The impact of government spending on the private sector: Crowding-out versus crowding-in effects *

Transcription

1 The mpac of governmen spendng on he prvae secor: Crowdng-ou versus crowdng-n effecs * Davde Furcer and Rcardo M. Sousa Summary We conrbue o he emprcal leraure on he effec of governmen spendng on economc acvy by assessng he mpac of changes n governmen spendng-gdp rao on (he shor-erm growh raes of) prvae consumpon and nvesmen. We do hs by analysng a panel sample of 15 counres from 1960 o The resuls of our paper sugges ha governmen spendng produces mporan crowdng-ou effecs by negavely affecng boh prvae consumpon and nvesmen. The resul s broadly robus o boh counry and me effecs and dfferen economerc specfcaons. In addon we show ha he effec of governmen consumpon on prvae consumpon and nvesmen does no depend on he phase of he busness cycle bu dffers subsanally among regons. The dfferenaed effecs of governmen consumpon on prvae consumpon and nvesmen among geographcal areas are exremely mporan and need o be furher nvesgaed. In parcular would be neresng o assess o whch exen he effec of governmen spendng on consumpon and nvesmen depends on polcal and nsuonal varables (e.g. democracy corrupon polcal sably) as well as macroeconomc varables (ncome neres raes degree of openness). We leave hs challengng avenue for fuure research. * Davde Furcer. Inernaonal Moneary Fund h Sree NW 2031 Washngon DC USA. Emal: dfurcer@mf.org;unversy of Palermo. Vale delle Scenze 90128Palermo Ialy. Emal : furcer@economa.unpa.. Rcardo Sousa. Unversy of Mnho Deparmen of Economcs and Economc Polces Research Un (NIPE) Campus of Gualar Braga Porugal; London School of Economcs Fnancal Markes Group (FMG) Houghon Sree London WC2 2AE Uned Kngdom. E-mals: rsousa@eeg.umnho.p; rsousa@alumn.lse.ac.uk. The opnons expressed heren are hose of he auhors and do no necessarly reflec hose of he IMF or s member counres. 1

2 Absrac The am of hs paper s o analyze he mpac of governmen spendng on he prvae secor assessng he exsence of crowdng-ou versus crowdng-n effecs. Usng a panel of 15 counres from 1960 o 2007 he resuls sugges ha governmen spendng produces mporan crowdng-ou effecs by negavely affecng boh prvae consumpon and nvesmen. Moreover whle he effecs do no seem o depend on he dfferen phases of economc cycle hey vary consderably among regons. The resuls are economcally and sascally sgnfcan and robus o several economerc echnques. Keywords: Fscal Polcy Governmen Spendng Crowdng-ou Crowdng-n. EL: E0 E6. 2

3 I. INTRODUCTION The heorecal and emprcal leraure has provded n he las year exensve analyss on he effec of governmen spendng on economc acvy. Despe hs here s no consensus on he effecs of governmen spendng on prvae consumpon and nvesmen (boh n he shor and n he long erm) neher from a heorecal nor from an emprcal pon of vew. Indeed from a heorecal perspecve he effec of an ncrease of governmen spendng on hose varables can be of boh sgns. The Real Busness Cycle (RBC) model predcs a declne n prvae consumpon n response o a rse n governmen spendng: wh nfnely-lved Rcardan households an ncrease n governmen spendng lowers he presen value of afer-ax ncome and hus generaes a negave wealh effec on consumpon (Ayagar e al. 1990; Baxer and Kng 1993; Chrsano and Echenbaum 1992). In conras he IS-LM model predcs ha consumpon should rse n response o a posve governmen spendng shock: when consumers behave n a non-rcardan fashon her consumpon s a funcon of her curren dsposable ncome hus an ncrease n ncome wll generae an ncrease n prvae consumpon (Blanchard 2001). Smlarly o consumpon he wo heores also predc dfferen oucomes for nvesmen. The RBC model clams ha an ncrease of governmen consumpon wll have a posve effec on nvesmen: an ncrease of governmen consumpon nduces a rse n employmen whch f suffcenly perssen leads o a rse n he expeced reurn o capal and herefore may rgger a rse n nvesmen. On he conrary he IS-LM model predcs ha nvesmen should declne n response o a posve governmen spendng shock: an ncrease n governmen consumpon (f no followed by an accommodang ncrease of money supply) leads o an ncrease n he neres rae whch n urn wll ranslae no a decrease n nvesmen. From hs dscusson emerges ha he predcons of he above menoned heores are orhogonal o each oher. These conrasng vews gave rse o several emprcal sudes aempng o assess he mpac of publc expendures on consumpon and prvae nvesmen. Unforunaely he predcons of he emprcal evdence are also que mxed n suppor of one heory or he oher as can be seen from Table 1. The man purpose of hs paper s o conrbue o he emprcal leraure by analyzng he mpac of changes n governmen spendng on prvae consumpon and nvesmen. By dong hs we provde an addonal es on wheher governmen 3

