015, Vol. 7, No. Empirical Evidence on Korea s Impor Demand Behavior Revisied Jungho, Baek 1,* 1 Deparmen of Economics, School of Managemen, Universiy of Alaska Fairbanks, AK, USA *Correspondence: Deparmen of Economics, School of Managemen, Universiy of Alaska Fairbanks, AK, USA Received: February 5, 015 Acceped: March 18, 015 Published: June 5, 015 doi:10.596/rae.v7i.7056 URL: hp://dx.doi.org/10.596/rae.v7i.7056 Absrac This paper aemps o re-examine Korea s impor demand behavior wih an enhanced economeric echnique and an up-o-dae daase. To achieve he goal, an auogressive disribued lag (ARDL) approach is adoped. Our resuls show he exisence of he long-run relaionship beween Korea s impors and is major deerminans such as income and price. I is also found ha income plays an imporan role in influencing Korea s impors in boh he shor- and long-run. On he oher hand, price is found o have a significan impac on Korea s impors only in he shor-run. Keywords: ARDL, Coinegraion, Impors, Korea 11 www.macrohink.org/rae
015, Vol. 7, No. 1. Inroducion Numerous sudies have invesigaed Korea s impor demand behavior. This lis includes Mah (1993 and 000), Bahmani-Oskooee and Rhee (1997), Sanos-Paulino (00), Tang (005), Baek (01), Bahmani-Oskooee e al. (01) and Baek (013). Mah (1993), for example, applies OLS o quarerly daa for he period 1971-1988 in esimaing he deerminans of impor demand equaion for Korea; he finds ha Korean impors are more sensiive o price changes over he sample period. Using he same daa se, Bahmani-Oskooee and Rhee (1997) conduc Johansen (1990) coinegraion analysis; hey conclude ha income plays more imporan role in deermining Korea s impors. Mah (000) employs he bounds esing mehod o examine Korea s impor demand for informaion echnology producs over he period 1980-1997; he repors ha boh income and price changes have significan impacs on he impors of hose producs. More recenly, Baek (01) analyzes macroeconomic facors affecing expors and impors in Korea; he finds ha Korean impors are more sensiive o changes in domesic income han oher facors (i.e., exchange rae). This paper aemps o conribue o he exising lieraure by reexamining Korea s impor demand equaion wih an enhanced ime series economerics and an up-o-dae daase. More specifically, as Bahmani-Oskooee and Rhee (1997) did, we use a coinegraion approach in ackling he issue. Unlike hem, however, we employ an auoregressive disribued lag (ARDL) bounds esing approach o coinegraion (hereafer ARDL coinegraion mehod) o quarerly daa for he 1989-014 period. The ARDL coinegraion approach is known o be more efficien and is well suied for small sample size han convenional coinegraion analysis (i.e., Johansen coinegraion). In addiion, we explicily incorporae srucural breaks ino our ARDL modeling. Srucural breaks in ime series are likely o affec esimaed resuls bu have been mosly negleced by previous sudies.(noe 1) I is expeced ha hese effors would lend confidence in he robusness and reliabiliy of our empirical findings.. The Empirical Model In examining impor demand funcions, i is a common pracice o relae he volume of impors demanded o a measure of domesic income (i.e., GDP) and domesic prices relaive o he price of impor subsiues (e.g., Sanos-Paulino, 00; Tang, 005). If he price and income elasiciies of demand are assumed o be consan, he impor demand equaion is defined as: * p m ( e) y (1) p where m is he impor volume; p* is he foreign prices; p is he domesic prices; e is he nominal exchange rae; φ is he price elasiciy of impor demand; y is he domesic income; and τ is he income elasiciy of impor demand. Afer aking logs, Equaion 1 can be expressed as follows: 1 www.macrohink.org/rae
