The Macroeconomic Effects of Fiscal Policy in Portugal: a Bayesian SVAR Analysis *

Transcription

1 The Macroeconomic Effecs of Fiscal Policy in Porugal: a Bayesian SVAR Analysis * Anónio Afonso # and Ricardo M. Sousa $ December 2010 Absrac Wih a new quarerly daase we esimae a Bayesian Srucural Auoregression model and a Fully Simulaneous Sysem approach o analyze he macroeconomic effecs of fiscal policy. Resuls show ha posiive governmen spending shocks, in general, have a negaive effec on real GDP; lead o crowding-ou effecs of privae consumpion and invesmen; have a persisen and posiive effec on he price level and a mixed impac on he average financing cos of governmen deb. Explicily considering he governmen deb dynamics in he model is also imporan. A VAR couner-facual exercise confirms ha unexpeced posiive spending shocks creae relevan crowdingou effecs. Keywords: B-SVAR, Fully Simulaneous Sysem, fiscal policy, deb dynamics, Porugal. JEL classificaion: E37, E62, H62, G10. * We are graeful o Ad van Rie, o paricipans o he 4 h Meeing of he Poruguese Economic Journal, and o an anonymous referee for helpful commens. The opinions expressed herein are hose of he auhors and do no necessarily reflec hose of he ECB or he Eurosysem. # European Cenral Bank, Direcorae General Economics, Kaisersraße 29, D Frankfur am Main, Germany. ISEG/TULisbon - Technical Universiy of Lisbon, Deparmen of Economics; UECE Research Uni on Complexiy and Economics; R. Miguel Lupi 20, Lisbon, Porugal. UECE is suppored by FCT (Fundação para a Ciência e a Tecnologia, Porugal), financed by ERDF and Poruguese funds. s: anonio.afonso@ecb.europa.eu, aafonso@iseg.ul.p. $ Universiy of Minho, Deparmen of Economics and Economic Policies Research Uni (NIPE), Campus of Gualar, Braga, Porugal; London School of Economics, Financial Markes Group (FMG), Houghon Sree, London WC2 2AE, Unied Kingdom. s: rjsousa@eeg.uminho.p; rjsousa@alumni.lse.ac.uk. 1

2 1. Inroducion In he las weny years, public spending conrol has been a major problem in Porugal. The gains from he drop in ineres raes and, consequenly, in he ineres paymens on he ousanding governmen deb were no accompanied by a susained consolidaion of public finances. Moreover, he episodes of fiscal improvemen ha occurred in he 1980s and in he 1990s have been shor-ermed and mosly no successful. Following he inroducion of he Sabiliy and Growh Pac (SGP), Porugal was he firs counry in he Economic and Moneary Union o breach he 3% of GDP reference value for he governmen defici in Consequenly, i became subjec o he Excessive Defici Procedure (EDP) in 2002, a siuaion ha occurred again in Therefore, given pas performance and oucomes, i seems fair o say ha afer enering he European Union (EU) in 1986, joining he Exchange Rae Mechanism (ERM) of he European Moneary Sysem (EMS) in 1992 and enering EMU in January 1999, Porugal s fiscal rack record could have been beer. In his conex, he evaluaion of he effecs of fiscal policy on economic aciviy in Porugal becomes relevan and is he major goal of his paper. Addiionally, we look a is impac on he composiion of GDP, herefore, analyzing poenial crowding-ou effecs on privae consumpion and privae invesmen. Fiscal policy shocks are idenified using a recursive parial idenificaion scheme 1 and we assess he poserior uncerainy of he impulse-response funcions by esimaing a Bayesian Srucural Vecor Auoregression (B-SVAR) model. We also accoun for he auomaic response of fiscal policy o he economic aciviy, and use a Fully Simulaneous Sysem approach in line wih he works of Blanchard and Peroi (2002), Leeper and Zha (2003) and Sims and Zha (1999, 2006). In addiion, we consider 1 Chrisiano e al. (2005) idenify he moneary policy shock using he same procedure. 2

3 he response of fiscal variables o he level of he governmen deb following Favero and Giavazzi (2007) and Afonso and Sousa (2009a). Anoher imporan conribuion of he paper is he use of a se of quarerly fiscal daa, which we build by drawing on he higher frequency (monhly) availabiliy of fiscal cash daa. This allows us o idenify more precisely he effecs of fiscal policy. The findings of his paper can be summarized as follows. On he one hand, governmen spending shocks: (i) have a negaive effec on real GDP; (ii) generae subsanial crowding-ou effecs and lead o a fall in boh privae consumpion and privae invesmen; (iii) have a persisen and posiive impac on he price level; and (iv) have mixed impacs on he average cos of refinancing he deb. Therefore, and from a policymaking perspecive, increasing governmen spending does no emerge as an obvious insrumen o help fosering economic aciviy. On he oher hand, governmen revenue shocks: (i) have a negaive impac on GDP; (ii) crowd-ou privae consumpion and privae invesmen, alhough he response emerges wih a lag of abou four quarers; and (iii) is normally followed by a somewha less disciplined fiscal policy. The consideraion of he feedback from governmen deb makes he effecs of fiscal policy on (long-erm) ineres raes and GDP more persisen and hese variables are also more responsive o he shocks. Moreover, he resuls do no seem o suppor he exisence of a significan sabilizing response of he budge balance o he deb level. In fac, here is only weak evidence suggesing ha: (i) governmen spending falls when he deb-o-gdp raio is above is mean (in paricular, in he period 1979:1-1993:3); and (ii) governmen revenue increases when he deb-o-gdp raio is above is mean (namely, in he period 1993:4-2007:4, ha is, afer he Maasrich Treay). Therefore, here was a possible Ricardian behaviour afer he beginning of he 1990s, alhough he 3