4 o The resuls show ha governmen spendng produces mporan crowdng-ou spendng generaes "crowdng-ou" or "crowdng-n" effecs on he prvae secor. In addon we can also dscrmnae beween he sandard RBC and IS-LM model. Whle mos of he ess of he crowdng-ou versus crowdng-n hypohess ha have been carred n prevous papers focus on a me seres or cross-counry approach hs work exends such analyss o a panel daa se of 15 counres from 1960 effecs by negavely affecng boh prvae consumpon and nvesmen. The emprcal evdence also suggess ha neher he predcon of he sandard RBC model nor he one of IS-LM model can be aken overall as vald. In fac our resuls are n conras wh he RBC predcon of a rse n nvesmen and wh he IS-LM predcon of a rse n consumpon. In addon we analyze possble asymmeres of he effec of governmen consumpon on prvae consumpon and nvesmen. In parcular we es: ) wheher he effec vares among regons; and ) wheher depends on o he phase of he economc cycle. We fnd ha he effec vares subsanally among regons bu does no seem o depend on he phase of he economc cycle. We show ha all resuls are economcally and sascally sgnfcan and robus o several economerc echnques. The res of he paper s organzed as followng. Secon wo descrbes he daa. Secon hree shows he emprcal mehodology used o assess he crowdng-ou versus he crowdng-n effecs and dscusses he maor resuls. Secon four provdes addonal robusness resuls and addresses he exsence of poenal asymmeres on he effec of governmen consumpon on prvae consumpon and nvesmen. Fnally Secon fve concludes wh he man fndngs and suggesons for furher research. II. DATA Ths secon provdes a summary descrpon of he daa employed n he emprcal analyss. 1 Aukeren (2005) also analyzes he ssue of crowdng-ou effecs bu from a dfferen perspecve n ha he auhor looks a he lnkages beween publc and prvae nvesmens and emphaszes he role of he share of governmen nvolvemen n he economy. Smlarly Gonzalez-Paramo and De Cos (2005) fnds ha publc ownershp negavely affecs producvy whle prvazaon has a posve effec on effcency.

5 The daa cover 15 counres and are obaned from he World Bank s World Developmen Indcaors for he me perod We consder annual daa for GDP prvae consumpon prvae nvesmen and governmen spendng. Due o daa avalably (boh n erms of me and counry dmenson) we decded o proxy prvae nvesmen and governmen spendng by usng gross fxed capal formaon and publc consumpon (whch represens he larges share of oal governmen spendng) respecvely. All varables are expressed n real per capa erms where we use he GDP deflaor o conver nomnal n real consan erms. The focus of he analyss s on he exsence of crowdng-n versus crowdngou effecs of governmen spendng. Consequenly we sudy he mpac of changes n he rao of governmen spendng o GDP on he growh of real per capa prvae consumpon and prvae nvesmen. Table A n he Appendx provdes he ls of he counres used n he sudy and Table 2 summarzes he descrpve sascs of he above-menoned varables. We can see ha he prvae secor s a very mporan componen of GDP: prvae consumpon represens almos wo hrds (68.22%) of GDP whle he rao of prvae nvesmen o GDP corresponds o an average of 21.55%. By s urn governmen consumpon represens 15.7% of GDP. Over he overall sample real per capa prvae consumpon grew a an annual rae of 1.3% whle he growh rae of prvae nvesmen was 0.72%. Despe he lower growh rae prvae nvesmen exhbed a much larger volaly as expressed by he sandard devaon. The change n governmen spendng n percenage of GDP was negave (-0.02). Fnally Table 3 summarzes he correlaon coeffcens beween he growh rae of prvae consumpon he growh rae of prvae nvesmen and he change n he rao of governmen spendng o GDP. I shows ha alhough prvae consumpon and prvae nvesmen are posvely correlaed her correlaons wh governmen spendng are negave and small. III. EMPIRICAL METHODOLOGY AND RESULTS We analyze he relaonshp beween prvae consumpon growh and he change n he rao of governmen spendng o GDP and esmae a model smlar o he emprcal specfcaon used n Romer and Romer (2007) o esmae he mpac of ax changes on economc acvy: 5