015, Vol. 7, No. lnm 0 1 lnrp ln y () where α 1 =φ and α =τ; rp is he relaive prices. (Noe ) In he empirical model adoped in his paper we modify Equaion o capure a possible srucural break ha may resul in changes in Korea s impor demand. Hence, he following specificaion is used for he empirical analysis: ln m dum 0 1 lnrp ln y 3 where m is he impor volume in Korea; y is he real income of Korea; rp is he relaive price of Korean impors, which is defined as rp / ip cp, where ip is he impor price index for Korea andcp is he consumer price index for Korea as a proxy for domesic price;(noe 3) dum is he dummy variable capuring a possible srucural break in his paper, aking one for 008:4 and zero for oherwise; (Noe 4) and is he error erm including all oher facors affecing he impor demand. If a rise in Korea s real income resuls in an increase in demand for impored producs, 1 is expeced o be posiive. If impor price increases a a faser han domesic price and has a negaive effec on impor demand, is expeced o be negaive. In order o carry ou he ARDL coinegraion procedure, following Pesaran e al. (001), Equaion 3 is reformulaed as a condiional error correcion model (ECM) as follows: lnm dum 0 ln m 1 1 1 p i1 ln y ln m i 1 i lnrp 3 p i0 1 ln y i i p i0 ln rp i i (4) Wihou lagged level variables ha is, m 1, y 1and rp 1, Equaion 4 would be he same a sandard VAR model. The linear combinaion of lagged level variables is replaced he lagged error-erm from Equaion 3, which resuls in an error-correcion specificaion expressed in Equaion 4. The ARDL coinegraion procedure consiss of he following wo seps. The firs sep of he modeling is o idenify he presence of he long-run (coinegraion) relaionship among he hree variables by conducing he join significance es of m 1, y 1and rp 1 in Equaion 4. For his purpose, he sandard F-es can be used o es he null hypohesis of none-exisence of he long-run relaionship (no coinegraion) - ha is, H 0 0 : 1 3 13 www.macrohink.org/rae
015, Vol. 7, No. agains H 0, 0, 0 0 : 1 3. Under he null hypohesis, however, he (asympoic) disribuion of his F-saisic is non-sandard, irrespecive of wheher he variables are inegraed of order zero ( I (0) ) or inegraed of order one ( I (1) ) processes. Pesaran e al. (001) hus abulae wo new ses of criical values ha accoun for inegraing properies of all variables. If he calculaed F -saisic lies above he criical value of band, a conclusive decision can be reached wihou carrying ou uni roos ess on he variables; for example, if he calculaed F-saisic is higher (lower) han he upper (lower) criical value, hen he null hypohesis can (canno) be rejeced. Unlike convenional coinegraion mehods ha require classifying he variables ino I(0) and I (1), herefore, his procedure does no require pre uni roo esing. A he second sep he long-run effecs and he associaed shor-run effecs are simulaneously esimaed in he seleced ARDL framework. The long-run esimaes of he seleced variables are derived from esimaes of and 3 normalized on 1. The shor-run effecs come from he esimaes of coefficiens relaed o firs-differenced variables. 3. Empirical Resuls Because he ARDL coinegraion approach assumes he variables mus be I(0) or I (1) compued F-saisic is no valid wih I() variables. The es o make sure ha no variable in Equaion 4 is I () series is conduced using he Dickey Fuller generalized leas squares (DF-GLS) es (Ellio e al., 1996). The resuls indicae ha he null hypohesis of a uni roo canno (can) be rejeced for any of he levels (firs differences) of he variables a he 10% level (Table 1), suggesing ha all he series are I(1) processes. Because of inabiliy of he DF-GLS o capure he possibiliy of a srucural change, however, he power of he DF-GLS es is likely o decrease wih an undeeced srucural break in he series, hereby providing misleading resuls. For compleeness, herefore, we invesigae uni roos in he exisence of a srucural break using he Zivo and Andrews (ZA) es. The resuls shows ha he null, he hypohesis canno (can) be rejeced for he levels (firs differences) of im and y bu can be rejeced for he level of rp (Table 1), indicaing ha im and y are I(1) processes and rp is I(0) process, respecively. Unlike convenional coinegraion mehods, herefore, he F-es is sill applicable afer aking ino accoun a srucural break in he series, proving ha he use of he ARDL model is indeed desirable o deal wih he curren issue. 14 www.macrohink.org/rae
