COINTEGRATION AND CAUSALITY AMONG EXCHANGE RATE, EXPORT, AND IMPORT: EMPIRICAL EVIDENCE FROM TURKEY SEKMEN, Fuat * SARIBAS, Hakan

Transcription

1 Applied Economerics and Inernaional Developmen Vol.7-2 (2007) COINTEGRATION AND CAUSALITY AMONG EXCHANGE RATE, EXPORT, AND IMPORT: EMPIRICAL EVIDENCE FROM TURKEY SEKMEN, Fua * SARIBAS, Hakan Absrac This paper examines he coinegraion and causaliy among exchange rae, expor, and impor for Turkey during he period of The economerics resuls show ha here is a coinegraion beween expors and impor, bu direcion of causaliy is bidirecional beween hese wo variables. The impulse response funcions also suppors ha here is a rade-off beween expors and impors; for example, when impors are high, here is smaller expors a ha ime. This sudy suppors few invesigaors who find no negaive effec of exchange rae volailiy on rade volume since i is found ha exchange raes canno deermine he variaion in expors and impors. JEL Classificaions: F4, C3/C32 Key Words: Expor, Impor, Exchange Rae, Coinegraion, and Impulse Response Funcion. Inroducion This sudy examines relaionship among exchange rae, expor, and impor for Turkish economy. Previous sudies have invesigaed he effec of exchange rae on rade volume bu hey have no reached an agreemen among hemselves. Some economiss, such as Ehier (973), Hooper and Kohlhagen (978), Kumar and Dhawan (99), Gagnon (993), Broll (994), Caporale and Dorodioon (994), Wolf (995), Dell Ariccia (998), Rose (2000), and Vergil (2002) illusrae ha exchange rae volailiy reduces inernaional rade. The main idea behind he inuiion of exchange rae volailiy decreases rade volume is ha exchange rae volailiy increases uncerainy, which in urn decreases rade volume. On he oher hand, De Grauwe (988), and Giovannini (988), Franke (99), De Grauwe (988), Franke (99), Grobar (993), McKenzie and Brooks (997), Dewlin e al. (200) saed ha exchange rae volailiy or risk may acually simulae rade flows since uncerainy is considered as an opion held by firms, which increases profiabiliy. Kroner and Lasrapes (993), McKenzie e al. (998), and Arisoelous (200), in conras o oher sudies, find no cerain evidence for he relaionship beween exchange rae volailiy and rade volume. Moccero and Winograd (2006) invesigae he effec of exchange rae volailiy by examining he inra and exra regional expors and hey conclude ha reducing volailiy has a posiive impac on expors o Brasil, bu a derimenal effec on expors o he res of he world. This resul also suppors he view ha here is an ambiguiy in he role of exchange rae volailiy on rade volume. There are a few sudies which examine separaely he negaive effec of exchange rae volailiy on impors, for example, Gour (985) and Cushman (986) invesigaed he * Fua Sekmen, Sakarya Universiy, Deparmen of Economics, Ass. Prof.Dr., Sakarya, Turkey, sekmen@sakarya.edu.r. Hakan Saribas, Zonguldak Karaelmas Universiy, Deparmen of Public Finance, Ass. Prof.Dr., Zonguldak, Turkey.

2 Applied Economerics and Inernaional Developmen Vol.7-2 (2007) effec of exchange rae volailiy on impors. Mos of he sudies have examined he effec of exchange rae volailiy on expors and on rade volume. This paper has focused on he associaion among exchange raes, expor, and impor raher han he effec of exchange rae. This sudy will also shed ligh he direcion of causaliy among exchange rae, expor, and impor. The organizaion of he paper will be as followed: secion 2 represens mehodology and daa, secion 3 shows he resuls and discussions, and secion 4 presens conclusion. 2. Daa and mehodology Monhly daa for Expors (X) and Impors (M) for , a curren prices and exchange raes, million US Dollars, Source: Turkish Treasury. Nominal exchange rae: The end of each monh is aken ino accoun, Purchasing of USD agains Turkish New Liras. Afer 2005, New Turkish Liras has been used insead of Turkish Liras. Source: Turkish Cenral Bank(2006). Graph shows he evoluion of Expors and Impors per inhabian in Turkey in dollars a curren prices, and graph 2 presen he evoluion of he exchange rae in new liras. 600 Graph. Foreign rade in Turkey (curren dollars per inhabian).6 Graph 2. Exchange rae of Turkey (new Turkish liras per dollar) Impors per inhabian Expors per inhabian Economeric mehodology firsly examines ime series seleced if hey exhibi saionariy or no because hese ess are necessary. For his purpose, his paper uses he Augmened Dickey Fuller (ADF) and Phillips-Perron (PP) ess. A mehodological noe on hese ess is included in he Annex. Secondly Coinegraion is analysed wih Johansen es and causaliy wih Granges es (see mehodological noe in he Annex). Afer being found coinegraion among series aken in his sudy, he paper will examine he relaionship among expors, impors, and exchange rae for Turkey using variance decomposiion mehods. For his purpose, The paper uses nominal exchange rae daa. Nex sudy can compare he resuls of nominal exchange rae and real exchange rae. 72

