Revisiting Budget and Trade Deficits in Lebanon: A Critique

Transcription

1 Universiy of Wollongong Research Online Faculy of Business - Economics Working Papers Faculy of Business 2006 Revisiing Budge and Trade Deficis in Lebanon: A Criique H. Marashdeh Universiy of Wollongong A. S. Saleh Monash Universiy Malaysia Publicaion Deails Marashdeh, H and Saleh, AS, Revisiing Budge and Trade Deficis in Lebanon: A Criique, Working Paper 06-07, Deparmen of Economics, Universiy of Wollongong, Research Online is he open access insiuional reposiory for he Universiy of Wollongong. For furher informaion conac he UOW Library: research-pubs@uow.edu.au

2 Universiy of Wollongong Economics Working Paper Series 2006 hp:// Revisiing Budge and Trade Deficis in Lebanon: A Criique Hazem Marashdeh and Ali Salman Saleh WP April 2006

3 Re-visiing Budge and Trade Deficis in Lebanon: A Criique by Hazem Marashdeh School of Economics and Informaion Sysem Faculy of Commerce Universiy of Wollongong, Wollongong, NSW Ausralia, haahm19@uow.edu.au and Ali Salman Saleh * School of Business, Monash Universiy Malaysia No. 2, Jalan Kolej, Bandar Sunway Pealing Jaya, Selangor, Malaysia saleh.asalman@buseco.monash.edu.my Absrac This sudy re-examines he relaionship beween he budge defici and he rade defici in Lebanon. In conras o earlier sudies, we sar by esing for a uni roo in he presence of srucural change using he Innovaional Oulier (IO) model. This sudy also uilizes he newly proposed auoregressive disribued lag (ARDL) approach o examine such a relaionship. The resuls show ha he endogenously deermined imes of he breaks coincide wih observed real evens occurring during he years of Civil War in Lebanon and especially afer he Israeli invasion of Beiru in This sudy finds, as well, ha he rade defici in Lebanon has a long run impac on he budge defici. JEL Classificaion: C13, C22, E62 Keywords: Budge defici, rade defici, srucural break, ARDL and Lebanon * Corresponding auhor

4 I. Inroducion Lebanon has never experienced high budge deficis prior o 1975 as i was always in balance and he governmen had never resored o borrowing. This siuaion deerioraed during he civil war period of , which had a severe effec on he Lebanese economy. As shown in Figure 1, he increase in public deficis in Lebanon occurred afer The larges increases occurred afer 1980 and peaked in During he period , here was marked deceleraion in economic growh and privae invesmen (Eken e al., 1995). Increased governmen expendiure and declining governmen revenues were boh responsible for he seep increase in he public secor deficis 1. However, he Lebanese economy sared he reconsrucion period afer he ending of he civil war in early 1990s. As a resul of rebuilding he infrasrucure and he economy as a whole, here was acceleraion in he growh of governmen capial expendiure, ogeher wih large and expanding curren expendiure. This led o sizable fiscal imbalances. Consequenly, governmen budge deficis increased from 9.2% of GDP in 1993 o 20.6% in 1996 before declining o around 15% in 2003 (BDL, 2003). Figure 1. Lebanese Budge Defici as a percenage of GDP: BD/GDP TD/GDP Sources: Banque du Liban (various years); Minisry of Finance (various years); Eken e al. (1995); Eken and Helbling (1999) and Auhors calculaions. 1 See Saleh and Harvie (2005a) for a deailed analysis of he public secor deficis and deb in Lebanon during he period

