Inernaional Trade and Finance Associaion Inernaional Trade and Finance Associaion 15h Inernaional Conference Year 2005 Paper 46 THE EFFECT OF EXCHANGE RATE CHANGES ON TRADE BALANCES IN NORTH AFRICA: EVIDENCE Marial A. Yol Universii Pura, Malaysia Ahmad Z. Baharumshah Universii Pura, Malaysia This working paper sie is hosed by The Berkeley Elecronic Press (bepress) and may no be commercially reproduced wihou he publisher s permission. hp://services.bepress.com/ifa/15h/ar46 Copyrigh c 2005 by he auhors.
THE EFFECT OF EXCHANGE RATE CHANGES ON TRADE BALANCES IN NORTH AFRICA: EVIDENCE Absrac This paper examines he effecs of exchange rae changes on he bilaeral rade balances of Morocco and Tunisia vis-à-vis he Unied Saes and Japan. I also sheds some ligh on unresolved issues surrounding he J-curve phenomenon. Presened a he 15h Inernaional Conference,Isanbul, Turkey, May 2005.
INTRODUCTION Accurae undersanding of rade balance response o changes in exchange rae is a crucial facor in he coordinaion and implemenaion of rade and exchange rae policies. The convenional wisdom is ha a nominal depreciaion/devaluaion of exchange rae improves he rade balance in he long run while causing i o worsen in he shor run. A change in he exchange rae has wo effecs on he rade balance; he price effec and volume effec. While he price effec works o make impors more expensive o foreigners, i causes domesic expors o be cheaper for foreign buyers, a leas, in he shor run and as he volume of expors and impors do no adjus insananeously in he shor run, rade balance may iniially experience some deerioraion in he shor run, following he currency depreciaion/devaluaion. However, following evenual adjusmen process of expors and impors o he currency depreciaion/devaluaion, price effec will end o dominae he volume effec in he shor run whereas volume effec will dominae in he long run, hence, reversing he overall effec in favor of he rade balance improvemen, assuming ha he Marshall-Lerner condiion holds. In he lieraure, he wo main heories ha aemp o explain he effec of exchange rae depreciaion/devaluaion on he rade balance are he elasiciy approach and he moneary approach. The elasiciy approach focuses on demand condiions by assuming ha he supply of domesic expors and foreign impors are perfecly elasic so ha changes in demand volumes have no unambiguous effec on prices. In effec, he assumpion implies ha domesic and foreign prices are fixed so ha changes in relaive prices are solely caused by changes in nominal exchange rae. In addiion, he approach idenifies wo direc effecs of depreciaion/devaluaion on he curren accoun balance, one ha works o reduce he defici and he oher ha works o worsen he defici more han before. The Marshall-Lerner condiion upon which he heory is buil saes ha, saring from a posiion of equilibrium in he curren accoun; a depreciaion/devaluaion of a domesic currency improves he curren accoun only as long as he sum of elasiciy of foreign demand for expors and elasiciy of domesic demand for impors exceeds uniy. However, if his sum is less han uniy hen devaluaion will worsen he curren accoun. This follows from he belief 1 ha ransacion compleed a he ime of depreciaion/ devaluaion may dominae a shor-erm change in he rade balance, leading o he deerioraion of rade balance during he conrac period, before quaniies of expors and impors adjus. As he elasiciies of expors and impors improve over ime, quaniies adjus o he new effecive prices, causing he rade balance o rise. However, 1 See Krueger (1983). Hosed by The Berkeley Elecronic Press
Williamson (1983) poins ou ha such higher impor prices resuling from depreciaion/devaluaion could raise domesic prices of non-raded goods, generaing an overall inflaion ha may raise he effecive real exchange rae, hence, eliminaing he every poenial for improving he rade balance. On he oher hand, he moneary approach which views he balance-ofpaymens surpluses and deficis as moneary flow resuling from sock disequilibrium in he money marke, assumes ha he balance of paymens is essenially a moneary phenomenon in inernaional moneary economy, which requires he analysis in erms of moneary conceps. The approach conends ha a defici in he balance of paymens is due o an excess supply of sock of money in relaion o money demand whereas a surplus is a flow resuling from an excess demand for money in relaion o he sock of money supply. Based on his premise, he moneary approach regards he balance of paymens disequilibrium as a reflecion of disequilibrium in he money marke. The immediae effec of depreciaion/devaluaion is o make domesic goods compeiive in relaion o PPP compared wih foreign goods, raising he demand for domesic currency. As he demand for domesic currency exceeds is supply, he comparaive advanage of deprecaion/devaluaion would improve he balance of paymens as domesic demand for foreign goods falls while foreign demand for more domesic goods increases. In he monearis s view, he effec of he currency depreciaion/devaluaion on he balance of paymens should be ransiory only as long as he moneary auhoriies do no simulaneously engage in expanding money supply hrough open marke operaions. However, if he moneary auhoriies increase money supply via open marke operaions, aggregae demand and domesic prices will rise, eliminaing he very comparaive advanage he currency depreciaion/ devaluaion was mean o generae. The objecives of his paper are wofold: firs, o examine he effecs of exchange rae changes on he bilaeral rade balance; and secondly, o aemp o esablish he exisence of a J-curve phenomenon. The sudy employs he Johansen coinegraion and error-correcion model on he annual daa on hree non-oil exporing Norh African counries of Egyp, Morocco and Tunisia vis-à-vis he US and Japan in he period 1970-2003. The choice of hese Sub-Saharan Norh African counries is moivaed by a number of facors. Firs, hese counries were among he firs on he Sub-Saharan African coninen in implemening he IMF Srucural Adjusmen Programs (SAPs) since early 1980s in which exchange rae adjusmen was among he insrumens argeed for bringing abou he inended economic liberalizaion. Hence hey may offer invaluable lessons in he area of exchange rae rade balance adjusmen. Secondly, flowing he implemenaion of he menioned programs, hese counries, among ohers, were hough o have experienced he J-curve kind of adjusmen ha characerizes he behaviour of rade balance ha follows exchange-rae adjusmen. For hese reasons, hey meri hp://services.bepress.com/ifa/15h/ar46
