TRADE BALANCE IN THE PRESENCE OF STRUCTURAL BREAKS

Transcription

1 PROSIDING PERKEM IV, JILID 1 (2009) ISSN: X TRADE BALANCE IN THE PRESENCE OF STRUCTURAL BREAKS ABU HASSAN SHAARI MD NOR, SITI HAMIZAH MOHD, TAMAT SARMIDI & UGUR ERGUN ABSTRACT The main objecive of his paper is o deermine he long run relaionship beween rade balance and erm of rade in Malaysia in he presence of srucural breaks. The US and Japan are seleced for he sample rading counries. The resuls from he uni roo es and Johansen coinegraion es show ha here exis is long run relaionship beween rade balance and erm of rade when he srucural breaks poins are aken ino accoun. In addiion, using he auoregressive disribued lag (ARDL) coinegraion framework, he resuls confirm ha a sable, long run relaionship exiss beween rade balance and erm of rade beween Malaysia and her major rading parner, Japan. However, we failed o find any long run relaionship beween he variables for he US case. Keywords: rade balance; srucural breaks; ne impor; ne expor. 1. Inroducion The sudy on he relaionship beween rade balance and exchange rae movemens is one of umos imporan especially for a counry like Malaysia whose economy is heavily depended on inernaional rade. I is because if currency depreciaion would improve he rade balance in he shor run or long run, i may be desirable o accommodae o exchange rae depreciaion o simulae ne expors and expand he economy. On he oher hand if currency depreciaion deerioraes he rade balance, hen sabiliy in exchange rae is suggesed o be vial objecive in a moneary and fiscal policy formulaion. However, presen empirical findings on he relaionship beween rade balance and exchange rae movemens are sill inconclusive (Bahmani-Oskooee and Niroomand, 1998, Bahmani-Oskooee and Raha, 2004 and Narayan, 2004), and for Malaysia case, i is very limied. This has invied o ineresing and on-going debae in he inernaional economic lieraure. This issue of responsiveness of rade balance o exchange rae flucuaions has been widely sudied by esing Marshall-Lerner condiion (Arize, 1990). In lieraure here are wo approaches o sudy he validiy of his condiion. Firs is radiional approach. This mehod requires he esimaion of boh expor and impor demand model (Arize, 1990; Bahmani- Oskooee and Niroomand, 1998). According o radiional approach, Marshall-Lerner condiion holds, i.e. he depreciaion of domesic currency improves rade balance, only if he sum of he absolue values of impors and expor demand price elasiciy exceed uniy (Gómez and Álvarez-Ude, 2006). Anoher approach ha has been widely used o assess Marshall-Lerner condiion is direc es as suggesed by Haynes and Sone (1982). This approach ries o creae a direc relaionship beween erms of rade and rade balance under he unconsrained disribued lag model. ML condiion holds only if he sum of he esimaed coefficiens of erms of rade is non-negaive i.e. escalaion in erms of rade has enhanced rade balance (Arize and Ndubizu, 1992; Arize, 1994). The aim of his sudy is o deermine wheher a long- and shor-run relaionship exiss beween he rade balance and he erms of rade. Persidangan Kebangsaan Ekonomi Malaysia (PERKEM IV) Memacu Pembangunan Ekonomi Dalam Keidakenuan Persekiaran Global Kuanan, Pahang, 2-4 Jun 2009

