Testing for Long-Run Relation between Economic Growth and Export Earnings of Cocoa in Ghana using Co-Integration Techniques*

Transcription

1 Tesing for Long-Run Relaion beween Economic Growh and Expor Earnings of Cocoa in Ghana using Co-Inegraion Techniques* S. Twumasi-Ankrah and E. N. Wiah Twumasi-Ankrah, S. and Wiah, E. N., (216), Tesing for Long-Run Relaion beween Economic Growh and Expor Earnings of Cocoa in Ghana using Co-Inegraion Techniques, Ghana Mining Journal, pp Absrac In his sudy, we explore he causal influence of expor earnings of cocoa owards economic growh (i.e., gross domesic produc (GDP)). The augmened Dickey Fuller (ADF) and he Phillips Perron (PP) uni roo ess indicae ha he wo series are inegraed of order one, I (1). The resuls of he race and he maxi-eigen coinegraion es based on Johansen s procedure indicae he exisence of a coinegraion beween expor earnings of cocoa and GDP. Thus, he wo variables of he sudy have a long-run equilibrium relaionship. The vecor error correcion model of order wo, VECM (2), was considered as he bes model afer evaluaing oher compeing models. I is observed ha, in he long-run, previous year expor earnings of cocoa is posiively relaed o economic growh. In he shor run, he resuls revealed ha he previous GDP has posiive effec on curren GDP; and higher expor earnings of cocoa have posiive effec on GDP. Feedback causaliy is observed beween economic growh (GDP) and expor earnings of cocoa. This suggess a bi-direcional causaliy from expor earnings o economic growh (GDP). Keywords: Coinegraion, VECM, Johansen, ADF, Causaliy. 1 Inroducion Counries are concerned wih improving he economic prosperiy of heir people. Therefore, i is known ha increasing gross domesic produc (GDP) will be able o achieve he goal of economic prosperiy. One of he ways of increasing GDP is o promoe expors, since i is a deerminan of GDP. Expor is ofen seen as an imporan caalys for economic growh in developing economies. According o Ramos (21), expor growh is ofen considered o have a direc influence on producion and employmen growh of an economy. In Ghana, cocoa is one of he mos imporan crops in he economy. Ghana is repored o be he second larges producer in he world, accouning for abou 21% of he oal producion (ICCO, 26). Cocoa expors from Ghana are made up of five producs, which are classified ino: raw, semi-processed and processed producs. Ghana expors cocoa beans, cocoa buer, cocoa powder, cocoa pase and cocoa husks (shells), wih expor of he laer commencing in he year Cocoa remains he counry s mos imporan agriculural expor crop, accouning for approximaely 23% of oal expor earnings (ICCO, 212) and 11% of agriculural gross domesic produc (GDP). In spie of Ghana s heavy dependence on he proceeds of cocoa expor, no deail research has been conduced o review he role of cocoa expor on economic growh in Ghana. No deailed sudy could be found on he causal relaionship beween cocoa expor earnings and economic growh in Ghana. However, here are deailed empirical sudies on he relaionship beween economic growh and expors. There are conradicing evidences abou he dynamic relaionship beween expors and economic growh. This has pu many developing counries in dilemma wheher o open up heir economies o promoe inernaional rade. Amavilah (23) examined he role of expors in Nambia s economic growh. I was revealed ha here was a general imporance of expors, bu he sudy finds no apparen sign of acceleraed growh due o expors. Thurayia (24) sudied he relaionship beween expors and economic growh in Saudi Arabia and Sudan. The coinegraion and error correcion models in he sudy showed ha expors had a posiive effec on GDP in he shor- and long- run. Dawson (25) sudied he conribuion of agriculural expors o economic growh in less developed counries. The sudy concluded ha he *Manuscrip received February 16, 216 Revised version acceped June 6, 216 hp://dx.doi.org/1.4314/gmj.v16i GMJ Vol. 16, No. 1, June, 216

