The survival probabiliies are derived from he age disribuion of populaion in boh counries. The survival probabiliy for he age group in year is calculaed as he raio of he number of he age group in year +30 o he number of he age group in year. The repored numbers in Table 2 are he average of he survival probabiliies for he age groups from year 950 o 960. The benchmark value for σ, he inverse of ineremporal elasiciy of subsiuion, is 2. I is assumed ha one period of he model is 30 years, and he firs period sars a age 25, which implies some die a age 55 and ohers survive o age 85. The oher four parameers, α, B, λ and β, are solved from he equilibrium seady sae condiions of our model, using he Japanese and he U.S. daa. The echnology parameers, α and B, are derived from he following equaions, using he informaion on he gross naional saving rae and he growh raes of per capia income and capial; α = log log( g) ( g ) log( + n) k [ ] [( ) ] ( ) α α α ( δ) B = + SAV + n g where g is he per capia income growh rae, and g k is he capial growh rae, and SAV is he gross naional saving rae. If we add he following age-consumpion raio o equaions (7), (9), and (5), we have four equaions o solve for k, r b, λ, and β. ϕj J + ( ϕ) J 2 3 = CONS c c old young where c old and c young are he per capia consumpions of old and young

13 generaion. Table 2 presens he values of economic indices ha are used o calibrae hese four parameers. The gross naional saving raes in Japan and he Unied Saes are from Boskin and Robers (988). These raes are measured in he OECD mehod which includes governmen nonmiliary invesmen. 2) The per capia GDP growh raes are calculaed from Summers and Heson (988). The esimaes of capial sock in boh counries are drawn from Maddison (99). They include all non-residenial srucures, machinery, equipmen and vehicles. They exclude residenial srucures (housing), non-reproducible iems (naural resources, land), foreign asses, invenories, and inangibles (human capial). The age-consumpion raios in boh counries are derived from Hayashi e al. (988) which repor he age-consumpion profile from wo ses of cross-secion household daa, he 984 "Naional Survey of Family Income and Expendiure" for Japan and he 983 "Survey of Consumer Finance" for he Unied Saes. 3) The las four rows in Table 2 repor he calibraed values of he four parameers. Japan appears less capial-inensive, bu more producive han he Unied Saes. Japanese people seem o have a sronger alruisic linkage beween generaions, and o be more paien. IV. Simulaion Resuls The simulaion resuls for he benchmark case when σ = 2 are presened in Table 3. In an aemp o measure he conribuion of he Japanese ax sysem o higher saving rae and faser growh, we firs

14 calculae he income growh rae, he saving rae and he ineres rae for Japan when Japan's ax parameers ake he values of he Unied Saes wih oher parameers inac a he values for Japan. The resuls in Table 3 illusrae a favorable saving incenive of he Japanese ax sysem. I shows ha he saving rae in Japan would drop by 9 percenage poins (from 34 o 25 %) and he annual growh rae would fall by 0.5 percenage poins (from 5.3 o 4.8 %) if Japan adops he U.S. ax sysem. This implies ha he unique ax sysem in Japan alone can explain 5 % of he growh rae difference and 58 % of he saving rae difference beween wo counries. Among he individual axes, higher ax rae on personal income in he Unied Saes has he mos deerren effec on growh and savings. 4) The second experimen considers demographic facors. In he life-cycle model, an increase in he populaion growh rae implies relaively more young savers and less old dissavers in he economy, which increases he saving rae. However, an increase in n may lead o a decline in per capia savings, and herefore per capia income growh because i can lower he real wage rae. The effec of populaion growh in he alruism model criically depends upon he funcional form of alruism. In he previous secion, we assumed γ equals one when he generaional discoun facor is γ λ( + n ). If we ake his assumpion in he pure alruism model, he change in n has no impac on growh and savings. Therefore, he effec of n in his simulaion is induced by he life-cycle par of he model. Since he difference in he populaion growh raes is small, an increase in n o he U.S. parameer value does no change he saving and growh raes much in our simulaion. When he survival probabiliy ϕ decreases o he U.S. level, he growh and he saving raes increase in our model. In fac, i can be shown ha hey are maximized in he neighborhood of 0.5 because

