Paper for Annual Meeing 2015 Absrac World Trade Flows in Phoovolaic Cells: A Graviy Approach Including Bilaeral Tariff Raes * Asuko Masumura (Tokyo Inernaional Universiy) Absrac This paper invesigaes he deerminans of world rade flows in one of he principal environmenal goods, phoovolaic cells, focusing on he period of growing rade of his indusry beween 2000 and 2004, wih he emphasis on he effecs of bilaeral ariffs. Our cross-secion analysis based on he graviy model including bilaeral ariffs as one of he rade coss demonsraes he sysemaic effecs of bilaeral ariff raes as well as disance effec and APEC bloc effec on he bilaeral rade flows in he phoovolaic cells secor among 51 counries in each year beween 2000 and 2004. Our panel analysis based on a fixed effec model of he graviy equaion by allowing rading-pair heerogeneiy using he daa of 2000-2004 consolidaes he significan effec of bilaeral ariff raes on he rade flows, and demonsraes he globalizaion effec on rade growh in his secor. As for he paern of bilaeral rade of his secor, we find from our cross-secion analysis ha he bilaeral rade flows are larger for he counry-pairs wih similar GDP per capia by exchanging he differeniaed producs, which can be explained by he monopolisic compeiion model. The resuls of our graviy analysis for phoovolaic cells sugges he effeciveness of he curren effors of APEC and he WTO for ariff reducion on various kinds of environmenal goods for expanding bilaeral rade flows of hose goods among he covered counries of eiher developed or newly indusrialized counries, which would lead o environmenal proecion hrough worldwide diffusion of hose goods. * This research is funded by Grans -in-aid Scienific Research(C), Japan Sociey for Promoing Science, as research wih no.25380319. 1
Ⅰ. Inroducion This paper examines he deerminans of he world rade nework beween 2000 and 2004 focusing on one of he principal environmenal goods, phoovolaic cells, wih emphasis on he effecs of impor ariffs. The imporance of rade liberalizaion in environmenal goods has been discussed since he end of he Uruguay Round in he World Trade Organizaion (WTO); and Paragraph 31 (ⅲ) of he Doha Miniserial Declaraion in 2001 considers he reducion or eliminaion of ariffs and non-ariff barriers o rade in environmenal goods and services o pursue win-win-win resuls for rade, economic developmen and he environmen. Members of he WTO have been working o liberalize rade in goods and services ha can benefi by proecing he environmen, and he alks for his kind of rade liberalizaion ake place in Special Sessions of he Trade and Environmen Commiee of he WTO. On he oher hand, he members of Asia-Pacific Economic Cooperaion (APEC) are making effors o ensure he realizaion of heir landmark commimen o reduce ariffs on an agreed lis of 54 environmenal goods o 5 % or less by he end of 2015. The purpose of his paper is o invesigae how he ariff barriers affeced rade growh in one of he imporan environmenal goods, phoovolaic cells, in order o evaluae he effeciveness of curren effors for ariff reducion in environmenal goods in APEC and he WTO. Based on he graviy model esimaion including ariff raes as one of he rade coss, we rigorously invesigae he effecs of ariffs as well as he oher deerminans on rade volume for he secor of phoovolaic cells among 51 counries, including APEC counries, European counries, and he newly indusrialized counries. 1 I seems ha he ariff reducion has been playing an imporan role in he growh of world rade since World War Ⅱ, ogeher wih income growh, ranspor cos declines, and oher facors. For example, he empirical resuls of Baier and Bersgrand (2001) sugges ha income growh explains abou 67%, ariff reducions abou 25%, and ranspor cos declines abou 8% of average world rade growh beween he 1950s and 1980s, using he graviy model esimaion. This paper is organized as follows. Secion Ⅱ provides he movemen of world ariff reducion of phoovolaic cells from he lae 1990s, and shows how he rade of phoovolaic cells increased wih he ariff reducion. Secion Ⅲ presens he analyical model and equaions used in he empirical sudy of he phoovolaic cells secor, and he las par of secion Ⅲ discusses he characerisics of daa used in he economeric analysis. Secion Ⅳ invesigaes he esimaion resuls of boh cross-secion analysis 1 The seleced counries are indicaed in Table 2. 2
