Public Investment and Growth in the Eastern Caribbean

Transcription

1 WP/7/124 Public Invesmen and Growh in he Easern Caribbean Shaun K. Roache

2

3 27 Inernaional Moneary Fund WP/7/124 IMF Working Paper Wesern Hemisphere Deparmen Public Invesmen and Growh in he Easern Caribbean Prepared by Shaun K. Roache 1 Auhorized for disribuion by Paul Cashin May 27 Absrac This Working Paper should no be repored as represening he views of he IMF. The views expressed in his Working Paper are hose of he auhor(s) and do no necessarily represen hose of he IMF or IMF policy. Working Papers describe research in progress by he auhor(s) and are published o elici commens and o furher debae. This paper quanifies he effec of public invesmen on growh in he ECCU. The resuls, emerging from panel vecor auoregressions, indicae ha he reurn on public invesmen, as defined by Perreira (2), is very likely negaive. This means ha he oal change in real oupu induced by one EC dollar of public invesmen, due o is shor-run impac on demand, or he longer-run impac on supply, is below one EC dollar. Public invesmen shocks also appear o appreciae he real exchange rae, suggesing ha he shor-run demand impac is larger han he long-run supply response. JEL Classificaion Numbers: E22, H54, 16, 54 Keywords: Public invesmen, growh, Caribbean Auhor s Address: sroache@imf.org 1 The auhor is graeful o Paul Cashin, Mario Dehesa, colleagues in he Caribbean I Division, and seminar paricipans in he Wesern Hemisphere Deparmen of he Inernaional Moneary Fund for helpful commens and suggesions.

4 2 Conens I. Inroducion...3 II. Lieraure Review...3 III. Daa Descripion and Trends IV. Esimaion...5 A. Panel Vecor Auoregressions Pros and Cons for he ECCU...5 B. Selecion of Endogenous Variables...6 C. Exogenous Variables OECD Growh, Aid, Naural Disasers and Elecions...6 D. Model Specificaion and Esimaion...7 E. Impulse Responses...7 V. Resuls...7 A. The Effecs of Aid Flows...7 B. The Impac of Public Invesmen on Growh...8 C. Granger Causaliy Tes Public Invesmen on Oupu...1 D. The Impac of Growh on Public Invesmen...1 E. The Impac of Public Invesmen and Growh on he Bilaeral RER...11 F. Effecs of Naural Disaser and Elecions...11 VI. Conclusions...11 Tables 1. ECCU: Rae of Reurn on Public Invesmen ECCU: Rae of Reurn Compared...1 Appendices I. ECCU Public Invesmen Model: Summary Saisics...12 II. Correlaion Marix of Model Variables...12 III. VAR Coefficiens on Naural Disaser and Elecion Dummy Variables...12 Figures 1. ECCU: Public Invesmen, Growh and Aid Inflows, ECCU: Public Secor Invesmen, ECCU: Impulse Response for GDP and he Bilaeral Real Exchange Rae o One Sandard Deviaion Public Invesmen and GDP Shocks ECCU: Accumulaed Impulse Responses o a One Sandard Deviaion Shock o Public Invesmen Accumulaed Impulse Responses o a One Sandard Deviaion Shock o GDP Accumulaed Impulse Responses of he Bilaeral RER o a One Sandard Deviaion Shock o GDP and Public Invesmen...18 References...19