4 pas fiscal sabilizaion aemps have been mosly biased owards increases in governmen revenue, wihou ackling sufficienly he spending side. Moreover, and as we will see below, he breaching of he SGP by Porugal in 2002 canno be disconneced from such pas fiscal responses. Finally, a VAR counerfacual exercise shows ha unexpeced increases in governmen spending lead o imporan crowdingou effecs. The res of he paper is organized as follows. Secion wo reviews he relaed lieraure. Secion hree presens fiscal developmens in Porugal. Secion four explains he empirical sraegies used o idenify he effecs of fiscal policy shocks. Secion five describes he daa and provides he empirical analysis. Secion six concludes. 2. Lieraure Review Despie he large lieraure on he impac of moneary policy on economic aciviy, he imporance of fiscal policy for economic sabilizaion has received less aenion. This secion provides a brief review of he exising evidence of he effecs of fiscal policy on GDP, he aggregae price level and he composiion of oupu, ha is, privae consumpion and privae invesmen. For he U.S., differen approaches have been used in he idenificaion of he fiscal policy shock. The narraive approach developed by Ramey and Shapiro (1998) isolaes poliical evens and finds ha, afer a brief rise in governmen spending, durables consumpion falls while nondurable consumpion displays a small decline. Faás and Mihov (2001) use a Cholesky ordering and show ha increases in governmen expendiures are expansionary, bu lead o imporan changes in he composiion of oupu in he form of an increase in privae invesmen ha more han compensaes for he fall in privae consumpion. Blanchard and Peroi (2002) idenify he auomaic 4

5 response of fiscal policy by using informaion abou he elasiciy of fiscal variables, and find ha fiscal shocks are expansionary, have a posiive effec on privae consumpion, and a negaive impac on privae invesmen. More recenly, Mounford and Uhlig (2009) relying on sign resricions for he fiscal impulse-response funcions find a negaive effec in residenial and non-residenial invesmen for he U.S.. A he inernaional level, he evidence is scarce due o he limied availabiliy of quarerly public finance daa. Peroi (2004) finds ha fiscal policy leads o no response of privae invesmen and a relaively large and posiive effec on privae consumpion in a se of five counries (Ausralia, Canada, Germany, he U.S. and he U.K.). For France, Biau and Girard (2005) find a posiive effec on boh privae consumpion and privae invesmen. For Spain, De Casro and Hernández de Cos (2008) show ha a posiive spending shock lead o higher inflaion and lower oupu in he medium and long-erm, bu can be expansionary in he shor-erm. Heppke-Falk e al. (2006) and Giordano e al. (2007) find ha governmen spending has expansionary effecs on boh oupu and privae consumpion for, respecively, Germany and Ialy. Afonso and Sousa (2009a, 2009b) show ha, for he U.S., he U.K., Germany and Ialy, quarerly fiscal policy shocks have imporan macroeconomic effecs while also impacing on housing and sock prices. In addiion, Burriel e al. (2010), using a quarerly sandard SVAR, repor ha expendiure shocks are more persisen in he US han in he euro area, while he negaive response from ne ax increases is shorer lived in he euro area. As can be inferred from he abovemenioned sudies, daa availabiliy in paricular, high frequency daa, remains a major drawback in he lieraure on fiscal policy. Therefore, we ry o overcome his issue, by building also a fiscal quarerly daase for Porugal. 5

6 3. Recen Fiscal Developmens in Porugal According o Afonso and Claeys (2008), he main reasons for he hree iniial breaches of he SGP in 2002 and 2003 were expendiure rises in France and Porugal, while large revenue reducions unmached by expendiure cus in Germany pushed he budge defici beyond 3 per cen of GDP. Therefore, his evidence, poins o some difficulies relaed o expendiure conrol in Porugal. In 2003 and 2004, Porugal used sizable emporary measures, amouning respecively o 2.5 and 2.3 per cen of GDP, in order o keep he budge defici below he 3 per cen limi. Overall, such emporary measures implemened in he period, added up o 6.2 per cen of GDP. 2 The adopion of emporary correcive measures, even if i prevened he budge defici from going above he 3 per cen limi, did no address he srucural facors underlying Poruguese fiscal imbalances. Addiionally, afer he 2002 EDP, he consolidaion sraegy also included an increase in he sandard VAT rae (from 17 o 19 per cen, which was again raised o 21 per cen in July 2005, bu cu back o 20 per cen in July 2009), while primary spending coninued rising. 3 A second EDP was sared in June 2005 and was abrogaed in June Among he several measures hen proposed and implemened by he auhoriies o conrol primary spending, one can menion as some of he more srucurally oriened ones, for insance, he revision of he civil servans pension schemes, and he reform of he healh care secor. 4 Ineresingly, boh EDP episodes ha occurred in Porugal were characerised by fiscal easing, bu while he 2001 episode was coupled wih more 2 For insance, pension funds ransfers from public secor enerprises o he civil servans pension sysem in 2004; securiizaion of ax credis in 2003; and ax amnesy in Guichard and Leibfriz (2006) also survey he reasons for he non-successful fiscal adjusmen in he period Since 2002, some public hospials have been ransformed ino public corporaions, o increase efficiency and decrease coss in he Naional Healh Service (NHS). However, Afonso and Fernandes 6