and panel analysis o clarify he main findings for he world rade of phoovolaic cells in he firs half of he 2000s when he rade of phoovolaic solar cells increased jus afer he ariff removal in he developed counries and he ariff reducion of developing counries. In Secion Ⅴ, he conclusion of his paper is presened. Ⅱ. World Trade and Tariffs of Solar Cells The world rade and ariffs of phoovolaic cells are classified in HS 854140, and his HS code is included in he lis of environmenal goods of boh he WTO and APEC as he principal iem in he caegory of renewable energy. The conen of HS code 854140 is phoosensiive semiconducor devices, including boh phoovolaic cells wheher or no assembled in modules or made up ino panels and ligh emiing diodes. In order o sor ou phoovolaic cells by excluding ligh emiing diodes, we have o invesigae more deailed rade daa han provided by he six digi HS. As he HS code of hose disaggregaed daa depends on he counries, we colleced he deailed daa for phoovolaic cells by examining rade daa of each of 51 counries included in he analysis. According o he descripion in he APEC lis of environmenal goods, ANNEX C of APEC (2012), he environmenal benefi of solar phoovolaic cells is o generae elecriciy in an environmenally benign manner (wih no emissions, noise or hea generaed). Table 1 provides he MFN applied ariff raes in phoovolaic cells in he counries wih a leas one non-zero ariff rae during 1998 and 2005, included in he graviy analysis. The ariff barriers were removed in he lae 1990s in many counries including he counries such as Ausralia, USA, and EU members, and also in he newly developed counries such as Souh Korea, Philippines, Taiwan and Thailand. The counries which joined EU in 2004 such as Cyprus, Czech Republic, Hungary, Mala and Slovakia had removed ariffs a few years earlier han heir enry o EU. The counries which sill had high ariff raes in solar cells in he firs half of 2000s are Brazil, Chile, China, India, Peru, and Russia, alhough China removed is ariff on phoovolaic cells in 2002. 3
Table 1 MFN Applied Tariffs for Solar Cells (Average), 1998-2005 (%) Year 1998y 1999y 2000y 2001y 2002y 2003y 2004y 2005y Ausralia 1.03 0.00 0.00 0.00 0.00 0.00 0.00 0.00 Brazil 10.18 9.50 9.22 7.50 6.89 3.78 3.67 4.44 Chile 11.00 10.00 9.00 0.00 7.00 0.00 6.00 6.00 China 10.00 10.00 10.00 10.00 0.00 0.00 0.00 0.00 Cyprus 1.08 0.00 0.00 0.00 0.00 0.00 0.00 0.00 Czech Republic 1.60 0.00 0.00 0.00 0.00 0.00 0.00 0.00 European Union 1.08 0.00 0.00 0.00 0.00 0.00 0.00 0.00 Hungary 5.00 5.00 5.00 5.00 0.00 0.00 0.00 0.00 India 20.00 15.00 16.50 15.00 15.00 15.00 10.00 0.00 Indonesia 2.50 2.50 0.00 0.00 0.00 0.00 0.00 0.00 Souh Korea 2.36 0.00 0.00 0.00 0.00 0.00 0.00 0.00 Mala 10.46 10.46 11.05 8.00 8.00 0.00 0.00 0.00 Peru 12.00 12.00 12.00 8.00 8.00 8.00 8.00 8.00 Philippines 3.00 3.00 0.00 0.00 0.00 0.00 0.00 0.00 Russia 20.00 20.00 20.00 20.00 20.00 20.00 20.00 20.00 Slovakia 1.60 0.80 0.00 0.00 0.00 0.00 0.00 0.00 Taiwan 0.28 0.21 0.00 0.00 0.00 0.00 0.00 0.00 Thailand 35.00 35.00 0.00 0.00 0.00 0.00 0.00 0.00 USA 0.06 0.00 0.00 0.00 0.00 0.00 0.00 0.00 Source : WTO, Tariff Download Faciliy Noe: The seleced counries are hose wih a leas one non-zero ariff rae beween 1998 and 2005. 6,000,000 Solar cells expors of 50 counries (1,000US$) 5,000,000 4,000,000 3,000,000 2,000,000 1,000,000 0 1999y 2000y 2001y 2002y 2003y 2004y Source: Daa by Global Trade Alas 4
Figure 1 shows he world rade in phoovolaic cells from 1999 o 2004, by aggregaing he oal expor value o he world from 51 counries included in his analysis. We can confirm ha he rade value in his environmenal good has increased in he early 2000s in various counries in accordance wih he worldwide rade liberalizaion. The years 2000-2004 were seleced in his analysis for he following reasons. The esimaion of he graviy equaion for 1999 was no saisfacory because of he oo few observaions of rade pairs in he of phoovolaic cells secor, and afer 2005, i is difficul o idenify he influence of bilaeral ariffs as India, a significan counry, removed ariffs on phoovolaic cells in 2005. Ⅲ. Analyical Framework and Daa This secion presens he mehodological base of his research on world rade in he phoovolaic cells secor, and he daa used for he se of variables. The graviy model is used for he analysis, which Bayoumi and Eichengreen (1997) described as he workhorse for empirical sudies of he paern of rade. Also, Rauch (1999) expressed he graviy equaion as he sandard empirical framework used o predic how counries mach up in inernaional rade 2. The radiional form of he graviy model can be expressed as equaion (1), according o Bergsrand and Egger (2010). ln PX M m 0 + β1 ln GDPi + β 2 ln GDPj + β m ln( Z ) ε (1) m= 3 = β + where PX is he value (in curren prices) of he merchandise rade flow from exporer i o imporer j, GDPi (GDPj) is he level of nominal gross domesic produc in counry i (counryj), Z m ( m = 3,..., M ) is a se of observables o which bilaeral rade barriers are relaed, including he bilaeral physical disance beween he economic ceners of counries i and j, and ε is a disurbance erm, wih expeced signs for coefficiens β, β > 0, andβ 0. ln refers o he naural logarihm. 1 > 0 2 m < A saisical overview in Cheng and Wall (2005) clarifies various forms of he graviy model o esimae bilaeral rade flows, using boh cross-secion and panel analysis. By saring wih a general model (GM) shown in equaion (2) wih no resricions on he parameers, he sandard single-year cross-secion model (CS), a pooled-cross-secion model (PCS), and a fixed effecs model (FE) are inroduced wih some specific resricions on he parameers. The following overview of he four ypes of he graviy model is quoed from Cheng and Wall (2005). 2 Baier and Bergsrand (2001) review how he researchers have expressed he inrinsic qualiies of he graviy model in inernaional rade. 5
