The Growth and Determinants of Vertical Trade in Korea

Transcription

1 March 2008 The Growth and Determnants of Vertcal Trade n Korea Young Kyung Suh* The vews expressed heren are those of the author and do not necessarly reflect the offcal vews of the Bank of Korea. When reportng or ctng t, the author's name should always be explctly stated. Insttute for Monetary and Economc Research, Bank of Korea Insttute for Monetary and Economc Research The Bank of Korea

2 The Growth and Determnants of Vertcal Trade n Korea Young Kyung Suh* The vews expressed heren are those of the author and do not necessarly reflect the offcal vews of the Bank of Korea, When reportng or ctng t, the author s name should always be explctly stated. * Economst, Internatonal Economcs Team, Insttute for Monetary and Economc Research, Bank of Korea ( , yksuh@bok.or.kr) I would lke to thank Ouk Heon Song, Seung Je Hong, Bong Geul Chun and Hyejoon Im and semnar partcpants at the Bank of Korea for ther helpful comments. Any errors appearng here are mne.

3 Contents I. Introducton II. Vertcal Trade n Korea Measurement of Vertcal Trade 2.2 Vertcal Specalzaton n Korea s Trade III. The Determnants of Vertcal Trade An Overvew of the Lterature 3.2 Theoretcal Model 3.3 Emprcal Estmaton IV. Concluson References Appendxes <Abstract n Korean>

4 The Growth and Determnants of Vertcal Trade n Korea Ths paper measures the volume of vertcal trade n Korea and nvestgates ts determnants. The contrbutons of ths paper are as follow. Frst, the measurement method n ths paper s related to Hummels et al. (2001), but t mproves upon ther method n that t uses nformaton on ntermedate goods trade as well as the nput-output tables of tradng countres. Second, ths paper derves a functonal form for the determnants of vertcal trade from a monopolstc competton model of trade and multnatonal producton. The emprcal results show that Korea s trade has been becomng more vertcally specalzed over last ten years. In partcular, vertcal exports to Chna have grown rapdly snce the early 2000 s, and thus the surplus on vertcal trade has wdened dramatcally n recent years. The emprcal estmaton wth 32 tradng partners durng the perod of provdes evdences that the determnants of Korea's vertcal trade are broadly consstent wth the theoretcal predctons: t has been affected by relatve labor costs and productvty, as well as by the gravty-related varables such as trade cost and market accessblty. Korea s vertcal trade was also ndrectly affected by relatve fxed costs through the change n the stock of FDI. Based on these results, t can be sad that, for the last ten years, the growth of Korea s trade has been largely attrbuted to the expanson of nternatonal producton fragmentaton accompaned wth ntermedate goods trade, and also that a monopolstc competton model of trade and producton accounts qute well for the pattern of vertcal trade n Korea. Keywords: Vertcal Specalzaton, Intermedates Trade, Producton Fragmentaton JEL Classfcaton: F14, F15, F23

5 I. Introducton In recent decades, the nternatonalzaton of producton has become more common as transportaton costs and trade barrers have decreased, and the busness envronment has been ncreasngly compettve. Partcularly when the producton process conssts of varous separable stages whch requre dfferent factor ntenstes, producton chans are slced thnner nto many stages n dfferent locatons where the producton cost s relatvely low. The producton of a fnshed good thus nvolves a partcpaton of many countres, specalzng n dfferent stages of vertcal producton. Internatonal trade s then ncreasngly domnated by vertcal trade, whch s defned as a trade n ntermedate goods that are part of an nternatonal producton network 1. Accordng to Hummels, Ish and Y (herenafter HIY, 2001), the growth rate of vertcal trade has outpaced that of tradtonal nternatonal trade flow and current vertcal trade conssts of around one-thrd of world trade. Increase of vertcal trade affects the way n whch trade flows and ther major 1 Hummels, Rapoport and Y (HRY, 1998) defned three condtons for vertcal trade to occur: ⑴ A good s produced n two or more sequental stages. ⑵ Two or more countres provde value-added durng the producton of the good. ⑶ At least one country must use nputs n ts stage of the producton process, and some of the resultng output must be exported. Under ths defnton, vertcal trade s smlar to processng trade, but dfferent from outsourcng or border-crossng trade. The cted paper provdes an example whch llustrates the dstncton between vertcal trade and outsourcng. Suppose a frm relocates producton of components to another country and mports these components from that country. If the frm completes the producton of the fnal goods and does not export, outsourcng occurs. If the country does, however, export the fnal goods, both outsourcng and vertcal trade occurs. Vertcal trade s also dfferent from border-crossng trade of goods that are merely shpments n transt, e.g., Chnese goods gong through Hong Kong's port on ther way to the thrd countres

6 determnants nteract. For example, cross-border producton sharng pushes specalzaton from the level of products to that of components. The tradtonal laws of comparatve advantage contnue to work here: labor-abundant countres would be expected to specalze n labor-ntensve components and assembly, whle captal- and skll-rch countres have comparatve advantages n captal- and skll-ntensve components and assembly (Arndt and Huemer, 2005). Cross-border producton sharng also alters the response of trade to changes n trade costs. The response of vertcal trade to trade costs s lkely very senstve due to the hgh ncdences of trade costs. HIY (2001) suggested that a relatvely small reducton n trade barrers snce the md 1980s was able to brng about a rapd growth n trade due to the convex response of vertcal trade to decreases n trade costs. On the other hand, there s the tradtonal debate as to whether trade and foregn drect nvestment (FDI) are complements or substtutes. In a global stuaton characterzed by growng producton fragmentaton, t s plausble that we should be able to observe smultaneously an ncrease n FDI and an ncrease n trade (Carpenter, 2005). As vertcal trade has expanded, the queston of how to analyze ths phenomenon has come to be an mportant subject n the recent lterature on nternatonal trade. However, because data on the producton process and drecton of trade flow for every producton stage are mpossble to obtan, a number of studes have been undertaken to measure the vertcal trade ndrectly ether by one-way trade of parts and components (Kmura et al. 2005, Athukorala and Yamashta 2006) or by the ndex of ntra-ndustry trade (Fukao et al. 2003, Okubo 2004, Lee and Byun 2005). Furthermore, these studes lean heavly on the gravty model and could not provde an explanaton consstent wth theory for the relaton between the explanatory varables and the vertcal trade

