Tracing value-added and double counting in sales of foreign affiliates and domestic-owned companies

Transcription

1 MPRA Munch Personal RePEc Archve racng value-added and double countng n sales of foregn afflates and domestc-owned companes Sébasten Mroudot and mng ye rade and Agrculture Drectorate, OECD, rade and Agrculture Drectorate, OECD 14 March 2018 Onlne at MPRA Paper No , posted 5 Aprl :49 UC

2 RACIN VALUE-ADDED AND DOUBLE COUNIN IN SALES O OREIN AILIAES AND DOMESIC-OWNED COMPANIES Sébasten Mroudot and Mng Ye, OECD March 2018 ABSRAC: he lterature on trade n value-added has emphassed that gross trade flows do not adequately measure the ncome generated by trade when many ntermedate nputs are mported. Whle ths lterature has deepened our understandng of global value chans, t s stll mssng an mportant element when analysng ncome generaton along the value chan: the fact that domestc value-added often results from actvtes of foregn-owned companes. Studes that look at actvtes of multnatonal enterprses (MNEs) stll rely on the concept of sales of foregn afflates, whch s a gross concept also subect to double countng when t comes to the use of ntermedate nputs. In ths paper, we propose a new accountng framework for the decomposton of value-added nto domestc, foregn and double countng terms n domestc sales. In ths framework, we show where the value-added double countng s derved from and gve an explct expresson of domestc and foregn double countng terms based on the Inter-Country Input-Output (ICIO) tables hosh nsght. We can dstngush domestc sales from exports and trace the value added and double countng n sales of foregn afflates and domestc-owned enterprses. Based on ths framework, we then calculate the value-added by foregn-owned and domestc-owned frms n exports and n domestc sales by usng an Inter-Country Input-Output table splt accordng to ownershp. Prelmnary results suggest that there s much more double countng n sales of foregn afflates than n exports and that more value-added s created through exports than through sales of foregn afflates n world DP. Keywords: nter-country nput-output, value-added decomposton, global value chans, foregn afflates JEL Codes: E01, E16, 23, L14

3 1. Introducton he lterature on trade n value-added (Johnson and Noguera, 2012; Koopman et al., 2014; Los et al., 2016; Nagengast and Stehrer, 2016) as well as emprcal datasets, such as the rade n Value-Added (VA) ndcators released by OECD and WO n 2013, have emphassed that gross trade flows do not adequately measure the ncome generated by trade n a world charactersed by global supply chans where ntermedate products are traded across countres. New accountng frameworks have been developed to dentfy the domestc value-added n gross exports and n fnal demand and to remove the double countng of ntermedate nputs that cross nternatonal borders more than once. But trade s only one dmenson n the actvtes of frms nvolved n global producton. Some of these frms are multnatonal enterprses (MNEs) that rely on foregn afflates to source nputs or produce abroad. Accordng to UNCAD (2013), 80% of global trade s co-ordnated by these MNEs (when ncludng ther arm s length trade transactons as well as trade flows related to franchsng, contract manufacturng and strategc allances). he emprcal lterature analysng actvtes of MNEs reles on the concept of sales of foregn afflates (Dunnng, 1980; Branard, 1997; Bergstrand and Egger, 2007), whch s also a gross concept and ncludes some double countng wth respect to foregn and domestc nputs. Somehow ths concept has not yet been through the knd of value-added revoluton that has sgnfcantly changed the analyss of trade. Also, the concept of sales of foregn afflates covers both ther domestc sales and ther exports. When exportng, the output of foregn afflates can be decomposed wth the same tools created for VA analyss. But t s dfferent n the context of domestc sales. In ths paper, we are nterested n decomposng not only trade but also domestc sales n a consstent framework that can allow us to dentfy the actvtes of foregn-owned frms. Such a decomposton can shed lght on the reasons why frms engage n DI. he lterature suggests that foregn afflates can be nvolved: (1) n the producton of (fnal) goods for domestc consumers n the case of horzontal DI (Markusen, 1984); (2) n the producton of (fnal) goods for foregn consumers n the case of export platform DI (Ekholm et al., 2007); or (3) n the producton of nputs for other afflates n the host economy or abroad n the case of vertcal DI (Helpman, 1984). 1 More recent work ndcates that n many nstances frms engage n complex DI combnng horzontal and vertcal motves (Alfaro and Charlton, 2009), or set up afflates for other purposes than contrbutng to the producton process such as conglomerate DI or DI for fnancal purposes (Herger and 1 hs lterature tends to not dstngush fnal from ntermedate goods. Whle vertcal DI s more about nvestment n order to supply nputs, the case of fnal assembly by the afflate s stll ambguous. 2

