Macroeconomic Policy in Japan *

Transcription

1 Macroeconomc Polcy n Japan * Warwck J. McKbbn The Australan Natonal Unversty and The Brookngs Insttuton October 6, Paper prepared for the Harvard Unversty Asan Economc Panel Meetng to be held n Seoul, Korea, October 4-5,. Ths paper draws partly on an earler jont paper wth Tm Callen of the IMF. The vews expressed n the paper are those of the author and should not be nterpreted as reflectng the vews of the trustees, offcers or other staff of the Brookngs Insttuton or The Australan Natonal Unversty.

2 . Introducton As Japan enters the twenty frst century mred n a perod of slow economc growth and wth predctons of another recesson, t s worth explorng the possble mpacts of macroeconomc polcy on the Japanese economy. Polcy optons for Japan are becomng ncreasngly mportant both because of the current state of the Japanese economy, but also because of the synchronzed global economc slowdown. A wde range of polces across the economy s lkely to be necessary n comng years. The role of ths paper s to assess the contrbuton that macroeconomc polcy can play n revtalzng the Japanese economy. Ths paper also brefly explores the lkely mpacts of macroeconomc polcy changes n Japan on other countres n the Asa Pacfc regon. Ths paper s a complement to prevous papers that have used the same modelng framework (McKbbn (996) and Callen and McKbbn ()) to explore the lkely causes of current economc problems n Japan. The paper s structured as follows. Secton summarzes the recent macroeconomc experence of Japan from 99 to the present and sets out some stylzed facts. Secton 3 ntroduces the general equlbrum, modelng framework that s used to assess the mplcatons of changes n macroeconomc polcy n Japan. Secton 4 focuses on two ssues related to monetary polcy and three ssues related to fscal polcy. The frst queston regardng monetary polcy s what s the mpact (and the transmsson mechansm) of the Bank of Japan (BoJ) buyng Japanese government bonds (JGBs) n the open market n a one-off monetarzaton of some of the fscal defct, of a magntude suffcent to rase the money supply by 5% forever (relatve to what t otherwse would have been)? Although focusng on government bonds, the actual assets purchased could be foregn exchange or as suggested by Fukao, Hor and Ito () a range of hgh yeldng assets such as real estate etc. The second queston related to monetary polcy s what s the lkely mpact (and channels of transmsson) of a credble announcement of an nflaton target of % n, % n and 3% from 3 forever? The polcy of announcng an nflaton target (and carryng through wth the quanttatve monetary easng) s very dfferent to the partal debt monetzaton, because the nflaton target approach leads to a permanent rse n the rate of nflaton where the debt purchase smulaton s a temporary rse n the rate of nflaton but a permanent rse n the level of prces n Japan. The fscal smulatons explored

3 are: a permanent ncrease n Japanese government spendng fnanced by a permanent ncrease n the fscal defct of 5% of GDP forever; a temporary ncrease n Japanese government spendng fnanced by a rse n the fscal defct of 5% of GDP; and fnally a phased and credble reducton n government spendng of % of GDP n wth cuts rsng lnearly to 5% of GDP by 5 and then mantaned at the new lower level forever. Ths selecton of polces s made to llustrate why the large sustaned fscal packages of the 99s n Japan were counterproductve and to show what a further temporary fscal stmulus mght acheve n the near term as well as showng the mplcaton of a serous fscal consoldaton. The results can be adjusted by the reader to calculate the mpacts of a moderate short term fscal stmulus combned wth a credble future fscal consoldaton.. The Macroeconomc Experence of Japan Snce 99. Economc growth slowed sharply n Japan durng the 99s n the aftermath of the burstng of the asset prce bubble (Fgure ). Snce ther peak asset prces n Japan have fallen dramatcally mpactng both on the supply sde and demand sde n Japan. The ssue of asset prce collapses s dscussed further n Callen and McKbbn (). Wth the economc reform process movng slowly and very lttle done to address the excess capacty bult-up durng the bubble perod, growth n total factor productvty (TFP) slowed sharply. The collapse n TFP could also partly reflect the demand contracton from large negatve wealth effects whch would also show up as excess capacty and a fall n measured TFP. Monetary and fscal polces were eased as polcymakers sought to renvgorate growth durng the 99s. The fscal balance moved from a surplus of close to 3 percent of GDP n 99 to a defct that s estmated to have exceeded 8 percent of GDP n, and publc sector debt has now rsen to over 3 percent of GDP. The yen contnued to apprecate aganst the U.S. dollar n real effectve terms untl the md-99s, but has snce retreated from these peak levels. There s a wde rangng debate on the prmary causes of the growth slowdown. Ths debate s concsely summarzed n Boltho and Corbett (). Some authors put the decade of low growth Hayash and Prescott () argue that the TFP shock was due to legslated changes n the length of the work week n Japan. Ths argument s dffcult to reconcle wth evdence from other countres that experenced smlar declnng work weeks and yet dd not experence a smlar collapse n TFP growth.

4 down to a bankng system collapse (see Bayoum (999)). However Hayash and Prescott () convncngly show that ths s hard to reconcle wth the large growth n nternal fnancng by Japanese frms even whle bank fnancng was fallng sharply. Hayash and Prescott () and Yoshkawa () argue the prmary problem was a sharp fall n productvty over the perod although n contrast to Hayash and Prescott (), Hoshkawa () argues that a sharp fall n demand caused the productvty declne. The nsuffcent demand argument s also made by Krugman (998) where hs focus s on the mpotence of monetary polcy caused by a lqudty trap n Japan. Nomnal nterest rates are unable to fall below zero and wth deflaton, real nterest rates are too hgh for the economy to recover. Wlson () argues that t s hard to separate the lqudty trap from the structural ssues behnd the fall n productvty because the supply ssues mpact on real nterest rate. Yoshno and Sakakbara () argue that the reason for the lack of response to Japanese fscal and monetary polcy s because nvestment s nterest nsenstve and government spendng has been on job creaton wth a very low multpler. Indeed the authors need to appeal to large changes n parameters n ther econometrc model n order to explan the Japanese experence. One possble reason for ths s that the model doesn t capture the role of expectatons formaton that a more structural model (such as n the model used n ths paper) would nclude. In McKbbn and Callen () the model used n the current paper s used to assess the varous arguments summarzed above. One of the key macroeconomc shocks n Japan has been the sgnfcant declne n TFP growth snce 99. Hayash and Prescott dentfed ths as the major explanaton of Japan s poor performance for the decade of the 99s. In McKbbn and Callen () we also fnd ths to be a sgnfcant part of the poor macroeconomc performance of Japan however t s by no means the entre explanaton. It s found that napproprate monetary and fscal polces are also to blame. Ths argument s developed further below. It s also the case that that earler study dentfes a range of shocks as explanatons, whch mples the search for a sngle answer to understandng Japan and a sngle soluton s probably unhelpful. 3. A Modelng Framework for Evaluatng Polcy Optons n Japan