5 3 I. INTRODUCTION Public invesmen has a high profile in he small islands of he Easern Caribbean Currency Union (ECCU). 2 In common wih many small saes, i accouns for a large share of GDP, averaging around 1 percen, compared o 3 percen for OECD counries (Figure 1). In par his no only reflecs he limis of economies of scale in he provision of public goods, bu also he regional percepion ha public invesmen is one of he main caalyss for economic growh and developmen. This paper quanifies he effec of public invesmen on growh and he real exchange rae in he ECCU. Avoiding he imposiion of oo many priors on he daa, an aheoreic approach is used. This has is drawbacks; he model will provide no informaion on why he resuls emerge as hey do. However, his approach also has advanages. I allows us o make inferences based on pas invesmen performance. I does so while avoiding imposing heoreical resricions on an invesmen-growh process ha, due in par o daa limiaions, is poorly undersood. The resuls, emerging from panel vecor auoregressions indicae ha he rae of reurn on public invesmen, as defined by Pereira (2), is very likely negaive. This means ha he oal change in real oupu induced by one EC (Easern Caribbean) dollar of public invesmen, due o is shor-run impac on demand, or he longer-run impac on supply, is below one EC dollar. This is no he only (or perhaps he bes) way o measure he reurns on public invesmen. However, i allows us o compare he resuls of he ECCU o hose of oher counries sudied in he lieraure. The plan of his paper is as follows: Secion II provides a selecive lieraure review. Secion III describes he daa. Secion IV deails he esimaion procedures. Secion V presens he resuls, and Secion VI provides a brief conclusion. II. LITERATURE REVIEW Early work on he growh effecs of public invesmen used he single-equaion saic producion funcion approach and aggregae daa. In is mos basic form, his mehod assumes a producion funcion in which oupu is a funcion of public capial K G, privae capial K P, labor N, and echnology A: G P ( A, K, K N ) Y = F, (1) Assuming a generalized Cobb-Douglas form, and allowing lower case variables o denoe logs, and ε o represen he elasiciy of oupu wih respec o facor i, his can be wrien as: 2 In his paper, he ECCU refers o he six independen saes of Anigua and Barbuda, Dominica, Grenada, S. Kis and Nevis, S. Lucia, and S. Vincen and he Grenadines.

6 4 y = a + ε k + ε k + ε n (2) G G P P N Celebraed early resuls emerged from Aschauer (1989a, 1989b) wih public capial seemingly having a powerful influence on oupu growh and produciviy in he Unied Saes. Using a variaion on equaion (2), one specificaion used he produciviy of privae capial as he lef-hand side variable, wih capaciy uilizaion represened by cu: G P P ( k k ) + a3 ( n k ) + a cu u P y k = a + a1 + a2 4 + (3) Subsequen lieraure, surveyed exensively by Munnell (1992), presened more ambiguous resuls; in many cases, growh effecs were absen. Oher noable examples include Khan and Kumar (1997), which used he neo-classical growh model, a cross-secion of 95 developing counries and decade-long sample periods. Per capia growh raes were regressed on iniial GDP per capia, populaion growh and echnical change, and invesmen raes for he public and privae secors. Resuls sugges ha while boh ypes of invesmens have a posiive and significan effec on growh, privae invesmen ended o have a larger impac. Kavanagh (1997) focuses on Ireland, ofen cied as a role model for small saes, over he period. Using a producion funcion approach and error-correcion mehodology, resuls suggesed ha public capial had a posiive, bu insignifican effec on oupu per capia. There are hree criicisms of he srucural modeling of aggregae producion funcions approach. Firs, here remain ousanding economeric problems, such as non-saionariy in he possible absence of coinegraion and endemic endogeneiy (see Jorgenson, 1991). Second, i ignores he dynamic relaionship beween public invesmen and growh. Third, research based on aggregae daa does no allow for he varying effecs of differen ypes of public capial sock; public invesmen in infrasrucure and social services may impac growh in very differen ways, boh in erms of scale and iming. More recen research aemps o address hese issues, wih much use made of vecor auoregressions (VARs), a field comprehensively surveyed by Kamps (24). Noable conribuions include Pereira (2), which focusing on he Unied Saes, uses daa on aggregae and five specific ypes of public invesmen, oupu, privae invesmen, and privae employmen. The esimaed long-run effec on oupu of a one percen, one-ime random shock o aggregae public invesmen was.4, wih variaon according o he ype of capial. Pereira and Faima-Pinho (26) used a similar VAR framework for he welve euro-area counries. The same long-run effecs on oupu were esimaed o average.6 and ranged from -.2 in he Neherlands o.2 in Ialy. Kamps (24) assesses public invesmen in 22 OECD counries. Using vecor errorcorrecion models and recursive idenificaion prcedures, impulse responses showed ha he effec of public capial on oupu is posiive in he long-erm, bu wih very large sandard errors. The average long-run elasiciy of oupu o a public invesmen shock is.12, alhough his varies widely across counries.

