RIETI Dscusson Paper Seres 15-E-022 Agng, Interregonal Income Inequalty, and Industral Structure: An emprcal analyss based on the R-JIP Database and the R-LTES Database FUKAO Kyoj RIETI MAKINO Tatsuj Htotsubash Unversty The Research Insttute of Economy, Trade and Industry http://www.ret.go.jp/en/
RIETI Dscusson Paper Seres 15-E-022 February 2015 Agng, Interregonal Income Inequalty, and Industral Structure: An emprcal analyss based on the R-JIP Database and the R-LTES Database * FUKAO Kyoj (Htotsubash Unversty and RIETI) and MAKINO Tatsuj (Htotsubash Unversty) Abstract By mergng two newly created databases for the analyss of prefecture-level productvty the R-JIP Database 2013 and the R-LTES Database 2013 wth other regonal statstcs, we examne how and why aged prefectures dffer from other prefectures. Our man fndngs can be summarzed as follows: 1. The hgh aged populaton rato of some prefectures such as Akta and Shmane s due to a large out-mgraton experenced durng Japan s hgh-speed growth era from 1955 to 1970. 2. Aged prefectures tend to have lower labor productvty. At the same tme, we fnd that populaton agng does not systematcally reduce local total factor productvty (TFP) levels. We therefore argue that, rather than populaton agng reducng TFP levels, the causalty runs n the opposte drecton. Most prefectures wth a hgh aged populaton rato today had a low TFP level 30 40 years ago; as low TFP levels mean lower wage rates, such prefectures experenced an out-mgraton of the young. Gven that TFP dfferences across prefectures are stable over tme (prefectures wth a low relatve TFP level mantan ths condton), we observe a negatve correlaton between current TFP levels and current aged populaton ratos. Ths mples that there s no need for concern about Japan s average labor productvty declnng n the future as a result of populaton agng. 3. Aged prefectures tend to have large net mports of goods and servces. Ther large net mports are manly the result of large negatve government savngs. Actve government captal formaton n aged prefectures also contrbutes to some extent to ther net mports. Snce large transfers n the form of recepts of publc pensons and medcal care from less aged prefectures to more aged ones are not sustanable, t seems that resdents n prefectures that are less aged now should expect a post-retrement lfe that wll be much less prosperous than what resdents n Akta and Shmane enjoy today. Keywords: Aged populaton rato, Mgraton, Regonal nequalty, Total Factor Productvty, Income transfer, R-JIP Database, R-LTES Database JEL classfcaton: D24, J11, N35, N95, R11, R23 RIETI Dscusson Papers Seres ams at wdely dssemnatng research results n the form of professonal papers, thereby stmulatng lvely dscusson. The vews expressed n the papers are solely those of the author(s), and nether represent those of the organzaton to whch the author(s) belong(s) nor the Research Insttute of Economy, Trade and Industry. * Ths study was conducted as part of the project Regonal-Level Japan Industral Productvty Database: Database Refnement and Its Analyss at the Research Insttute of Economy, Trade and Industry. The authors are grateful for helpful comments and suggestons by Dscusson Paper semnar partcpants at RIETI. 1
1. Introducton Japan s populaton s agng rapdly. The share of those aged 65 and over n the total populaton n 2011 stood at 23.3%, the hghest n the world, and the speed of populaton agng n Japan s much faster than n the advanced European countres and the Unted States (Statstcal Research and Tranng Insttute 2012). However, populaton agng n Japan s proceedng at an uneven pace across regons. In some prefectures, such as Akta and Shmane, populaton agng as measured by the rato of those aged 65 and over s about 15 years ahead of Japan as a whole and 25 years ahead of the Tokyo Metropoltan Area. The economc stuaton n these prefectures lkely s a precursor of thngs to come n Japan as a whole. To formulate approprate land and macroeconomc polces, whch have been put n place to deal wth agng n Japan as a whole, t s mportant to understand the agng process n prefectures that are partcularly advanced along ths course. In ths paper, we examne how and why aged prefectures dffer from other prefectures by mergng two newly created databases for the analyss of prefecture-level productvty, the Regonal-Level Japan Industral Productvty Database 2013 (R-JIP 2013) and the Regonal-Level Long Term Economc Statstcs Database 2013 (R-LTES 2013), 1 wth other regonal statstcs, such as the Populaton Projectons by Regon (March 2012): 2011 to 2060 by the Natonal Insttute of Populaton and Socal Securty (IPSS), varous ssues of the Populaton Census by the Mnstry of Internal Affars and Communcatons (MIC), and varous ssues of the Annual Report of Prefectural Accounts by the Cabnet Offce. 2 The structure of the paper s as follows. In the next secton, we examne ntertemporal changes n the regonal concentraton of the aged populaton and the relatonshp between agng and regonal economc performance ndcators such as per capta ncome, labor productvty, and net exports of goods and servces. We try to extract some stylzed facts on these ssues from long-term prefectural data. We wll show that per capta ncome and labor productvty n aged prefectures are substantally lower than n other prefectures. We wll also show that aged prefectures are net mporters of goods and servces. In Secton 3, we then examne why n some prefectures, the populaton share of the old s so hgh and how the regonal concentraton of the aged populaton 1 The R-JIP s compled by the Research Insttute of Economy, Trade and Industry and Htotsubash Unversty. The R-JIP 2012 contans sectoral and macro-level data for each prefecture for the analyss of total factor productvty (TFP) for the perod 1970 2008. The R-JIP 2012 (n Japanese) s avalable at <http://www.ret.go.jp/jp/database/r-jip2012/ndex.html>. The R-JIP 2013 contans macro-level data for each prefecture for TFP analyss for the perod of 1955-2009 and each prefecture s sectoral-level data for TFP analyss for the perod 1970 2009. For detal on the R-JIP Database, see Toku et al. (2013). The R-LTES s compled by Htotsubash Unversty. The R-LTES 2013 contans sectoral data for each prefecture for the analyss of ndustral structures and labor productvty for the perod 1874 2009. For detals on the part of the R-LTES Database for the prewar perod, see Bassno et al. (2010). 2 Data reported n the Annual Report of Prefectural Accounts by the Cabnet Offce (n Japanese) are avalable at <http://www.esr.cao.go.jp/jp/sna/data/data_lst/kenmn/fles/fles_kenmn.html.> 1