7 favourable moneary condiions, in he 2005 episode moneary condiions were more sringen (see Figure 1). [Figure 1] Afer enering he EU in 1986, boh inflaion and ineres raes in Porugal decreased seadily and converged owards he lower levels ha were more common in oher counries already in he EU. This was an obvious benefi from enering he EU, wih capial markes adjusing expecaions vis-à-vis Porugal, which also allowed for beer and more sable sovereign deb raings aribued o he counry. Regarding he pas experiences in erms of fiscal consolidaions, fiscal episodes can be idenified based on he change in he cyclically adjused primary budge balance. For his purpose, Afonso (2010) deermines for he EU counries he periods when he change in he primary cyclically adjused budge balance is a leas 2 percenage poins (pp) of GDP in one year or a leas 1.5 pp poins on average in he las wo years. For he case of Porugal, wo episodes of fiscal expansion ( , 2005) and hree episodes of fiscal conracion can be repored ( , 1986, and 1992). 5 Following such approach, we can also observe an addiional fiscal conracion in The abovemenioned fiscal consolidaion episodes were, on he one hand shorermed, and on he oher hand mosly unsuccessful. During he consolidaion boh expendiures and revenues increased, as a share of GDP, while he deb-o-gdp raio kep on increasing a he same ime. 6 In he 1986 consolidaion (he year of Porugal s enry in he EU), one observes a cerain sabilizaion of revenues as a share of GDP, a decrease in he expendiure-o-gdp raio, and also a decrease in he deb raio in (2008) did no find, for he period , significan differences in efficiency beween hospials wih public corporaions saus and Adminisraive Public Secor hospials. 5 Blanchard (2007) argues ha discreionary fiscal policy was expansionary in Porugal in The Poruguese Escudo effecive exchange rae faced a devaluaion of 17 and 23 per cen respecively in 1982 and 1983, while Porugal underook an IMF sabilisaion programme. 7

8 he following hree years. Addiionally, he primary balance was also in surplus for he firs ime in hireen years. Finally, he 1992 episode was very shor-ermed, aking place in a difficul environmen, following revenue and expendiure increases wih he deb raio rising immediaely aferwards. 7 Moreover, he 1993 economic downurn in Europe did no play in favour of prolonging he consolidaion, wih he primary spending-o-gdp raio increasing more significanly in ha year. Indeed, a commonly known feaure of fiscal policies in Porugal in he pas has been he pro-cyclical behaviour of primary spending, which conribued o preven he implemenaion of successful fiscal consolidaions (see, for insance, Pina, 2004). Such pro-cyclical behaviour would again be presen in 2001, wih he budge defici going once more above he 3 per cen limi. 4. Modelling Sraegies 4.1 The Bayesian Srucural VAR The firs mehodology used o analyze he macroeconomic effecs of fiscal policy consiss on he esimaion of he following Srucural VAR (SVAR) ( L) X id 1 0 X 1 X 1... id 1 c (1) nn n1 d 1 i ( 1 π )( 1 μ d ) 1 G T PY (2) 1 0 v, (3) 7 Addiionally, in 1992 and in 1993 privaisaion revenues amouning respecively o 1.5 and 0.4 per cen of GDP were used for deb redempion. Coincidenally Porugal enered he ERM in April 1992, precisely he year were several currencies in he ERM sysem undergone speculaive aacks, forcing boh he Ialian Lira and he Briish Pound ou of he sysem afer Sepember

9 where, s ~ (0, ), Γ(L) is a marix valued polynomial in posiive powers of X s he lag operaor L, n is he number of variables in he sysem, ε is he fundamenal economic shock, and v is he VAR innovaion. P, Equaion (2) describes he governmen s budge consrain, and i, G, T, π, Y, and d represen, respecively, he average cos of deb refinancing, governmen primary spending, governmen revenues, inflaion, GDP, price level, real growh rae of GDP, and he deb-o-gdp raio a he beginning of he period. The specificaion follows Favero and Giavazzi (2007) in ha we include he governmen deb dynamics, namely, by appending he non-linear budge ideniy o he VAR. follows: We use a recursive idenificaion scheme and characerize fiscal policy as G f ( ) (4) G T g( ) (5) T where, G is he governmen primary spending, T is he governmen revenue, f and g are linear funcions, is he informaion se, and G and T are, respecively, he governmen spending shock and he governmen revenue shock, which are orhogonal o he elemens in. We assume ha he variables in X can be separaed ino 2 groups: (i) a subse of n 1 variables, X 1 (GDP, GDP deflaor, consumpion, invesmen, cos of deb), whose conemporaneous values appear in he policy funcion and do no respond conemporaneously o he fiscal policy shocks; and (ii) he policy variables in he form of governmen primary expendiure, G, and/or governmen revenue, T. 9

10 Therefore, he recursive assumpions can be represened by ' X X 1, T, G and 0 11 n 1n1 21 2n1 0 n (6) This approach is also used by Chrisiano e al. (2005) in he conex of idenificaion of unexpeced variaion in moneary policy. Finally, we follow Sims and Zha (1999) and assess he poserior uncerainy abou he impulse-response funcions by using a Mone Carlo Markov-Chain (MCMC) algorihm. 4.2 The Fully Simulaneous Sysem Approach The second mehodology used in he idenificaion of fiscal policy shocks relies on a Fully Simulaneous Sysem of Equaions approach in a Bayesian framework. We sar by considering he srucural VAR represened by (1), (2) and (3). The X G, T, Y, P, i se of variables included in he sysem is ', G, he governmen primary expendiures, T, he governmen revenue, Y, he GDP, P, he GDP deflaor, and i, he average cos of deb financing. The economy is divided ino wo secors: a public and a producion secor. The public secor ha allows for simulaneous effecs, comprises he equaions for governmen primary spending and governmen revenue, and links hem wih he log real GDP, he GDP deflaor, and he average cos of financing deb. The producion secor consiss of log real GDP, and he GDP deflaor. Addiionally, we follow Blanchard and Peroi (2002) and Peroi (2004), and assess he auomaic response of axes and governmen primary spending o economic 10