being chosen for sudy o analyse he effec of exchange rae changes on bilaeral rade balance during his adjusmen period. In general, he findings confirm he exisence of boh shor-run dynamics and long-run causal relaionships beween rade balance and he se of specified independen variables. However, he resuls produce mixed evidence abou he exisence of he J-curve effec. A classic J-curve paern is observed only in case of Morocco/Japan in which he rade balance deerioraes almos immediaely following he devaluaion of he real exchange rae and recovers afer wo years. In cases of Egyp/Japan and Tunisia/US, we could observe some kind of delayed J-curve effecs in which, following he currency devaluaion, rade balances experience shor-erm improvemen firs ha lass for abou one year in boh counries. However, in cases of Egyp/US, Morocco/US and Tunisia/Japan, no clear evidence of he J-curve phenomenon is found. The res of he sudy will be organized as follows. Secion II describes he exchange rae policies ha exised in Egyp, Morocco and Tunisia prior o and during his period of sudy. Secion III presens he empirical model and economeric mehodology ha will be employed in invesigaing he effecs of exchange rae changes on he rade balance. While secion IV presens he discussion of empirical resuls, secion V summarizes he conclusions. EXCHANGE RATE POLICIES IN EGYPT, MOROCCO AND TUNISIA IN THE PERIOD 1970S 1990S During 1970 and much of 1980s, Egyp and he wo Maghreb naions of Morocco and Tunisia, like many of heir counerpars in developing counries by he ime, pursued inward-looking developmen sraegies in which he financial sysems were heavily conrolled. Ineres raes were se adminisraively and consequenly, were someimes negaive in real erms. Moneary policy was conduced chiefly hrough direc allocaion of credi and refinancing. In addiion, money markes were in nascen sage of developmen wih complee absence of bond and equiy markes. Commercial banks were forced o lend o he so-called prioriy secors wih lile aenion paid o he borrowing firm s profiabiliy and defaul risk. Capial flows were ighly regulaed wih enry of foreign invesmen o he financial secor severely resriced. When hese inefficiencies and disorions exacerbaed he economic problems and almos crippled he prospecs of economic growh, he governmens of hese counries underook comprehensive sabilizaion and srucural reform programs in 1986 wih IMF and World Bank suppor. The reforms aim a reducing direc governmen inervenion and srenghening he role of marke forces in he allocaion of financial resources, improving he capaciy of financial Hosed by The Berkeley Elecronic Press
insiuions o mobilize domesic savings, enhancing compeiion among banks, and srenghening heir financial soundness. In Egyp, he second 2 larges economy in he Arab world wih a GDP of $85.5 billion (2002 esimaes), he exchange rae has been pegged o he US dollar since he 1960s. In he early 1980s, he counry was operaing a a muliple exchange rae regime in which here was he Cenral Bank s rae, which coninued unil 1979 when he pound was devalued in 1990; he commercial banks rae ha coninued unil is aboliion in1989 and lasly, he rae a which foreign exchange was raded a a premium in he so-called exchange rae pool. Wih beginning of he economic reform program in 1991, he unificaion of exchange rae in managed floaing regime was an imporan componen of he policy package. To sreamline he economy furher, successions of exchange rae devaluaion were underaken beween 1991 and 2000. Following he policy of gradual devaluaion of he currency in he firs eigh monh of 2001, he governmen announced anoher new exchange rae regime, known as he cenral rae, around which banks and exchange bureaus were o buy and sell wihin a rading band of 1%. No long enough, in June of he same year, he Cenral Bank adjused he cenral rae and increased he rading band o 1.5% for dollars and 2% for oher currencies. Finally, in January 2003, he Egypian currency was finally floaed and by mid-ocober, he exchange rae had depreciaed by abou 33% o sele a LE 6.15/US$. The esablishmen of inerbank foreign exchange markes in Tunisia in 1994 and in Morocco in 1996 marked an imporan sep oward decenralizing he managemen of foreign exchange and allowing marke forces o play a greaer role in exchange rae deerminaion. The naional currencies of boh counries (he Moroccan dirham and he Tunisian dinar) experienced significan upheavals in early 1970 due o he insabiliy of French franc (FF) o which he wo currencies were pegged. In case of he Moroccan dirham, he link o FF was severed in 1973 and he currency underwen a managed-floa regime. The move was o mainain dirham s relaive effecive exchange rae in relaion o a baske of major foreign currencies. Like in many developing counries, a dual exchange rae sysem was se up for dirham in 1973 and 1978 respecively. In 1980, in order o accommodae he changes in Moroccan foreign rade paern and srucure of currencies used in exernal selemen, he weighs in he currency baske were adjused, paving he way for a gradual depreciaion of he dirham. In he subsequen years he worsening erms of rade and he accompanying deerioraion in he balance of paymen posiion ha sared in 1986 forced he real depreciaion of he dirham by abou 9% in 1990. 2 Second only o he Kingdom of Saudi Arabia, which has he GDP of US$ 191 billion (2002 esimaes). hp://services.bepress.com/ifa/15h/ar46