changed over tme n Japan. Next, n Secton 4, we examne why per capta ncome and labor productvty n aged prefectures are lower than n other prefectures. Further, n Secton 5, we examne why aged prefectures are net mporters of goods and servces and whether the ndustral structures of aged prefectures dffer from those of other prefectures. Secton 6 concludes the paper. 2. How Are Aged Prefectures Dfferent from Other Prefectures? In ths secton, we examne ntertemporal changes n the regonal concentraton of the aged populaton. We also examne the relatonshp between agng and regonal economc performance ndcators such as per capta ncome, labor productvty, and net exports of goods and servces. Fgure 1 compares rato of the aged populaton (those aged 65 and over) n 2010 across prefectures. Akta has the hghest aged populaton rato at 29.6%, whch s 7 percentage ponts hgher than the natonal average and 9 percentage ponts hgher than that of Tokyo. The prefecture wth the lowest aged populaton rato s Oknawa, probably because of ts hgh brth rate. Except for Oknawa, all the 12 prefectures that have a lower aged populaton share than the natonal average are ether metropoltan areas, such as Tokyo, Osaka, Kanagawa, Ach, Fukuoka, and Myag, suburbs of metropolses, such as Satama, Chba, and Shga, or ndustral dstrcts around Tokyo, such as Ibarag and Tochg. % 35.0 30.0 25.0 Fgure 1. Aged Populaton Rato by Prefecture n 2010 Aged populaton rato Japan average = 23.0% 20.0 15.0 10.0 5.0 0.0 Akta Shmane Koch Yamaguch Yamagata Wakayama Iwate Tokushma Ehme Ota Nagano Kagoshma Tottor Ngata Toyama Nagasak Kagawa Myazak Aomor Kumamoto Okayama Fuku Fukushma Yamanash Hokkado Saga Me Gfu Nara Hroshma Shzuoka Ishkawa Gunma Kyoto Hyogo Ibarag Osaka Myag Fukuoka Tochg Chba Shga Tokyo Satama Ach Kanagawa Oknawa Source: Populaton Census 2010, Statstcs Bureau, MIC. 2
Fgure 2 shows how the aged populaton ratos of Akta, Shmane, Tokyo, and for Japan as a whole have changed over tme and how they are expected to change n future. The populaton projectons were conducted by IPSS. As already mentoned, Akta and Shmane are about 15 years ahead of Japan as a whole and about 25 years ahead of the Tokyo Metropoltan Area n terms of ther aged populaton ratos. % 50 Fgure 2. Aged Populaton Rato:1884 2040 45 40 35 30 25 Japan average Tokyo Akta Shmane 20 15 10 5 0 1884 1903 1908 1913 1918 1920 1925 1930 1935 1940 1947 1950 1955 1960 1965 1970 1975 1980 1985 1990 1995 2000 2005 2010 2015 2020 2025 2030 2035 2040 Sources: Data from 1884 to 1918 were estmated on the bass of the permanent domcle populaton. Data from 1920 to 2010 are based on the Populaton Census (varous years), Statstcs Bureau, MIC. Data from 2015 are projectons by the Natonal Insttute of Populaton and Socal Securty. Fgure 2 also shows that from a hstorcal perspectve, Akta and Shmane are qute dfferent from Japan as a whole. Shmane s aged populaton rato has been more than double that of Tokyo and more than 40 percent above the natonal average throughout the entre perod from 1884 to 2010. On the other hand, Akta s hgh aged populaton rato s a more recent phenomenon. Untl 1965, the rato for Akta was below the natonal average, but t subsequently rapdly pulled ahead of the natonal average. In the next secton, we wll examne why the populaton rato s so hgh n some prefectures such as Akta and Shmane. Another nterestng pont whch Fgure 2 shows s that the dfference n the aged populaton ratos across prefectures s projected to declne n the future. In fact, when we measure the dfference 3
n the aged populaton ratos across prefectures usng the coeffcent of varaton, we fnd that ths dfference show a long-term downward trend (Fgure 3). Why has the dfference n the aged populaton rato across prefectures been on a declnng trend and s projected to declne further n the future? Ths s an ssue we wll also examne n Secton 3. 0.25 0.20 Fgure 3. Coeffcent of Varaton over Tme Aged populaton rato Crude brth rate Death rate 0.15 0.10 0.05 0.00 1884 1903 1908 1913 1918 1920 1925 1930 1935 1940 1947 1950 1955 1960 1965 1970 1975 1980 1985 1990 1995 2000 2005 2010 2015 2020 2025 2030 2035 2040 Sources: See Fgure 2. Next, let us examne the relatonshp between agng and regonal economc performance, such as per capta ncome, labor productvty, and net exports of goods and servces. Fgure 4 shows the correlaton between the aged populaton rato (n logarthm) and per capta ncome (logarthm of per capta ncome/average per capta ncome of all prefectures) for 1955, 1970, 1990, and 2008. In ths fgure, r denotes the correlaton coeffcent. The correlaton s negatve and statstcally sgnfcant (at the 5% level) for all four benchmark years. Per capta ncome can be decomposed n the followng two ways: Per capta ncome = (GDP per worker) (number of workers) / (total populaton) + (net recepts of factor ncome from other prefectures and abroad) / (total populaton) = (GDP per hour) (man-hour nput) / (total populaton) + 4
(net recepts of factor ncome from other prefectures and abroad) / (total populaton) Per capta ncome/average per capta ncome of all prefectures (log) 0.8 0.6 0.4 0.2 0.0 0.2 0.4 0.6 1.0 1.5 2.0 2.5 3.0 3.5 Sources: R-JIP Database 2013 and Populaton Census, Statstcs Bureau, MIC. Labor productvty/average labor productvty of all prefectures (log) 0.8 0.6 0.4 0.2 0.0 0.2 0.4 r= 0.30 Fgure 4. Agng and Per Capta Income r= 0.59 Aged populaton rato (log) Fgure 5. Agng and Labor Productvty r= 0.41 r= 0.53 1955 1970 1990 2008 r= 0.60 r= 0.53 0.6 1.0 1.5 2.0 2.5 3.0 3.5 Aged populaton rato (log) r= 0.43 r= 0.55 1955 1970 1990 2008 Sources: See Fgure 4. 5