2 Trade Balance and Srucural Breaks Theoreical consideraion and model specificaion A flucuaion of currency could improve, worsen or has no affec o rade balance depending on he size of he demand price elasiciies for he counry s expor and impors. I is because depreciaion in local currency makes foreign goods more expensive and reduces impor spending. A he same ime i will reduce price of domesic goods and may increase demand for expor goods and hence raises expor revenue. Theoreically, his siuaion can be explained using Marshall-Lerner condiion. This condiion sugges ha a. If he sum of expor and impor price elasiciies greaer han uniy, hen devaluaion (appreciaion) will improve (worsen) he rade balance, b. If he sum of expor and impor price elasiciies is less han uniy, hen devaluaion (appreciaion) will worsen (improve) he rade balance, c. If he sum of expor and impor price elasiciies is equal o uniy, hen devaluaion (appreciaion) will no affec he rade balance. The long run specificaion model are given by: where TB is rade balance (raio of expor o impors), TOT is erm of rade (raio of impor price o expor price), is error erm. If Marshall-Lerner condiion holds, increase in he price of expors relaive o he price of impors is expeced o decrease he TOT and he TB. I is because impors are encouraged and expors are discouraged (β < 0).This condiion also holds for he reverse (β > 0). This condiion says ha, for a currency devaluaion o have a posiive impac in rade balance, he sum of price elasiciy of expor and impors (in absolue value) mus be greaer han 1. The principle is named for economiss Alfred Marshall and Abba Lerner. As a devaluaion of he exchange rae means a reducion on price of expors, quaniy demanded for hese will increase. A he same ime, price of impors will rise and heir quaniy demanded will diminish. The ne effec on he rade balance will depend on price elasiciies. If goods expored are elasic o price, heir quaniy demanded will increase proporionaely more han he decrease in price, and oal expor revenue will increase. Similarly, if goods impored are elasic, oal impor expendiure will decrease. Boh will improve he rade balance. 3. Daa and Mehodology Empirical analysis was carried ou using monhly daa from 1995:01 o 2008:09. The rade balance (TB) is represened by he raio of expors o impors and he erm of rade (TOT) is he raio of he impor prices o he expor prices, and all daa measured in Ringgi Malaysia. All of hese daa were obained from he Deparmen of Saisics, Malaysia. Firs, saionary ess which are Augmened Dickey Fuller (ADF) (Dickey & Fuller, 1981), Phillips-Perron (PP) (Phillips & Perron, 1988) and uni roo es wih srucural breaks are applied o deermine he order of inegraion. If here is a shif in he ime series, i should be aken ino accoun in esing for a uni roo because he ADF es may be disored if he shif is simply ignored. Saikkonen and Lukepohl (2002) and Lanne e al. (2002) have proposed he following model: A shif funcion, which is f ( ), is added in he equaion; y f ( ) e 0 1 (2) (1)

3 384 Abu Hassan Shaari, Sii Hamizah, Tama Sarmidi & Ugur Ergun where and are unknown parameers or parameer vecors and he errors e are generaed by an AR(p) process.. And shif funcion is defined as; f ( 1) d1 0, 1, T B T T B The shif funcion, f (θ)γ is a simple shif dummy variable wih shif daa T B. Saikkonen and Lukepohl (2002) and Lanne e al. (2002) have proposed uni roo ess based on esimaing he deerminisic erm by he generalised leas squares (GLS) procedure and subracing i from he original series. Thereafer an ADF-ype es is performed on he adjused series. Nex, Johansen coinegraion wih srucural break (Johansen e al. 2000) is performed in order o examine he impac any breaks on he coinegraion relaionship. Johansen (1998, 1991, 1992, 1994, 1995) has proposed likelihood ess if dependen variable is normally disribued. These ess are known as race ess because of he special form of he es saisic. The disribuions of he es saisics under heir respecive null hypoheses depend on he deerminisic erms. Considering he relaively small sample size, we also esimae he model using he auoregressive disribued lag (ARDL) coinegraion es. We began by esing for he presence of a long-run relaionship using he F-es o deermine he join significance of lagged levels of he variables involved. The F-saisics hen compared wih he bound criical value documened in Pesaran e al. (2001) and Narayan (2005). 4. Empirical Resuls Saionariy (uni roo) wih srucural break es resuls which is repored in Table 1 indicae ha all series are nonsaionary in log levels bu saionary in log firs differences. Table 1 Uni Roo Tes Resul Levels Firs Differences ADF UR wih Sr. Break ADF UR wih Sr. Break TB-JP * TB-US * * TOT-JP * * TOT-US * * * denoes rejecion of he null hypohesis a leas 0.05 level All variables are nonsaionary in levels and inegraed wih same order I(1) according o he resuls from ADF and uni roo wih srucural break ess. Table 2 JJ Tes Resul Variables Lag Inervals wihou break wih break TB-JP, TOT-JP * 38.76* TB-US, TOT-US * * Indicae 1 coinegraing equaion a he 0.01 level