015, Vol. 7, No. Table 1. Resuls of DF-GLS and Zivo-Andrew uni roo ess ln im ln y ln rp ln im ln y ln rp Variable Variable DF-GLS es Level Firs difference -.3-4.83** (1) (7) -0.95-5.68** (1) () -1.4-5.30** (4) Zivo-Andrew Tes Level Time break Firs difference Time break -3.99-5.33** 008:4 1998:4 -.8-5.65** 008:4 1991:1-5.0** 007:4 Noes: ** and * denoe rejecion of he null hypohesis a he 5% and 10% levels, respecively. The 5% and 10% criical values for he DF-GLS (Zivo-Andrew), including a consan and rend, are -3.03 (-4.80) and -.74 (-4.58), respecively. Parenheses are lag lenghs. As discussed above, he firs sep of he ARDL approach requires he applicaion of he F-es in order o idenify wheher he long-run (coinegraion) relaionship among he hree variables ( m, y and rp ) exiss or no. The resuls show ha he calculaed F-saisic (4.39) is above he upper criical value (4.19) a he 10% level, (Noe 5) hereby rejecing he null hypohesis of no coinegraion, suggesing here exiss he long-run equilibrium relaionship beween Korea s impor demand and is major deerminans over he period 1989-014. Table. Resuls of esimaed long-run coefficiens Variable Coefficien -raio ln y 1.88 1.07** ln rp -0.01-0.03 dum -0.53-1.83* consan -4.18-9.89** Noes: ** and * represen saisical significance a he 5% and 10% levels, respecively. Wih he idenified long-run relaionship among he hree variables, he long-run coefficiens and he associaed shor-run coefficiens are hen esimaed based on he seleced ARDL 15 www.macrohink.org/rae
015, Vol. 7, No. model defined by Equaion 4. To ha end, he maximum lag of six is chosen based on he Akaike Informaion Crierion (AIC). The esimaes of he long-run coefficiens are shown in Table. The esimaed coefficien on he real GDP carries a posiive sign and is saisically significan a he 5% level, indicaing ha economic growh in Korea ends o increase impored producs for Korea in he long-run; for example, a 1% increase in economic growh pushes he Korean impors o increase by 1.88%. This empirical evidence subsaniaes he findings of Mah (1993), Bahmani-Oskooee and Rhee (1997), Tang (005), Baek (01 and 013). On he oher hand, he esimaed coefficien on he relaive price carries a negaive sign as expeced bu is found o be saisically insignifican even a he 10% level, suggesing ha changes in relaive price do no seem o have a significan effec on Korean impors in he long-run. This finding is a odds wih oher sudies (e.g., Mah, 1993 and 000), which argue for a srong, negaive effec of he relaive price on Korean impors. Finally, he esimaed coefficien on he dummy variable is significanly negaive a he 10% level, implying ha he recen financial crisis indeed reduces Korean impors. The esimaes of he shor-run (including an error-correcion erm) are summarized in Table 3. The resuls show ha, as seen in he long-run findings, he real income of Korea has a significan effec on Korean impors in he shor-run. Similarly, he recen financial crisis is also found o have a significan shor-run effec on Korean impors. Unlike he long-run resuls, however, he relaive price is found o be saisically significan a leas a he 10% level, indicaing ha he relaive price is an imporan deerminan of Korea s impors in he shor-run. I is imporan o noe ha he error-correcion erm ( ec 1 ) is negaive and saisically significan a he 5% level, which is anoher sign of coinegraion (Kremers e al., 199; Banerjee e al., 1998). In addiion, he esimaed coefficien of ec 1 reflecs he adjusmen speed oward he long-run equilibrium. The coefficien of -0.15 in he model, for example, means ha approximaely 15% of he adjusmen akes place wihin one quarer; in oher words, i akes more han 6 quarers (e.g., 1/0.15 =6.67 quarers) in order o achieve he long-run equilibrium. Finally, he esimaed ARDL model passes all he diagnosic ess (Table 3). For example, he Lagrange Muliplier (LM) and Ramsey s RESET saisics are used o es for serial correlaion and for model specificaion, respecively. Boh saisics are disribued as χ wih differen degrees of freedom. The calculaed LM saisic using four degrees of freedom is found o saisically insignifican a he 10% level, supporing serial correlaion free residuals. The calculaed RESET saisic wih on degree of freedom is also found o be saisically insignifican, suggesing ha our ARDL model is correcly specified. 16 www.macrohink.org/rae