General Model (GM): ln X = α 0 + α + α + β ' Z + ε, =1,,T, (2) where X is expors from counry i o counry j in year, Z =[z1, z2, ] is he 1 k vecor of graviy explanaory variables. The inercep has hree pars: one common o all years and counry pairs, α 0, one specific o year and common o all pairs, α, and one specific o he counry pairs and common o all years, α. The disurbance erm, ε is assumed o be normally disribued wih zero mean and consan variance for all observaions, and he disurbances are pairwise uncorrelaed. The sandard single-year cross-secion model imposes he resricions ha he slopes and inerceps are he same across counry pairs, ha is α =0, and β = β as specified in equaion (3). Cross-Secion Model (CS) : ln X = α 0 + α + β ' Z + ε =1,,T, (3) where α 0 and α canno be separaed. The pooled cross-secion model imposes he resricion on he general model ha he parameer vecor is he same for all, alhough i normally allows he inerceps o differ over ime, and akes he form of equaion (4). Pooled Cross-Secion Model (PCS) : ln X = α 0 + α + β ' Z + ε =1,,T, (4) In he following fixed effec model, he resricion ha he counry-pair inercep erms equal zero is removed, and his model can be esimaed using leas squares wih dummy variables for each of he counry pairs. In his specificaion, he explanaory variables such as disance and common language are eliminaed, because hey are fixed over ime, even hough hey are no collinear wih he counry-specific effecs. Fixed Effec Model (FE) : ln X = α 0 + α + α + β ' Z + ε =1,,T, (5) This model is also considered o be he preferred mehod in Feensra (2004), in order o ake accoun of he unobserved bilaeral price indexes, mulilaeral resisance, ermed by Anderson and van Wincoop (2003). In Ⅲ-1, he esimaion model of cross-secion analysis is presened, in Ⅲ-2, he specificaions for panel analysis are developed, and inⅢ-3, an explanaion of he daa is presened. Ⅲ-1.Specificaions of Cross-Secion Analysis Bergsrand and Egger (2010) poin ou wo aspecs of equaion (1). Firsly, acual bilaeral rade flows can be explained quie well by his specific and simple log-linear 6
equaion, and secondly, bilaeral rade flows are srongly influenced by economic rade fricions (or rade coss). They show ha rade fricions have wo forms, one being naural rade coss and he oher is unnaural (or policy-based) rade coss. Naural rade coss are made of ranspor coss and oher relaed coss such as ime, conrac enforcemen coss, coss associaed wih he use of differen currencies, legal and regulaory coss, and local disribuion coss. On he oher hand, unnaural rade coss can be decomposed ino ariffs, axes on goods crossing inernaional borders, and nonariff barriers on inernaional rade. In his paper, we explicily include ariff raes of he phoovolaic cells secor, as well as disance beween wo pair counries, wih dummy variables for common language and APEC bloc o ake ino accoun he effecs relaing o rade coss in equaion (1). According o he graviy esimaion for manufacuring rade by Ando and Kimura (2013), in order o ake ino accoun he effec of differences in facor endowmens on bilaeral rade, he difference in GDP per capia beween wo counries is inroduced. The coefficien will be posiive if he difference in facor endowmens is he imporan deerminan for rade increase of he phoovolaic cells secor as Heckscher-Ohlin Model predics, and i will be negaive if he similariy of paern of facor endowmen is imporan for rade increase in his secor as prediced by he new rade heory based on monopolisic compeiion wih differeniaed producs. ln PX β ln ln ln ln (1 AP + ε = 0 + β1 GDPi + β 2 GDPj + β 3 GDPPC + β 4 DIST + β 5 + TAR ) + β 6CL + β 7 (6) where GDPPC is he absolue value of he difference in GDP per capia beween counry i and counry j, DIST is he geographical disance beween capials of counry i and counry j, TAR is bilaeral ariff raes for phoovolaic cells, CL is a dummy variable aking uniy if an official language is common o boh counries and zero oherwise, and AP is a dummy variable wih uniy if boh of exporing counry and imporing counry are members of APEC. We esimae he graviy equaion for phoovolaic cells rade among 51 counries, including developed and newly indusrialized counries, for each of 5 years from 2000 o 2004, by OLS according o equaion (2). The expeced signs in equaion (6) are as follows; β 1 > 0, β 2 > 0, β 4 < 0, β < 0, β > 5 6 0, β 7 > 0. The sign of he absolue value of he difference in GDP per capia beween counry i and counry j, β 3, depends on he characerisics of he goods as already menioned. In order o examine he effecs of he combinaion of he explanaory variables, four ypes of specificaions are esimaed selecing some of he 7
