Assessing the Effects of Fiscal Policy in Japan with Estimated and Calibrated DSGE Models

Transcription

1 Bank of Japan Working Paper Series Assessing he Effecs of Fiscal Policy in Japan wih Esimaed and Calibraed DSGE Models Takuji Fueki * Ichiro Fukunaga ** ichirou.fukunaga@boj.or.jp Masashi Saio *** masashi.saiou@boj.or.jp No.11-E-9 Augus 2011 Bank of Japan Nihonbashi-Hongokucho, Chuo-ku, Tokyo , Japan * Research and Saisics (currenly Personnel and Corporae Affairs) Deparmen ** Research and Saisics (currenly Financial Markes) Deparmen *** Research and Saisics (currenly Moneary Affairs) Deparmen Papers in he Bank of Japan Working Paper Series are circulaed in order o simulae discussion and commens. Views expressed are hose of auhors and do no necessarily reflec hose of he Bank. If you have any commen or quesion on he working paper series, please conac each auhor. When making a copy or reproducion of he conen for commercial purposes, please conac he Public Relaions Deparmen (webmaser@info.boj.or.jp) a he Bank in advance o reques permission. When making a copy or reproducion, he source, Bank of Japan Working Paper Series, should explicily be credied.

2 Assessing he Effecs of Fiscal Policy in Japan wih Esimaed and Calibraed DSGE Models Takuji Fueki, Ichiro Fukunaga, and Masashi Saio Research and Saisics Deparmen, Bank of Japan Augus 2011 Absrac In his paper, we assess he effecs of fiscal policy in Japan using wo dynamic sochasic general equilibrium (DSGE) models. One is a medium-scale DSGE model of Japan s economy ( M-JEM, Fueki e al., 2010) esimaed using Bayesian echniques. The oher is he IMF s muli-region GIMF (Global Inegraed Moneary and Fiscal) model (Kumhof e al., 2010) calibraed o daa for Japan and oher counries. The governmen consumpion muliplier calculaed from he former model is larger han ha from he laer, mainly because he negaive effecs of he resuling increase in he ineres rae are larger in he laer model. In boh models, however, he effec of a posiive governmen consumpion shock on real GDP becomes subsanially smaller when he governmen arges a fiscal surplus by raising ax raes. The effeciveness of endogenous adjusmen of he ax policy in response o non-fiscal shocks is no so much differen beween he wo models and is no much affeced by changes in he ineres rae. Keywords: Fiscal Policy; Dynamic Sochasic General Equilibrium Model; Global Inegraed Moneary and Fiscal (GIMF) Model JEL classificaion: E62; E17 The auhors are graeful o Pelin Berkmen, Inernaional Moneary Fund, for her kind suppor in using he GIMF model and for fruiful discussions. They would also like o acknowledge helpful commens from Kosuke Aoki, Michel Juillard, Esuro Shioji, he saff a he Bank of Japan, and he seminar paricipans a he Inernaional Moneary Fund, he Cenral Bank Macroeconomic Modeling Workshop in Manila, he Inernaional Conference on Compuing in Economics and Finance in San Francisco, and he Asian Meeing of he Economeric Sociey in Seoul. Views expressed in his paper are hose of he auhors and do no reflec hose of he Bank of Japan. Corresponding auhor. ichirou.fukunaga@boj.or.jp

3 1 Inroducion Many developed counries recenly launched aggressive fiscal simulus packages o ackle he global financial crisis. Japan is no excepion, bu had already deployed large-scale fiscal simulus long before his crisis. In he 1990s, when Japan s economy experienced a slowdown and a domesic financial crisis, fiscal policy measures were aggressively pursued, resuling in large governmen budge deficis and a rapid increase in public deb. Given he already high level of public deb, he governmen s policy sance on fiscal consolidaion may affec he effeciveness of fiscal policy measures and he impacs of financial crises on he economy. In his paper, we assess he effecs of fiscal policy in Japan using wo dynamic sochasic general equilibrium (DSGE) models. One is a mediumscale DSGE model of Japan s economy (Fueki e al., 2010) esimaed using Bayesian echniques. The oher is he Inernaional Moneary Fund (IMF) s Global Inegraed Moneary and Fiscal (GIMF) model (Kumhof e al., 2010) calibraed o daa for Japan and oher counries. The laer, hough no esimaed, is a muli-region model ha can consider he effecs of foreign shocks and he inernaional spillover of fiscal policy effecs. Many recen sudies have used DSGE models o analyze he effecs of fiscal simulus packages. 1 The main advanage of using srucural models is he large amoun of idenifying informaion ha allows us o race he ransmission mechanisms of various fiscal policy measures. Coenen e al. (2011) compare seven srucural models ha have been regularly used in policymaking insiuions, and show ha here is subsanial agreemen across models on he sizes of fiscal mulipliers and ha he sources of differences across hese mulipliers are fairly sraighforward o idenify. Cogan e al. (2010) compare a sandard New Keynesian DSGE model wih a pracically used old Keynesian model, and show ha he fiscal mulipliers are very differen beween hese models due o differences in underlying heories. Oher sudies examine how he sizes of fiscal mulipliers depend on he fiscal shock 1 Meanwhile, here have also been empirical sudies using srucural vecor auoregression models on his issue, including Blanchard and Peroi (2002). 1