3 Sekmen, F., Saribas, H. Coinegraion and Causaliy Among Exchange Rae, Expor & Impor monhly daa for expors, impors, and nominal exchange raes of Turkey since 998 unil 2006 will be used in his sudy. 3. Resuls and Discussions 3.. Resuls of he uni roo ess: The resuls of he ADF and he PP ess for saionariy properies of he variables are presened in Table A. The able A in he Annex shows ha he τ saisics for all he variables (X, M, and ER) are no significanly negaive since hey are posiive and greaer han he criical values a, respecively, %, 5%, and 0% levels from boh he ADF and he PP ess. Thus, i is no possible o rejec he null hypohesis of he presence of he uni roo for he variables. However, he resuls of he firs differenced variables indicae ha he ADF es saisics for all he variables are significanly negaive, herefore, he null hypohesis of uni roo can be rejeced in all variables a %, 5%, and 0% levels. The PP es also shows ha afer differencing all he variables are saionary, meaning ha all he variables are inegraed of order I() Resuls of ADF Coinegraion Tess: In order o find ou if he variables are coinegraed, he ADF uni roo es was applied on he residuals from he hree equaions, where M was regressed on X in he firs equaion, hen ER was regressed on X in he second equaion, and in he hird equaion ER was again regressed on M. Table A2 in he Annex presens he resuls from his analysis. From he able, τ saisics on residuals for M regressed on X is less han he criical value a 0 % level, hereby null uni roo hypohesis is rejeced for he equaion. However, he null uni roo hypohesis canno be rejeced since τ saisics on residuals are no less han he criical values a %, 5 %, and 0 % level. These resuls sugges ha only M is coinegraed wih X. This resul is meaningful in respec of economics heory because if a counry s rade balance is no good oday, his does no mean ha his counry may no achieve a susainable growh in he fuure. For example, if inermediae goods which are crucial for expors are impored and here is a rade balance defici, his counry can succeed expor-lead growh in he long run. Thus, in my view, Turkish rade defici is no so big problem as long as oday s impors can lead o higher expors in he fuure Resuls of Johansen Coinegraion Tes: Johansen (988) and Johansen and Juselius (990) developed coinegraion mehods in order o obain long-run relaionship among he series. According o Johansen coinegraion es, non-saionary series, which are obained from able, are esed wheher hese series reach long-run equilibrium. Table shows Johansen coinegraion es resuls. Table. Johansen Coinegraion Tes Resuls H 0 H Eigenvalue Trace Saisic %5 Criical Max-Eigen %5 Criical Value Saisic Value r =0 r r = r r = 2 r Noe. Sample: No deerminisic rend. Lags inerval (in firs differences): o. 73