On he oher hand, he Tunisian dinar was also linked o he German mark (DM) in addiion o he FF and he US dollar in 1978. Numerous aemps, including changing he weighs of he currency baskes, were embarked upon o sabilize he currency, however, he rade and balance of paymens difficulies ha cropped up in he mid-1980s forced he auhoriies in Augus 1996 o fully liberalize he currency. THE EMPIRICAL MODEL AND METHODOLOGY Trade balance, someimes referred o as visible balance, represens he difference beween receips for expors of goods and expendiure on impors of goods. In oher words, i is he difference beween he moneary value of oal expors and impors. As suggesed 3 ha rade balance is deermined by a number of macro variables such as real oupus, exchange raes, money supplies, ec., he exisence of direc or indirec casual feedback beween rade balance and such macro variables canno be ruled ou. Wih his in mind, he Krugman-Baldwin (1987) model is adoped 4 for his sudy. An advanage of his model, in addiion o is simpliciy, is is abiliy o capure he effec of all he specified macro variables on rade balance. This wo-counry model sipulaes ha he demand for impored goods by domesic residens is posiively relaed o real domesic income and negaively o relaive price of impored goods. Hence, rade balance is given as * TB = TB( Q, Y, Y ) (1) where Q e.p * /P. In his expression, an increase in e would denoe a depreciaion of he domesic real exchange rae and vice-versa. Equaion (1) can be esimaed in a log-linear form by aking he logs on boh sides o obain he following esimaing equaion * b = 0 + 1 q + 2 y + 3 y + µ (2) 1 > 0, 2 > 0 or < 0 and 3 > 0. where lower-case leers represen logarihm of a variable and µ is whie-noise process. In he model, a four-variable sysem will be se up, denoed by he noaions: log of rade balance is b, log of real effecive exchange rae (real exchange rae for Egyp) is q, log of domesic income (GDP) is y, and he log of foreign income is y *, compacly wrien as x = {b, q, y, y * }. In his sudy, rade balance is expressed as a raio of expors o impors (e.g., Weixian, 1998; Gupa-Kapoor and Ramakrishnan, 1999; Baharumshah, 2001; Bahmani-Oskooee 2002; and Onafowora, 2003). The advanage of his raio is ha i is insensiive o he uni of 3 See Baharumshah (2001) for more deails. 4 For deailed discussion of he model, see Rose and Yellen (1989), Rose (1991), Bahmani- Oskooee (1991), Brada e al. (1993), and Baharumshah (2001) Hosed by The Berkeley Elecronic Press
measuremen as i can be easily inerpreed as nominal or real exchange rae. Secondly, i is emphasized (Boyd e al., 2001) ha in logarihmic model i gives he Marshall-Lerner condiion exacly raher han approximaing i. Furhermore, following sudies by Rose (1991), Bahmani-Oskooee (1991) and Baharumshah (2001), his sudy will use real effecive exchange rae insead of nominal bilaeral exchange rae. Finally, rade balance, domesic and foreign incomes will be in real erms while consumer price index will be used as he price deflaor. The daa series for his sudy, which are seasonally unadjused, are drawn mainly from he IMF Inernaional Finance Saisics (IFS. A dummy variable is included wih value uniy before 1986 when he exchange raes were fixed and a zero oherwise. As a sandard pracice, his sudy employs he coinegraion and vecor error-correcion model because of is abiliy o examine he shor-run dynamics and long-run relaionship beween rade balance and he specified se of independen variables. The model is based on four crucial seps. Firsly, he properies of individual ime series are invesigaed by applying he Augmened Dickey-Fuller (ADF) and he Phillip-Perron (PP) ess for uni roos in he variables. Esimaing vecor error-correcion mechanism (VECM) requires he use of saionary daa in order o avoid spurious resuls. Secondly, Johansen maximum likelihood mehod (Johansen, 1988 and Johansen and Juselius 1990) is applied o es he degree of inegraion among he variables. In addiion o yielding maximum likelihood esimaors of unconsrained coinegraing vecors, he procedure allows for explici esing for he number of coinegraing vecors. In his mehod, Johansen and Juselius wrie he disribued lag model of a vecor of variables, X, in he following vecor error-correcion form X k n = X + X + (3) k = 1 k k + 1 where i = - + 1 + 2 +... + i (i = 1,..., k 1) and = -( - 1-2 -... - k ). I is he erm ha conains he informaion concerning he long-run relaionship among X variables. This erm is a n x n marix of long-run parameers whose rank deermines he number of independen coinegraing vecors; say r (< n). If his marix is of full rank, all variables in he sysem are saionary and he model may be esimaed wih variables in levels. However, if he marix is of zero rank, here is no coinegraing relaionship among he variables and, hence, he model could be esimaed in firs difference wihou error-correcion erm. When he marix is of reduced rank, i.e., 0 < r < n, here are r (linearly independen) coinegraing vecors. Marix can be decomposed ino wo (n x r) dimensional hp://services.bepress.com/ifa/15h/ar46