4 McCorrston, 2016; Ray, 2016). here s therefore a need for more emprcal work on value creaton n relaton to actvtes of foregn afflates. Moreover, by applyng to sales of foregn afflates the same knd of treatment appled to trade flows n the context of value-added analyses, we would lke to provde a more accurate measurement of the mportance of MNEs n global producton. gures such as the one proposed by UNCAD seem to overstate the true mportance of MNEs n trade and output as they are based on gross fgures and not on a value-added decomposton. he paper s organsed as follows. Secton 2 detals the methodology, frst presentng an alternatve mathematcal framework to derve the domestc and foregn value-added multpler coeffcents n an nter-country nput-output framework. Based on ths, we provde a full decomposton of the valueadded n domestc sales followng the hosh nsght and dentfyng double countng terms n addton to domestc and foregn value-added. In Secton 3, we use ths conceptual framework to decompose DP n the context of ICIO tables that have an ownershp dmenson (splttng data for domestcowned and foregn-owned frms). Secton 4 concludes. 2. Methodology hs secton ntroduces a new type of value-added decomposton n the nter-country nput-output (ICIO) framework. he startng pont s that gross output conssts of domestc sales (.e. domestc shpments) and exports (.e. shpments to foregn countres). Some mportant efforts have been devoted to tracng domestc and foregn value-added, as well as double countng, n gross exports (followng Koopman et al., 2014), but not for domestc shpments. Domestc sales are also nterestng n terms of ther domestc and foregn value-added content, especally when these domestc sales result from actvtes of foregn-owned companes. Our obectve s to provde a full decomposton of DP n a gven economy that would allow us to dentfy the domestc and foregn value-added n domestc sales and n exports, and ultmately to compare the foregn value-added comng from exports wth the value added by foregn-owned frms n the domestc economy (whch s part of the domestc valueadded n current decompostons). 2.1 Clarfyng the domestc sales n IO tables and value added multpler coeffcents Leontef (1936) establshed that the amount and type of ntermedate nputs needed n the producton of one unt of output can be estmated based on the nput-output (IO) structure across ndustres. Usng the lnkages across ndustres, one can trace output n all stages of producton needed to produce one unt of fnal goods. When the gross output flows assocated wth a specfc level of fnal demand are 3

5 known, value-added producton and trade can be smply derved by multplyng these flows wth the value added to gross output rato n each ndustry. In the IO framework, all gross output must be used ether as an ntermedate good or as a fnal good, X AX Y (1) where, X s the N 1 gross output vector, Y s the N 1 fnal demand vector, and A s the N I-O coeffcents matrx. he accountng relatonshp between domestc sales H and fnal demand n an Inter-Country Input- Output (ICIO) model can be expressed as: N H AH Y (2) Here, A A I A D 1 ( ), wth D A the domestc coeffcents n the global ICIO table (.e. the blockdagonal matrx of the A matrx n the ICIO table). A s the export matrx of A ndcatng the use of D D ntermedate nputs from one country nto another country. In addton, Y Y AY, Y denotng domestc fnal demand and Appendx I (Lemma 1). Y fnal demand n foregn countres. he full dervaton s provded n Each element of the A matrx descrbes how domestc ntermedate goods are sent abroad (or transported domestcally) to produce one unt of domestc sales n foregn countres (or n the domestc economy). or example, the element A means that n order to produce one unt of domestc sales n country, country needs to produce A unts of ntermedate nputs that are then emboded n domestc sales n country. AH means that country needs to produce AH ntermedate nputs for domestc sales H n country, so we can call A as the drect domestc sales requrements matrx. Re-arrangng equaton (2) above, we obtan H BY, and B I A 1 ( ), smlar to B I A 1 ( ) n the IO model. We can defne matrx B as the total domestc sales requrements matrx. Wth respect to B, we have B ( I A) [ I A ( I A ) ] [( I A )( I A ) A ( I A ) ] 1 D D 1 1 D [( I A A )( I A ) ] 1 1 D D ( I A ) B( I A A ) B I A B 4

6 So, for any element n matrx B, B I AB. AB or H, the domestc sales n country, all the ntermedate nputs needed are A H. We can thus calculate the value-added n domestc sales n country as VaH () H A H. hs valueadded does not only nclude country s value-added but also other countres value-added. We can then express the value-added multpler coeffcents n domestc sales n the form of a matrxv, defned as: V u I A u I A I A V I A 1 1 ( ) ( )( ) ( ) (3) where V s a 1 N drect value-added coeffcents vector. Each element of V gves the share of drect domestc value-added n total output. It s equal to one mnus the ntermedate nput share from all countres (ncludng domestcally produced ntermedates): V u[ I A ], where u s a 1 N unt vector. If we use the notaton B I A 1 ( ), we obtan the expresson for value-added coeffcents n domestc sales n country : V VB V B. hey can be dvded nto two parts: the value- added from country (domestc part) VB and the value-added from other countres (foregn part) VB. 2.2 Measurement of value added n domestc sales Moreover, we also can derve a consstent measure of the domestc and foregn value-added (or DP) n domestc sales from the ntal ICIO model. In the ICIO model, gross exports and gross output can be wrtten as: E I A A H I A Y 1 1 ( ) ( ) (4) X H E I A H I A Y B H B Y 1 1 ( ) ( ) (5) Based on the expresson above, for country s output X : 5