5 One of the mportant aspects of analyzng the Japanese experence s that a general equlbrum framework s essental snce there are a varety of dentfable shocks httng Japan durng the 99s. In lookng forward at possble polcy optons facng Japan, t s mportant to capture the most mportant general equlbrum lnkages n the Japanese economy, ncludng allowng for some sectoral ds-aggregaton, because monetary and fscal polces have dfferental effects on dfferent sectors dependng on a range of factors such as the degree of captal ntensty (wth dfferent nterest rate mpacts), the exposure to foregn trade (wth dfferent exchange rate mpacts). Secondly the model should not be vulnerable to the Lucas crtque n the sense that changes n the polcy regme and expectatons should be modeled explctly so that they do should not affect the structural parameters of the model. The G-Cubed (Asa Pacfc) model, based on the theoretcal structure of the G-Cubed model outlned n McKbbn and Wlcoxen (999), s deal for such analyss havng both a detaled country coverage of the regon and rch lnks between countres through goods and asset markets. A number of studes summarzed n McKbbn and Vnes () show that the G-cubed model (and ts predecessor, the McKbbn-Sachs Global model) have been useful n assessng a range of ssues across a number of countres snce the md-98s. 3 A summary of the model coverage s presented n Table. Some of the prncpal features of the model are as follows: The model s based on explct ntertemporal optmzaton by the agents (consumers and frms) n each economy. In contrast to statc CGE models, tme and dynamcs are of fundamental mportance n the G-Cubed model. In order to track the macro tme seres, however, the behavor of agents s modfed to allow for short run devatons from optmal behavor ether due to myopa or to restrctons on the ablty of households and frms to borrow at the rsk free bond rate on government debt. For both households Full detals of the model ncludng a lst of equatons and parameters can be found onlne at: 3 These ssues nclude: Reagonomcs n the 98s; German Unfcaton n the early 99s; fscal consoldaton n Europe n the md-99s; the formaton of NAFTA; the Asan crss; and the productvty boom n the US.

6 and frms, devatons from ntertemporal optmzng behavor take the form of rules of thumb, whch are consstent wth an optmzng agent that does not update predctons based on new nformaton about future events. These rules of thumb are chosen to generate the same steady state behavor as optmzng agents so that n the long run there s only a sngle ntertemporal optmzng equlbrum of the model. In the short run, actual behavor s assumed to be a weghted average of the optmzng and the rule of thumb assumptons. Thus aggregate consumpton s a weghted average of consumpton based on wealth (current asset valuaton and expected future after tax labor ncome) and consumpton based on current dsposable ncome. Ths s consstent wth the econometrc results n Campbell and Mankw (987) and Hayash (98). Smlarly, aggregate nvestment s a weghted average of nvestment based on Tobn s q (a market valuaton of the expected future change n the margnal product of captal relatve to the cost) and nvestment based on a backward lookng verson of Q. There s an explct treatment of the holdng of fnancal assets, ncludng money. Money s ntroduced nto the model through a restrcton that households requre money to purchase goods. The model also allows for short run nomnal wage rgdty (by dfferent degrees n dfferent countres) and therefore allows for sgnfcant perods of unemployment dependng on the labor market nsttutons n each country. Ths assumpton, when taken together wth the explct role for money, s what gves the model ts macroeconomc characterstcs. (Here agan the model's assumptons dffer from the standard market clearng assumpton n most CGE models.) The model dstngushes between the stckness of physcal captal wthn sectors and wthn countres and the flexblty of fnancal captal, whch mmedately flows to where expected returns are hghest. Ths mportant dstncton leads to a crtcal dfference between the quantty of physcal captal that s avalable at any tme to produce goods and servces, and the valuaton of that captal as a result of decsons about the allocaton of fnancal captal. As a result of ths structure, the G-Cubed model contans rch dynamc behavor, drven on the one hand by asset accumulaton and, on the other by wage adjustment to a neoclasscal steady state. It

7 embodes a wde range of assumptons about ndvdual behavor and emprcal regulartes n a general equlbrum framework. The nterdependences are solved out usng a computer algorthm that solves for the ratonal expectatons equlbrum of the global economy. It s mportant to stress that the term general equlbrum s used to sgnfy that as many nteractons as possble are captured, not that all economes are n a full market clearng equlbrum at each pont n tme. Although t s assumed that market forces eventually drve the world economy to a neoclasscal steady state growth equlbrum, unemployment does emerge for long perods due to wage stckness, to an extent that dffers between countres due to dfferences n labor market nsttutons. Although there are a number of arbtrary parameters such as the shares of forward and backward lookng agents n the model, most parameters are the deep parameters of the utlty and producton technologes and the polcy regmes are modeled as structural equatons. In the followng secton, the model s used to explore the mplcatons for Japan and other countres n the regon, of changes n monetary and fscal polcy n Japan. 4. The Impact of Macroeconomc Polcy n Japan In the remander of ths paper, a number of macroeconomc polcy changes are consdered. They are dvded nto monetary polcy and fscal polcy. In the monetary polcy secton we consder the mpact of an outrght purchase of Japanese government debt by the BOJ suffcent to rase the money supply by 5% as well as the credble announcement of a new nflaton target of % n, % n and 3% n 3. These two polces are qute dfferent n the sense that the frst polcy s a rse n the prce level or a temporary rse n the nflaton rate whereas the second polcy s a permanent rse n the nflaton rate. Some commentators argue that the frst polcy cannot be mplemented due to the techncal problem of the lqudty trap. Ths may be true for purchases of Japanese government bonds, but the results to follow could also be nterpreted as a purchase of foregn currency by the BOJ or ndeed as a purchase of a range of assets n the Japanese economy (see Fukao et al ()). The results n the followng fgures are calculated by usng the model plus projectons of sector specfc productvty growth rates, country specfc populaton growth rates and a range of