11 aciviy by compuing he elasiciy of governmen revenue and spending o macroeconomic variables. The resricions on he marix of conemporaneous effecs, Γ 0, ha allow us o idenify he fiscal policy shock can be defined as G, Y T, Y G, T, G G, i 11 T, i 22 T 0 Y, (7) 0 P 55 i where he parameers ij represen he elasiciy of he fiscal policy insrumen i wih respec o he macroeconomic variable j. Table 1 repors he elasiciies used in he idenificaion procedure. Table 1 Elasiciies of Governmen Spending and Revenue. Elasiciies of Governmen Spending Elasiciies of Governmen Revenue G,Y G, G, i T, Y T, T, i Noe: The esimaes of he elasiciies are based on Girouard and André (2005), alhough he resuls are no sensiive o changes in parameers. Finally, we use Bayesian inference o assess he poserior uncerainy abou he impulse-response funcions in he Fully Simulaneous sysem of equaions as in Leeper and Zha (2003), and Sims and Zha (1999, 2006), and consider a Mone Carlo Imporance Sampling Normalized Weighs algorihm. 5. Empirical analysis 5.1 Daa This secion provides a summary descripion of he daa employed in he empirical analysis. A deailed descripion is provided in he appendix. All variables are 11

12 in naural logarihms unless saed oherwise and he daa covers he period 1978:1-2007:4. In he recursive parial idenificaion scheme, he variables ha are predeermined wih respec o fiscal policy innovaions are he GDP, privae consumpion, privae invesmen, and GDP deflaor. To hese variables, we add he average cos of governmen deb financing (or he yield o mauriy of long-erm governmen bonds). 8 As measure of he fiscal policy insrumens, we use eiher he governmen expendiures (and governmen revenues are included in X 1 ) or he governmen revenues (in which case, governmen expendiures are included in X 1 ). In he Fully Simulaneous Sysem approach, we resric he se of variables o he GDP, GDP deflaor, he average cos of deb financing and he fiscal policy insrumens. In boh frameworks, we include a consan (or quarerly seasonal dummies), and he governmen deb-o-gdp raio in he se of exogenous variables. All variables were deflaed by he GDP deflaor (2000=100). Figure 2 plos he average deb cos servicing and he nominal (annualized) GDP growh. I shows ha Porugal, in general, has moved from a siuaion where nominal GDP growh exceeded he cos of financing he deb o a siuaion where he converse has been rue. In addiion, Figure 3 displays he observed deb-o-gdp raio and he implici deb-o-gdp raio, ha is, he one ha emerges from he governmen deb s feedback. As can be seen, he implici series for he deb-o-gdp raio racks prey well he acual series. [Figure 2] [Figure 3] 8 The average governmen deb cos is obained by dividing he ne ineres paymens in by he governmen deb a ime 1. 12

13 Finally, he quarerly series of governmen spending and revenues are compued using he monhly Cenral Governmen s cash daa. 5.2 Resuls The Bayesian Srucural VAR We sar by esimaing a B-SVAR model ha does no include he feedback from governmen deb, ha is, where equaion (2) is no considered. Then, we compare he resuls wih he ones ha emerge from esimaing specificaions (1), (2), and (3). Figure 4 shows he impulse-response funcions o a fiscal policy shock. The solid line refers o he median response when he VAR is esimaed wihou he feedback from governmen deb, and he dashed lines are, respecively, he median response and he 68 per cen poserior confidence inervals from he VAR esimaed by imposing he deb dynamics. The confidence bands are consruced using a Mone Carlo Markov-Chain (MCMC) algorihm based on draws. [Figure 4] Figure 4a displays he impulse-response funcions of all variables in X o a posiive shock in governmen primary spending. In he case we do no include he deb feedback, i can be seen ha governmen spending declines seadily following he shock, and he effec roughly vanishes afer eigh quarers. The effecs on GDP are negaive and reveal ha governmen spending has a srong crowding-ou effec on he privae secor. In fac, boh privae consumpion and privae invesmen fall afer he shock. In paricular, a 6% shock in governmen spending leads o a 0.5% fall in GDP six quarers ahead. Similarly, privae consumpion falls by 0.6%, while he effecs in privae invesmen are more pronnounced (a fall of 1% a he horizon of six quarers). These resuls are in line wih he works of Giavazzi and Pagano (1990) and Alesina and 13

14 Ardagna (1998) who uncovered he presence of non-keynesian effecs (i.e., negaive spending mulipliers) during large fiscal consolidaions. In addiion, here is a posiive effec on he average cos of deb ha reaches is peak afer six quarers. The price level is also impaced persisenly and posiively by he shock in governmen spending. Finally, he resuls sugges ha afer a governmen spending shock, here is an increase in governmen revenue which is, however, small. Therefore, his suggess ha an expansion of governmen spending is associaed wih a episode of fiscal deerioraion. When we include he deb dynamics in he model, he effecs of a governmen spending shock on he average cos of deb become somewha larger while he impac on GDP is marginally smaller. Addiionally, invesmen consisenly falls much more han before and he posiive impac on he price level is aenuaed by he feedback from governmen deb. Figure 4b shows he impulse-response funcions o a posiive shock in governmen revenue. The resuls sugges ha governmen revenue declines afer he shock which erodes in abou eigh quarers. The effecs on GDP, privae consumpion and privae invesmen are slighly posiive over he four quarers following he shock, bu hey quickly mean rever and become negaive. In fac, a 5% shock in governmen revenue has is maximum impac on GDP (0.3%), privae consumpion (0.2%) and privae invesmen (0.7%) a he four quarers horizon, afer which he effecs erode and change sign. These resuls are in accordance o he findings of Giavazzi e al. (2000) who show ha a rise of axes can have a posiive impac on privae consumpion in periods of fiscal consolidaion. In conras, he price level falls for abou four quarers, hen recovers, and becomes posiive. This evidence is closely relaed o he reacion of governmen spending, which increases afer he shock. In fac, an increase in 14