7 The am of ths paper s to nvestgate the pattern and determnants of vertcal trade n Korea, where the expanson of export related to producton sharng s far faster than n other countres. The contrbutons of ths paper are as follow. Frst, t measures vertcal trade usng data on trade and producton of tradng countres. Ths method s related to HIY (2001), but t mproves upon ther method n that t uses nformaton on the ntermedate goods trade as well as nput-output tables. In ths way, we can estmate vertcal trade not only by ndustry but also by tradng partner. Second, ths research provdes an explct functonal form for the determnants of vertcal trade, whch s derved from a monopolstc competton model of trade and multnatonal producton. It shows that vertcal trade s affected by the comparatve advantage n labor costs and productvty as well as the geographcal advantage measured by trade costs and market accessblty. Relatve fxed costs also play an mportant role n determnng vertcal trade va the change n the stock of FDI. The emprcal results presented below show that Korea s trade has been becomng more vertcally specalzed over last ten years. In partcular, vertcal exports have grown rapdly snce the early 2000's, and thus the surplus on vertcal trade has wdened dramatcally. The swft ncrease n vertcal exports s largely attrbutable to the expanson of vertcal exports to Chna, whch emerged as a global producton hub by attractng a large amount of foregn drect nvestment. On the other hand, the share of vertcal mports n total mports has dropped slghtly due to a slowdown of FDI nflows and ntermedates mports, especally from the US and Japan. The estmaton results for Korea's trade data durng the 1995 to 2005 perod show that ncreases n outward FDI, relatve wage and productvty played mportant roles n expandng vertcal exports. On the other hand, world demand and nward FDI were the man factors that determned vertcal mports. In contrast to the case of vertcal exports, t - 3 -

8 appears that relatve wage had no sgnfcant effect on changes n vertcal mports. The paper conssts of the followng sectons. Secton 2 descrbes a new method to measure the volume of vertcal trade and sets out the recent characterstcs of vertcal trade n Korea. In Secton 3, we buld a smple model to examne the determnants of vertcal trade, and provde the estmaton results. Some conclusons and possble polcy mplcatons are put forward n Secton 4. II. Vertcal Trade n Korea 1. Measurement of Vertcal Trade A. The HIY measure Vertcal trade (VT) refers to the trade n ntermedate goods that are part of an nternatonal producton networks (HRY, 1998). Fgure 1 llustrates an example of vertcal trade nvolvng three countres. Country 1 produces ntermedate goods and exports them to country 2. Country1 combnes the mported ntermedate goods wth captal, labor, and domestc ntermedates to produce fnal goods. Fnally, country 2 exports some of the fnal goods to country 3. Vertcal trade occurs only when there are both mported ntermedate goods and exports

9 Fgure1. Example of Vertcal Trade and Stages of Producton <Stage1> <Stage2> Source: Y (2003, p. 57). To construct emprcal measures of vertcally fragmented trade, deally we would use data on the producton process and drecton of trade flow for every stage of each good traded. Unfortunately, these data are mpossble to obtan. As an alternatve, prevous studes estmate the volume of VT usng varous measures such as ntermedate goods trade, ntra-frm trade or ntra-ndustry trade 2. The most popular and easy way to measure VT s to capture the one-way trade of ntermedates. The man dsadvantage n connecton wth ths method, however, s that the mported ntermedates data may nclude amounts used for the producton of domestc goods 3. 2 For ntermedate goods trade and ntra-ndustry trade n Korea, see Appendx A and B. 3 In addton, classfcaton schemes for ntermedate goods can be somewhat arbtrary. For example, tres, flour and motherboards are ntermedates when they are mported and used to make cars, bread, and computers, respectvely, but they are fnal goods when they are purchased by households (HIY 2001, p.4)

10 More precse measurement of fragmentaton trade requres the nput-output tables of tradng natons whch provde data on mported ntermedates and on each ndustry s output exported or used domestcally. The general way of computng vertcal trade wth I/O tables was proposed by HIY (2001) as follows: VT = A M I A D 1 (1) k [ ] k x where VT k s vertcal trade n ndustry k of country, M A s the mported nput matrx, [ ] 1 I s the nverse matrx coeffcents to capture the mported nput A D used n mult-stages for the producton of export goods, and x k s the export share of ndustry k (= X / X ). k The natonal I/O table, however, only provdes the amount of vertcal trade for the country overall. It cannot be used to calculate blateral vertcal trade because transnatonal I/O tables are not avalable for most pars of countres 4. B. A new measure of vertcal trade: HIY2 I mprove on HIY s method by usng the mported ntermedate nputs nstead of the mported nput matrx ( M A ) on the nput-output tables. The blateral vertcal export (VE) and vertcal mport (VI) can be calculated as a multplcaton of traded ntermedates by export share of each ndustry's output n the I/O transacton table. 4 Recognzng the strong nterdependence of the Asa-Pacfc regon, I.D.E and JETRO n Japan jontly publshed the Asan nternatonal nput-output table n 1995 and They have nevertheless only released lmted tables manly focused on the fnal I/O tables of ten Asan countres, and the transnatonal I/O tables between subgroup of countres are avalable for only a few cases