8 assumptons about tax rates etc, to generate a projecton of the global economy. Ths baselne projecton s then perturbed by the varous polcy changes consdered. The results shown n each fgure are presented as devaton (ether percent, percentage pont or percent of GDP) relatve to what would have been the outcome. Thus a zero observaton would be a smulaton n whch the value of the varable n a gven year s equal to ts underlyng baselne value. a) Relaxaton of Monetary Polcy ) A Quanttatve Monetary Easng Wth nomnal short-term nterest rates n Japan havng been at, or near zero, for a number of years, debate has focused on whether the Bank of Japan (BoJ) should seek to undertake quanttatve easng, ncludng through ncreased rnban operatons, to provde addtonal lqudty to the economy. Whle there s a debate about the vablty of such a polcy, the G-cubed model s useful n provdng an nsght nto the lkely transmsson mechansms of such a polcy both n Japan and across the regon more broadly. Of course, the numercal results themselves are subject to consderable uncertanty but they gve the man mechansm that such a polcy would lkely produce. To argue that t wll be neffectve requres a focus on the varous steps n the logcal relatonshps between varables as emboded n the model. In the smulaton, the BoJ s assumed to ease monetary polcy by purchasng government bonds suffcent to brng about a permanent 5 percent ncrease n the money supply relatve to the baselne. Fgures a-h show the results for Japan and for other countres. The monetary njecton rases nflaton expectatons n the near term and consequently lowers short-term real nterest rates. Ths declne n real nterest rates temporarly stmulates prvate nvestment and the rse n expected nflaton causes the yen to deprecate, whch temporarly stmulates net exports. Lower real nterest rates and the rse n equty prces also temporarly ncreases consumpton through a postve wealth effect. These temporary ncreases n demand put n moton a multpler process. The result s a temporary rse n real GDP through standard Keynesan channels a demand stmulus accompaned by a fall n real wages and real nterest rates temporarly ncreasng aggregate supply.

9 Over tme, however, prce adjustment removes the real effects of the monetary shock and the economy settles down to the orgnal baselne wth hgher prces, but not hgher nflaton due to the shock beng a rse n the level of money balances (a shock to the rate of growth of money results n a larger stmulus to demand, but also a permanent change n the underlyng nflaton rate n Japan). Long-term nterest rates change very lttle because the nflatonary mpulse s only temporary, whle the change n the real exchange rates that stmulates net exports s largely eroded by the second year. The effects on the rest of Asa are relatvely small. The temporary boost to aggregate demand leads to an ncrease n the demand for Asan goods n Japan, but ths s offset by the rse n the prce of these goods when converted nto yen wthn the Japanese economy. Indeed, n the frst year the exchange rate effect domnates, and exports from each Asan economy to Japan and nto thrd markets n whch they compete wth Japanese goods falls. In the second year, the demand stmulus n Japan has not declned as quckly as the real exchange rate, and therefore Asan exports are hgher than baselne for several more years. Despte the export response beng negatve for growth n Asan economes n the frst year, real GDP s broadly unchanged because Asan equty prces rse antcpatng the growth n perods through 5, whch rases prvate wealth and consumpton suffcently to offset the export declne. ) Adopton of a 3% Inflaton Target A number of economsts ncludng Fukao et al () have advocated the adopton of an explct nflaton target by the BoJ. In ths secton we consder the mpact of announcng an explct and credble nflaton target of % n, % n and 3% from 3 forever. The actual polcy rule to acheve ths s mplemented by specfyng the rate of consumer prce nflaton as the target and the money supply as the nstrument and usng a tme-consstent polcy optmzaton routne to calculate a numercal feedback rule that exactly hts the target each year forever. The results are presented n fgures 3a through 3h. In contrast to the results for the smulaton of the one-off monetzaton presented above, the nflaton target s qute stmulatve. Indeed real GDP rses above baselne by up to.9% by the thrd year of the polcy and s sustaned above baselne for over a decade. The announcement of the nflaton target stmulates the stock market through hgher

10 expected real actvty and through a fall n long term real nterest rates. Ths stmulates nvestment through a rse n Q (fgure 3c) and stmulates consumpton (fgure 3a) through a rse n real wealth. The long term nomnal bond rates jumps by.8% ntally and gradually rses by 3%. The nomnal and real exchange rates deprecate on the announcement and ths stmulates net exports. The Keynesan demand multplers then causes a long ncrease n demand. On the supply sde, the nvestment stmulus rases capacty output (only temporarly) and the rse n nflaton and lagged response of nomnal wages cause a fall n real wages, whch s sustaned for a number of years. Thus both the supply and demand sdes of the economy are stmulated for at least a decade. From ths smulaton, t s also possble to see why a rse n expected deflaton as currently beng experenced n Japan can be depressng for economc actvty. The crtcal problem for the BoJ s how to make the polcy credble snce the credblty s a crtcal part of the effectveness of the polcy. Ths s where the aggressve purchase of a range of assets by the BoJ as advocated by Fukao et al () would be a potentally mportant ngredent n establshng the polcy regme shft. b) Japanese Fscal Polcy One of the key characterstcs of the Japanese experence n the 99s s that despte a large a sustaned fscal stmulus, there seemed to be very lttle response of the real economy. Sakakbara and Yoshno () put ths down to a non-responsveness of nvestment to real nterest rate changes, as well as the specfc nature of the government spendng undertaken whch n ther vew effectvely had lttle multpler value n the economy. However there are other nterpretatons that are consstent wth the observed lnk between fscal polcy and real economc actvty n Japan. In the model used n ths paper the role of expectatons of future taxes and crowdng out of prvate expendture through changes n real exchange rates and real nterest rates and the explanaton. It s not that nvestment s unresponsve to nterest rate (ndeed qute the opposte) but that there are mportant expectatons channels of transmsson that offset the nterest rate effects. To explore ths further a number of fscal polcy smulatons are presented. Before dong so t s worth brefly dscussng what fscal polcy was over ths perod (a detaled dscusson can be found n