15 governmen revenue is followed by a somewha less disciplined fiscal policy and, as a resul, here is a deerioraion of he fiscal balance. This also seems o be he reason for he posiive impac on he average cos of deb, implying ha he auhoriies should be aware of such marke reacions when implemening fiscal policy. In erms of he forecas error-variance decomposiion of he variables in he sysem o a shock in governmen spending or revenue (no shown for space sake), boh shocks accoun for a large fracion of heir own forecas-error variance decomposiion, and play a negligible role for he remaining variables The Fully Simulaneous Sysem Approach Figure 5 displays he impulse-response funcions o a fiscal policy shock in he Fully Simulaneous Sysem approach. The solid line refers o he median response when he sysem is esimaed wihou imposing he dynamics from governmen deb, and he dashed lines are, respecively, he median response and he 68 per cen poserior confidence inervals from he sysem esimaed by including he feedback from governmen deb. The confidence bands are consruced using an Imporance Sampling Normalized Weighs algorihm based on draws. [Figure 5] Figure 5a displays he impulse-response funcions of all variables o a posiive shock in governmen primary spending. When we do no ake ino accoun he feedback from governmen deb, one can see ha he shock o governmen spending erodes afer four quarers. The effecs on GDP are negaive herefore, denoing he presence of non-keynesian effecs - and he rough is reached afer four quarers. A his horizon, a 3% shock in governmen spending leads o a fall of GDP of beween 0.2% and 0.4%. 15

16 As in he case of he recursive parial idenificaion scheme, he impac on he price level is posiive, alhough somewha less persisen, and here is also a fiscal deerioraion as governmen revenues fall afer he shock. In conras, here is a negaive effec on he average cos of deb ha reaches is rough afer six quarers. When he deb dynamics is included in he model, he effecs of a governmen spending shock: (i) on GDP are smaller; (ii) become larger in he case of he average cos of deb; and (iii) are somewha smaller for he price level. Figure 5b shows he impulse-response funcions o a posiive governmen revenue shock. The (negaive) response of GDP is no lagged, despie being gradual. The rough is achieved afer eigh quarers, where a 3% shock in governmen revenue reduces GDP by abou 0.4%. Similarly, he price level does no reac wih a lag bu gradually increases insead, and he effec is very persisen. These pieces of evidence can no be separaed from he behaviour of governmen spending, which increases afer he shock, herefore, suggesing a deerioraion of public finances. In fac, he average cos of deb is posiively impaced, reflecing he less disciplined fiscal policy. 5.3 Fiscal shocks and governmen deb feedback In his sub-secion, we consider he poenial deb feedback by esimaing he following srucural VAR: 0 X 1 X 1... i ( d 1 d*) c, (8) d 1 i ( 1 π )( 1 μ d ) 1 G T PY. (9) Specificaion (8) is suggesed by Bohn (1998) who considers a fiscal reacion funcion in which d* is he uncondiional mean of he deb raio. In he same spiri, Romer and Romer (2007) sugges ha he effec of a ax shock on oupu may be 16

17 condiional on he governmen s aim o sabilize he deb. Therefore, we model he arge level of he deb as a consan on he basis of he evidence of saionariy of d. The esimaed coefficiens on (d 1 d*) in he srucural equaions of he SVAR (governmen spending and governmen revenue) are repored in Table 2. We consider he full sample and wo sub-samples: 1979:1 1993:3, corresponding o he period before he Maasrich Treay enered ino force; and 1993:4 2007:4, hereafer. Table 2 The effec of (d -1 -d*) in a VAR. (d -1 -d*) 1979:1-2007:4 (N=116) 1979:1-1993:3 (N=59) 1993:4-2007:4 (N=57) T (0.001) *** (0.002) 0.006* (0.03) G (0.002) * (0.003) (0.004) Noe: sandard errors in brackes. *, **, *** - saisically significan respecively a he 10%, 5%, and 1% levels. In general, he resuls do no show a significan response of revenue and primary spending o deviaions of he deb-o-gdp raio from is sample average for he full sample. In he firs sub-sample (1979:1 1993:3), here is some evidence suggesing a weak sabilizing effec ha works mainly hrough governmen spending: when he debo-gdp raio is above is hisorical mean, governmen primary spending decreases (he coefficien associaed o (d 1 d*) is negaive (-0.005), alhough some desabilizing effec hen also occurs via he revenue side. In he second sub-sample (1993:4 2007:4), he empirical findings show ha governmen revenue plays some sabilizing effec: when he deb-o-gdp raio is above is hisorical mean, governmen revenue increases as he coefficien associaed o (d 1 d*) is posiive (0.006). Therefore, we can conclude ha when faced wih high governmen indebedness and fiscal imbalances in he Maasrich sub-period, he fiscal sabilizaion responses have been biased owards increases in governmen revenue. 17

18 5.4 A VAR couner-facual exercise We now conduc a VAR couner-facual exercise aimed a describing he effecs of shuing down he shocks in governmen spending or governmen revenue. In pracice, afer esimaing he VAR summarized by (1), (2) and (3), we consruc he couner-facual (CFT) series as follows: CFT ( L) X id nn n1 1 0 X CFT X 1 CFT 1... d i 1 c CFT (10) d 1 i ( 1 π )( 1 μ d ) 1 G T PY (11) CFT 1 CFT 0 v. (12) This is equivalen o consider he following vecor of srucural shocks CFT T Y P i I C 0,,,,,, (13) ' CFT G Y P i I C,0,,,,,, (14) ' where we shu down, respecively in (13) and in (14), he governmen primary spending and he governmen revenue unexpeced variaion and hen use he couner-facual srucural shocks o build he couner-facual series for all endogenous variables of he sysem. This empirical exercise allows us o quanify he magniude of fiscal policy shocks and is impac on a se of macroeconomic variables. Moreover, i helps us undersanding wha he dynamics of he economy would be in he absence of unexpeced variaion in fiscal policy. Figure 6a plos he acual and he couner-facual series for GDP, privae consumpion, privae invesmen, and governmen spending in he case of a shock o governmen spending. Figure 6b displays he acual and he couner-facual series for 18