4 Applied Economerics and Inernaional Developmen Vol.7-2 (2007) In he able, he null hypoehesis of r =0, here is no one coinegraing vecor, is esed agains he alernaive r. For he es based on he race saisic, i is 35.2 so he null is rejeced a 5% level, since he race saisic is calculaed as In he case of maximum eigenvalue saisic, he criical value is 22.3 so ha he null hypohesis can be rejeced a he 5% level, since he maximum eigenvalue saisic is calculaed as Thus, he resul based on he race saisic and maximum eigenvalue saisic represens ha here is a leas one coinegraing vecor. The nex sep is o es he null hypohesis of r = agains he alernaive hypohesis of r 2, meaning here migh be wo coinegraing vecors. In his case, he null canno be rejeced using eiher race saisic (he 5% criical value is 20.3 while calculaed value is 7.7) or he maximum eigenvalue saisic (he 5% criical value is 5.9 while he calculaed value is 2.2), and so i can be concluded ha here is exacly one coinegraing vecor. Anoher sep is o es he null of r = 2 agains he alernaive of r 3. Here, he null canno be rejeced using eiher he race saisic (he 5% criical value is 9.2 while he calculaed value is 5.5) or he maximum eigenvalue saisic (he 5% value is 9.2 while he calculaed value is 5.5, and again i can be concluded ha here is exacly one coinegraing vecor Resuls of he Error Correcion Models: Table 2 presens he resuls of he ECM which conains hree equaions and each equaion includes adjusmen coefficien. Equaion akes ino accoun of expor (X) which has a negaive adjusmen coefficien ( ), his is also significan (saisic is -3.). Negaive coefficien means here is a endency from shor erm flucuaions o long erm equilibrium condiion. Thus, no coinegraing hypohesis can be rejeced and alernaive hypohesis is acceped. Table 2. Error Correcion Esimaes D(X) D(M) D(ER) Adj. coefficien Sandard Dev. (0.95) (0.255) (26.) -saisic (-3.2) (0.975) ( ) 3.5. The Resuls of Variance Decomposiion and Impulse Response Funcion: Variance decomposiion gives informaion abou he proporion of he movemens in he dependen variables ha are due o heir own shocks, versus shocks o he oher variables. A shock o any variable, for example a shock o expor, will direcly affec ha variable (expor), bu his shock will also be ransmied o all of he oher variables in he sysem (here impors and exchange raes) hrough he dynamic srucure of he Vecor Auoregression (VAR). Variance decomposiion deermine how much of he s-sep-ahead forecas error variance of a given variable is explained by innovaions o each explanaory variable for s =, 2, (Brooks, C., pp.342). The firs par of he able A3 shows he variance decomposiion of expors following a shock o expor innovaions of $ 274 million. In he firs round, he enire change in expor is explained only by a shock o he expor innovaion. This shock also causes an immediae change in impor and exchange rae, bu he resuling changes in hese variables have no effec on expor a his ime, since curren impors and exchange raes 74

5 Sekmen, F., Saribas, H. Coinegraion and Causaliy Among Exchange Rae, Expor & Impor have no effec on curren expors. In round wo, exchange rae variables accouns for 0.5% of he change in expors, however, impor accouns for 0.38% of he change in expors. When he enire 0-year period is aken ino accoun, he effec of exchange rae on expor, following he iniial shock o he expor innovaion, is negligible, never geing larger han 3%, bu he effec of impors on expors is geing larger han 50% afer he 0-year period. The second par of he able A3, which races he variance decomposiion of impors, presens reacions following a shock o impor innovaions of $ million. Because of he ordering expors-impors-exchange raes, a shock o impor innovaions has an immediae effec on impors and curren expor which has a negligible impac on curren impors a his poin. In round one, a shock o he impor innovaion accouns for 99% of he variaion in he impor variable, while expor accouns for he res of he variaion. In round wo, expor accouns for 9% of he variaion in impors and impors iself accoun for 9% of is own variaion. The influence of expors on impors increases in ime and by he end of he 0-year period i is accouning for 5% of he oal variaion in impors. Impors accouns for 83% of is own variaion in he round 0. The hird par of he able A3 shows he variance decomposiion of exchange rae following a shock o he exchange rae innovaions of $ million. In he firs round, a shock o he exchange rae innovaion accouns for almos he enire change in exchange rae variaion (99.63%). This siuaion is almos he same for all period, for example, a he end of he 0-year period, he variaion in exchange rae because of he iniial shock is accouned by exchange rae iself (97.58%). In sum, he variance decomposiion resuls indicae ha impors and exchange rae are he mos exogenous variables as a high proporion of heir shocks are explained by heir own innovaions compared wih he conribuions of own shocks o innovaions for expors. As saed above, a he end of 0 years, he forecas error variance for impor and exchange rae explained by heir own innovaions are 83.35% and 97.58%, respecively. An alernaive way of obaining informaion regarding he relaionship among he variables, here expor, impor, and exchange rae, is via generalized impulse response funcions. Figures A-A3 in he Annex presen impulse response funcions The Direcion of he Causaliy; Resuls from he Granger-Causaliy Tes: The resuls of he Granger-causaliy es (GC) are presened in he able 3. The GC es resuls show ha he null hypohesis of impors does no Granger cause expors is rejeced a % level. Also, here is reverse causaion from X o M, since he F value is saisically significan. This conclusion also suppors he variance decomposiion resuls, which are presened in he Annex. On he oher hand, as saed before, here is no any associaion beween exchange rae and expors and impors, he null hypohesis of exchange rae does no Granger cause expor and impors canno be rejeced. In he able 3, here is a relaionship beween exchange rae (ER) and expors (X), bu he null hypohesis of ER does no Granger cause X canno be rejeced since he F value is no saisically significan. 75