11 McKbbn (996)). The nature of the Japanese fscal expanson durng the 99s s open to some nterpretaton. It seems lkely that the ntal ncrease n the defct was vewed as beng temporary n nature, respondng to a perceved cyclcal downturn n the economy, and gven Japan s relatvely strong fscal poston at the begnnng of the decade, the defct may have been vewed as havng lttle mplcaton for future fnancng costs (ths vew s consstent wth the declne n real long-term bond yelds durng the frst half of the 99s). However, as the defct contnued to wden, partcularly over the past three years, t s lkely to ncreasngly have been vewed as a permanent fscal expanson, partcularly n the absence of a credble polcy to brng about medum-term fscal consoldaton. Agan, ths vew s consstent wth the ncrease n long-term bond yelds snce 998. Consequently, two dstnctly dfferent fscal smulatons are consdered n ths secton; a permanent expanson and a temporary expanson. We gnore the ssue of the mx between spendng and tax changes. These are explored n McKbbn and Callen (). ) Permanent Increase n Government Spendng The frst fscal smulaton s a permanent rse n government spendng on goods and servces of 5 percent of GDP forever, fnanced by ssung government debt. Over tme, the fscal closure rule n the model ensures that lump sum taxes on households rse to cover the servcng costs of the addtonal debt ssued. Because of the sectoral dsaggregaton n the model we need to specfy what the government purchases. The addtonal spendng s assumed to be dstrbuted across sectors n the rato:.5 percent of GDP on durable manufacturng sector; percent of GDP on non-durable manufacturng sector; and 3.5 percent of GDP on servces. The results are shown n Fgures 4a to 4e for Japan and Fgures 4f to 4h for the other regons n the model. The mpact of the fscal expanson on the economy s made up of a number of offsettng factors. Frst, the addtonal spendng on goods and servces rases aggregate demand through conventonal Keynesan channels n the short run. As there s some stckness n wages, producer prces rse, real wages fall, and addtonal labor s forthcomng to temporarly satsfy the addtonal demand. However, the fnancal effects of the future fscal defcts are also mportant. Expected future defcts lead to some ncrease n savng and reducton n consumpton by Japanese households n antcpaton of

12 hgher future taxes. However, ths effect s relatvely small and the addtonal resources requred to fnance the defcts wll requre hgher future real nterest rates as the government competes wth the prvate sector for domestc and foregn savngs. The hgher expected future real nterest rates cause long term real nterest rates to rse, and as foregn fnancng (ether the repatraton of Japanese captal from abroad or new foregn captal nflows) s attracted by the hgher real nterest rates the exchange rate also apprecates. These fnancal adjustments hurt equty market confdence and crowd out prvate nvestment and exports. Thus, the short-term stmulus from the hgher government spendng, only lasts for a year and then GDP falls below baselne as the debt burden rses and crowds out economc actvty. From year onwards, growth s roughly.5 percent per year lower for a decade (the growth rate effects can be calculated from the slope of the GDP lne n Fgure 4a). In other countres, the fnancng effects of hgher real nterest rates and the relatve trade relance on Japan determne the transmsson of the shock. In the short run, there s a postve effect on some Asan economes because the temporary stmulus to demand n Japan rases the demand for nputs from these countres. However, the mpact quckly turns negatve (both drectly through hgher real nterest rates and because equty prces n Asa fall affectng prvate consumpton and nvestment). The longer run negatve effects on Asa reflect the long run negatve effects on Japan. It s worth notng that the US, Europe and Chna are the least mpacted over tme, whereas most Asan economes experence a sgnfcant negatve effect from the Japanese fscal polcy. In contrast to much popular dscusson, the model suggests that a fscal expanson whch s permanent n nature only offers a very short run stmulus to the Japanese and regonal economes and proves to be negatve for all economes over tme because t extracts needed savngs from the world economy. ) Temporary Increase In Government Spendng Now consder a temporary fscal stmulus n whch the ncrease n government spendng s the same n the frst year as for the permanent shock, but then returns to baselne n subsequent years. Ths s clearly stylzed but t provdes a useful dagnostc for comparson wth the permanent shock and offers mportant nsghts aout how the Japanese

13 government mght respond to the currently emergng economc crss. Results are presented n fgures 5a to 5h. Gven that the change n government spendng n the frst year s the same for the temporary shock as for the permanent shock, the key dfferences n the results n the frst year are therefore due to the mpacts of the long term fnancng effects of the permanent versus the temporary shocks (.e. n the permanent shock there are expected future fscal defcts whereas n the temporary shock there are none). In Fgure 5a, the rse n GDP s now close to.8 percent relatve to baselne, more than double the.7 percent rse for the permanent fscal expanson. Thus, the mpact on long term nterest rates (Fgure 5c) and the real exchange rate apprecaton (Fgure 5e) are substantally reduced for the temporary fscal stmulus compared to the permanent stmulus. The spllover effects to other countres n the regon are also much more postve n the frst year wth Malaysa and Indonesa experencng GDP ncreases of.6 percent relatve to baselne. These countres both rely on the Japanese market for exports as well as beng mpacted by hgher real nterest rates under the permanent spendng case. ) Phased-In Credble Fscal Consoldaton The new Prme Mnster, Mr. Kozum, has ndcated hs ntenton to move toward fscal consoldaton, commttng to lmt the budget defct to 3 trllon Yen n FY and suggestng that over the medum-term t would be approprate to elmnate the central government s prmary defct n the General Account. At the tme of wrtng ths paper, no tmeframe had been set for ths. In ths smulaton, the mpact of a phased, fully credble, fscal consoldaton s consdered wth government spendng reduced by.7% of GDP n year one, 3.4% of GDP n year two, and 5.% of GDP from