15 uncerainy, and herefore he precauionary saving, disappears a he value of zero or one. The magniude of changes in he saving or growh rae, however, seems o be very small when ϕ is changed. The sronger is he alruism oward fuure generaions, he more he presen generaion saves and he faser he economy grows. If Japanese families had he same level of inergeneraional alruism as he U.S. families (i.e. he same value of λ), he annual growh rae of per capia income in Japan would fall by. percenage poins, and he saving rae would also fall by 6 percenage poins. Compared o oher experimens, he difference in he alruism parameer appears o explain a fairly big porion of he saving and growh rae differences beween he wo counries33 % and 00 %, respecively. Anoher culural facor considered in our analysis is he difference in paience. Even hough he subjecive ime preference parameer β differs much beween wo counries, changes in he growh and he saving raes are very small when β is changed. In he simple alruism model wihou uncerainy, we can show ha a change in β srongly affecs he consumpion disribuion beween generaions, bu i has no effec on growh or savings. The channel hrough which β affecs savings and growh is he precauionary saving in his model. The res of Table 3 repors he effecs of changes in oher parameers in he model, including producion funcion parameers, he inflaion rae, and he depreciaion rae. Differences in echnology parameers, B and α, have opposing effecs on he growh and he saving rae. While a larger value of α for he U.S. increases he growh rae and decreases he saving rae, he decline of produciviy decrease of B reduces he growh rae and increases he saving rae. 5) In his calibraion, he laer effec dominaes he former, and he ne effec of echnology difference is a decrease in growh (0.7 percenage poins) and an increase in savings (7 percenage poins). The direcion and magniude of he effec of a change in he inflaion

16 rae is dependen on he size of oher parameer values, and especially, various ax raes because mos axes are levied on nominal values. In our simulaion, a decline in π o he U.S. level reduces he growh rae and he saving rae in a small magniude. An increase in he depreciaion rae o he U.S. level slows down capial accumulaion and herefore income growh while i increases he invesmen raio or he saving rae. However, he new growh and saving rae for Japan wih he U.S. depreciaion rae are no much differen from he acual values. The magniude of changes in he growh rae and he saving rae from differen parameer values should vary wih he value of he ineremporal elasiciy of subsiuion, which is he only free parameer in our model. Table 4 repors he sensiiviy analysis resul for differen values of he ineremporal elasiciy of subsiuion. For mos parameers of he model, he simulaion resul is no so sensiive o he value of σ. For insance, ax sysem difference can explain 0 o 9 % of he growh rae difference and 39 o 68 % of he saving rae difference beween Japan and he U.S. for he range of σ from (log uiliy) o 4. The excepion is he alruism parameer λ even hough is explanaory power for he saving and growh difference beween wo counries is fairly big for σ values in his range. In recen years he role of inergeneraional ransfers in savings and wealh accumulaion has been he subjec of numerous sudies. Numerous evidences suggess ha inergeneraional ransfers play an imporan role, if no he mos dominan one, in wealh accumulaion in such developed counries as Japan and he Unied Saes (Hayashi, 986; Kolikoff and Summers, 98). Two simulaion experimens are carried ou in Table 5 in an aemp o find ou wha porion of Japan's saving and income growh rae can be aribued o inergeneraional ransfers in our model. In he firs experimen, we derive he saving rae and he growh rae of Japan in a hypoheical siuaion where Japanese people have no inergeneraional alruism, and hence no inenional beques. The same values for oher parameers in he previous secion are used excep in his calculaion. See appendix for he mahemaical seups of his and he following model. I appears ha 70 percen of savings and 40 percen of income growh in

