CAUSALITY BETWEEN EXCHANGE RATE AND ECONOMIC GROWTH IN BANGLADESH

Transcription

1 CAUSALITY BETWEEN EXCHANGE RATE AND ECONOMIC GROWTH IN BANGLADESH Kazi Mohammed Kamal Uddin PhD Candidae School of Economics Huazhong Universiy of Science and Technology Wuhan, Hubei, P.R.China Mohammad Morshedur Rahman PhD Candidae School of Managemen Huazhong Universiy of Science and Technology Wuhan, Hubei, P.R.China G. M. Azmal Ali Quaosar Assisan Prof., Deparmen of Managemen Sudies, Comilla Universiy, Comilla, Bangladesh Absrac This paper aemps o examine he relaionship beween Exchange Rae (ER) and Economic Growh (EG) proxied by Real Gross Domesic Produc (RGDP) in Bangladesh for a period of 4 years ranges from 973 o 203 by using ime series economeric echnique. The empirical resuls show ha here is a significan posiive correlaion beween ER and EG. The resuls also advocae he presence of long-run equilibrium relaionship beween ER and EG. This is evidenced from Granger s Causaliy Tes ha here is a bi-direcional causaliy runs hrough ER o EG and EG o ER. Keywords: Exchange Rae, Economic Growh, Real Gross Domesic Produc, Bangladesh, Uni Roo, Co-inegraion, Causaliy Inroducion Bangladesh is said o be a counry wih grea poenials. Though many consider i over burdened wih huge populaion, many oher consider his populaion as asse. Talking he posiive aspecs, his populaion can easily conribues economic growh (Rahman e al., 2006).Talking abou he negaive aspecs, his populaion requires a huge amoun of daily necessiies ha may no be possible for Bangladesh o produce and hus hey fare bound o go for foreign rade. Over he pas few decades, he nexus beween Exchange rae and economic growh have drawn exensive aenion of macroeconomiss, policy makers and he cenral bankers of boh developed and developing counries. Specifically, he issue ha wheher exchange rae

2 is necessary for economic growh or i is harmful generaes a significan debae boh heoreically and empirically. The exchange rae is defined as he domesic price of a uni of foreign currency. Exchange rae can be called he conversion facor ha deermines he rae of change of currencies. Real and nominal exchange raes are differen from each oher. If we leave ou he inflaion influence hen i is real exchange rae and if we incorporae he inflaion influence hen i is nominal exchange rae. The nominal exchange rae can be expressed in bilaeral and mulilaeral erm. Real exchange rae volailiy means he shor erm oscillaion of he real exchange rae. A differen paern of exchange rae behavior ino caagories is known as exchange rae regime. In which exchange rae remains fixed is called fix exchange rae regime and in which exchange rae flucuaes is known as floaing exchange rae regime. The middle of fix and floaing exchange rae is called managed floa regime. The gross domesic produc (GDP) is one he primary indicaors used o deermine he healh of a counry's economy. I represens he oal dollar value of all goods and services produced over a specific ime period - you can hink of i as he size of he economy. Usually, GDP is expressed as a comparison o he previous quarer or year. For example, if he year-o-year GDP is up 3%, his is hough o mean ha he economy has grown by 3% over he las year. Economiss have long known ha poorly managed exchange raes can be disasrous for economic growh. Avoiding significan overvaluaion of he currency is one of he mos robus imperaives ha can be gleaned from he diverse experience wih economic growh around he world, and one ha appears o be srongly suppored by cross-counry saisical evidence (Razin and Collins, 997). Cheung and Lai (998) considered he influence of relaive facors such as he per capia GDP o he foreign exchange reserves. Johansen (988) hough ha he foreign exchange reserve of some counries wih a rapid increase is he by-produc of he undervalued real exchange rae policy carried ou by hem aiming a promoing he expor, no ha hese naional moneary and financial auhoriies are inended. Review of Recen Lieraure A large number of research sudies relaing o various aspecs of exchange rae have been published home and abroad. However, criical reviews of some of he imporan research sudies/aricles have been made in his sudy. Glasure and Lee (999) examined he expor-led growh hypohesis for Korea in five-variable vecor auoregressive and vecor error correcion models from 973: o 994:4. Resuls of he vecor auoregressive models indicae economic growh Granger-causing expor growh, regardless of he sample period. However, resuls of he vecor error 2