11 (2) VE j = IE (3) VI j = IM k k jk jk B B jk k where IE jk s the exported ntermedate from country to country j, B jk s a share D 1 I A X / of country j's producton that s exported ( [ j ] jk j = X ), IM jk s exported ntermedates from country j to country, B k s a share of country 's producton that D 1 I A X / = X ). s exported ( [ ] k Producton nducement coeffcents and export shares of Korea are avalable from I/O transacton table n 1995, 1998, 2000 and 2003 and for the between years, averaged ndustry shares are appled. Korea s trade data n parts and components by tradng partner and by ndustry are come from the database of Mnstry of Commerce, Industry and Energy of Korea. It s more dffcult to measure vertcal exports than vertcal mports, because the measurement requres blateral export data matchng the nput-output tables of foregn countres. I use the OECD STAN nput-output database (2000) and I/O tables from Chna (2000), Tawan (2001), Russa (2000) and Indonesa (2000) 5. Yearly trends are acqured by adjustng the ndustry share of the 2000 s wth the yearly ncreases of total export share n real GDP. Intermedates export data are also acqured from the database of the Mnstry of Commerce, Industry and Energy of Korea. 5 For the countres whose I/O tables are not avalable, the share of total exports n real GDP s used

12 Table1. Comparson of Measurement Methods for Vertcal Trade Intermedates Trade Trade n Parts and Components Producton and Export Informaton n I/O Tables Coverage Trade by country and ndustry HIY Trade by ndustry HIY2 Trade by country and ndustry 2. Vertcal Specalzaton n Korea s Trade A. Rapd Growth of Vertcal Trade for the Last Decade The HIY2 method descrbed above s employed to estmate the volume of vertcal trade n Korea 6. Table2 llustrates that vertcal trade n Korea has grown rapdly over the past ten years and played a key role n drvng overall trade growth. The second row of the table shows the value of vertcal exports and mports of Korea for the selected years from 1995 to The value of vertcal exports almost quntupled 6 The table below shows that there are sgnfcant dfferences n the volume of vertcal trade generated by the three methods. Snce the HIY and HIY2 methods defne vertcal trade as the subset of ntermedates trade whch s used to product exports, they yeld smaller values than the method of ntermedates trade. The amounts of vertcal trade based on HIY are bgger that those based on HIY2 because the ntermedate nput n I/O table s larger than mport of parts and components. Value (100ml.$) Share n total Trade (%) Comparng Vertcal Trade Calculated by the Three Methods (2006) Intermedate Goods Trade 1,490.9 <45.8> Exports HIY 1,210.2 <37.2> HIY2 1,129.5 <34.7> Imports Intermedate Goods Trade HIY 1, <36.8> <31.4> HIY <29.1> - 8 -

13 from 1995 to represent about 35% of total exports as of The value of vertcal mports has more than trpled snce 1995 and presently makes up some 29% of totals mport. As the growth rate of vertcal exports exceeded that of vertcal mports, the surplus related to vertcal trade has wdened dramatcally n recent years and represented about one and half tmes the overall trade surplus as of Table 2 also provdes the sectoral composton of vertcal trade n Korea. It shows that vertcal exports occur mostly n two key sectors: machnery and electroncs. The respectve shares of machnery and electroncs n vertcal exports surged from 9% and 54% n 1995 to 15% and 62% n In these sectors, the surplus on vertcal trade account also has wdened rapdly n recent years. Table2. Trend of Vertcal Trade n Korea (100ml.$,%) Vertcal Export Vertcal Import Trade Balance Total Export s(a) 1, , , , , , Vertcal Trade (B) B/A, % <17.5> <29.6> <34.7> <16.9> <29.4> <29.1> <32.4> <140.9> Share, % Textles Chemcals Rubber Non-metallc mnerals Metals Machnery Electroncs Vehcles

14 B. Sharp Increase of Vertcal Exports to Chna Table 3 llustrates the regonal composton of vertcal trade n Korea. It s notable that vertcal trade wth Chna has expanded very rapdly, wth ts share of exports rsng from 13% n 2000 to 41% n 2006 and that of mports from 9% to 20% durng the same perod. On the other hand, the shares of the U.S, Japan and other developng countres n Asa have been mostly decreased. Ths mples that the recent ncrease n Korea s vertcal exports was manly attrbutable to the expanson of vertcal exports to Chna. It also shows that the ncrease of vertcal mports from Chna was also larger. Table3. Trend of Vertcal Trade by Country (100ml.$,%) Vertcal Exports Vertcal Imports Trade Balance Total Asa Developng (Chna) (Hong Kong) (Tawan) (ASEAN) U.S Japan EU Other countres Through ts emergence as a regonal producton hub, Chna has become the most mportant destnaton for Korea ntermedates exports and FDI. The lberalzaton of trade and nvestment n Chna, prompted by jonng the WTO n 2001, has led to a surge of foregn drect nvestment nto Chna and changes n Chna s role wthn regonal and global producton chans. Korean frms have been establshng a manufacturng base n Chna to take advantage of ts low producton costs and preferental treatment for processng trade. Snce 2002, Chna has replaced the US as

15 Korea s largest destnaton for overseas nvestment, accountng for 40 percent of Korea s manufacturng outward FDI n Fgure2. Vertcal Exports and FDI to Chna (%) (Chna's Share n Korea's Export) Total Export Intermedates Export Vertcal Export (ml.$) (FDI to Chna) Annual FDI to Chna(left) Share of Chna n Manufacturng FDI (rght) (%) Equaton (4) dvdes the ncrease n total exports (A) nto that of vertcal trade (B) and other exports. Furthermore, equaton (5) decomposes the change n vertcal trade nto two stages: a change n ntermedates trade (C) and that n the producton (export) of fnal goods (D). Table 4 shows that the ncrease n vertcal exports explans more than 80% of the rse n Korea s export/gdp rato durng 2002 and It also provdes the results of the breakdown of the changes n vertcal trade wth Chna, the U.S., and Japan nto major sectors. The rse n vertcal exports to Chna s largely attrbutable to the ncrease n Chna's producton and export of fnal goods, whle the rse n those to the U.S. and Japan s caused by the ncrease n ntermedates exports themselves. (4) ( EX / Y) = ( VE / Y ) + ( EX VE ) Y / (A) (B) (5) VE = IE B + IEt B D 0, where IE: ntermedates export, B: [ I A ] X k (C) (D)