5 The main aim of his sudy is o re-examine he relaionship beween he budge defici and he rade defici in Lebanon during he period This relaionship has been examined by Paichis (2004) who used he radiional Augmened Dickey-Fuller (ADF) uni roo es as well as he radiional coinegraion echniques proposed by Johansen (1988, 1991). However, Paichis s sudy could be criicized as follows. Firsly, i did no address he issue of srucural change when dealing wih ime series daa. I is now well known ha convenional uni roo ess such as ADF and Phillips-Perron (PP) ess are biased owards he non-rejecion of he uni roo null hypohesis in he presence of srucural breaks. Secondly, a radiional coinegraion procedure, such as Johansen (1991), is no appropriae for small sample sizes. However, our sudy ess for a uni roo in he presence of srucural change (a an unknown ime of he break). I also uilizes he newly proposed auoregressive disribued lag (ARDL) approach o coinegraion, proposed by Pesaran (1997) and Pesaran e al. (2001). This approach is applicable irrespecive of wheher he underlying regressors are purely I(0) or purely I(1), and performs well for small sample sizes. The res of he paper is organized as follows. Secion 2 presens he heoreical background and mehodology. Secion 3 presens he empirical findings. Conclusions and policy implicaions are discussed in Secion 4. II. Theoreical Background and Empirical Sudies A posiive relaionship beween he governmen budge and rade balance can be seen in he conex of a simple Keynesian open-economy model. In an open economy, gross domesic produc, Y, is he sum of privae consumpion expendiures, C, gross privae domesic invesmen expendiures, I, governmen expendiures, G, and expors, X, over impors, M: Y = C + I + G + X M (1) Alernaively, Y equals privae consumpion expendiures, C, savings, S, and axes, T: Y = C + S +T (2) Subsiuing (2) in (1) and rearranging erms yields: (X-M) = (S-I) + (T-G) (3) 2

6 Equaion (3) suggess ne expors equal privae and public savings. Assuming here is a balanced fiscal budge (T-G = 0) and balanced rade (X-M = 0, ha, is, ne expors are 0), hen (3) suggess ha privae domesic saving equals privae domesic invesmen. This is necessarily he case in a closed economy where domesic invesmen is consrained by domesic saving. However, in an open economy, such a relaionship may no always exis. An economy wih a foreign secor has access o inernaional financial markes. The Mundell-Fleming Model argues ha he increase in he governmen s budge defici could lead o an increase in he rade defici hrough increased consumer spending (Fleming, 1962; Mundell, 1963). Furhermore, he Keynesian absorpion heory argues as well ha an increase in he budge defici would induce domesic absorpion and herefore impor expansion, causing a curren accoun defici (Saleh e al., 2005). There have been exensive empirical sudies esing he views of he above argumens 2. The majoriy of hem used he radiional coinegraion echniques. Sudies by Abell (1990), Tallman and Rosensweig (1991), Piersani (2000), Akbosanci and Tunc (2002) and Saleh e al. (2005) provide suppor for he Keynesian view ha here is a long run relaionship beween he budge defici and curren accoun deficis. In conras, oher sudies such as Evans (1988) and Bachman (1992) found no evidence for he link beween he wo deficis. Hence hese findings neiher suppor he Ricardian Equivalence hypohesis which holds ha he wo deficis are nor relaed. III. Mehodology Innovaional Oulier Models The curren sudy employs Perron s (1997) Innovaional Oulier model (IO2). This allows for he occurrence of boh changes in he inercep and in he slope of he rend funcion. y = μ + β + θdu + γdt + δdt B + αy k 1 + ciδy 1 + e i= 1 (4) 2 For a more heoreical and empirical evidence on he relaionship beween budge deficis and curren accoun deficis, please see Saleh and Harvie (2005b). 3

7 where T B denoes he unknown ime of break (1< T B <T ), DU = 1 if > TB, 0 oherwise, DT = 1() if >, 0 oherwise, DT = 1 if T + 1 TB B = B, 0 oherwise, y is any general ARMA process and e is he whie noise residual erm. The null hypohesis of a uni roo is rejeced if he absolue value and he -saisics for esing α = 1 is greaer han he criical value. In order o deermine he imes of he break endogenously a an unknown break poin, wo mehods have been suggesed by Perron (1997); he firs mehod is ha he ime of he break T b is seleced as he value, which minimizes he -saisic for esingα = 1. The reason for his selecion is o make i more likely o rejec he null hypohesis of α = 1 (Wilson, 2004, p.16). The second mehod is ha T b is seleced as o minimize eiher he -saisic on he parameer associaed wih he change in he inercep ( θ ), or he -saisic on he change in he slope ( γ ), or boh of hem. Augmened Disribued Lag (ARDL) Approach o Coinegraion. This sudy uilizes he newly proposed auoregressive disribued lag (ARDL) approach. The error correcion represenaion of he ARDL model is given as follows: Δ Δ k i ki a 2Δ ln BD i + a3δ ln TD i + a 4 ln BD 1 + a5 TD 1 + a6 DumBD + i= 1 i= 1 ln ε (5) BD = a1 + ln k i i k ln TD = b1 + b2 Δ ln TD i + b3δ ln BD i + b4 ln TD 1 + b5 ln BD 1 + b6 DumTD + ε (6) i = 1 i = 1 where, ln BD and ln TD are he naural log of he Budge and rade deficis in Lebanon, a 2, a3, b2, b3 are he shor-run dynamic coefficiens of he wo underlying ARDL models represened by equaions (5) and (6), while a 4, a5, b4, b5 are he longrun coefficiens. The ARDL procedure involves wo sages. In he firs sage, he null hypohesis of no coinegraion for equaion (5) defined as H a = a 0 is esed 0 : 4 5 = agains he alernaive of H a a 0. Also, he null hypohesis of no 1 : 4 5 coinegraion for equaion (6) defined as H b = b 0 is esed agains he 0 : 4 5 = alernaive of H b b 0. These null hypohesises are compued using he 1 : 4 5 familiar F-saisic. The calculaed F-saisic is compared wih criical value abulaed by Pesaran and Pesaran (1997). If he compued F-saisics is greaer han he upper 4