Notes: Labor productvty n 1955 s measured n terms of nomnal value added per worker. Labor productvty n 1970, 1990 and 2008 s measured n terms of nomnal value added per hour. The negatve correlatons observed n Fgure 4 are lkely partly due to the fact that aged prefectures tend to have lower workers/total populaton and man-hour nput/total populaton ratos. But t s mportant to note that, as Fgure 5 shows, the correlaton between the aged populaton rato and labor productvty s also negatve n all the benchmark years. 3,4 If populaton agng causes labor productvty declne, ths has grave mplcatons for agng socetes. It s well known that populaton agng goes hand n hand wth a lower percentage of workers n the total populaton and a reducton n per capta ncome. But the mpact of populaton agng on labor productvty has not been well analyzed yet. In Secton 4, we examne ths ssue n more detal. Accordng to the lfe-cycle model of household consumpton over tme, an ncrease of the aged populaton wll reduce prvate savngs. An ncrease of the aged populaton wll also reduce government tax ncome by reducng per capta ncome. On the other hand, an ncrease of the aged populaton wll have a negatve mpact on prvate fxed captal formaton because of the slow or negatve growth of the workforce. 5 If the negatve mpact of agng on prefectural savng s larger than the negatve mpact of agng on prefectural nvestment, agng prefectures wll have a negatve savng-nvestment balance and become net mporters of goods and servces. 6 Fgure 6 shows the cross-prefectural relatonshp between agng and the percentage of net exports of goods and servces n total gross prefectural expendture for benchmark years. The correlaton s postve and statstcally sgnfcant (at the 5% level) for all the benchmark years except 1955. The fgure also shows that some aged prefectures have a very hgh net mport rato. In Secton 5, we examne what factors are responsble for the large net mports of goods and servce, a low prvate (or government) savng rate or a hgh prvate (or government) nvestment. Prefectures wth a negatve savng-nvestment balance can meet supply shortage by mportng goods and servces. However, they cannot meet through mports any supply shortage of non-tradable 3 As explaned n the notes for Fgure 5, labor productvty n 1955 s measured n terms of nomnal value added per worker, whle labor productvty n 1970, 1990, and 2008 s measured n terms of nomnal value added per hour. 4 If many workers of one prefecture commute to other prefectures and f we measure labor nput on a resdent bass, we wll underestmate ths prefecture s labor productvty. Therefore, we use labor nput data on a workplace bass. A detaled dscusson of ths ssue s avalable n Toku et al. (2013). 5 If nterregonal captal flows are lmted, a declne n savng n a prefecture mght reduce that prefecture s nvestment. However, snce there are no barrers to nterregonal captal flows wthn Japan and such captal flows are lkely very actve, we thnk that the effect of populaton agng on prefectural savng s unlkely to have a large mpact on prefectural nvestment. 6 These ssues have been dscussed manly at the natonal level. See Ontsuka (1974) and Auerbach and Kotlkoff (1990). 6
goods and servces. As open economy macroeconomcs tells us, countres or regons wth negatve a savng-nvestment balance (net mporters of goods and servces) therefore need to allocate more resources to the producton of non-tradable goods and servces. We would therefore expect a declne of tradable sectors, such as manufacturng and the prmary sector, n agng prefectures. We wll also examne ths ssue n greater detal n Secton 5. Net exports of goods and servces/gross domestc expendture % 40 30 20 10 0 10 20 Fgure 6. Relatonshp Between Agng and Net Exports of Goods and Servces r= 0.18 r= 0.43 r= 0.31 1955 1970 1990 2008 r= 0.38 30 1.0 1.5 2.0 2.5 3.0 3.5 Aged populaton rato (log) Sources: R-JIP Database 2013 and Populaton Census, Statstcs Bureau, MIC. 3. Why s the Aged Populaton Rato of some Prefectures so Hgh? In ths secton, we examne why the aged populaton rato of some prefectures such as Akta and Shmane s so hgh. We also examne why the dfference n the aged populaton rato across prefectures has been on a declnng trend and s projected to declne further n future. Fgure 7 compares varous populaton statstcs for Akta and Shmane wth those of Tokyo: the rates of natural and socal populaton change, the crude brth rate, the death rate, and the total fertlty rate. The fgure shows that there were substantal dfferences n rate of socal populaton change across the three prefectures from 1920 to 1970. Shmane and Akta experenced large out-mgraton durng ths perod, and the out-mgraton rate was partcularly hgh durng Japan s hgh growth era from 1955 to 1970. 7
Fgure 7. Comparson of Populaton Change Statstcs across Akta, Shmane and Tokyo: 1920-2010 % 12 10 8 6 4 2 0 2 4 Rate of Natural Populaton Change Akta Shmane Tokyo % 25 20 15 10 5 0 5 10 15 Rate of Socal Populaton Change Akta Shmane Tokyo 50 Crude Brth Rate 30 Death Rate 40 30 20 Akta Shmane Tokyo 25 20 15 10 Akta Shmane Tokyo 10 5 0 1920 1940 1950 1960 1970 1980 1990 2000 2010 0 1920 1940 1950 1960 1970 1980 1990 2000 2010 Total Fertlty Rate 7 6 5 4 Akta Shmane Tokyo 3 2 1 0 1925 1930 1950 1960 1970 1980 1990 2000 2010 Source: Populaton Census (varous ssues), Statstcs Bureau, MIC. Data of the total fertlty rate were provded by Professor Norko Tsuya of Keo Unversty. On the other hand, Tokyo experenced extremely rapd n-mgraton untl the 1960s. It s nterestng to note that the n-mgraton rate of Tokyo was already hgh before the Second World War. Populaton nflows to Tokyo gradually declned n the 1960s and even turned negatve durng the 1970s and 1980s, probably because of congeston phenomena, such as hgh land prces, ar polluton, etc., and regulatons on the startup of new factores and unversty campuses to stop overconcentraton. People who wanted to work or study n the Tokyo area started to choose lvng n adjacent prefectures such as Chba, Satama, and Kanagawa. We should pont out that Fgure 7 does not help to explan why Shmane s aged populaton rato was so hgh even n the pre-war perod (Fgure 2). One possble explanaton s that Shmane s 8