7 (6) X B H B H B Y herefore, country s DP can be measured as: DP V X V B H V B H V B Y (7) Accordng to equaton (7), country s DP can be dvded nto 3 parts. V BY s the share of DP n exports of fnal products, VB H s DP n country s domestc sales and B H V s DP n foregn countres domestc sales va exports of ntermedates from country to other countres and measured as foregn value-added for other countres. Smlarly, we can also express domestc value-added n domestc sales n a consstent way and regard coeffcents VB as the domestc valueadded multpler coeffcents for a country s domestc sales. Symmetrcally, we can obtan the expresson of country s DP as k k k k k k DP V X V B H V B H V B Y the part. rom the pont of vew of country s DP, VB H s ncluded n country s domestc sales. If we sum up the value-added from all countres (except country ) n country s domestc sales, we also obtan an expresson for the foregn value-added part of country s domestc sales, measured as VB H. herefore, the value-added n country s domestc sales should be equal to the domestc part plus the foregn part: VB H V B H VH. 2.3 racng value added n domestc sales: the hosh nsght he hosh model (hosh, 1958), n turn, s also known as the supply drven nput-output model, snce value-added s the exogenously specfed drvng force of the model. Although the hosh model s generally nterpreted as a prce model (Detzenbacher, 1997), t can be appled to the analyss of the structure of value-added flows as an alternatve to Leontef s demand-drven model. he supply drven accountng dentty states that country s total nput (whch should be equal to domestc sales here) s equal to the value of ts ntal nputs (domestc and foregn value-added) plus ts ntermedate nput flows from all other countres (whch can be nterpreted as value-added double countng terms). 6

8 In the IO table, the output coeffcent s defned as l x / x. Output coeffcents gve the output percentage of ndustry that s sold to ndustry. he accountng equaton can be rewrtten as: X VA X L VA (8) where I L 1 ( ) denotes the hosh nverse; Meanwhle, ˆ ˆ, where ˆX s a N N 1 X BX dagonal matrx wth output on the dagonal. Smlarly, n the domestc sales nput-output table, domestc sales can be wrtten as H VaH H L VaH. Here ˆ ˆ 1, ˆ L H AHˆ 1 and ˆ L H A Hˆ. L gves 1 H BH the share of country s goods n country s domestc sales. o llustrate the relatonshp between domestc sales and value-added, we can refer to the aylor expanson: H VaH I L L L 2 3 ( ) (9) In the value-added nput VaH, the domestc sales value s H, whch s decomposed nto three value-added terms: an ntal nput VaH, a drect nput 2 3 n subsequent rounds amountng tovah ( L L ). VaH L n the frst round and ndrect nput ollowng the hosh nsght, we can gve the full decomposton for country s domestc sales: H VaH () VaH () L VaH ( ) L VaH ( ) [ L] VaH ( ) [ L] VaH ( ) [ L] VaH ( ) [ L] (10) he above expresson provdes an explct nterpretaton of the decomposton of domestc sales accordng to the hosh nsght. Every sub-term has an economc nterpretaton. he ntal effect s the value-added n country s domestc sales. Accordng to the equaton above, ths term s equal to () ˆ VaH VH ( VB V B ) Hˆ, whch s the ntal value-added nput n the hosh nsght. As value-added s fully measured n ths round, the value-added found n the later rounds s part of double countng terms (.e. value-added that has already been measured and further 7

9 goes to sectors and/or countres as ntermedate nput). If we expand ths term, t ncludes domestc value-added ntal nputs VB H ˆ and foregn value-added ntal nputs VB ˆ H whch are contaned n goods mported from country. In the frst round, t means that the value-added term whch s already counted n the ntal round propagates through the matrx 2 3 L Hˆ A Hˆ (havng n mnd that 1 A A B A A [ A ] A [ A ], ths value-added propagaton route ncludes not only what has gone across the country border but also across domestc sectors as ntermedate nputs). Because ths value-added was already measured n the ntal round, t should be part of the valueadded double countng terms n later rounds. he drect effect can be dvded nto two parts, the effect from country s nputs (whch s not domestc value-added here) and from other country s nputs (whch s not foregn value-added here). Country s nput s equal to: ˆ ˆ 1 () ˆ ˆ VaH L V H H A H V A H (11) he other countres value-added wthn ntermedate nputs s mported from country. hese terms are equal to: ˆ ˆ 1 VaH ( ) L ˆ ˆ VH H AH VAH (12) In the second round, the addtonal value-added has a smlar nterpretaton and can also be dvded nto the domestc nput and foregn nput parts. It stll accounts for value-added double countng terms passed from country s domestc propagaton to the other countres and returned back home. hs mples that for the domestc nput part, the value-added comng from country s VaH () Lk Lk, k reflectng value-added from country VaH () L k propagated to country k. he L k part n country k returned back home. hs part of value-added has already been measured n the ntal round, so t should stll be counted as value-added double countng term (domestc). We have 2 () [ ] () k k k VaH L VaH L L ˆ ˆ 1 ˆ ˆ 1 ˆ ˆ k k k k k k k k VH H AH H AH V AAH (13) 8

10 or the value-added contrbuted by country, we have VaH ( ) k L L k k, reflectng the value-added from country VaH ( ) L k propagated to country k. L k s the part n country k that has returned back to country. hs part has also already been measured n the ntal round as foregn value-added nput, so t should be counted as value-added double countng term (foregn). Also, we have 2 ( ) [ ] ( ) k k k VaH L VaH L L ˆ ˆ 1 ˆ ˆ 1 ˆ ˆ k k k k k k k k VH H A H H AH V A AH (14) herefore, n round 2, the whole double counted value-added n the foregn part sv ˆ Ak Ak H. And for the domestc part, the domestc double counted value-added s: k VaH () L VaH () [ L] VaH () [ L] 2 3 V ( A A A A A A ) Hˆ V ( B I) Hˆ V A B Hˆ k k k (15) Whle the foregn double counted value-added s: 2 3 VaH( ) L VaH( ) [ L] VaH( ) [ L] V ( A A A A A A ) Hˆ V B Hˆ k k k kt t k t k (16) Mergng the expresson of value-added coeffcents V VB V B n the domestc sales ICIO framework, the domestc value-added n country s domestc sales should be equal to country s value-added porton n the ntal round: VB H ˆ. Moreover, country s value-added double countng term n domestc sales should be equal to the sum of the country s value-added porton n the double countng content (ncludng the domestc nput term and foregn nput term): VB A B Hˆ VB B Hˆ. heorem 1: In the value-added decomposton of domestc sales, the sum of the domestc value-added and the double countng term s equal to the domestc content n domestc sales. 9