4 processes, he zero lower bound on he nominal ineres rae, and several key feaures of model srucures. 2 Following hese recen sudies, we assess and compare he effecs of fiscal policy calculaed from he wo differen ypes of DSGE models under differen condiions and siuaions. In paricular, aking ino accoun he curren sae of Japan s economy, we examine how he governmen s policy sance on fiscal consolidaion influences he effecs of fiscal policy shocks. Moreover, unlike previous sudies, we also assess he effeciveness of endogenous adjusmens of fiscal insrumens in response o non-fiscal shocks. As for he effecs of exogenous fiscal policy shocks, we firs examine he muliplier in a siuaion wihou a fiscal policy rule ha considers fiscal consolidaion. The governmen consumpion muliplier calculaed from our esimaed model is above one while ha from he GIMF model is below one. The laer is smaller mainly because he increase in he ineres rae caused by he fiscal expansion has a larger negaive effec on domesic privae invesmen and also leads o a decline in expors hrough a real appreciaion of he domesic currency. Meanwhile, in our esimaed model, he esimaed share of non-ricardian households is around 0.3, which conribues o generaing a muliplier above one. We hen assess he effec of a posiive shock o governmen consumpion on real GDP in a siuaion wih an endogenous fiscal policy rule. In boh models, he effec becomes subsanially smaller when he governmen arges a fiscal surplus by raising ax raes. The effec becomes even smaller or negaive when he governmen adjuss he speed of consolidaion depending on business cycles. This is because under his business-cycle sabilizaion policy, he governmen addiionally raises ax raes o dampen he boom caused by is spending increase. The impac of non-fiscal shocks on real GDP also depends on he endoge- 2 For insance, Corsei e al. (2009) show ha governmen spending mulipliers depend on expecaions abou offseing fiscal measures in he fuure. Chrisiano e al. (2009), Eggersson (2011), and Woodford (2011) show ha he governmen spending mulipliers become larger when he moneary policy is consrained by he zero lower bound on he nominal ineres rae. Fujiwara and Ueda (2010) examine mulipliers in an environmen where wo counries are caugh simulaneously in a liquidiy rap. 2

5 nous adjusmen of he governmen s ax policy. We assess he effeciveness of he endogenous ax policy by focusing on he change in real GDP when he governmen considers business-cycle sabilizaion in addiion o argeing a surplus. Under his business-cycle sabilizaion policy, he negaive impac of an adverse non-fiscal shock on real GDP becomes smaller as ax raes are cu more aggressively. The effeciveness of his endogenous adjusmen of ax raes is no very differen beween he wo models because changes in he ineres rae, which cause a subsanial difference in he governmen consumpion muliplier beween he wo models, affec he responsiveness of he endogenous ax raes o non-fiscal shocks raher han he policy effeciveness iself on real GDP. Furher, he GIMF model suggess ha he effeciveness of he endogenous ax policy does no differ much irrespecive of wheher a domesic produciviy shock or a foreign demand shock is considered. The remainder of he paper is organized as follows. Secion 2 shows he resuls from our esimaed model, while Secion 3 shows he resuls from he GIMF model. Secion 4 concludes. 2 Resuls from Esimaed DSGE Model In his secion, we assess he effecs of fiscal policy using a Bayesian-esimaed DSGE model of Japan s economy. We firs provide an overview of he model and hen show he impulse responses o fiscal and non-fiscal shocks. 2.1 Model Overview Our esimaed model is a varian of he Medium-scale Japanese Economic Model (M-JEM), which has been developed a Research and Saisics Deparmen, Bank of Japan (Fueki e al., 2010). The model shares many feaures wih recen New Keynesian DSGE models in he lieraure and hose pracically used in cenral banks, especially he Federal Reserve Board s Esimaed, Dynamic, Opimizaion-based (EDO) model (Chung, Kiley, and Lafore, 2010). The M-JEM is a wo-secor growh model ha akes ino accoun growh rae shocks including invesmen-goods secor specific ech- 3

6 nological progress. The wo-secor producion srucure reflecs he rends in relaive prices and caegories of real expendiure apparen in he Japanese daa. Meanwhile, he M-JEM does no explicily consider foreign counries and assumes ha expor demand follows an exogenous sochasic process. We add a few feaures o he M-JEM for he analyses of fiscal policy. Firs, we inroduce liquidiy-consrained households who do no have access o asse markes and are forced o consume heir afer-ax income in every period. Owing o he exisence of his ype of households ( non-ricardian households) in addiion o households who opimize heir consumpion plans subjec o ineremporal budge consrains, fiscal policies including ransfers have subsanial real effecs. 3 Second, we assume a simple fiscal policy rule o ensure a non-explosive deb-o-oupu raio and o specify he governmen s fiscal policy sances. Following he GIMF model, which we will compare wih his model in he nex secion, we specify he fiscal policy sance as a argeing rule for he surplus-o-oupu raio. 4 The governmen budge consrain relaes he surplus in period o he evoluion of he deb level. B R = B 1 S, (1) where B is he real governmen deb, R is he real ineres rae, and S is he real governmen surplus. The fiscal policy rule is hen specified as S X = s + ω X, (2) where X is real GDP, s is he arge surplus-o-gdp raio, and X is he GDP gap (oupu gap). 5 Apar from surplus argeing, his rule incorporaes a policy-feedback mechanism: seing ω > 0 allows he governmen o run a emporary fiscal defici (or below-arge fiscal surplus) when economic aciviy falls below he normal level (he GDP gap is negaive). Thus, he pa- 3 Non-Ricardian households are inroduced in many DSGE models for he analysis of fiscal policy, as in Galí e al. (2007) and Coenen and Sraub (2005). 4 Various kinds of policy rules are assumed in DSGE models for he analysis of fiscal policy. Leeper e al. (2010) esimaes a DSGE model ha incorporaes various specificaions of fiscal policy rules. 5 The GDP gap is defined as he gap beween he acual and efficien levels of real GDP. 4