out-mgraton started much earler than that of Akta, but to confrm ths we need more long-term data than that shown n Fgure 7. 7 People are partcularly lkely to mgrate n ther teens or twentes. Therefore, out-mgraton wll reduce the workng age populaton and the crude brth rate, and wll ncrease the aged populaton rato for the followng 30-40 years. Fgures 8 and 9 show the correlaton between the rate of socal populaton change over each 5 year nterval and the aged populaton rato 40 years later by prefecture for both the post-war and the pre-war perod. We fnd a negatve and statstcally sgnfcant correlaton (at the 5% level) between n-mgraton and the aged populaton rato 40 years later for all perods analyzed. 3.5 Fgure 8. Mgraton and Aged Populaton Rato 40 Years Later: Post War Perod Rate of Socal Populaton Change 1950 55 and Aged Populaton Rato n 1990 Aged populaton rato (log) 3.0 2.5 r= 0.61 r= 0.55 r= 0.82 Rate of Socal Populaton Change 1960 65 and Aged Populaton Rato n 2000 Rate of Socal Populaton Change 1970 75 and Aged Populaton Rato n 2010 2.0 15 10 5 0 5 10 15 20 25 Rate of Socal Populaton Change % 7 Fgure 7 also shows that Akta and Shmane have experenced large negatve rates of natural populaton change n recent years. The rate of natural populaton change s equal to the crude brth rate mnus the death rate. Fgure 7 ndcates that Akta and Shmane have consderably hgher death rates than Tokyo, lkely because they have older populatons. Moreover, although Akta and Shmane have hgher total fertlty rates than Tokyo.e., women of chldbearng age n the two rural prefectures tend to have more chldren than those n Tokyo, the crude brth rates n the three prefectures are very smlar, lkely because there are relatvely fewer women of chldbearng age n the two rural prefectures due to ther aged populatons. Ths means that populaton agng reproduces tself through the declne n the crude brth rate and that, as a consequence, the rate of natural populaton declne wll contnue to be faster n prefectures wth an older populaton. 9
Source: Populaton Census (varous ssues), Statstcs Bureau, MIC. 2.5 Fgure 9. Mgraton and Aged Populaton Rato 40 Years Later: Pre War Perod Rate of Socal Populaton Change 1920 25 and Aged Populaton Rato n 1960 Aged populaton rato (log) 2.0 1.5 r= 0.66 r= 0.45 Rate of Socal Populaton Change 1930 35 and Aged Populaton Rato n 1970 1.0 15 10 5 0 5 10 15 20 Rate of Socal Populaton Change % Source: See Fgure 8. Next, we examne the slowdown of mgraton wthn Japan. Fgure 10 shows the rate of socal populaton change n the top 10% and bottom 10% n-mgraton prefectures n terms of cumulatve populaton. We fnd that mgraton as measured by the rate of socal populaton change was hghest n the hgh growth era from 1955 to the early 1970s and declned markedly thereafter. Table 1 shows that for a long tme, people n Japan, lke people n other countres, mgrated from low ncome to hgh ncome regons. Lke Fgure 7, ths table shows that mgraton was most pronounced durng the hgh growth era,.e., n 1955 and 1970 n our table. However, mgraton, as measured by the rate of socal populaton change n hgh and low ncome prefectures, has slowed n recent decades. There was more actve mgraton n 1925-30 than n 1990 and 2008. Ths fact s related to our second queston, namely, why dfferences n the aged populaton rato across prefectures has been declnng and s projected to declne further n the future. 10
Fgure 10. Rate of Socal Populaton Change n the Top 10% and Bottom 10% In mgraton Prefectures n Terms of Cumulatve Populaton Socal ncrease rate of populaton n each fve years (%) 20 15 10 5 0 5 10 15 Top 10% Source: Populaton Census (varous ssues), Statstcs Bureau, MIC. Table 1. Rate of Socal Populaton Change n Hgh and Low Income Prefectures: 1925-2008 (%, Annual Rate) Rate of socal populaton change n low per capta ncome prefectures (bottom 20% prefectures n terms of cumulatve populaton) Rate of socal populaton change n hgh per capta ncome prefectures (top 20% prefectures n terms of cumulatve populaton) 1925-30 1955 1970 1990 2008-0.6-0.7-1.3-0.4-0.4 1.6 1.7 0.0-0.4 0.4 1925-30 1955 1970 1990 2008 Rate of socal populaton change n low per capta ncome prefectures (bottom 10% prefectures n terms of cumulatve populaton) Rate of socal populaton change n hgh per capta ncome prefectures (top 10% prefectures n terms of cumulatve populaton) -0.6 2.2-0.7-1.7-0.5-0.4 2.3-0.9-0.4 0.6 Sources: Populaton Census, and Annual Report on Internal Mgraton n Japan Derved from the Basc Resdent Regstraton, Statstcs Bureau, MIC; R-LTES 2013; Annual Report of Prefectural Accounts, Cabnet Offce. 11
Notes: The selecton of the top and bottom prefectures n 1925 s based on the per capta gross prefectural product (GPP). The rate of socal populaton change for 1925-1930 s the annual average rate of change for that perod. We can pont out two causes of ths slowdown of mgraton from low ncome to hgh ncome prefectures. Frst, as the agng of the populaton proceeds, the percentage of the populaton at an age when ndvduals are most lkely to mgrate, that s, the populaton n ther teens and twentes n the total populaton, declnes. Second, ncome dspartes across prefectures, whch are the man engne of mgraton, have gradually declned. In order to confrm that not only the frst factor but also the second factor has been responsble for the slowdown of mgraton n recent years, we measured the net mgraton rate of each cohort from when they are aged 10 14 to when they are aged 30 34 for each prefecture. For example, to derve the 20-year net mgraton rate of the cohort that was aged 10 14 n 1955, we used the followng equaton: Net n-mgraton rate of prefecture from 1955 to 1975 = {(Populaton of 30 34 year olds n prefecture n 1975) (Populaton of 10 14 year olds n prefecture n 1975) (Natonal average survval rate of ths cohort from 1955 to 1975)}/ (Populaton of 10 14 year olds n prefecture n 1975) Fgure 11 shows our result. Prefectures are ordered n terms of ther net n-mgraton rate from 1955 to 1975. The fgure shows that young generatons mgrated from low ncome prefectures to hgh ncome prefectures. We fnd that mgraton of younger generatons as measured by the net mgraton rates of cohorts from when they are aged 10 14 to when they are aged 30 34 declned substantally. As already explaned, one lkely explanaton s that as ncome and wage dspartes decreased over tme younger generatons n low ncome prefectures had less ncentve to move to hgh ncome prefectures. Another lkely explanaton s that the declne of the brth rate created many one-chld famles and chldren of such famles tended to stay wth ther parents. Ths change may have contrbuted to reducng out-mgraton from low ncome regons. However, more research s needed to determne what the man reasons for the declne of the mgraton rate among younger generatons are. 12