3 correcion models show bidirecional causaliy beween expor growh and economic growh when he mulivariae generalizaions of he Granger causaliy ess are used. In he variance decomposiions, he real exchange rae conains mos informaion regarding fuure flucuaions in economic growh and expor growh followed by money supply and governmen expendiure in he subsample and he full sample, wih economic growh as a dependen variable. However, when he dependen variable is expor growh, hen he order of he magniude in he full sample becomes he exchange rae, followed by economic growh, governmen spending, and money supply. The findings in his paper sugges ha he omied variables have masked or oversaed he effec of expors on income or income on expors in prior sudies. Mundaca and Srand (2005) derive he opimal exchange rae policy for a small open economy subjec o erms-of-rade shocks. Firm owners and workers are risk averse bu workers more so. Wages are given or parially indexed in he shor run, and capial markes are imperfec. The governmen ses he exchange rae o allocae risk beween workers and owners. Wih less risk-averse firms, and greaer difference in risk aversion beween workers and firms, he opimal exchange rae should vary lile wih pure erms-of-rade shocks bu more wih general shocks o prices. Opimal exchange rae variaion is greaer wih indexed wages, bu is smaller when firms behave monopolisically and when wage axes (profi axes) change procyclically (counercyclically) wih expor prices (impor prices). The model gives policy rules for deermining opimal variaions of he exchange rae, and indicaes when i is, and is no, opimal o join a currency union wih rading parners, implying zero exchange rae variaion. Iō (2005) provided of pas work covering use of he yen on PPP, covered and uncovered ineres rae pariy, he unbiasedness of expeced fuure exchange raes, volailiy spillover across borders and he effeciveness of inervenion. He ried o discuss he role of he yen in he inernaional financial srucure and is fuure role in global and regional financial markes. Gala (2008) inended o conribue o he debae by bringing more heoreical elemens and providing new economeric evidence o he connecions beween real exchange rae levels and developmen. Aghion e al. (2009) offered empirical evidence ha real exchange rae volailiy can have a significan impac on produciviy growh. However, he effec depends criically on a counry s level of financial developmen. The resuls appear robus o ime window, alernaive measures of financial developmen and exchange rae volailiy, and ouliers. They also offer a simple moneary growh model in which real exchange rae uncerainy exacerbaes he negaive invesmen effecs of domesic credi marke consrains. 3

4 Alba e al (200) examined he impac of exchange raes on foreign direc invesmen (FDI) inflows ino he Unied Saes in he conex of a model ha allows for he inerdependence of FDI over ime. Inerdependence is modeled as a wo-sae Markov process where he wo saes can be inerpreed as eiher a favorable or an unfavorable environmen for FDI in an indusry. They use unbalanced indusry level panel daa from he US wholesale rade secor and heir analysis yields wo main resuls. Firs, hey find evidence ha FDI is inerdependen over ime. Second, under a favorable FDI environmen, he exchange rae has a posiive and significan effec on he average rae of FDI inflows. Ok e al. (200) sudied sources of flucuaions in real and nominal US dollar exchange raes in Cambodia and Lao PDR by decomposing hem ino he componens induced by real and nominal facors. These shocks affecing real and nominal exchange raes are idenified by using a srucural vecor auo-regression (SVAR) model wih he long-run neuraliy resricion. The empirical analysis demonsraed ha real shocks in direcion of depreciaion lead o real and nominal depreciaion, while nominal shocks induce long-run nominal depreciaion bu real appreciaion in he shor-run. Ba and Shen (200) choose six indusries o divide hem ino hree groups based on per capia possession of capial, and hen employ he monhly daa from 200 o 2008 o carry ou EG wo-sep co-inegraion es, and finally analyze he impacs of he US economic growh and he exchange rae variabiliy on differen expor indusries. Empirical resuls show ha he labor-inensive indusries are mos suscepible o flucuaions brough by economic growh and real exchange rae, while hose indusries wih higher per capia possession of capial are less suscepible o exernal facors. In he shor run, he expor of labor-inensive producs gives an advanage o China s foreign rade developmen, bu in he long-run, hese indusries will be affeced grealy by various uncerain facors and he advanages of China s labor-inensive expor indusries will disappear wih he shif of he inernaional division. Therefore, he only way o guaranee he dominan posiion of China s foreign rade is o develop capial and echnology inensive expor indusries and upgrade expor srucure. Aman e al. (203) aemped o explore he relaionship beween exchange rae and economic growh in Pakisan for period They employ wo, hree sage leas square (2SLS and 3SLS) echniques and found ha exchange rae has a posiive associaion wih economic growh hrough he channel of expor promoion incenives, enlarging he volume of invesmen, enhancing FDI inflow and promoing impor subsiue indusry. He ricour and Ponce(203) sudied how firm-level expor performance is affeced by Real Exchange Rae (RER) volailiy and invesigae wheher his effec depends on exising financial consrains. 4