6 C C G 0 Y. (1) We also look a he mpac of a change n he governmen spendng o GDP rao on prvae nvesmen growh by esmang he followng model: I I G 0 Y. (2) In he above specfcaons C represens prvae consumpon I s he prvae nvesmen Y s he GDP he s are parameers o be esmaed s ndexng over counres and over me and represen counry- and me-specfc effecs s he frs dfference operaor s he number of lags (se equal o four) and s he error erm. Panel A and Panel B n Table presen he resuls for he esmaon of respecvely equaons (1) and (2). The columns of he ables show he resuls obaned usng dfferen economerc specfcaons namely: ) OLS wh me fxed effecs; ) OLS wh counry fxed effecs; ) OLS wh boh counry and me fxed effecs; and v) counry Random effecs esmaor. Sarng wh he analyss of he effec of governmen consumpon on prvae consumpon (Panel A) we can mmedaely see ha s negave and sascally sgnfcan. The resuls also sugges ha no only conemporaneous changes n he governmen consumpon-gdp rao maer bu also s pas lags (specfcally he 2 nd and 3 rd ones). In parcular he cumulave effec of governmen spendng on prvae consumpon s abou 1.9 % of whch abou 1.2% capured by conemporaneous changes n he governmen consumpon-gdp rao and 0.7 % by s lags. Ths resul can be nerpreed as follows: an ncrease of governmen consumpon by 1 % of real GDP mmedaely reduces consumpon by approxmaely 1.2% wh he declne connung for abou four years when he cumulave decrease n consumpon has reached approxmaely 1.9 %. The resul s broadly robus o boh counry and me effecs and boh o Fxed and Random effecs specfcaon. In Panel B we repor he resuls obaned esmang he nvesmen equaon (2). Smlarly o wha we obaned for prvae consumpon boh curren and lagged changes n governmen consumpon-gdp rao have a negave and sgnfcan effec on prvae nvesmen wh a cumulave effec of approxmaely 1.8%. The man dfference beween he effec on consumpon and nvesmen s ha whle conemporaneous 6

7 change n he governmen consumpon-gdp rao seems o have a bgger effec on consumpon lagged changes are more dermenal for nvesmen. IV. ROBUSTNESS ANALYSIS AND ASYMMETRIC EFFECTS.1 Exogeney Snce our measure of he change n he rao of governmen spendng o GDP may no be compleely exogenous here s he rsk ha he esmaed s n models (1) and (2) are based (and nconssen). A frs aemp o address hs ssue s carred ou by elmnang he conemporaneous change n he rao of governmen spendng o GDP n models (1) and (2). In fac snce boh he growh rae of consumpon (and nvesmen) and our ndependen varable are (for he vas maory of he counres n he sample) saonary and no perssen we should expec ha he lagged values of our ndependen varables are no nfluenced by he curren value of consumpon (and nvesmen) growh raes. Followng hs approach we revse models (1) and (2) o: and C C G 1 Y (3) I I G 1 Y. () respecvely hereby smply excludng he conemporaneous change n he rao of governmen spendng o GDP from he orgnal equaons. Due o space consrans n Table 5 we only repor he esmaed sums of he s obaned wh hs analyss and we compare hem wh he ones obaned prevously (Table 5). Lookng a he able we can sll see ha governmen consumpon crowds-ou boh prvae consumpon and nvesmen. However as already poned ou n he prevous secon he cumulave effec of lags of changes n governmen consumpon-gdp rao (Panel B) s lower han he one due o conemporaneous changes (Panel A). In parcular he cumulave effec of (lagged) changes n governmen consumpon-gdp rao on prvae consumpon (nvesmen) s approxmaely 0.10% (0.35%). A second aemp o correc for possble endogeney problems s carred ou by usng he GMM esmaor proposed by Blundell and Bond (1998). The resuls repored n he able clearly show ha he esmaed mpac of governmen spendng on boh 7