19 GDP, privae consumpion, privae invesmen, and governmen revenue in he case of a shock o governmen revenue. [Figure 6] The resuls show ha fiscal policy shocks play a minor role as he difference beween he acual and he counerfacual series are relaively small. Neverheless, one can see ha in he absence of governmen spending shocks, privae consumpion and privae invesmen would have been higher, for insance, in he period and, more recenly, since Therefore, such evidence suggess and confirms ha unexpeced increases in governmen spending generae relevan crowding-ou effecs, a useful insigh o bear in mind when resoring o fiscal insrumens o boos he economy. In he case of governmen revenue, he difference beween he acual and he counerfacual series are negligible, a feaure ha may be relaed wih he relaive size of he governmen revenue shocks. In fac, while unexpeced variaion in governmen spending seems o be large as follows from he larger differences beween he acual governmen spending and he couner-facual governmen spending, governmen revenue shocks are generally small. 6. Conclusion This paper evaluaes he macroeconomic effecs of fiscal policy in Porugal for he period 1979:1-2007:4, drawing on a new se of quarerly daa buil from he monhly Cenral Governmen s cash daa. We idenify fiscal policy shocks using: (i) a recursive parial scheme, and esimae a Bayesian Srucural Vecor Auoregression; and (ii) a Fully Simulaneous Sysem approach, where we accoun for he auomaic response of fiscal policy o he economic aciviy. Consequenly, he use of such high frequency 19

20 fiscal daa, which is quie valuable from a perspecive of monioring he fiscal posiion, urns ou o be also raher useful from a modelling poin of view. The empirical evidence suggess ha governmen spending shocks: (i) have, in general, a negaive effec on GDP; (ii) lead o a fall of boh privae consumpion and privae invesmen; (iii) rise persisenly and posiively he price level; and (iv) have mixed effecs on he average cos of refinancing he deb. More specifically, a 1% posiive shock in governmen spending has a maximum impac on GDP of -0.1%, while crowding-ou boh privae consumpion and privae invesmen, which fall by 0.12% and 0.2%, respecively. In addiion, governmen revenue shocks have a negaive impac on GDP, on privae consumpion and on privae invesmen, alhough he response emerges wih a lag of abou four quarers. In fac, a his horizon, a 1% posiive shock in governmen revenue is able o generae a posiive response of GDP by 0.06%, and an increase of boh privae consumpion and privae invesmen of 0.04% and 0.14%, respecively. Afer his, he macroeconomic effecs of he ax shock erode and even become negaive. These resuls sugges ha an expansion of governmen spending is associaed o an episode of fiscal deerioraion. Similarly, an increase in governmen revenue is followed by a somewha less disciplined fiscal policy. This helps explaining he differen reacion of he cos of deb o he fiscal policy shock across he wo idenificaion mehods. Neveheless, he response of he oher macroeconomic variables included in he differen frameworks is qualiaive and quaniaive similar, which gives suppor o he policy implicaions of he paper and allows us o be confiden on he correc idenificaion of he unexpeced variaion in fiscal policy. When we explicily consider he feedback from governmen deb, (long-erm) ineres raes become more responsive, GDP is marginally less responsive, and he 20

21 effecs of fiscal policy on hese variables also become more persisen. In addiion, he resuls provide weak evidence of sabilizing effecs of he deb level on he primary budge balance. In a nushell, i is adequae o say ha he resuls reinforce he idea ha fiscal behaviour in Porugal came shor of effecively consolidaing public finances. Moreover, expansionary spending shocks have ended o lower real GDP growh. In addiion, a VAR couner-facual exercise shows ha unexpeced spending shocks are responsible for imporan crowding-ou effecs. Therefore, he curren sudy provides relevan policy implicaions, noably from he perspecive of he fiscal auhoriies. Mos imporan is he fac ha expansionary fiscal policies in Porugal do no emerge as an obvious insrumen o foser economic growh. On he oher hand, and from an inflaion conrol poin of view, governmen spending also pus upwards pressure on he price level. Anoher significan resul, o bear in mind by policymakers when implemening fiscal policy measures in a small open economy as Porugal, is he fac ha when faced wih high governmen indebedness and fiscal imbalances, afer he Maasrich Treay enered ino force, he fiscal sabilizaion responses have been biased owards increases in governmen revenue. Given he associaed excess burden of axaion of such policies and he uncovered crowding-ou effec of governmen spending, i would be advisable o consider designing fiscal policies from a perspecive of reducing oal governmen spending. Such approach, when pursued in good imes, would conribue o consolidae public finances in a more susained fashion, while allowing more leeway for an expansionary behaviour in crises, and sill wihin he EU fiscal framework. 21