6 Applied Economerics and Inernaional Developmen Vol.7-2 (2007) Table 3. Granger Causaliy Tes Direcion of causaliy F-Saisic Probabiliy M X E-6 X M ER X X ER ER M M ER As seen in Guisan (2005) and (2006), he bilaeral relaionship beween Impors and Expors is of uppermos imporance for economic developmen. Expors increase he capaciy o Impor and besides real Impors have usually an imporan posiive impac on indusrial and non-indusrial real Gross Domesic Produc. The esimaed effec of an increase of 00 dollars in real Impors and Expors on real GDP was: 62 in Canada, 79 in Mexico, 76 in he pool of 4 European Union counries, 9 in Spain, 32 in Turkey and 40 in he USA. All hese values are expressed in dollars a 2000 prices and exchange raes bu in he case of he USA hey are expressed a 990 prices. This auhor recommends indusrial developmen for hose counries wih low levels of foreign rade in order o induce direc and indirec posiive effecs on rade and real GDP per inhabian. Alhough Turkey has experienced a high increase of indusrial developmen and foreign rade per inhabian for he period , we may sae ha he levels are sill very low in comparison wih oher OECD counries, and ha an increase in hese variables will be posiive for economic developmen in Turkey. 4. Conclusion This paper examined coinegraion and causaliy among exchange rae, expor, and impor, a curren prices, using Turkish monhly daa for period exraced from Turkish Treasury. The resuls show ha expor (X) is coinegraed wih impor (M). I migh be hough ha negaive rade balance (he difference beween impors and expors) can be reversed in he long-run by increasing expors if impored goods bring new echnology and a differen enrepreneur spiri o home counry. Therefore, expors and impors are coinegraed. This resul is suppored by coinegraion es resuls; for example, Johansen coinegraion es has concluded ha here is exacly one coinegraed vecor. Also, he resul of he variance decomposiion has presened a relaionship beween expors and impors since a shock o expor or impor can be explained by expors and impors. The direcion of causaliy has suppored coinegraion beween expors and impors because bi-direcional causaliy has been found beween hese wo variables. Anoher resul from his paper is ha here is no relaionship beween exchange rae and rade a curren prices. The variance decomposiion es explains ha he power of exchange rae o explain he change in expors and impors a curren prices is no larger han 3% afer he 9-year period. Exchange rae accouns for he change in iself only when here is a shock o exchange rae innovaions. 76

7 Sekmen, F., Saribas, H. Coinegraion and Causaliy Among Exchange Rae, Expor & Impor References Afzal, M. (2006). Causaliy beween expors, world income, and economic growh in Pakisan, Inernaional Economic Journal, 20(), Arisoelous, K. (200). Exchange rae volailiy, exchange rae regime, and rade volume: evidence from he UK-US expor funcion ( ), Economics Leers, 72, Brooks, C. (2005). Inroducory economerics for finance, Cambridge Universiy Press, UK. Broll, U. (994). Foreign producion and forward markes, Ausralian Economic Papers, 33(62), -6. Caporale, T. and K. Doroodian (994). Exchange rae variabiliy and flow of inernaional rade, Economics Leers, 46, Chuah, M. K. L., Collen, D., VandenDriessche, T. and McKenzie, M. D. (998). The impac of exchange rae volailiy on Ausralian rade flows, Journal of Inernaional Financial Markes, Insiuions and Money, 8 (), Cushman, D. O. (986). Has exchange rae risk depressed inernaional rade? The impac of hirdcounry exchange risk, Journal of Inernaional Money and Finance, 5, De Grauwe, P. (988). Exchange rae variabiliy and he slowdown in he growh of inernaional rade, IMF Saff Papers, 35, Dell Ariccia, G. (998) Exchange rae flucuaions and rade flows: evidence from he European Union, IMF Research Paper (07). Dewlin, R., Esevadeordal, A., Moneagudo, J. and S. Raul (200). Macroeconmic sabiliy, rade and inegraion, Inegraion and Trade, 3, Ehier, W. (973). Inernaional rade and forward exchange marke, American Economic Review, 63(3), Franke, G. (99). Exchange rae volailiy and inernaional rade sraegy, Journal of Inernaional Money and Finance, 0, Gagnon, J. E. (993). Exchange rae variabiliy and he level of inernaional rade, Journal of Inernaional Economics, 34, Gour, P. (985). Effecs of exchange rae volailiy in rade; some furher evidence, IMF Saff Papers, 32, Grobar, L. M. (993). The effec of real exchange rae uncerainy on LDC manufacured expors, Journal of Developmen Economics, 4, Guisan, M.C. (2005). Human Capial, Populaion Growh and Indusrial Developmen in Mexico and Turkey. A Comparaive Analysis wih Oher OECD Counries, , Working Paper Series Economic Developmen, number 85 77