5 Their empirical analysis relies on expor daa for more han 00,000 Chinese exporers over he periods. They confirmed a radedeerring effec of RER volailiy. They also found ha firms decision o begin exporing and he expored value decrease for desinaions wih higher exchange rae volailiy and ha his effec is magnified for financially vulnerable firms. As expeced, financial developmen seems o dampen his negaive impac, especially on he inensive margin of expor. These resuls provide micro-founded evidence suggesing ha he exisence of welldeveloped financial markes allows firms o hedge exchange rae risk. The resuls also suppor a key role of financial consrains in deermining he macro impac of RER volailiy on real oucomes. Thus i appears from he preceding discussions ha relaionship beween exchange rae and economic growh in Bangladesh have no been addressed in Bangladesh. I would, herefore, no be unjusified o sae ha presen sudy is he firs of is kind in Bangladesh and can be used for guidelines for he similar sudies in years ahead. Mehodology and Daa Collecion To avoid he seasonal biases annual daa are used in his sudy. A long run span of daa is required for giving he ess for co-inegraion more power han merely increasing he daa frequency. This is because he coinegraion is a long run concep (Hakkio and Rush, 99). Secondary daa are mainly used in his sudy. The daa were colleced from he websie of Bangladesh Bank (BB), Bangladesh Bureau of Saisics (BBS), websie of World Bank (WB) and Asian Developmen Bank (ADB), Key Indicaors (KI), World Developmen Indicaors (WDI), Inernaional Financial Saisics (IFS). Finally, for he analysis he economeric sofware, namely Microfi 4. and Eviews 7 are used.i is seen from he lieraure of he ime series ha if he series are non-saionary or I() process, he regression resuls wih variables a level will be spurious (Granger and Newbold, 974; Phillips, 986). Thus, we sar wih examining he ime series properies of he series hrough he Augmened Dickey-Fuller (ADF) saionariy ess. The Economeric Model The model ha ries o esablish he relaionship beween ER and EG in Bangladesh can be expressed in he following basic bivariae model: Y = α + β X + ε () where, Y is real GDP and X is he ER and ε is error erm. Logarihmic ransformaion of he above equaion and inclusion of a rend variable would leave he basic equaion as follows: 5

6 LY = α 0 + α + β XE + ε (2) where, is he rend variable. The sandard Granger causaliy es (Granger, 988) seeks o deermine wheher pas values of a variable helps predic changes in anoher variable. In he conex of his analysis he Granger mehod involves he esimaion of he following equaions: q q (3) LY β + β i LY i + β 2i LX i + = 0 ε r r = ϕ 0 + ϕi i LX i + ϕ 2i LY i + ε LX 2 (4) where, LY and LX represen real GDP and ER, respecively, ε and ε 2 are uncorrelaed saionary random process, and subscrip denoes he ime period. Failing o rejec H 0 : β 2 = β 22 =... = β 2q = 0 implies ER does no Granger cause real income aciviies. On he oher hand, failing o rejec H0 : ϕ2 = ϕ22 =... = ϕ2r = 0 implies ha real GDP does no Granger cause ER. Empirical works based on ime series daa assume ha he underlying ime series is saionary. However, many sudies have shown ha majoriy of ime series variables are non-saionary or inegraed of order (Engle and Granger, 987). The ime series properies of he daa a hand are herefore sudied in he ouse. The above specificaion of he causaliy es assumes ha he ime series a hand are mean revering process. However, i is highly likely ha variables of his sudy are non-saionary. Formal ess will be carried ou o find he ime series properies of he variables. If he variables are I (), Engle and Granger (987) asser ha causaliy mus exis in, a leas, one direcion. The Granger causaliy es is hen augmened wih an error correcion erm (ECT) as shown below: q q LY = β 0 + β i LY i + β 2i LX i + αz + ε r r = ϕ 0 + ϕi i LX i + ϕ 2i LY i + λz + ε (5) LX 2 (6) where Z is he ECT obained from he long run co-inegraing relaionship beween real GDP and ER. The above error correcion model (ECM) implies ha possible sources of causaliy are wo: lagged dynamic regressors and lagged co-inegraing vecor. Accordingly, by equaion (5), q ER Granger causes real GDP, if he null of eiher β2i = 0 or α = 0 is 6