16 Table4. Decomposton of Increase n Korea s Export (Contrbuton to the ncrease n exports/gdp) Increase of Export/GDP (A) Contrbuton Rate (%) (Decomposton of ncrease n vertcal export) Increase (95~06) (100ml.$) C (%) D (%) Machnery ~06 02~06 95~06 02~06 Chna Total Export Electroncs Intermedates Machnery US Vertcal Export Electroncs (Chna) Machnery Japan (ASEAN 1) ) Electroncs Note: 1) ncludng Hong Kong, Tawan As a result of the rapd ncrease of ntermedates exports to Chna, the share of Korea n total Chnese mports of ntermedates also rose substantally, reachng a level slghtly lower than those of Japan and Tawan n Korea s share n the US market, however, remaned flat durng , whle that of Chna rose sharply over the same perod. Fgure3. Major Countres' Share n Intermedates Imports of Chna and the US (Chna s Intermedates Imports by Country) (US s Intermedates Imports by Country) 20 (%) Chna Canada Mexco Japan 8 4 Korea

17 III. The Determnants of Vertcal trade 1. An Overvew of the Lterature The early lterature on vertcal trade mostly focused on vertcal ntra-ndustry trade and explaned a rse n vertcal trade by a wdenng of the gap n comparatve advantage 7. Fragmentaton-based vertcal trade has only recently receved attenton. HIY (2001) was the frst research paper to develop a comprehensve framework to analyze the phenomenon of vertcal specalzaton. Usng nput-output tables from the OECD and emergng market countres, the authors estmated that vertcal specalzaton accounts for up to the 30% of world exports. By developng an extenson of the Dornbusch-Fsher-Samuelson Rcardan trade model, they showed that even small reductons n tarffs and transport costs can lead to large trade growth due to the double ncdence of trade-costs n vertcal trade. By calbratng and smulatng a two-stage dynamc Rcardan trade model, Y (2003) showed that the model generates a nonlnear trade response to tarff reductons and t can explan over 50 percent of the growth of trade. He also showed that the welfare gans to tarff reductons n the vertcal model are hgh, compared to the standard one-stage model. Carpenter (2005) emphaszed that ntermedates trade occurs n combnaton wth 7 Falvey (1981) was the frst researcher who nvestgated vertcal ntra-ndustry trade n fnal goods. He demonstrated that hgher qualty goods are exported by a country that has a hgher captal-labor-rato technology, whle lower qualty goods are exported by a low-technology country. Accordng to hm, VIIT s lkely to be actve when two countres have large dfferences n factor endowments and technology. Based on ths perspectve, Flam and Helpman (1987) and Grossman and Helpman (1991) presented the North-South VIIT model, n whch the North (hgh-ncome countres) export hgh qualty, hgh prce goods and the South (low-ncome countres) export low qualty, low prce goods

18 vertcal FDI. Incorporatng an ntermedate good and endogenous multnatonals nto a monopolstc competton trade model, she showed that relatve trade costs and fxed costs are the man factors whch determne the export and locaton decsons of vertcal frms. The soluton of the model showed that under certan combnaton of relatve trade costs and fxed costs, countres engage n both ntra-ndustry FDI and ntra-ndustry trade. She also developed the export functon of vertcal frms, but ths though only covers one-way trade n ntermedates. On the other hand, the recognton of the expanson of producton sharng has been far faster n East Asa than n any other regon and many studes have been undertaken to analyze the magntude and motvaton of producton fragmentaton wthn t. Ng and Yeats (1999) were the frst to analyze the nature of producton sharng n East Asa. By usng the trade data for parts and components (SITC revson 2), they found that nternatonal producton sharng n East Asa s expandng at a faster pace than n ether North Amerca or Europe. The extent of producton sharng s measured by the RCA ndex of parts and component trade. Fukao, Ishdo and Ito (2003) emprcally estmated Japan s blateral trade wth developng countres n Asa usng a gravty model. They proposed that resource-seekng or export orented FDI played the most sgnfcant role n the rapd ncrease n Japanese VIIT. Usng the trade data between 1991 and 1999, Lee and Byun (2005) demonstrated that varatons of the level of IIT n Korea are better explaned by vertcal IIT. The estmaton results on the coeffcents of ndependent varables supported the conventonal hypotheses derved from the prevous theores of the horzontal IIT and vertcal IIT. Usng a standard gravty model and the recent trade data (SITC revson 3) for 36 major countres, Athukorala and Yamashta (2006) also presented the estmaton result

19 that nternatonal producton fragmentaton s determned by the level of GDP, dstance, relatve wage dfferental and the formaton of RTA. Based on the gravty equaton approach and dsaggregated trade data, Kmura, Takahash, and Hayakawa (2007) compared the trade patterns between East Asa and Europe, partcularly trade n machnery parts and components. They argued that fragmentaton theory s sutable for explanng the mechancs of nternatonal producton networks n East Asa, n contrast wth the tradtonal horzontal product dfferentaton model ftted for ntra-ndustry trade n Europe. To sum up, explanng vertcal trade has been an mportant feld n the recent lterature, both theoretcal and emprcal. However, prevous studes of vertcal trade n Asa countres measured vertcal trade ether by one-way trade n parts and components (Ng and Yeats 1999, Kmura et al. 2005) or by the ndex of ntra-ndustry trade (Fukao et al. 2003, Athukorala and Yamashta 2005, Lee and Byun 2005), whch are not exactly consstent wth the defnton of vertcal trade. Furthermore, these studes reled on gravty equatons for emprcal estmatons and could not provde an explanaton consstent wth theory for the relaton between the explanatory varables and the vertcal trade. Ths paper adds to the lterature by dervng the functonal form of vertcal trade from a monopolstc competton framework and estmatng the equaton usng more refned measurements of vertcal trade, whch has been barely attempted so far. 2. Theoretcal Model To llustrate frms' decson to export vertcally, we buld on Carpenter (2005) whch