8 8 prvae consumpon and nvesmen s qualavely (n erms of sgn) and quanavely (n erms of magnude) unchanged..2 Seral correlaon Anoher possble problem wh specfcaons (1) and (2) s he presence of seral correlaons. To ackle hs ssue we modfy models (1) and (2) o: u GDP G C C v w C C 0 1 (5) and u GDP G I I v w I I 0 1 (6) where we add lags of he dependen varables o he se of explanaory varables so ha s are parameers o be esmaed. The resuls (Table 5 Panel C) confrm he robusness of he prevous fndngs and herefore sugges ha our orgnal specfcaons do no suffer from seral correlaon...3 Idenfcaon problem We repea our emprcal exercse usng he changes n he defc-gdp rao as an addonal conrol: Y D Y G C C 0 0. (7). 0 0 Y D Y G I I (8) The ncluson of hs varable allows us o conrol for a possble msspecfcaon of he model. In fac o he exen ha changes n governmen revenue and spendng are correlaed our resuls could be capurng he effec of ax changes on he economc acvy. For nsance Cebula (1978) emphaszes he mporance of he governmen sze n producng crowdng-ou effecs. Thus he ncluson of he defc allows us o overcome hs denfcaon problem. Ths approach has however a poenal cavea: f changes n governmen revenue and spendng are no correlaed hen changes n governmen spendng may be correlaed wh changes n defc whch would lead o mulcollneary problems.

9 The resuls are repored n Panel D of Table 5 and once agan confrm he exsence of crowdng-ou effecs. 2. Asymmerc regonal effecs The analyss presened so far has shown robus evdence on he exsence of crowng-ou effecs. Bu s he effec smlar for dfferen regons and counres? To answer hs queson we replcae he esmaons for specfc geographcal areas and counres. Table 6 presens he resuls for egh dfferen areas: ) Afrca ) Asa and Pacfc ) Europe v) Mddle-Eas v) Norh Amerca v) Souh Amerca and Wes Indes v) OECD and v) Developng Counres. We fnd ha he effec vares subsanally beween areas. In parcular whle here are sascally sgnfcan crowdng-ou effecs n Afrca Europe and Souh Amerca governmen spendng does no seem o have (sacally) sgnfcan affecs n he oher areas consdered. We also assess wheher he effec s dfferen beween developed (OECD) and developng counres. The resuls sugges ha he mpac of governmen spendng on boh prvae consumpon and nvesmen s more dermenal n he OECD group. However also among OECD counres here seems o be some degree of heerogeney. In fac analyzng he resuls n Table 7 emerges ha he crowdng-ou effecs of governmen consumpon are larges n relavely less developed counres (such as Mexco and Turkey) and n hose counres wh a hgh share of governmen spendng (such as Fnland Sweden and Norway)..5 Asymmerc effecs over he busness cycle The effec of governmen spendng on economc acvy may also dffer beween dfferen phases of he economc cycles (Pero 200). To address hs ssue we now look a he effecs of governmen spendng on he prvae secor condonng on he nformaon abou he busness cycle. To be more specfc we consruc he followng dummy varables: og = 1 f og>0 and 0 oherwse and og =1 f og<0 and 0 oherwse where og s a measure of oupu gap consruced as he dfference beween our seres and s rend (compued usng he HP fler wh a smoohness parameer equal o 6.25 as suggesed by Ravn and Uhlg2002). 2 The absence of sascally sgnfcance of he esmaed coeffcens n he nvesmen equaon s due o he hgh correlaon (0.15 for he enre sample) beween our explanaory varables. 9

10 Then we nerac he dummy varables wh he change n he rao of governmen spendng o GDP ha s we esmae he followng models: C C w v G G og og u (9) 0 GDP 0 GDP and I I w v G G og og u (10) 0 GDP 0 GDP where and measure he effec of governmen spendng durng upurns and downurns respecvely. Table 8 summarzes he resuls and shows ha he effec of governmen spendng on boh prvae consumpon and nvesmen does no sgnfcanly vary accordng o dfferen phases of he cycles. 3 V. CONCLUSIONS We conrbue o he emprcal leraure on he effec of governmen spendng on economc acvy by assessng he mpac of changes n governmen spendng-gdp rao on (he shor-erm growh raes) of prvae consumpon and nvesmen. We do hs by analysng a panel sample of 15 counres from 1960 o The resuls of our paper sugges ha governmen spendng produces mporan crowdng-ou effecs by negavely affecng boh prvae consumpon and nvesmen. Consequenly he predcons of boh he sandard RBC and IS-LM models canno be aken overall as vald: our resuls are n conras wh he RBC predcon of a rse n nvesmen and wh he IS-LM predcon of a rse n consumpon. We fnd ha he cumulave effec of governmen spendng on prvae consumpon (nvesmen) s abou 1.9 % (1.8 %) of whch abou 1.2 % (0.6 %) s capured by he conemporaneous change n he governmen consumpon-gdp rao and 0.7% (1.2%) by s lags. Ths resul s nerpreed as follows: an ncrease of governmen consumpon by 1% of real GDP mmedaely reduces consumpon (nvesmen) by approxmaely 1.2% (0.6%) wh he declne connung for abou four years when he cumulave decrease n consumpon has reached approxmaely 1.9% (1.8%). The resul s broadly robus o boh counry and me effecs and dfferen economerc specfcaons. 3 The resuls are quanavely unchanged f we use he average growh rae as he measure of rend nsead of he HP rend. 10

