NBER WORKING PAPER SERIES GIVE CREDIT WHERE CREDIT IS DUE: TRACING VALUE ADDED IN GLOBAL PRODUCTION CHAINS

Transcription

1 NBER WORKING PAPER SERIES GIVE CREDIT WHERE CREDIT IS DUE: TRACING VALUE ADDED IN GLOBAL PRODUCTION CHAINS Robert Koopman William Power Zhi Wang Shang-Jin Wei Working Paper NATIONAL BUREAU OF ECONOMIC RESEARCH 1050 Maachuett Avenue Cambridge, MA September 2010 The view in the paper are olely thoe of the author and may not reflect the view of the USITC, it Commiioner, or of any other organization that the author are affiliated with. The view expreed herein are thoe of the author and do not necearily reflect the view of the National Bureau of Economic Reearch by Robert Koopman, William Power, Zhi Wang, and Shang-Jin Wei. All right reerved. Short ection of text, not to exceed two paragraph, may be quoted without explicit permiion provided that full credit, including notice, i given to the ource.

2 Give Credit Where Credit I Due: Tracing Value Added in Global Production Chain Robert Koopman, William Power, Zhi Wang, and Shang-Jin Wei NBER Working Paper No September 2010 JEL No. F1,F2 ABSTRACT Thi paper aim to integrate and generalize the many attempt in the literature at tracing value added by country in international trade. We provide a conceptual framework to etimate value added that i more comprehenive than what i available in the exiting literature. We fully detail the ource of value added accounting for 100 percent of gro export, including everal component of dometic value-added in export. The decompoition of value-added export examine both export that are aborbed by the direct importer and thoe that are proceed and ubequently exported to other countrie. To empirically implement the framework, we contruct an international databae of value-added production and trade baed on the GTAP databae. Our reult document ubtantial difference between major regional upply network, with le integration between Eat and Wet Europe than within North America or Aia in Many meaure how that emerging Aia differ coniderably from other emerging economie. Much of Aia, including China itelf, end their export to countrie that provide final aembly on behalf of conumer in other countrie. Aian countrie alo have relatively dipered ourcing of imported intermediate input. Robert Koopman Reearch Diviion Office of Economic US International Trade Commiion 500 E Street SW Wahington, DC Robert.Koopman@uitc.gov William Power Reearch Diviion Office of Economic US International Trade Commiion 500 E Street SW Wahington, DC william.power@uitc.gov Zhi Wang Reearch Diviion Office of Economic US International Trade Commiion 500 E Street SW Wahington, DC zhi.wang@uitc.gov Shang-Jin Wei Graduate School of Buine Columbia Univerity Uri Hall Broadway New York, NY and NBER hangjin.wei@columbia.edu

3 1. Introduction Worldwide trade ha become increaingly fragmented, a different tage of production are now regularly performed acro multiple countrie. A input cro border multiple time, traditional tatitic on trade volume meaured in gro value term become increaingly le reliable a a gauge of value added being exported and imported. Thi paper aim to integrate and generalize the many attempt in the literature at tracing value added by country in international trade. We provide a conceptual framework to etimate value added that i more comprehenive than what i available in the exiting literature. By deign, thi i an accounting exercie, and doe not directly examine the caue and the conequence of global production chain. However, an accurate accounting of value added chain by ource country i a neceary tep toward a better undertanding of all thee iue. Supply chain can be decribed a a ytem of value-added ource and detination within a globally integrated production network. 2 An international upply chain ditribute value-added hare among countrie in a particular global indutry. Within the upply chain, each producer purchae input and then add value, which i included in the cot of the next tage of production. At each tage in the proce, a good cro an international border, the value-added trade flow i equal to the value added paid to the factor of production in the exporting country, and the final good conumed by end uer are equal to the um of all accumulated value added in the chain. However, all official trade tatitic are meaured in gro term, which include both intermediate good and final product. Official trade flow will therefore be overtated becaue they double count the value of intermediate good that cro international border more than once. Cae tudie of value chain in indutrie uch a electronic, apparel, and motor vehicle have provided detailed example of the dicrepancy between gro and value-added trade value. According to a commonly cited tudy of the Apple ipod (Dedrick, Kraemer, and Linden, 2008), the Chinee factory gate price of an aembled ipod i $144. Of thi, a little a $4 may be Chinee value added. 3 Nor i thi a particularly iolated example, at leat for Chinee electronic. Koopman, Wang, and Wei (2008) how that on average, foreign countrie contribute 2 Wang, Power, and Wei (2009). 3 The ipod exported from China contain about $100 in Japanee value added (for the hard drive, diplay, and battery), and about $15 of U.S. value added (for the proceor, controller, and memory). Korea alo make a mall contribution. China may contribute ome additional value added in the $22 of unpecified part. 2

4 80% or more of the value added embodied in Chinee export of computer, office equipment, and telecom equipment. There are numerou other cae tudie of pecific chain that how imilar dicrepancie, including Barbie doll, Chinee hard drive, North American automobile, and Aian apparel. Hummel, Ihii, and Yi (2001), denoted HIY in the following dicuion, provided the firt general meaure of vertical pecialization in trade that meaure foreign value added in a country export. Thi eminal paper provided key definition and the firt meaure of the extent of vertical pecialization in trade. While thee meaure have ince been revied, they continue to be central to the dicuion of vertical pecialization in trade. 4 HIY defined meaure of both direct value-added trade and indirect value-added trade that pa through third countrie. Foreign content in direct export, what HIY define a VS, ha received more attention in the literature than indirect value added trade flow through third countrie, or VS1 in HIY terminology. A a conequence, important upplier like Japan, that lie uptream in global upply chain and whoe intermediate export are embodied in further intermediate export by other countrie, have ometime received le attention than large downtream aembler, uch a China, that hip more finihed product. And, in extended upply chain, where intermediate cro border more than once, the HIY meaure are no longer accurate meaure of vertical trade. Our ability to track ource and detination of value added within pecific chain ha improved a detailed inter-region input output (IRIO) table have become available for pecific countrie and region. Several paper have invetigated value-added trade in Aian upply chain uing the Aian input-output (IO) table produced by the Intitute of Development Economie. Such paper include Koopman, Wang, and Wei (2009), Pula and Peltonen (2009), and Wang, Power, and Wei (2009). Thee tudie have provided a careful breakdown of the ource and detination of value-added in Aian production network. For pecific chain, they have noted large difference in the organization and ditribution of production acro product (e.g., apparel, automobile, and electronic). However, thee tudie reliance on the Aian IO table preclude them from tracking value-added to and from countrie outide of Aia, although flow to the United State are included. 4 See Chen et al. (2005) and Yi (2003) for revied etimate of the extent of vertical pecialization in trade; ee Daudin, Rifflart, and Schweiguth (2009), Johnon and Noguera (2009), and Wang, Power, and Wei (2009) for refined definition of vertical pecialization in trade. 3

5 Truly global analye have become poible recently, with the advent of global IRIO table baed on the GTAP databae. 5 Such table provide global etimate of double-counted intermediate in trade (about 25% of gro flow), and allow comparion of production network in different region. Global data can change our undertanding of value-added trade for important countrie uch a the United State, which, a we how below, ha an unuually large amount of it own exported value added returned to it producer and conumer after proceing abroad. Though uefully global in cope, the GTAP databae doe not eparate imported intermediate and final good trade flow, o ome important parameter have to be etimated. 6 Effort are underway to produce more accurate and up-to-date global IRIO table with le etimation of unknown parameter baed on a compilation of ingle-country (or -region) IO table and detailed bilateral trade tatitic. 7 Thi paper provide the firt unified framework that integrate the older literature on vertical pecialization with the newer literature on value added trade. It completely decompoe gro export and connect official gro tatitic to value-added meaure of trade. The framework ditribute all value-added in a country export to it original ource, and it expree individual ource and detination of value added at either the country-wide or indutry average level. Depite the breadth of the framework, it i alo quite parimoniou, expreing major global value-added flow a the product of only three matrice. Thi paper alo provide new detailed decompoition of each country value-added export that highlight it uptream or downtream poition in global value chain. Thi paper i alo related to Daudin, Rifflart, and Schweiguth (2009) and Johnon and Noguera (2009). Each of thee paper highlight inaccuracie in HIY meaure of value-added export. They analyze global value-added trade flow uing an etimated IRIO table baed on the GTAP databae, in which they proportionally allocate gro trade flow into intermediate and final good and ditributing acro uer. Each how that countrie and ector differ widely in their ratio of value added to gro trade. Thi paper expand upon their analyi in the following four apect: 5 See Daudin, Rifflart, and Schweiguth (2009), Johnon and Noguera (2009), and Bem, Johnon, and Yi (2010). 6 See ection 3 for additional ditinction between the IO tructure underlying the GTAP databae and IRIO table required for global value-added analyi. 7 See Wang, Tiga, Mora, Xin, and Xu (2010) for the contruction of one uch databae. The World Input-Output Databae Conortium ( i producing a imilar et of table. 4