20 ncorporates ntermedates trade and vertcal FDI decsons nto a monopolstc competton framework. We extend the model to nclude asymmetry n producton costs and to predct the functonal form for the dependence of vertcal trade on comparatve advantage, geographcal advantage and FDI stock. (Frms' maxmzaton problem) The representatve consumer's behavor s modeled as a blateral verson of a CES (0<σ<1) utlty functon. The consumer n country j maxmzes the utlty functon subject to the expendture constrants: (6) max = N U j x j = 1 ρ ρ N s.t. pj x = 1 j = Y j where x j denotes the demand for a varety produced n country, p j s the c..f. prce, Y s the total expendture n country j and ρ 1/(1 σ ) s the demand j elastcty (ρ >1). Tme and ndustry subscrpts are omtted for notatonal smplcty. The constraned utlty maxmzaton yelds the mport demand functon: (7) x j ρ = E j pj where E j s the total demand of country j dvded by the prce of all other varetes 1 ρ avalable to the consumer n that market ( E Y P ). j j / j The representatve frm n country produces each varety for proft maxmzaton. N (8) max = ( p c ) xj F j=1-16 -

21 where p s producer's prce, c s a unt varable cost, and F s a fxed cost. Under monopolstc competton, prce s a markup over the unt varable cost. (9) p = / σ c Then, the operatng proft can be expressed as π = 1 c ρ j A j where ρ 1 A j (1 σ ) σ E j. (Frms' locaton decson) We now assume that producton of a good comprses two stages: the frst stage conssts of the producton of an ntermedate good I and the second stage, whch we denote as A, nvolves further processng of the ntermedate good. Each stage requres ts own plant. Frms can be classfed nto three groups by the locaton of producton. Frst, there can be natonal frms, whch produce all components and fnsh them n Home. We denote these frms as HH-types. Secondly, there may be frms that produce all components n Home and fnsh them at a second assembly plant n Foregn. We denote these frms as HF-types. Fnally, there can be horzontal multnatonal frms, whch produce whole products n foregn countres. We denote these frms as FF-types. The table below summarzes the characterstcs of each type of frm

22 Table5. Characterstcs of each type of frm Type of frm Producton Trade FDI HH-type tradtonal and natonal frms Produces all components and fnshes them n Home. Horzontal Specalzaton (Exports fnal good) None HF-type vertcally fragmented frms Produces all components n Home; fnshes them at an assembly plant n Foregn. Vertcal Specalzaton (Exports ntermedates) Vertcal FDI (Assembly Plant) FF-type horzontal multnatonals Produces components and fnshes them n foregn countres. Source: Modfcaton of Carpenter (2005, p.6) None Horzontal FDI (Assembly Plant and Components Factory) Gven that the producton costs and fxed costs are dfferent dependng on where producton s located, we assume that frms pck the locaton confguraton that maxmzes ther proft. In an ntal equlbrum, when domestc producton costs are very low, there wll not be an ncentve for a frm to move a plant abroad. In ths equlbrum, all frms are HH-types and there s no FDI. The number of HH-type frms s determned endogenously to yeld zero profts for each frm only covers fxed costs. Now magne rsng producton costs at home. As the relatve producton costs rse, the prce charged to foregn consumers for exported goods ncreases, foregn demand falls and frm's profts are lower. If the fxed cost of the assembly plant s relatvely small and there are pure profts to be made from operatng as a HF-type, then the frm moves the assembly plant abroad (vertcal FDI) and becomes a HF-type. If the producton costs rse steadly, t becomes optmal for the frm to move all the producton facltes ncludng the components factory to foregn countres and become a FF-type (horzontal FDI). A frm would want to shft from HF-type to

23 FF-type f the ncreased profts from the foregn sales are more than the ncrease n fxed costs. To sum up, the frm's locaton decson s determned manly by the relatve producton cost and dfferences n fxed costs between the home country and foregn countres. (Frms' Exports) The case we are nterested n s the HF-type frms that are engaged n both vertcal trade and FDI. To derve export functon of the vertcal frms, we add two asymmetres. Frst, we allow for trade costs between countres. If τ j and t j denote ceberg transport costs ( τ j >1) and ad valorem tarff ( t j >1), respectvely, the consumers n country j are charged p j for fnal goods from country as follows. (10) p = τ t c / σ j j j Now we consder the cost functon when the producton costs dffer dependng on where the stages of producton are located. The ntermedate I s manufactured by usng γ a unts of labor, where weght,γ, reflects the mportance of ntermedates n the producton of x. It takes one unt of I, plus further processng whch requres ( 1 γ )a unts of labor, to produce one unt of x. The wage n home ( w ) s dfferent from that n foregn ( w ). The varable cost for producton can be expressed as a weghted combnaton of the costs of each stage. (11) c = ( 1 γ ) a j w j + γa w j

24 Now we turn to the export functons of each type of frm. From equaton (7) and (10), the export functon of an HH-type frm can be defned as the foregn demand for fnal goods n country. What we are nterested n s the export functon of HF-type frms. The export functon of afflates n country j s expressed as the world demand for fnal goods n country j. Because the value of the ntermedate s defned as fracton γ of the fnal good x, the export functon of ntermedates s expressed as fracton γ of demand for fnal goods n country j. For FF-type frms, there s no trade from the home country. 1 ρ (12) HH-type (export of fnal goods): X j = m pj xj = ρm ( τ jtja w ) E j (13) HF-type (export of ntermedates): I j γx jh = γρmj ( τ jht (14) FF-type: No Export h jh c ) 1 ρ E h Where m and m j are the respectve numbers of afflates n country and n the partner country and E j and E h measure the effectve market sze of country j 1 ρ ( = Y P ) and of country h ( = Y j / j 1 ρ h / Ph ), respectvely. 3. Emprcal Estmaton A. Model and Varables From equaton (13), we can see that vertcal exports (VE) from country to country j are determned manly by the share of ntermedates n producton, the number of country s afflates n country j, the market access of country j and relatve