8 Applied Economerics and Inernaional Developmen Vol.7-2 (2007) Guisan, M.C. (2006). Indusry, Foreign Trade and Developmen: Economeric Models of Europe and Norh America, , Inernaional Journal of Applied Economerics and Quaniaive Sudies, Vol. 3-, pp Hooper, P. and S.W. Kohlhagen (978). The Effecs of exchange rae uncerainy on he prices and volume of inernaional rade, Journal of Inernaional Economics 8, Johansen, S. (988). Saisical analysis of coinegraion vecors, Journal of Economic Dynamics and Conrol, 2, Johansen, S. and K. Juselius (990). Maximum likelihood esimaion and inference on coinegraion-wih Applicaion o he Demand for Money, Oxford Bullein of Economics and Saisics, 52, Kroner, K. F. and W.D. Lasrapes (993). The impac of exchange rae volailiy on inernaional rade: reduced form esimaes using he GARCH-in-mean model, Journal of Inernaional Money and Finance, 2, Kumar, R. and R. Dhawan (99). Exchange rae volailiy and Pakisan s expors o he developed world, , World Developmen, 9 (9), McKenzie, M. D. and R. D. Brooks (997). The impac of exchange rae volailiy on German-US rade flows, Journal of Inernaional Financial Markes, Insiuions and Money, 7, Moccero, D. N. and C. Winograd (2006). Real exchange rae volailiy and expors: ArgeninePerspecives, Phillips, P.C.B. and P. Perron (988). Tesing for a Uni Roo in Time-Series Regrssion, Biomerica, 75, Rose, A. K. (2000). One money, one marke: esimaing he effec of common currencies on rade, Economic Policy, 5, Turkish Cenral Bank(2006). hp: Turkish Treasury(2006). hp: Vergil, H. (2002). Exchange rae volailiy in Turkey and is effec on rade flows, Journal of Economic and Social Research, 4, Wolf, A. (995). Impor and hedging uncerainy in inernaional rade, Journal of Fuure Markes, 5, 0-0. * On line Annex a he journal websie Journal published by he EAAEDS: hp:// 78

9 Sekmen, F., Saribas, H. Coinegraion and Causaliy Among Exchange Rae, Expor & Impor Annex The ADF es is based on he following regressions: n 0 + αy + αi i= y = α yi + e () 0 i i= n y = α + α y + α yi+ δ+ e (2) where y is a ime series, is a linear ime rend, is he firs difference operaor, α is a consan, n is he opimum number of lags on he dependen variable, and e is 0 random error erm. The difference beween he equaion and 2 is ha he firs equaion includes jus drif, however, he second one includes boh drif and linear ime rend. The null hypohesis for esing nonsaionariy is H α 0, meaning economic series are nonsaionary. Tha is y 0 : = is a random walk and i has a uni roo. If he -saisic associaed wih esimaed coefficien, hereα, is less han he criical values for he es, he null hypohesis of no-coinegraion canno be rejeced a or 5 or 0 % level of significance. This sudy also employs he PP es since he possibiliy of he presence of srucural breaks makes he ADF es unreliable for esing saionariy. The presence of srucural breaks will end o bias he ADF es owards non rejecion-of he null hypohesis of a uni roo. Therefore, his paper has used boh Phillips-Perron es (PP) suggesed by Phillips and Perron (988) and he ADF es o examine he saionariy of he daa, whereas, oher sudies have used eiher he ADF or he PP es. Table A. Resuls of he ADF and PP uni roos ess Variable Augmened Dickey- Philips-Perron es (ADF) Fuller es (PP)es Level Firs Level Firs Form differences Form differences X M ER Criical values a %, 5% and 0% significance levels: -3.50, and Having esablished he hypohesis of non-saionariy for he underlying variables, here expor, impor, and exchange rae, he ime series daa will be examined for coinegraion using he ADF coinegraion ess. This paper will also employ Johansen s coinegraion es because, as saed by Afzal (2006), his es can esimae several coinegraion relaionship and fully capures he underlying ime series properies of he daa. Because of he number of coinegraion vecor is no known and only endogenous 79