22 References Afonso, A. (2010), Expansionary fiscal consolidaions in Europe: new evidence, Applied Economics Leers, 17 (2), Afonso, A., and P. Claeys (2008), The dynamic behaviour of budge componens and oupu, Economic Modelling, 25, Afonso, A., and S. Fernandes (2008). Assessing hospial efficiency: non-parameric evidence for Porugal, Deparmen of Economics, ISEG-UTL, Working Paper Nº. 07/2008/DE/UECE. Afonso, A., and R. M. Sousa (2009a), The macroeconomic effecs of fiscal policy, ECB Working Paper Nº Afonso, A., and R. M. Sousa (2009b), Fiscal policy, housing and sock prices, ECB Working Paper Nº Alesina A., and S. Ardagna (1998), Tales of fiscal adjusmen, Economic Policy, 27, Beesma, R., and H. Jensen (2005), Moneary and fiscal policy ineracions in a Micro- Founded Model of a Moneary Union, Journal of Inernaional Economics, 67, Biau, O., and E. Girard (2005), Poliique budgéaire e dynamique économique en France: l'approche VAR srucurel., Économie e Prévision, , Blanchard, O. (2007), Adjusmen wihin he Euro: The Difficul Case of Porugal, Poruguese Economic Journal, 6(1), Blanchard, O., and R. Peroi (2002), "An empirical characerizaion of he dynamic effecs of changes in governmen spending and axes on oupu", Quarerly Journal of Economics, 117(4), Bohn, H. (1998), "The Behaviour of U.S. public deb and deficis", Quarerly Journal of Economics, 113, Burriel, P., de Casro, F., Garroe, D., Gordo, E., Paredes, J. and Pérez, J. (2010). Fiscal policy shocks in he euro area and he US: An empirical assessmen, Fiscal Sudies, 31(2), Chrisiano, L. J., Eichenbaum, M., and C. L. Evans (2005), "Nominal rigidiies and he dynamic effecs of a shock o moneary policy", Journal of Poliical Economy, 113(1),

23 De Casro Fernández, F., and P. Hernández De Cos (2008), The economic effecs of exogenous fiscal shocks in Spain: a SVAR approach, Journal of Macroeconomics, 30, Faás, A., and I. Mihov (2001), The effecs of fiscal policy on consumpion and employmen: heory and evidence, CEPR Discussion Paper Nº Favero, C., and F. Giavazzi (2007), "Deb and he effecs of fiscal policy", Universiy of Bocconi, Working Paper Nº Giavazzi, F., and M. Pagano (1990), Can severe fiscal conracions be expansionary? Tales of wo small european counries, in Blanchard, O. J.; Fischer, S. (eds.), NBER Macroeconomics Annual, MIT Press, Giavazzi, F., Jappelli, T., and M. Pagano (2000), "Searching for non-linear effecs of fiscal policy: evidence from indusrial and developing counries", European Economic Review, 44(7), Giordano, R., Momigliano, S., Neri, S., and R. Peroi (2007), The effecs of fiscal policy in Ialy: Evidence from a VAR model, European Journal of Poliical Economy, 23, Girouard, N., and C. André (2005), Measuring cyclically-adjused budge balances for OECD counries, OECD Economics Deparmen Working Paper No Guichard, S., and W. Leibfriz (2006). The fiscal challenge in Porugal, OECD Working Paper Nº Heppke-Falk, K. H., Tenhofen, J., and G. B. Wolff (2006), The macroeconomic effecs of exogenous fiscal policy shocks in Germany: a disaggregaed SVAR analysis, Deusche Bundesbank, Discussion Paper Nº. 41. Leeper, E. M., and T. Zha (2003), Modes policy inervenions, Journal of Moneary Economics, 50(8), Mounford, A., and H. Uhlig (2009), "Wha are he effecs of fiscal policy shocks?", Journal of Applied Economerics, 24 (6), Peroi, R. (1999), Fiscal policy in good imes and bad, Quarerly Journal of Economics, 114, Peroi, R. (2004), "Esimaing he effecs of fiscal policy in OECD counries", Bocconi Universiy, IGIER Working Paper Nº Pina, A. (2004), Fiscal policy in Porugal: discipline, cyclicaliy and he scope for expendiure rules, proceedings of he 2 nd Conference on Poruguese Economic 23

24 Developmen in he European Conex, held by he Bank of Porugal in Lisbon, March 2004, Ramey, V., and M. Shapiro (1998), Cosly capial reallocaion and he effecs of governmen spending, Carnegie Rocheser Conference on Public Policy, 48, Romer, C., and D. H. Romer (2007), "The macroeconomic effecs of ax changes: esimaes based on a new measure of fiscal shocks", American Economic Review, 100(3), Sims, C., and T. Zha (2006), "Does moneary policy generae recessions?", Macroeconomic Dynamics, 10(2), Sims, C., and T. Zha (1999), "Error bands for impulse-responses", Economerica, 67(5),

25 Appendix Daa descripion and sources GDP Daa for GDP are quarerly, seasonally adjused, and comprise he period 1978:1-2007:4. The source is he Bank of Porugal. Privae Consumpion The source is he Bank of Porugal. Consumpion is defined as he household consumpion expendiure including non-profiable insiuions serving households. Daa are quarerly, seasonally adjused, and comprise he period 1978:1-2007:4. Price Deflaor All variables were deflaed by he GDP deflaor (2000=100). Daa are quarerly, seasonally adjused, and comprise he period 1978:1-2007:4. The source is he Bank of Porugal. Privae Invesmen The source is he Bank of Porugal. Privae Invesmen is defined as oal gross fixed capial formaion. Daa are quarerly, seasonally adjused, and comprise he period 1978:1-2007:4. Governmen Spending The source is he Bank of Porugal, colleced from he Monhly Bullein of he Direcorae-General of Public Accouning. Governmen Spending is defined as Cenral Governmen primary spending (on a cash basis), ha is, he difference beween auhorized expendiure and deb ineres paymens. We seasonally adjus quarerly daa using Census X12 ARIMA, and he series comprise he period 1978:1-2007:4. Ineres Paymens The source is he Bank of Porugal, colleced from he Monhly Bullein of he Direcorae-General of Public Accouning. Ineres Paymens is defined as Cenral Governmen deb ineres paymens (on a cash basis). We seasonally adjus quarerly daa using Census X12 ARIMA, and he series comprise he period 1978:1-2007:4. Governmen Revenue The source is he Bank of Porugal, colleced from he Monhly Bullein of he Direcorae-General of Public Accouning. Governmen Revenue is defined as Cenral Governmen oal revenue (on a cash basis). We seasonally adjus quarerly daa using Census X12 ARIMA, and he series comprise he period 1978:1-2007:4. Governmen Deb The source is he Bank of Porugal, he Direcorae-General of Treasury, and he Direcorae-General of Public Credi. Governmen Deb is defined as he sock of Direc Sae Deb. The original series are available as follows: 1. a) Toal Inernal Deb, for he period 1997: :6, on a quarerly basis; b) Inernal Direc Deb, for he period 1997: :6, on a quarerly basis; c) Toal Exernal Deb, for he period 1997: :6, on a quarerly basis; d) Direc Exernal Deb, for he period 1997: :6, on a quarerly basis; 25