7 rejeced. On he oher hand, by equaion (6), real GDP Granger causes ER, if r λ is significan or i = ϕ 2i are joinly significan. Real GDP and ER granger causes each oher (i.e. presence of bidirecional causaliy), if causaliy exiss in boh direcions. Resuls and Discussion In his sudy annual daa on ER (X) and RGDP (Y) of Bangladesh are used. ER wih US dollar is exraced from he BB, BBS, WB, ADB, KI, WDI and IFS. RGDP is calculaed by GDP a curren marke price divided by consumer price index and boh are exraced from BB, WB, ADB, KI, WDI and IFS. Daa are used in original as well as in naural logarihms. To undersand he endency of economic aciviy, a primary analysis of he daa is done. The following Table shows he descripive saisics of boh variables in original and naural logarihmic form. From he Table i is seen ha average RGDP and ER are and respecively, whereas he RGDP ranges from a maximum o a minimum On he oher hand, he ER ranges from a maximum 8.82 o a minimum The average of LNRGDP and LNER are and respecively. Table : Descripive Saisics RGDP ER LNRGDP LNER Mean Median Maximum Minimum Sd. Dev Skewness Kurosis Jarque-Bera Probabiliy Observaions The maximum values of LNRGDP and LNER are and respecively. The following Table 2 represens he pair-wise correlaion beween RGDP and ER and Table 3 depics he pair-wise correlaion beween LNRGDP and LNER. The ables show here is a srong posiive relaion beween RGDP and ER a original value as well as naural logarihmic value and which is saisically significan a % level of significance. 7

8 RGDP Sig. (2-ailed) Table 2: Correlaion beween ER and RGDP RGDP ER 0.982** ER ** Correlaion is significan a he % level. LNRGDP Sig. (2-ailed) Table 3: Correlaion beween LNER and LNRGDP LNRGDP LNER 0.973** LNER ** Correlaion is significan a he % level. The esimaion procedure sars wih esing he ime series properies of RGDP, ER, LNRGDP and LNER. Table 4 (See in Appendices) shows he Correlogram RGDP a level. I is seen from he able ha Auocorrelaion Coefficien (AC) and Parial auocorrelaion coefficien (PAC) cross he boundary line. Moreover, in he case of AC a s o 6 h lag crosses boundary lines. A s lag PAC crosses boundary line. Apar from ha all he values of Box Pierce Ljung saisic (Q saisic) more han 25 and p value for hypohesis ha all auocorrelaion coefficien o his poin are zero and i shows significan a percen level. Therefore i indicaes non saionary rend and i follows sochasic rend. Table 5 (See in Appendices) depics he Correlogram of RGDP a s difference. The able shows ha AC and PAC do no cross he boundary line excep a lag 5. The p-value for all auocorrelaion coefficiens are more han 0.05 excep a lag5 and lag6 and i shows insignifican a percen level. Tha s why i indicaes saionary rend due o s difference and i follows deerminisic rend. Table 6 (See in Appendices) shows he Correlogram of LNRGDP a level. Table shows ha AC crosses boundary lines a s o 6 h lag. A s lag PAC crosses boundary line. Apar from s lag all he values of Box Pierce Ljung saisic (Q saisic) are more han 28 and p value for hypohesis for all auocorrelaion coefficiens o his poin are zero and i shows significan a percen level. Tha s why i indicaes non saionary rend and i follows sochasic rend. Table 7 (See in Appendices) shows he Correlogram of LNRGDP a s difference. Table shows ha AC crosses boundary lines a s, 4 h and 5 h lag. A s and 4 h lag PAC crosses boundary line. Apar from s o 5 h lag all he values of Box Pierce Ljung saisic (Q-saisic) are more han 22 and p value for hypohesis for mos of he auocorrelaion coefficiens o his poin are insignifican. Tha s why i indicaes saionary rend and i follows deerminisic rend. 8