17 Japan are due o inenional wealh ransfer beween generaions, and ha he magniude of his effec does no vary much wih differen values of σ. The second experimen deals wih he addiional reducion in he saving rae and he growh rae when we assume furher ha here is a perfec annuiy marke in addiion o no alruism. Therefore, in his experimen, we have neiher inenional beques nor uninenional beques, and he sole moive for saving is for old age. The saving rae and he growh rae in he life-cycle model wihou any inenional or uninenional beques are repored in Table 5 o be only 7 % and 39 % of he acual raes in he benchmark case. V. Concluding Remarks Japan's high saving rae from he 960's hrough he 980's has drawn aenion from economiss and policy makers for many years. Among possible facors brough up by life-cycle models, Japan's income ax sysem is considered o provide srong saving incenives because personal incomes from capial and wealh ransfers are lighly axed. Demographic facors like he growh and age disribuion of Japanese populaion and srong inergeneraional alruism among Japanese families have also been claimed o be imporan for Japan's high saving rae. This paper quanifies he explanaory power of hese facors. For ha purpose, we calibrae an endogenous growh model ha incorporaes hese facors o he Japanese and he U.S. economy, and use he simulaion mehod o accoun for he conribuion of each facor in Japan's high saving and growh raes. Among various facors, Japan's srong inergeneraional alruism and unique ax sysem which is favorable o savers are shown o be imporan in explaining he differences in he growh rae and he saving rae beween wo counries. We also show ha inergeneraional wealh ransfer in Japan is an imporan facor as well. Our heoreical model does no incorporae several popular explanaions for Japan's high saving rae. One is he underdeveloped social securiy sysem in Japan which is said o have increased he need o save more for

18 life during reiremen. However, Japan's social securiy has expanded dramaically since 973 and is now comparable o ha of oher developed counries. Furhermore, Hayashi (986) shows ha savings of he elderly have no decreased bu increased slighly since he large expansion of Japan's social securiy. Anoher explanaion offered in he lieraure is he bonus sysem in Japan. Japanese workers receive large lump-sum paymens wice a year. This bonus sysem was sared in he large firms afer World War II and has been spread ou o he whole economy hereafer. This explanaion is based on he idea ha he bonus income is a ransiory income and hus largely saved according o he permanen income hypohesis. However, in recen years, he bonus income has been insiuionalized and he dae and amoun of bonus is no longer unpredicable. The bonus hypohesis enjoyed populariy in 980's wih he observaion of he close co-movemen of bonus-income raio and he personal saving rae since 973. This observaion, however, can be explained easily. Bonuses are ransfers of corporae savings o households, and if hey can see hrough he corporae veil, he personal saving rae should rise wih bonus raio. Therefore, he bonus sysem can no be a good explanaion for high naional saving rae. The reiremen age for Japanese workers in he privae secor is radiionally 55 which is quie early by inernaional sandards. If people reire early, hey do need more savings for he old age afer reiremen. Alhough Japanese workers formally reire early, hey, however, ypically coninue working unil very lae in life. In fac, he labor force paricipaion rae of he aged in Japan is one of he highes among he developed counries. Many opinion surveys consisenly find ha housing purchase is a major saving moive for he Japanese. The land and housing prices are very high in Japan and a down paymen for a house coss up o 40 % of he purchase price of a house. Alhough his facor is likely o conribue o high saving rae for he young generaion, evidence does no suppor his explanaion. Hayashi (986), for insance, showed ha in urban areas where he price of housing is much higher, he saving rae is lower han ha in rural areas. In

19 sum, alhough hese four aforemenioned facors enjoy some currency in he lieraure, heir explanaory power is apparenly limied and our analysis ignored hese facors accordingly. Since he Shoup recommendaions in 949, he Japanese ax sysem has been revised occasionally bu has failed o keep up wih he changes in Japanese sociey and economy. Japanese people in homes and businesses feel more heavily and unfairly axed and preferenial ax reamen of personal savings has given rise o especially sharp complains. The ongoing ax reform is hence aimed o revise unfair axaion on capial income. No longer has his new ax sysem he srong saving incenives, and his may parially explain why Japanese saving and growh raes have declined in recen years.