11 ˆ ˆ VB H VB A B H VB B Hˆ VB Hˆ Smlarly, the foregn value-added for country s domestc sales should be equal to the sum of foregn countres value-added n the ntal round s foregn nput: VB Hˆ. oregn value-added double countng term n country s domestc sales should be equal to the foregn value-added porton n the. double countng content: ˆ VB ABH VB B Hˆ s s s heorem 2: In the value-added decomposton of domestc sales, the sum of the foregn value-added and the double countng term s equal to the foregn content n domestc sales. ˆ ˆ VB H VB ABH VB B Hˆ VB Hˆ s s s See Appendx I for the proof of these theorems. 2.4 he meanng of double countng terms: further clarfcatons In the nput-output framework, double countng comes from the concept of ntermedate nput. Output s equal to (domestc) value-added plus ntermedate nputs (domestc value-added double countng, foregn value-added and foregn double countng), so the double countng term n output decomposton s a subset of ntermedate nputs. hs translates n the hosh model nto an ntal round where all (domestc) value-added s captured and subsequent rounds that correspond to double countng n output. oregn value-added, when relevant, frst appears n the frst round. When t comes to value-added n exports, there s no consensus yet on the defnton of double countng. Some authors, such as Koopman et al. (2014), Nagengast and Stehrer (2016) and Born and Mancn (2017) propose to base the defnton on the number of nternatonal border crossngs. Also, Los and mmer (2018) pont out that the double counted domestc value-added s the sum of the blateral domestc value-added across all partners mnus the unlateral one (.e. wth partner world). Alternatvely, Mroudot and Ye (2017) rely on the hosh nsght. In ther framework, domestc valueadded s the ntal round nput and foregn value-added s the frst round, whch s consstent wth the analyss of output. In addton, every round n ths framework s conceptually value-added crossng the 10

12 supply sde s country border (.e. not consderng the destnaton where t s absorbed) 2. hs defnton based on the hosh framework s consstent wth Los and mmer (2018), as the double countng can also be expressed as the sum of the value-added across all blateral relatonshps mnus value-added n unlateral trade. he decomposton for domestc sales presented n the prevous secton s smlar wth the exports decomposton n Mroudot and Ye (2017). However, exports and domestc sales are dfferent concepts. If we compare the double countng n domestc sales wth the one n exports, we fnd that double countng terms n domestc sales are sgnfcantly hgher. It comes from the fact that rather than the country border for nternatonal transactons, the double countng for domestc sales starts wth the sector border. When value-added crosses the sector border as ntermedate nput more than once, t becomes part of double countng. Let assume that a car s exported wth all nputs produced domestcally except one whch s mported. Only the mported nput can create some double countng n the case of exports (for example f t ncludes domestc value-added comng back). If the car s sold on the domestc market, all the nputs requred (and the nputs used to produce these nputs) are also domestc sales and the double countng wll be sgnfcantly hgher. Secton 3 ncludes emprcal results that llustrate ths dfference n double countng. rom a conceptual pont of vew, t seems however mportant to have a consstent defnton of double countng n exports and n domestc sales, partcularly f we start to decompose DP nto both exports and domestc sales as n the next sub-secton. 2.5 DP decomposton nto exports and domestc sales (wth an overlap) As prevously hghlghted, the accountng relatonshp between domestc sales H and fnal demand n destnaton n the ICIO model can be wrtten as H AH Y. In a smlar way, we can also obtan the accountng relatonshp between gross exports E and fnal demand n dfferent destnatons n the Inter-Country Input-Output (ICIO) model (see Appendx I): 2 he country border n the hosh framework can be seen as conceptually dfferent from the border crossed by products n the real world. As the model s supply-drven, value-added crosses the border from the supply sde rather than the demand sde and cannot be merged wth value-added fnally absorbed n the destnaton country. Concretely, t means that n the ntal round of the hosh model appled to the decomposton of exports, value-added crosses the domestc border only once whle from the demand sde t would have crossed countres borders for ndefnte tmes. We beleve ths dfference explans why the decomposton n Mroudot and Ye (2017) cannot be descrbed as arbtratry as Los and mmer (2018) suggest for decompostons that focus on border crossngs from the demand sde. 11