7 rameer ω represens he policy sance on business-cycle sabilizaion. 6 Given policy rule (2), he governmen can use various policy insrumens such as axes, ransfers, and spending, o conrol he surplus. In wha follows, however, we assume ha he governmen adjuss only he labor income ax rae o conrol he surplus. All oher model specificaions are essenially he same as in Fueki e al. (2010). The Appendix provides a brief descripion of his version of he M-JEM. In he esimaion of he model, we use quarerly daa for Japan from 1981:Q1 o 2009:Q4. 7 The esimaion resuls are summarized in Table 1. A key parameer ha deermines he effecs of fiscal policy is he share of non-ricardian households. Our poserior mean esimae of his parameer is 0.31, which is generally comparable wih he resuls of previous empirical sudies, including Iwaa (2009) who esimaed a DSGE model using Bayesian echniques. As for he fiscal policy rule, we do no esimae he policy sance on business-cycle sabilizaion, ω in equaion (2), bu calibrae i o zero, following he esimaion resul by Iwaa (2009). 8 Our oher esimaion resuls are generally consisen wih Fueki e al. (2010), in which he non-ricardian households and he fiscal policy rule are no incorporaed Fiscal Policy Effecs Using he M-JEM, we firs assess he effecs of an unanicipaed increase in governmen consumpion. 10 Before considering he policy rule, we check he 6 The parameer ω may also capure he effeciveness of buil-in sabilizers such as he progressive ax srucures. 7 As in Fueki e al. (2010), we esimae he model using daa ha include he period afer he shor-erm nominal ineres rae effecively hi he zero lower bound. However, he esimaed parameer values in he M-JEM on he moneary policy rule remain largely unchanged irrespecive of wheher daa for he full period or daa up o he period jus before he Bank of Japan sared he zero ineres rae policy are used. 8 In Iwaa (2009), he poserior mean esimae of he non-ricardian share is 0.25, while ha of he fiscal policy response o he oupu gap is 7 and no significanly differen from zero, alhough he specificaion of fiscal policy rule is differen from ours. 9 We use he same calibraed parameer values as Fueki e al. (2010). For comparison wih he GIMF model, we also calibrae he parameers for he processes of he produciviy and he governmen consumpion shocks. 10 The M-JEM assumes ha governmen spending is all on consumpion goods. 5

8 radiional muliplier, ha is, he effec on real GDP when governmen consumpion is increased on a susained basis by one percen of nominal GDP above he baseline level and policy rule (2) is absen (he labor income ax rae is held fixed regardless of he level of he surplus-o-gdp raio). 11 Table 2 shows he mulipliers calculaed from various ypes of models. The governmen consumpion muliplier calculaed from he M-JEM is 1.15 in he firs year (he average of he firs four quarers), which is larger han ha calculaed from he large-scale macro-economeric model, Quarerly Japanese Economic Model (Q-JEM), developed a Research and Saisics Deparmen, Bank of Japan (Fukunaga e al., 2011). 12 We hen consider he effecs of he governmen consumpion increase under policy rule (2). Figure 1 shows he responses o an increase in governmen consumpion by one percen of GDP, which follows an AR(1) process and gradually reurns o he iniial level as shown in panel (7). All variables shown in he figure are percen deviaions from heir baseline levels. The hree responses for each variable correspond o he policy ha fixes he labor income ax ( no fiscal rule ) and he wo policy sances wih he fiscal policy rule: surplus argeing rule under which he governmen simply mees he arge surplus-o-gdp raio in each period wihou policy feedback from he GDP gap (ω = 0 in equaion (2)) and sabilizaion rule under which he governmen considers business-cycle sabilizaion (ω > 0) in addiion o surplus argeing. 13 In response o he posiive governmen consumpion shock, he governmen simulaneously raises he labor income ax rae o conrol he surplus following he policy rule, or keeps he ax rae fixed under no fiscal rule, as shown in panel (2). When he governmen considers business-cycle sabilizaion (under he sabilizaion rule), i addiionally raises he ax rae o dampen he boom caused by is spending increase. 11 We assume ha he increase on a susained basis is considered by privae agens as he resul of repeaed unanicipaed shocks. 12 For comparison wih oher models, he mulipliers calculaed from he Q-JEM shown in Table 1 are hose under he assumpion ha oher fiscal policy insrumens (public invesmen and ax raes) are fixed. 13 ω is se o 0.9 in Figure 1 (and he following figures). 6

9 The effecs on real GDP (panel (1)) under he policy rule are much smaller han he effec under no fiscal rule. Under he sabilizaion rule, he addiional ax increase causes real GDP o fall below is baseline level. The labor income ax increase especially dampens he demand for privae consumpion (panel (3)). 14 Privae invesmen (panel (4)) is also dampened bu is sill above he baseline level under he surplus argeing rule. The posiive response of privae invesmen o he governmen consumpion shock is parly due o a modes increase in he policy ineres rae (panel (6)). A he same ime, he increase in inflaion (panel (5)) is also very small. 2.3 Domesic Produciviy Shock Nex we assess he effeciveness of he fiscal policy rule in response o nonfiscal shocks. We focus on a negaive produciviy shock, which may capure he decline in produciviy during Japan s los decade in he 1990s. Alhough here are persisen echnology growh rae shocks and emporary echnology level shocks in he M-JEM, wha we consider here for comparison wih he GIMF model is a persisen echnology level shock. Figure 2 shows he impulse responses o a one-percen negaive economywide produciviy shock ha causes a permanen decrease in he level of oal facor produciviy (TFP) as shown in panel (7). Accordingly, real GDP decreases permanenly from is baseline level. Meanwhile, inflaion slighly acceleraes and he policy ineres rae is raised modesly. The governmen cus he labor income ax rae boh under he surplus argeing and he sabilizaion rules, which booss he demand for privae consumpion and parly offses he negaive impac of he produciviy shock on real GDP. Under he sabilizaion rule, he ax rae cu is larger and he negaive impac on real GDP is smaller han under he surplus argeing rule, excep in he firs hree quarers afer he shock While boh he Ricardian and he non-ricardian households reduce heir consumpion under he policy rule ha raises he labor income ax rae, he reducion in consumpion of non-ricardian households is much larger han ha of Ricardian households. 15 For he firs few quarers afer he shock, he GDP gap emporarily becomes posiive because he efficien level of GDP falls immediaely when he produciviy shifs down. 7