4 Trade Balance and Srucural Breaks 385 The series for variable TOT-US has significan srucural break and need o be aken ino accoun in he analyses. To ignore i can cause spurious resul. As indicaed in Table 2, here are significan linkage beween TB and TOT for Japan and US in he coinegraion analyses when we include srucural break. However, when we ignore he srucural break in our analyses he relaionship disappear for US. ECM es resul repored in Table 3 reveal ha no evidence for dynamic linkages is deeced beween TB and TOT for each sample counries. Table 3 Esimaion of Error Correcion Model c 1 y 1 y F-sa. 1 s 11(1) 11(2) 12(1) 12(2 Wih Japan ) Wih Japan (SB) Wih US Wih US (SB) The bound ess resuls are repored in Table 4. For Malaysia-Japan case, we find ha here exiss long run relaionship beween erms of rade and rade balance of he lag order one, wo and hree under he unresriced rend model. However, we failed o esablished long run coinegraion for he Malaysia-US case. The compued F-saisic was also compared wih he criical values ha accoun for small sample sizes provided by Narayan (2005). Model/ Table 4 The Bounds Tes Resuls Model wihou rend Lag C.V. i C.V. ii MS-US (5%) (5%) MS-JP (10%) (10%) Model/ Model wih rend Lag C.V. i C.V. ii US (5%) (5%) Japan a,b a,b a,b (10%) (10%) Noe: MS-US and MS-JP represens Malaysia-US and Malaysia-Japan model, respecively. The superscrip a,b denoes ha he saisic lies above he upper bound, a leas a 10% significance level based on (C.V.) i and (C.V. ii ), respecively. Criical values, (C.V.) i are based on Pesaran e al. (2001) and (C.V. ii ) are sourced from and Narayan (2005). Using he ARDL approach, we obained he long-run parameer esimaes for erms of rade. As observed in Table 5, he resul shows ha he coefficien carry he expeced signs (so called Marshall-Lerner condiion holds) wih reasonable magniudes bu saisically insignifican a usual significance level. The posiive signs of he erms of rade parameer implies ha he depreciaion of Malaysian ringgi (ha is, an increase in erms of rade) leads o an improvemen in he bilaeral rade wih is major rading parners, in his case, Japan. The insignifican of he coefficien implies ha he impac is relaively weak or small. This

5 386 Abu Hassan Shaari, Sii Hamizah, Tama Sarmidi & Ugur Ergun finding indicaes ha currency depreciaion or devaluaion in Malaysia will affec he bilaeral rade balance. However, he magniude is quie small. Model Lag Coefficien Table 5 Long Run Esimaes Diagnosic Tes TOT Inercep A: LM B: RESET C: JB D: ARCH MS-JP AIC (1,0) (0.5230) ( ) [0.7440] [0.3290] [0.9200] [0.5400] Noe: MS-US and MS-JP represens Malaysia-US and Malaysia-Japan model, respecively. Figures repored in bracke ( ) and [ ] are -raio and p-value, respecively. The es saisics are: LM = Lagrange muliplier es for he auocorrelaion; RESET = Ramsey s es for funcional form misspecificaion, JB = Jarque-Bera es for normaliy of residuals; and ARCH = Engle s auoregressive condiional heeroscedasiciy es. Table 6 presens he error correcion esimaions for he Marshall-Lerner condiion model based on he ARDL approach. The coefficiens of he lagged difference are presened in he second column and he -raio indicaing he significance level of he coefficiens are shown in bracke. The error correcion erm represens he poenial effecs of deparures from he long-run equilibria. The size and he significance of he error correcion erm in he equaion show he endency of he variable o resore equilibrium in he rade balance. Based on he resuls, he lagged error-correcion erm for Malaysia-Japan ( ) carries is expeced negaive sign and is highly significan. The coefficiens reveal ha he speed of adjusmen is raher high for he previous monh s discrepancy beween he acual and equilibrium value of TB is correced each monh. In oher words, his means ha he adjusmen process o an exogenous shock is relaively high. Variable Table 6 Shor Run Esimaes MS-JP AIC (1, 0) TB -1 - TB -2 - TOT (0.5220) TOT -1 - TOT -2 - Inercep ( ) ECM ( ) Noe: MS-US and MS-JP represens Malaysia-US and Malaysia-Japan model, respecively. Figures repored in bracke ( ) are -raio. Δ means he firs difference and ECM -1 is he error correcion erm. Nex, we assess he sabiliy of he long-run relaionship beween rade balance and erms of rade. We relied upon he CUSUM and CUSUM-squared ess proposed by Brown e al. (1975) o es for consancy of long-run parameers. If he plo of he CUSUM saisics says wihin he 5% significance level, hen he esimaes are said o be sable. The same applies o he CUSUM-squared saisics, which are based on he squared recursive residuals. As can be seen in Figure 1, he plo of he CUSUM and CUSUM-squared saisics sayed wihin he criical bounds indicaing he sabiliy in he model (for Japan case). These findings jusify he sabiliy of he model and we can conclude ha he relaionship beween rade balance and erms of rade are sable for he Malaysia-Japan model.