Fgure 11. Net Mgraton Rate of Each Cohort n the 20 Years from Age 10 14 to Age 30 34 by Prefecture 120 100 80 60 40 1955 75 1970 90 1990 2010 20 0 20 40 60 Kanagawa Satama Osaka Chba Tokyo Ach Nara Hyogo Kyoto Shzuoka Hroshma Shga Fukuoka Gfu Hokkado Ibarag Ishkawa Me Wakayama Okayama Myag Gunma Toyama Tochg Fuku Kagawa Aomor Yamaguch Nakano Koch Iwate Ngata Ehme Tottor Yamanash Ota Myazak Tokushma Nagasak Fukushma Akta Kumamoto Yamagata Saga Shmane Kagoshma Oknawa Sources: Populaton Census (varous ssues), Statstcs Bureau, MIC. Table 2 shows the correlaton coeffcents between the log of labor productvty relatve to the natonal average and the net n-mgraton rate of each cohort n the 20 years from age 10 14 to age 30 34. We fnd that the correlaton between labor productvty and subsequent mgraton s qute hgh. Table 2. Correlaton Coeffcent Between the Log of Labor Productvty Relatve to the Natonal Average and the Net In Mgraton Rate of Each Cohort n the 20 Years from Age 10 14 to Age 30 34 1955 1970 1990 1975 1990 2008 Correlaton coeffcent 0.628 0.750 0.710 Sources: R-JIP Database 2013 and Populaton Census (varous ssues), Statstcs Bureau, MIC. Note: Oknawa s not ncluded n the calculaton for 1955 75 and 1970 90. Next, we examne how ncome dspartes across prefectures, whch are lkely to have been the man engne of mgraton, have changed over tme. Fgure 12 shows that regonal ncome dspartes measured n terms of the coeffcent of varaton for per capta (nomnal) gross prefectural product (GPP) has declned substantally. 13
As mentoned n footnote 5, as commutng across prefectures has ncreased n recent years, t s not approprate to measure regonal ncome dspartes n terms of per capta GPP. In order to take account of ths problem, Fgure 13 compares the coeffcents of varaton of per capta prefectural ncome, GPP per worker on a workng place bass, and GPP per hour on a workng place bass, n addton to coeffcent of varaton of per capta GPP. Regonal economc dspartes measured n terms of the coeffcent of varaton become smaller when ncome data, whch nclude factor ncome from other prefectures, are used n place of GPP data. Regonal economc dspartes measured n terms of the coeffcent of varaton show a sharp downward trend when we use data on a workng place bass. Fgure 12. Regonal Economc Dspartes n Japan: 1890 2008 0.5 0.4 0.4 0.3 0.3 0.2 0.2 0.1 0.1 Coeffcent of varaton of per capta GPP (R LTES) Coeffcent of varaton of per capta GPP (Cabnet Offce) Gn coeffcent 0.0 1890 1895 1900 1905 1910 1915 1920 1925 1930 1935 1940 1945 1950 1955 1960 1965 1970 1975 1980 1985 1990 1995 2000 2005 Sources: R-LTES 2013 and Annual Report of Prefectural Accounts, Cabnet Offce. Note: The Gn coeffcent s calculated assumng that all resdents n the same prefecture have the same ncome. 14
Fgure 13. Regonal Economc Dspartes n Japan: 1970 2008 0.25 0.20 0.15 0.10 Per capta nomnal GPP 0.05 Nomnal GPP per hour on a workng place bass Per capta nomnal prefectural ncome Nomnal GPP per worker on a workng place bass 0.00 1970 1975 1980 1985 1990 1995 2000 2005 Sources: R-JIP Database 2013 and Annual Report of Prefectural Accounts, Cabnet Offce. 4. Why Is Labor Productvty n Aged Prefectures Low? In ths secton, we examne why labor productvty tends to be lower n aged prefectures. Followng Caves et al. (1982), we decompose labor productvty dfferences across prefectures nto dfferences n the TFP level, dfferences n captal ntensty, and dfferences n labor qualty: K K Z r H r 1 L L Qr s s ln ln s s ln v 1 ln RTFPr r r (1) v 2 Z H 2 Q where v, Z, H, Q, and RTFP denote nomnal value added per hour, captal servce nput, man-hour K L labor nput, labor qualty, and the relatve TFP level of prefecture. s and s denote the captal and labor cost shares n prefecture. Varables wth an upper bar denote the geometrc mean of that varable across prefectures. Usng the above equaton we can decompose the covarance between the log of the aged populaton rato and the log of labor productvty relatve to the natonal average nto (1) the covarance between the log of the aged populaton rato and the log of the relatve TFP level (the frst term on the rght-hand sde of the above equaton), (2) the covarance between the log of the aged populaton rato and the contrbuton of the relatve captal-labor rato to the labor productvty 15
gap (the second term on the rght-hand sde of the above equaton), and (3) the covarance between the log of the aged populaton rato and the contrbuton of relatve labor qualty to the labor productvty gap (the thrd term on the rght-hand sde of the above equaton). Table 3 shows the result of ths decomposton. We used the R-JIP Database 2013. Table 3. Decomposton of the Covarance Between the Log of the Aged Populaton Rato and the Log of Labor Productvty Relatve to the Natonal Average and Each Factor's Contrbuton (n Parentheses) 1955 1970 1990 2008 Covarance between log of aged populaton rato and log of labor productvty relatve to natonal average Covarance between log of aged populaton rato and log of relatve TFP level Covarance between log of aged populaton rato and contrbuton of relatve captal labor rato to labor productvty gap Covarance between log of aged populaton rato and contrbuton of relatve labor qualty to labor productvty gap 0.020 0.017 0.012 0.005 (100.0) (100.0) (100.0) (100.0) 0.015 0.007 0.007 0.003 (74.9) (39.8) (56.3) (55.3) 0.003 0.008 0.003 0.001 (13.8) (45.4) (21.8) (21.5) 0.002 0.003 0.003 0.001 (11.3) (14.8) (21.8) (23.1) Sources: R-JIP Database 2013 and Populaton Census (varous ssues), Statstcs Bureau, MIC. As Table 3 shows, the covarance between the log of the aged populaton rato and all three factors underlyng labor productvty dfferences TFP, the captal-labor rato, and labor qualty s always negatve. Moreover, the correlaton coeffcents are all statstcally sgnfcant at the 5% level except n two cases: the correlaton wth the captal-labor rato n 1955 and 2008. We can thus say that all three factors contrbuted to the lower labor productvty n prefectures wth a larger aged populaton rato. Among the three factors, TFP dfferences made the largest contrbuton n 1955, 1990, and 2008. In other words, labor productvty n aged prefectures s relatvely low manly because of ther low TFP. Fgure 14 shows the relatonshp between the relatve TFP level and the aged populaton rato. 16