8 bound criical value, hen we rejec he null hypohesis of no coinegraion and conclude ha here exiss seady sae equilibrium beween he variables. If he compued F-saisics is less han he lower bound criical value, hen we can no rejec he null of no coinegraion. If he compued F-saisics falls wihin he lower and upper bound criical values, hen he resul is inconclusive. In his case, following Kremers, e al. (1992) he error correcion erm will be a useful way of esablishing coinegraion. The second sep is o esimae he long-run coefficiens of he same equaion and he associaed ARDL error coercion models. The second sep is o esimae he long-run and he shor-run coefficiens for boh equaions. IV. Empirical Resuls Srucural Change This paper used annual ime series daa for he period of The yearly daa was obained from IMF, Inernaional Financial Saisics Yearbook (various years) and Banque du Liban (various years).two variables are used in his sudy, he budge defici (BD) and he rade defici (TD). These wo variables were measured in erms of heir raios o gross domesic produc. The resuls of he Perron (1997) IO2 model (equaion 4) are repored in able 1. They show ha he ime series of budge defici as a percenage of he GDP (BD/GDP) is saionary as he αˆ is more han he criical values. However, he ime series of he rade defici as a percenage of GDP (TB/GDP) is non-saionary as he αˆ is less han he criical values. The ime of he break for he budge defici occurred during 1983 afer he Israeli invasion in 1982 which had a severe impac and was a major urning poin for he Lebanese economy. The invasion resuled in massive desrucion in he counry in erms of infrasrucure, loss of human resources ec. Therefore, he year 1983 was associaed wih massive fiscal deficis (24 % of oal expendiure) and he governmen spen a massive amoun on miliary services and weapons which regisered US$800 million during his ime (Al-Khalil, 1991). In he case of he rade defici, i is found ha he ime of he break occurred in A ha ime Lebanon was in he middle of civil war and boh he fighing and violen became unconrolled in he capial ciy Beiru. This had a severe impac on he Lebanese economy and is infrasrucure. Moreover, i severely affeced he expor secor in Lebanon during ha ime. 5

9 Table 1. IO model wih a change in inercep and slope Variables Time of he break ( T b ) k β θ γˆ α αˆ Inference BD/GDP Saionary TB/GDP Nonsaionary * The resuls are significan a 5%. Criical values = -6.07, and 4.94 for 1%, 5% and 10%, respecively. Auoregressive Disribued Lag (ARDL) The ARDL model requires a priori knowledge or esimaion of he orders of he exended ARDL. This appropriae modificaion of he orders of he ARDL model is sufficien o simulaneously correc for residual serial correlaion and he problem of endogenous regressors. The order of he disribued lag on he dependen variable and he regressors is seleced using he Akaike Informaion Crierion (AIC), as i is known for selecing he maximum relevan lag lengh (Pesaran and Shin, 1998). As has been menioned before, a significan F-saisic for esing he join level significance of he lagged level indicaes he exisence of a long-run relaionship. When he budge defici is he dependen variable, he calculaed F-saisic equals o This value is greaer han he lower bound and lower han he upper bound. This resul indicaes ha we have an inconclusive oucome since he calculaed F- saisics is less han he upper bound criical value bu greaer han he lower bound. In his case, following Kremers, e al. (1992), we coninue wih he ARDL procedure and our decision on wheher he long-run relaionship exiss or no depends on he significance of he error correcion erm. When rade defici is dependen variable, he calculaed F-saisic equals 3.424, which is lower han he lower bound. This indicaes ha here is no long-run relaionship when rade defici is dependen variable. However, we move o he second sage, and esimae equaion (5). This esimaion shows he long-run and shor-run relaionships beween he budge defici and he rade defici in Lebanon over he las hree decades. 3 The relevan criical values are obained from Pesaran and Pesaran (1997). The criical values in he case of one regressor are a he 90% significance level. 6