13 (17) wth and. Re-arrangng equatons (2) and (17), we can express gross exports and domestc sales as: E I A I A A I A Y Y D 1 1 D 1 D [ ( ) ] [ ( ) ] (18) H I A I A A I A Y Y D 1 1 D 1 D [ ( ) ] [ ( ) ] (19) herefore, n the ICIO model, gross output can be wrtten as: Or Rearrangng equatons (20) and (21), we get: D D X A X Y A X Y A X Y E (20) D X A X Y A X Y H A X Y (21) D 1 D D 1 X ( I A ) Y ( I A ) E (22) And 1 1 X ( I A ) Y ( I A ) H (23) he expresson ( I A ) D 1 s sometmes descrbed as the local Leontef nverse n the ICIO. DP can then be calculated as follows: D 1 D D 1 DP VX V( I A ) Y V( I A ) E (24) Or 1 1 DP VX V( I A ) Y V( I A ) H (25) Accordng to equaton (24), DP can be dvded nto two parts. he frst part s the share of DP that does not partcpate n nternatonal trade and s ust for domestc fnal demand. he second part, D 1 V( I A ) E, s the share of DP n exports. DP n exports ncludes some value-added that can return home. hs s why the splt s not based on whether fnal consumpton takes place n the domestc economy or abroad. Exports nclude both ntermedate and fnal products. rom equaton (25), DP can also be decomposed nto two parts along another dmenson: 1 V( I A ) H reflects the value-added n domestc sales whle 1 ( V I A ) Y corresponds to value-added for the foregn fnal demand. Agan, t does not ndcate where value-added s ultmately gong as the concept of domestc sales s stll a mx of ntermedate and fnal products. Mergng equatons (18), (19), (24) and (25), we obtan the followng DP decomposton: 12

14 DP VBA ( I A ) Y V( I A ) Y D 1 D D 1 D VBA ( I A ) Y V( I A ) Y D 1 1 (26) or a specfc country, the equaton can be wrtten as follows: 1 D 1 D s s ( ) ( ) s DP V B A I A Y V I A Y s ss s s V B A B Y V B Y (27) Equaton (27) allows us to dvde DP nto 4 terms that we can nterpret n the followng way. he D s s s 1 frst term, V B A ( I A ) Y, measures the value-added that has propagated n domestc and global value chans 3 and s absorbed by destnaton countres as an ntermedate product. he 1 second term, V ( I A ) Y D, measures the purely domestc value-added whch has not been part of nternatonal trade and ends up n domestc fnal demand. he thrd term, V BsAssBsY s, measures the value-added that has partcpated n domestc and global value chans and s ultmately absorbed by destnaton countres as a fnal product. he last term, V BY, measures the valueadded that has not partcpated n the domestc propagaton and s absorbed by foregn countres. erms 1, 3 and 4 are equal to the domestc value-added n exports, as measured by Koopman et al. (2014) or by Los et al. (2016), whch ncludes the value-added n exports comng back to the domestc economy. he second term corresponds to value-added gong nto domestc fnal demand wthout havng transted through other countres. heorem 3: he DP decomposton n equaton (27) s consstent wth DP decomposton accordng to fnal demand. We have: VBA ( I A ) Y V( I A ) Y VBY D 1 D D 1 D D VBA ( I A ) Y V( I A ) Y VBY D In ths sub-secton, we use the concept of global value chan the way t s used n the trade lterature to descrbe the producton of goods and servces across dfferent countres that provde nputs at dfferent stages of producton before fnal producton takes place. he Leontef nverse n a global ICIO can be understood as a matrx descrbng the fragmentaton of producton across countres and ndustres. 13

15 rom the above decomposton, we can also provde expressons for the value-added n exports and n domestc sales as follows: V ( I A ) E VBA ( I A ) Y VBA ( I A ) Y V ( I A ) Y D 1 D 1 D D 1 1 V ( I A ) H VBA ( I A ) Y VBA ( I A ) Y V ( I A ) Y 1 D 1 D D 1 D 1 D (28) hese equatons hghlght an mportant feature of ths value-added decomposton. here s an overlap between the value-added n exports and n domestc sales (as some domestc sales are ntermedates D 1 D that are then ncorporated nto exports). he overlap can be seen n VBA ( I A ) Y and VBA ( I A ) Y D 1, as these two terms not only partcpate n the domestc propagaton but also n nternatonal trade. Comng back to equaton (27), t means that whle terms 1, 3 and 4 are equal to domestc value-added n exports, value-added n domestc sales s equal to terms 1, 2 and 3. he overlappng terms are 1 and 3. hey correspond to value-added that has transted through the domestc value chan before beng exported and absorbed abroad (or back to the domestc economy). In gross terms, we can dstngush gross exports from domestc sales but n value-added terms there s by defnton on overlap. here s no reason for the value-added reflected n terms 1 and 3 to be attrbuted to exports or domestc sales only. And t hghlghts that decompostons of gross exports nclude value-added from domestc sales. hs fndng has mplcatons for authors comparng exports and sales of foregn afflates or n the case of trade n servces the comparson between cross-border trade (Mode 1, 2 and 4 n the eneral Agreement on rade n Servces) and trade through commercal presence (Mode 3). 3. Emprcal results In order to llustrate how our framework can be used to analyse domestc sales and compare sales of foregn afflates wth exports, we rely on the World Input-Output Database (WIOD) nter-country nput-output tables. Some nformaton on WIOD can be found n mmer et al. (2015). We use the 2016 update of the database (mmer et al., 2016). he World Input-Output ables (WIOs) do not have nformaton on domestc and foregn ownershp of frms but the OECD has splt these tables accordng to ownershp as part of ts analytcal AMNE database (Cadestn et al., 2018a). In ths proect, WIOD data have been merged wth statstcs on the output, value-added and exports of domestcowned and foregn-owned frms n each country and ndustry. ull nput-output tables are avalable where n each ndustry the nformaton s splt between domestc-owned and foregn-owned frms across columns and rows. 14