6 Firt, we develop a ingle, unified, tranparent conceptual framework that incorporate all meaure of value-added trade. Thee meaure include analogue of Johnon and Noguera ratio of value-added trade to gro trade (VAX ratio), HIY meaure of vertical pecialization (both VS and VS1), and Daudin et al. meaure of dometic value added in import returned from abroad (VS1*). 8 Second, we completely decompoe each country' gro export into it value-added component, thu etablihing a formal relationhip between value-added meaure of trade and tandard trade tatitic. We begin by plitting gro export into dometic value added ent abroad, dometic value added ent abroad that ultimately return in home import, and foreign value added embodied in gro export. Our decompoition tie HIY original meaure of vertical pecialization (for the firt time, a far a we know) with newer meaure of value-added trade in a way that completely account for all element of gro export. 9 Third, we plit dometic value-added export into four part that allow u to more clearly ditinguih each country poition and role in global value chain. 10 Dometic value-added export are either aborbed by the importing country or exported on to third countrie. The portion that i aborbed by the directly importing country include dometic value added in: (i) final good export, and (ii) intermediate export that are tranformed into final good and aborbed by the direct importer. The portion not aborbed by the direct importer (i.e., VS1) include dometic value-added in two additional part: (iii) intermediate input that are tranformed into final good and exported to a third country for conumption, and (iv) intermediate input that are ued to produce other intermediate and ent to a third country for further proceing. Fourth, our etimated global IRIO better capture the international ource and ue of intermediate good than in previou databae in two way. Firt, we improve the etimate of intermediate import by matching detailed bilateral import data to end-ue claification o a to 8 Other framework have been le complete or le fully pecified. Johnon and Noguera (2009) did not examine HIY VS1 meaure. In addition, their value added decompoition i preented fully only for trade with one combined world region, and they do not pecify how trade within the ret of the world i incorporated. Daudin, Rifflart, and Schweiguth (2009) computed value of VS, VS1, and VS* for multiple countrie, but they calculated each term eparately, and did not pecify the connection between thee term and gro or value-added trade flow, which i very different from the integrated value-added trade decompoition reported in our paper. 9 Each of thee term ha been modified from it original definition, however, to correctly pecify it hare of value added in a multicountry framework or to account for all element of gro trade. 10 Thi provide a decompoition of the VAX ratio from Johnon and Noguera (2009). 5

7 determine the intermediate good hare of import in each ector for each exporter and importer. 11 The additional detail provide a ubtantial improvement over earlier approache that aumed the hare of intermediate good in import matched the hare of intermediate in total aborption. Similarly, proceing trade regime can foter import that have dramatically different (i.e., higher) intermediate content than dometic ue in ome countrie. We explicitly account for thee difference in Mexico and China, uing the expanded Chinee IO table with eparate account for proceing export from Koopman, Wang, and Wei (2008) and the 2003 Mexican IO table with eparate dometic and Maquiladora account from Mexico tatitical agency. While other tudie have ued a imilar correction for Chinee export, the new Mexican IO table provide improved accuracy in etimate of NAFTA trade flow by ditinguihing dometic and Maquiladora production. Thi paper i organized a follow. Section 2 preent key meaure of vertical pecialization and value-added trade in global upply chain, and it pecifie the global IO model that generate thee key meaure. Section 3 dicue how the required global IO model can be made operational, given the limited information in current databae with linked IO table. Section 4 applie the model to the contructed databae to decompoe each country' gro export and to decribe each country' participation in global value chain in term of the compoition of their value-added export. Section 5 conclude. 2. Value Chain in Global Production Network: Concept and Meaurement 2.1 Concept With modern international production chain, value added embodied in export originate in many place. Detailing thee ource and meauring their contribution ha been the main focu of both the vertical pecialization and value-added trade literature. A noted, HIY provided the firt empirical framework to meaure participation in vertically pecialized trade: a country can ue imported intermediate input to produce export, or it can export intermediate good that are ued a input producing good exported by another country. HIY derived VS a a 11 Feentra and Jenen (2009) ue a imilar approach to eparate final good from intermediate in U.S. import. They concord HS import to end-ue categorie provided by the BEA. We concorded HS import to UN Broad Economic Categorie, which are more applicable to international trade flow. We integrated the Chinee and Mexican IO table with verion 7 of the GTAP databae uing a quadratic mathematical programming model to minimize the deviation between the reulting new data et and the original GTAP databae. 6

8 meaure of the value of import embodied in a country export, and VS1 a a meaure of the value of exported good that are proceed and further exported by other countrie. However, HIY original meaure were inufficient for full analyi of upply chain. Their meaure of VS1 i valid only in a pecial cae, they did not mathematically define VS1, and thee two meaure do not capture all ource of value added in gro export. Thi ection examine each of thee hortcoming, and provide a fully generalizable olution for each. 13 Two key aumption are needed for the HIY meaure to accurately reflect value-added trade. Firt, intenity in the ue of imported input mut be the ame whether good are produced for export or for dometic final demand. Thi aumption i violated when proceing export are pervaive due to policy incentive, a in China and Mexico. 14 The econd key aumption i that all imported intermediate input mut contain 100% foreign value added and no more than one country export intermediate. 15 In the HIY model, a country cannot import intermediate input, add value, and then export emi-finihed good to another country to produce final good. Nor can a country receive intermediate import that embody it own value added, returned after proceing abroad. Thee aumption generally do not hold in today increaingly globalized world, and the multi-country, back-and-forth nature of many current global production network i often not well captured by the HIY meaure. 16 To accurately track the ource of global value-added, a framework i needed that combine VS and VS1, and alo capture the remaining ource of value added. Such a framework i given by our VAS matrix decribed in the next ection. Meauring value-added embodied in export require contruction of a databae detailing international production and ue for all flow of value added. To preciely define uch chain acro many countrie, the databae mut quantify the contribution of each country to the total value-added generated in the proce of upplying final product. 13 The firt hortcoming ha been addreed in the literature (Johnon and Noguera, 2009), but the other have not. 14 Proceing regime provide incentive to ue imported intermediate import, but require all reulting product to be 100% exported. Hence, proceing export have a much greater imported intermediate input intenity than dometic product. A dicued in ection 3, we follow Koopman, Wang, and Wei (2008) to re-compute dometic and foreign value added in China and Mexico, but in a multi-country global etting. 15 Thi i equivalent to the aumption that the firt exporting country export have to be 100% dometic ourced when compute VS1 in HIY framework. 16 See Johnon and Noguera (2009) for a comparion of HIY VS and a more generalized meaure of foreign value in export. 7

9 A global IRIO table provide ufficient information to completely lice up the value chain acro all related countrie at the indutry average level. 17 In detail, an IRIO table contain a number of ub-matrice that pecify (a) tranaction flow of intermediate product and final good within and between each country at the indutry level, (b) the direct value-added in production of each indutry in all countrie, and (c) the gro output of each indutry in all countrie. In other word, the IRIO table provide the origin and detination of all tranaction flow by indutry. It alo pecifie every intermediate and/or final ue for all uch flow. For example, an IRIO table would decribe the number of electronic component produced in Japan that were hipped to China. It would ditinguih the number that were ued a intermediate input in each Chinee ector and the number that were ued in Chinee private houehold conumption and capital formation. The IRIO would alo allow u to determine the amount of proceed electronic that were then exported to the United State and ued by different ector of the U.S. economy. IRIO exit, however, for a limited number of countrie and region. A uch, we have etimated a global IRIO table baed on verion 7 of the GTAP databae (2004). Section 3 preent the detail of our databae contruction. In the ret of ection 2, we will illutrate how a global IO model can allocate the value added in a multi-country production chain to each participating country. We will incorporate previou meaure of value added and vertical pecialization into a complete decompoition of gro trade flow. To preent the major concept of our decompoition and how how they differ from earlier meaure, we tart with a two-country cae and then extend to a world with many countrie Two-country cae Aume a two-country (home and foreign) world, in which each country produce good in N differentiated tradable ector. Good in each ector can be conumed directly or ued a intermediate input, and each country export both intermediate and final good to the other. 17 There are alo product-level approache to etimating the financial value embedded in a product and quantifying how the value i ditributed among participant in the upply chain, moving from deign and branding to component manufacturing to aembly to ditribution and ale (Dedrick, Kraemer, and Linden, 2008). 18 The author are very grateful for the contructive dicuion with Dr. Kei-Mu Yi at the Federal Reerve Bank of Philadelphia in developing the two- and three-country cae and the relationhip between our new meaure of vertical pecialization and the original HIY meaure. 8

10 All gro output produced by country i mut be ued a an intermediate good or a final good at home or abroad, or X r A X A X Y Y, r, = 1,2 (1) rr r r rr r where X r i the N 1 gro output vector of country r, Y r i the N 1 final demand vector that give demand in country for final good produced in r, and A r i the N N IO coefficient matrix, giving intermediate ue in of good produced in r. The two-country production and trade ytem can be written in block matrix notation a X X 1 2 A A and rearranging, X X 1 2 I A A A A 22 X X A 1 2 I A Y Y Y Y Y Y 22 Y Y, (2) 22 B B B B 22 Y Y 1 2. (3) where B r denote the N N block Leontief invere matrix, which i the total requirement matrix that give the amount of gro output in producing country required for a one-unit increae in final demand in country r. Y i i an N 1 vector that give global ue of i final good. Thi ytem can be expreed uccinctly a: 1 X ( I A) Y BY, (4) where X and Y are 2N 1 matrice, and A and B are 2N 2N matrice. Having defined the Leontief invere matrix, we turn to meaure of dometic and foreign value added, firt for production, and then applied to trade. Let V be the 1 N direct value-added coefficient vector. Each element of V give the hare of direct dometic value added in total output. Thi i equal to one minu the intermediate input hare from all countrie (including dometically produced intermediate): V r u( I Ar ), (5) where u i a 1 N unity vector. To make the analyi conitent with the multiple-country dicuion below, we will alo define V, the 2 2N matrix of direct dometic value added for both countrie, V1 0 V. (6) 0 V2 9