11 In addon we show ha he effec of governmen consumpon on prvae consumpon and nvesmen does no depend on he phase of he busness cycle bu dffer subsanally among regons. The dfferenaed effecs of governmen consumpon on prvae consumpon and nvesmen among geographcal areas are exremely mporan and need o be furher nvesgaed. In parcular would be neresng o assess o whch exen he effec of governmen spendng on consumpon and nvesmen depends on polcal and nsuonal varables (e.g. democracy corrupon polcal sably) as well as macro economc varables (ncome neres raes degree of openness). We leave hs challengng avenue for fuure research. References Afonso A. and Sousa R. M. (2011a). The macroeconomc effecs of fscal polcy Appled Economcs forhcomng. Afonso A. and Sousa R. M. (2011b). The macroeconomc effecs of fscal polcy n Porugal: a Bayesan SVAR Analyss Poruguese Economc ournal. 10(1): Ayagar R. Chrsano L. and Echenbaum M. (1990). Oupu employmen and neres rae effecs of governmen consumpon ournal of Moneary Economcs. 30: Argmón. M. González-Páramo. M. and Roldán. M. (1997). Evdence of publc spendng crowdng-ou from a panel of OECD counres Appled Economcs. 29: Aschauer D. A. (1989). Does publc capal crowd-ou prvae capal? ournal of Moneary Economcs. 2: Aukeren E. (2005). Ineracons beween publc and prvae nvesmen: Evdence from developng counres Kyklos. 58(3): Barro R. (1991). A cross-counry sudy of growh savng and governmen n Bernhem D. Shoven. (eds) Naonal savng and economc performance NBER Unversy of Chcago Press Chcago IL: Baxer M. and Kng R. (1993). Fscal polcy n general equlbrum Amercan Economc Revew. 83: Bau O. and Grad E. (2005). Polque budgéare e dynamque économque en France: l'approche VAR srucurel Économe e Prévson : 1 2. Blanchard O. (2003). Macroeconomcs 3rd ed. Prence Hall. 11

12 Blanchard O. and Pero R. (2002). An emprcal characerzaon of he dynamc effecs of changes n governmen spendng and axes on oupu Quarerly ournal of Economcs. 117(): Blundell R. and Bond S. (1998). Inal condons and momen resrcons n dynamc panel daa models ournal of Economercs. 87: Burnsde C. Echenbaum M. and Fsher. (200). Fscal shocks and her consequences ournal of Economc Theory. 115(1): Cebula R. (1978). An emprcal analyss of he crowdng ou of fscal polcy n he Uned Saes and Canada Kyklos. 31(3): Chrsano L. and Echenbaum M. (1992). Curren real busness cycles heores and aggregae labor marke flucuaons Amercan Economc Revew. 82: Coenen G. and Sraub R. (2005). Does governmen spendng crowd n prvae consumpon? Theory and emprcal evdence for he euro area Inernaonal Fnance. 8(3): Edelberg W. Echenbaum M. and Fsher. (1999). Undersandng he effecs of a shock o governmen purchases Revew of Economcs Dynamcs. 2: Easerly W. and Rebelo S. (1993). Fscal polcy and economc growh: an emprcal nvesgaon ournal of Moneary Economcs. 32: Erenburg S.. (1993). The real effecs of publc nvesmen on prvae nvesmen Appled Economcs. 25: Erenburg S.. and Wohar M. E. (1995). Publc and prvae nvesmen: are here causal lnkages? ournal of Macroeconomcs. 17: Faás A. and Mhov I. (2001). The effecs of fscal polcy on consumpon and employmen: heory and evdence CEPR Dscusson Paper Nº Gordano R. Momglano S. Ner S. and Pero R. (2007). The effecs of fscal polcy n Ialy: Evdence from a VAR model European ournal of Polcal Economy Gonzalez-Paramo. M. and De Cos P. H. (2005). The mpac of publc ownershp and compeon on producvy Kyklos. 58(): Grer K. G. and Tullock G. (1989). An emprcal analyss of cross-naonal economc growh ournal of Moneary Economcs. 2: Heppke-Falk K.H. Tenhofen. and Wolff G. B. (2006). The macroeconomc effecs of exogenous fscal polcy shocks n Germany: a dsaggregaed SVAR analyss. Deusche Bundesbank Dscusson Paper Nº