16 3.1 Value added n domestc sales by foregn owned and domestc owned frms We frst decompose domestc sales accordng to the equatons presented n Secton 2.3 (able 1). hs decomposton allows us to dentfy the domestc value-added, domestc double countng, foregn value-added and foregn double countng n domestc sales. he nformaton s further splt between domestc-owned and foregn-owned companes. As prevously hghlghted, domestc sales feature a hgh level of double countng due to the fact that most of the ntermedate nputs used n the producton process are also contrbutng to domestc sales. It s an mportant dfference wth the decomposton of gross exports where only nputs crossng nternatonal borders can create double countng. he foregn double countng s also mportant n domestc sales as mported nputs also go through processng n the domestc economy. Country Chna ermany rance Japan able 1. Decomposton of domestc sales for selected economes, 2014 Ownershp Domestc sales (mo USD) Domestc VA (%) 15 Domestc double countng (%) oregn VA (%) oregn double countng (%) Domestc-owned 27,192, oregn-owned 2,126, Domestc-owned 4,383, oregn-owned 1,000, Domestc-owned 3,750, oregn-owned 510, Domestc-owned 7,563, oregn-owned 287, Unted Domestc-owned 26,677, States oregn-owned 2,366, Source: Authors calculatons based on WIOD and the OECD analytcal AMNE database. Another result llustrated n able 1 s that afflates of foregn frms also rely to a large extent on domestc nputs to produce n the host economy. We fnd that foregn afflates rely relatvely more on foregn VA as compared to domestc-owned frms, but the dfference s qute small and domestc VA s n all countres hgher than foregn VA for foregn-owned frms. 3.2 Decomposton of DP: Value added n exports and n domestc sales by domestc owned and foregn owned frms Usng equaton (27) n the ICIO splt accordng to ownershp, we can provde a full decomposton of world DP ndcatng whether value-added s derved from domestc sales or from exports and then whether ths value-added s generated by domestc-owned frms or foregn-owned frms. he results

17 are summarsed n gure 1 through a chart that ndcates the relatve mportance of the value-added generated n each case, decomposed between domestc-owned frms (D) and foregn-owned frms (). able 2 (n Appendx II) reports detaled results by country for the four terms of equaton (27). gure 1 llustrates the overlap between value-added n exports and n domestc sales, through 1 and 3 (.e. the value-added that propagates through domestc and global value chans and s fnally absorbed n the destnaton country as an ntermedate or as a fnal good). 1 and 3 belong both to DP n exports and n domestc sales. hs overlap amounts to 16% of world DP. gure 1. Decomposton of world DP accordng to equaton (27), domestc-owned and foregn-owned frms, 2014 DP n Exports 2 (D) = 73.8% 1 (D) 8.4% 3 (D) 4.8% 4 (D) 2.9% 2 () = 6% 1 () 2.1% 3 () 0.7% 4 () 1.3% DP n Domestc Sales Source: Authors calculatons based on WIOD and the OECD analytcal AMNE database. D = Domestc-owned frms; = oregn-owned frms. 1 = value-added that has propagated n domestc and global value chans and s absorbed by destnaton countres as an ntermedate product; 2 = purely domestc value-added whch has not been part of nternatonal trade and ends up n domestc fnal demand, 3 = value-added that has partcpated n domestc and global value chans and s ultmately absorbed by destnaton countres as a fnal product, and 4 = value-added that has not partcpated n the domestc propagaton and s absorbed by foregn countres. gure 1 hghlghts that more value-added s created through trade (exports of domestc-owned and foregn-owned frms) than through the sales of foregn afflates (domestc sales and exports of foregn-owned frms), a result maybe n contradcton wth the lterature based on gross flows that overstate the mportance of sales of foregn afflates when gnorng the double countng. he valueadded n exports by domestc-owned frms and foregn-owned frms (8.4%+4.8%+2.9%+2.1%+0.7%+1.3%=20.2%) s double the sze of the value-added n domestc sales and exports by foregn-owned frms (6%+2.1%+0.7%+1.3%=10.1%). he overlap between trade and 16

18 nvestment s relatvely small.e. 4.1% of world DP - whch corresponds to the value-added n exports by foregn-owned frms (2.1%+0.7%+1.3%). hese 4.1% seem to suggest that at the world level foregn afflates account for a rather small share of trade n value-added terms. It s an ndcaton that global value chans operate wth many arm s length trade transactons and maybe less wthn pure MNE networks (e.g. nputs transferred between afflates). gure 1 also ponts out that 73.8% of world DP s value-added created by domestc-owned frms n domestc sales wthout transtng through global value chans (2). or most products and partcularly for servces, a hgh share of value-added s purely domestc. However, MNEs can stll play a role n ths domestc value-added as they are nvolved n domestc sales n the parent country and ther operatons can n some cases not rely on nternatonal trade Concludng remarks hs paper has ntroduced a new type of DP decomposton that allows us to trace value-added and double countng not only n exports but also n domestc sales. he motvaton s that tradtonal I-O analyss lookng at value-added n fnal demand s not suffcent to dscuss trade and nvestment n global value chans. Lookng at trade n value-added terms and decomposng gross exports has brought many nterestng analytcal results and led to new polcy mplcatons. We beleve the same wll happen when lookng at actvtes of multnatonal enterprses n value-added terms and decomposng sales of foregn afflates. hs s why we need for domestc sales tools smlar to what was developed for gross exports. By usng our methodology to compare the value-added n exports and n sales of domestc-owned and foregn-owned frms, there are already nterestng fndngs. In partcular, t seems that the double countng n sales of foregn afflates s much more pronounced than n exports, as afflates of foregn frms rely even more on nputs from the host economy and nputs transactons n the domestc economy lead to double countng n sales. he whole lterature on the benefts of DI and the mpact of actvtes of MNEs wll have to be revsted n lght of ths value-added analyss whch can provde a better ndcaton of how ncome s generated and who really benefts from producton by foregnowned frms. 4 When addng the contrbuton of MNEs n the parent economy, the share of MNEs n world DP s 31% accordng to Cadestn et al. (2018b). 17