marices and : wrien as =, where, also called loadings, is he marix conaining he adjusmen coefficiens, and comprises he coinegraion vecors. Following Engle and Granger (1987), a variable is inegraed of order d, or I(d), if i has o be differenced d-imes o become saionary. A vecor x is inegraed of order d if he maximum order of inegraion of he variables in x is d, while a vecor x is coinegraed, or CI (d, b) if here exiss a linear combinaion x ha is inegraed of lower order (d b) han x. I has been shown (Johansen and Juislius, 1990) ha can be inerpreed as a marix of coinegraing vecors and can be esimaed as he eigenvecor associaed wih he larges r, saisically significan eigenvalues. The rank of a marix is equal o he number of is characerisic roos ha differ from zero, deermined by esing he significance of he characerisic roos of. Thirdly, if he variables are coinegraed, hen ess involving differenced variables will be misspecified and some imporan informaion los unless a lagged error-correcion erm is included. Therefore, we esimae he errorcorrecion model in which he error-correcion erms (ECT), derived from longrun coinegraing vecors, are included as independen explanaory variables in he esimaion process of equaion (3) in order o recover all he long-run informaion ha was los in he original esimaion process. Assuming ha equaion (3) is coinegraed wih all is non-saionary variables, he sysem can be rewrien as: ln( b) = + µ + n i = 1 1 + ln( y) i ln( b) i i i = 1 i = 1 i k + + l ln( q) i i + m i = 1 ln( y i * ) (4) i where all he variables in equaion (4) remain defined as before. µ -1 is he errorcorrecion erm (one lagged error) generaed from Johansen mulivariae process and is he disurbance erm. The coefficien of he lagged error-correcion erm,, is he adjusmen coefficien ha defines he proporion by which long-run disequilibrium in dependen variable is correced in a given period of ime. The applicaion of he Johansen mehod produces wo ypes of relaionships; shor-run dynamics and long-run causal relaionships beween he rade balance and he specified se of independen variables. The long-run causal relaionship is capured by he coefficiens of he lagged value of he errorcorrecion erm, µ -1, while he shor-run causal effecs are implied by he coefficiens of he lagged explanaory variables (in firs difference). If only he vecor error-correcion erm (ECT) is saisically significan, his implies ha he variable is weakly endogenous wih respec o he long-run parameers. Similarly, Hosed by The Berkeley Elecronic Press
if only he coefficiens of he lagged explanaory variables (in firs difference) are saisically significan, his implies ha he variable is weakly endogenous wih respec o he shor-run parameers. Finally, if he lagged explanaory variables and error-correcion erms are no saisically significan, hen he dependen variable is economerically srongly exogenous. Finally, in order o discover wheher he J-curve effecs exis, generalized impulse response funcions mus be compued and ploed o idenify he response of rade balance o shocks in he real effecive exchange rae for hese hree counries vis-à-vis he US and Japan. This is one way of visualizing he response of endogenous series o various shocks. Expressing vecor auoregression in vecor moving average allows he ime pah of he various shocks on he variables included in VAR sysem o be raced ou. There is no unanimiy on he empirical evidence in lieraure concerning he relaionship beween exchange rae and rade balance, o say he leas. For example, Miles (1979) for seven European, hree Lain American and wo Souh Asian counries in addiion o Israel and New Zealand, Rose and Yellen (1989) for G-7, Rose (1991) for UK, Canada, Japan and USA, Upadhyaya and Dhakal (1997) for eigh developing counries from Europe, Africa and Lain America and Wilson (2001) for Singapore, Malaysia and Korea and Wilson and Ta (2001) for Singapore have empirically shown ha currency depreciaion/devaluaion does no necessarily lead o improvemen in he rade balance. On he oher side, Himarios (1985) for en Lain American, European, Asian and Middle Easern counries, Rawlins and Praveen (1993) for seleced nineeen Sub-Saharan African counries, Bahmani-Oskooee (1991) for eigh Lain American and Asian counries in addiion Greece, Weixian (1998) and Zhang (1999) for China, Gupa-Kapoor and Ramakrishnan (1999) for Japan, Baharumshah (2001) for Malaysia and Thailand, Lal and Lowinger (2002) for five Souh Asian counries, Suka (2003) for Croaia, Onafowora (2003) for Indonesia, Malaysia and Thailand, and Narayan (2004) for New Zealand, provide an empirical evidence ha currency depreciaions/devaluaions in general do lead o improvemens in rade balances. Neverheless, differences over he exisence of he J-curve phenomenon remain common as well. Baharumshah (2001), Wilson (2001) and Onafowora, (2003), for example, reach differen conclusions regarding he exisence of J- curve effecs in case of Malaysia. While he firs wo find no evidence of he exisence of J-curve for Malaysia, Singapore and Thailand, wih he excepion of Korea, he laer confirms he evidence of shor-run J-curve effec in case of Indonesia and Malaysia in heir rade wih boh US and Japan. On he oher hand, Rawlins and Praveen (1993), in a sudy of nineeen Sub-Saharan African hp://services.bepress.com/ifa/15h/ar46