2 less developed counries provided evidence o suppor he heory of expor led growh. Aurangzeb (26) sudied he relaionship beween economic growh and expors in Pakisan. The sudy indicaed ha expor oriened and ouward looking approach was needed for high raes of economic growh in Pakisan. Tang (26) saed ha here is no shor- and longrun causaliy beween expor expansion and economic growh in China. Jordaan (27) examined he causaliy beween expors and GDP of Namibia. The sudy revealed ha expors Granger-cause GDP, and suggesed ha he expor-led growh sraegy hrough various incenives has a posiive influence on growh. Rangasamy (28) examined he expors and economic growh relaionship for Souh Africa. The sudy provided an evidence of unidirecional Granger causaliy runs from expors o economic growh. Sanjuan-Lopez and Dawson (21) assessed he conribuion of agriculural expors o economic growh in developing counries. The resuls of he sudy indicaed ha here exised long run relaionship and he agriculure expor. From he above lieraures, i is obvious ha a number of sudies suppor he expor-led economic growh while ohers do no. As indicaed earlier, here are no sudies on expor earnings of cocoa and economic growh in Ghana. Thus, his paper is an aemp o invesigae he expor earnings of cocoa-economic growh nexus for Ghana. This sudy shall provide useful informaion which will be helpful o policy makers. Thus, he purpose of his sudy is o examine he causal relaionship, if any, beween he expor earnings of cocoa and economic growh (GDP) using a vecor error correcion model (VECM), and o ascerain he economic implicaions of such a causal relaionship. 2 Resources and Mehods Used 2.1 Daa Source The expor earnings (US$) daa were obained from he Ghana Cocoa Board and GDP (per capia) from he World Bank Developmen Indicaor websie. All daa span from 198 o Uni Roo Tes In order o find an appropriae model for a series, we mus check he series for saionariy. A nonsaionary ime series will have a ime-varying mean or a ime-varying variance or boh. Thus, i is always proper o ransform a non-saionary ime series o a saionary series before doing any meaningful analysis. The uni roo es is a formal way of esing for saionariy. Among he various mehods of uni roo es, he es developed by Dickey and Fuller, known as he augmened Dickey-Fuller (ADF) es, is commonly used. The hypohesis of he es is: H : series has a uni roo or no saionary H 1 : series does no have a uni roo or saionary The ADF es consiss of esimaing he following regression model: m Y Y Y i 1 i 1 1 where is a pure whie noise error erm and he ADF es follows an asympoic disribuion. 2.3 Error Correcion Model The error correcion model is used when he ime series are no saionary and are coinegraed. Tess for Coinegraion Two I (1) ime series y 1, and y 2, are said o be coinegraed if here exiss a linear relaionship of he form Z 1y1, 2y such ha Z 2, is I(). If we define he vecors y, 2 1 1, y y 2 2, so ha he coinegraing relaionship is wrien Z y, hen is called he coinegraing vecor. The coinegraion relaionship is ofen inerpreed as being a long run or equilibrium relaionship beween he variables. The number of linearly independen coinegraing vecors is called he coinegraing rank. The wo common ess o deermine he coinegraing rank are he race and he maximum eigens ess. The hypohesis of he es is: H : he number of coinegraing vecors is r, H 1 : he number of coinegraing vecors is (r+1) 9 GMJ Vol. 16, No. 1, June, 216

3 The wo saisics are: Trace Max ( r) T g ir1 ln(1 ˆ ) ( r, r 1) T ln(1 ˆ i r1 ) (3) where ˆi is he esimaed for he i h ordered eigen. Vecor Error Correcion Model The appropriae model for coinegraed ime series is called a Vecor Error Correcion Model (VECM) and is a rearranged resriced form of a VAR. In general, he VECM is of he form: y y y y p1 p1 (4) A VECM hus consiss of a mixure of variables in levels and firs difference form. If we applied he univariae modeling sraegy of aking firs differences of any I(1) ime series, and hence fiing a VAR in firs differences, he resuling model would be mis-specified because of he omied error correcion erm. Conversely we canno use a VAR in levels o model coinegraed ime series because he resuling inference in he presence of he non-saionariy would no be valid. In he presence of coinegraion, a VECM is required. 