Fgure 14. Aged Populaton Rato and TFP 0.8 0.6 1955 1970 1990 2008 0.4 Relatve TFP level (log) 0.2 0.0 0.2 r= 0.44 r= 0.38 r= 0.50 r= 0.32 0.4 0.6 1.0 1.5 2.0 2.5 3.0 3.5 Aged populaton rato (log) Sources: R-JIP Database 2013 and Populaton Census (varous ssues), Statstcs Bureau, MIC. Table 3 suggests that n 1970, the contrbuton of captal-labor rato dfferences was slghtly larger than that of TFP dfferences. Compared wth these two factors, the contrbuton of labor qualty dfferences was relatvely small. Ths s partly because of smaller dfferences n labor qualty across prefectures (Toku et al. 2013). Next, we nvestgate why aged prefectures tend to have a lower TFP level. One possble explanaton s that agng places a greater burden on local government and the local workng age populaton through a declne n tax revenues and an ncrease n the need to provde care for the elderly, medcal servce, etc. Another explanaton s that, as populaton agng proceeds, demand for care for the elderly and medcal servces expands and the producton share of these sectors n the local economy ncreases. Snce TFP growth n these sectors tends to be low, growth of these sectors may reduce macro-level TFP growth of aged prefectures. Ths s a knd of Baumol effect. Fgure 15 shows the relatonshp between the aged populaton rato n 1955, 1970, and 1990 and TFP growth n the succeedng 20 years. The correlaton s not statstcally sgnfcant for 1970-90 and 1990-2008, but t s postve and sgnfcant for 1955-75. These results suggest that populaton agng does not reduce TFP growth. We also checked whether there s a statstcally sgnfcant negatve correlaton between the change of the aged populaton rato and the change of the TFP level n each perod,.e., 1955 75, 1970 90, and 1990 2008. However, as Fgure 16 shows, we do not observe such a relatonshp. 17
Fgure 15. Aged Populaton Rato and TFP Growth 0.6 Change of log of TFP level relatve to natonal average 0.4 0.2 0.0 0.2 0.4 r=0.46 (1955 1975) r= 0.02 (1970 1990) r=0.08 (1990 2008) 1955 1975 1970 1990 1990 2008 0.6 1.2 1.4 1.6 1.8 2.0 2.2 2.4 2.6 2.8 3.0 Aged populaton rato (log) Sources: R-JIP Database 2013 and Populaton Census (varous ssues), Statstcs Bureau MIC. 0.6 Fgure 16. Change of Aged Populaton Rato and TFP Growth Change of log of TFP level relatve to natonal average 0.4 0.2 0.0 0.2 0.4 r= 0.01 (1955 1975) r=0.03 (1990 2008) 1955 1975 r= 0.21 (1970 1990) 1970 1990 1990 2008 0.6 0.1 0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 Change of log of aged populaton rato Sources: See Fgure 15. 18
Employng a fxed effect model, we further estmated how the rato of the aged populaton to the workng age populaton affects the TFP level and TFP growth usng prefectural level data for the perod 1955-2008. The estmaton results are reported n Table 4 and ndcate that there s a statstcally sgnfcant negatve relatonshp between ths rato and the TFP level. However, n the case of TFP growth, we fnd a postve and statstcally sgnfcant relatonshp. Therefore, although there s a negatve correlaton between agng and the TFP level, t seems that agng does not slow down TFP growth. These results are consstent wth our fndngs from Fgures 14 and 15. Table 4. Relatonshps between Agng and the TFP Level/TFP Growth: Estmaton Results Based on Fxed Effect Models, 1955 2008 Dependent varable Log of TFP level TFP growth Log of (aged populaton/workng age populaton 0.138 ** 0.018 *** (0.065) (0.008) Number of observatons 184 138 Number of prefectures 46 46 Housman statstc 4.684 2.359 p value 0.321 0.501 Adjusted R 2 0.042 0.120 Notes: In the case of the TFP level regresson, we used data for 1955, 1970, 1990, and 2008. In the case of the TFP growth regresson, we used data for 1955-1970, 1970-1990, and 1990-2008. Oknawa s not ncluded n the regressons. Fgures n parentheses are t-statstcs. Year and prefecture dummes are ncluded n each regresson, but ther coeffcents are not reported. ** p<0.05 and *** p<0.01. Fgure 17 shows our result on the Baumol effect. Usng sectoral TFP data, whch are only avalable for 1970 2008 n the present R-JIP Database, we can decompose the macro-level TFP growth of each prefecture nto the contrbuton of the expanson of sectors wth hgh TFP growth and the contrbuton of TFP growth wthn each sector. We call the frst term the Baumol effect. The vertcal axs of Fgure 17 measures the percentage of the Baumol effect n the total TFP growth of each prefecture for the perods 1970 1990 and 1990 2008. The horzontal axs shows the log of the aged populaton rato n the startng year of each perod, 1970 and 1990 respectvely. In the case of the perod 1970 1990, we fnd a statstcally sgnfcant (at the 5% level) negatve correlaton between the aged populaton rato n the startng year and the percentage of the Baumol effect n total TFP growth n the succeedng 20 years. However, even n the perod 1970 1990, the Baumol effect accounts for less than 10% of the total TFP growth n most of prefectures. To sum up our results on the effects of populaton agng on TFP growth, t seems that populaton agng does not systematcally reduce local TFP levels. 19
Fgure 17. Aged Populaton Rato and Baumol Effect 60 Percantage of contrbuton of Baumol effect to total TFP growth 50 40 30 20 10 0 10 20 1970 1990 1990 2008 r= 0.33 (1970 1990) r=0.13 (1990 2008) 30 1.2 1.4 1.6 1.8 2.0 2.2 2.4 2.6 2.8 3.0 Aged populaton rato (log) Sources: R-JIP Database 2013 and Populaton Census (varous ssues), Statstcs Bureau, MIC. Let us consder the possblty that the causalty s the reverse of what we have assumed so far. That s, f a lower TFP level means lower wage rates and out-mgraton of the younger generaton, prefectures wth a hgh aged populaton rato may have suffered low TFP levels 30 40 years ago, so f TFP dfferences across prefectures are stable over tme (prefectures wth a low relatve TFP level contnue to have a low relatve TFP level), we wll observe a negatve current correlaton of the TFP level and the aged populaton rato. Table 5 shows the nter-temporal correlaton coeffcent of the relatve TFP level. All the coeffcents are statstcally sgnfcant at the 5% level, ndcatng that TFP dfferences across prefectures are qute stable over tme. Fgure 4 shows the standard devaton of the relatve TFP level (log) across prefectures. We fnd that from 1955 to 1970, TFP dfferences across prefectures declned substantally, but after that cross-regonal dfferences dd not change (Table 6). Table 5. Intertemporal Correlaton Coeffcent of Relatve TFP Level 1955 1970 1990 1970 0.61 1990 0.60 0.68 2008 0.63 0.53 0.65 20