10 We assess he effeciveness of he fiscal policy rule in response o he produciviy shock by focusing on he difference in he response of real GDP under he surplus argeing and he sabilizaion rules. Figure 3 shows wo indicaors ha represen he responsiveness of he labor income ax rae and is effeciveness on real GDP. responsiveness is defined as he difference in he response of he labor income ax rae under he sabilizaion and he surplus argeing rules (panel (2) of Figure 2) divided by he size of he produciviy shock. effeciveness is defined as he difference in he response of GDP under he sabilizaion and he surplus argeing rules (panel (1) of Figure 2) divided by he difference in he response of he labor income ax rae under he wo policy sances. Afer he sabilizaion responsiveness becomes posiive (i.e., he ax rae cu becomes larger under he sabilizaion rule), he sabilizaion effeciveness is sable a around For reference, Figure 3 also shows he above indicaors in he case where he policy ineres rae is consan. Since he response of he policy ineres rae in he M-JEM is modes, he wo indicaors are no much differen beween he variable and consan ineres rae cases. 3 Resuls from he GIMF Model We also use he IMF s large-scale Global Inegraed Moneary and Fiscal (GIMF) model (Kumhof e al., 2010) o assess he effecs of fiscal policy in Japan. Alhough he GIMF model is no esimaed, i has several imporan feaures ha are no considered in our esimaed model (M-JEM) in Secion 2. Mos of all, he GIMF is a muli-region model ha can consider he effecs of foreign shocks and he inernaional spillover of fiscal policy effecs. Moreover, he GIMF considers overlapping generaions households as well as liquidiy consrained households, he financial acceleraor mechanism for he non-financial corporae secor, ec. In his secion, we firs provide a brief overview of he GIMF model and hen show he impulse responses o fiscal and non-fiscal shocks. 16 By definiion, he sabilizaion effeciveness becomes very unsable when he sabilizaion responsiveness is near zero. 8

11 3.1 Model Overview The version of he GIMF model we use in his paper is he 5-block version, consising of he Unied Saes, he Euro Area, Japan, emerging Asia, and oher counries. Many parameers, including hose on fiscal and moneary policy rules, differ across blocks. The deails of he model specificaion and he benchmark calibraion are described in Kumhof e al. (2010). 17 The fiscal policy specificaions of he GIMF are very rich, which is one of is main feaures. However, for simpliciy and for he purpose of comparison, we use he same policy rule (2) as in our esimaed model. 18 Moreover, as in Secion 2, we coninue o assume ha he governmen adjuss he labor income ax rae o arge a surplus following policy rule (2), alhough he GIMF conains a rich se of fiscal policy insrumens, including lump-sum axes and ransfers, redisribuion beween agens, and public invesmen and consumpion spending. On he oher hand, we mainain he GIMF s assumpion ha here are overlapping generaions households as well as liquidiy consrained households, which makes he siuaions of fiscal policy effecs more realisic. The GIMF model we use in his paper is an annual model. In wha follows, we compare he annual impulse responses in he GIMF wih he quarerly impulse responses in he M-JEM shown in Secion Fiscal Policy Effecs Using he GIMF model, we assess he effecs of an unanicipaed increase in governmen spending, as in Secion 2. We begin by looking a he mulipliers in Table 2 again, which shows boh he governmen consumpion muliplier and he governmen invesmen muliplier calculaed from he GIMF. The former is 0.67 in he firs year, which is smaller han he governmen con- 17 For comparison wih he M-JEM, we se differen parameer values from he benchmark calibraion for he fiscal policy rule. We also change he calibraion on elasiciy parameers, inernaional spillover parameers, and moneary policy rule parameers, following a revised version of he GIMF used by he IMF. 18 The governmen budge consrain in he GIMF is B = R 1 B 1 S, which is slighly differen from (1) in he M-JEM. 9

12 sumpion mulipliers calculaed from he M-JEM and he Q-JEM. Meanwhile, he laer is 1.09 in he firs year, which is slighly larger han he governmen invesmen mulipliers calculaed from he Q-JEM and he ESRI s large-scale macro-economeric model (Sakuma e al., 2011). 19 We hen consider he effecs of he governmen consumpion increase under policy rule (2). Figure 4 shows he responses o an increase in governmen consumpion by one percen of GDP, which follows an AR(1) process and gradually reurns o he iniial level as shown in panel (7). In response o his governmen consumpion shock, real GDP increases from he baseline level in he firs year bu hen diminishes more quickly han in he M-JEM. A key difference from he M-JEM is he larger increase in he policy ineres rae (panel (6)). This increase in he policy ineres rae causes he negaive response of privae invesmen (panel (4)), which conrass wih he posiive response in he M-JEM. Moreover, unlike in he M-JEM, he high ineres rae also causes a decline in expors (panel (8)) hrough he real appreciaion of he domesic currency (no shown in he figure). If he policy ineres rae were consan, he response of privae invesmen would be posiive and he decline in expors would be modes. As a resul, he response of real GDP under no fiscal rule would be larger by 24 and 92 percen in he firs and second year, respecively. 20 Under he surplus argeing and he sabilizaion rules, he labor income ax increase dampens he demand for privae consumpion, as in he M-JEM. As a resul, real GDP reurns o he baseline level in he second year. Moreover, using he GIMF model, we examine anoher policy sance labeled sabilizaion & high deb under which he governmen follows he sabilizaion rule bu sars wih a high level of deb ha pus he surpluso-gdp raio below he arge when a shock occurs. Considering he curren fiscal siuaion in Japan, we assume ha he governmen deb-o-gdp raio is 180% and abou 10% above he arge level, 21 which implies ha he 19 Unlike he Q-JEM, he GIMF model akes accoun of he accumulaion of public capial ha serves as a facor of producion. 20 In he M-JEM, he corresponding figures are only 7 and 12 percen. 21 Under oher siuaions, we calibrae boh he iniial and arge levels of he debo-gdp raio o 105%, following a revised version of he GIMF model used by he IMF. 10