7 5 III. DATA DESCRIPTION AND TRENDS Esimaions use annual daa from 1975 o 25. Public invesmen refers o capial expendiure underaken by cenral governmen, public enerprises, and sauory corporaions. The sources of he daa are IMF saff repors and World Bank Counry Economic Memorandums; his laer source proved paricularly imporan for he pre-1995 period in which he Bank played an acive role in developing public secor invesmen plans (PSIPs). In almos all cases, hese are aken from he presenaion of he fiscal accouns, where i was no possible o idenify he ype of invesmen spending. Even when he World Bank helped develop he PSIP, he secoral breakdown of invesmen daa is on an approved and proposed basis only; in many cases, execuion of he PSIP may be quie differen from ha proposed in he PSIP or he budge. Public invesmen was deflaed using he average OECD deflaor of capial expendiure for he federal governmen (excluding defence), and sae and local governmens. Summary saisics are presened in Table 1. Figure 2 shows he raio of public invesmen o GDP for he six ECCU counries (in curren EC dollars). There are a leas wo ineresing poins. Firs, his raio is relaively high. Second, his raio is volaile, wih he average for he ECCU showing a sandard deviaion of 3 percen (compared o.3 percen for he OECD average). Boh he log of real GDP and he bilaeral RER are non-saionary. The evidence ha public invesmen is also I(1) is weaker, bu his appears o be due o he effec of one counry in he panel. Tess for coinegraion, based on Pedroni (1999) were mixed, bu in he mos general specificaions of he es here was lile evidence of panel coinegraion (resuls of all uni roo ess are available upon reques). IV. ESTIMATION A. Panel Vecor Auoregressions Pros and Cons for he ECCU The model used in his paper is a panel vecor auoregression (PVAR). A reduced-form VAR enails cerain limiaions on he conclusions one may draw from he resuls. I is an aheoreic mehod ha allows one o draw some inferences regarding he daa-generaing process, bu i provides lile guidance on he underlying economic relaionships. The advanages ouweigh he disadvanages when assessing public invesmen in he ECCU, for four main reasons. Firs, many srucural models rely on he exisence of a fairly rich daa se, including esimaes of he public capial sock. While i is possible o derive esimaes for he ECCU (e.g. Roache, 26), he high incidence of naural disasers inroduce huge uncerainy relaed o he rae of depreciaion (or desrucion) in paricular years. Second, popular srucural approaches end o model oher facor inpus, such as labor; he daa for which simply does no exis in mos of he ECCU counries. Third, a PVAR explicily models dynamics, useful when considering he shor- o medium-run impac of public invesmen, paricularly in small economies in which he governmen plays a large role. Finally, and more generally, a PVAR approach avoids he horny issue of endogeneiy.