25 producton costs. The relatve fxed costs ndrectly affect the amount of vertcal trade va the change n the number of afflates. The foregn country's market access, 1 ρ 1 ρ 1 ρ ( τ jht jh ) Eh, can be decomposed as ( τ jt j ) E + ( τ jht jh ) Eh. The frst h h term s the demand sze of the home country dscounted by blateral trade costs and the second term s country j's ablty to access the world market. In the estmaton of VE, we lnearze equaton (13) to llustrate more drectly the mpact of explanatory varables on vertcal exports. The VI of country from country j can be defned n a smlar way. VI s expressed as a lnear functon of number of afflates, relatve wages, blateral and multlateral trade costs and world market access of country. (15) VE + m + D + t + Y + D + jt = β1γ t β 2 jt β3 j β 4 ln jt β 5 ln t β 6 ω β β β ω β 7 tω t + 8Yωt + 9 jt + 10ajt + jt = β 1γ jt β 2 jt β3 j β 4 ln jt β 5 jt (16) VI + m + D + t + lny + β β β ω β 6 tω t + 7Yωt + 8 jt + 9ajt + Next, we take nto account the endogenety of the number of afflates. The determnants of vertcal trade and FDI are possbly correlated and ths may cause the effects of FDI on vertcal trade to be based. I adjust for the endogenety of FDI by usng an nstrumental varable. ε jt ε jt In the frst-stage, we estmate m j and m j on nstrument varables. The relatve fxed costs for buldng a plant at home and those for buldng a plant abroad are used as nstrument varables. To capture the fxed cost of establshng a plant n a foregn

26 country, we use the Dong Busness Index (DBI) of the World Bank 8. We consder three of the ten areas of the DBI that are closely related to the entry of new frms: startng a busness, regsterng property and protectng nvestors. We average rankngs of the ease of dong busness of three areas and take the nverse of the average for a proxy of fxed costs. The emprcal estmaton draws data from varous sources. Internatonal comparable data on wages come from WDI and UNIDO. The calculaton of the TFP ndex follows Caves et al. (1982). Data on value added, labor nput, and captal nput are acqured from the INSTAT database of UNIDO, whch provdes value-added, gross fxed captal formaton and number of employees accordng to ISIC revson 2 at three-dgt manufacturng level. For each par of countres, the dstance measure s calculated as a great crcle dstance between the countres captals, whch s gven by Haveman s webste ( Trade barrers are evaluated by appled MFN tarffs. The trade-weghted MFN tarffs are taken from the Unted Natons Conference on Trade and Development s Trade Analyss & Informaton System (TRAINS). To measure market sze, constant GDP from WDI s used. The stock of FDI s used as a proxy for number of afflates. The outward FDI data are taken from the database of the Export-Import Bank of Korea, and the nward FDI data from the database of Mnstry of Commerce, Industry and Energy of Korea. 8 DBI has some lmtatons as a proxy for fxed costs: t covers only busness regulatons. For example, t does not account for the qualty of nfrastructure servces, the transparency of government procurement, or the underlyng strength of nsttutons

27 Table6. Defnton and Sources of Varables Varables Notaton Expected Sgns VE VI Source Relatve wages Relatve productvty Share of ntermedates Home GDP World GDP Blateral Dstance w j +/- +/- WDI, UNIDO a j +/- +/- γ, γ j + + Y, value added, captal stock, employment from UNIDO I/O tables from OECD, country home page Y j + + WDI Y w + + WDI D j - - Haveman: Weghted Dstance D w - Haveman, PWI 6.1 Blateral tarff t j +/- +/- WITS based on TRAINS World tarff t w - - WITS based on TRAINS Outward FDI outfdi + EXIM Bank of Korea Inward FDI nfdi + MCIE of Korea Fxed costs F j + - World Bank s DBI ndex B. Estmaton Results The equatons (15) and (16) are estmated wth blateral trade data wth 32 tradng partners for the 1995 to 2005 perod. The dependent varables are the level of vertcal trade and ntermedates trade. The explanatory varables are categorzed nto four groups. The frst conssts of the comparatve advantage measured by relatve wage and technology, the second ncludes gravty-related varables such as dstance, trade barrers and mporter s demand measured by real GDP, the thrd s the stock of FDI and the fourth s the nput share of ntermedates. Two dfferent methods of estmaton are employed. Frst, I estmate wth two stage least squares (2SLS) to solve the possble endogenety problem. I also apply a generalzed method of moment (GMM) to get the approxmate consstency and effcency of estmates

28 Table 7 presents the estmaton results where the dependent varables are vertcal exports and ntermedates exports. It uses tme-seres and cross-sectonal nformaton for a panel of 32 tradng partners. Robust standard errors are reported n parentheses. The frst and second columns report the estmaton results where the dependent varable s vertcal exports. Accordng to the theory, the sgn of the coeffcent on the relatve comparatve advantage can be ether postve or negatve. In the short run, producton costs are ncreasng n conjuncton wth the rses n both home and foregn wages, and thus the sgn of the coeffcent s ndetermnate. In the long run, the relatve wage s effects on the frm's decson on FDI become to overwhelm, and under certan condtons, t should be postvely correlated wth vertcal exports. There s also no a pror expectaton on the coeffcent of the relatve productvty, because the short-run and long-run effects of relatve productvty are also dfferent. The regresson results show that coeffcents on relatve wages n columns 1 and 2 are postve and statstcally sgnfcant at the ten percent level. The next row of the table show that coeffcents on the productvty of the exporter country also have postve sgns and that they are statstcally sgnfcant at the ten percent level. These results mply that the larger share of Korea's vertcal exports takes place wth low-productvty countres and labor abundant countres. The thrd to eghth rows of the table ndcate that gravty-related varables are generally approprate to explan Korean vertcal exports. Blateral dstance as well as the weghted dstance to the world market turns to be negatvely correlated wth vertcal exports, as the theory suggests. The parameters of world demand are also found to be all postve and statstcally sgnfcant at the one percent level. The coeffcents on blateral tarffs can be ether postve or negatve because vertcal trades are two-way trades. The results on blateral tarffs are all negatve and