13 surplus-o-gdp raio is abou 0.5% below he arge. Under his siuaion, he governmen raises he labor income ax rae by more han under he sabilizaion rule. As a resul, he firs-year real GDP, inflaion, and policy ineres rae do no increase as much as under he sabilizaion rule. 3.3 Domesic Produciviy Shock Nex we assess he effeciveness of he fiscal policy rule in response o nonfiscal shocks. As in Secion 2, we firs focus on a negaive produciviy shock. Figure 5 shows he impulse responses o a one-percen negaive domesic produciviy shock ha causes a persisen decrease in he level of TFP as shown in panel (7). 22 In response o his shock, real GDP, consumpion, and invesmen decrease persisenly from heir baseline levels. Meanwhile, inflaion slighly acceleraes and he policy ineres rae is raised. The high ineres rae causes a persisen decline in expors hrough he real appreciaion of he domesic currency. Under he surplus argeing rule, he governmen persisenly raises he labor income ax rae in order o compensae for he decrease in he ax base. In conras, under he sabilizaion rule, he governmen cus he labor income ax rae, which parly offses he negaive impac of he produciviy shock on real GDP. Therefore, he governmen s policy sance on business-cycle sabilizaion conflics wih surplus argeing. Meanwhile, under he sabilizaion & high deb siuaion, he governmen raises he labor income ax rae so much ha real GDP decreases more han under he oher policy sances. Figure 7 shows he sabilizaion responsiveness and he sabilizaion effeciveness in he GIMF model. Compared wih he M-JEM (Figure 3), he sabilizaion responsiveness is larger and always posiive, while he sabilizaion effeciveness is slighly smaller in absolue value. Compared wih he consan ineres rae case, he sabilizaion responsiveness in he variable Berkmen (2011) analyzes he impac of fiscal consolidaion and srucural reforms in Japan using he GIMF model wih a arge (and seady-sae) deb-o-gdp raio of 87%. 22 As in he M-JEM, we consider an economy-wide produciviy shock, assuming ha he same produciviy shocks occur in boh radable and non-radable secors. 11

14 ineres rae case is larger because he governmen needs o cu he ax rae more under he sabilizaion policy in order o offse he negaive effec of he ineres rae increase on real GDP. Meanwhile, he sabilizaion effeciveness is no so much differen beween he variable and consan ineres rae cases. 3.4 Foreign Demand Shock Using he GIMF model, we can also consider a negaive foreign demand shock, which corresponds o he recen global financial crisis. Figure 6 shows he impulse responses o such a shock, which is a combinaion of he shocks o consumpion and invesmen demand in he Unied Saes and he Euro Area. 23 The shock causes persisen declines in foreign GDP and in Japan s expors as shown in panels (7) and (8), respecively. In response o his shock, domesic real GDP decreases a firs, he inflaion rae falls, and he policy ineres rae is reduced subsanially. Meanwhile, he low ineres rae booss privae invesmen. Then, real GDP reurns o he baseline level four years afer he shock, and inflaion and he policy ineres rae sar increasing. Under he surplus argeing rule, he governmen raises he labor income ax rae a firs in order o compensae for he decrease in he ax base, while under he sabilizaion rule, he governmen cus he labor income ax rae from he firs year of he shock. Under he sabilizaion & high deb siuaion, he governmen raises he labor income ax rae so much ha real GDP, and especially privae consumpion, decrease more han under he oher policy sances. Figure 7 shows he sabilizaion responsiveness and he sabilizaion effeciveness. Compared wih he produciviy shock scenario, he sabilizaion responsiveness is smaller (and urns negaive four years afer he shock), while he sabilizaion effeciveness is similar (excep in he hird year). Compared wih he consan ineres rae case, 24 he sabilizaion responsiveness in he 23 In response o his shock, real GDP in he Unied Saes and he Euro area is assumed o decrease by abou 1.5% and 1%, respecively. 24 The consan ineres rae case could be seen as represening a siuaion where he moneary policy is consrained by he zero lower bound on he nominal ineres rae. Using 12

15 variable ineres rae case is much smaller because he subsanial cu in he policy ineres rae booss he economy and herefore he governmen does no need o cu he ax rae more under he sabilizaion policy. Meanwhile, he sabilizaion effeciveness is very similar (excep in he hird year) beween he variable and consan ineres rae cases. 4 Concluding Remarks In his paper, we assessed he effecs of fiscal policy in Japan using an esimaed and a calibraed DSGE models, he M-JEM and he GIMF. The governmen consumpion muliplier calculaed from he former model is larger han ha from he laer, mainly because he negaive effecs of he resuling increase in he ineres rae are larger in he laer model. In boh models, however, he effec of a posiive governmen consumpion shock on real GDP becomes subsanially smaller when he governmen arges a fiscal surplus by raising ax raes. The effeciveness of endogenous adjusmen of he ax policy in response o non-fiscal shocks is no so much differen beween he wo models and is no much affeced by changes in he ineres rae. Given he curren sae of Japan s economy, i is imporan o ake ino accoun he governmen s policy sance on fiscal consolidaion in assessing he effecs of fiscal policy. In his paper, we only considered a very simple fiscal policy rule wih an endogenous adjusmen of he labor income ax rae. We did no consider he welfare implicaions of he policy rule, he ineracion beween fiscal and moneary policy sances, non-linear dynamics of fiscal consolidaion wih a fiscal limi, ec. Clearly, here is much room for furher invesigaion of endogenous fiscal policy rules, and DSGE models like hose used in his paper would be a useful ool for ha purpose. he Q-JEM, Fukunaga e al. (2011) show ha foreign demand shocks in his case have larger effecs on he domesic economy han usual. 13