Source: R-JIP Database 2013. Note: Oknawa s not ncluded n the analyss. Table 6. Standard Devaton of Relatve TFP Level: 1955 2008 1955 1970 1990 2008 0.18 0.09 0.08 0.08 Source: R-JIP Database 2013. Note: Oknawa s not ncluded n the calculatons for 1955 and 1970. What factors are responsble for ths persstence n TFP gaps across regons? Snce frms carry out a lot of nvestment across regons and large frms have networks of afflates throughout Japan, t seems to be dffcult to explan ths persstence as the result of slowness n technology dffuson. Instead, there must be other reasons and there are a number of possble explanatons for the persstence n TFP gaps. One of these s agglomeraton effects. Hgh ncome regons enjoy postve agglomeraton effects n certan ndustres or wth regard to the populaton overall. And snce frms and households are attracted by such postve effects, agglomeraton does not dsappear. Another possble explanaton s locatonal advantages as a result of natural or polcy factors. A prefecture wth a good sea port may enjoy hgher TFP for a long perod. Next, let us examne the theoretcal relatonshp between dfferences n TFP across prefectures and dfferences n wage rates across prefectures. We assume a constant returns to scale neoclasscal producton functon wth two producton factors, captal and labor. Captal moves wthout frcton across prefectures and the real rate of return on captal s always equalzed across prefectures. On the other hand, we assume that t takes tme for labor to move and workers decson about where to lve depends not only on wage rates but also other factors such as amentes, ther personal hstory, the locaton of ther parents, etc. Therefore, we assume that labor supply n each regon s gven at least n the short run and wage rate dfferences across prefectures contnue to exst. All prefectures produce dentcal products and all markets are characterzed by perfect competton. To smplfy our analyss, we assume that labor qualty s constant and does not dffer across prefectures, although our man results would reman unchanged f we relaxed these assumptons regardng labor qualty. We assume the followng neoclasscal constant returns to scale producton functon: 8 8 We assume f (k, A ) 0, f (k, A )/ k >0, 2 f (k, A )/ k 2 <0, f (k, A )/ A >0 for any k 0 and A >0, and lm f ( k, A ) / k k 0, and lm f ( k, A ) / k 0. k 21
K, N, A N f k A V F,, (2) where V, A, K, N, and k denote the real value added, the productvty ndex, the captal servce nput, the labor nput, and the captal-labor rato (K /N ) of prefecture. Because of free captal movement, we have f k, A k r, (3) where r denotes the rate of return on captal, whch s dentcal across regons. The real wage rate of prefecture, w, s determned by the margnal productvty of labor n ths prefecture: w f k f k, A, A k. (4) k To smplfy our analyss, we ntroduce a small prefecture assumpton, just lke the small country assumpton n nternatonal economcs. We assume that prefecture s so small that a change of the TFP level n ths prefecture does not affect the natonal rate of return on captal, r. We assume that r s gven and constant. Under ths assumpton, we can regard equaton (3) as an mplct facton, whch shows how k s determned for a gven value of A. Let us express ths relatonshp by k = k(a ). By dfferentatng equaton (3) wth regard to A, we obtan 2 2 k, A dk A f k, A f k 2 da Ak 0. (5) Usng k = k(a ), equaton (4) can be regarded as showng how changes n A affect the wage rate w : w f ka A ka f k A, A,. (6) k 22
By dfferentatng equaton (6) wth regard to A and w and usng equaton (5), we obtan f dw f f ka, A dka f ka 2 2 ka, A dka f ka, A dka f ka k k A A k, A da da da da da k A k, A 2 da da da k, A Ak da (7) Keepng N and k constant and dfferentatng producton functon (2) wth regard to V and A, we obtan the followng relatonshp: dv V N V f k, A A da (8) The left-hand sde of the above equaton denotes percentage change of TFP, d TFP / TFP. Therefore, the above equaton shows the relatonshp between a change n TFP and a change n the productvty ndex. From equatons (7) and (8), we obtan the followng relatonshp, whch we have been lookng for: dw V dtfp (9) w w N TFP On the rght-hand sde, the coeffcent, V /w N denotes the nverse of the cost share of labor, whch s larger than 1. Therefore, equaton (9) means that when TFP ncreases, the wage rate wll ncrease more than proportonately under our assumpton of a constant rate of return on captal. Wll hgher TFP brng a hgher captal-labor rato? The relatonshp of these two varables s gven by equaton (5). Under the assumptons we have made above, the relatonshp s ambguous. However, we can show that f technology mprovements are ether Hcks neutral or Harrod neutral, one of whch s assumed n most studes on economc growth, hgher TFP wll brng a hgher captal-labor rato. 23
What we have done above s a comparatve statc analyss of a small prefecture. But we can easly apply our results for the comparson of wage rates between two prefectures wth dfferent TFP levels, at least when the two prefectures are not very large. Fgure 18 shows the cross-prefectural relatonshp between relatve TFP levels and wage rates. When we measure wage rates, dfferences n labor qualty are not taken nto account. Therefore, there s a rsk that we may overestmate the wage rates of hgh ncome prefectures, snce the labor qualty of such prefectures tends to be hgh (Toku et al., 2013). The fgure shows that there s a statstcally sgnfcant (at the 5% level) postve correlaton between TFP levels and wage rates. The slopes of the estmated ftted lnes are less than one for 1955 and close to one for later years. Therefore, the slopes are smaller than equaton (9) mples. However, takng account of the potentally large measurement errors n TFP, especally for 1955, we can probably say that the emprcal results n Fgure 18 are consstent wth theory. 0.5 Fgure 18. TFP and Wage Rate Wage rate/average wage rate of all prefectures (log) 0.4 0.3 0.2 0.1 0.0 0.1 0.2 r=0.73 (2008) r=0.49 (1970) r=0.54 (1955) 1955 1970 1990 2008 r=0.73 (1990) 0.3 0.4 0.3 0.2 0.1 0.0 0.1 0.2 0.3 0.4 0.5 0.6 Relatve TFP level (log) Sources: R-JIP Database 2013 and Monthly Labour Survey (Prefectural Survey), Mnstry of Health, Labour and Welfare. Dd a lower TFP level cause the out-mgraton of younger generatons? In Fgure 19, we compare each prefecture s relatve TFP level n 1955, 1970, and 1990 and the out-mgraton of younger generatons n the perods 1955 75, 1970 90, and 1990 2010, respectvely. In the same 24