variables, as shown in Table 4. Ⅲ-2.Specificaions of Panel Analysis As Bergsrand and Egger (2010) sae, researchers have urned more o panel daa o esimae he effecs of each deerminan of he graviy equaion in order o avoid unobservable heerogeneiy across counry pairs. According o Cheng and Walls (2005), we esimae a specificaion wih a counry-pair fixed effec and a ime effec, by inroducing dummy variables for each of he counry pairs and ime dummy variables. The dependen variable is real expors of phoovolaic cells, and independen variables are log of real GDP of exporing counry i, (lnrgdpi), log of real GDP of imporing counry j (lnrgdpj), and log of ariff raes of phoovolaic cells in he impor of counry j from counry i, ln(1+tar). ln RX = α + α + β1 ln( RGDP ) + β 2 ln( RGDP ) + β 3 ln(1 + TAR ) + ε (7) i j where α is he specific counry-pair effec beween he rading parners. The counry-pair inerceps include he effecs of all omied variables ha are cross-secionally specific bu remain consan over ime, such as disance and common language. We also esimae he following equaion of he PCS model by OLS wih ime dummies, o compare he esimaion resuls wih he Fixed Effecs model. ln RX α + α + β ln( RGDP ) + β ln( RGDP ) + β ln(1 + TAR ) + β ln( DIST + β AP + β CL + ε = 0 1 i 2 j 3 4 ) 4 5 (8) where ln RX is log of real expors of phoovolaic cells as dependen variable, α 0 denoes he par of he inercep common o all years and rading pairs, α is he year-specific effec common o all rading pairs, lnrgdpi is log of real GDP of imporing counry j, lnrgdpj is log of real GDP of imporing counry j, TAR is ariff raes of phoovolaic cells, lndist is log of disance beween counry i and counry j, AP is a dummy variable wih 1 if boh counries are members of APEC and 0 oherwise, and CL is a dummy variable if boh counries share a common official language. Explanaory variables oher han lnrgdpi and lnrgdpj ake he same form as in equaion (6). Ⅲ-3.Daa The daa used for he esimaion are described briefly as follows. The nominal bilaeral rade values of he secor of phoovolaic cells (in 1,000 US dollars) for 51 rading counries are aken from he Global Trade Alas online daa providing cusoms 8
rade daa repored by he governmen of each counry. We examined he daa of each counry in order o sor ou phoovolaic cell daa, by excluding he daa for ligh emiing diodes from HS 854140 daa. In cross-secion analysis, zero rade flows are excluded, and in panel analysis, hese are excluded from he daa se if all of he five years daa are zero. The daa source of bilaeral ariffs of phoovolaic cells in all five years for each counry is he WTO Tariff Download Faciliy. Nominal GDP (in 1,000 US dollars), real GDP (in 1,000 US dollars), and nominal GDP per capia (in 1,000 US dollars) are aken from he World Bank s World Developmen Indicaors. Trade daa of phoovolaic cells are scaled by GDP deflaors o generae real rade flows for he panel analysis. The daa of bilaeral disance beween capials of he pair counries (in kilomeers) and common official language are from he CEPII (Cenre d Eude Prospecives d Informaions Inernaionales) daabase. Daa of GDP and GDP per capia for Taiwan are from he JETRO daabase originally from Direcorae-General of Budge, Accouning and Saisics, Execuive Yuan, Republic of China, (Taiwan). The lis of 51 counries included in his analysis is shown in Table 2. All APEC counries excep Brunei Darussalam and Papua New Guinea are included, along wih 28 European counries, and Brazil, India, Souh Africa, and Turkey, which are significan newly indusrialized counries. Table 2 Lis of counries in he analysis APEC Europe Oher counries Ausralia Ausria Brazil Canada Belgium India Chile Bulgaria Souh Africa China Croaia Swizerland Hong Kong Cyprus Turkey Indonesia Czech Republic Japan Denmark Korea Esonia Malaysia Finland Mexico France New Zealand Germany Peru Greece Philippines Hungary Russia Ireland Singapore Lihuania Taiwan Mala Thailand Ialy USA Norway Vienam Neherlands Poland Porugal Romania Slovakia Slovenia Spain Sweden Swizerland UK 9