8 6 Using a panel VAR improves he degrees of freedom of he resuls, bu obviously imposes he resricion ha he dynamics (bu no necessarily he fixed-effecs) are homogenous across he ECCU counries. Almos all previous sudies have ended o use single-equaion VARs. In his case, he panel approach was used on he assumpion ha he role of he public secor in he ECCU economies is relaively similar. B. Selecion of Endogenous Variables Three endogenous variables were used in he PVAR real GDP, real public invesmen, and he bilaeral real exchange rae (RER) wih he Unied Saes. This small lis of domesic variables was dicaed by reliable daa availabiliy. The selecion of he RER was moivaed by he possible relaionships beween public invesmen and compeiiveness, of which here are many. In he shor erm, a large rise in public invesmen could cause he inernal real exchange rae o appreciae; since mos physical capial is impored, he main price effec would be hrough greaer demand for wages and a higher price for nonradable goods. In he long run, a higher public capial sock could encourage more privae invesmen, boosing he domesic supply of radables and nonradables and allowing he inernal RER o depreciae. Conversely, Balassa-Samuelson effecs migh dominae, wih higher radable secor produciviy leading o an appreciaed RER. C. Exogenous Variables OECD Growh, Aid, Naural Disasers and Elecions The OECD growh rae was used as a conrol variable, conemporaneously and wih a number of lags equal o hose of he PVAR. Aid flows were also included as an exogenous variable o assess he effec of changes in exernal donor assisance. This is measured as explici aid, raher han he implied benefis obained from preferenial rade agreemens, an area explored in Mlachila and Cashin (27). These daa were deflaed by he invesmen deflaor, since i was assumed ha mos aid flows are used for invesmen projecs raher han curren spending. The model was run wih and wihou he aid variable. The inclusion of aid in he model may be conroversial and likely raises as many quesions as i answers. For example, is aid endogenous o any of he domesic variables, including he incidence of naural disasers? Correlaions sugges ha aid has only a weak conemporaneous relaionship wih he oher variables. Also, a raf of panel regressions in which aid flows are on he lef-hand side and domesic variables, including naural disaser dummy variables, are on he righ-hand side suggesed no relaionship beween domesic variables, including naural disasers. Exogenous domesic variables include dummy variables for boh naural disasers and elecions. Naural disasers ha affeced eiher a leas 25 percen of he populaion or caused more han 25 percen of GDP in damage were idenified from he EMDAT daabase. Dummies were included for he year he disaser hi, and he following wo years o conrol for he effec of disaser-relaed reconsrucion, whose growh effecs may be very differen from regular public invesmen. There has been research suggesing ha public spending in he ECCU may be affeced by he elecion cycle (Duagupa and Tolosa, 26). To conrol

9 7 for his, dummies for he year before and he year of naional elecions in each counry were also included. D. Model Specificaion and Esimaion To allow for heerogeneiy, a fixed-effecs model is esimaed. As is well known, i is inappropriae o rea individual effecs as consans o be esimaed for a dynamic model (Nickell, 1981). Alhough he bias goes o zero as he ime dimension ges very large, Judson and Owen (1996) show his bias may remain significan even wih a ime dimension of 3, close o ha used in his paper. Following Holz-Eakin, Newey, and Rosen (1988), he model is esimaed in he firs difference of growh raes, o sweep away he fixed effecs. Denoing he (3x1) vecor of endogenous variables (growh raes) as x, he OECD growh rae as w, and he vecor of dummies as f, his model may be wrien as: m Δx = a + a Δx + γ Δw + Θf + υ l= 1 l l m l= 1 l l (4) The vecor of insrumenal variables z ha is available o idenify he model is: [, x, K, x, w,, w f ] z =, (5) K 1 The model was esimaed using he Generalized Mehod of Momens (GMM) and a weighing marix ha conrolled for heeroscedasiciy of an unknown form. The se of insrumens was chosen o ensure ha he es of overidenifying resricions was saisfied. The lag lengh of wo was chosen on he basis of log-likelihood ess and, a he margin, informaion crieria. E. Impulse Responses The ordering used o obain orhogonalized impulse responses was ha public invesmen leads GDP, which leads he RER. This ordering is ypical of he lieraure and was jusified by Pereira (2) on he basis ha: (i) shocks o economic growh ake ime o feed hrough he policy funcion influencing public invesmen; (ii) public invesmen has an immediae impac on growh. These assumpions are all fairly unconroversial. Impulse response sandard errors were derived using Mone Carlo simulaions, as oulined in Hamilon (p.337, 1994). V. RESULTS A. The Effecs of Aid flows Aid, as noed above, is included in he sudy as a conrol variable; i is no he inenion here o assess he effec of aid on growh. As migh be anicipaed from simple correlaions, he inclusion of aid flows had lile effec, only marginally reducing he public invesmen muliplier on growh.