14 year 3 onwards (all relatve to baselne).4 The results are presented n Fgure 6a to 6h. To compare the dfference between phased n fscal cuts and sharp unannounced cuts the reader only needs to nvert the results n fgures 3a to 3g (whch were the permanent fscal spendng ncrease) snce both shocks are scaled to 5% of GDP changes n the long run. The ntal mpact of the announcement of fscal consoldaton s to reduce real nterest rates and deprecate the yen by around 5% whch leads to a rse n net exports. As households antcpate the lower future tax oblgatons they wll face, consumpton also rses. The rse n consumpton and net exports more than offsets the declne n government expendture and n prvate nvestment, whch declnes due to the lower expected growth n the next few years, and real GDP rses. The key pont for growth n the frst year s that the fnancng gans are larger than the fscal cuts. Gven the polcy reacton functon, nterest rates spke upwards n response to the large deprecaton of the yen. If the BoJ dd not rase nterest rates (.e. mposng smulaton n fgures a to h), the ntal output response would be even more postve. However, n the second and thrd years, the postve mpact from the fnancng gans s more than offset by the mpact of the actual declne n government expendture, pushng real GDP below that n the baselne, and t s only from the ffth year that real GDP agan moves above the baselne as the postve mpact of the declne n real nterest rates and the real exchange rate on consumpton, nvestment, and net exports s fully felt. The smulaton, when compared to (nverse) of the temporary fscal expanson consdered n the prevous secton, shows the potental benefts of announcng a fully credble fscal consoldaton strategy as aganst one that s not beleved. Whle even n the case of a credble consoldaton there are short-run costs to output as government demand s wthdrawn from the economy, these are partly mtgated by the postve announcement effect on consumpton and nvestment brought about by the rse n equty prces, declne n long-term nterest rates, and the lower future tax labltes of households. In the case of the temporary consoldaton, none of these offsettng factors are apparent. 4 Whle t s unlkely that any consoldaton would happen ths quckly, for the purposes of the smulatons t s useful to have t occurrng n a relatvely short perod of tme so that the competng effects of the polcy become more clearly vsble.

15 The mpact on the other Asan economes are smlar (but opposte n sgn) to the results dscussed earler for a fscal expanson n Japan. In the frst year, the mpact agan depends on the relatve mportance of trade and fnancal lnks. Those wth mportant trade lnks (such as Malaysa) suffer a modest declne n real GDP as the deprecaton of the yen offsets the rse n demand n Japan, but countres wth hgh debt levels (such as Indonesa) actual see an ncrease n real GDP. In the second year, all the Asan economes are ganng more from lower captal costs than they are losng from a temporary slowdown n Japan and the weaker yen. 6. Conclusons and Polcy Implcatons The results presented n ths paper hghlght a number of mportant ssues n understandng the mpacts of macroeconomc polcy n Japan. Frstly, the proposals for macroeconomc polcy outlned by Fukao et al () and others are strongly supported by the emprcal model presented n ths paper. Secondly as Japan moves toward fscal consoldaton over the medum-term, the results gve some grounds for optmsm that the economc mpact can be lmted. Whle undoubtedly there wll be some short-term negatve mpact on economc actvty from a permanent phased fscal consoldaton, ths would be farly lmted f the announcement were credble--perhaps n the form of an announced and detaled medum term strategy and would be qute quckly replaced by the postve mpact on actvty from the declne n long term real nterest rates and the rse n equty prces. The negatve short-run mpact could also be offset by a more expansonary monetary polcy through the central bank s purchase of government debt or even better, the announcement of a credble nflaton target. As argued n detal n McKbbn and Callen () ths paper also shows that the exstence of fnancal as well as trade lnkages means that the negatve effects of the demand contracton n Japan wll be transmtted less negatvely to the regon, makng any fscal consoldaton n Japan less problematc for the other countres n Asa unless the captal markets lnkages are constraned because of other polcy objectves. The model suggests that a quanttatve easng of monetary polcy by the BoJ through the outrght purchase of government bonds would stmulate the economy n the short-run, and from a

16 poston of nsuffcent demand would help close the output gap. However despte the relatvely optmstc assessment of the mpacts of macroeconomc polcy n Japan, a range of other problems also need to be addressed such as the bank loan problems and structural reforms to the Japanese economy n order to rase productvty growth back towards pre 99 levels. The agng of the Japanese populaton and the mpacts of ths on potental growth and fscal balances5 suggest that Japanese polcymakers need to act sooner rather than later to address the serous economc problems now facng the Japanese economy. 5 See McKbbn and Nguyen () for some results usng a smlar framework to the model used n ths paper.