13 Karras G. (199). Governmen spendng and prvae consumpon: some nernaonal evdence ournal of Money Cred and Bankng. 26: Mounford A. and Uhlg H. (2009). Wha are he effecs of fscal polcy shocks? ournal of Appled Economercs. 2(6): Pero R. (200). Esmang he effecs of fscal polcy n OECD counresunversy of Boccon Workng Paper Nº Ravn M. O. and Uhlg H. (2002). On adusng he Hodrck-Presco fler for he frequency of observaons The Revew of Economcs and Sascs. 8(2): Ramey V. and Shapro M. (1998). Cosly capal reallocaon and he effecs of governmen spendng Carnege Rocheser Conference on Publc Polcy. 8: Romer C. and Romer D. H. (2007). The macroeconomc effecs of ax changes:esmaes based on a new measure of fscal shocks NBER Workng Paper Nº

14 Table 1. References. Auhors (Year) Mehodology Counry Effec on Consumpon Effec on Invesmen Afonso and Sousa (2011a) VAR US UK No sgnfcan Negave Germany and Ialy Afonso and Sousa (2011b) VAR Porugal Negave Negave Argmón e al.(1997) Panel OECD Posve Aschauer (1989) Tme seres US - Posve Barro (1991) Cross-counry Developed and - Negave Developng counres Bau and Grard (2005) VAR France Posve Posve Blanchard and Pero (2002) VAR US Posve Posve Burnsde e al. (200) Narrave US No sgnfcan Posve Coenen and Sraub (2005) VAR Euro area Negave - Easerly and Rebelo (1993) Tme seres and - Posve cross counry US and Developed and Developng counres Edelberge al. (1999) Narrave Approach US Negave Posve Erenburg (1993) Tme Seres US - Posve Erenburg and Wohar (1995) Tme Seres US - Posve Faás and Mhov (2001) VAR US Posve No sgnfcan Gordano e al. (2007) VAR Ialy Posve Posve Grer and Tullock (1989) Cross-counry Developed and Developng counres - Posve Hepke-Falk e al. (2006) VAR Germany Posve Posve Karras (199) Tme Seres Developed and Developng counres Posve Mounford and Uhlg (2009) VAR US No Sgnfcan Negave Pero (200) VAR Ausrala Canada Germany and UK Posve No sgnfcan Ramey and Shapro (1998) Narrave Approach US Negave - 1

15 Table 2. Summary Sascs. Varable # Observaons Mean Sandard Devaon Prvae Consumpon (% of GDP) Prvae Invesmen (% of GDP) Governmen Spendng (% of GDP) Prvae Consumpon Growh Prvae Invesmen Growh Change n Governmen Spendng (% of GDP) Table 3. Correlaon Coeffcens. Prvae Consumpon (% of GDP) Prvae Invesmen (% of GDP) Change n Governmen Spendng (% of GDP) Varable Prvae Consumpon (% of GDP) 1 Prvae Invesmen (% of GDP) Change n Governmen Spendng (% of GDP)

16 Table. Effecs of Governmen Spendng on Prvae Consumpon and Invesmen Growh Baselne Model. Panel A. Consumpon Panel B. Invesmen OLS OLS OLS FE RE OLS OLS OLS FE RE G -1.25*** -1.23*** -1.2*** -1.23*** -1.2*** -0.59** -0.62** -0.61** -0.62*** -0.61*** (0.16) (0.17) (0.17) (0.07) (0.07) (0.30) (0.33) (0.32) (0.20) (0.19) Y G * ** -0.0** 1 (0.12) (0.12) (0.13) (0.07) (0.07) (0.31) (0.32) (0.33) (0.20) (0.20) Y 1 G -0.27** -0.2** -0.26** -0.2*** -0.25*** -0.8** -0.57** -0.53** -0.57*** -0.53*** 2 (0.11) (0.12) (0.13) (0.07) (0.07) (0.23) (0.2) (0.2) (0.19) (0.19) Y 2 G -0.17* * -0.15** -0.16** -0.39** -0.9** -0.6** -0.9*** -0.3*** 3 (0.11) (0.12) (0.11) (0.07) (0.11) (0.22) (0.22) (0.22) (0.19) (0.19) Y 3 G (0.09) (0.09) (0.09) (0.07) (0.07) (0.21) (0.23) (0.23) (0.19) (0.18) Y Tme FE Yes No Yes Yes No Yes Counry No Yes Yes No Yes Yes FE R Noes: OLS denoes Ordnary Leas Squares RE refers o Random Effecs. Esmaed sandard errors n parenheses. *** ** and * denoe sascal sgnfcance a he 1% 5% and 10% sgnfcance levels. 16

