Tropical Agriculural Research Vol. 0: 4-49 (008) Trade Liberalisaion and Expor Demand Funcion: The Case of Souh Asia D. Sajeewani and A. Wijeweera School of Business, Economics and Public Policy Universiy of New England Armidale, NSW 35, Ausralia ABSTRACT. A large number of sudies have esimaed expor price and income elasiciies for many differen counries uilising various economerics echniques. This paper seeks o add o his lieraure by esimaing expor demand elasiciies of seleced Souh Asian counries, India, Bangladesh, Pakisan, Sri Lanka, and Nepal by using coinegraion and error correcion mehods. The sudy aemps o find if he sample counries share he same feaures in heir expor demand funcion afer heir rade reforms. The income elasiciy varies considerably among he sample counries ranging from 0.7 o 4.3 in he long run and from 0.79 o 3.6 in he shor run. Price elasiciy esimaes for all sample counries are inelasic, varying from 0.7 o 0.5 in he long run and from 0.3 o 0.78 in he shor run. INTRODUCTION Numerous sudies have aemped o esimae expor demand elasiciies across a wide range of developed and developing counries (Johnson, 958, Sanos-Paulino and Thirlwall, 004). Houhakker and Magee (969) showed ha for a small counry expor demand is a funcion of he world price, he price of expor producs and he income levels of counries where expors are sen. A beer undersanding of expor price and income elasiciies is imporan because hey direcly impac on rade policies and sraegies. For example, if a counry enjoys an elasic expor income elasiciy i can expec considerable growh in expor earnings and expor volumes whenever here is an economic boom in is major purchasing counries. On a similar noe, we can expec increase in expor earnings in response o currency devaluaion if he counry has an elasic expor price. This sudy endeavours o empirically esimae and analyse he aggregae expor price and expor income elasiciies for some seleced Souh Asian counries, namely; Bangladesh, India, Pakisan, Nepal and Sri Lanka considering rade liberalisaion reforms of each counry using Engle and Granger Coinegraion and an Error Correcion mehods. Alhough Nguyen and Bhuyan (977), as well as Ahmed e al. ( 993) esimaed expor and impor demand elasiciies for some Souh Asian counries, none of he invesigaed counries have iniiaed rade liberalisaion reforms a he ime of sudy. Among oher noable sudies, Choudhury (00) conduced a single counry analysis and esimaed he expor demand funcion of Socio Economics and Planning Division, Deparmen of Agriculure, Peradeniya, Sri Lanka Deparmen of Ars and Sciences, The Peroleum Insiue, PO Box 533, Abu Dhabi, UAE
Sajeewani and Wijeweera Bangladesh while Nur e al. (007) examined he likely impacs of rade liberalizaion policies on he disaggregaed expor funcion in Bangladesh for he period 973 o 004. Model specificaion METHODOLOGY This sudy esimaes income and price elasiciies associaed wih he expor demand funcion for five Souh Asian counries which have undergone considerable changes in heir rade policies in he recen pas. Due o is flexibiliy in accommodaing many economic siuaions, and following he exising lieraure (Pacheco-Lopez 005; Sahinbeyoglu and Ulasan 999; Sanos-Paulino and Thirlwall 004), The sandard expor demand funcion is used o obain he expor price and income elasiciies. The expor demand is modelled following he sandard expor demand funcion which is wrien as: rexp = p f er p d α rgdp where, rexp is he real expors of a counry; represens a consan ime rend; P f is he foreign price, er is he nominal exchange rae measured as he domesic price of foreign currency; P d is he domesic price which are used o derive real exchange rae (rer); rgdp is he real foreign income and α and α are coefficiens o be esimaed. The sandard expor demand funcion is augmened by incorporaing a dummy (lib) o capure he effec of rade liberalisaion policies of he sample counries. The dummy variable akes a zero value prior o he effecive liberalisaion year and a value of one in he pos liberalisaion period. The exended expor demand funcion wih he double log ransformaion is expressed as: ln r exp + α = β 0 ln rgdp 3lib e () Variables are self explanaory wih log values. The Unied Saes (US) Consumer