10 The resuls repored in Table 2 show ha he rade defici in Lebanon has a longrun impac on he budge defici over he period ( ). An increase in he rade defici by 1% will have a significan long-run impac on he budge defici by 0.22%. Also, he dummy variable is found o be saisically significan a 5% level. This indicaes ha he srucural break happened in 1983 as a resul of he Israeli invasion had a significan long-run impac on he budge defici in Lebanon during he same period. The model also passes diagnosic ess, supporing he overall validiy of he shor run model. Table 2. Long-Run Resuls from he ARDL Model Dependen Variable: BD/GDP Regressors Coefficien Sandard Error T-Raio Inercep TT/GDP * DumBD ** Diagnosic Tess 2 R =0.53 F [0.000] ( 3.29) = Diagnosic Tess : Serial Correlaion: F ( 1,28) = [0.886] Funcional Form: F ( 1,28) = [0.635] Heeroscedasiciy: F(1,31) = [0.306] * Significan a 10% level, ** Significan a 5% level The shor-run coefficiens and he error correcion erm (ECT) are repored in Table 3. These resuls show ha he error correcion erm is saisically significan a 5% wih he expeced negaive sign. The ECM represens he speed of adjusmen of he change in he budge defici o is long-run equilibrium following a shock. Moreover, he significance of he ECM confirms he exisence of a sable long-run relaionship beween he significan regressors and he dependen variable. The ECM, in Table 3, suggess ha 73% of he adjusmen back o long-run equilibrium is correced afer one year. The large magniude of he coefficien of he ECM suggess ha once shocked, convergence o equilibrium is very fas. 7

11 Table 3. Shor-Run Resuls from he ARDL Model Dependen Variable: ΔBD/GDP Regressors Coefficien Sandard T-Raio Error Inercep ΔTD/GDP * DumBD ** ECM ** * Significan a 10% level, ** Significan a 5% level V. POLICY IMPLICATIONS AND CONCLUSIONS Our sudy re-examined he work of Paichis (2004) on budge and rade deficis in Lebanon using he convenional coinegraion echnique. We found by using he IO2 model ha he ime of srucural breaks for boh variables, he rade defici and he budge defici, occurred during 1980 and 1983 respecively, which is consisen wih he years of he civil war in Lebanon and he pos era of he Israeli invasion of Beiru in In addiion, we found as well by using he ARDL model ha he rade defici in Lebanon has a long run impac on he budge defici. This empirical finding is quie consisen wih previous sudies. The empirical resuls in his paper suppor he Keynesian view ha here is a srong linkage beween budge defici and rade defici during he period of Using he ARDL approach and he Bounds es, our resuls show ha he budge defici and rade defici have a posiive significan relaionship in he long run. The above resuls are no surprising for he Lebanese economy, as from he counry was under he impac of he civil war. Much of he counry s economy was deerioraed in he process. To finance governmen expendiure and he war during he civil war period, governmen borrowed from boh inernal and exernal sources. Therefore, he win-defici problem susained during he pos war period as well, because of increase in impor of capial goods (among consumpion goods), mainly o rebuild he counry s infrasrucures and economic developmen. The rade deficis in he counry were mainly financed from domesic and oher exernal resources. As such, boh he domesic and exernal public deb has been on an upward rend. The increase in naional deb over he period also increased he cos of servicing he naional deb, resuling in he governmen running larger budge defici in his period. 8