9 Table 8 (See in Appendices) shows he Correlogram of ER a level. I is seen from he able ha AC crosses boundary lines a s o 6 h lag and PAC crosses boundary line a s lag. Apar from s lag all he values of Box Pierce Ljung saisic (Q saisic) are more han 50 and p value for hypohesis ha all auocorrelaion coefficiens o his poin are zero and i shows significan a percen level. Tha s why i indicaes non saionary rend and i follows sochasic rend. Table 9 (See in Appendices) shows he Correlogram of ER a s difference. I is observed from he able ha AC and PAC do no cross he boundary line. Apar from s lag all he values of Box Pierce Ljung saisic (Q saisic) are more han 5 and p value for hypohesis ha mos of AC is insignifican. Tha s why i indicaes saionary rend and i follows deerminisic rend. Table 0 (See in Appendices) shows he Correlogram of LNER. I is seen from he able ha AC crosses boundary lines a s o 6 h lag and PAC crosses boundary line a s lag. Apar from s lag all he values of Box Pierce Ljung saisic (Q saisic) are more han 50 and p value for hypohesis of all auocorrelaion coefficiens o his poin are zero and i shows significan a percen level. Tha s why i indicaes non saionary rend and i follows sochasic rend. Table (See in Appendices) shows he Correlogram of LNER a s difference. The able exhibis ha AC and PAC do no cross he boundary line. Apar from s lag all he values of Box Pierce Ljung saisic (Qsaisic) are more han 2 and p-values for hypohesis of he mos of auocorrelaion coefficiens are insignifican. Tha s why i indicaes saionary rend and i follows deerminisic rend. In he 2 nd sep of esing he ime series properies of he daa he uni roo es is done. The following Table 2 shows he resuls of uni roo es by using Augmened Dickey-Fuller (ADF). For ADF, boh wih consan and consan and rend, one is unable o rejec a null hypohesis a level bu is able o rejec he null hypohesis when s differenced series are used. Table 2: Uni roo es of he variables Variables Augmened Dickey-Fuller Tes Criical Value (ADF) Tes Pro Uni cess Saisics P-values A % A 5% A 0% Roo Tes Equaion: Inercep LGDP Yes I() GDP * No I(0) LLNGDP Yes I() LNGDP * No I(0) LER Yes I() ER *** No I(0) LLNER Yes I()

10 LNER *** No I(0) Tes Equaion: Trend and Inercep LGDP Yes I() GDP *** No I(0) LLNGDP Yes I() LNGDP *** No I(0) LER Yes I() ER *** No I(0) LLNER Yes I() LNER *** No I(0) Source BB, WB, ADB, KI, WDI, IFS. Noe: L sands for level, Δ denoes he firs difference of he variable. The null hypohesis saes ha he variable has a uni roo. P-values are used o decide he uni roos a he %, 5% and 0% significance level. The criical values and deails of he ess are presened in Dickey and Fuller (979, 98). The AIC deermines he lag lengh (P) in he ADF ess (see Sock and Wason 2007:56 for deails). Tes equaion: rend and inercep. *,**, and *** denoe rejecion of null a 0%, 5%, and % level of significance. I is observed from he Table 2 ha all he examined series are inegraed of order one, I (). These resuls are consisen wih he noion ha mos of he macroeconomic variables are non-saionary a level, bu become saionary afer firs differencing (Nelson and Plosser, 982). Once i is esablished ha variables are I (), he nex sep is o es for exisence of any co-inegraion relaionship beween RGDP and ER. To es he coinegraion, he Johansen (99) LR es is applied and resuls are showed in Table 3. The appropriae VAR lag lengh is seleced using BIC. Table 3: Coinegraion es beween RGDP and ER (Johansen Coinegraion Tes) [VAR lag k = 2, [Y,X] Null Eigen values Trace Tes Max Eigen value Tes λ race p value λ max p value r r Since race saisic is 4.7 and p-value is which means ha he saisics is significan a 0% level and we can rejec he null hypohesis of no co-inegraion vecor and accep he alernaive of one co-inegraing vecor. Again race saisic is 3.06 and p-value is significan a 0% level, so we can rejec he null hypohesis of one co-inegraing vecor and accep he alernaive hypohesis of more han one co-inegraing vecor. Same hing is no happened when λ-max es is used. Therefore we have o es coinegraion beween LNRGDP and LNER. 20