17 References Bayoum T. (999) The Mornng After: Explanng the Slowdown n Japanese Growth n the 99s IMF Workng Paper WP/99/3. Boltho A. and J. Corbett () The Assessment: Japan s Stagnaton Can Polcy Revve the Economy Oxford Revew of Economc Polcy, vol 6, no, page -7. Callen T. and W. McKbbn () Polces and Prospects n Japan and the Implcatons for the Asa Pacfc Regon IMF Workng Paper WP//3. Fukao, M., Hosh, T. and T. Ito () What Japan Needs: Unconventonal Monetary Polcy and Aggressve NPL Wrte-offs (mmeo) Hayash, F. (98) Tobn s Margnal q and Average q: A Neoclasscal Interpretaton. Econometrca 5, pp.3-4. Hayash F. and E. Prescott () The 99s n Japan: A lost Decade paper presented at the Sprng Internatonal Forum of Collaboraton Projects sponsored by the Economc and Socal Research Insttute of the Japan Cabnet Offce n Tokyo, Japan, March 8-,. Krugman P. (998) Its Baaack: Japan s Slump and the Return of the Lqudty Trap Brookngs Papers on Economc Actvty,, p Lucas, R. E.(967) "Adjustment Costs and the Theory of Supply." Journal of Poltcal Economy 75(4), pp Lucas, Robert (976) Econometrc Polcy Evaluaton: A Crtque n Brunner Karl and Alan Meltzer (eds.) The Phllps Curve and Labour Markets, Amsterdam: North-Holland. McKbbn W.J. (996) The Macroeconomc Experence of Japan 99 to 995: An Emprcal Investgaton Brookngs Dscusson Paper n Internatonal Economcs #3. McKbbn () Documentaton of the G-Cubed Asa Pacfc Model verson 46n McKbbn Software Group Pty Ltd, Canberra, Australa. McKbbn W and T. Callen () The Impact of Japan on the Asa Pacfc Regon paper presented at the conference on Economc Interdependence Shapng Asa-Pacfc n the st Century held n Tokyo durng March -3.

18 McKbbn W. and W. Martn (998) The East Asa Crss: Investgatng Causes and Polcy Responses Workng Paper n Trade and Development #98/6, Economcs Department, Research School of Pacfc and Asan Studes, Australan Natonal Unversty and Brookngs Dscusson Paper n Internatonal Economcs #4, The Brookngs Insttuton, Washngton DC. McKbbn W. and J. Nguyen () The Impact of Demographc Change n Japan: Some Prelmnary Results from the MSG3 Model paper presented to the Fall Internatonal Forum of Collaboraton Projects, Economc and Socal Research Insttute of the Japanese Cabnet Offce, Tokyo Sept 7-9. McKbbn W. and A. Stoeckel (999) East Asa s Response to the Crss: A Quanttatve Analyss, paper prepared for the ASEM Regonal Economst s Workshop: From Recovery to Sustanable Development The World Bank. Brookngs Dscusson Paper n Internatonal Economcs #49, The Brookngs Insttuton, Washngton DC. McKbbn W.J. and J. Sachs (99) Global Lnkages: Macroeconomc Interdependence and Cooperaton n the World Economy, Brookngs Insttuton, June. McKbbn W.J. and D. Vnes () Modellng Realty: The Need for Both Intertemporal Optmzaton and Stckness n Models for Polcymakng Oxford Revew of Economc Polcy vol 6, no 4. (draft verson ) McKbbn W. and P. Wlcoxen (998) The Theoretcal and Emprcal Structure of the G- Cubed Model Economc Modellng, 6,, pp 3-48 (ISSN ) (workng paper verson: ) Motonsh T and H. Yoshkawa (999) Causes of the Long Stagnaton of Japan durng the 99s: Fnancal or Real Journal of the Japanese and Internatonal Economes, 3(4), pp 8-. Posen A. (998) Restorng Japan s Economc Growth, Insttute for Internatonal Economcs, Washngton DC. Sakakbara, E. and N. Yoshno () Understandng Recent Slow Growth n Japan: The Way Forward paper presented at the Harvard Unversty Asan Economc Panel, Cambrdge Mass, Aprl 6-7. Treadway, A. (969) "On Ratonal Entrepreneural Behavor and the Demand for Investment." Revew of Economc Studes 36(6), pp Yoshkawa H. () Techncal Progress and the Growth of the Japanese Economy Past

19 and Future Oxford Revew of Economc Polcy, vol 6, no, page Wlson D. (() Japan s Slowdown: Monetary versus Real Explanatons Oxford Revew of Economc Polcy, vol 6, no, page 8-33.

20 9 Appendx I: A stylzed country G-Cubed model In ths secton a stylzed country model s presented whch dstlls the essence of the G- Cubed model and n partcular how the ntertemporal aspects of the model are handled. The reader s referred to chapters and 5 of McKbbn and Wlcoxen () for greater detal. In ths stylzed model there are symmetrc countres (based essentally on US data adjusted to create symmetry). Each country conssts of several economc agents: households, the government, the fnancal sector and frms, one each n the producton sectors. The two sectors of producton are energy and non-energy (ths s much lke the aggregate structure of the MSG model). The followng gves an overvew of the theoretcal structure of the model by descrbng the decsons facng these agents n one of these countres. Throughout the dscusson all quantty varables wll be normalzed by the economy's endowment of effectve labor unts. Thus, the model's long run steady state wll represent an economy n a balanced growth equlbrum. Frms We assume that each of the two sectors can be represented by a prce-takng frm whch chooses varable nputs and ts level of nvestment n order to maxmze ts stock market value. Each frm s producton technology s represented by a constant elastcty of substtuton (CES) functon. Output s a functon of captal, labor, energy and materals: o σ o o /( σ ) o ( σ o ) / σ () Q ( ) = A δ j xj j=k,l,e,m o σ o / where Q s the output of ndustry, x j s ndustry 's use of nput j, and A o, o o δ j, and σ are parameters. A o reflects the level of technology, σ o s the elastcty of substtuton, and the o δ j parameters reflect the weghts of dfferent nputs n producton; the superscrpt o ndcates that the parameters apply to the top, or output, ter. Wthout loss of generalty, we constran theδ j o 's to sum to one.