29 statstcally sgnfcant. The results on the producton share of ntermedates of exporter country are consstent wth the theory. The coeffcents show postve sgns, even though some of them are not statstcally sgnfcant. Turnng to the effects of FDI on vertcal trade, the estmaton results dsplay overall consstency wth the theory. The parameters of FDI are found to be all postve and statstcally sgnfcant at the one percent level. The model was re-estmated usng ntermedates export as a dependent varable (columns 3 and 4). The results are smlar overall to those obtaned when vertcal exports were used, but they are less sgnfcant. Table 8 reports the same regresson as n Table 7, except that the dependent varables are now panel data of vertcal mports and ntermedates mports. The second row of the table show that relatve wage are negatvely correlated wth vertcal mports but do not have statstcally sgnfcant effects. The coeffcents on relatve TFP's are negatve and statstcally sgnfcant. Agan, the country-par gravty varables are found to be very plausble n explanng vertcal mports. The parameters for dstance are all negatve and statstcally sgnfcant. The coeffcents on world and exporter s GDPs are correctly sgned, even though some of them are not statstcally sgnfcant. The sgns of coeffcents on the producton share of ntermedates of the exportng country are the opposte of what was expected. The coeffcents show negatve sgns, even though they are not statstcally sgnfcant. The coeffcents of FDI also are found to be all postve and statstcally sgnfcant at the one percent level

30 Table7. Estmaton Results on Vertcal Exports: Panel Data by Country and Industry Dependent Varable: Vertcal Exports Intermedates Exports notaton explanaton Mean (Std. Dev.) 2SLS GMM 2SLS GMM w / w j Relatve wage 4.55 (0.77) 0.065* (0.036) 0.340** (0.162) 0.045* (0.029) 0.021* (0.011) a / a j Relatve productvty 4.42 (0.45) 0.378* (0.311) 0.967* (0.530) (0.274) (0.040) D j Blateral dstance 8.73 (0.68) * (0.239) ** (0.276) (0.341) (0.340) t / t Relatve tarff k jk 2.26 (0.18) *** (0.048) * (0.163) *** (0.039) (0.143) Y Korea s GDP 8.53 (0.14) ** (0.672) (0.879) (0.933) (0.191) D w GDP-weghted dstance to major markets 26.5 (1.33) * (0.432) (1.103) (1.569) (1.322) t w World tarff 1.97 (0.18) ** (0.346) *** (0.329) (0.292) (0.217) Y w World GDP (0.09) (2.033) (1.601) 1.657* (1.811) (2.096) γ k Intermedates nput rato 4.24 (0.07) 0.634* (0.345) 0.408* (0.228) (0.314) (0.404) Outward FDI outward FDI (2.15) 3.821** (1.521) 4.450*** (1.460) 3.345** (1.401) 4.068*** (1.189) R-squared Observatons 1,724 1,724 1,724 1,

31 Table8. Estmaton Results on Vertcal Import: Panel Data by Country and Industry Dependent Varable: Vertcal Imports Intermedates Imports notaton explanaton Mean (Std. Dev.) 2SLS GMM 2SLS GMM w / w j Relatve wage 4.55 (0.77) (0.214) (0.130) (0.132) (0.313) a / a j Relatve productvty 4.42 (0.45) *** (0.103) * (0.149) *** (0.131) (0.221) D j Blateral dstance 8.73 (0.68) * (0.121) * (0.101) * (0.118) ** (0.131) t / t k jk Relatve tarff 2.26 (0.18) ** (0.210) * (0.476) ** (0.043) (0.062) Y j Korea s GDP 8.05 (1.33) 0.121** (0.052) 0.532* (0.317) (0.231) (0.085) t w World tarff 1.97 (0.18) (0.232) (0.191) (0.121) (0.131) Y w World GDP (0.09) 2.121* (1.091) 2.426* (1.473) (1.452) (1.163) γ jk Intermedates Input rato 4.22 (0.07) (1.313) (2.542) (2.315) (1.453) Inward FDI nward FDI (3.45) 0.321*** (0.092) 0.334*** (0.089) 0.231** (0.095) 0.451*** (0.052) R-squared Observatons 1,721 1,721 1,721 1,

32 IV. Concluson Whle Korea s vertcal trade, fostered by foregn drect nvestment, has been growng rapdly, t has been left unexamned untl recently. Ths paper provded the mproved measurement based on HIY (2001) and quantfed the vertcal trade n Korea by tradng partner and by ndustry. The emprcal results show that Korea s trade has been more vertcally specalzed over the last ten years. In partcular, vertcal exports to Chna have grown rapdly snce the early 2000 s, and thus the surplus on vertcal trade has wdened dramatcally n recent years. Ths paper also derves a functonal form for the determnants of vertcal trade from a monopolstc competton model of trade and multnatonal producton. The emprcal estmaton wth 32 tradng partners durng the perod of provdes evdence that the Korea's vertcal trade has been affected by the relatve comparatve advantage as well as by the geographcal advantage. Vertcal trade s also ndrectly affected by relatve fxed costs through the change n the stock of FDI. The evdence presented here suggests some polcy mplcatons. As nternatonal producton fragmentaton and vertcal trade have expanded rapdly, t s becomng more mportant for frms to have a comparatve advantage n "stages" wth hgh value added and to focus on the producton of parts and components n those stages. It can be sad that, untl recently, Korea has been successful n establshng ts regonal producton network and remanng a major suppler of ntermedate goods for processng trade to developng countres n Asa. Despte the strong performance of vertcal exports, however, concerns about ther sustanablty have recently been rased. Korea s perceved by some as beng sandwched between Chna, whch s catchng

33 up rapdly n terms of product technology, and Japan, whch s stll technologcally more advanced. In partcular, the domestc producton capacty of Chna has ncreased substantally n recent years, whch has enabled ncreased local content n ts processng trade and weakened the demand for ntermedate goods from Korea 9. Vertcal mport of ntermedate goods from Chna s also ncreasng, whle the dependency on Japan for hgh-tech ntermedate mports s stll hgh. To cope wth these challenges, the government and frms n Korea need to put more effort nto contnung to move ts manufacturng sector up the value chan. 9 As the tradtonally tght lnk between exports and mports has weakened, Chna s trade surplus has expanded sharply n recent years. Durng , the trade surplus grew over fve -fold n dollar terms and more than trpled as a percentage of GDP. As of 2006, the trade surplus was more than 8 percent of GDP. The rapd wdenng of the trade surplus has manly been drven by a sgnfcant slowdown n mports. Imports of ntermedate goods slowed the most, wth parts and components and sem-fnshed goods accountng for almost half of the slowdown n mport growth between 2003 and 2006 (Cu and Syed, 2007)