counries 5, fail o repor any evidence supporing he exisence of a J-curve effec. Neverheless, Narayan (2004) finds ha New Zealand s rade balance exhibis a J- curve effec. RESULTS OF THE EMPIRICAL MODEL The empirical analysis is carried ou using annual daa on expors, impors, real effecive exchange rae and/or real exchange rae, domesic oupu (GDP) and foreign oupu (GDP) for Egyp, Morocco and Tunisia in he period 1970-2003. The lengh of he sample period was dicaed by he availabiliy of he relevan daa. We employ he sandard Dickey-Fuller and Phillips-Perron ess for all he variables a levels and firs differences wih rend and wihou. While he es resuls a levels indicae ha he null hypohesis of a uni roo could no be rejeced for all he variables, he saisical significance of absolue value of es saisics a leas a five percen level indicaes ha he series are saionary a firs differences and should be employed in VAR/VECM process in heir presen form. To conserve a space, he resuls of uni-roo ess are no repored here bu available wih he auhor upon reques. Afer confirming he saionariy of he variables a I(1), he Johansen maximum likelihood mehod (Johansen, 1988 and Johansen and Juselius 1990) is applied o examine he quesion of coinegraion among he variables. Before conducing he coinegraion ess, he dynamic specificaion of he Johansen model is deermined firs by esimaing he iniial equaions using he full daa se, seing k, he opimal lag lengh, arbirarily. The unresriced model is hen esed agains a resriced version where k varies from 3 o 1. In case of Egyp, we employ a lag lengh of 3 while for Morocco and Tunisia, a lag lengh of 2 was found o be appropriae as indicaed by Akaike and Schwarz informaion crieria. The compued es saisic is asympoically disribued as 2 wih n 2 ( 1-0 ) degrees of freedom. As he Johansen procedure is noorious in over-rejecing he null hypoheses in small samples, however, Reimers (1992) and Psaradakis (1994) recommend making adjusmen for he degrees of freedom by placing T by T- nk in race and Maximal eigenvalue es equaions where T is he sample size, n is he number of variables in he model and k, is he lag lengh. This adjusmen could improve he small-sample behavior of likelihood raio saisics. Therefore, we use 22 in case of Egyp and 26 in oher wo cases raher han 34 observaions in his sudy. In addiion, a consan was included in he coinegraing vecor for all he 5 The counries were: Burkina Faso, Cameroon, Cenral African Republic, Côe d Ivoire, Gabon, The Gambia, Ghana, Kenya, Madagascar, Mauriius, Niger, Nigeria, Rwanda, Senegal, Sierra Leone, Tanzania, Togo, Zaire and Zambia. Hosed by The Berkeley Elecronic Press
models excep in cases of Morocco/Japan and Tunisia/Japan where consans and deerminisic rends were included in coinegraing vecors. The resuls of coinegraion ess are repored in able 1 in which he Johansen es rejecs he null hypohesis of no coinegraion in favour of a leas one coineraing vecor in all cases. Whereas boh he Max-eigen and race saisics indicae wo coinegraion vecors in case of Egyp/US, only he Maxeigen saisic idenifies wo coinegraing vecors in case of Egyp/Japan. In cases of Morocco/US and Morocco/Japan, boh saisics indicae he exisence of a single coinegraing vecor. However, in case of Tunisia/US, he race saisic indicaes one coinegraing vecor whereas in case of Tunisia/Japan, he Maxeigen saisic indicaes a single coinegraing vecor. Table 1: Resuls of Johansen s Tes for Mulivariae Coinegraing Vecors H o Egyp/US Max-Eigen Saisic 95% criical value Trace Saisic 95% criical value r 0 40.00 *** 27.078 71.073 *** 47.21 r 1 26.34 *** 20.97 31.07 ** 29.68 r 2 4.25 14.07 4.73 15.41 r 3 0.48 3.76 0.48 3.76 Egyp/Japan r 0 25.63 31.46 68.68 ** 62.99 r 1 21.75 25.54 43.05 ** 42.44 r 2 14.21 18.96 21.30 25.32 r 3 7.07 12.25 7.07 12.25 Morocco/US R 0 27.24 ** 27.07 50.58 ** 47.21 R 1 13.02 20.97 23.35 29.68 R 2 9.06 14.07 10.33 15.41 R 3 1.27 3.76 1.27 3.76 Morocco/Japan R 0 69.62 *** 31.46 108.69 *** 62.99 R 1 21.30 25.54 39.07 42.44 R 2 13.86 18.96 17.77 25.32 R 3 3.91 12.25 3.91 12.25 Tunisia/US R 0 24.82 27.078 47.44 ** 47.21 R 1 14.78 20.97 22.62 29.68 R 2 7.60 14.07 7.83 15.41 R 3 0.23 3.76 0.23 3.76 Tunisia/Japan hp://services.bepress.com/ifa/15h/ar46
R 0 32.03 ** 31.46 61.85 62.99 R 1 14.51 25.54 29.82 42.44 R 2 11.27 18.96 15.31 25.32 R 3 4.04 12.25 4.04 12.25 Noes: Noes: ***, ** and * indicae 1%, 5% and 10% significance level respecively. In models where here is one coinegraing vecor, idenifying he equaion ha represens he rade balance is sraighforward. By normalizing on he log of he rade balance 6, we could obain he coinegraing vecor, which can be inerpreed as he long-run relaionship beween rade balance and he independen variables. However, in case of he Egyp/US and Egyp/Japan models where wo coinegraing vecors were idenified, alhough he choice of which coinegraing vecor o represen he rade balance equaion is debaable, we normalize on he rade balance variable since i is he variable of he ineres. The resuls of he esimaed coinegraing equaions are repored in able 2. Table 2: Esimaed Coinegraing Vecors in Johansen Sysem COUNTRY/VAR b q y * y Consan Egyp/US 1.00-0.111-5.421 *** 4.263 *** 45.004 Egyp/Japan 1.00-1.181 *** 4.214 *** 0.925 ** -136.654 Morocco/US 1.00-0.036 *** 0.001-0.015-1.227 Morocco/Japan 1.00-0.031 *** -0.032 ** 0.093 *** -1.950 Tunisia/US 1.00-0.028 *** 0.009 0.013-1.475 Tunisia/Japan 1.00-0.059 *** 0.001 0.053 *** -2.189 Noes: ***, **, and * denoe 1%, 5% and 10% significance level. In all cases, he coefficiens of real exchange raes have correc posiive signs and significan a 1% level as he heory predics excep in case of Egyp/US, suggesing ha a depreciaion/devaluaion of real exchange rae would simulae expors and discourage impors, hence, improving he rade balance of hese counries. The insignificance of he coefficien of RER in case of Egyp/US 6 See Charemza and Deadman (1992). Hosed by The Berkeley Elecronic Press