11 trade. 19 Variation of thi framework have been ued in a number of recent tudie of value-added However, none of thee paper clearly illutrate how to apply thi framework to completely decompoe export and capture all ource of value added. We turn to thi tak next. Combining thee direct value-added hare with the Leontief invere matrice produce the 2 2N value-added hare (VAS) matrix, our baic meaure of value-added hare. V1B11 V1B VAS VB. (7) V 2B21 V2B22 Within VAS, each column of V 1 B 11 denote dometic value-added hare of dometic produced product in a particular ector at home. Similarly, column of V 2 B 21 denote country 2 valueadded hare for thee ame good. Together, each of the firt N column in the VAS matrix include all value added, dometic and foreign, needed to produce one additional unit of dometic product at home. The econd N column preent value-added hare for production in country 2. Becaue all value added mut be either dometic or foreign, the um along each of the 2N column of VAS i unity: 20 V1 B11 V2B21 V1B V2B22 u. (8) The VAS matrix i mot naturally applicable to final good trade, becaue of the definition of the invere Leontief matrix. To compare to other meaure of vertical pecialization that have been developed in the literature and link our meaure with official trade tatitic, however, we will apply the VAS matrix to export of both final and intermediate good. 21 Let E r be i the N 1 vector of intermediate and final export from r to, and Ê r be the N N diagonal matrix of thi export vector. For conitency with the multi-country analyi below, 22 alo define E E ( A X Y ), r, = 1,2 (9) r r r r r E 0 E, and (10) 0 E21 19 Daudin et al. (2008), Johnon and Noguera (2009), and Wang, Power, and Wei (2009). 20 To how thi, ubtitute the value of the B ij from equation (9) into equation (4). KWW (2008) how thi mut hold in the general cae with any number of countrie and ector. 21 Mathematically, the application to intermediate good trade preent no problem, becaue value added in a product doe not depend on how it i ued. In other word, value-added hare in intermediate good match the hare in final good within the ame ector. 22 In the two-country cae, all trade i bilateral trade, o E r =E r. For clarity, we ue E r intead of E r in the twocountry cae, although we will generally ue E r with multiple countrie. 10

12 ˆ diag( E) 0 E, (11) 0 diag( E21) where E i a 2N 2 matrix and Ê i a 2N 2N diagonal matrix. The combination of the value-added hare matrix (VAS) and the export matrix produce VAS_E, our main meaure of value-added trade and vertical pecialization in global value chain: ˆ ˆ ˆ V1B11E V1BE21 VAS_Ê VB E ˆ ˆ. () V 2B21E V2B22E21 VAS_ Ê i a 2 2N matrix. Thi matrix i a diaggregated meaure of value-added export, ince it expree value added embodied in export of each ector. Thi meaure allow comparion to cae tudie of upply chain in ingle ector, uch a apparel, electronic, or motor vehicle. However, for implicity, and to match our empirical focu on aggregate trade, we will focu on the aggregate verion of thi meaure throughout the ret of thi ection. 23 The aggregate (2 2) meaure of value-added export i given by V1B11E V1BE21 VAS_E VBE. (13) V 2B21E V2B22E21 Although rather elementary with only two countrie, VAS_E expree the major concept of our vertical pecialization meaure. Diagonal element of VAS_E define the dometic value-added hare in a unit of each country export. Off-diagonal element give the hare of foreign value-added embodied in a unit of each country export. The off-diagonal element allow u to improve upon HIY two major value-added concept: foreign value added exported directly (VS), and dometic value added exported indirectly through third countrie (VS1). In the two-country cae, the explicit olution of the four B ij block matrice are not too cumberome, and nicely illutrate the difference between our general meaure and the HIY meaure. Applying the algebra of the partitioned matrix invere, 24 we have B B B B 22 ( I A11 A ( I A22 ) 1 ( I 2 A22 ) A21B 1 11 A ) 1 21 ( I A 11 B A A ( I A 21 ( I A ) ) 23 All reult continue to hold with full dimenionality, and can be expreed imply by replacing an export vector with the relevant diagonal matrix. 24 See, for example, Simon and Blume (1994, 182); B 22 follow from the ymmetry of countrie 1 and 2. A ) 1 (14)

13 Therefore, gro export can be decompoed into foreign value-added (VS) and dometic value-added (DV) a follow: V 2B VS V1 B 21 V1B DV V 2B E E E E 21 u( A u( A A A V1 ( I A V 2 ( I A ( I A ( I A A A 21 ) ) A A ( I A ( I A 21 ) )( I A )( I A ) A A ) ) A A E E ( I A ( I A ) ) A A ) ) They are both 2 1 matrice. Uing the ame notation, HIY VS meaure can be expreed a another 2 1 matrix: ua VS_HIY ua 21 ( I A ( I A ) ) E E 21 E E 21 (15) (16). (17) Comparing equation (15) and (17), we can ee that the HIY VS meaure accurately capture value added in trade only when A =0 or A 21 =0; i.e., in the cae when only one country intermediate good are ued abroad. A Johnon and Noguera (2009) alo point out, whenever two or more countrie export intermediate product, the HIY meaure diverge from the true meaure of value added in export. The true meaure capture an important element omitted from HIY implified formula. For the home country, both dometic and foreign value-added differ from their true value by the 1 adjutment term A ( I A22 ) A21. The term conit of the product of A 1 ( I A22 ), which a (16) how, repreent country 1 value added embodied in country 2 export, time A 21, which give country 1 intermediate ue of thoe good. Thu the true meaure can account for value added when a country import it own value added. In a more general context, VAS will properly attribute foreign and dometic content to multiple countrie when intermediate product cro border in even more complicated pattern. Thi will become clearer when we extend the meaure to three or more countrie. The econd HIY meaure of vertical pecialization i VS1, which meaure the amount of a country value-added that i ent indirectly through third countrie, or the value added in intermediate export ued to produce other countrie export. Although thi term ha not been previouly defined mathematically, VS1 can be pecified preciely baed on our meaure of value-added in VAS. In a two-country world, the home country VS1 meaure can be defined a VS1 1 V B E. (18) 1 21

14 Thi product of three term give the home country value-added in the good it produce to generate a dollar worth of increaed production abroad time the amount of that foreign production that i then exported back home. In a two-country framework, home country VS1 i identical to foreign country' VS in export. Thi will not be true in the multi-country model that we turn to next Three or more countrie The previou analyi can be generalized to any arbitrary number of countrie. Production, value-added hare, and value-added export are given uccinctly by: X ( I A) VAS VB 1 Y BY VAS _ E VBE. (19) With G countrie and N ector, X and Y are GN 1 vector; A, B, and Ê are GN GN matrice; V and VAS are G GN matrice; E i a GN G matrix; and VAS_E i a G G matrix. In the multiple-country cae, accurately calculating value added require adjutment for intermediate input that cro multiple border. Examining a three-country cae in ome detail i ueful for two reaon: (i) it exhibit nearly all the richne of the fully general multi-country analyi, and (ii) analytical olution remain available and continue to have intuitive explanation. For example, home dometic value-added hare in production i given by 26 V B 1 11 V1{ I A A [ I A A A 32 [ I A ( I A A ) ( I A A A23] [ ) A A ] [ A ( I A A ) A21 ( I A 1 ]} 33 ) 1 A 31 ]. (20) Thi include adjutment for country 2 export to country 3 that are ubequently exported and ued in country 2 or country 1, and adjutment for country 3 export to country 2 that are ubequently ued in country 3 or country 1. Similar adjutment are made to all VS and VS1 meaure. Thee adjutment uccinctly capture value added in production chain tretching acro multiple border. 25 But, conitent with the two-country cae, VS will be meaured along column and VS1 will be meaured along row. 26 Thi expreion i derived by iteratively applying the expreion for the invere of a partitioned matrix (ee appendix in Wang, Power, and Wei, 2009 for the derivation). 13

15 A before, the value-added hare can be applied to gro export to produce VAS_E. With three countrie, VAS_E can be meaured either with aggregate or bilateral trade. With total gro trade, VAS_E i the 3x3 matrix: V1B11E1 V1BE2 V1B13E3 VAS _ E VBE V2B21E1 V2B22E2 V2B23E3. (21) V 3B31E1 V3B32E2 V3B33E3 The ditinction between VS and VS1 i much clearer with three countrie than with two. The um of off-diagonal element along a column i the hare of foreign value-added embodied in a unit of a particular country export, which i the true meaure of VS. The um of offdiagonal element along a row provide information on the hare of a country value-added export embodied a intermediate input in third countrie export, which i the true meaure of VS1. In mathematical notation: DV r V r B rr E r. (22) DV meaure the total dometic value-added embodied in country r' export. VS V B E. (23) r r r r VS meaure other countrie value-added embodied in country r' export. VS1 r r V r B r E. (24) VS1 meaure country r' indirect value-added export through third countrie. Thi i the firt explicit derivation of the VS1 meaure provided in the literature. A noted above, it i traightforward to extend equation (21) to multiple ector. However, with multiple importer, exporter, and ector, dimenionality become a problem when decribing global value chain. In the reult ection of thi paper, we have thu aggregated acro indutrie to ummarize the poition of each country in global production network. 27 To provide ome comparion of upply chain for different type of good, we plit out final good from intermediate in export. A noted above, final good are more naturally applicable to the derivation of value-added flow, but total gro flow (final plu intermediate) 27 To diaggregate acro indutrie uing the notation above, replace the diagonal export matrix E with matrix Ê, the reulted value-added export matrix of ize G GN. Beide the dimenionality problem, there alo data limitation that may cat doubt on the diaggregated ector reult a we dicued in ection 3, o we decide do not report diaggregate reult in current paper but it i available upon requet. 14

16 better match official trade tatitic and the value-added meaure already developed in the literature. Within intermediate, we further divide thoe good that are conumed by the direct importer from thoe good that are proceed or finihed in one country and exported for conumption or further proceing in another. To do thi, we divided gro export into final good and intermediate flow a follow: E r Y r A r X Y r Final good Ar X Finihed and conumed in A r r X r export back to r A t r X t export to third countrie, (25) where X r denote country r output conumed in country. Such a gro export decompoition help ditinguih countrie that lie downtream in upply chain, i.e., countrie that directly upply the final demander of their product, from thoe that are uptream in the upply chain, largely upplying intermediate for later incorporation into final good. Finally, we normalize mot meaure of value-added trade by their correponding gro flow, to report value-added trade a hare of gro export. When we combine the mathematical definition of dometic value-added (DV) in equation (22) and the export decompoition in equation (25), we will be able to decompoe a country' dometic value-added (value-added export) into four part: DV r V r B rr E r V r B rr r Y r V r B rr r A r X V r B rr r The firt term i the value-added trade embodied in country r' final good export, which i conumed by countrie that directly import them; the econd term i the value-added embodied in country r' intermediate good ued by countrie to produce their own dometic final demand, which are alo conumed by the direct importer; the third term i country r' intermediate export to countrie producing intermediate or final good that are then hipped back home (VS1*). The lat term i the value-added embodied in country r' intermediate good ued by a econd country to produce intermediate or final good for a third country (VS1). A indicated by equation (26), the portion of DV that i aborbed by the directly importing country conit of dometic value in (i) final good export, and (ii) intermediate export that tranformed to final good and aborbed by the direct importer. The portion of DV not aborbed by the direct importer (i.e., VS1) conit of (i) intermediate input that are tranformed into final good and exported to a third country for conumption, and (ii) A r X r V r B rr t A r X t (26) 15