16 Appendix: Descripion of he M-JEM This appendix provides a brief descripion of he version of he M-JEM used in his paper. In his version, we inroduce non-ricardian households in addiion o opimizing (Ricardian) households, and a simple fiscal policy rule ha arges a fiscal surplus. All oher specificaions are essenially he same as in Fueki e al. (2010), where a more deailed descripion of he M-JEM and liss of variables and parameers are provided. Final goods producers Final goods producers in he slow-growing secor (secor c) produce he consumpion goods X c, and hose in he fas-growing secor (secor k) produce he invesmen goods X k. They face compeiive markes and produce he final goods, X s, s {c, k}, combining a coninuum of s secor-specific inermediae goods, X s (j), j [0, 1], according o he following Dixi-Sigliz ype echnology. ( 1 X s = X s (j) 0 Θ x,s 1 Θ x,s dj ) Θx,s Θ x,s 1, s = {c, k} where Θ x,s is he sochasic elasiciy of subsiuion beween he differeniaed inermediae goods inpus. Subjec o he above aggregaion echnology, final goods producers in each secor choose he opimal level of each inermediae good o minimize he cos of purchasing hem, aking heir prices as given. Inermediae goods producers Inermediae goods producers in boh secors face a monopolisically compeiive marke and produce secor-specific inermediae goods X s (j), s {c, k} wih he following producion funcion. X s (j) = [K u,s (j)] α [AZ m AZ s L s (j)] 1 α where K u,s (j) and L s (j) respecively are he effecive capial inpu and he labor inpu of firm j. Leing U s (j) be he capial uilizaion rae in secor s, 14

17 he effecive capial inpu is wrien as K u,s (j) K s (j) U s (j). Furher, he labor inpu of a firm j is he coninuum of he differeniaed labor inpu, L s (j) = [ 1 0 Ls (i, j) (Θl 1)/Θl di] Θ l /(Θl 1), where Θ l is he sochasic elasiciy of subsiuion. AZ m is an economy-wide echnology shock and AZ s is a secor-specific echnology shock. We assume ha each of he echnology shocks conains wo separae sochasic componens: one ha is saionary in levels and he oher ha is saionary in growh raes. An inermediae goods producer j in secor s {c, k} maximizes he discouned fuure profi, E 0 =0 β Λ c P c { P s (j)x s (j) MC s (j)x s (j) 100 χp 2 ( ) } P s 2 (j) P 1(j) s ηp Π p,s 1 (1 η p )Π p,s P s X s, subjec o he final goods producers demand schedule, ( ) P X s s Θ x,s (j) = (j) P s X s, aking as given he marginal cos of producion, MC s (j), he aggregae price level for is secor, P s = { 1 [P s 0 (j)] (Θs 1)/Θs dj} Θ s /(Θs 1), and Ricardian households valuaion of a uni of nominal income in each period, Λ c /P c where Λ c is he marginal uiliy of consumpion. The second line in he above discouned fuure profi represens he quadraic price adjusmen cos where Π p,s = P s /P 1 s and Π p,s Capial owners is ime-invarian rend inflaion. Capial sock owners provide capial services o inermediae goods producers in boh secors, receive he renal cos of capial in exchange, and accumulae invesmen goods. Each capial sock owner k chooses invesmen expendiure I (k) and he amoun and uilizaion of capial in boh secors, K c (k), U c(k), Kk (k), and Uk (k), o maximize his discouned profi, E 0 =0 β Λ c [ R c P c U c (k)k(k) c + R k U k (k)k k (k) P k I (k) ], 15

18 subjec o a capial evoluion process wih quadraic invesmen adjusmen cos and he coss from higher uilizaion raes, K+1(k) c + K+1(k) k = (1 δ) ( K c (k) + K k (k) ) + I (k) [ ] 100 χ I (k)a ϕ I 1 (k)γ z,m Γ z,k 2 (K c 2 K (k) + K k (k) ) s=c,kκ [( Z U U s (k) ) ] 1+ψ 1 K s (k), 1 + ψ where A ϕ and Z U are sochasic variaions in he adjusmen cos and he uilizaion cos, respecively. Γ z,m and Γ z,k are he growh-rae-saionary componens of he economy-wide and k secor-specific echnology shocks. Ricardian households Each Ricardian household i [λ, 1], 0 < λ < 1, chooses is purchase of consumpion goods, C R (i), and is holdings of bonds, B (i), o maximize he lifeime uiliy funcion, E 0 β Ξ β {ς c ln ( C R (i) hc R 1(i) ) [( L c ς l (i) + L k (i))] } 1+ν 1 + ν =0 subjec o is budge consrain, 1 B +1 (i) = B (i) + (1 τ L R )W s(i)ls (i) + Ω (i) P ccr (i) s=c,k { } 100 χ w W s 2 (i) 2 W 1(i) s ηw Π w,s 1 (1 η w )Π w,s W s L s s=c,k 100 χl 2 ( L c W L c + L k c + ( L c (i) L k (i) ηllc 1 (1 η l ) Lc L k 1 L k ) Lk W k L c + L k ) 2 L k, where R is he nominal ineres rae on he bonds, τ L is he labor income ax rae, W s (i) and L s (i) are i s wage and labor supply for secor s, and Ω (i) is i s capial and profis income. The fifh erm (in he second line) of he righ hand side is he quadraic wage adjusmen cos imposed on he 16 L c