11 Table 4: Co-inegraion Tes beween LNRGDP and LNER (Johansen Co-inegraion Tes) VAR lag k = 2, [Y,X] Null Eigen values Trace Tes Max Eigen value Tes λ race p value λ max p value r r Since race saisic is and p-value is which means ha he race saisic is significan and we can rejec he null hypohesis of no co-inegraion vecor and accep he alernaive of one co-inegraing vecor. Again race saisic is and p-value is which means ha he saisic is significan a % level and we can rejec he null hypohesis of one co-inegraing vecor and accep he alernaive hypohesis of more han one co-inegraing vecor. Same hing is happened when λ-max es is used. Boh p-values for r=0 and r= are significan. The Eigen value ess based on sochasic marix indicae exisence of he co-inegraion relaionship beween RGDP and ER. Therefore, he Granger causaliy ess are o be modeled using ECM as explained in equaions (5) and (6). Table 5: Granger causaliy es beween RGDP and ER Pairwise Granger Causaliy Tess Lags: 2 Null Hypohesis: Obs F-Saisic Probabiliy GDP does no Granger Cause ER ER does no Granger Cause GDP Table 5 summarizes he Granger Causaliy resuls beween RGDP and ER of Bangladesh from 973 o 203. F-saisic and probabiliy values are consruced under he null hypohesis of no causaliy. Here we canno rejec he null hypohesis because p-values are more han 0.0, i.e., RGDP does no cause ER and ER does no cause RGDP. Table 6: Granger causaliy es beween LNRGDP and LNExchange Rae Pairwise Granger Causaliy Tess Lags: 2 Null Hypohesis: Obs F-Saisic Probabiliy LNGDP does no Granger Cause LNER LNER does no Granger Cause LNGDP Table 6 summarizes he Granger Causaliy resuls beween LNRGDP and LNER of Bangladesh from 973 o 203. F-saisic and probabiliy values are consruced under he null hypohesis of no causaliy. Here we can rejec he null hypohesis a % level of significance. I is eviden from he resuls ha here is a bi-direcional causal relaionship 2

12 beween he variables i.e. he boh way causaliy runs hrough LNRGDP o LNER and LNER o LNRGDP. Conclusion The main objecive of his paper is o examine empirically he relaionship beween EG and ER in Bangladesh wih he laes ime series economeric mehod. Time series economeric ools are used o examine he relaionship beween he variables. We use he ADF es, Granger Causaliy Tes and Johansen Co-inegraion models by aking care of sochasic properies of he variables. From he resuls of uni roo es i is seen ha boh he variables are inegraed of order in boh original values and logarihmic values of he variables. By using naural logarihmic form of he variables we find ha here is a long-erm equilibrium relaionship beween EG and ER. The Eigen value ess based on sochasic marix indicae he exisence of he co-inegraion relaionship beween EG and ER. The resul of Granger s Causaliy es denoes ha here is bi-direcional causaliy runs hrough EG o ER and hrough ER o EG. References: Alba, J.D., Park, D., & Wang, P. (200). The Impac of Exchange Rae on FDI and he Inerdependence of FDI over Time. The Singapore Economic Review, 55(4), Aghion,P., Bacchea, P., Rancière, R., & Rogoff, K. (2009). Exchange Rae Volailiy and Produciviy Growh: The Role of Financial Developmen. Journal of Moneary Economics, 56, Aman, Q., Ullah, I., Khan, M.I., & Khan, S.(203). Linkages beween exchange rae and economic growh in Pakisan (an economeric approach): European Journal of Law and Economics. DOI: 0.007/s Ba, S., & Shen, S. (200). Research on China s Expor Srucure o he US: Analysis Based on he US Economic Growh and Exchange Rae. Fronier of Economics in China, 5(3), Cheung, Y., & Lai, K.S. (998). Economic growh and saionary of real exchange raes: Evidence from some fas-growing Asian Counries: Pacific Basin Finance Journal, 6, Engle, R.F., & Granger, C.W.J. (987). Co-Inegraion and Error Correcion: Represenaion, Esimaion, and Tesing. Economerica, 55(2), Gala, P. (2008). Real exchange rae levels and economic developmen: heoreical analysis and economeric evidence. Cambridge Journal of Economics, 32, Granger, C.W.J., & Newbold, P. (974). Spurious Regressions in Economerics. Journal of Economerics, 2,