implies ha Egypian rade balance does no respond o changes in real exchange rae vis-à-vis he US dollar in he long run, bu is driven by shor-run adjusmens. The finding could be an implicaion of he shrinking rade beween Egyp, which is a non-oil exporing counry, and he USA. More imporanly, his could be he case since coon, which used o be he major foreign exchange earner in he Egypian economy vis-à-vis he USA, has los is imporance in inernaional rade since he 1970s. Foreign real income variable is significan a leas a 1% level in cases of Egyp/US, Egyp/Japan and Morocco/Japan, bu carries correc posiive sign only in cases of Egyp/US and Morocco/Japan, implying ha an increase in American and Japanese incomes improves rade balances of Egyp and Morocco in he long run. Alhough significan a 1% level, foreign real income surprisingly carries posiive sign in case of Egyp/Japan, which suggess ha an increase in Japanese income causes rade balance o deeriorae in Egyp in a long run. In cases of Egyp/US, Egyp/Japan, Morocco/Japan and Tunisia/Japan, domesic real incomes carry correc negaive signs and are significan a leas a 5% level, suggesing ha a rise in domesic income of hese counries encourages heir consumers o demand more foreign goods, leading o a deerioraion of rade balance in favor of US and Japan, which is consisen wih he heory. Furhermore, he fac ha he esimaed long-run exchange rae elasiciies are less han 1 implies ha he Marshall-Lerner condiion is no saisfied excep in case of Egyp/Japan where i is greaer han 1 (1.181). Having esablished he exisence of long-run relaionship among he relevan variables by employing he Johansen es, he nex ask is o se up he error-correcion model in which error correcion erms (ECT) are included. The resuls of he shor-run version of he model, specified as a vecor error correcion model, are presened in Table 3, in which all he insignifican variables are omied from he model. In addiion, since he inerpreaion of hese shor-run relaions does no have any significance, we ignore hem. In his regard, Masih and Masih (2004, p.597) cauion agains aaching oo much significance o such shor-run relaionships since he nonsignificance or oherwise of any of he differenced explanaory variables, which reflec only shor-run relaionship, do no violae he heory. Tha is because he heory says lile abou such shor-run relaionships. The value of Durbin-Wason saisics lying around 2 rejecs he presence of any serious problem of serial correlaion among he residuals. The speed of adjusmen, defined by he coefficien associaed wih he error-correcion erm, are correcly negaive and significan a leas a 5% level in each model. Since his value lies beween 0 and 1, (excep in case of Egyp/Japan), he correcion of disequilibrium in he presen period is a fracion of he error in he previous period. In his case, he ECT ends o cause rade balance o converge monoonically o is long-run equilibrium pahs in relaion o changes hp://services.bepress.com/ifa/15h/ar46
in he exogenous variables in each case. This implies ha abou 37%, 62%, 64%, 98% and 60% of disequilibrium in bilaeral rade beween Egyp and US, Morocco and US, Morocco and Japan, Tunisia and US and Tunisia and Japan respecively is being eliminaed per a year. From our esimaes in his sudy, o dissipae 50% of a shock o rade balance requires abou one-and-half years, eigh monhs, seven monhs, wo monhs and nine mohs beween Egyp and US, Morocco and US, Morocco and Japan, Tunisia and US and Tunisia and Japan respecively. In case of Egyp/US, disequilibrium in he rade balance does no respond o long-run adjusmen, however, in case of Tunisia/US, he speed of adjusmen is considered o be very fas compared o oher models. Table 3: Resuls of Error-Correcion models wih Trade balance as he dependen variable Independen Variables Egyp/ US Egyp/ JP Morocco/ US Morocco/ JP Tunisia /US Tunisia/ JP Consan -0.211-0.197 0.000-0.002 ** 0.001 0.001 ECT -1-0.208 0.507-0.619 *** -0.640 *** -0.977 *** -0.603 *** ECT -2-0.369 *** - - - - - TB(-1) - -0.804 ** - - 0.608 *** 0.328 * TB(-2) - -0.443 * - -0.206 * 0.517 *** 0.451 ** TB(-3) - - - - - - E(-1) 2.061 ** 0.506 * - -0.030 *** -0.026 *** - E(-2) - - -0.017 * -0.06 *** - -0.036 *** Y * (-1) - 11.937 *** -0.065 ** - - - Y * (-2) - - 0.098 *** 0.113 *** - - R 2 0.71 0.68 0.70 0.76 0.70 0.60 Hosed by The Berkeley Elecronic Press
D-W Sa 2.224 1.968 2.064 2.493 2.077 2.160 Noes: ***, ** and * indicae 1%, 5% and 10% significance level respecively. When he final models were subjeced o a number of diagnosic ess, here was neiher serious misspecificaion problem nor any evidence of auocorreclaion in he disurbance erm. The resuls of he ARCH and Jarque- Bera normaliy ess confirm he homoscedasiciy of errors, independence of he included regressors and he normaliy of he esimaed equaions. In addiion, o check wheher he esimaion regression equaions were sable hroughou he sample period, he plos of CUSUM and CUSUM (cumulaive sum) of squares ess (Brown e al; 1975) fall inside 5% criical lines 7 as shown in Figures 1-6. The imporance of hese ess is ha a movemen of he CUSUM and CUSUM squared residuals ouside he criical lines is suggesive of he insabiliy of he esimaed coefficiens and parameer variance over he sample period. In his sudy, he ess could no rejec he null hypohesis ha he regression equaions are correcly specified a 5% level of significance, implying ha here has no been sysemaic changes in he regression coefficiens. Figure 1: CUSUM and CUSUM of Squares Tess (Egyp/US) 12 1.6 8 1.2 4 0.8 0-4 0.4-8 0.0-12 1990 1992 1994 1996 1998 2000 2002-0.4 1990 1992 1994 1996 1998 2000 2002 CUSUM 5% Significance CUSUM of Squares 5% Significance 7 The es resuls are no repored here. hp://services.bepress.com/ifa/15h/ar46