Price Index and Gross Domesic Produc are aken o derive foreign price level and foreign income in he respecive counries expor demand equaions. The coefficiens o be esimaed are β 0, β, β, β 3 and e is he disurbance/noise erm. Due o his double log ransformaion, he coefficiens are he elasiciies. For a normal good, price and income elasiciy (β and β ) are expeced o be posiively relaed o expors. A posiive sign is expeced for β 3, considering rade liberalisaion may promoe economic growh hrough increasing expor demand of counries. Daa Daa for his research were mainly obained from he Inernaional Financial Saisics (IFS) daa base of he Inernaional Moneary Fund (IMF). The effecive year of rade liberalisaion was obained from each counry s exernal rade repors of he World Trade Organizaion. The income level of he USA i.e GDP and he price level i.e CPI were aken () 43
Trade liberalisaion and expor demand funcion: Souh Asia as proxies for he res of he world income and price respecively, considering he conribuion of expor share by each sample counry o he USA 3. The ime period of analysis depends on he daa availabiliy for each counry. Accordingly, he annual daa periods for each counry for he periods are specified in he parenhesis: Bangladesh (97-005); India (950-005); Nepal (964-005); Pakisan (970-005) and Sri Lanka (950-005). Uni roo esing Since he sudy uses ime series daa, i is imporan o idenify wheher he series is saionary before using hem in he esimaion process. Saionary process necessiaes having a consan mean, variance and auo-covariance. The Augmened Dickey Fuller (ADF) uni roo es is used o examine saionariy of he variables. The null hypohesis of he ADF es is ha he series conains a uni roo. Since he rue daa generaing process of hese variables was no known, ADF ess were conduced on he hree model specificaions; as a pure random walk (ADF), a random walk wih drif (ADF) and a random walk wih drif and a rend (ADF3). The Schwarz Bayesian Crieria is used o deermine he opimal lag lengh. ADF es resuls sugges ha all of he variables conain a uni roo in eiher one of he model specificaions in levels and is non-saionary, bu are saionary in firs differences (Table ). Table. Augmened Dickey Fuller uni roo es Counry lnrexp lnrgdp lnrer level ADF ADF ADF3 ADF ADF ADF3 ADF ADF ADF3 Sri -.06-0.87 -.80.03*.5 -.3.66 -.03 -.03 Ban 0.5 -.04 -.94 8.37* -0.04 4.96* 0.57 -.6 -.5 Ind.4-0. -.66 0.86* -.54-4.9*.0-0.79 -.86 Nep -0.56-3.68* -3.9** 0.04* -.00-4.99*.44 -. -3. Pak 0.9 -.9 3.80** 0.73* -0.9 -.5 0.96-0.79-3.6** Counry Δ lnrexp Δ lnrgdp Δ lnrer s dif ADF ADF ADF3 ADF ADF ADF3 ADF ADF ADF3 Sri -7.49* -7.63* -7.55* -3.7* -7.0* -7.7* -6.3* -6.5* -6.5* Ban -4.5* -4.46* -4.77*.35** -4.86* -4.8* -6.3* -6.47* -6.36* Ind -5.* -5.7* -5.46*.6** -4.85* -4.98* -8.* -8.56* -8.46* Nep -7.* -7.04* -7.* -.48** -5.3* -5.5* -7.* -7.74* -7.7* Pak -5.63* -5.59* -5.53* -.54** -5.9* -5.95* -8.07* -8.6* -8.8* *, ** denoes saisical significance a he and 5% levels, respecively. The criical values for 50 observaions as repored in Engle and Yoo (987) a, 5 and 0% levels of significance : ADF(τ saisics) -.6, -.95, and -.6 : ADF (τ μ saisics) -3.58, -.93 and -.60: ADF3(τ ι saisics) -4.5, -3.50 and -3.8. 3 In year 000 he proporion of expor shares sen o USA from Bangladesh, India, Nepal, Pakisan and Sri Lanka were 3.8, 3.6, 5., 40. and 44.% respecively. (Direcion of Trade Saisics, IMF daa, various issues). 44
Sajeewani and Wijeweera Engle-Granger (EG) mehodology Engle and Granger (987) suggesed ha a linear combinaion of nonsaionary variables may produce a saionary series. In order o empirically esimae he coinegraion relaionship, he Engle and Granger (EG) mehod suggess a wo sep procedure. Firsly, if he variables are nonsaionary and are inegraed in he same order, hen he bes linear equaion is esimaed using sandard regression echniques. Secondly, he residuals from he bes linear equaion are esed for he uni roo. If he residuals do no conain a uni roo, i can be deermined ha he variables are coinegraed and share a long run equilibrium relaionship. As discussed before, he firs sep of EG mehod is o esimae he bes possible linear equaion for he sandard expor demand funcion. However, our iniial diagnosic ess sugges