Table 3(a) Descripive Saisics(Cross Secion) ln(export) ln(gdpi) ln(dgdppc) ln(dist) ln(1+tariff) APEC CL 2000 Observaions 899 899 899 899 899 899 899 Mean 4.799 19.812 9.109 8.190 0.013 0.176 0.133 Sd.Dev. 2.563 1.534 1.222 1.135 0.039 0.381 0.340 Max 12.107 23.054 10.519 11.700 0.182 1 1 Min 0.693 15.191 3.176 4.088 0.000 0 0 2001 Observaions 951 951 951 951 951 951 951 Mean 4.851 19.777 9.074 8.214 0.012 0.198 0.126 Sd.Dev. 2.549 1.430 1.174 1.131 0.036 0.398 0.332 Max 12.177 23.087 10.514 11.700 0.182 1 1 Min 0.077 15.181 2.853 4.088 0.000 0 0 2002 Observaions 1027 1027 1027 1027 1027 1027 1027 Mean 4.772 19.848 9.171 8.189 0.010 0.187 0.124 Sd.Dev. 2.610 1.401 1.206 1.141 0.034 0.390 0.329 Max 12.554 23.119 10.641 11.700 0.182 1 1 Min 0.693 15.273 2.256 4.088 0.000 0 0 2003 Observaions 1075 1075 1075 1075 1075 1075 1075 Mean 4.854 19.964 9.363 8.195 0.009 0.173 0.126 Sd.Dev. 2.644 1.447 1.135 1.141 0.034 0.378 0.332 Max 12.806 23.167 10.793 11.700 0.182 1 1 Min 0.693 15.449 3.546 4.088 0.000 0 0 2004 Observaions 1126 1126 1126 1126 1126 1126 1126 Mean 5.108 20.074 9.459 8.206 0.008 0.164 0.123 Sd.Dev. 2.677 1.395 1.100 1.143 0.032 0.371 0.329 Max 13.103 23.231 10.918 11.700 0.182 1 1 Min 0.693 15.546 3.757 4.088 0.000 0 0 Table 3(b) Descripive Saisics (Panel) ln(real EXPORT) ln(real GDPi) ln(1+tariff) Observaions 7775 7775 7775 Mean 3.149 19.787 0.011 Sd.Dev. 3.131 1.448 0.036 Max 13.160 23.263 0.182 Min -1.068 15.544 0.000 Table 3(a) and 3(b) presens he descripive saisics of he variables used in cross-secion analysis and panel analysis respecively. We exclude he zero from rade daa in cross-secion analysis. I is clear from he rade enlargemen of he secor of phoovolaic cells from 2000 o 2004 ha he observaions of non-zero rade are increasing year by year. In panel analysis, we include he rade of he counry-pairs wih a leas one non-zero rade in he five years daa. Ⅳ. Empirical Resuls This secion presens esimaion resuls of he graviy analysis for he secor of phoovolaic cells. In Ⅳ-1, he resuls of cross-secion analysis are presened and in Ⅳ -2, hose of panel analysis are repored. Ⅳ-1. Esimaion Resuls for Cross-Secion Analysis Table 4 shows he resuls in he cross-secion analysis for he esimaion of 10
equaion (6) using OLS for he secor of phoovolaic cells. In his able, four cases are shown wih differen specificaions, for each of five years. If we focus on case 4 whose specificaion includes all he explanaory variables excep he difference of GDP per capia beween counry pairs, all he coefficiens are saisically significan a he 1 % level excep ha of bilaeral ariff raes in 2002, which is significan a he 5% level. The main findings of cross-secion analysis of each of five years can be summarized as follows. A firs, he coefficien esimaes for GDPs in boh exporing and imporing counries are posiive and saisically significan a he 1 % level, and he former is bigger han he laer in all he cases for all he years. The coefficiens of GDPs in exporing counries ake values in he range of 0.73-0.82, and hose in imporing counries are in he range of 0.64-0.71, suggesing ha an increase in a counry s GDP will lead o a less-han proporional increase in is impors and expors. The coefficiens of he difference beween wo counries GDP per capia are always negaive and significan a he 1 % level in he more recen years of 2003 and 2004, and significan a he 5 % level in 2000 and 2002. Also, he absolue values of hese coefficiens are bigger in 2003 and 2004 han hose of he earlier years, wih 0.168 and 0.174 respecively in case 3. This suggess ha rade is acceleraed beween counries wih similar paerns of facor endowmen, reflecing ha rade in he secor of phoovolaic cells can be explained by a new rade heory based on monopolisic compeiion wih differeniaed goods, and conradics he predicion of he radiional Heckscher-Ohlin heory of rade. Furhermore, his endency is srenghened in 2003 and 2004. The coefficiens for disance are negaive as expeced and saisically significan in all cases for all years. If we focus on he esimaes in he case 4, he absolue values of hese coefficiens are relaively small in 2000 wih 0.49, bu afer 2001 hose absolue values ended o increase, wih nearly 0.54 in 2001 and 2002, 0.57 in 2003, suggesing ha he counry pairs wih a shorer disance ended o increase rade more han in 2000. In 2004, however, he effec of he disance on rade decreased slighly o 0.52 in case 4. The coefficiens for bilaeral ariff raes are negaive as expeced and saisically significan a he 1% level excep hose in 2002, which are negaive and saisically significan a he 5% level. Also, he absolue value of he coefficien in 2002 is smalles wih 4.245 in case 4, bu his increased o 6.738 and 7.399 in 2003 and 2004 respecively. Hayakawa (2011) shows ha he absolue value of he esimae of he coefficien of bilaeral ariff rae in he analysis of manufacuring rade is nearly 4 which can be 11