34 References Ando, Mtsuyo and Fukunar Kmura (2007), "Fragmentaton n East Asa: Further Evdence", mmeo. Amt, Mary (2005), "Locaton of Vertcally Lnked Industres: Agglomeraton versus Comparatve Advantage", European Economc Revew 49, pp Arndt, Sven W. and Alex Huemer (2005), "Trade, Producton Networks and the Exchange Rate", mmeo. Athukorala, Prema-chandra and Nobuak Yamashta (2006), "Producton Fragmentaton and Trade Integraton: East Asa n a Global Context", The North Amercan Journal of Economcs and Fnance, 17, pp Bernard, B., J. Jenson, S. Reddng and Peter Schott (2007), "Intra-Frm Trade", mmeo. Carpenter, Theresa (2005), "Multnatonals, Intra-Frm Trade and FDI: a Smple Model", HEI WP No 1/2005. Dean, Judth M, K.C. Fung and Zh Wang (2007), Measurng the Vertcal Specalzaton n Chnese Trade, U.S. Internatonal Trade Commsson Workng Paper no A. Cu, L and Murtaza Syed (2007), The Shftng Structure of Chna s Trade and Producton, IMF Workng Paper WP/07/214. Feenstra, Robert C. and Gordon Hansen (2003), "Global Producton Sharng and Rsng Inequalty: A Survey of Trade and Wages", n Kwon Cho and James Harrgan, eds., Handbook of Internatonal Trade, Blackwell, pp Fukao, Ishdo and Ito (2003), "Vertcal Intra-Industry Trade and Foregn Drect Investment n East Asa", RIETI Dscusson Paper Seres 03-E-001. Gauler, Lemone and Nal-Kesenc (2005), "Chna s Integraton nto East Asa: Producton Sharng, FDI and Hgh-Tech Trade", CEPII WP No Greenaway D., Hne, R., and Mlner C. (1994), "Country-Specfc Factors and the Pattern of Horzontal and Vertcal Intra-Industry Trade n the U.K.", Weltwrtschaftlches Archv, Vol. 130, pp Greenaway D. and J. Tortensson (2000), "Economc Geography, Comparatve Advantage and Trade wthn Industres: Evdence from the OECD", Journal of Economc Integraton, 15, pp Gullstrand, Joakm (2001), "Demand Patterns and Vertcal Intra-Industry Trade wth Specal Reference to North-South Trade", CFE Workng Paper Seres no. 9. Hanson, Gordon H., Raymond Matalon and Matthew J. Slaughter (2004), Multnatonal Frms", Brookngs Trade Forum 2001, pp "Expanson Strateges of U.S. Hayakawa, Kazunobu (2007), "Growth of Intermedate Goods Trade n East Asa", Pacfc Economc Revew, 12, pp Hummels, Davd, Dana Rapoport, and Ke-Mu Y (1998), Vertcal Specalzaton and the Changng Nature of World Trade", FRBNY Economc Polcy Revew, June, pp Hummels, Davd, Jun Ish, and Ke-Mu Y (2001), "The Nature and Growth of Vertcal Specalzaton n World Trade", Journal of Internatonal Economcs, 54 (1), pp

35 Jones, R. W., and H. Kerzkowsk (1990), "The Role of Servces n Producton and Internatonal Trade: a Theoretcal Framework. In R. W. Jones and A. O. Krueger, eds., The Poltcal Economy of Internatonal Trade: Essays n Honor of R. E. Baldwn. Oxford, Basl Blackwell Kmura, F., Y. Takahash, and K. Hayakawa (2007), "Fragmentaton and Parts and Components Trade: Comparson between East Asa and Europe", The North Amercan Journal of Economc and Fnance, Vol. 18 (1) pp Kose, M. Ayhan and Ke-Mu Y (2006), "Can the Standard Internatonal Busness Cycle Model Explan the Relaton Between Trade and Co-movement?", Journal of Internatonal Economcs, pp La, Huwen Susan and Susan Zhu (2006), "U.S. Exports and Multnatonal Producton", Revew of Economcs and Statstcs, pp Martn, Joan A. and Vcente Orts (2001), "Vertcal Specalzaton and Intra-Industry Trade: The Role of Factor Endowments", Weltwrtschaftlches Archv, Vol. 138(2), pp Markusen, James R. and Anthony J. Venables (2006), "Interactng Factor Endowments and Trade Costs: a Mult-Country, Mult-Good Approach to Trade Theory", Internatonal Trade Workshop, Unversty of Otago, March Markusen, James R. and Keth E. Maskus (2001), "A Unfed Approach to Intra-Industry Trade and Drect Foregn Investment", NBER WP Ng, Francs and Alexander Yeats (1999), "Producton Sharng n East Asa: Who Does What for Whom, and Why?", World Bank Polcy Research Workng Paper Okubo, Toshhro (2004), "Intra-ndustry Trade and Producton Networks", HEI Workng Paper no.13/2004. Okamoto, Yumko (2005), "The Emergence of "Intra-Medate Trade": Implcatons for the Asa-Pacfc Regon", Paper presented at Secton 2 of PAFTAD 30, Does Trade Delver What It Promses?, Honolulu on February Wakasug, Ryuhe (2006), "The Growth of Trade and Heterogenety n East Asa: An Emprcal Evdence", Asan Economc Panel Meetng n Seoul, March Y, Ke-Mu (2003), "Can Vertcal Specalzaton Explan the Growth of World Trade?" Journal of Poltcal Economy, February, pp