19 References Alfaro, L. and A. Charlton (2009), Intra-Industry oregn Drect Investment, Amercan Economc Revew, 99(5), pp , Bergstrand, J. and P. Egger (2007), A knowledge-and-physcal-captal model of nternatonal trade flows, foregn drect nvestment, and multnatonal enterprses, Journal of Internatonal Economcs, 73(2), pp , Born, A. and M. Mancn (2017), ollow the Value Added: rackng Blateral Relatons n lobal Value Chans. MPRA Paper, No Branard, S. L. (1997), An Emprcal Assessment of the Proxmty-Concentraton rade-off Between Multnatonal Sales and rade, Amercan Economc Revew, 87(4), pp , Cadestn, C., K. de Backer, I. Desnoyers-James, S. Mroudot, D. Rgo and M. Ye (2018a), Multnatonal enterprses and global value chans: the OECD analytcal AMNE database, OECD rade Polcy Papers, No. 211, OECD Publshng, Cadestn, C., K. de Backer, I. Desnoyers-James, S. Mroudot, D. Rgo and M. Ye (2018b), Multnatonal enterprses and global value chans: new nsghts from the trade-nvestment nexus, OECD Scence, echnology and Industry Workng Papers, 2018/04, OECD Publshng. Detzenbacher, E. (1997), In vndcaton of the hosh Model: A renterpretaton as a prce model, Journal of Regonal Scence, 37(4), pp , Dunnng, J.H. (1980), oward an Eclectc heory of Internatonal Producton: Some Emprcal ests, Journal of Internatonal Busness Studes, 11(1), pp. 9 31, Ekholm, K., R. orsld and J. Markusen (2007), Export Platform oregn Drect Investment, Journal of the European Economc Assocaton, 5(4), pp , hosh, A. (1958), Input-Output Approach n an Allocaton System, Economca, 25(97), pp , Helpman, E. (1984), A Smple heory of Internatonal rade Wth Multnatonal Corporatons, Journal of Poltcal Economy, 92(3), pp , Herger, N. and S. McCorrston (2016), Horzontal, Vertcal, and Conglomerate Cross-Border Acqustons, IM Economc Revew, 64(2), pp , Johnson, R. and. Noguera (2012), Accountng for ntermedates: Producton sharng and trade n value added, Journal of Internatonal Economcs, 86(2), pp , 18

20 Koopman, R., Z. Wang and S.-J. We (2014), racng value-added and double countng n gross exports, Amercan Economc Revew, 104(2), pp , Leontef, W. (1936), Quanttatve Input and Output Relatons n the Economc System of the Unted States, Revew of Economc and Statstcs, 18(3), pp , Los, B., M. P. mmer and. J. de Vres (2016), racng Value-Added and Double Countng n ross Exports: Comment, Amercan Economc Revew, 106(7), pp , Los, B. and M. P. mmer (2018), Measurng Blateral Exports of Value Added: A Unfed ramework, paper prepared for CRIW Conference on he Challenge of lobalzaton n the Measurement of Natonal Account. Washngton DC, USA. Markusen, J. (1984), Multnatonals, Mult plant Economes, and the ans rom rade, Journal of Internatonal Economcs, 16(3), pp Mroudot, S. and M. Ye (2017), Decomposton of Value-Added n ross Exports: Unresolved Issues and Possble Solutons MPRA Paper, No Nagengast, A. J. and R. Stehrer (2016), Accountng for the Dfferences Between ross and Value Added rade Balances, World Economy, 39(9), pp , Ray, A. (2016), Expandng Multnatonals Conglomerate M&A and Actvty-Basket Proxmty, paper presented at the 19th Annual Conference on lobal Economc Analyss, Washngton DC, USA. mmer, M. P., E. Detzenbacher, B. Los, R. Stehrer and. J. de Vres (2015), An Illustrated User ude to the World Input Output Database: the Case of lobal Automotve Producton, Revew of Internatonal Economcs, 23, pp , mmer, M. P., B. Los, R. Stehrer and. J. de Vres (2016), An Anatomy of the lobal rade Slowdown based on the WIOD 2016 Release, DC research memorandum number 162, December. UNCAD (2013), World Investment Report lobal Value Chans: Investment and rade for Development, Unted Natons Conference on rade and Development, New York and eneva. 19

21 Appendx I Lemma 1:he accountng relatonshp between domestc shpments H and fnal demand n destnaton n an Inter-Country Input-Output (ICIO) model can be expressed as: H AH Y Proof: ross output X s the sum of gross exports E and gross domestc shpments H. rom the accountng dentty n equaton (1), X = AX + Y, we can express E the vector of exports and H the vector of gross domestc shpments as: Solvng for E, we obtan: E A ( EH) Y D H A ( EH) Y D E ( I A ) A H ( I A ) Y 1 1 Mergng the expresson for H and for E, we obtan: D H A ( EH) Y D 1 1 D [ ( ) ( ) ] A [ I ( I A ) A ] H A ( I A ) Y Y D 1 D 1 D A ( I A ) H A ( I A ) Y Y D A H I A A H I A Y Y D 1 D 1 D AH Y wth A A I A D 1 ( ) and D Y AY Y. Lemma 2:he accountng relatonshp between gross exports E and fnal demand n destnaton n an Inter-Country Input-Output (ICIO) model can be expressed as: Proof: Smlar to Lemma1. Lemma 3:In the domestc sales accountng framework, we have 20