ha he same specificaion would no be appropriae for esimaing expor elasiciies for he seleced counries. Therefore, he es was carried ou incorporaing a rend () and Auo Regressive (ar) variable ino he model. This resuled in hree alernaive model specificaions (Model, Model, and Model 3) for he expor demand funcion. Finally, based on he expecaion of he sign of he coefficiens, Model for India and Pakisan, Model for Bangladesh and Sri Lanka, and Model 3 for Nepal were seleced in esimaing expor demand funcions of hese counries. Model : ln r exp = β o ln rgdp 4lib + e (3) Model : ln r exp = β o ln rgdp 4lib + α + e (4) Model 3: ln r exp = β o ln rgdp 4lib + α ar + α + e (5) Error correcion model (ECM) Coinegraion is only relaed o he long run relaionship of he variables. In order o examine he shor run relaionship we esimae an ECM. The EG coinegraion mehodology provides an easier way o obain he error correcion model. Once he coinegraion relaionship of he variables is confirmed, he saved residuals {e - } are used o esimae he ECM as shown in equaion (6). rexp = ln rgdp + γ e ln α + ε (6) All of he variables, excep he error correcion erm, are differenced once o make hem saionary. The coefficien of he error correcion erm (γ) in his model is referred o as he speed of adjusmen parameer. RESULTS AND DISCUSSION The aim of his sudy is o esimae shor run and long run expor elasiciies for some seleced Souh Asian counries. Engle and Granger coinegraion mehod indicaes ha longrun equilibrium relaionship in fac exiss for all five counries. The coinegraion resuls and ADF uni roo es on residuals obained for each counry are summarised in Table. ADF uni roo es on residuals (e ) show ha coinegraion relaionships do exis for all five 45
Trade liberalisaion and expor demand funcion: Souh Asia counries. Table 3 summarises error correcion coefficiens. As he resuls indicae, he error correcion erm is saisically significan for all of he counries excep India. As far as long run equilibrium resuls are concerned, expor demand equaion of Sri Lanka explains 77 percen of variaion by he variables and boh price and income elasiciies carry expeced signs, bu are saisically insignifican a 5 percen level of significance. A low price elasiciy of expor demand which is 0.9 suggess ha Sri Lanka is basically exporing inelasic goods. Hence, price cuing policies such as devaluaion may no be appropriae for increasing expor earnings. The income elasiciy of demand is only 0.36 which suggess ha expors prospecs from high income growh counries are gloomy. I is generally expeced ha rade reforms would have a significan impac on he expor demand of Sri Lanka as reforms were iniiaed well ahead of he oher Souh Asian counries in he region alhough rade liberalisaion coefficien is negaive bu saisically insignifican. This migh be due o is inabiliy o arac considerable amouns of foreign invesmen inflows, is frequen inernal poliical insabiliy and is long running inernal conflics wih a separais movemen hampering he counry s abiliy o reap he real benefis of he rade liberalisaion policies. However, error correcion model resuls show relaively high shor run income and price elasiciies which sugges ha expor opporuniies exis in he shor run and any deviaion from he long run equilibrium is parially adjused wih he adjusmen speed of 0.9. Table. Resuls of he coinegraed relaionship ADF on Counry Consan lnrer lnrgdp lib R resid (e ) Sri Lanka 0.5 (0.09) 0.9 (.50) 0.36 (0.63) -0.0 (0.07) 0.77 -.96+ Bangladesh -37.7 (.4) 0.04 (0.5) 3.86 (.37)* 0.58 (4.4)* 0.66 -.99+ India 3.4 (4.3) 0.7 (0.68) 0.7 (.57) 0.06 (0.37) 0.43 -.++ Nepal -43.5 (.45) 0.50 (.46) 4.3 (.49)* 0.8 (0.99) 0.4-6.73+ Pakisan.7 (.57) 0.9 (.06)* 0.0 (.36) -0.3 (3.86)* 0.5-4.40+ * 5%, ** 0%, significance in OLS esimaes, saisics are given in parenhesis +, ++ and saisical significance a he and 5% according o he Dickey Fuller criical values. The criical values for τ saisics as repored in Engle and Yoo (987) for 50 observaions are -.6, -.95, and -.6 a, 5 and 0% levels of significance respecively. Table 3. Resuls of error correcion model Counry Consan Δlnrer Δlnrgdp e - Sri Lanka -0.06 (-.53) 0.3 (.47)*.08 (.88)** -0.9 (-3.4)* Bangladesh -0. (-.65) 0.5 (0.87) 3.59 (3.8)* -0.9 (-.9)* India -0.0 (-0.40) 0.8 (.00) 0.79 (.05) -0. (-.9) Nepal -0.4 (-.67) 0.78 (.69)* 3.6 (.50)* -0.5 (-3.08)* Pakisan -0.03 (-0.73) 0.3 (0.9) 0.94 (0.99) -0.34 (-.75)* * 5%, ** 0%, level of significance In Bangladesh, he income elasiciy of he expor demand has been found o be highly income elasic. To illusrae, our resuls sugges ha a percen increase in he foreign income level would increase Bangladesh expor demand by 3.86 percen alhough he saisically insignifican smaller price elasiciy implies ha depreciaion of he currency in Bangladesh may no lead o higher expor earnings. Therefore, frequen devaluaion is no 46