015, Vol. 7, No. Table 3. Resuls of error-correcion equaion of Korea s impors Variable Coefficien -raio ln y.85 9.88** ln y 1 0.63.19** ln y -0. -0.86 ln y 3-0.36-1.51 ln y 4 0.61.69** lnrp -0.14-1.89* lnrp 1 0..9** dum -0.08 -.49** ec 1-0.15 -.89** (4) =.14 [0.71], (1) =1.50 [0.], () =1.17 [0.56], (1) =0.06 [0.80] SC FF N H Noes: ** and * denoe saisical significance a he 5% and 10% levels, respecively. ec 1 indicaes an error-correcion erm. (4), (1), (), and (1) denoe chi-square SC saisics o es for no serial correlaion, no funcional form misspecificaion, normaliy and homoskedasiciy, respecively. Brackes are p-values. FF N H 4. Conclusions and Policy Implicaions In his shor paper, we empirically reexamine Korea s impor demand equaion. To address his issue adequaely, we adop an enhanced ime series economerics - an auogressive disribued lag (ARDL) bounds esing approach o coinegraion. Furhermore, we pay close aenion o an imporan ime series issue relaed o how we should incorporae a poenial srucural break in our modeling. Alhough he use of he ARDL mehod does no radically change he findings of previous sudies on he issue, i does make a subsanive difference o he esimaes of some variables, which is he main conribuion of his paper. The resuls show ha here exiss he long-run relaionship beween Korea s impor demand and is major deerminans such as domesic income and price. We also find ha economic growh plays a pivoal role in influencing Korea s impors in boh he shor- and long-run. On he oher hand, price is found o be an imporan deerminan in Korea s impors in he shor-run, bu no in he long-run. Finally, he marke shock such as he recen financial crisis is found o significanly reduce Korea s impors in he shor- and long-run. An imporan implicaion of our findings is ha, given he significan income impac on Korea s impors, Korea s recovery from he recen slow growh an average GDP growh rae of 0.6% over he pas hree years - is likely o increase Korea s demand for impors, hereby causing he rade surplus o deeriorae. Anoher imporan implicaion is ha, since Korea s impors seem o significanly respond o changes in relaive prices in he shor-run, a 17 www.macrohink.org/rae
015, Vol. 7, No. depreciaion of he Korean won may lead o an increase in inflaion as Korean prices of impored producs end o increase. This hus explains why simulaneous analysis of he shor- and long-run is crucial in modeling he deerminans of Korea s impor demand. I should be poined ou ha, since our analysis is conduced using aggregae impor daa beween Korea and he res of he world, he findings may suffer from he so-called aggregaion bias problem (Baek, 011; Baek, 014); ha is, wihin aggregae impors some of significan facor impacs (e.g., income and relaive prices) are likely o be offse by oher insignifican effecs, hereby resuling in insignifican impacs. Fuure research should address his issue by employing disaggregae rade daa in a modeling. Acknowledgemen The auhor is graeful o an anonymous reviewer s commen and he edior s assisance for improving he manuscrip from an earlier version. Any remaining errors are he auhor s. References Baek, J. (011). How Sensiive is U.S. Agriculural Trade o he Bilaeral Exchange Rae?: Evidence from Bulk, Inermediae, and Consumer-Oriened Producs, Agriculural Economics, 4, 387-403. hp://dx.doi.org/10.1111/j.1574-086.010.0055.x Baek, J. (01). Exchange Rae Sensiiviy of Korea-U.S. Bilaeral Trade: Evidence from Indusrial Trade Daa, Journal of Korea Trade, 16, 1-1. hp://www.dbpia.co.kr/issue/134441. Baek, J. (013). Does he Exchange Rae Maer o Bilaeral Trade beween Korea and Japan? Evidence from Commodiy Trade Daa, Economic Modelling, 30, 856-86. hp://dx.doi.org/10.1016/j.econmod.01.11.00 Baek, J. (014). Exchange Rae Effecs on Korea-U.S. Bilaeral Trade: A New Look, Research in Economics, 68, 14-1. hp://dx.doi.org/10.1016/j.rie.014.05.00 