22 Here, B I A 1 ( ) and B I A B B B 1 ( ), B s the total requrements matrx n the ICIO table whch s B I A 1 ( ). Proof: Expandng the expresson of B and B, we obtan: B B ( I A ) ( I A) [( I A)( I A )] {[ I A ( I A ) ]( I A )} 1 D 1 1 {[( I A )( I A ) A ( I A ) ]( I A )} D 1 1 [( I A A )( I A ) ( I A )] 1 ( I A) B D 1 1 heorem 1: In the value-added decomposton of domestc sales, the sum of the domestc value-added and the double countng term s equal to the domestc content n domestc sales. ˆ ˆ VB H VB A B H VB B Hˆ VB Hˆ Proof: Accordng to lemma 3, we can obtan the submatrx s expresson as B B B B B B B mergng the expresson of matrx B I AB, we have B B A B B B B heorem 2: In the value-added decomposton of domestc sales, the sum of the foregn value-added and the double countng term s equal to the foregn content n domestc sales. Proof: Smlar to theorem 1. ˆ ˆ VB H VB ABH VB B Hˆ VB Hˆ s s s heorem 3: he four terms DP decomposton s consstent wth DP decomposton n the fnal demand. We have: 21

23 VBA ( I A ) Y V ( I A ) Y VBY D 1 D D 1 D D VBA ( I A ) Y V ( I A ) Y VBY D 1 1 Proof: or the frst equaton, we have BA ( I A ) ( I A ) ( BA I)( I A ) D 1 D 1 D 1 [ BA B( I A)]( I A ) [ B( I A A )]( I A ) B he same can be done wth the second equaton. D 1 D 1 22

24 Appendx II able 2. our terms DP decomposton, domestc-owned and foregn-owned frms, 2014 Country Ownershp DP (mllon USD) AUS Domestc-owned 1,214, oregn-owned 170, AU Domestc-owned 327, oregn-owned 77, BEL Domestc-owned 396, oregn-owned 105, BR Domestc-owned 38, oregn-owned 14, BRA Domestc-owned 2,040, oregn-owned 222, CAN Domestc-owned 1,408, oregn-owned 289, CHE Domestc-owned 520, oregn-owned 177, CHN Domestc-owned 9,722, oregn-owned 676, CYP Domestc-owned 20, oregn-owned 1, CZE Domestc-owned 127, oregn-owned 70, DEU Domestc-owned 3,003, oregn-owned 616, DNK Domestc-owned 272, oregn-owned 44, ESP Domestc-owned 1,112, oregn-owned 191, ES Domestc-owned 17, oregn-owned 6, IN Domestc-owned 215, oregn-owned 32, RA Domestc-owned 2,342, oregn-owned 306, BR Domestc-owned 2,216, oregn-owned 589, RC Domestc-owned 200, oregn-owned 15, HRV Domestc-owned 41, oregn-owned 9, HUN Domestc-owned 78, oregn-owned 47, IDN Domestc-owned 773, oregn-owned 106, IND Domestc-owned 2,050, oregn-owned 55, IRL Domestc-owned 130, (%) 2 (%) 3 (%) 4 (%)

25 IA JPN KOR LU LUX LVA MEX ML NLD NOR POL PR ROU RUS SVK SVN SWE UR WN USA ROW oregn-owned 105, Domestc-owned 1,783, oregn-owned 214, Domestc-owned 4,317, oregn-owned 171, Domestc-owned 1,301, oregn-owned 64, Domestc-owned 35, oregn-owned 10, Domestc-owned 44, oregn-owned 16, Domestc-owned 22, oregn-owned 6, Domestc-owned 1,069, oregn-owned 157, Domestc-owned 9, oregn-owned Domestc-owned 661, oregn-owned 167, Domestc-owned 372, oregn-owned 103, Domestc-owned 372, oregn-owned 139, Domestc-owned 179, oregn-owned 36, Domestc-owned 120, oregn-owned 65, Domestc-owned 1,661, oregn-owned 63, Domestc-owned 67, oregn-owned 29, Domestc-owned 37, oregn-owned 7, Domestc-owned 415, oregn-owned 117, Domestc-owned 677, oregn-owned 72, Domestc-owned 467, oregn-owned 66, Domestc-owned 16,149, oregn-owned 1,266, Domestc-owned 9,801, oregn-owned 886, Source: Authors calculatons based on WIOD and the OECD analytcal AMNE database. he decomposton s based on equaton (27). 1 = value-added that has propagated n domestc and global value chans and s absorbed by destnaton countres as an ntermedate product; 2 = purely domestc value-added whch has not been part of nternatonal trade and ends up n domestc fnal demand, 3 = value-added that has partcpated n domestc and global value chans and s ultmately absorbed by destnaton countres as a fnal product, and 4 = value-added that has not partcpated n the domestc propagaton and s absorbed by foregn countres. 24