20 Appendix The following is he maximizaion problem for a family which is no alruisically linked. The maximizaion problems for firms are he same as in secion II. ( ) + β ϕ ( ) max uc uc s.. where + ( τ ) ( τ ) c + s = w l+ g + T c = R s + T ( + n) g = R s + T + + n g when alive in he 2nd period when dead in he 2nd period If we assume ha here is a perfec annuiy marke, people do no leave any uninenional beques o heir children. People who are no alruisic and have a perfec annuiy marke solve he following problem ( ) + β ϕ ( + ) ( τ ) max uc uc s.. c + s = w l+ T n c = R s + T ϕ + + +

21 References Abel, Andrew B "Precauionary savings and accidenal bequess," American Economic Review, 75, Abel, Andrew B., and Laurence J. Kolikoff "Does he Consumpion of Differen Age Groups Move Togeher? A New Nonparameric Tes of Inergeneraional Alruism," NBER Working Paper 2490, Cambridge, MA. Alonji, Joseph G., Fumio Hayashi, and Laurence J. Kolikoff "Is he Exended Family Alruisically Linked? Direc Tess Using Micro Daa," American Economic Review, 82, Barro, Rober J "Are Governmen Bonds Ne Wealh?" Journal of Poliical Economy, 82, Barhold, Thomas A., and Takaoshi Io "Beques Taxes and Accumulaion of Household Wealh: U.S.-Japan Comparison," In The Poliical Economy of Tax Reform, edied by Takaoshi Io and Anne O. Krueger, The Universiy of Chicago Press, Chicago, IL. Becker, Gary S "A Theory of Social Ineracions," Journal of Poliical Economy, 82, Blinder, Alan S., Roger Gordon, and Donald E. Wise. 98. "Social Securiy, Bequess, and he Life Cycle Theory of Saving: Cross-Secional Tess,"NBER Working Paper 69, Cambridge, MA. Boskin, Michael J., and John M. Robers "A closer look a saving raes in he Unied Saes and Japan," In Governmen Policy owards Indusry in he Unied Saes and Japan, edied by John B. Shoven, Cambridge Universiy Press, Cambridge, England. Brock, William A., and Sephen J. Turnnovsky. 98. "The Analysis of Macroeconomic Policies in Perfec Foresigh Equilibrium," Inernaional Economic Review, 22, Dekle, Rober "Do he Japanese Elderly Reduce Their Toal Wealh? A New Look wih Differen Daa," Journal of he Japanese and Inernaional Economies, 4, Hayashi, Fumio "Why Is Japan's Saving Rae So Apparenly High?" NBER Macroeconomics Annual,, Hayashi, Fumio, Alber Ando and Richard Ferris "Life Cycle and Beques Savings: A Sudy of Japanese and U.S. Households Based on Daa from he 984 NSFIE and he 983 Survey of Consumer Finances," Journal of he Japanese and Inernaional Economies, 2, Horioka, Charles Yuji "The Applicabiliy of he Life-Cycle Hypohesis of Saving o Japan," The Kyoo Universiy Economic Review, 54, Hubbard, R. Glenn "Uncerain Lifeimes, Pensions, and Individual Saving," In Issues in Pension Economics, edied by Zvi Bodie e al., Universiy of Chicago Press, Chicago, IL.