2.4 Model Selecion Crieria The informaion crieria used in his sudy are he following: Akaike informaion crierion (AIC): AIC = -2lnL( ˆ q k ) + 2k ( 5) Baysian informaion crierion (SIC) : SIC = -2ln L + k ln n ( ) ( 6) where L ˆk is he likelihood of he fied model, k = p + 1 (which is he model size), p = number of parameers and n = number of observaion. 3 Resuls and Discussion A correlaion analysis on expor earnings of cocoa and GDP is performed, he Pearson s correlaion coefficien beween he wo variables is calculaed over he sample period, and is significance is esed by he p-. The of Pearson s correlaion coefficien (r) is.86 wih p- of.. This shows ha expor earnings of cocoa and GDP are posiively relaed in Ghana and his relaionship is saisically significan. Correlaion, however, does no say anyhing abou long-run relaionship, and ha does no help us o sele he debae concerning he long-run relaionship beween expor earnings of cocoa and economic growh (GDP). Saionariy Tes The resuls of he ADF and PP uni roo ess are repored in Table 1. These wo uni roo ess are performed on boh he level and firs differences of GDP and expor earnings. The variables GDP and Expor earnings of cocoa are I(1) processes according o ADF and PP. However, i is eviden ha hese variables became saionary afer firs difference. Table 1 Tes of Uni Roo Tes Hypohesis wih Inercep Variables ADF Tes PP Tes Conclusions Tes Saisic Tes Saisic Expor Earnings Level Level Expor earnings is no saionary a level Firs Firs Firs difference of expor earnings is saionary Log GDP Level Level GDP is no saionary a level Firs Firs Firs difference of GDP is saionary 91 GMJ Vol. 16, No. 1, June, 216

4 Coinegraion Tes In Table 2, we give he resuls of he Johansen Coinegraion Tes. According o he resuls of he ADF and PP in Table 2,he variables have he same order of inegraion, i.e., I(1) and he Johansen Coinegraion Tes was used o find ou he coinegraion rank and he number of coinegraing vecors. The null hypohesis of r = (i.e., here is no coinegraion) is rejeced agains he alernaive hypohesis of r =1 a he 5% level of significance in case of he Max-Eigen saisic. Similarly, going by he resul of he Trace saisic, he null hypohesis of r = is rejeced agains he alernaive hypohesis of r 1. In Table 2, boh race saisic and maximum eigen saisic show ha here is one coinegraion equaion a.5 level of significan. We fi VECM because he wo condiions for using he vecor error correcion model are me. Table 2 Unresriced Coinegraion Rank Tes for GDP and Expor Earnings Ho: No. of CE(s) Trace Saisics Max-Eigen Saisics None A mos 1* Model Selecion In order o capure he impac of variables observed in he pas ime period in explaining he fuure performance, he ideal lag lengh p (which deermines he appropriae model) is chosen. Firsly, he informaion crieria in Table 3 is use o selec he bes model for he daa. The lag lengh wih he minimum informaion crierion is seleced as he bes model. Thus, according o Table 3, he hree informaion crieria are no consisen in selecing a unique model as he bes model. In specific, AIC selecs lag lengh 8 as he bes model, SIC selecs order 6 and HQ selecs order 7. For proper assessmen, we fied eigh models wih lag lengh from 1 o 8, and sandard diagnosics checking are conduced on hese fied models. Table 3 Lag Order Selecion [GDP, Expor Earnings] Lag AIC SIC HQ * * * * indicaes lag order seleced by he crierion Models VECM (1), VECM (3), VECM (4), VECM (5), VECM (6), VECM (7) and VECM (8) violaed he whie noise es (inverse roos of AR characerisic polynomial, serial correlaion and normaliy es).therefore, VECM(2) which passed he whie noise es was considered as he bes model for our daa. In Table 4, he coinegraion equaions are given along wih he equaion for changes in GDP [firs column, D(GDP)] and changes in Expor earnings (second column). In his sudy, our ineres is he firs column (GDP as he endogenous variable). The coefficiens of he coinegraing equaion conain informaion abou wheher he pas s affec he curren s of he variable under sudy in he long run. In he coinegraing equaion in Table 4, he previous year expor earnings variable [EXE(-1)] is posiive and saisically significan. This means ha, in he long-run, previous year expor earnings is posiively relaed o economic growh. For he vecor error correcion model, a significan lagged co-efficien implies ha pas equilibrium errors has a role in deermining he curren oucomes in he shor run. The lagged coefficiens of change in GDP are posiive bu only he second lag is saisically significan a.5. This indicaes ha higher second lag of GDP has posiive effec on curren GDP in he shor run. However, his sudy revealed ha previous GDP has posiive effec on curren GDP. The lagged coefficiens of change in expor earnings are posiive bu only he second lag is saisically significan a.5. This suggess ha higher expor earnings have posiive effec on GDP in he shor run 92 GMJ Vol. 16, No. 1, June, 216