13 Granger, C.W.J. (988). Some Recen Developmens in he Concep of Causaliy: Journal of Economerics, 39, Hakkio, C.S., & Rush, M. (99). Co-inegraion: How Shor Is he Long- Run? Journal of Inernaional Money and Finance, 0, Iō, T. (2005). The Exchange Rae In The Japanese Economy: The Pas, Puzzles, And Prospecs. The Japanese Economic Review, 56(), -38. Johansen, S., (988). Saisical Analysis of Co-inegraion Vecors. Journal of Economic Dynamics and Conrol, 2(2-3), Mundaca, B.G., & Srand, J. (2005). A Risk Allocaion Approach To Opimal Exchange Rae Policy: Oxford Economic Papers, 57, Nelson, C.R. and Plosser, C.L. (982). Trends and random walks in macroeconomic ime series: Some evidence and implicaions. Journal of Moneary Economics, 0, Ok, S., Kakinaka, M., & Miyamoo, H. (200). Real Shock or Nominal Shock? Exchange Rae Movemens in Cambodia and Lao PDR: The Singapore Economic Review, 55(4), Phillips, P.C. (986). Undersanding Spurious Regressions in Economerics: Journal of Economerics, 33, Rahman, M.H., & Barua, S. (2006). Recen Experiences in he Foreign Exchange and Money Markes Policy: Noe 703, Bangladesh Bank Quarerly, III (4). Razin, O., & Collins, S.M. (997). Analysis of real Exchange rae and Economic growh in Leas Developing Counry. Journal of Applied Saisics and Economics, 4(), Appendices Table 4: Correlogram of Real GDP a Level Auocorrelaion Parial Correlaion AC PAC Q-Sa Prob. ******. ****** ***** **** *** ** ** ** * * * * *

14 Table 5: Correlogram of Real GDP a s Difference Auocorrelaion Parial Correlaion AC PAC Q-Sa Prob. *.. * *.. * **..** ***. *** *.. * * * ** * * Table 6: Correlogram of LN Real GDP a Level Auocorrelaion Parial Correlaion AC PAC Q-Sa Prob. ******. ****** ***** **** ***. * *** ** * * * * Table 7 Correlogram, LN Real GDP, s Difference Auocorrelaion Parial Correlaion AC PAC Q-Sa Prob ***. *** * ***. *** ****. ** * *

15 . *.. * * * * Table 8: Correlogram of Exchange Rae(level) Auocorrelaion Parial Correlaion AC PAC Q-Sa Prob. *******. ******* *****. * **** *** *** ** ** * * * * Table 9: Correlogram of Exchange Rae( s Difference) Auocorrelaion Parial Correlaion AC PAC Q-Sa Prob. *.. * ***. *** * * *.. * * * * * *.. * * *.. * *.. * *.. *

16 Table 0: Correlogram of LN Exchange Rae(level) Auocorrelaion Parial Correlaion AC PAC Q-Sa Prob. *******. ******* *****. * **** ****. * ***. * ** ** *.. * * Table : Correlogram of LN Exchange Rae( s Difference) Auocorrelaion Parial Correlaion AC PAC Q-Sa Prob. *.. * **..** * *.. * *.. * **.. * *.. * *.. * * *.. * *.. * *