Figure 2: CUSUM and CUSUM of Squares Tess (Egyp/Japan) 15 1.6 10 5 0-5 -10 1.2 0.8 0.4 0.0-15 84 86 88 90 92 94 96 98 00 02-0.4 84 86 88 90 92 94 96 98 00 02 CUSUM 5% Significance CUSUM of Squares 5% Significance Figure 3: CUSUM and CUSUM of Squares Tess (Morocco/US) 12 1.6 8 4 1.2 0.8 0-4 0.4-8 0.0-12 1990 1992 1994 1996 1998 2000 2002-0.4 1990 1992 1994 1996 1998 2000 2002 CUSUM 5% Significance CUSUM of Squares 5% Significance Figure 4: CUSUM and CUSUM of Squares Tess (Morocco/Japan) Hosed by The Berkeley Elecronic Press
12 1.6 8 4 1.2 0.8 0-4 -8 0.4 0.0-12 1992 1994 1996 1998 2000 2002-0.4 1990 1992 1994 1996 1998 2000 2002 CUSUM 5% Significance CUSUM of Squares Figure 5: CUSUM and CUSUM of Squares Tess (Tunisia/US) 12 1.6 5% Significance 8 1.2 4 0.8 0-4 0.4-8 0.0-12 1990 1992 1994 1996 1998 2000 2002-0.4 1990 1992 1994 1996 1998 2000 2002 CUSUM 5% Significance CUSUM of Squares 5% Significance Figure 6: CUSUM and CUSUM of Squares Tess (Tunisia/Japan) 12 1.6 8 4 0-4 -8 1.2 0.8 0.4 0.0-12 1992 1994 1996 1998 2000 2002-0.4 1990 1992 1994 1996 1998 2000 2002 CUSUM 5% Significance CUSUM of Squares 5% Significance hp://services.bepress.com/ifa/15h/ar46
In order o esablish he exisence or absence of a J-curve phenomenon in hese counries, we employ impulse response funcion o race he effec of real effec exchange rae on he rade balance of hese counries. As impulse response funcions (IRFs) are generaed from a coinegraed sysem, a shock in any variable is expeced o exer a permanen effec on he sysem, which gradually adjuss o a new equilibrium. The impulse response funcions ploed in Figures 7-12 indicae 10-year response of rade balance o an iniial and one-ime shock in real effecive exchange rae. As o wheher he lagged effecs presen he J-curve phenomenon, he evidence in his paper is as mixed. A classic J-curve paern is observed in case of Morocco/Japan in which he rade balance deerioraes almos immediaely following he devaluaion of he real exchange rae and recovers afer wo years. Long-run equilibrium is esablished afer abou 7 years. Furhermore, in cases of Egyp/Japan and Tunisia/US, we could observe some kind of delayed J-curve effecs in which, following he currency devaluaion, rade balances experience shor-erm improvemen ha lass for abou one year in boh counries. Afer he second year, rade balances elapses ino deerioraion and coninue o worsen up o he 7 h year in case of Egyp and 10 h year in case of Tunisia. In case of Egyp/US, Morocco/US and Tunisia/Japan, here seem o be no clear evidence of he J-curve phenomenon. So, in general, he resuls of impulse response funcion analysis in his sudy reveal ha he effecs of exchange rae changes on he rade balance are more pronounced in cases of Egyp/Japan, Morocco/Japan and Tunisia/US. Hosed by The Berkeley Elecronic Press
Figure 7: Impulse Response of TB o RER (Egyp/US).10.05.00 DTrade Balance (in log) -.05 -.10 1 2 3 4 5 6 7 8 9 10 Time (Years) Figure 8: Impulse Response of TB o RER (Egyp/Japan) hp://services.bepress.com/ifa/15h/ar46
.16.12.08.04 DTrade Balance (in log).00 -.04 2 4 6 8 10 12 14 Time (Years) Figure 9: Impulse Response of TB o RER (Morocco/US).0000 -.0004 -.0008 -.0012 DTrade Balance (in log) -.0016 -.0020 2 4 6 8 10 12 14 Time (Years) Figure 10: Impulse Response of TB o REER (Morocco/Japan) Hosed by The Berkeley Elecronic Press
.0004.0000 -.0004 -.0008 DTrade Balance (in log) -.0012 -.0016 2 4 6 8 10 12 14 Time (Years) Figure 11: Impulse Response of TB o REER (Tunisia/US).0020.0015.0010.0005 DTrade Balance (in log).0000 -.0005 -.0010 2 4 6 8 10 12 14 Time (Years) Figure 12: Impulse Response of TB o RER (Tunisia/Japan) hp://services.bepress.com/ifa/15h/ar46
.0020.0015.0010.0005 D T rade B alance (in log).0000 -.0005 -.0010 2 4 6 8 10 12 14 Time (Years) CONCLUSIONS This paper examines he effecs of exchange rae changes on he bilaeral rade balance of Egyp, Morocco and Tunisia vis-à-vis he US and Japan, in addiion o aemping o esablish wheher hese effecs produce a J-curve phenomenon. The sudy employs he Johansen coinegraion and error-correcion model on he annual daa in he period 1970-2003. In general, he findings confirm he exisence of boh he shor-run dynamics and long-run causal relaionships beween rade balance and he se of specified independen variables. However, he resuls produce mixed evidence abou he exisence of he J-curve effec. A classic J-curve paern is observed only in case of Morocco/Japan in which he rade balance deerioraes almos immediaely following he devaluaion of he real exchange rae and recovers afer wo years. In cases of Egyp/Japan and Tunisia/US, we could observe some kind of delayed J-curve effecs in which, following he currency devaluaion, rade balances experience shor-erm improvemen firs ha lass for abou one year in boh counries. However, in cases of Egyp/US, Morocco/US and Tunisia/Japan, no clear evidence of he J-curve phenomenon is found. Hosed by The Berkeley Elecronic Press
The findings ha exchange rae depreciaion/devaluaion improves rade balance (in cases of Egyp/Japan, Morocco/Japan and Tunisia/US) corroborae he conjecure in he lieraure ha a counry s rade balance may deeriorae following a currency depreciaion/devaluaion before improving in he long run. The finding of J-curve in hree cases ou of six in his paper does no help much in clearing he murk surrounding he issue of he J-curve phenomenon in he lieraure. These findings underscore he imporan implicaion ha he effec of real exchange rae changes on he rade balance of any given counry depends largely on he volume of rade aciviies beween he counries concerned. However, more research is needed o find conclusive evidence abou he exisence of J-curve phenomenon ha have commonly characerized he responses of rade balances of many counries o exchange rae changes. REFERENCES Baharumshah, A. Zubaidi. 