10 Applied Economerics and Inernaional Developmen Vol.7-2 (2007) variables are need o be known, using of he one equaion model is resriced. However, Johansen (988) has exended he echnique of vecor auoregression (VAR) model by which i is possible o idenify he correc coinegraion among he series. If series are coinegraed, he sandard Granger bivariae causaliy es will be performed. Esimaing he following equaions performs he sandard Granger causaliy es: X = α o + αi X i + δ M i i + µ (3) M = β 0 + ϕim i + βi X i + e (4) According o equaion 3, one variable M is said o Granger cause anoher variable, X, if X can be explained by using pas values of X. Here, X (expor) and M (impor) are wo separae economic ime series. Equaion 4 also represens he same relaionship wih equaion 3. Causaliy can be found by esing he null hypohesis H δ = β 0. If 0 = i i = δ iandβ i are significan, here will be bi-direcional causaliy. On he oher hand, X Granger causes M if β i is saisically significan bu δ i is no; and X Granger causes M if δ i is saically significan and β i is no. This is called unidirecional causaliy. However, Granger (988) saes ha he sandard Granger or Sims ess are likely o provide invalid causal inferences when he ime series are coinegraed. This is because error correcion model mus be used insead of sandard Granger-causaliy es. As o Engle and Granger (987), he usual ECM echnique may ake he following form: m X = γ e γ X + Φ M + 0 i i i i= i= e V m V (5) where denoes firs difference operaor, is he error correcion erm, m is he number of lags o obain whie noise and is anoher random disurbance erm. If he coefficien of he error correcion erm is significanly differen from zero 2, in and his case, his will sugges ha boh series, X and M, exer a long-run relaionship. Table A2. Resuls of ADF coinegraion ess Regression Tes saisic R- D.W Tes R- D.W. equaion (Coefficien) square saisic square (residuals) M on X ER on X ER on M Criical values a %, 5% and 0% significance levels: -3.50, and The absolue value of he error correcion erm is aken ino accoun. 80

11 Sekmen, F., Saribas, H. Coinegraion and Causaliy Among Exchange Rae, Expor & Impor Figure. Impulse response of expor (X) o a one-sandard deviaion shock in expor, impor (M), and exchange rae (EX) Figure 2. Impulse response of impor o a one-sandard deviaion shock in impor, expor, and exchange rae 8

12 Applied Economerics and Inernaional Developmen Vol.7-2 (2007) Figure 3. Impulse response of exchange rae o a one-sandard deviaion shock in exchange rae, expor, and impor Figure plos he response of expor o a shock in expor, impor, and exchange rae. The mos ineresing resul from he figure is he response of expor o a shock in impor; expor falls whenever impor increases. Meanwhile he response of expor o a shock in exchange rae is he opposie. Afer he firs wo years, expor increases wih exchange rae bu i is sable afer eigh-year period. Figure 2 plos he response of impors o a shock in impor, expor, and exchange rae. In response o a shock in expor, impor declines, he response of impors o a shock in expor is similar o figure. However, in response o a shock in exchange rae, impor is no affeced. Figure 3 plos he response of exchange rae o shocks in expor, impor, and exchange rae. Own shocks explain he variaion in exchange rae, bu shocks in expor and impor slighly affec variaion in exchange rae. 82

13 Sekmen, F., Saribas, H. Coinegraion and Causaliy Among Exchange Rae, Expor & Impor Table A3. Decomposiion of variance (percenage of forecas variance explained by innovaions) Variance decomposiion of expors Period S.E. Expor Impor Exchange Rae Variance decomposiion of impors Period S.E. Expor Impor ExchangeRae Variance decomposiion of exchange rae Period S.E. Expor Impor ExchangeRae