considered a reasonable size. 3 Since we noice ha he adjused R-squared is higher in case 2 as compared o case 1 for all years, he imporance of bilaeral ariff rae in he specificaion of he graviy equaion for he secor of phoovolaic cells can be clarified. The coefficien of he dummy variable for common official language in he phoovolaic cells is saisically significan in all cases of all years. If we focus on case 4, he coefficien value increased unil 2002, wih 1.063 in 2000, 1.152 in 2001, and 1.323 in 2002. On he oher hand, he coefficien reduced afer 2003, wih 0.806 in 2003 and 0.698 in 2004. Conrary o he resuls for he dummy variable for common official language, he effecs of he dummy variable for he APEC bloc are higher in 2003 and 2004 wih 1.495 and 1.291 respecively, as compared o his effec in 2001 and 2002 wih 1.040 and 1.049 respecively, in case 4. The coefficien esimae for 2004, 1.291 suggess ha on average, wo APEC members rade phoovolaic cells abou 3.3 imes as much as oher counry pairs. Surprisingly, his esimae of 1.29 is he same value esimaed by Frankel (1997) for he APEC bloc dummy variable in he graviy model for 1992, including he effecs of prospecive rading blocs shown in Table 4-3 of Frankel (1997). 3 Theoreically, he coefficien of bilaeral ariff rae in he graviy equaion can be shown as 1-σ, where σ is elasiciy of subsiuion among varieies. Concerning he value of σ, Hayakawa (2011) refers he value in he range of 5 o 10 concluded by Anderson and Wincoop (2004). 12
Table4. Esimaion Resuls (Cross-Secion Analysis) Equaion(2) Equaion(2) Equaion(2) 2000 2001 2002 case1 case2 case3 case4 case1 case2 case3 case4 case1 case2 case3 case4 Consan -18.048 *** -18.206 *** -17.284 *** -18.316 *** -19.090 *** -19.321 *** -18.780 *** -19.521 *** -20.697 *** -20.678 *** -19.838 *** -20.705 *** (-12.484) (-12.577) (-12.338) (-14.037) (-13.673) (-13.787) (-13.886) (-15.315) (-14.746) (-14.688) (-14.528) (-15.950) ln(gdpi) 0.756 *** 0.756 *** 0.739 *** 0.730 *** 0.807 *** 0.809 *** 0.809 *** 0.803 *** 0.824 *** 0.823 *** 0.808 *** 0.804 *** (14.665) (14.728) (15.047) (14.776) (16.145) (16.213) (16.786) (16.583) (16.322) (16.244) (16.547) (16.334) ln(gdpj) 0.668 *** 0.665 *** 0.638 *** 0.638 *** 0.662 *** 0.666 *** 0.649 *** 0.647 *** 0.738 *** 0.732 *** 0.700 *** 0.700 *** (14.158) (14.117) (14.139) (14.018) (14.267) (14.354) (14.469) (14.371) (16.153) (15.997) (15.627) (15.548) ln(gdppc) -0.158 *** -0.146 ** -0.143 ** -0.121 ** -0.110 ** -0.107 * -0.145 *** -0.137 *** -0.114 ** (-2.624) (-2.396) (-2.412) (-2.024) (-1.821) (-1.834) (-2.758) (-2.574) (-2.277) ln(dist) -0.450 *** -0.431 *** -0.484 *** -0.494 *** -0.468 *** -0.456 *** -0.529 *** -0.540 *** -0.483 *** -0.474 *** -0.532 *** -0.544 *** (-7.183) (-6.858) (-7.864) (-8.035) (-7.678) (-7.488) (-8.760) (-8.993) (-8.236) (-8.098) (-9.136) (-9.430) ln(1+tariff) -5.842 *** -5.290 *** -5.591 *** -6.063 *** -5.386 *** -5.584 *** -4.496 ** -3.971 ** -4.245 ** (-3.278) (-2.896) (-3.078) (-3.321) (-2.896) (-3.099) (-2.373) (-2.097) (-2.249) CL 1.004 *** 1.063 *** 1.125 *** 1.152 *** 1.287 *** 1.323 *** (4.666) (5.070) (5.673) (5.851) (6.504) (6.724) AP 1.268 *** 1.227 *** 1.058 *** 1.040 *** 1.062 *** 1.049 *** (5.649) (5.523) (5.541) (5.486) (5.507) (5.468) Adjused R 2 0.274 0.281 0.345 0.341 0.292 0.298 0.345 0.355 0.297 0.300 0.361 0.359 Observaions 899 899 899 899 951 951 951 951 1027 1027 1027 1027 Equaion(2) Equaion(2) 2003 2004 case1 case2 case3 case4 case1 case2 case3 case4 Consan -17.871 *** -18.156 *** -17.216 *** -18.853 *** -19.887 *** -19.991 *** -19.239 *** -20.882 *** (-11.778) (-11.972) (-11.842) (-13.969) (-13.454) (-13.618) (-13.358) (-15.463) ln(gdpi) 0.763 *** 0.765 *** 0.753 *** 0.757 *** 0.826 *** 0.825 *** 0.808 *** 0.809 *** (14.393) (14.472) (14.684) (14.642) (16.153) (16.210) (16.173) (16.036) ln(gdpj) 0.681 *** 0.688 *** 0.652 *** 0.662 *** 0.708 *** 0.712 *** 0.689 *** 0.696 *** (14.668) (14.835) (14.336) (14.556) (15.862) (15.979) (15.614) (15.685) ln(gdppc) -0.202 *** -0.191 *** -0.168 *** -0.194 *** -0.192 *** -0.174 *** (-3.519) (-3.344) (-3.085) (-3.326) (-3.300) (-3.027) ln(dist) -0.484 *** -0.476 *** -0.547 *** -0.574 *** -0.483 *** -0.443 *** -0.496 *** -0.517 *** (-8.091) (-7.969) (-9.271) (-9.833) (-7.793) (-7.607) (-8.495) (-8.903) ln(1+tariff) -6.927 *** -6.474 *** -6.738 *** -7.616 *** -7.343 *** -7.399 *** (-4.106) (-3.670) (-3.811) (-4.521) (-4.198) (-4.169) CL 0.756 *** 0.806 *** 0.673 *** 0.698 *** (3.558) (3.863) (3.293) (3.419) AP 1.501 *** 1.495 *** 1.285 *** 1.291 *** (7.555) (7.542) (6.618) (6.685) Adjused R 2 0.276 0.283 0.345 0.341 0.281 0.289 0.332 0.327 Observaions 1075 1075 1075 1075 1126 1126 1126 1126 Noe: values are in parenheses. ***: significan a he 1% level; ** a 5 % level; * a 10 % level. 13