Sajeewani and Wijeweera advisable for Bangladesh o ry and enhance is expor earnings. However, rade liberalisaion has been found o be very effecive which suggess ha rade liberalisaion policies have been able o raise demand for expors by 0.58 percen on average. The coefficien of he error correcion erm implies ha 9 percen of he discrepancy beween he acual and he equilibrium value of he expor demand is correced wihin a year. The shor erm income elasiciy of 3.59 is slighly less han long erm income elasiciy of demand, implying he shor run effecs of world income on Bangladesh expors are less han he long run effecs. The foreign income elasiciy of Nepal expor demand is highly income elasic. As foreign income increases by percen, he demand for Nepal expors increases by 4.3 percen. This is mainly due o he naure of he expor baske of Nepal. More han one hird of is expor baske consiss of rising sar goods such as, ready-made garmens, exiles and jewellery which have a high demand, mosly from rich counries such as USA and Germany. Following marke oriened economic reforms during he 990s, Nepal subsanially increased is inegraion ino he world economy. However, he rade liberalisaion dummy is saisically insignifican. Ou of he many reasons, rade resricions imposed by some developed counries and frequen inernal conflics migh have exered some impacs on effecive rade liberalisaion policies. Therefore, i seems ha he rade liberalisaion dummy variable is unable o explain he rue impac of rade liberalisaion on expor demand in Nepal. Shor run esimaes sugges ha devaluaions are relaively effecive as percen change in price level would increase expor demand by 78 percen. The relaively high rae of adjusmen parameer indicaes ha how well consumers and producers respond o marke signals which is a an annual rae of 5 percen. Pakisan expors are price inelasic. A one per cen increase in he price will increase expor demand by 0.8 percen. The relaively low price elasiciy means ha Pakisan does no gain much in expor demand hrough price changes. Therefore, devaluaion alone canno boos expor demand in Pakisan. The coefficien of rade liberalisaion is surprisingly negaive and saisically significan. This implies liberalisaion has reduced Pakisan expor demand by 0.3 percen on average. This unexpeced resul can be explained in wo ways. Firsly, he curren daa se is no sufficien o capure he full impac of rade reforms on expor growh. Secondly, even hough Pakisan declared rade reforms in 998, he exisence of rade barriers remain high. Tariffs and oher proecive insrumens sill provide subsanial proecion o domesic indusries creaing srong disincenives for expors, causing significan ani expor bias. None of he coefficiens are saisically significan for India. Therefore, care should be aken in inerpreing hese elasiciies and using hem in rade policies. The sandard expor demand funcion may no be he bes specificaion o model he Indian expor secor. Our doub over he suiabiliy of he model is furher confirmed by he error correcion model resuls. We found ha he speed of adjusmen parameer is insignifican a convenional levels, conradicing he coinegraion resuls. India began is rade liberalisaion policies in he mid 990s. Therefore, he curren daa se may no be large enough o capure he long run properies and shor run dynamics of he expor demand funcion. CONCLUSIONS 47