2001. The Effec of Exchange Rae on Bilaeral Trade Balance: New Evidence from Malaysia and Thailand. Asian Economic Journal, 15, (3):291-312. Bahmani-Oskooee M. 1991. Is here a long-run relaionship beween he rade balance and he real effecive exchange rae o LDCs? Economics Leers, 36:403-407. Bahmani-Oskooee, M. and Farhang Niroomand. 1998. Long-run Price Elasiciy and Marshall-Lerner Condiion Revisied. Economics Leers, 61:101-109. Bahmani-Oskooee M. 2001. Nominal and real exchange raes of Middle Easern counries and heir rade performance. Applied Economics, 33:103-111. Bahmani-Oskooee, M. and Gour Gombinda Goswami. 2004. Exchange rae sensiiviy of Japan s bilaeral rade flows, Japan and he World Economy, 16: 1-15. Buluswar, Murli D., Henry T. and K. P. Upadhaya. 1996. Devaluaion and he rade balance in India: saionariy and coinegraion. Applied Economics, 28: 429-432. Canei, E. and Joshua G. 1991. Moneary Growh and Exchange Rae Depreciaion as Causes of Inflaion in African Counries: An Empirical Analysis. IMF Working Paper, WP/91/67. hp://services.bepress.com/ifa/15h/ar46
Charemza, W. W. and D. F. Deadman. 1992. New Direcions in Economeric Pracice: General o Specific Modelling, Coinegraion and Vecor Auoregression, Aldersho: Edward Algar Publishing Limied, England. Chiang, T.C. and D. Kim. 2000. Shor-erm Eurocurrency Rae and Specificaion of Coinegraing Process. Inernaional Review of Economics and Finance, 9:157-179. Demirden, T. and I. Pasine 1995. Flexible exchange rae and J-curve: An alernaive approach. Economics Leers, 48:373-377. Domaç, I. and G. Shabsingh 1999. Real Exchange Rae Behavior and Economic Growh: Evidence from Egyp, Jordan, Morocco and Tunisia. IMF Working Paper, WP/99/40. Enders, W. 1995. Applied Economeric Time Series, New York: John W. & Sons, INC. Granger, C. W. J. 1969. Invesigaing Causal Relaions by Economeric Models and Cross-Specral Mehods. Economerica, 37 (3):424-438. Gujarai, D. N. 1995. Basic Economerics (3 rd Ediion). Singapore: McGraw-Hill Book Company. Gupa-Kapoor, A. and Uma R. 1999. Is here a J-curve? A New Esimaion for Japan, Inernaional Economic Journal, 13 (4):71-79. Hamilon, J. D. 1994. Time Series Analysis. New Jersey: PRINCETON UNIVERSITY PRESS, NEW JERSEY. Himarios, D. 1985. The Effecs of Devaluaion on he Trade Balance: A Criical View and Re-examinaion of Miles s New Resuls, Journal of Inernaional Money and Finance, 4:553-563. Krueger, A. O. 1983. Exchange Rae Deerminaion, Cambridge, Cambridge Universiy Press. Lal, A. K. and T. C. Lowinger 2002. Nominal effecive exchange rae and rade balance adjusmen in Souh Asia counries. Journal of Asian Economics, 13: 371-383. Hosed by The Berkeley Elecronic Press
Marinez, J. De La C. 1999. Mexico's Balance of Paymens and Exchange Raes: A Coinegraion Analysis, Norh American Journal of Economics and Finance, 10:401-421. Masih, Abdul M. M. and R. Masih 1996. Modeling he Dynamics of Macroeconomic Aciviy: New Evidence from a Developing Economy. Journal of Quaniaive Economics, 12 (2):85-105. Masih, Abdul M. M. and R. Masih 2004. Fracional coinegraion, low frequency dynamics and long-run purchasing power pariy: an analysis of he Ausralian dollar over is recen floa. Applied Economics, 36:593-605. Miles, M. A. 1979. The effecs of devaluaion on he rade balance and he balance of paymens: Some new resuls. Journal of Poliical Economy, 87:600-620. Narayan, P. K. 2004. New Zealand s rade balance: evidence of he J-curve and granger causaliy. Applied Economics Leers, 11:351-354. Onafowora, O. 2003. Exchange Rae and rade balance in Eas Asia: Is here a J- curve? Economic Bullein, 5 (18):1-13. Pilbeam, K. 1992. Inernaional Finance. MACMILLAN PRESS LTD, Hampshire and London. Rahman, M. and Muhammad M. 1997. Dynamics of he yen-dollar real exchange rae and he US-Japan real rade balance. Applied Economics, 29:661-664. Rawlins, G. and John P. 1993. Devaluaion and he Trade Balance: The Recen Experience of Seleced African Counries. Cenre for Economic Research on Africa, School of Business, Monclair Sae Universiy, Upper Monclair, New Jersey. Rose, A. K. and Jane L. Y. 1989. Is here a J-curve? Journal of Moneary Economics, 24:53-68. Rose, A. K. 1991. The Role of exchange raes in a popular model of inernaional rade: Does he Marshall Lerner condiion hold? Journal of Inernaional Economics, 24:301-316. Singh, T. 2002. India s rade balance: he role of income and exchange raes. Journal of Policy Modeling, 24:437-452. hp://services.bepress.com/ifa/15h/ar46
Suc ka, T. 2003. The Impac of Exchange Rae Changes on he Trade Balance in Croaia. Working Papers, Croaian Naional Bank, Zagreb. Swif, R. 1998. Exchange Rae Pass-Through: How much do Exchange Rae Changes affec Prices of Ausralian Expors? Ausralian Economic Papers, 37 (2):169-184. Upadhyaya, K. P. and Dharmendra D. 1997. Devaluaions and he Trade Balances: Esimaing he long-run effec. Applied Economic Leers, 4:343-345. Upadhyaya, K. P., Franklin G. M. and Dharmendra D. 1999. Do Devaluaions Improve Trade Balances? Evidence from Four Souh Asian Counries. Indian Economic Journal, 46 (3):91-97. Weixian, W. 1999. An Empirical Sudy of he foreign rade balance in China. Applied Economics Leers, 6:485-490. Williamson, J. 1983. The Open Economy and he World Economy. Basic Books, New York: John Wiley & Sons, INC. Wilson, P. 2001. Exchange Raes and he Trade Balance for Dynamic Asian Economies Does he J-Curve Exis for Singapore, Malaysia and Korea? Open Economies Review, 12:389-413. Wilson, P. and Kua Choon Ta 2001. Exchange rae and rade balance: he case of Singapore 1970 o 1996. Journal of Asian Economics, 12:47-63. Zhang, Z. 1999. Foreign Exchange Rae Reform, he rade balance and economic growh: An empirical analysis for China. Journal of Economic Developmen, 24 (2):143-162. Hosed by The Berkeley Elecronic Press