Ⅳ-2. Esimaion Resuls for Panel Analysis Table 5 repors he esimaion resuls of he panel analysis of he graviy equaion for he rade of phoovolaic cells among 51 counries. The regression resuls for he pooled cross-secion model esimaed according o equaion (8) are shown in he firs column, and hose for he fixed effecs model according o equaion (7) are shown in he second column. In panel analysis, he variables of expors and GDP are in real values insead of nominal values used in he cross-secion model. For he resuls of he pooled cross-secion model, he signs of all he coefficiens of he included variables are as expeced and are saisically significan, excep he ime dummy of 2001. According o he esimaes of he pooled cross-secion model, he following major findings can be clarified. Firsly, an increase in a counry s real GDP leads o an approximaely proporional increase in is real expors, and less han proporional increase in is real impors. Secondly, he esimaed coefficien for bilaeral ariff raes is -6.304 showing a srong effec. Thirdly, he esimaed coefficien on he log of disance is -0.693, which means ha when he disance beween wo counries is increased by 1.0 percen, rade beween hem falls by abou 0.69 percen. Fourhly, regarding he esimae of he coefficiens on he dummy variables, ha for a common language is 1.28, which implies ha wo counries sharing an official linguisic link end o rade 3.6 imes as much as wo oherwise similar counries and he esimaed coefficien on he dummy for he APEC bloc is 1.33, which implies he wo of he APEC counries rade 3.8 imes as much as wo oherwise similar counries. 4 Accordingly, he effecs of common official language and APEC bloc can be hough o be quie similar. Lasly, he year dummies are saisically significan excep in 2001, and his implies he common rend oward greaer real rading volumes, independen of he sizes of he economies. As Cheng and Wall (2005) define hese year dummies as an indicaor of he exen of globalizaion, we find ha globalizaion is an imporan facor for rade growh of phoovolaic cells. Regarding he esimaion resuls for he fixed effecs model of equaion (7), he signs of he coefficien of he exporing counry s real GDP and bilaeral ariff raes are as expeced and are saisically significan. The negaive coefficien on he imporing counry s real GDP is no as expeced, alhough i is no saisically differen from zero. In he fixed effecs model, by allowing for rading-pair heerogeneiy, he esimaed coefficien for bilaeral ariff raes, in absolue value, is smaller han ha in he pooled cross-secion model, represening he value of 4.97 consisen wih he resuls in Hayakawa (2011). On he oher hand, he esimaed coefficien of real GDP of exporing 4 The exponenial of 1.28 is equal o 3.60, and he exponenial of 1.33 is equal o 3.78. 14
counry and he esimaed role of globalizaion in he fixed effecs model are bigger han in he pooled cross-secion model. An increase in a counry s real GDP by 1% leads o increasing is expors by 1.5%, and globalizaion has increased he real volume of rade by 63% beween 2000 and 2004 according o he definiion of Cheng and Wall (2005). Table5. Esimaion Resuls (Panel Analysis) Pooled Cross-Secion Fixed Effecs 2000-2004 2000-2004 Consan -25.790 *** -24.662 *** (-43.949) (-4.002) ln(gdpi) 1.005 *** 1.558 *** (49.967) (-5.898) ln(gdpj) 0.727 *** -0.165 (37.594) (-1.423) ln(dist) -0.693 *** (-26.132) ln(1+tariff) -6.304 *** -4.973 *** CL 1.281 *** (-36.604) (-3.659) (13.827) AP 1.331 *** (16.344) Time Dummies 2001 0.117 0.132 *** (1.307) (13.145) 2002 0.214 ** 0.264 *** (2.388) (11.881) 2003 0.341 *** 0.380 *** (3.799) (11.522) 2004 0.569 *** 0.627 *** (6.332) (13.098) Counry Pair YES Adjused R 2 0.363 0.7544 Observaions 7775 7775 Noe: values are in parenheses. ***: significan a he 1% level; ** a 5 % level; * a 10 % level. 15
Ⅴ. Conclusions This paper invesigaed he imporance of reducion of bilaeral ariff raes on rade of one of he represenaive environmenal indusries for renewable energy, he secor of phoovolaic cells, in order o assess effors for ariff reducion for environmenal goods among APEC counries and he WTO. We clarify hree poenial conribuions of our sudy. Firsly, we have sored he bilaeral rade flow daa in he secor of phoovolaic cells using deailed daa among 51 counries, as well as collecing he bilaeral ariff raes of phoovolaic cells from WTO daa. Secondly, we have shown ha he graviy model is very useful in esimaing rade flows in he phoovolaic cells secor by using cross-secion daa, including difference beween per capia income of exporing and imporing counries, as well as bilaeral ariff raes. We have found ha bilaeral ariff raes sysemaically affeced he rade flows in all he years beween 2000 and 2004, and he rade flows of phoovolaic cells can be explained by he new rade heory based on monopolisic compeiion wih differeniaed goods, since he counry pairs wih less difference of GDP per capia rade more acively. Thirdly, using he fixed effec model wih panel daa by allowing rading-pair heerogeneiy, we