26 e) Toal Public Deb, for he period 1997: :6, on a quarerly basis; f) Effecive Public Deb, for he period 1997: :6, on a quarerly basis; a) Inernal Effecive Direc Deb, for he periods 1991:12, 1992:12, and 1993:6-1995:11, on a monhly basis; b) Toal Effecive Direc Deb, for he periods 1991:12, 1992:12, and 1993:6-1995:11, on a monhly basis a) Inernal Direc Deb, for he period 1995:7-1998:12, on a monhly basis; b) Toal Direc Deb, for he period 1995:7-1998:12, on a monhly basis a) Direc Sae Deb, for he period 1998: :4, on a monhly basis. We build he series for he Direc Sae Deb as follows: 1) For he period 1998: :4, as he series of Direc Sae Deb iself; 2) For he period 1995:7-1997:12, we use he raio of Direc Sae Deb o Toal Sae Deb in 1998:12 (ha is, a scale facor of ), o back-ou he series of Direc Sae Deb; 3) For he period 1993:6-1995:6, we use he raio of Toal Effecive Direc Sae Deb o Toal Direc Sae Deb in he period 1995:7-1995:11 (ha is, a scale facor of ), o back-ou he series of Toal Direc Deb; 4) For he period 1977: :3, we use he raio of (Effecive Public Deb minus Non-Direc Deb) o Toal Effecive Direc Deb in he period 1993:6-1994:6 (ha is, a scale facor of ), o back-ou he series of Toal Effecive Direc Deb. Given ha he scale facors are very close o one, he ime series of he Direc Sae Deb is smooh over ime and we guaranee ha here are no srucural breaks. We build he quarerly series using monhly daa (where available) and seasonally adjus i using Census X12 ARIMA. The consruced series comprise he period 1977:4-2007:4. Average Cos of Financing Deb The average cos of financing deb is obained by dividing ineres paymens by deb a ime -1. Long-Term Ineres Rae Long-Term Ineres Rae corresponds o he yield o mauriy of 10-year governmen securiies. Daa are quarerly, and comprise he period 1957:1-2007:4. Daa for he period 1974:2-1975:4 is no available. Therefore, we linearly inerpolae he daa for ha period using he observaions a 1974:1 and 1976:1. The source is he IMF, Inernaional Financial Saisics (series " IFS.Q $$$.Z.F.$$$"). 26

27 1978Q4 1980Q1 1981Q2 1982Q3 1983Q4 1985Q1 1986Q2 1987Q3 1988Q4 1990Q1 1991Q2 1992Q3 1993Q4 1995Q1 1996Q2 1997Q3 1998Q4 2000Q1 2001Q2 2002Q3 2003Q4 2005Q1 2006Q2 2007Q3 Change in real shor-erm ineres rae (Euribor 6m) Figure 1 Moneary condiions and fiscal balances in Porugal ( ). EDP Fiscal easing, Moneary ighening Fiscal ighening, Moneary ighening Fiscal easing, Moneary easing Change in cyclically adjused primary balance Fiscal ighening, Moneary easing Source: EC, Eurosa, Banco de Porugal, and own calculaions. Noes: HICP, Sepember 2008, Euribor, Ocober 2008, and EC Auumn 2008 forecass for CAPB in EDP Excessive Defici Procedure. Figure 2 - Average cos of servicing deb and (annualized) nominal GDP growh. Figure 3 - Implici deb raio and observed deb raio, percenage of GDP. Deb Raio % Implici Deb Raio Observed Deb Raio Noe: Implici deb raio compued via equaion (9). 27

28 Figure 4 Impulse-response funcions: he Bayesian Srucural VAR. 4a spending shock 4b revenue shock Noes: The solid line corresponds o he median esimae when he model is esimaed wihou accouning for he feedback from governmen deb; he doed line corresponds o he median esimae and he dashed lines indicae he 68 per cen poserior confidence inervals esimaed by using an Imporance Sampling Normalized Weighs algorihm based on draws, when he deb dynamics is included in he model. 28

29 Figure 5 Impulse-response funcions: he Fully Simulaneous Sysem approach. 5a spending shock 5b revenue shock Noes: The solid line corresponds o he median esimae when he model is esimaed wihou accouning for he feedback from governmen deb; he doed line corresponds o he median esimae and he dashed lines indicae he 68 per cen poserior confidence inervals esimaed by using an Imporance Sampling Normalized Weighs algorihm based on draws, when he deb dynamics is included in he model. 29

30 Figure 6 VAR counerfacual. 6a spending shock 6b revenue shock Noe: All series are in logs of real erms. 30