17 intermediate input that are ued in other intermediate and ent to a third country for further proceing. Thee four part of DV repreent an important apect of the poition of an exporter in a value chain: e.g., far downtream with a high hare of final good, or far uptream with a high hare of intermediate ubequently exported by other countrie in the chain. In a imilar manner, VS alo can be expreed a foreign value that i exported in final product or foreign value embodied in exported intermediate. Equation (26) make it poible to integrate the older literature on vertical pecialization with the newer literature on value added trade, thu completely decompoing gro export and connecting official gro tatitic to value-added meaure of trade. The vertical pecialization literature ha proven that gro export conit of dometic value added (DV) and foreign value added embodied in imported intermediate input (VS). Equation (26) how that dometic valueadded can be further broken down into additional component that reveal the ultimate detination of a country exported value added. Our complete decompoition of gro export i diagraphed in figure 1. Section 4 of thi paper will report the major breakout of the value-added in a country gro export to the world baed on thi figure, providing detail on the uptream or downtream poition of pecific countrie in global value chain. (Inert figure 1 here) 3. Databae Contruction 3.1 Overview A world production chain characterize the ditribution of value-added hare among countrie in a particular global indutry. To preciely define uch chain acro many countrie one need to be able to quantify the contribution of each country to the total value-added generated in the proce of upplying final product. A noted in the previou ection, IRIO table provide ufficient information to allow u to model global value-added generation at the indutry average level, but thee table are not available on a global bai, and in fact are rarely available at the national or regional level. The available global IO databae, uch a the GTAP Multi-Region Input-Output (MRIO) table, do not have enough detail on the cro-border upply and ue of good to be directly applied to upply chain analyi. To provide a workable dataet for our global value chain analyi, we contructed a global IRIO table for 2004 baed on verion 7 of the GTAP databae a well a upplemental 16

18 detailed trade data and the proportionality aumption decribed below. To reflect the reality and importance of Export Proceing Zone (EPZ) in developing economie and their role in global value-added trade and production network, 28 we alo incorporated an expanded Chinee IO table with eparate account for proceing export and a 2003 Mexican IO table with eparate dometic and maquiladora account. 29 We integrated the GTAP databae and the additional information with a quadratic mathematical programming model that (a) minimized the deviation of the reulting new data et from the original GTAP data, (b) enured that upply and ue balance for each ector and every country, and (c) kept all ectoral bilateral trade flow in the GTAP databae contant. The new databae cover 26 countrie and 41 ector and wa ued a the major data ource of thi paper. 30 IRIO table report imported intermediate ue coefficient that pecify country r ue in ector i of import from ector j from ource country. To etimate thee coefficient, we need to (i) ditinguih intermediate and final ue of import from all ource in each ector, and (ii) allocate intermediate good from a particular country ource to each ector it i ued within all detination countrie. We addreed the firt of thee deficiencie uing detailed trade data and United Nation (UN) Broad Economic Categorie (BEC), a decribed in ection 3.2 No additional information i available to properly allocate intermediate of a particular ector from a pecific ource country to it ue indutrie at the detination economy, however. Thu, ector j imported intermediate input of a particular product are initially allocated to each ource country by auming they are conitent with the aggregate ource tructure of that particular product. 31 The data requirement for final good in our analyi alo exceed the detail available in MRIO databae. In particular, GTAP provide information on the allocation of final good into 28 WTO reported that around 20% of developing country export are out of EPZ while the hare on the import ide i about 13% baed on balance of payment tatitic. In the period, China alone accounted for about 67% of all reported proceing export while Mexico repreent another 18 percent. See Maurer and Degain (2010) for more detail. 29 Proceing regime provide incentive to ue imported intermediate input, provided the reulting final good are entirely exported. Proceing export violate the aumption that export have the ame intermediate input intenity a dometic ue. Failure to account for proceing import can dramatically overtate the dometic content of export (Dean, Fung, and Wang, 2009). The ource of the Chinee proceing data i Koopman, Wang, and Wei (2008), and the Mexican table i from the Mexican tatitical agency Intituto Nacional de Etadítica, Geografía e Informática (INEGI). 30 The country claification and it concordance with GTAP region and ector are given in Table For example, if 20% of U.S. imported intermediate teel come from China, then we aume that each U.S. indutry uing imported teel a an intermediate input obtain 20% of it imported teel from China. Such an aumption ignore the heterogeneity of imported teel from different ource. In reality, 50% of the imported teel ued by the U.S. contruction indutry may come from China, while only 5% of the imported teel ued by auto maker may be Chinee. Further, thee proportion may be revered for teel imported from Japan. 17

19 three type of final demand (private conumption, government conumption, and invetment) but doe not ditinguih how import from pecific ource are ditributed to thee component. The initial allocation again aumed that final good import were allocated to thee component conitent with the aggregate upply tructure of thoe final good. After integrating the new data into GTAP with the quadratic programming model, the reulting dataet i ufficient to calculate value added generated by every country at each tage of the global value chain for each product. We next turn to the reult of thee calculation. 3.2 Ditinguihing import of intermediate input and final good Although the GTAP databae provide bilateral trade flow, it doe not ditinguih the ue of thee good. Our initial allocation of bilateral trade flow into intermediate and final ue i baed on the UN BEC method, determined by detailed trade tatitic at the 6-digit HS level from COMTRADE. Thi differ from the approache in Johnon and Noguera (2009) and Daudin, Rifflart, and Schweiguth (2009). Thee tudie alo tranfer the MRIO table in the GTAP databae into an IRIO table. However, they do not ue trade data to identify intermediate good in each bilateral trade flow. Intead, they applied a proportion method directly to the GTAP trade data; i.e., they made an additional aumption that the proportion of intermediate to final good i the ame for dometic upplied and imported product. The ue of end-ue categorie to ditinguih import by ue i becoming more widepread in the literature and avoid ome noted deficiencie of the proportional method. Feentra and Jenen (2009) ue a imilar approach to eparate final good from intermediate in U.S. import in their recent re-etimation of the Feentra-Hanon (1999) meaure of offhoring. Dean, Fung, and Wang (2009) have hown that the proportionality aumption underetimate the hare of imported good ued a intermediate input in China proceing trade. Norda (2005) tate that the larget indutrial countrie have a higher hare of intermediate in their export than in their import, while the oppoite i true for large developing countrie. Thee reult imply that the intermediate content of import differ ytematically from the intermediate content in dometic upply. Conitent with the literature, our tudy how that the proportional method may overetimate the hare of intermediate good in import of mot developed countrie and underetimate the proportion of final good in export of many developing countrie (uch a 18

20 China and Vietnam). BEC method thu provide a le ditorted picture of the value added ditribution in global value chain. Table 1a report the hare of intermediate good in each country' total export and import by the two different method of allocating gro trade flow into intermediate and final good. Column (2) and (3) how that for mot developing countrie, the BEC method produced a much lower intermediate hare in export. Thi indicate thee developing countrie (e.g., Vietnam, China, South Aia, Thailand, and Mexico) may export ubtantially more final good than i indicated by previou value-added databae. The exception are the natural-reource exporting countrie uch a Brazil, Ruia and the ret of the world the BEC method produced higher intermediate hare in their export. (Inert table 1a here) The pattern for import are more nuanced. Column (4) and (5) how that the BEC claification alo reult in higher hare of intermediate in ome developing countrie import (e.g., Mexico, India, Indoneia, and South Aia), indicating thee developing countrie may import ubtantially more intermediate product. At the national level, China appear to be an exception, becaue the BEC claification give a lower intermediate hare etimate for China import than the proportion method. However, when we ditinguih proceing and normal trade, we find that intermediate good make up a much higher hare for China proceing import than for it normal import (89 v. 71%), very imilar to Mexico (Table 1b). The difference at the national level between China and Mexico i due to the role of normal nonproceing ector in China i large than that in Mexico. (Inert table 1b here) The difference between the BEC and proportion method at the indutry level are much greater than the national aggregate. Table 2 illutrate the difference in U.S. import (econd panel) and Chinee export (third panel) of electronic machinery. Becaue the proportion method aign the intermediate hare in total aborption to all foreign input ource, the econd panel report the ame intermediate good hare (54% percent) for all ource. Since the BEC claification i applied to each bilateral trade flow, it allow for product compoition difference from different ource countrie. The BEC claification identified more final product in U.S. import from almot all Eat Aian developing countrie a well a Brazil and Mexico, but more intermediate input from developed countrie and natural reource exporter. Over all, the 19

21 proportion method allocate 7 percentage point more U.S. electronic machinery import into intermediate good than the BEC claification, which may be inaccurate given that the United State ha the world larget conumer electronic market. (Inert table 2 here) The third panel of table 2 report Chinee export of electronic machinery. The proportion method reult in ubtantially more Chinee intermediate good export to almot all countrie in our databae. In total, the hare of intermediate good in Chinee electronic machinery export i 21 percentage point higher under the proportion method than the BEC claification, which doe not reflect China poition a one of the larget exporter of conumer electronic in the world. A tatitical ummary of the difference in etimate of intermediate hare between the two method for all countrie and indutrie in the databae are reported in appendix table A2 and A3. The table include both the imple and trade-weighted mean difference, along with their tandard error for all bilateral merchandie trade flow. At the ector level, the proportion method reulted in lower intermediate good hare for primary and manufacturing product, but higher intermediate good hare in conumer and capital product. Therefore, depite it hortcoming, 32 we believe that the UN BEC claification provide a better identification of intermediate in gro trade flow than the proportion method. It provide a better row total for each block matrix of A r in the IO coefficient matrix A, thu improving the accuracy of the mot important parameter, the IO coefficient in an IRIO model. However, it till doe not olve how to properly allocate particular intermediate good imported from a pecific ource country to each uing indutry (the coefficient in each cell of a particular row in each block matrix A r till have to be etimated by proportion aumption). Thi allocation i epecially important to preciely etimate value-added by ource in a particular indutry, although it may be le critical at the country aggregate level becaue total import of intermediate from a particular ource country are fixed o miallocation will be canceled out. 32 The literature dicue that the hortcoming of the UN BEC claification, particularly it inability to properly identify dual-ue product uch a fuel, automobile, and ome food and agricultural product. 20