17 Y 0 Table 5. Effecs of Governmen Spendng on Prvae Consumpon and Invesmen Growh - Robusness Checks. Consumpon Invesmen OLS FE RE GMM OLS OLS FE RE GMM Counry Tme FE Counry FE FE OLS Tme FE G -0.38*** (0.06) -0.36*** (0.06) OLS Tme- Counry FE -0.38*** (0.06) -0.36*** (0.0) -0.37*** (0.03) Panel A. Baselne -0.1*** (0.06) -0.37*** (0.12) -0.*** (0.1) OLS Tme & Counry FE -0.2*** (0.15) -0.*** (0.10) -0.0*** (0.10) -0.5*** (0.12) R Y G -0.12** 1 (0.06) (0.06) (0.06) -0.09** (0.0) -0.11*** (0.0) Panel B. Exogeney -0.10* (0.06) -0.29*** (0.13) -0.37*** (0.1) -0.3*** (0.15) -0.37*** (0.10) -0.3*** (0.10) -0.36** (0.15) R Panel C. Seral Correlaon G -0.1** -0.39** -0.1** -0.39** -0.0** -0.36*** -0.36*** -0.36*** -0.36*** -0.36*** 0 Y (0.06) (0.06) (0.06) (0.03) (0.03) (0.13) (0.13) (0.13) (0.13) (0.13) R Panel D. Idenfcaon Y G -0.3** 0 (0.08) -0.35** (0.09) -0.36** (0.09) -0.35** (0.05) -0.35** (0.10) R Noes: OLS denoes Ordnary Leas Squares RE refers o Random Effecs GMM denoes Generalzed Mehod of Momens. Esmaon mehod Blundell-Bond (1998). Esmaed sandard errors n parenheses. *** ** and * denoe sascal sgnfcance a he 1% 5% and 10% sgnfcance levels. Baselne refers o models (1) and (2). Exogeney refers o models (3) and () where only lagged growh raes of governmen consumpon are consdeor+ed; Seral correlaon refers o models (5) and (6) where lagged growh raes of he dependen varables are ncluded; Idenfcaon refers o model (7) and (8) where curren and lagged growh raes of he budge defc are ncluded (0.19) (0.23) (0.23) (0.13) (0.12) 17

18 Table 6. Effecs of Governmen Spendng on Prvae Consumpon and Invesmen Growh - Asymmerc Regonal Effecs. Afrca Asa and Pacfc Europe Mddle Eas Norh Amerca Souh Amerca and Wes Indes OECD Developng counres Panel A. Consumpon Y G -0.36*** 0 (0.06) (0.16) -0.39*** (0.17) 0.02 (0.19) (0.2) -0.66*** (0.15) -0.59*** (0.07) -0.37*** (0.06) R Panel B. Invesmen Y G (0.22) (0.) -0.9*** (0.23) 0.22 (0.37) -0.7 (0.61) -0.91*** (0.28) -1.50*** (0.20) -0.37*** (0.15) R Noes: We esmae he model usng OLS denoes Ordnary Leas Squares and ncludng boh counry and me effecs. Esmaed sandard errors n parenheses. *** ** and * denoe sascal sgnfcance a he 1% 5% and 10% sgnfcance levels. 18

19 Table 7. Panel A. Effecs of Governmen Spendng on Prvae Consumpon Growh - Asymmeres Across Counres. G Y 0 Ausrala Ausra Belgum Canada Czech Denmark Fnland Republc -0.81* *** -0.** ** (0.5) (0.) (0.17) (0.18) (0.8) (0.70) (0.38) R France Germany Greece Hungary Ireland Ialy apan G Y ** -0.71*** *** (0.17) (0.17) (0.32) (0.60) (0.21) (0.23) (0.16) R Korea Luxembourg Mexco New Zealand Norway Poland Porugal G Y *** -1.36*** ** (0.8) (0.59) (0.98) (0.0) (0.29) (0.17) (0.53) R G Y 0 Slovak Span Sweden Swzerland Turkey Uned Uned Republc Kngdom Saes ** -0.76*** -1.03*** -3.6** -0.68*** (0.31) (0.37) (0.18) (0.25) (1.23) (0.23) (0.18) R Noes: We esmae he model usng OLS denoes Ordnary Leas Squares and ncludng boh counry and me effecs. Esmaed sandard errors n parenheses. *** ** and * denoe sascal sgnfcance a he 1% 5% and 10% sgnfcance levels. 19