19 deviaion of he opimum wage growh from pas wage inflaion, Π w,s 1, and from rend wage inflaion, Π w,s. The las erm of he righ hand side is he labor reallocaion cos. Non-Ricardian households Each non-ricardian household i [0, λ] does no opimize and simply chooses is nominal consumpion equal o he afer-ax labor income in every period. P ccn (i) = (1 τ L)W s(i)ls (i) Unions s=c,k Following Galí e al. (2007) and Coenen and Sraub (2005), we assume ha here is a coninuum of monopolisically compeiive unions wihin all households i [0, 1]. Each union ses is wages for boh secors, W c (i) and W k (i), and supply of labor consisen wih each wage, L c (i) and L k (i), given he demand schedule for he differeniaed labor supply, L c (i) = (W c (i)/w c ) Θl L c and L k k (i) = (W (i)/w k) Θl L k, o maximize he above lifeime uiliy funcion, which is assumed o be shared by boh Ricardian and non-ricardian households. Fiscal policy rule As explained in he ex, he fiscal policy rule is specified as follows. S X = s + ω X, where S is he real governmen surplus, X is real GDP, s is he arge surplus-o-gdp raio, and X is he GDP gap (oupu gap). The governmen budge consrain is given by B R = B 1 S, where S = s=c,k(1 τ L ) 1 0 W s (i)l s (i)di P c G. 17

20 Moneary policy rule The moneary policy rule is specified as a Taylor-ype feedback rule wih ineres rae smoohing, as follows. R = (R 1 ) φr ( R ) 1 φr exp(ǫ r ( ) ( ) φ h,gdp Π R p,gdp = R X Π p,gdp Π p,gdp ) φ π,gdp is he inflaion rae of he GDP deflaor implicily defined by Π p,gdp H gdp = where H gdp H gdp = P c X c + P k X k P 1 c Xc 1 + P, 1 k Xk 1 is he growh rae of real GDP calculaed as [ ( X c X c 1 Marke clearing ) P c X c ( X k X k 1 ) P k X k ] 1 P c Xc +P k Xk A he symmeric equilibrium, each marke clears. X c = λ 0 1 C N(i)di + C R(i)di + G λ 100 χw [ + Π w,c χp [ + Π p,c 2 ( L c 100 χl + 2 η w Π w,c 1 (1 η w )Π w. ] 2 W c L c η p Π p,c 1 (1 ] ηp )Π p,c 2 P cxc ) { Lk L c W L c + L k c + W L c + L k k L k where G is governmen consumpion which is assumed o be sochasic. X k = 1 0 I (k)dk + F χw 2 η llc 1 (1 η l ) Lc L k 1 L k [ Π w,k η w Π w,k 1 (1 η w )Π w 100 [ ] χp 2 + Π p,k η p Π p,k 1 (1 η p )Π p,k P k X k, 2 where F is ne expors which is assumed o be sochasic. 1 0 L s (i) = U s (k)k s (k)dk = L s (i, j)dj, i [0, 1], s {c, k} K u,s (j)dj, s {c, k} 18 ] 2 W k L k } 2 L k, L c

21 References [1] Berkmen, S. P. (2011), The Impac of Fiscal Consolidaion and Srucural Reforms on Growh in Japan, IMF Working Paper WP/11/13. [2] Blanchard, O. and R. Peroi (2002), An Empirical Characerizaion of he Dynamic Effecs of Changes in Governmen Spending and Taxes on Oupu, Quarerly Journal of Economics, 117(4), [3] Chrisiano, L., M. Eichenbaum, and S. Rebelo (2009), When is he Governmen Spending Muliplier Large?, NBER Working Paper [4] Chung, H. T., M. T. Kiley, and J.-P. Lafore (2010), Documenaion of he Esimaed, Dynamic, Opimizaion-based (EDO) Model of he U.S. Economy: 2010 Version, Finance and Economics Discussion Series [5] Coenen, G., C. Erceg, C. Freedman, D. Furceri, M. Kumhof, R. Lalonde, D. Laxon, J. Lindé, A. Mourougane, D. Muir, S. Mursula, C. de Resende, J. Robers, W. Roeger, S. Snudden, M. Traband, and J. in Veld (2011), Effecs of Fiscal Simulus in Srucural Models, American Economic Journal: Macroeconomics (forhcoming). [6] Coenen, G. and R. Sraub (2005), Does Governmen Spending Crowd in Privae Consumpion? Theory and Empirical Evidence for he Euro Area, Inernaional Finance, 8(3), [7] Cogan, J. F., T. Cwik, J. B. Taylor, and V. Wieland (2010), New Keynesian versus Old Keynesian Governmen Spending Mulipliers, Journal of Economic Dynamics and Conrol, 34(3), [8] Corsei, G., A. Meier, and G. Müller (2009), Fiscal Simulus wih Spending Reversals, IMF Working Paper WP/09/106. [9] Eggersson, G. B. (2011), Wha Fiscal Policy is Effecive a Zero Ineres Raes?, NBER Macroeconomics Annual 2010, [10] Fueki, T., I. Fukunaga, H. Ichiue, and T. Shiroa (2010), Measuring Poenial Growh wih an Esimaed DSGE Model of Japan s Economy, Bank of Japan Working Paper 10-E-13. [11] Fujiwara, I. and K. Ueda (2010), The Fiscal Muliplier and Spillover in a Global Liquidiy Trap, IMES Discussion Paper Series 2010-E-3, Insiue for Moneary and Economic Sudies (IMES), Bank of Japan. 19