22 4. Reult 4.1 Complete decompoition of gro export Table 3 preent a complete decompoition of each country gro export to the world by value-added component and the main breakdown of dometic value added. It elect key etimate of value-added export from the VAS_E matrix, a pecified in equation (22) to (26) and applied to our databae of international production and trade flow (the IRIO table). The firt three column of reult decompoe gro export into three term that integrate the vertical pecialization and value-added trade literature: the value-added trade to gro trade (VAX) ratio from Johnon and Noguera (2009), VS from HIY 2001, and VS1* from Daudin et al. (2008). A decribed in ection 2, thee term preent the value added by dometic factor in the production proce, the foreign value embodied in imported intermediate input, and dometic value-added embodied in imported good that have returned after proceing abroad. Each of thee term ha been modified from it original definition to correctly pecify it hare of value added in a multi-country framework. Although thee element have been independently computed baed on different element in the VAS_E matrix, they um to exactly 100 percent of gro export, thu verifying that the decompoition i complete. Thi i the firt uch decompoition of gro export in a global etting. Column (2) give the hare of value-added export in gro export the VAX ratio propoed by Johnon and Noguera (2009). For example, on average, each dollar of gro export from Autralia and New Zealand incorporate 88 cent of dometic value added. Thi hare i quite high for two type of exporter: natural reource producer (uch a Autralia, Brazil, Latin America, and Ruia) and high-income countrie, particularly the EU and Japan. Column (3) report the hare of foreign value-added in gro export the VS hare in HIY For example, on average, each dollar of gro export from China include 36 cent of foreign content. Thi hare i high for Mexico, China, ASEAN countrie and Eat Aia tiger, but low for indutrial countrie and natural reource exporter. An important difference between dometic value-added hare and VAX ratio need to be mentioned, there i a portion of gro export not belong to a country value-added export but are a part of it dometic value-added generated from it production factor, which i the valueadded embodied in it intermediate export ued by a foreign country producing product that hipped back to the home country (therefore cannot counted a the home country value-added 21

23 export), ued a intermediate input, inveted or conumed there. Thi component of dometic value-added a a hare of each country gro export i lited in column (4). Thi i called VS1* in Daudin, Rifflart, and Schweiguth (2009). For mot countrie, thi hare i low (under 1%), but it i quite high for the United State (.4%) 33 and the EU (7.4%), and relatively high for Japan (2.9%) and the ret of the world (2.5%). The world average i 4.0%. To account for all ource of value added, our meaure of VS1* include both imported intermediate and final good, o we report higher value than Daudin et al. (world average of 1.8%), which include final good only. 34 (Inert table 3 here) 4.2 Decompoition of dometic value added in gro export Different region of the world have harply different compoition of dometic value added embodied in export. Broadly, dometic value added in export can be apportioned to the value in final good, intermediate export aborbed by the direct importer, and intermediate export proceed and exported to further countrie (i.e., VS1). 35 VS1 can be further divided into value that i re-exported in final good and value re-exported in proceed intermediate good. Column (5) through (8) in table 3 decompoe total dometic value added in gro export into thee four component, i.e., value-added embodied in final good export and three type of intermediate-good export. Thi decompoition provide a more detailed look at valueadded export than ha previouly available in the literature, particularly Daudin, et al. Column (5) give the hare of value added incorporated in export of final good, relative to total gro export, which correpond to the firt term in equation (26) divided by gro export. Not urpriingly, natural reource exporter uch a Ruia have little of their value added in final good, becaue thee countrie produce relatively few final good. 36 Value-added embodied in final good contitute a large hare of mot Aian countrie' value-added export. For example, more than half of the value-added export come from final product in Vietnam, South Aia, 33 Thi may explain why Johnon and Noguera (2009)' VAX ratio i much maller than 1 minu the VS hare in HIY for the United State, but very cloe for other countrie, becaue they are different by definition. 34 If only final good are taken into account, our etimate i 1.9%, conitent with Daudin et al. 35 Correponding to column 5, 6, and 7+8 in table However, value added per dollar of gro export i imilar for export of final good and intermediate good from thee countrie. See dicuion below. 22

24 and proceing regime in China (a well a Mexico), indicating that thee countrie are located in the end of world value-added production chain. Dometic value-added in intermediate good export i plit out in the next three column. Column (6) report the hare of value added embodied in exported intermediate good that by the direct importer to produce final good and inveted or conumed (i.e., aborbed) there. Thi column correpond to the econd term in equation (26) divided by gro export. All the natural reource exporter generate about half of their value-added export through thi channel. For example, about 55% of value-added export from Ruia and the ret of the world (largely oil producing countrie) come from exporting intermediate directly to a detination country which ue thee intermediate input produce final good to atify their dometic demand. In contrat, mot Aian economie except Indoneia have a ubtantially le valueadded embodied in uch intermediate export (le than 25%). Column (7) and (8) how the hare of value added embodied in exported intermediate good that are ubequently exported a intermediate or final good and ued in a third country, which correpond to the lat term in equation (26) divided by gro export. Mot of Eat Aia countrie except Vietnam and proceing China produce a large hare of value-added in export via uch channel. Intermediate good ent to another country to produce intermediate or final good for export account for a high proportion of total valueadded export for much of Eat Aia, including Taiwan, the Philippine, Malayia, Korea, Indoneia, Thailand, Singapore, and Japan. Eat Aian countrie thu engage in longer production network and are located in the middle of the production chain, providing a large hare of manufactured intermediate to both advanced and emerging economie. Thi reult i conitent with cae tudie that have examined upply chain for electronic (Baldwin and Evenet, 2009) and automobile (Nag, 2007) in China. 37 Japan i an outlier among developed countrie. A paper uch a Dean, Lovely, and Mora (2009) have hown, a high hare of Japanee export are proceed in China and are then ent a finihed good to developed countrie uch a the United State. A comparion of Canada and Mexico i intructive about the difference in upply network preented in table 3. Both countrie end three-quarter or more of their export to the neighboring United State, and both can take advantage of nearly free trade with the United 37 See Baldwin and Evenet (2009) for dik drive, and Nag (2007) for automobile. 23

25 State under NAFTA. Analyi of value-added trade, however, how that thee countrie participate with ditinctly different upply chain. Of the two countrie, Canada ha much higher dometic value-added hare in it gro export (71% v. 51%). A with China, extenive proceing trade in Mexico generate a low dometic value-added hare of export. Column (6) how a major difference between the two countrie value-added export: Mexico end a much lower hare of export that are, at the end of the upply chain, finihed and conumed in the ame country (33%/53%). Thi implie that many Mexican good, if finihed in the United State, may not ued there, but ent to a third country for further proceing or final aembly and on to a different final detination. Thi ultimately implie that Mexican export proceed through a more lengthy upply network than Canadian export. Canadian firm may alo be part of lengthy chain, but they are further downtream in upplying product to the United State. The value-added decompoition contained in table 3 i quit rich, but may not be eaily aborbed by a reader unfamiliar with thee different value-added concept. Figure 2 diplay thee component by country group, including advanced economie, Aian NIC, emerging Aia, and other emerging economie. 38 It i clear from the figure that Aian countrie have much lower total dometic value added in export than other imilar region. Aian NIC have lower dometic value added than the advanced economie, and hare in emerging Aia are lower than in other emerging countrie. Depite their large overall hare, Aian countrie have lower hare of value added aborbed by the direct importer than their counterpart. Among the emerging economie, Mexico and the EU acceion economie appear mot imilar to the Aian countrie. Thee two region are ditinguihed, however, by the large hare of final good and intermediate aborbed directly by their large immediate neighbor. (Inert figure 2 here) Figure 3 compare thee component in another way, by examining each component hare of dometic value-added export rather than gro export. The figure relie on a property of an equilateral triangle, that the um of the perpendicular ditance to each ide i the ame for all point in the triangle and i equal to the height of the triangle. 39 Becaue the triangle can compare only three item at a time, we collape the two component of VS1 back into a ingle hare. In thi cae, the height of the triangle i 100, becaue the three trade hare um to Group defined by the IMF at 39 See Ley (1996) for an application of thi diagram (the Kolm triangle) to the proviion of public good. 24

26 percent of dometic value-added export. Each country i plotted in the triangle with the ditance to each ide correponding to the hare of each component in value-added export. For example, Ruia ha the lowet hare of final good in dometic value-added export (9.5/89.1=10.6 percent), o it lie cloet to the bottom edge of the triangle. Ruia alo ha the highet hare of intermediate aborbed by the direct importer (49.1/89.1=55.1 percent), o it lie the farthet from the left edge of the triangle. Region are repreented by color, with emerging and newly-indutrialized Aian countrie in light blue, other emerging countrie in green, and advanced countrie in red. The figure how the tark difference between Aian and non-aian emerging economie. Emerging Aia ha much lower hare of export aborbed by the direct exporter. Thee countrie are much more likely to be involved in longer upply chain, in which export are proceed and exported by multiple countrie. The Philippine are the mot extreme example of thi phenomenon. Alo we ee that primary good exporter, uch a Ruia, the ret of the world (largely the Middle Eat), Indoneia, and Latin America tend to have much lower hare of dometic value added in final good than other region. The advanced economie tend to lie in the middle of the triangle, although there are ome difference. Canada, with it proximity to the United State, ha relatively low export through third countrie, while a much higher hare of Japan export go through multiple border. (Inert figure 3 here.) 4.3 Decompoition of foreign value added in gro export Column (9) and (10) in table 3 further divide the foreign value-added hare in gro export into intermediate and final good. Thee hare correpond to the um of off-diagonal element along each column of VAX_E matrix, given by equation (23). Mot Aian developing countrie (China, Vietnam, Thailand, South Aia, and the ret of Eat Aia), a well a Mexico and EU acceion countrie ue ubtantial amount of imported content to produce final good export, while mot developed countrie and natural reource exporter ue imported value largely in the production of intermediate export. Column (11) and () how the dometic value-added hare in export for final good and intermediate good eparately. While natural reource producer and highly developed countrie both have high hare of dometic value added in their export, column (11) and () 25