21 The goods and servces purchased by frms are, n turn, aggregates of mported and domestc commodtes whch are taken to be mperfect substtutes. We assume that all agents n the economy have dentcal preferences over foregn and domestc varetes of each commodty. We represent these preferences by defnng composte commodtes that are produced from mported and domestc goods. Each of these commodtes, Y, s a CES functon of nputs domestc output, Q, and an aggregate of goods mported from all of the country s tradng partners, M : fd / fd fd fd ( ) / () fd σ ( ) + ( ) σ σ fd y = A δ Q δ d f fd / σ fd fd ( σ ) / σ M fd fd σ /( ) σ where σ fd s the elastcty of substtuton between domestc and foregn goods. For example, the energy product purchased by agents n the model are a composte of mported and domestc energy. The aggregate mported good, M, s tself a CES composte of mports from ndvdual countres, M c, where c s an ndex ndcatng the country of orgn: c= 7 ff (3) ff M = A ( δ ) c ff / σ ff fd ( σ ) / σ M c σ ff ff /( σ ) The elastcty of substtuton between mports from dfferent countres s σ ff. By constranng all agents n the model to have the same preferences over the orgn of goods we requre that, for example, the agrcultural and servce sectors have the dentcal preferences over domestc ol and mported ol. Ths accords wth the nput-output data we use and allows a very convenent nestng of producton, nvestment and consumpton decsons. In each sector the captal stock changes accordng to the rate of fxed captal formaton (J ) and the rate of geometrc deprecaton (/ ): (4) k = J δ k Ths approach follows Armngton (969). Ths does not requre that both sectors purchase the same amount of ol, or even that they purchase ol at all; only that they both feel the same way about the orgns of ol they buy.

22 Followng the cost of adjustment models of Lucas (967), Treadway (969) and Uzawa (969) we assume that the nvestment process s subject to rsng margnal costs of nstallaton. To formalze ths we adopt Uzawa's approach by assumng that n order to nstall J unts of captal a frm must buy a larger quantty, I, that depends on ts rate of nvestment (J/k): J (5) I= + J k φ where φ s a non-negatve parameter. The dfference between J and I may be nterpreted varous ways; we wll vew t as nstallaton servces provded by the captal-goods vendor. Dfferences n the sector-specfcty of captal n dfferent ndustres wll lead to dfferences n the value of φ. The goal of each frm s to choose ts nvestment and nputs of labor, materals and energy to maxmze ntertemporal net-of-tax profts. For analytcal tractablty, we assume that ths problem s determnstc (equvalently, the frm could be assumed to beleve ts estmates of future varables wth subjectve certanty). Thus, the frm wll maxmze:3 I (6) ( p I R s n) s t ( 4) ( ( ) ( ) ) π τ e ds t where all varables are mplctly subscrpted by tme. The frm s profts, π, are gven by: * e m (7) π = ( τ )( p Q w xl p xe p xm) ZKHUH LV WKH FRUSRUDWH LQFRPH WD[ 4 s an nvestment tax credt, and p * s the producer prce of the frm s output. R(s) s the long-term nterest rate between perods t and s: (8) R( s) = r( v) dv s t Because all real varables are normalzed by the economy's endowment of effectve labor unts, profts are dscounted adjustng for the rate of growth of populaton plus productvty growth, n. s t 3 The rate of growth of the economy's endowment of effectve labor unts, n, appears n the dscount factor because the quantty and value varables n the model have been scaled by the number of effectve labor unts. These varables must be multpled by exp(nt) to convert them back to ther orgnal form.

23 Solvng the top ter optmzaton problem gves the followng equatons characterzng the frm s behavor: (9) ( ) m} {l,e, j p p Q A x o o j o o j j = σ σ δ * () p k J = I 4 ) )( ( τ φ λ + () * ) ( ) ( ) ( I 4 k J p dk dq p r + = ds d φ τ τ λ δ λ where s the shadow value of an addtonal unt of nvestment n ndustry. Equaton (9) gves the frm s factor demands for labor, energy and materals and equatons () and () descrbe the optmal evoluton of the captal stock. Integratng () along the optmum trajectory of nvestment and captal accumulaton, )) ˆ( ), ˆ( ( t k t J, gves the followng expresson for : () + + = t t s s R I k J ds e k J p dk dq p t ) )( ) ( ( 4 ˆ ˆ, * ˆ ˆ ) ( ) ( ) ( δ φ τ τ λ Thus, λ s equal to the present value of the after-tax margnal product of captal n producton (the frst term n the ntegral) plus the savngs n subsequent adjustment costs t generates. It s related to q, the after-tax margnal verson of Tobn's Q (Abel, 979), as follows: (3) ( )p = q I τ 4 λ Thus we can rewrte () as: (4) ( ) q = k J φ

24 3 Insertng ths nto (5) gves total purchases of new captal goods: = φ (5) I ( q ) k Based on Hayash (979), who showed that actual nvestment seems to be party drven by cash flows, we modfy (5) by wrtng I as a functon not only of q, but also of the frm's current cash flow at tme t, π, adjusted for the nvestment tax credt: φ (6) I = α ( q ) k + ( α) ( ) I π τ4 Ths mproves the model s ablty to mmc hstorcal data and s consstent wth the exstence of frms that are unable to borrow and therefore nvest purely out of retaned earnngs. p So far we have descrbed the demand for nvestment goods by each sector. Investment goods are suppled, n turn, by a thrd ndustry that combnes labor and the outputs of other ndustres to produce raw captal goods. We assume that ths frm faces an optmzaton problem dentcal to those of the other two ndustres: t has a nested CES producton functon, uses nputs of captal, labor, energy and materals n the top ter, ncurs adjustment costs when changng ts captal stock, and earns zero profts. The key dfference between t and the other sectors s that we use the nvestment column of the nput-output table to estmate ts producton parameters. Households Households have three dstnct actvtes n the model: they supply labor, they save, and they consume goods and servces. Wthn each regon we assume household behavor can be modeled by a representatve agent wth an ntertemporal utlty functon of the form: (7) = c s + g s - s-t U t (ln ( ) ln ( )) θ ( ) e ds t where c(s) s the household's aggregate consumpton of goods and servces at tme s, g(s) s government consumpton at s ZKLFK ZH WDNH WR EH D PHDVXUH RI SXEOLF JRRGV SURYLGHG DQG LV WKH