22 Hurd, Michael D "Savings of he Elderly and Desired Bequess," American Economic Review, 77, King, Mervyn A., and Don Fulleron The Taxaion of Income from Capial: A Comparaive Sudy of he Unied Saes, he Unied Kingdom, Sweden, and Wes Germany, Universiy of Chicago Press, Chicago, IL. Kolikoff, Laurence J., and Lawrence H. Summers. 98. "The role of inergeneraional ransfers in aggregae capial formaion," Journal of Poliical Economy, 89, Kolikoff, Laurence J., John Shoven, and Avia Spivak "The Effec of Annuiy Insurance on Savings and Inequaliy," Journal of Labor Economics, 4, S83-S207. Kolikoff, Laurence J., John Shoven, and Avia Spivak "Annuiy Markes, Savings, and he Capial Sock," In Issues in Pension Economics, edied by Zvi Bodie e al., Universiy of Chicago Press, Chicago, IL. Maddison, Angus. 99. Dynamic Forces in Capialis Developmen: A Long-Run Comparaive View, Oxford Universiy Press, New York, NY. Shoven, John B., and Toshiaki Tachibanaki "The axaion of income from capial in Japan," In Governmen Policy owards Indusry in he Unied Saes and Japan, edied by John B. Shoven, Cambridge Universiy Press, Cambridge, England. Summers, Rober, and Alan Heson "A New Se of Inernaional Comparison of Real Produc and Price Levels: Esimaes for 30 Counries," Review of Income and Wealh, 34, -25.

23 Japan a USA a Personal saving rae (%) Privae saving rae (%) Naional saving rae (%) JAPAN USA Tax rae on Corporae income (τ p ) Ineres income (τ i) a Dividend income (τ d) a Capial gains (τ c ) a Gif (beques) income (τ g) Invesmen ax credi (Ω c) Depreciaion ax credi (Ω d ) 0 0 Deb-o-equiy raio (µ) Dividend yield (d) Depreciaion rae (δ ) b Inflaion rae (+π ) b Populaion growh rae (+n) b Survival probabiliy (ϕ) Ineremporal elasiciy of subsiuion (/σ) /2 /2 Gross naional saving rae (SAV) Per capia GDP growh rae (g) b Capial growh rae (g k) b Age-consumpion raio (CONS) Capial inensiy (α) Produciviy (B) Alruism parameer (λ) Subjecive ime preference (β )

24 Table 3 SIMULATION RESULTS FOR JAPAN (BENCHMARK CASE) Annual per capia income growh (%) Gross naional saving rae (%) Annual real ineres rae (%) Acual number All axes τ p only τ i only τ g only Ωc only n ϕ λ β B and α B only α only π δ σ =4 σ = Growh rae (annual %) Saving rae (%) Growh rae (annual %) Saving rae (%) Acual number All axes n ϕ λ β B, α π

25 Table 5 ALTRUISM AND BEQUEST Model Saving rae (%) Growh rae (%) σ = 2 Acual number 34. [00] 5.32 [00] No alruism 0.0 [ 30] 3.0 [ 58] No beques 5.79 [ 7] 2.07 [ 39] σ = 4 Acual number 34. [00] 5.32 [00] No alruism 0.44 [ 3] 3.6 [ 59] No beques 5.4 [ 5].85 [ 35] σ = Acual number 34. [00] 5.32 [00] No alruism 9.89 [ 29] 3.06 [ 58] No beques 6.6 [ 8] 2.8 [ 4]

26 일본의고성장시기인 960 년대초에서 980 년대말에이르는기간동안일본의저축율과경제성장율은미국보다훨씬높은수준을유지하였다. 내생성장이론 (endogenous growh heory) 에바탕을둔경제모델의 calibraion 을통하여, 이논문은이들두국가간의저축율과경제성장율이달랐던이유를밝히는데그목적을두고있다. 이논문은기존연구들에서중요하게다루어졌던저축의유산동기 (beques moive) 와예비금동기 (precauionary moive) 를동시에고려한포괄적인모델을제시하고있다. 고려가된여러사회적, 경제적요인들중에서일본인들의자식에대한강한 alruism 과일본의특수한세율제도가일본이 년대에빠른성장률과높은저축율을갖게한중요한요인이었음을이논문은밝히고있다. E2, O40 핵심주제어어 : 저축률, 저축동기, 경제성장, 일본