27 how ubtantial difference between thee group. Natural reource exporter tend to have very imilar hare of value added in final and intermediate good. Highly developed countrie, however, generally have much higher dometic value added per dollar of final good export than per dollar of intermediate good export. The United State ha a particularly large difference between value added embodied in final good and the value added in intermediate. Converely, normal Chinee and Mexican export embody much more value added per dollar of intermediate good than final good. Thee difference are mainly caued by compoition difference in thee countrie' export in intermediate and final good ince we do not differentiate the value-added content at the ector level for final and intermediate good. The mot notable feature of U.S. value-added trade appear in column (3) and (13), which correpond to the third term in equation (26) divided by each country total gro export and import repectively. The United State ha by far the highet hare of it own value-added export (8.3%) that return embodied in good it import. Thi high hare certainly reflect the large U.S. market ize, but it likely alo reflect the U.S. role a a key upplier of value added in many advanced product that it conume. 4.4 Poition of countrie within value chain: Evidence from broad ector Table 4 report the broad ector content of dometic value-added (DV), foreign valueadded (VS), and value-added exported indirectly through third countrie (VS1). Two type of countrie have high dometic value-added hare in their total export. Countrie that export much of their value added in raw material ector have the highet value-added hare. Countrie that export much of their value added in ervice, including Hong Kong, the EU, and the United State, alo have relatively high overall dometic value-added hare, but lower than the natural reource exporter. Converely, countrie that export mot of their value added in manufacturing ector, uch a Taiwan, Mexico, and the Philippine, all have quite low dometic value-added hare in their total export. (Inert table 4 here) Table 4 alo report the broad ector decompoition for VS1, value-added exported indirectly through third countrie. 40 At the global level, manufacturing account for about one- 40 VS1 i given by the um of column (4), (7), and (8) in table 3. Thi correpond to the um of off-diagonal element along each row of VAX_E matrix, given in equation (25). 26

28 half of export that are ent indirectly through third countrie, and raw material account for another one-third. Countrie with indirect export may be highly uceptible to global hock (Bem, Johnon, and Yi, 2010). Thi may be particularly true for thoe with high value of indirect export in manufacturing (e.g., the Philippine, Japan, Taiwan, and Korea). Thee countrie uffered major production and export downturn in 2008, even though ome of their major direct export detination (uch a China) were le affected by the global lowdown. It i difficult to find one meaure that ummarize difference in vertical trade acro countrie. The VS1/VS ratio ha been ued in thi regard in previou tudie (Daudin et al., 2009). At global level, VS1 and VS equal each other, therefore, the average VS1/VS ratio i equal to 1. A ratio larger than 1 indicate the country lie uptream in the global value-chain, either by providing raw material (uch a Ruia) or by providing manufactured intermediate (uch a Japan) or both. 41 A ratio le than one mean the country lie downtream in the global value-chain, uing more intermediate input from other countrie to provide final good. Column (14) in table 4 how that widely different countrie have high VS1/VS ratio, including both emerging market and highly developed economie. For intance, Ruia, Japan, Autralia and New Zealand, United State, the EU 15, the ret of the world, Brazil, the ret of Latin America, Indoneia, and South Africa all have a ratio larger than one. While South Aia, Thailand, Canada, EU acceion countrie, Vietnam, China, and Mexico have the lowet VS1/VS ratio (le than 0.5). Countrie like Hong Kong, Korea, Taiwan, Malayia, Philippine and EFTA have intermediate VS1/VS ratio, indicating they both ue other countrie intermediate input to produce final product, and they provide natural or manufactured intermediate input in the world production chain. Further detail on the type of good that countrie provide to value chain allow u to eparate thee two type of countrie. Column (15) to (17) how that countrie with higher than average VS1/VS fall into one of two ditinct group. The mot uptream emerging economie (e.g., Ruia and Indoneia) produce primary product for global upply chain. The uptream advanced economie produce both manufactured good and ervice for thee chain. (Almot uniformly, countrie with high VS1 in manufacturing have high VS1 in ervice a well.) Only a 41 Uptream here refer to the amount of value added provided relative to the amount received. Uptream in other literature may denote the extent of primary product production. A we will ee, ome countrie are uptream in both ene. 27

29 few countrie (Autralia, Brazil, and Ruia) appear uptream in both primary and non-primary product. 42 Although the VS1/VS ratio i familiar from the literature, the pecific meaure of dometic value added in table 3 and 4 are better decriptor of country poition in global chain. Firt, thi i true a a matter of mathematic, becaue the VS1/VS ratio i largely dependent on dometic value added by definition. 43 Second, the component of dometic valueadded provide more nuanced vie of each country poition. For example, it can be clearly een from table 3 that countrie near large market (e.g., Canada, the EU acceion countrie, and Southeat Aia) provide particularly high hare of their good directly to their final conumer rather than through countrie that proce and re-export product. A another example, figure 4 provide an alternative view of countrie poition uing the dometic value added export hare in table 4 (plit between raw material and other ector) a well a foreign value-added hare(including both VS and VS1*). Becaue thee hare um to 100 percent for each country, the triangle diagram i again appropriate. The figure clearly ditinguihe the advanced economie from the raw material exporter. Mot advanced countrie and all of the Aian NIC have very low export of primary material, and appear on the left edge of the triangle (with the major advanced economie toward the bottom). Non-Aian emerging market have very low foreign value added content in their export, and are located toward the bottom of the diagram. Emerging Aian countrie appear toward the middle. (Inert figure 4 here) 4.5 Source of value added in final good export Table 5 detail ource of value added in each country export of final good. For example, 88.1% of the value added in Autralia and New Zealand final good i ourced dometically. Thi hare correpond to column (11) in table 3, and i high relative to the world average. 44 Of the remaining % of foreign value added, table 5 how that only the EU and the US contribute more than 2% of value each in Autralia and New Zealand final good export. 42 Note that thee three region alo have a very low VS hare in column (9) of table A i clear in table 3, VS1 i a component of dometic value added, and VS i cloe to 100 minu dometic value added for mot countrie. Thu the ratio mut alo depend largely on dometic value added. The correlation coefficient between dometic value-added export and the VS1/VS ratio i Autralia and New Zealand have relatively high income and relatively large hare of value added in natural reource, both of which tend to increae dometic value-added hare. 28

30 The hare illutrate ome urpriing contrat acro region. Comparing Mexico and EU acceion countrie, for example, how that Mexico ha much lower dometic content and much higher value added from it large neighbor than the EU acceion countrie. Thi illutrate Mexico deep integration into North American production and the influence of Mexican proceing export. Thi contrat alo illutrate the importance of incorporating a eparate IO account for Mexican proceing trade into our databae. 45 (Inert Table 5 here) While table 5 i ueful to denote ource for particular countrie, it can alo ditinguih ourcing acro countrie. The final column preent each country Herfindahl Hirchman Index (HHI) for value-added hare. The EU, Japan, and the United State have the highet value, indicating high reliance on dometic ource of value-added for thee highly developed region. 46 In contrat, much of developing Aia ha far lower indexe, revealing a broader (or deeper) ource of international network for input. All countrie have much more concentrated ourcing than the world average, indicating trong home and regional biae in global upply chain. V. Concluion and Direction for Future Work We have developed a ingle, unified framework for vertical pecialization and valueadded trade baed on a tranparent conceptual framework including many countrie uing an international IO model. Thi new framework incorporate all previou meaure of value-added trade in the literature and adjut for the back-and-forth trade of intermediate acro multiple border prevalent in modern international production network. Uing thi meaure, we can completely decompoe gro export into it dometic and foreign content and further decompoe dometic value-added export into different value-added component that reveal a country uptream and downtream poition in a global value chain. While the analyi of value-added trade flow in an IRIO table i naturally applicable to final good trade, the 45 The abence of a eparate proceing IO table for emerging Europe may be underetimating the hare of developed EU value added in the region, particularly in ector uch a vehicle. 46 The EU HHI would likely drop if we better diaggregated European countrie in our region. The U.S. HHI might alo drop if we diaggregated Latin America. Japan core i unlikely to change much with further diaggregation, however, becaue much of Aia i included eparately. We plan to reviit thee core uing the full diaggregation available in the GTAP dataet. 29