20 Table 7. Panel B. Effecs of Governmen Spendng on Prvae Invesmen Growh - Asymmeres Across Counres. G Y 0 Ausrala Ausra Belgum Canada Czech Denmark Fnland Republc *** *** (1.39) (1.05) (1.19) (0.56) (0.88) (1.38) (1.17) R France Germany Greece Hungary Ireland Ialy apan G Y *** ** -2.56*** (0.66) (0.6) (1.13) (0.73) (0.66) (0.36) (0.73) R Korea Luxembourg Mexco New Zealand Norway Poland Porugal G Y * * * (1.57) (3.2) (2.85) (2.23) (1.06) (1.21) (1.62) R G Y 0 Slovak Span Sweden Swzerland Turkey Uned Uned Republc Kngdom Saes * -2.19** ** -1.28** -1.86*** (1.12) (1.20) (0.87) (1.88) (.50) (0.59) (0.53) R Noes: We esmae he model usng OLS denoes Ordnary Leas Squares and ncludng boh counry and me effecs. Esmaed sandard errors n parenheses. *** ** and * denoe sascal sgnfcance a he 1% 5% and 10% sgnfcance levels. 20

21 Table 8. Effecs of Governmen Spendng on Prvae Consumpon and Invesmen Growh - Asymmerc Effecs Over he Busness Cycle. Panel A. Consumpon Panel B. Invesmen OLS OLS OLS FE RE OLS OLS OLS FE RE G -0.1*** -0.39*** -0.0*** -0.39*** -0.1*** -0.38** -0.7** -0.3** -0.7*** -0.3*** og (0.08) (0.08) (0.08) (0.05) (0.05) (0.17) (0.33) (0.19) (0.13) (0.12) 0 Y G Y 0 og -0.35** (0.08) -0.33** (0.08) -0.35** (0.08) -0.33** (0.05) -0.33** (0.05) -0.35** (0.1) -0.1** (0.19) -0.0** (0.19) -0.1*** (0.1) -0.37** (0.13) Tme FE Yes No Yes Yes No Yes Counry FE No Yes Yes No Yes Yes R (0.6) (0.57) (0.75) (0.33) (0.2) (0.89) (0.79) (0.91) (0.73) (0.72) 0 0 (p-value) Noes: OLS denoes Ordnary Leas Squares RE refers o Random Effecs. Esmaed sandard errors n parenheses. *** ** and * denoe sascal sgnfcance a he 1% 5% and 10% sgnfcance levels. og + =1 durng upurns and zero oherwse; og - =1 durng downurns and zero oherwse. 21

22 Appendx. Table A. Counry Sample. Counry ls Albana Croaa ordan Porugal Uruguay Algera Cyprus Kazakhsan Puero Rco Venezuela Argua and Barbuda Czech Republc Kenya Romana Yemen Argenna Denmark Korea Russan Federaon Zamba Armena Dbou Kyrgyz Republc Rwanda Zmbabwe Ausrala Domnca Lava Sao Tome e Prncpe Ausra Domncan Republc Lesoho Senegal Azerbaan Ecuador Luxembourg Seychelles Bahamas Egyp Macao Serra Leone Bangladesh El Salvador Macedona Slovak Republc Barbados Esona Madagascar Slovena Belarus Ehopa Malaw Solomon Islands Belgum Fnland Malaysa Somala Belze France Mal Souh Afrca Benn Gabon Mala Span Bhuan Gamba Maurana Sr Lanka Bolva Germany Maurus S. Ks and Nevs Boswana Ghana Mexco S. Vncen and he Grenadnes Brazl Greece Moldova Sudan Brune Darussalam Grenada Morocco Swazland Bulgara Guaemala Mozambque Sweden Burkna Faso Gunea Namba Swzerland Camboda Gunea-Bssau Neherlands Syran Arab Republc Cameroon Guyana New Zealand Taksan Canada Ha Ncaragua Tanzana Cape Verde Honduras Nger Thaland Chad Hong Kong Ngera Togo Chle Hungary Norway Trndad and Tobago Chna Iceland Paksan Tunsa Colomba Inda Panama Turkey Comoros Indonesa Papua New Gunea Uganda Congo Dem. Rep. Iran Paraguay Ukrane Congo Rep. Ireland Peru Uned Arab Emraes Cosa Rca Ialy Phlppnes Uned Kngdom Côe d Ivore apan Poland Uned Saes of Amerca 22