10 8 B. The Impac of Public Invesmen on Growh Impulse responses sugges ha public invesmen has a posiive bu shor-run impac on growh (see Figures 3 and 4). Given he assumpion of no coinegraion, he model by definiion implies ha public invesmen canno permanenly affec he growh rae, bu can permanenly affec he level of GDP. The effec of a one-ime random one-sandard deviaion shock on growh effecively dies ou afer 4 years. This suggess ha public invesmen generaes a weak invesmen response from he privae secor, given he likely lags involved in privae secor decisions. In conras, he major effec of public invesmen on growh is direc (in he sense ha i booss domesic demand), wih relaively weak mulipliers. To measure he impac on GDP, we adap he rae of reurn meric suggesed by Pereira (2). The firs elemen in his calculaion is he accumulaed percenage poin impac on he level of GDP of a one-ime random shock o real public invesmen. Abusing erminology (given he coinegraion resuls), his is ermed he long-run elasiciy (σ) as in Pereira (2). σ = Y I Y I (6) To conrol for he size of public invesmen in he economy, his elasiciy σ is divided by he average public invesmen-o-gdp raio since 199 (I / Y) A. Since 199, his raio has exhibied saionariy, and he sample from which he average is chosen has lile effec. This obains he marginal produc of invesmen (MP): MP Y = σ I A = dy di (7) To make he calculaion simple, assume a one percen shock o public invesmen. The marginal oupu may hen be assumed o flow over some realisic projec horizon. In his case, a 2-year horizon wih a consan oupu flow, ha is (dy / 2) per period. The rae of reurn on public invesmen (r) is hen he discoun rae ha solves he presen value problem (5): 1 = 2 dy ( r) = 1 1+ (8) The able below applies his mehod o he resuls from he esimae panel-var and he hisorical ECCU daa. The baseline model is he wo-lag model including OECD growh as an exogenous regressor. Two oher specificaions, including aid flows and a ime rend, are shown.

11 9 ECCU: Rae of Reurn on Public Invesmen Long-run Marginal Annual Rae of Elasiciy 1/ Produciviy 2/ Reurn 3/ Baseline model Upper bound 4/ Cenral endency Lower bound 4/ Including aid flows as an exogneous regressor Cenral endency Including aid flows and a ime rend Cenral endency Source: Auhor's calculaions. 1/ Accumulaed log-change in he level of real GDP divided by he accumulaed log-change in he level of real public invesmen due o a one-ime one-percen shock in real public invesmen. 2/ Calculaed as he long-run elasiciy divided by he public invesmen - GDP raio and inerpreed as he long-run real EC dollar change in real GDP divided by he long-run change in real public invesmen. 3/ This is he inernal rae of average annual percenage rae of reurn implied by he marginal produc of capial over a 2-year period. 4/ Bounds based on one-sandard deviaion confidence inervals on he change in growh and public invesmen. These reurn esimaes are compared in he able below o recen resuls from he lieraure; he rae of reurn esimaed for he ECCU appears o be a he lower end of he range. Noe ha he reurn calculaion is slighly differen han in he able above. To ensure consisency wih he lieraure, he reurn calculaion below is based on he assumpion ha payoff o public invesmen arrives as a lump sum afer 2 years.

12 1 ECCU: Rae of Reurn Compared 1/ Counry/ region Sudy Rae of Reurn ECCU -2. U.S. 2/ Pereira (2) 7.8 Ausria Pereira and Faima-Pinho (26) -6.2 Belgium " -7.9 Finland " 2.7 France " 6.7 Germany " 1.3 Greece " 7.6 Ireland " 6.8 Ialy " 11.4 Porugal " 6. Spain " 3.8 Average (excl. ECCU) 4.5 Sandard deviaion (excl. ECCU) 6.2 Source: Auhor's calculaions; Pereira (2); and Pereira and Faima-Pinho (26) 1/ Rae of reurn calculaed as a "bulle" oupu flow afer 2-years. 2/ Measured in erms of privae oupu. C. Granger Causaliy Tes Public Invesmen on Oupu The convenional way o assess wheher pas changes in public invesmen affec he conemporaneous level of growh is a Granger causaliy es. For all hree specificaions, he hypohesis ha public invesmen did no Granger cause oupu growh could be rejeced a he 1 percen level of significance (using log likelihood ess, wih and wihou small sample adjusmens). These resuls should no be so surprising given ha public invesmen, on average, accouns for over 9 percen of GDP, suggesing ha demand effecs alone could be significan. D. The Impac of Growh on Public Invesmen The effec of a growh shock on public invesmen is posiive (see Figure 5). Higher growh may be supporive for ax and oher revenues and allow he public secor o increase capial expendiure from is own resources. Sandard errors are fairly wide, however. This is unsurprising. Ignoring he effec of naural disasers (which clearly play a role, even in he presence of some conrol variables), public invesmen in hese economies may be influenced by facors ouside of he model. The mos obvious example is he availabiliy of funds, wheher from donors in he form of grans or concessionary loans, or from privae sources.