showed he srong sysemaic effec of bilaeral ariff raes on he rade flows, and also he rade increase in he phoovolaic cells secor resuling from he globalizaion effec. Alhough many counries abolished ariffs on phoovolaic cells in he end of he 1990s, he resuls of our analysis shows ha he remaining ariffs in BRICs counries and Chile were clear rade impedimens in he firs half of he 2000s in his secor. Our analysis suggess ha furher ariff reducions are imporan o acivae rade flows of environmenal goods among he covered counries, wheher developed or newly indusrialized counries. The curren effors of ariff reducion of various kinds of environmenal goods wih differen environmenal purposes in APEC and WTO are imporan for rade increase and environmenal proecion hrough worldwide diffusion of hose goods. Furher research esimaing he deerminans of rade flows of various environmenal goods is required o consolidae he effecs of bilaeral ariff raes in hose goods. In paricular, a graviy analysis of environmenal goods having similar purposes as phoovolaic cells for elecriciy generaion from renewable resources, such as elecric generaing ses for wind-powered (HS850231), would be necessary for comparing resuls wih hose derived from he analysis for phoovolaic cells. 16
References Anderson, James and Eric van Wincoop (2003). "Graviy wih Gravias: A Soluion o he Border Puzzle," American Economic Review, 93(1), 170-192. Anderson, James and Eric van Wincoop (2004). "Trade Coss," Journal of Economic Lieraure, 42, Sepember, 691-751. Ando, Misuyo and Fukunari Kimura (2013). Producion Linkage of Asia and Europe via Cenral and Easern Europe, Journal of Economic Inegraion, 28(2), 204-240. APEC (Asia-Pacific Economic Cooperaion) (2012). Vladivosok Declaraion- Inegrae o Grow, Innovae o Prosper. Vladivosok, Russia. Baier, Sco and Jeffrey H.Bergsrand (2001). The Growh of World Trade: Tariffs, Transpor Coss, and Income Similariy, Journal of Inernaional Economics 53(1), 1-27. Baier, Sco and Jeffrey H.Bergsrand (2007). Do Free Trade Agreemens Acually Increase Members Inernaional Trade? Journal of Inernaional Economics 71(1), 72-95. Baldwin, Richard and Daria Taglioni (2011). Graviy Chains: Esimaing Bilaeral Trade Flows when Pars and Componens Trade is Imporan, NBER Working Paper Series 16672. Bayoumi, Tamin and Barry Eichengreen (1997). Is Regionalism Simply a Diversion? Evidence from he Evoluion of he EC and EFTA, In: Io, Takaoshi, Kruger, Anne O(Eds.), Regionalism vs. Mulilaeral Arrangemens. The Universiy of Chicago Press, Chicago. Cheng, I.-Hui, and Howard J.Wall (2005). Conrolling for Heerogeneiy in Graviy Models of Trade, Federal Reserve Bank of S.Louis Review, January/February, 49-63. Bergsrand, Jeffrey H. and Peer Egger (2010). A General Equilibrium Theory for Esimaing Graviy Equaions of Bilaeral FDI, Final Goods Trade and Inermediae 17
Goods Trade, In Peer A.G. van Bergek and Seven Brackman eds., The Graviy Model in Inernaional Trade: Advances and Applicaions, New York: Cambridge Universiy Press. 29-70. Bergsrand, Jeffrey H. and Peer Egger (2010). Graviy Equaions and Economic Fricions in he World Economy. Universiy of Nore Dame Working Papers. Feensra, Rober C. (2004). Advanced Inernaional Trade, Theory and Evidence. Princeon: Princeon Universiy Press. Frankel, Jeffrey H. (1997). Regional Trading Blocs in he World Economic Sysem. Washingon, DC, USA: Insiue for Inernaional Economics. Frankel, Jeffrey H., Erneso Sein and Shang-Jin Wei (1998) Coninenal Trading Blocs: Are They Naural or Supernaural? In Jeffrey H. Frankel ed., The Regionalizaion of he World Economy. Chicago: The Universiy of Chicago Press. Hayakawa, Kazunobu (2011). How Serious Is he Omission of Bilaeral Tariff Raes in Graviy? IDE Discussion Paper,No.311, Ocober. Ishida, Osamu (2011). Globalizaion and Trade Srucure (In Japanese), Bunshindo. Masumura, Asuko (2014a) Srucure and Characerisics of Japanese Trade of Environmenal Goods: Graviy Analysis Focusing on Inra-Indusry Trade and FTA effecs (In Japanese), Boueki o Kanzei (Trade Journal): Japan Tariff Associaion 62(7), 62-77. Masumura, Asuko (2014b) Deerminans of Trade Neworks of Environmenal Goods, Focusing on he Effecs of Regionalizaion and Fragmenaion (In Japanese), Paper for he 73h Annual Meeing of he Japan Sociey of Inernaional Economics (JSIE). Masumura, Asuko (2015) How are he Trade Neworks for Environmenal Goods Consruced? Analysis for Deerminans of Environmenal Goods by HS Group (In Japanese), Boueki o Kanzei (Trade Journal): Japan Tariff Associaion 63(9), 20-29. Rauch, J.E.(1999) Neworks versus Markes in Inernaional Trade, Journal of 18
Inernaional Economics (48) 7-35. World Trade Organizaion (2011) Universe of Environmenal Goods: Official HS Descripions, Commiee on Trade and Environmen Special Session TN/TE/20. 19