Bahmani-Oskooee, M., & Rhee, H.J. (1997). Srucural Change in Impor Demand Behavior, he Korean Experience: a Reexaminaion, Journal of Policy Modeling, 19, 187-193. hp://dx.doi.org/10.1016/0161-8938(95)00146-8 Bahmani-Oskooee, M., Hanafiah, H., & Hegery, S.W. (01). Exchange Rae Volailiy and Indusry Trade beween he U.S. and Korea, Journal of Economic Developmen, 37, 1-7. Banerjee, A., Dolado, J.J., & Mesre, R. (1998). Error-Correcion Mechanism Tess for Coinegraion in a Single-Equaion Framework, Journal of Time Series Analysis, 19, 67-84. hp://dx.doi.org/10.1111/1467-989.00091 Ellio, G., Rohenberg, T., & Sock, J. (1996). Efficien Tess for an Auoregressive Uni Roo, 18 www.macrohink.org/rae
Economerica, 64, 813 836. hp://dx.doi.org/ 10.307/171846. Research in Applied Economics 015, Vol. 7, No. Johansen, S. (1988). Saisical Analysis of Coinegraion Vecor, Journal of Economic Dynamics and Conrol, 1, 31-54. hp://dx.doi.org/10.1016/0165-1889(88)90041-3 Johansen, S., & Juselius, K. (1990). Maximum Likelihood Esimaion and Inference on Coinegraion wih Applicaions o he Demand for Money, Oxford Bullein of Economics and Saisics, 5, 169-10. hp://dx.doi.org/10.1111/j.1468-0084.1990.mp500003.x Kremers, J.J.M, Ericson, N.R., & Dolado, J.J. (199). The Power of Coinegraion Tess, Oxford Bullein of Economics and Saisics, 54, 35-48. hp://dx.doi.org/10.1111/j.1468-0084.199.b00005.x Mah, J.S. (1993). Srucural Change in Impor Demand Behavior: he Korean Experience, Journal of Policy Modeling, 15, 3-7. hp://dx.doi.org/10.1016/0161-8938(93)90017-k Mah, J.S. (000). An Empirical Examinaion of he Disaggregaed Impor Demand of Korea he Case of Informaion Technology Producs, Journal of Asian Economics, 11, 37-44. hp://dx.doi.org/10.1016/s1049-0078(00)00053-1 Pesaran, M.H., Shin, Y., & Smih, R.J. (001). Bounds Tesing Approaches o he Analysis of Level Relaionships, Journal of Applied Economerics, 16, 89-36. hp://dx.doi.org/10.100/jae.616 Sanos-Paulino, A.U. (00). The Effec of Trade Liberalizaion on Impors in Seleced Developing Counries, World Developmen, 30, 959-974. hp://dx.doi.org/10.1016/s0305-750x(0)00014-1 Tang, T.C. (005). Revisiing Souh Korea s Impor Demand Behavior: a Coinegraion Analysis, Asian Economic Journal, 19, 9-50. hp://dx.doi.org/10.1111/j.1467-8381.005.0003.x Wooldridge, J. (013). Inroducory economerics: A modern approach (5 h Souh-Wesern Cengage Learning. ed.). Ohio: Noes Noe 1. I should be poined ou ha several sudies for example, Tang (005), Bahmani-Oskooee e al. (01) and Baek (01 and 013) - use he ARDL coinegraion approach in examining Korea s impor demand funcion. However, hey do no incorporae srucural breaks in heir modelling. Furhermore, some sudies ackle he issue wih a relaively small size (i.e., Tang, 005; Bahmani-Oskooee e al., 01). Since a small sample size ends o increase sampling variances hrough a decrease in he sample variaion in each of explanaory variables, his problem may cause he esimaed coefficiens in a model o be very sensiive o is specificaions and even inefficien, hereby undermining he credibiliy of 19 www.macrohink.org/rae
015, Vol. 7, No. heir findings (Wooldridge, 013). Noe. I is worh menioning ha in his analyical framework, oher facors such as exchange rae, marke srucure and rade barriers are assumed o affec impor demand hrough changes in relaive prices (Tang, 005). Noe 3. Quarerly daa for he period 1989:1-014: are obained from he Organizaion for Economic and Cooperaion Developmen (OECD) saisical daabase. Noe 4. This break is idenified based on he ZA es (see he empirical resuls secion) and involves he recen financial crisis ha peaked in 008. Noe 5. The criical value bounds are generaed for he sample size (n=10) and 0,000 replicaions using he saisical sofware known as Microfi 5. Copyrigh Disclaimer Copyrigh for his aricle is reained by he auhor(s), wih firs publicaion righs graned o he journal. This is an open-access aricle disribued under he erms and condiions of he Creaive Commons Aribuion license (hp://creaivecommons.org/licenses/by/3.0/). 0 www.macrohink.org/rae