31 decompoition of gro trade flow align our meaure with official trade tatitic, and provide a better comparion with exiting meaure of vertical pecialization in the literature. We have contructed a global IRIO table for 2004 baed on verion 7 of the GTAP databae to empirically implement the conceptual framework. We make a refinement to the databae by uing the UN BEC claification method to ditinguih final and intermediate good in gro trade flow. The reulting dataet appear le ditorted than the reult produced under the widely ued proportion method. Thi paper ha provided the mot complete decription to date of the relationhip between gro trade and it value added component. It ha highlighted important relationhip within major regional upply network (the EU, North America, and Aia). It ha alo demontrated ubtantial difference between the region, with deeper integration apparent in Aia and North America. Many meaure how that countrie in Emerging Aia have different role in global upply chain than other emerging economie. Their export pa through more border than export from other emerging market. Much of Aia, including China itelf, end it export to countrie that provide final aembly on behalf of conumer in other countrie. Aian countrie alo have relatively dipered ourcing of imported intermediate input, and are le likely to be ource of raw material. In contrat, other developing countrie, which tend to lie either further uptream or further downtream in global upply chain, more commonly end their intermediate directly to the country of final conumption, and have much more concentrated ourcing network. The contribution of thi paper lie largely in it comprehenive framework, it approach to databae development, and the new detailed decompoition of value-added trade that ha revealed about each country role in global value chain. The creation of a databae that encompae detailed global trade in both gro and value-added term, however, will allow u to move from a largely decriptive empirical exercie to analyi of fragmented trade itelf in future work. In the dicuion in thi paper, clearly country ize and proximity to large market greatly affect the way in which countrie engage in global upply chain. In future work we plan to examine the extent to which thee economic force, which have been noted and meaured in gro trade for decade, apply to value-added export. 30

32 In addition, there may be fruitful inight gained on other related topic. While thi paper decompoe dometic value added (DVA) in export for jut one year, the reult for overall DVA and it compoition may add inight on whether a country role in global value chain affect growth. In thi ingle-year decompoition we ee that the pattern of overall DVA in export varie ubtantially among fat growing developing countrie uch a China, Brazil, and India. Thee are more imilar in the ectoral compoition of their DVA with Brazil and India having much higher DVA contribution from ervice than China, but all have imilar (and only moderate) DVA contribution from manufacturing, depite China reputation for being predominantly a manufacturing and aembly hub. In contrat, Japan, a country with particularly high DVA in manufacturing export, ha had low economic growth for decade. Evidence i accumulating that a country role in global upply chain can affect the variability of trade and national income (Bem, Johnon, and Yi, 2010). Might the poition in global upply chain alo affect the rate of growth? We leave thi for future reearch. 31

33 Appendix Section 2 derive our vertical pecialization meaure uing matrix notation applicable to any number of countrie and indutrie. While fully generalizable, thi notation doe not refer to IO coefficient for any pecific national or world IO account. Thi appendix provide calculation for each ub-element of the A, B, and VAS, DV, VS, and VS1 matrixe, for thoe perhap more familiar with calar um of the coefficient themelve. Aume there are G countrie, with N indutrie in each country. A world IO table provide a comprehenive account of annual product and payment flow within and between countrie. We ue the following notation to decribe the element of the world IO table (expreed in annual value): x r i i gro output of indutry i in country r; r v i i direct value added by production of indutry i in country r; r zij denote delivery of good i produced by country and ued a an intermediate by ector j in country r; and y r ik denote delivery of good i produced in country for final ue in final demand type k in country r. The number of final demand type, uch a private conumption or gro capital formation, i H. The following accounting identitie decribe the relationhip among element of each row (i, r) and column (j, ) of the international IO table: G N G r z ij 1 j=1 1 G N r1 i1 z r ij v H r r y = x (A.1) j = x j k=1 ik i (A.2) The two equation have traightforward economic interpretation. A typical row in equation (A.1) tate that total gro output of commodity i in country r i equal to the um of all deliverie to intermediate and final uer in all countrie (including itelf) in the world. Equation (A.2) define the value of gro output for commodity j in production country a the um of the value from all of it (dometic plu imported) intermediate and primary factor input. Equation (A.1) and (A.2) mut hold for all i, j N, k H and, r G in each year. Define rr ij rr zij rr x j a a the direct input coefficient of dometic product of country r, r zij rr x j a, r, a intermediate input/output coefficient of good i produced in ource country r ij 32

34 for ue in ector j by detination country r; and av j v j x j a each ector j ratio of direct value added to gro output for each producing country. Uing matrix notation, equation (A.1) and (A.2) could be re-written a equation (4): 1 X ( I A) Y BY, (A.4) where A i a GNxGN quare matrix with G 2 block ubmatrice of ize NxN. Although the analytical olution for the block matrix invere i too cumberome when the number of countrie exceed three, we can define each element of the block invere matrix B [ r ] where the r b ji upercript and r denote ource and detination country repectively, and ubcript i and j denote the ue and upply indutry repectively. Let u further define V 1 j n av av av a a 1xGN vector of direct value-added. matrix Share of Our baic meaure of value added in a global production network i the GxGN VAS V1B11... V1B1 G N r r VAS : V B : [ vai ] av jb ji (A.3) j1 VG BG 1... VG BGG The dometic value-added hare in the ource country total export i given by DV VB E ue vai ei N e i1 i N i1 N j1 N i1 av b e j i e ji i (A.4) which are the diagonal element of the VAS matrix weighted by the tructure of the ource country export. Sectoral gro export are given by Share of VS G e i e r1 r i rr, where e 0. The foreign value-added hare in the ource country total export (VS) become G r V B ue r E G va r N i1 e r i e r i G N N r i1 j1 N i1 av e i j b r ji e r i i. (A.5) which i the um of the off-diagonal element of the ource country column in the VAS matrix weighted by the ource country export tructure. 33

35 The true meaure of VS1, which i value-added embodied in exported intermediate that are proceed and ent to their final detination through third countrie, become G G r r r r r V B r E va e av jb ji e i i r r r i1 j1 Share of VS1 N N ue e e. (A.6) i1 i G N Thi hare i the um of the off-diagonal element of the ource country row in the VAS matrix weighted by the export tructure of each country (excluding the ource). Thee meaure could alo be defined at diaggregate level, for each ource or detination country and for each indutry. For example, for a particular indutry N i1 i Share of DV i va e i i N j1 av b e j i e ji i (A.7) Share of VS i G r va e i r i G r N j1 av b e i j e r r ji i (A.8) At a diaggregate level, however, VS1 may not be expreible a a hare of a country export, ince the country may not have direct export of the particular ector or direct export to a particular partner country. If a ector had zero direct export but poitive indirect value-added export via third countrie, the hare of VS1 in export would be infinite. 34

36 Reference Athukorala, Prema-chandra and Nobuaki Yamahita, 2006, Production Fragmentation and trade integration: Eat Aia in a global context North American Journal of Economic and Finance, 17: Bem, Rudolf, Robert Johnon, and Kei-Mu Yi, 2010, The Role of Vertical Linkage in the Propagation of the Global Downturn of 2008, Mimeo Prepared for IMF/Pari School of Economic Conference, January. Chen, Hogan, Matthew Kondratowicz, and Kei-Mu Yi, 2005, Vertical pecialization and three fact about U.S. international trade, North American Journal of Economic and Finance, 16(1): Daudin, Guillaume, Chritine Rifflart, and Danielle Schweiguth, 2009, Who Produce for Whom in the World Economy? OFCE Working Paper, Science Po Pari, July. Dean, Judith, K.C. Fung, and Zhi Wang, 2008, How Vertically Specialized i Chinee Trade? USITC Working Paper No D, September. Dean, Judith, Mary Lovely, and Jee Mora, 2009, "Decompoing China Japan U.S. trade: Vertical pecialization, ownerhip, and organizational form" Journal of Aian Economic 20(6): Dedrick, Jaon, Kenneth L. Kraemer, and Greg Linden, 2008, Who Profit from Innovation in Global Value Chain? A Study of the ipod and notebook PC, Paper preented at the Sloan Indutry Studie Annual Conference, Boton, MA, May. Feentra, Robert and J. Bradford Jenen, 2009, Evaluating Etimate of Material Offhoring from U.S. Manufacturing, prepared for the National Academy of Public Adminitration conference Meaurement Iue Ariing from the Growth of Globalization November 6-7, Feentra, Robert, Robert Lipey, Lee Brantetter, C. Fritz Foley, Jame Harrigan, J. Bradford Jenen, Lori Kletzer; Catherine Mann, Peter Schott, Greg C. Wright, "Report on the State of Available Data for the Study of International Trade and Foreign Direct Invetment." NBER Working Paper Hummel, D., J. Ihii, and K. Yi, 2001, The Nature and Growth of Vertical Specialization in World Trade, Journal of International Economic 54: Johnon, Robert, and Guillermo Noguera, 2009, Accounting for Intermediate: Production Sharing and Trade in Value-added, Mimeo, Princeton Univerity, May. 35

37 Koopman, Robert, Zhi Wang, and Shang-Jin Wei, 2008, How Much Chinee Export I Really Made in China Aeing Foreign and Dometic Value-added in Gro Export, NBER Working Paper Koopman, Robert, Zhi Wang and Shang-Jin Wei, 2009, A World Factory in Global Production Chain: Etimating Imported Value Added in Export by the People Republic of China, CEPR Dicuion Paper in International Trade and Regional Economic, No. 7430, forthcoming in Robert Barro and Jong-Wha Lee, ed, Cot and Benefit of Economic Integration in Aia, Oxford Univerity Pre, forthcoming Ley, Eduardo, 1996, On the Private Proviion of Public Good: A Diagrammatic Expoition, Invetigacione Económica, 20:1, Maurer, Andrea and Chritophe Degain, "Globalization and Trade Flow: What you ee i not what you get!" Staff Working Paper ERSD-2010-, World Trade Organization. Norda, Hildegunn, 2005, International Production Sharing: A Cae for a Coherent Policy Framework, Dicuion Paper No. 11, World Trade Organization. Pula, Gabor and Tuoma Peltonen, 2009, Ha Emerging Aia Decoupled? An analyi of production and trade linkage uing the Aian International Input-Output Table, Working Paper No. 993, European Central Bank. Wang, Zhi, William Power, and Shang-Jin Wei, 2009, Value Chain in Eat Aian Production Network, USITC Working Paper No C, October. Wang, Zhi, Marino Tiga, Jee Mora, Li Xin, and Daniel Xu, 2010, A Time Serie Databae for Global Trade, Production, and Conumption Linkage, Mimeo Prepared for GTAP Conference on Global Economic Analyi, Penang, Malayia, June. Yi, Kei-Mu, 2003, Can Vertical Specialization Explain the Growth of World Trade? Journal of Political Economy, 111(1):