way as n Fgure 11, we measure mgraton of younger generatons n terms of the net mgraton rates of a cohort from when they are aged 10 14 to when they are aged 30 34. Fgure 19 shows that the correlaton s postve and statstcally sgnfcant (at the 5% level) n all three cases. Let us summarze the results of ths secton. We found that populaton agng does not systematcally reduce prefectural TFP levels. Instead, we found that t s more plausble that causalty runs n the opposte drecton. That s, a low TFP level means a low wage rate and out-mgraton of younger generatons. Most prefectures that have a hgh aged populaton rato today had a low TFP level 30 40 years ago. And snce TFP dfferences across prefectures are stable over tme (prefectures wth a low relatve TFP level contnue to have a low relatve TFP level), we observe a negatve correlaton between current TFP levels and current aged populaton ratos. Net rate of n mgraton of each cohort n the 20 years from age 10 14 to age 30 34 by prefecture 120 100 80 60 40 20 0 20 40 Fgure 19. TFP and Net Mgraton Rate of Each Cohort n the 20 Years from Age 10 14 to Age 30 34 by Prefecture 1955 1970 1990 r=0.49 (1970) r=0.65 (1990) r=0.41 (1955) 60 0.4 0.3 0.2 0.1 0.0 0.1 0.2 0.3 0.4 0.5 0.6 Relatve TFP level (log) Sources: R-JIP Database 2013 and Populaton Census (varous ssues), Statstcs Bureau, MIC. 5. Aged Prefectures Net Exports and Industral Structure As seen n Fgure 6, aged prefectures tend to have large net mports of goods and servces. In ths secton, we examne what factors are responsble for these large net mports of goods and servces a low prvate (or government) savng rate or a hgh prvate (or government) nvestment rate. As mentoned above, accordng to open economy macroeconomcs, we would expect tradable sectors such as manufacturng and prmary sector to declne n agng prefectures. We also examne ths ssue n ths secton. 25
Fgures 20, 21, 22 and 23 plot the log of the aged populaton rato aganst the prvate gross savng/gross prefectural expendture rato, the prvate gross nvestment/gross prefectural expendture rato, the government gross savng/gross prefectural expendture rato, and the government gross nvestment/gross prefectural expendture rato, respectvely, for the years 1955, 1970, 1990, and 2008. As Fgures 20 and 21 show, n the case of prvate sector savng and nvestment, there s lttle correlaton wth the aged populaton rato. On the other hand, Fgure 22 shows, there s a statstcally sgnfcant (at the 5% level) negatve correlaton between the aged populaton rato and the government gross savng/gross prefectural expendture rato. Moreover, the negatve slope has been gettng steeper over tme and some aged prefectures recorded remarkably large negatve government savng ratos n 2008. The largest was that of Koch (34.6%) and the second largest was that of Shmane (31.9%). Fgure 23 shows there s also a statstcally sgnfcant (at the 5% level) postve correlaton between the aged populaton rato and the government nvestment/gross prefectural expendture rato. It s dffcult to understand why the central government concentrates nvestment n aged prefectures. To sum up our analyss, the large net mports of aged prefectures, whch we observed n Fgure 6, are manly the result of large negatve government savng n those prefectures. Government s actve captal formaton n agng prefectures also made some contrbuton to the net mports of agng prefectures. % Fgure 20. Aged Populaton Rato and Prvate Gross Savng Rato 70 Prvate gross savng/gross prefectural expendture 60 50 40 30 20 10 1955 1970 1990 2008 r= 0.30 r= 0.08 r= 0.05 r= 0.06 0 1.0 1.5 2.0 2.5 3.0 3.5 Aged populaton rato (log) 26
Sources: R-JIP Database 2013; Fukao and Yue (2000); Populaton Census (varous ssues), Statstcs Bureau, MIC; Prefectural Economc and Fscal Model/Database; and Annual Report of Prefectural Accounts (varous ssues), Cabnet Offce. % Fgure 21. Aged Populaton Rato and Prvate Gross Investment Rato Prvate gross nvestment/gross prefectural expendture 40 35 30 r= 0.14 1955 25 1970 r= 0.20 1990 20 2008 r= 0.08 15 r= 0.14 10 5 0 1.0 1.5 2.0 2.5 3.0 3.5 Aged populaton rato (log) Sources: See Fgure 20. % Fgure 22. Aged Populaton Rato and Government Gross Savng Rato Government gross savng/gross prefectural expendture 30 20 r= 0.61 10 r= 0.64 r= 0.38 0 r= 0.64 10 1955 20 1970 1990 2008 30 40 1.0 1.5 2.0 2.5 3.0 3.5 Aged populaton rato (log) 27
Sources: See Fgure 20. % Fgure 23. Aged Populaton Rato and Government Gross Investment Rato 20 Government gross nvestment/gross prefectural expendture 15 1955 1970 1990 2008 10 5 0 1.0 1.5 2.0 2.5 3.0 3.5 Aged populaton rato (log) Sources: See Fgure 20. What factors are responsble for such large negatve government savng n aged prefectures? Probably not surprsngly, t turns out the most mportant factor s net recepts of publc pensons and medcal care. As Fgure 24 shows, for some aged prefectures, the rato of net recepts to gross prefectural expendture s around 12-13%. Another factor s the low rato of tax payments to gross prefectural expendture, as shown n Fgure 25. We should note that these large government transfers must contrbute to rasng the prvate savng rato of agng prefectures. As we have seen n Fgure 20, there s no negatve correlaton between the aged populaton rato and prvate gross savng, whch seems nconsstent wth the lfe-cycle hypothess on household consumpton. However, the large government transfers lkely rase the ncome and savng of the prvate sector n aged prefectures, and we can probably explan the nconsstency by ths fact. What lessons can we draw from these fndngs on prefectural I-S balances for the future of Japan as a whole? As Fgure 2 showed, Japan s average aged populaton rato wll become smlar to the current level of Akta and Shmane n 15 years. However, Japan as a whole wll not be able to enjoy large net natonal mports and recepts of publc penson and medcal care, whch Akta and Shmane now get. Snce Japan s net foregn assets are equvalent to only 60% of ts GDP, Japan 28