Trade liberalisaion and expor demand funcion: Souh Asia This sudy esimaed boh shor run and long run income and price elasiciies of he expor demand funcions for five seleced Souh Asian counries. A number of conclusions can be drawn from his research. Firs, he long run income elasiciies have he expeced sign, and in mos cases are saisically significan. According o our esimaions, he income elasiciies vary from 0.7 o 4.3 in he long run and from 0.79 o 3.6 in he shor run. The relaively high variabiliy of income elasiciies beween hese five counries suggess he need for a number of differen and comprehensive policies o promoe expors of hese counries. This may be aribued o he naure of he commodiies expored by hose counries for a long ime. Imperfec income variable proxy could be anoher reason for his variaion. The USA income is used as a proxy for foreign income, so ha hese values direcly relaed o income changes of he USA. Second, we find ha he absolue values of price elasiciies are inelasic and vary from 0.7 o 0.5 in he long run and 0.3 o 0.78 in he shor run. These inelasic esimaes sugges ha frequen devaluaions are no proven o be an appropriae policy in promoing expors in hese counries. This is mainly due o he naure of he expor commodiies. A large par of expor baske of he sample counries sill consiss of primary and agriculural commodiies, which are relaively price inelasic. Diversificaion of expors ino relaively more manufacured goods is recommended for improving he compeiiveness of expors. Third, our resuls show ha he shor run income elasiciy esimaes are slighly higher han heir respecive long run income elasiciy counerpars. This is counerinuiive because economic agens are more flexible o he changes in income in he long run han in he shor run. Furher research is necessary o explain his. Neverheless, implemening expor qualiy assurance schemes and expor promoion fares ec. would assis hese counries o arac consisen response for heir major expored producs. Finally, we have found mixed resuls regarding he rade liberalisaion impacs on expor secor performance in he seleced counries. Trade liberalisaion produces posiive impacs only in he cases of Bangladesh, India and Nepal. We aribue hese unconvincing resuls o he smaller sample size, exising rade barriers, and inernal conflics. ACKNOWLEDGEMENT Auhors would like o hank hree anonymous referees for helpful commens on an earlier draf. REFERENCES Ahmed, S. M., Houque, N. and Talukder, S. I. (993). Esimaing Expor Demand Funcion for Bangladesh: An Applicaion of Coinegraion and Error-correcion Modelling. The Bangladesh Developmen Sudies (4): 89-04. Choudhury, R. Z. (00). Expor Demand Funcion of Bangladesh: An Economeric Exercise. The Bangladesh Developmen Sudies 7(3): 67-90. 48
Sajeewani and Wijeweera Dickey, D. and Fuller, W. A. (979). Disribuion of he Esimaes for Auoregressive Time series wih a Uni roo. J. Am. Sa. Assoc.74: 47-43. Direcion of Trade Saisics (005). Inernaional Moneary Fund daa base. hp://www. imfsaisics.org/imf (accessed on -5 March, 008). Engle, R. F. and Granger, C. W. J. (987). Coinegraion and Error Correcion: Represenaion, Esimaing and Tesing. Economerica 55: 5-7. Engle, R.F. and Yoo, B.S. (987). Forecasing and Tesing in Coinegraed Sysems. J. Economerics 35: 43-59. Houhakker, H. S. and Magee, S. P. (969). Income and Price Elasiciies in World Trade. Rev. Econ. Sa. 5(): -5. Inernaional Financial Saisics (005). Inernaional Moneary Fund daa base. hp:// www.imfsaisics.org/imf. (accessed on -5 March, 008). Johnson, H. G. (958). Inernaional Trade and Economic Growh Cambridge: Harvard Universiy Press, USA. Nguyen, D.T. and Bhuyan, A.R. (977). Elasiciies of Expor and Impor Demand in Some Souh Asian Counries: Some Esimaes. The Bangladesh Developmen Sudies 5: -6. Nur, M., Wijeweera, A. and Dollery, B. (007). Esimaion of he Expor Demand Funcion using Bilaeral Trade Daa: The Case of Bangladesh. Souh Asia Econ. J. 8(): 49-64. Pacheco-Lopez, P. (005). The Effec of Trade Liberalisaion on Expors, Impors, he Balance of Trade, and Growh: The Case of Mexico. J. Pos Keynesian Econ. 7(4): 596-69. Sahinbeyoglu, G. and Ulasan, B. (999). An Empirical Examinaion of he Srucural Sabiliy of Expor Funcion: The Case of Turkey. The Cenral Bank of he Republic of Turkey, Discussion Paper No. 9907: -4. Sanos-Paulino, A. and Thirlwall, A. P. (004). The Impac of Trade Liberalisaion on Expors, Impors, and he Balance of Paymens of Developing Counries. Econ. J. 4(): 50-7. 49