25 4 rate of tme preference.4 The household maxmzes (7) subject to the constrant that the present value of consumpton be equal to the sum of human wealth, H, and ntal fnancal assets, F:5 c ( R( s) n)( s t) (8) p ( s) c( s) e = H t + Ft t Human wealth s defned as the expected present value of the future stream of after-tax labor ncome plus transfers: G C I (9) = - W L + L + L + L TR ( R( s) n) ( s t) H t ( τ )( ( ) + ) e ds t where τ s the tax rate on labor ncome, TR s the level of government transfers, L C s the quantty of labor used drectly n fnal consumpton, L I s labor used n producng the nvestment good, L G s government employment, and L s employment n sector. Fnancal wealth s the sum of real money balances, MON/P, real government bonds n the hand of the publc, B, net holdng of clams aganst foregn resdents, A, the value of captal n each sector: MON I I c c () F = + B + A+ q k + q k + q k p Solvng ths maxmzaton problem gves the famlar result that aggregate consumpton spendng s equal to a constant proporton of prvate wealth, where prvate wealth s defned as fnancal wealth plus human wealth: () p c c= θ ( F + H ) However, based on the evdence cted by Campbell and Mankw (99) and Hayash (98) we assume some consumers are lqudty-frqvwudlqhg DQG FRQVXPH D IL[HG IUDFWLRQ RI WKHLU DIWHU-tax = = 4 Ths specfcaton mposes the restrcton that household decsons on the allocatons of expendture among dfferent goods at dfferent ponts n tme be separable. 5 As before, n appears n (8) because the model's scaled varables must be converted back to ther orgnal bass.

26 5 ncome (INC).6 Denotng the share of consumers who are not constraned and choose consumpton n accordance wth () by α 8, total consumpton expendture s gven by: c () p c= α 8θ ( Ft + H t) + ( α8) γ INC The share of households consumng a fxed fracton of ther ncome could also be nterpreted as permanent ncome behavor n whch household expectatons about ncome are myopc. Once the level of overall consumpton has been determned, spendng s allocated among goods and servces accordng to a CES utlty functon. 7 The demand equatons for captal, labor, energy and materals can be shown to be: c σc p, p c c (3) p x = δ y, { k, l, e m} o where y s total expendture, x c s household demand for good, σ o c s the top-ter elastcty of substtuton and the δ c are the nput-specfc parameters of the utlty functon. The prce ndex for consumpton, p c, s gven by: (4) c p = j = k, l, e, m δ o o c σ c σ c j p j Household captal servces consst of the servce flows of consumer durables plus resdental housng. The supply of household captal servces s determned by consumers themselves who nvest n household captal, k c, n order to generate a desred flow of captal servces, c k, accordng to the followng producton functon: 6 There has been consderable debate about the emprcal valdty of the permanent ncome hypothess. In addton the work of Campbell, Mankw and Hayash, other key papers nclude Hall (978), and Flavn (98). One sde effect of ths specfcaton s that t prevents us from computng equvalent varaton. Snce the behavor of some of the households s nconsstent wth (), ether because the households are at corner solutons or for some other reason, aggregate behavor s nconsstent wth the expendture functon derved from our utlty functon. 7 The use of the CES functon has the undesrable effect of mposng untary ncome elastctes, a restrcton usually rejected by data. An alternatve would be to replace ths specfcaton wth one derved from the lnear expendture system.

27 6 (5) k k c = c α where α s a constant. Accumulaton of household captal s subject to the condton: (6) c c k = J c c δ k We assume that changng the household captal stock s subject to adjustment costs so household spendng on nvestment, I c, s related to J c by: c c (7) = + J c c I φ J c k Thus the household's nvestment decson s to choose I C to maxmze: ck c I c ( R( s) n)( s t) (8) ( p α k p I ) e ds t where p ck s the mputed rental prce of household captal. Ths problem s nearly dentcal to the nvestment problem faced by frms and the results are very smlar. The only mportant dfferences are that no varable factors are used n producng household captal servces and there s no nvestment tax credt for household captal. Gven these dfferences, the margnal value of a unt of household captal, λ C, can be shown to be: ck I φ + (9) = + c Jˆ c ( R( s) δ )( s t) λ c ( t) p α p e ds t kˆ c where the ntegraton s done along the optmal path of nvestment and captal accumulaton, ( Jˆ ( ), ˆ c t kc ( t)). Margnal q s: (3) λ c qc= I p and nvestment s gven by:

28 Jc (3) = ( q ) kc φ c c 7 The Labor Market We assume that labor s perfectly moble among sectors wthn each regon but s mmoble between regons. Thus, wages wll be equal across sectors wthn each regon, but wll generally not be equal between regons. In the long run, labor supply s completely nelastc and s determned by the exogenous rate of populaton growth. Long run wages adjust to move each regon to full employment. In the short run, however, nomnal wages are assumed to adjust slowly accordng to an overlappng contracts model where wages are set based on current and expected nflaton and on labor demand relatve to labor supply. The equaton below shows how wages n the next perod depend on current wages; the current, lagged and expected values of the consumer prce level; and the rato of current employment to full employment: (3) α5 α5 c α6 p c t+ pt Lt wt + = wt c c p p L t t The weght that wage contracts attach to expected changes n the prce level s α 5 whle the weght assgned to departures from full employment ( L ) s α 6. Equaton (3) can lead to short-run unemployment f unexpected shocks cause the real wage to be too hgh to clear the labor market. At the same tme, employment can temporarly exceed ts long run level f unexpected events cause the real wage to be below ts long run equlbrum. The Government We take each regon's real government spendng on goods and servces to be exogenous and assume that t s allocated among nputs n fxed proportons, whch we set to 996 values. Total government outlays nclude purchases of goods and servces plus nterest payments on government debt, nvestment tax credts and transfers to households. Government revenue comes from sales taxes, corporate and personal ncome taxes, and from sales of new government bonds. In addton,