38 Figure 1 Decompoition gro export into it major value-added component 37

39 Figure 2 Compoition of value-added export, by country and region Autralia, New Zealand Japan EU 15 United State EFTA Canada Hong Kong Korea Taiwan Singapore India South Aia Ret of Eat Aia Indoneia China Vietnam Thailand Malayia Philippine Ruian Federation Brazil Ret Latin America Ret of the world South Africa EU acceion countrie Mexico World average Final good Intermed' aborbed by direct importer Intermed' reexported a intermediate Intermed' reexported a final good 38

40 Figure 3 Intermediate and final good, hare of value-added export Intermediate aborbed by direct importer S_ASIA VNM CHN THA EU KOR JPN E_ASIA USA EU_NEW MEX HKG IND TWN CAN PHL BRA SGPAUS EFTA MYS ZAF LATAM IDN Intermediate exported to third countrie ROW RUS Final good Note: For any country, the perpendicular ditance to each ide give the relevant hare in dometic value added export. The um of perpendicular ditance for any point in the triangle i

41 Figure 4 Dometic and foreign value-added hare of gro export Dometic value: Raw material SGP MEX Dometic value: Manufacturing and ervice TWN PHL MYS THA CHN KOR EU_NEW CAN HKG USA EFTA S_ASIA E_ASIA IND EU ZAF JPN BRA AUS VNM IDN LATAM RUS ROW Foreign value (including VS1*) Note: For any country, the perpendicular ditance to each ide give the relevant hare in gro output. The um of perpendicular ditance for any point in the triangle i

42 Table 1a Ditinguihing import of intermediate input from final good Proportion and BEC method, 2004, in percent of gro trade Country Share of intermediate in export Share of intermediate in import Share of Value-added in Gro export Proportion BEC Proportion BEC Proportion BEC (2) (3) (4) (5) (6) (7) Autralia, New Zealand Brazil Canada China EU acceion countrie EU Hong Kong Indoneia India Japan Korea Mexico Malayia Philippine Latin America & Carib. EFTA Ret of the world Ruian Federation Singapore South Aia Thailand Taiwan United State Vietnam Ret of Eat Aia South Africa World Source: Author etimate baed on verion 7 GTAP databae and UN BEC claification. 41

43 Table 1b Ditinguihing import of intermediate input from final good Normal and proceing trade in China and Mexico, 2004, in billion of dollar Country Gro export Share of intermediate in export Gro import Share of intermediate in import Share of Valueadded in Gro export China normal China proceing Total Mexico normal Mexico proceing Total Source: Author etimate by combine verion 7 GTAP databae and China 2002 IO table from Koopman, Wang and Wei and Mexico 2003 IO table from Intituto Nacional de Etadítica, Geografía e Informática (INEGI). Table 2 Ditinguihing import of intermediate input from final good Difference between the proportion and BEC method in electronic machinery trade, China and the United State, 2004 US electronic machinery import China electronic machinery export Country Proportion BEC Difference Value a Proportion BEC Difference Value a Autralia & New Zealand ,882 Brazil Canada , ,255 China ,357 EU , ,281 EU , ,018 Hong Kong ,382 Indoneia , India ,530 Japan , ,088 Korea , ,857 Mexico , ,698 Malayia , ,205 Philippine , ,772 Ret of America ,409 Ret of High Income countrie REST OF THE WORLD , ,824 Ruian Federation Singapore , ,904 South Aia Thailand , ,081 Taiwan , ,085 United State of America ,357 Viet Nam Ret of eat aia South Africa World , ,587 a Value of trade (not difference), in million of 2004 U.S. dollar. 42

44 Table 3 Decompoition of each country gro export Country Dometic valueadded = (5)+(6) +(7)+(8) Total Foreign valueadded = (9)+(10) Value added that return to ource a a % of gro export Channel of dometic value-added in export (hare of each value added component in total gro export) Exporte d final good Direct export of intermed to produce final good and conume there Export of intermed to third country producing intermediate good for another det Export of intermed to third country producing final good for another det. Share of foreign valueadded in gro export % Export Export of of final intermmed good good Dometic value-added hare % Dometic Dometic valueadded value- in added in final intermedi good ate export export Value added that return to ource a a % of gro import (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) () (13) Autralia, New Zealand Brazil Canada China EU acceion countrie EU Hong Kong Indoneia India Japan Korea Mexico Malayia Philippine Ret Latin America EFTA Ret of the world Ruian Federation Singapore South Aia Thailand Taiwan United State Vietnam Ret of Eat Aia South Africa World average Source: Author etimate. Note: a. Column (2) + (3)+(4) =100; b. Column (4)+(7) +(8) = HIY VS1. C. Column (4) equal the third term in equation (24) divided by a country gro export; Column (5) equal the firt term in equation (24) divided by a country gro export; column (6)+(7) equal the econd term in equation (24) divided by a country gro export; column (8) equal the lat term in equation (24) divided by a country gro export and um of column (5) to (8) equal column (2). Column (13) equal the third term in equation (24) divided by the country final good import. Column (11) are the firt term in equation (24) divided by a country final good export, while column () are the um of the econd, third, and lat term in equation (24) divided by a country intermediate export. 43

45 Table 4 Decompoition of value-added export from major ector, hare of gro export Country/Region DV: Sector generating dometic value-added Raw material Service Total Raw material VS: Sector content of embodied foreign value-added Service Total Raw material VS1: ector ouce of valueadded in intermediate exported via third countrie Service Total Total Raw material VS1/VS ratio Manufacturing Manufacturing Manufacturing Manufacturing Service (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) () (13) (14) (15) (16) (17) Autralia, New Zealand Brazil Canada China EU acceion EU Hong Kong Indoneia India Japan Korea Mexico Malayia Philippine Latin Am. And Carib EFTA Ret of the world Ruian Federation Singapore South Aia Thailand Taiwan United State Vietnam Ret of Eat Aia South Africa Average Source: Author etimate. 44

46 Table 5 Source of value-added in final good export Country/Region AUS BRA CAN CHN EU acce EU HKG IDN IND JPN KOR MEX MYS PHL Latin South Am. EFTA ROW RUS SGP Aia THA TWN USA VNM Ret E. Aia ZAF HHI a Autralia, NZL Brazil Canada China EU acceion EU Hong Kong Indoneia India Japan Korea Mexico Malayia Philippine Latin Am EFTA Ret of world Ruian Fed Singapore South Aia Thailand Taiwan United State Vietnam Ret of E. Aia South Africa Average a Herfindahl-Hirchman Index divided by

47 Appendix Table A1 Countrie in databae and correponding GTAP region Country or region Correponding GTAP region Autralia, New Zealand Autralia, New Zealand Brazil Brazil Canada Canada China China China normal N/A China proceing N/A EU acceion Bulgaria, Cypru, Czech Republic, Etonia, Hungary, Latvia, Lithuania, Malta, Poland, Romania, Slovakia, Slovenia EU 15 Autria, Belgium, Denmark, Finland, France, Germany, Greece, Ireland, Italy, Luxembourg, Netherland, Portugal, Spain, Sweden, UK Hong Kong Hong Kong Indoneia Indoneia India India Japan Japan Korea Korea Mexico Mexico Mexico normal N/A Mexico proceing N/A Malayia Malayia Philippine Philippine Latin America and the Caribbean Argentina, Bolivia, Caribbean, Chile, Colombia, Cota Rica, Ecuador, Guatemala, Nicaragua, Panama, Paraguay, Peru, Ret of Central America, Ret of North America, Ret of South America, Uruguay, Venezuela EFTA Switzerland, Norway, Ret of EFTA Ret of world Albania, Armenia, Azerbaijan, Belaru, Botwana, Central Africa, Croatia, Egypt, Ethiopia, Georgia, Iran, Ilamic Republic of, Kazakhtan, Kyrgyztan, Madagacar, Malawi, Mauritiu, Morocco, Mozambique, Nigeria, Ret of Eatern Africa, Ret of E. Europe, Ret of Europe, Ret of Former Soviet Union, Ret of North Africa, Ret of Oceania, Ret of South African Cutom Union, Ret of Wetern Africa, Ret of Wetern Aia, Senegal, South Central Africa, Tanzania, Tuniia, Turkey, Uganda, Ukraine, Zambia, Zimbabwe Ruian Federation Ruian Federation Singapore Singapore South Aia Bangladeh, Pakitan, Ret of South Aia, Sri Lanka Thailand Thailand Taiwan Taiwan United State United State Vietnam Vietnam Ret of Eat Aia Cambodia, Lao, Myanmar, Ret of Eat Aia, Ret of Southeat Aia, South Africa South Africa South Africa 46

48 Appendix Table A2 Ditinguihing import of intermediate input from final good The difference between the proportion and BEC method, by country, 2004, in percent Export Import Simple average Weighted Average Simple average Weighted Average Country N MEAN STD MEAN STD N MEAN STD MEAN STD Autralia & New Zealand Brazil Canada China EU EU Hong Kong Indoneia India Japan Korea Mexico Malayia Philippine Ret of America Ret of High Income countrie Ret of the world Ruian Federation Singapore South Aia Thailand Taiwan United State of America Viet Nam Ret of Eat Aia South Africa World 16, ,

49 Appendix Table A3 Ditinguihing import of intermediate input from final good The difference between the proportion and BEC method, by ector, 2004, percent Simple average Weighted Sector N MIN MAX Average MEAN STD MEAN STD Crop production Animal hubandry Foretry Fihing Coal Oil and ga Mineral nec Meat and Dairy product Food product nec Beverage and tobacco product Textile Wearing apparel Leather product Wood product Paper product publihing Petroleum coal product Chemical rubber platic product Mineral product nec Ferrou metal Metal nec Metal product Motor vehicle and part Tranport equipment nec Electronic equipment Machinery and equipment nec Manufacture nec

50 Figure A1 Databae contruction: MRIO to IRIO table Figure A2 Share of intermediate good in bilateral gro trade Flow: hitogram-difference between Proportion and BEC method 49