13 11 E. The Impac of Public Invesmen and Growh on he Bilaeral RER Impulse responses sugges ha he bilaeral RER appreciaes in he firs hree years following posiive public invesmen and growh shocks, bu he sandard errors are very wide (Figure 6). In he long-erm, higher levels of public invesmen appear o have lile effec in reversing his effec, suggesing he shor-run demand impac on prices is much more imporan han he long-run supply effec. F. Effecs of Naural Disaser and Elecions The conrols for naural disasers and elecions worked as anicipaed. The coefficiens on boh ses of dummies had he expeced sign (see Table 3). The effecs on growh were paricularly pronounced, wih growh falling sharply in he year of a disaser and hen rising srongly he following year. VI. CONCLUSIONS Public invesmen plays an imporan role in raising and susaining economic growh in he region. However, public invesmen mus be efficien if i is o have he desired growh effec and reduce he risks ha i will add o public secor indebedness. The evidence from he ECCU over he las 3 years suggess ha public invesmen has had only a emporary and limied growh effec. To he exen ha invesmen is financed by borrowing, his suggess ha public invesmen has had a larger impac on he deb sock han on GDP. This paper does no seek o explain why public invesmen is relaively inefficien. The mehods used only help o make inferences regarding is impac on growh. I also does no address he differenial effecs of various ypes of public invesmen. Aemping o explain why he rae of reurn is low and wha ypes of public invesmen are mos producive in he easern Caribbean remains a ferile area for fuure research.

14 12 APPENDIX Table 1. ECCU Public Invesmen Model: Summary Saisics Observaions Mean Max. Min. Sd. dev. Skew Public invesmen Counry GDP Bilaeral RER OECD GDP Aid flows Source: Auhor's calculaions 1/ All saisics based on he firs difference of he log of real variables muliplied by 1. Table 2. Correlaion Marix of Model Variables 1/ Public Counry Bilaeral OECD Aid invesmen GDP RER GDP flows Public invesmen - Counry GDP.28*** - Bilaeral RER OECD GDP *** Aid flows Source: Auhor's calculaions 1/ Correlaions use firs difference of he log of real variables. *** indicaes ha he correlaion coefficien is significan a he 1 percen level. Table 3. VAR Coefficiens on Naural Disaser and Elecion Dummy Variables Public Real Bilaeral Invesmen GDP RER Naural disaser (same year) coefficien sandard error Naural disaser (following year) coefficien sandard error Elecion (preceding year) coefficien sandard error Elecion (same year) coefficien sandard error Source: Auhor's calculaions.

15 13 Figure 1. ECCU: Public Invesmen, Growh and Aid Inflows, ECCU Public Invesmen (percen of GDP) OECD Public Invesmen, GDP, and he RER (consan prices, 198=1)) GDP RER righ axis Public invesmen Public Invesmen and Aid Inflows (percen of GDP) 25 Public Invesmen and Aid Inflows (percen of GDP) 2 ECCU public invesmen 2 R 2 = ECCU aid inflows Public Invesmen and GDP growh ECCU public invesmen (percen of GDP) ECCU GDP (percen change) aid inflows on horizonal axis GDP growh and Aid Inflows ECCU aid inflows (percen of GDP) ECCU GDP (percen change) Source: Naional auhoriies; Organizaion for Economic Co-operaion and Developmen; and auhor's calculaions.

16 14 Figure 2. ECCU: Public Secor Invesmen, (Percen of GDP) Anigua and Barbuda Dominica year average year average Grenada S. Kis and Nevis year average year average S. Lucia 25 2 S. Vincen and he Grenadines year average year average Source: Counry auhoriies; ECCB; World Bank; IMF; and auhor's calculaions.