22 [12] Fukunaga, I., N. Hara, S. Kojima, Y. Ueno, and S. Yoneyama (2011), Quarerly Japanese Economic Model (Q-JEM): 2011 Version, Bank of Japan Working Paper (forhcoming). [13] Galí, J., J. D. López-Salido, and J. Vallés (2007), Undersanding he Effecs of Governmen Spending on Consumpion, Journal of he European Economic Associaion, 5(1), [14] Iwaa, Y. (2009) Fiscal Policy in an Esimaed DSGE Model of he Japanese Economy: Do Non-Ricardian Households Explain All?, ESRI Discussion Paper Series No. 216, Economic and Social Research Insiue (ESRI), Cabine Office. [15] Kumhof, M., D. Laxon, D. Muir, and S. Mursula (2010), The Global Inegraed Moneary and Fiscal Model Theoreical Srucure, IMF Working Paper 10/34. [16] Leeper, E. M., M. Plane, and N. Traum (2010), Dynamics of Fiscal Financing in he Unied Saes, Journal of Economerics, 156(2), [17] Sakuma, T., M. Masujima, S. Maeda, K. Fukawa, and K. Iwamoo (2011) The ESRI Shor-Run Macroeconomeric Model of he Japanese Economy (2011 version): Basic Srucure, Mulipliers, and Economic Policy Analyses, ESRI Discussion Paper Series No. 259, Economic and Social Research Insiue (ESRI), Cabine Office. [18] Woodford, M. (2011), Simple Analyics of he Governmen Expendiure Muliplier, American Economic Journal: Macroeconomics, 3(1),

23 φ φ Table 1: Esimaed Parameer Values in he M-JEM Parameers Prior Disribuion Prior Mean Prior S.D. Pos. Mean Pos. Inerval (90%) Lower Upper h : Habi persisence bea ν : Inverse of labor supply elasiciy gamma λ : Non-Ricardian share bea p χ : Size of adjusmen cos in reseing prices gamma η p : Relaive imporance of lagged price inflaion bea w χ : Size of adjusmen cos in reseing wages gamma w η : Relaive imporance of lagged wage inflaion b bea l χ : Size of labor secoral adjusmen cos gamma l η : Relaive imporance of lagged labor supply bea χ : Invesmen adjusmen cos gamma ψ : Elasiciy of uilizaion cos normal φ r φ π,gdp φ : Smoohing parameer in moneary policy rule h,gdp : Coefficien on inflaion in moneary policy rule : Coefficien on he oupu gap in moneary policy rule bea normal normal

24 Table 2: Governmen Spending Muliplier Effecs on real GDP when: Governmen spending is increased by 1% of nominal GDP every period (on a susained basis). Tax raes and ransfers are held fixed. The policy ineres rae follows a Taylor-ype moneary policy rule. M-JEM GIMF Q-JEM ESRI (2011) CG CG IG CG IG IG Firs Second Noe 1: CG refers o he case of a shock o governmen consumpion, while IG refers o he case of a shock o governmen invesmen. Noe 2: Q-JEM refers o "Quarerly Japanese Economic Model" developed a Research and Saisics Deparmen, Bank of Japan. Noe 3: ESRI (2011) refers o he model developed a Economic and Social Research Insiue (ESRI), Cabine Office.

25 Figure 1: Posiive Governmen Consumpion Shock in he M-JEM (1) Real GDP (2) Labor Income Tax Rae Quarer No fiscal rule Surplus argeing Quarer No fiscal rule Surplus argeing (3) Consumpion (4) Invesmen No fiscal rule Surplus argeing Quarer Quarer (5) Inflaion (annual) (6) Policy Ineres Rae (annual) Quarer No fiscal rule Surplus argeing Quarer No fiscal rule Surplus argeing (7) Governmen Consumpion Quarer

26 Figure 2: Negaive Produciviy Shock in he M-JEM (1) Real GDP (2) Labor Income Tax Rae Surplus argeing Quarer Surplus argeing Quarer (3) Consumpion (4) Invesmen Surplus argeing Quarer Quarer (5) Inflaion (annual) (6) Policy Ineres Rae (annual) Quarer Surplus argeing Quarer Surplus argeing (7) TFP Quarer

27 Figure 3: Responsiveness and Effeciveness in he M-JEM (1) Responsiveness Labor Income Tax Rae under Rule Labor Income Tax Rae under Surplus Targeing Rule Size of Shock Produciviy Shock Variable ineres rae Consan ineres rae Quarer (2) Effeciveness GDP under Rule GDP under Surplus Targeing Rule (% deviaion) Labor Income Tax Rae under Rule Labor Income Tax Rae under Surplus Targeing Rule Produciviy Shock Quarer Variable ineres rae Consan ineres rae

28 Figure 4: Posiive Governmen Consumpion Shock in he GIMF (1) Real GDP (2) Labor Income Tax Rae No fiscal rule Surplus argeing & high deb No fiscal rule Surplus argeing & high deb (3) Consumpion (4) Invesmen (5) Inflaion (6) Policy Ineres Rae No fiscal rule Surplus argeing & high deb (7) Governmen Consumpion (8) Expors no feedback

29 Figure 5: Negaive Produciviy Shock in he GIMF (1) Real GDP (2) Labor Income Tax Rae (3) Consumpion (4) Invesmen Surplus argeing & high deb (7) TFP (8) Expors Surplus argeing & high deb Surplus argeing & high deb (5) Inflaion (6) Policy Ineres Rae Non-radables TFP

30 Figure 6: Negaive Foreign Demand Shock in he GIMF (1) Real GDP (2) Labor Income Tax Rae (3) Consumpion (4) Invesmen 0.2 Surplus argeing & high deb Surplus argeing & high deb (5) Inflaion (6) Policy Ineres Rae (7) Foreign GDP (8) Expors Surplus argeing & high deb Surplus argeing & high deb

31 Figure 7: Responsiveness and Effeciveness in he GIMF (1) Responsiveness Labor Income Tax Rae under Rule Labor Income Tax Rae under Surplus Targeing Rule Size of Shock Produciviy Shock Foreign Demand Shock Variable ineres rae Consan ineres rae (2) Effeciveness GDP under Rule GDP under Surplus Targeing Rule (% deviaion) Labor Income Tax Rae under Rule Labor Income Tax Rae under Surplus Targeing Rule Produciviy Shock Foreign Demand Shock Variable ineres rae Consan ineres rae