5 Table 4 Vecor Error Correcion Model Esimaes [VECM (2)] Coinegraing Coin. Eq1 Equaion: GDP(-1) 1. EXE(-1) 8.55E-7 (9.2E-8) [ ] C Error Correcion: D(GDP) D(EXE) CoinEq (.12659) ( ) [ ] [.3474] D(GDP(-1)) (.16184) ( ) [ ] [ ] D(GDP(-2)) (.17256) ( ) [ ] [ ] D(EXE(-1)) 1.88E (1.2E-7) (.32251) [ ] [.69266] D(EXE(-2)) 5.2E (1.4E-7) (.36267) [ ] [ ] C ( ) (4.6E+7) [ ] [ 1.235] Log likelihood Akaike informaion crierion Schwarz crierion (a) Sandard errors in ( ) and -saisics in [ ] EXE denoes expor earnings of cocoa Causaliy Tes wih VECM The causaliy es wih VECM (2) is presened in Table 5. The null hypohesis ha expor earnings does no Granger cause GDP is esed using changes in GDP (D(GDP), and changes in expor earnings (D(expor earnings). These variables are saionary in heir firs difference form in sandard Granger causaliy regression. The null hypohesis is acceped or rejeced based on he F- es o deermine he join significance of he resricions under he null hypohesis. In his sudy, our ineres is o esablish he direcion of influence beween economic growh (GDP) and expor earnings and if here is a feedback influence. In Table 5, all he p s of he variables are less han.5; his indicaes ha he coefficien of Expor earnings is no zero in he equaion for GDP. Thus, he null hypohesis ha expor earnings does no granger cause economic growh (GDP) can be rejeced and a bi-direcional causaliy is observed from expor earnings o economic growh (GDP). Table 5 Resuls of Granger causaliy es Null Hypohesis: EXPORT Earnings does no Granger Cause GDP GDP does no Granger Cause EXPORT Earnings EXPORT Earnings GDP F-Saisic Conclusion Rejec Ho Rejec Ho GDP EXPORT Earnings I is observed ha here is exisence of a long-run relaionship beween economic (GDP) and expor earnings, since here is presence of bi-direcional causaliy beween he wo variables. Impulse Response Here, we wan o know how economic growh will reac or behave as a resul of one sandard deviaion shock or impulse or innovaion in expor earnings and iself. We presen he graphical resuls in Figure 1. The firs graph is he response of economic growh o iself as a resul of one sandard deviaion shock. The graph shows ha when here is one sandard deviaion shock, he influence of economic growh on iself decreases bu is posiive unil he fourh year. A he fifh year, is influence is negaive bu begins o increase unil i becomes posiive a he sevenh year and coninue o rise. The second graph (horizonally) is he response of economic growh (GDP) as resuls of one sandard deviaion shock or innovaion o expor earnings of cocoa. The response of GDP on expor earnings increases rapidly and becomes consan a he fifh and sixh year bu sill rising. In he hird graph, he response of expor earnings o GDP iniially decreases and i is negaive bu sars rising before he second year (and becomes posiive a he second year). Is influence sars o decrease negaively a he sixh year and never becomes posiive up o he enh year. 93 GMJ Vol. 16, No. 1, June, 216

6 Response o Nonfacorized One S.D. Innovaions Response of GDP o GDP Response of GDP o EXE Response of EXE o GDP Response of EXE o EXE 4,, 4,, 3,, 3,, 2,, 2,, 1,, 1,, -1,, ,, Fig 1 Impulse Response graph for GDP and Expor Earnings 4 Conclusions In his sudy, we explore he causal influence of expor earnings of cocoa owards economic growh (GDP). Here, he ADF and he PP uni roo ess indicae ha he wo series are I(1). The resuls of he race and he maxi-eigen coinegraion ess based on Johansen s procedure indicae he exisence of a coinegraion beween expor earnings of cocoa and GDP. Thus, here is a longrun equilibrium relaionship beween he wo