6 Trade Balance and Srucural Breaks 387 Figure 1 Plo of CUSUM and CUSUMSQ for Japan Plo of Cumulaive Sum of Recursive Residuals (CUSUM) M7 2006M5 2007M3 2008M1 2008M9 The sraigh lines represen criical bounds a 5% significance level Plo of Cumulaive Sum of Squares of Recursive Residuals (CUSUMSQ) M7 2006M5 2007M3 2008M1 2008M9 The sraigh lines represen criical bounds a 5% significance level 5. Conclusion In his sudy, we examined he long run relaionship beween rade balance and erm of rade for Malaysia and her major rading parner, he US and Japan. The his end, we use he mehod ha ake ino accoun he srucural break poin, he Johansen coinegraion wih srucural break es (Johansen e al. 2000). Our resuls failed o rejec he null hypohesis ha here is no long run relaionship beween rade balance and erm of rade. The resul holds for all he cases under invesigaion. This resul indicae ha he exchange rae movemen plays an imporan role and have significan impac on he value of ne impor and expor in Malaysia for he case where he US and Japan are he rading parner. Addiionally, employing bounds coinegraion esing approach as proposed by Pesaran e al. (2001, ARDL), we found ha here exis a posiive and significan long-run saisical equilibrium beween rade balance and erms of rade for Malaysia-Japan case bu no for Malaysia-US. References Arize, A. C An economeric invesigaion of expor behaviour in seven Asian developing counries. Applied Economics, 22,

7 388 Abu Hassan Shaari, Sii Hamizah, Tama Sarmidi & Ugur Ergun Arize, A. C. and Ndubizu, G. (1992). Error-Correcion Represenaion and Impor Demand Funcion wih Implicaions for Inernaional Finance & Accouning. Review of Quaniaive Finance and Accouning, 2 (Dec.), Bahmani-Oskooee, M. and Niroomand, F Long-run price elasiciies and he Marshall-Lerner condiion revisied, Economics Leers, 61, Bahmani-Oskooee, M. and Raha, A., 2004, The J-Curve: A Lieraure Review, Applied Economics, 36, Bahmani-Oskooee, Nohsen and Massomeh Hajilee The J-Curve a indusry level: Evidence from Swedeb- US rade. Economic Sysems 33(1), Brown, R. L. J. Durbin and J. M. Evans (1975). Techniques for esing he consancy of regression relaionships over ime (wih Discussion). J. Roy. Saisical Sociey, Series B. 37, Dickey, D. A. & Fuller, W. A Likelihood raio saisics for auoregressive ime series wih a uni roo. Economerica, 49, (July). Engle, R. and Granger C.W.J Coinegraion and Error-Correcion: Represenaion, Esimaion and Tesing, Economerica, 55 (2), Gomez, David Maesanz and Guadalupe Fugarolas Alvarez-Ude Exchange rae policy and rade balance. A coinegraion analysis of he Argenine experience since MPRA paper 151. Halicioglu, F The bilaeral J-curve: Turkey versus her 13 rading parners. Journal of Asian Economics 19(3), Haynes, Sephan E. and Joe A. Sone Impac of he Terms of Trade on he U.S. Trade Balance: A Reexaminaion. The Review of Economics and Saisics 64(4), Johansen, S Saisical analysis of coinegraion vecors, Journal of Economic Dynamics and Conrol 12, Johansen, S Esimaion and hypohesis esing of coinegraion vecors in Gaussian vecor auoregressive models, Economerica 59, Johansen, S Deerminaion of coinegraion rank in he presence of a linear rend, Oxford Bulein of Economics and Saisics 54, Johansen, S The role of he consan and linear erms in coinegraion analysis of nonsaionary ime series, Economeric Reviews 13, Johansen, S Likelihood-based Inference in Coinegraed Vecor Auoregressive Models, Oxford Universiy Press, Oxford. Johansen, S., Mosconi, R. and Nielsen, B Coinegraion analysis in he presence of srucural breaks in he deerminisic rend, Economerics Journal 3, Lanne, M., P.Saikkonen and H. Lükepohl, Comparison of uni roo ess for ime series wih level shifs. Journal of Time Series Analysis, 23, Paresh Kumar Narayan New Zealand's rade balance: evidence of he J-curve and granger causaliy, Applied Economics Leers, Taylor and Francis Journals, 11(6), Narayan, P.K., Smyh, R., Trade liberalisaion and economic growh in Fiji. An empirical assessmen using he ARDL approach. J. Asia Pacific Economy 10(1), Pesaran, M.H., Shin, Y., Smih, R.J., Bounds esing approaches o he analysis of level relaionships. J. App. Economerics. 16, Pesaran, M.H., Shin, Y., An auoregressive disribued lag modelling approach o coinegraion analysis, in: Srom, S., Holly, A., Diamonds., (Eds.), The Ragnar Frisch Cenennial Symposium. Cambridge: Cambridge Universiy Press. Phillips, P. C. B., & Perron, P Tesing for a uni roo in ime series regression. Biomerika, 75, Saikkonen, P. and Lükepohl, H Tesing for a uni roo in a ime series wih a level shif a unknown ime. Economeric Theory, 18,