17 15 Figure 3. ECCU: Impulse Response for GDP and he Bilaeral Real Exchange Rae o One Sandard-Deviaion Public Invesmen and GDP Shocks 6. Response of GDP o Public Invesmen and GDP shocks 5. Percen Change o GDP o Public Invesmen Years Response of he RER o Public Invesmen and GDP shocks o GDP Percen Change o Public Invesmen Years Source: Auhor's calculaions.

18 16 Figure 4. ECCU: Accumulaed Impulse Responses o a One Sandard Deviaion Shock o Public Invesmen Response of Public Invesmen 1/ Percen Change Years Response of GDP 1/ Percen Change Years Source: Auhor's calculaions. 1/ Sandard error bounds calculaed using Mone Carlo simulaions.

19 17 Figure 5. ECCU: Accumulaed Impulse Responses o a One Sandard Deviaion Shock o GDP Response of Public Invesmen 1/ Percen Change Years Response of GDP 1/ Percen Change Years Source: Auhor's calculaions. 1/ Sandard error bounds calculaed using Mone Carlo simulaions.

20 18 Figure 6. ECCU: Accumulaed Impulse Responses of he Bilaeral RER o a One Sandard Deviaion Shock o GDP and Public Invesmen Response o Public Invesmen 1/ Percen Change Years Response o GDP 1/ Percen Change Years Source: Auhor's calculaions. 1/ Sandard error bounds calculaed using Mone Carlo simulaions.

21 19 REFERENCES Aschauer, D.,1989a, Is Public Expendiure Producive,? Journal of Moneary Economics, 23, pp Aschauer, D., 1989b, Does Public Capial Crowd Ou Privae Capial, Journal of Moneary Economics, 24, pp Duagupa, Rupa, and Guillermo Tolosa, 26, Fiscal Discipline and Exchange Rae Regimes: Evidence from he Caribbean, IMF Working Paper, 6/119, (Washingon D.C.: Inernaional Moneary Fund). Hamilon, James, D., 1994, Time Series Analysis, Princeon Universiy Press, Princeon, New Jersey. Holz-Eakin, D., Newey, W., and H. S. Rosen, 1988, Esimaing Vecor Auoregressions wih Panel Daa, Economerica, 56, pp Jorgenson, D., 199, Fragile Saisical Foundaions, The Public s Capial, pp Judson, R., and A. L. Owen, 1996, Esimaing Dynamic Panel Daa Models: A Pracical Guide for Macroeconomiss, Federal Reserve Board of Governors, Washingon D.C. Kamps, Chrisophe, 24, The Dynamic Macroeconomic Effecs of Public Capial Theory and Evidence for OECD Counries, Springer-Verlag, Berlin. Kavanagh, Caherine, 1997, Public Capial and Privae Secor Produciviy in Ireland, , Journal of Economic Sudies, Vol. 24, No. l/2, pp Khan, Mohsin S., and Manmohan S. Kumar, 1997, Public and Privae Invesmen and he Growh Process in Developing Counries, Oxford Bullein of Economics and Saisics, Vol. 59, No. 1, pp Mlachila, Monfor, and Paul Cashin, 27, The Macroeconomic Impac of Trade Preference Erosion on he Windward Islands, Easern Caribbean Currency Union: Seleced Issues, IMF Counry Repor No. 7/97, pp (Washingon D.C.: Inernaional Moneary Fund). Munnell, A., 1992, Infrasrucure Invesmen and Economic Growh, Journal of Economic Perspecives, 6, pp Nickell, Sephen, 1981, Biases in Dynamic Models wih Fixed Effecs, Economerica, 49, pp

22 2 Pereira, Alfredo M., 2, Is All Public Capial Creaed Equal?, Review of Economics and Saisics, Vol. 82, No. 3, pp Pereira, Alfredo M., and Maria de Faima-Pinho, 26, Public Invesmen, Economic Performance, and Budgeary Consolidaion: VAR Evidence for he 12 Euro Counries, College of William & Mary, Working Paper No. 4. Roache, Shaun K., 26, Domesic Invesmen and he Cos of Capial in he Caribbean, IMF Working Paper 6/152 (Washingon D.C.: Inernaional Moneary Fund).