variables in his sudy. The vecor error correcion model of order wo, VECM (2) was considered as he bes model. We observed ha, in he long-run, previous year expor earnings of cocoa is posiively relaed o economic growh. In he shor run, he resuls revealed ha he previous GDP has posiive effec on curren GDP. Again, i was observed ha higher expor earnings of cocoa have posiive effec on GDP in he shor run. Feedback causaliy beween economic growh (GDP) and expor earnings of cocoa is observed. This suggess ha a bi-direcional causaliy is observed from expor earnings of cocoa o economic growh (GDP). The policy implicaion of such empirical evidence may be ha policies aimed a increasing he produciviy and qualiy of cocoa should be implemened. For insance, mos of he producers lack he maerial and financial means o increase he producion of cocoa. Thus, as a policy, he governmen should encourage farmers o form cooperaives so ha hey could be open o loan schemes, which will go a long way o increase produciviy. Also addiional should be added o cocoa before exporing. When his is done, i will lead o a higher rae of economic growh in Ghana. References Amavilah, V. H., (23), Expors and Economic Growh in Namibia, , Economic Working Paper Series, pp Aurangzeb (26), Expors, produciviy and economic growh in Pakisan: a ime series Analysis, The Lahore Journal of Economics, Vol. 11, No. 1, pp Dawson, P. J. (25), Agriculural expors and economic growh in less developed Counries, Agriculural Economics, No. 33, pp Glasure, Y. U. and Lee, A., (1997), Coinegraion, error correcion and he relaionship beween GDP and energy: The case of Souh Korea and 94 GMJ Vol. 16, No. 1, June, 216

7 Singapore, Resource and Energy Economics, Vol. 2, pp Hwang, D. and Gum, B., (1991), The causal relaionship beween energy and GNP: The case of Taiwan, Journal of Energy Developmen, Vol. 16, pp Hyndman R. J., Koehler A. B., Ord J. K. and Snyder R. D. (28), Forecasing wih Exponenial Smoohing: The Sae Space Approach., Springer-Verlag Berlin Heidelberg, pp Jordaan, A. C. and Eia, J. H., (27), Expor and Economic Growh in Namibia: A Granger Causaliy Analysis, Souh African Journal of Economics, Vol. 75, No. 3, pp Paul, S. and Bhaacharya, R. N., (24), Causaliy beween energy consumpion and economic growh in India: A noe on conflicing resuls, Energy Economics, Vol. 26, pp Phung, T. B. (211), Energy Consumpion and Economic Growh in Vienam: Threshold Coinegraion and Causaliy Analysis, Inernaional Journal of Energy Economics and Policy, Vol. 1, No. 1, 211, pp Rangasamy, L., (28), Expors and Economic Growh: The Case of Souh Africa, Journal of Inernaional Developmen, Vol. 21, No. 5, pp Sanjuan-Lopez, A. I. and Dawson, P. J. (21), Agriculural expors and economic growh in developing counries: A panel co-inegraion approach, Journal of Agriculural Economics, Vol. 61, No. 3, pp Shumway R. H. and Soffer D. S. (21), Time Series Analysis and Is Applicaions Wih R Examples, Second Ediion, Springer Science Business Media, LLC, USA, pp Tang, T. C., (26), New Evidence on Expor Expansion, Economic Growh and Causaliy in China, Applied Economics Leers, Vol. 13, pp Yu, E. S. H. and Choi, J. Y. (1985), The causal relaionship beween energy and GNP: An inernaional comparison, Journal of Energy and Developmen, Vol. 1, pp Eric Neebo Wiah is a Lecurer a he Deparmen of Mahemaics, Universiy of Mines and Technology (UMaT), Tarkwa, Ghana. He holds Bachelor of Science degree and Maser of Philosophy degree in Mahemaics from he Kwame Nkrumah Universiy of Science and Technology, Kumasi, Ghana. He obained his Docor of Philosophy from UMaT. His curren research ineres focuses on applying discree and coninuous elemen modeling o vegeaion paerns. Also, ineresed in numerical weaher and climae predicion and regional climae modeling using WRF. Auhors Sampson Twumasi-Ankrah is a Lecurer a he Deparmen of Mahemaics, Universiy of Mines and Technology (UMaT), Tarkwa, Ghana. He holds Ph.D in Saisics from he Universiy of Peradeniya, Sri Lanka. His curren research ineres focuses on model selecion, diagnosics analysis, medical saisics, agriculural saisics, pharmaceuical saisics, exreme evens and economerics. 95 GMJ Vol. 16, No. 1, June, 216