12 The empirical resuls from his paper have imporan policy implicaions in erms of managing effecively he win-defici problem. From his sudy, we indicae ha sabilising he rade defici problem could assis in managing he budge defici problem in Lebanon. In his case, our resuls provides a view ha a policy measures which can be aken o reduce he rade defici could assis in reducing he budge defici in Lebanon. Furhermore, Lebanon economy depends heavily on he service secor such as ourism, finance, among ohers. Many of he services are exporable, and have he poenial o conribue huge sums in foreign currency. Policies and sraegies o srenghen he supply side and demand side in hese key secors menioned above will improve he erms of rade in favour of Lebanon. In addiion o ha, incenives (e.g. ax credis for R&D) should be provided o improve he level of innovaion among indusries and small medium enerprises (SME) in he counry. Policies o arac foreign direc invesmen (FDI) in key secors of he economy should be enhanced by inroducing various fiscal incenives such as lowering corporae ax in he counry. References Abell, John D., (1990), The Role of he Budge Defici During he Rise in he Dollar Exchange Rae from , Souhern Economic Journal, Vol. 57, No. 1, pp Akbosanci, E. and Tunc (2002) Turkish Twin Deficis: An Error Correcion Model of Trade Balance, ERC Working paper no: 01/06. Al-Khalil, Y. (1991), Economic Developmens in Lebanon Since 1982, Discussion Paper, No. 3, Lebanese Cenre for Policy Sudies, LCPS, Beiru, pp Bachman, D. (1992), Why is he US Curren Accoun Defici So Large? Evidence From Vecor Auoregressions, Souhern Economic Journal, Vol. 59, pp Banque du Liban, BDL (various years), Monhly, Quarerly and Yearly Repors, Beiru. Banque du Liban, BDL (2003), Annual Repor, Beiru, Lebanon. Eken, S., Cashin, P., Erbas, S. N., Marelino, J., and Mazarei, A. (1995), Economic Dislocaion and Recovery in Lebanon, IMF Occasional Papers, No. 120, Washingon, DC. 9

13 Eken, S. and Helbling T. (eds.), (1999), Back o Fuure: Reconsrucion and Sabilizaion in Lebanon, IMF Occasional Papers, No. 176, Washingon, DC. Evans, P. (1988), Do Budge Deficis Affec he Curren Accoun? Unpublished, Ohio Sae Universiy, Ohio. Fleming, J. M. (1962), Domesic Financial Policies Under Fixed and Under Floaing Exchange Raes, Inernaional Moneary Fund Saff Papers, No. 10, pp Inernaional Moneary Fund, IMF (Various Years), Inernaional Financial Saisics Yearbooks, Washingon DC. Johansen, S. (1988), Saisical analysis of coinegraing vecors, Journal of Economic Dynamics and Conrol, 12, Johansen, S. (1991), Esimaion and Hypohesis Tesing of Coinegraing Vecors in Gaussian Vecor Auoregressive Models, Economerica, Vol. 59, pp Kremers, J. J. M., Neil R. Ericsson and Juan J. Dolado, (1992), The Power of Coinegraion Tess, Oxford Bullein of Economics and Saisics, 54, pp Minisry of Finance, Annual Repor, (various years), Beiru. Mundell, R. A. (1963), Capial Mobiliy and Sabilizaion Policy Under Fixed and Flexible Exchange Raes, Canadian Journal of Economics and Poliical Science, Vol. 29, pp Paichis, C. (2004), Budge and Trade Deficis in Lebanon, Applied Economerics Leers, Vol. 11, pp Perron, P. (1997), Furher Evidence on Breaking Trend Funcions in Macroeconomic Variables, Journal of Economerics, Vol. 80, No. 2, pp Pesaran, M.H., (1997), The Role of Economic Theory in Modelling he Long Run, The Economic Journal, 107, Pesaran, M.H., Shin, Y. and Smih, R.J. (2001), Bounds Tesing Approaches o he Analysis of Level Relaionships, Journal of Applied Economerics, 16, Piersani, G. (2000), Curren Accoun Dynamics and Expeced Fuure Budge Deficis: Some Inernaional Evidence, Journal of inernaional Money and Finance, Vol.19, No. 2, pp

14 Saleh, A. S., Nair, M., and Agalewaee, T. (2005), The Twin Deficis Problem in Sri Lanka: An Economeric Model, Souh Asia Economic Journal, Vol. 6, No. 2, pp Saleh, A. S. and Harvie, C. (2005a), An Analysis of Public Secor Deficis and Deb in Lebanon: , Middle Eas Review of Inernaional Affairs (MERIA), Vol. 9, No. 4 (December), pp Saleh, A. S. and Harvie, C. (2005b), The Budge Defici and Economic Performance: A Survey, Singapore Economic Review, Vol. 50, No. 2, pp Tallman, W. and Rosensweig (1991), Invesigaions U.S. Governmen and Trade Deficis, Economic Review (Federal Reserve Bank of Alana), pp Wilson, Edgar, (2004), Noes on he Consequences of Uni Roo Processes and Tess for Saionariy, presened in he economeric workshop held on he 1 s of November 2004, Economic Discipline, Universiy of Wollongong. Ausralia. 11