Rules of Origin for Preferential Trading Arrangements: Implications for the ASEAN Free Trade Area of EU and US Experience

Transcription

1 Journl of Economic Integrtion 22(2), June 2007; Rules of Origin for Preferentil Trding Arrngements: Implictions for the ASEAN Free Trde Are of EU nd US Experience Olivier Cdot University of Lusnne Jime de Melo University of Genev Alberto Portugl-Pérez University of Genev Abstrct With free trde res (FTAs) under negotition between Jpn nd the ASEAN Free Trde Are (AFTA) members nd between the Republic of Kore nd AFTA members, preferentil mrket ccess will become more importnt in Asin regionlism. Protectionist pressures will likely increse through rules of origin, the nturl outlet for these pressures. Bsed on the experience of the Europen Union nd the United Sttes with rules of origin, this pper rgues tht, should these FTAs follow in the footsteps of the EU nd the US nd dopt similr RoO, trding prtners in the region would incur unnecessry costs. Using EU trde with GSP nd ACP prtners, the pper estimtes how the utiliztion of preferences would likely chnge if AFTA were to veer wy from its current uniform RoO requiring 40% locl content rte. Depending on the smple used, 10 percentge point reduction in the locl vlue content requirement is estimted to increse the utiliztion rte of preferences by between 2.5 nd 8.2 percentge points. JEL clssifiction: F13, F15 *Corresponding ddress: Olivier Cdot, University of Lusnne, CERDI nd CEPR: E-mil: hec.unil.ch, Jime de Melo, University of Genev, CERDI nd CEPR: E-mil: Alberto Portugl-Pérez, University of Genev,: E-mil: 2007-Center for Interntionl Economics, Sejong Institution, All Rights Reserved.

2 Rules of Origin for Preferentil Trding Arrngements:~ 289 Keywords: Rules of Origin, Preferentil Trde Agreements, Mrket Access, NAFTA, PANEURO, AFTA, ASEAN. I. Introduction Rules of Origin (RoO) re integrl to the proliferting Free Trde Ares (FTAs) or the non-reciprocl Preferentil Trding Arrngement (PTAs) such s the Generlized System of Preferences (GSP). RoO re necessry in ny PTA to prevent trde deflection, by which is ment tht the country with the lowest externl triff cts s port of entry for the entire bloc s imports, depriving prtners of triff revenue. RoO re bout to become importnt in the trde policy lndscpe in Asi s t lest 45 but up to 70 discrimintory trde dels will be in existence t the end of Typiclly RoO re defined t the HS-6 triff level. However, the Hrmonized System (HS) ws not designed s vehicle for conferring origin, its purpose being to provide unified commodity clssifiction for defining triff schedules nd for the collection of sttistics. As result, devising methods for determining sufficient processing or substntil trnsformtion hs turned out to be very complex in ll existing PTAs, notbly for the two big plyers, the EU nd the US, which use three ctegories of criteri to determine if sufficient trnsformtion hs tken plce in ctivities tht require processing: (i) chnge of triff clssifiction; (ii) criticl threshold for vlue-dded (in short vlue content (VC) rule); nd (iii) specific mnufcturing processes. Currently under AFTA, ASEAN - Chin Free Trde Are (ACFTA) nd the other regionl Asin FTAs under negotition involving the Kore nd Jpn, bilterls with ASEAN ntions, to obtin originting sttus (i.e. to fulfill the criterion of sufficient processing), either of two criteri is used. The wholly obtined criterion is used for few griculturl products. For the vst mjority of products, single vlue content rule requiring tht t lest 40% of the vlue of the finl product must originte from the countries belonging to the FTA (i.e. nonoriginting imports cnnot exceed 60% of the vlue of the finl product) is used in combintion with digonl cumultion (see below). This criterion is remrkbly simple compred with the criteri described here which re used notbly by the two mjor plyers, the EU nd the US. 1 This pper rgues tht it should be kept so. 1 The EU uses such RVC rule-only for 11.46% of its triff lines, while the US does not use ny RVC rule-only.

3 290 Olivier Cdot, Jime de Melo, nd Alberto Portugl-Pérez However, keeping the present simple system might prove to be difficult becuse intr-regionl trde is likely to rise shrply s the economies integrte. Given the limits to other forms of protection imposed by WTO membership, the cndidte outlet for rising protectionist pressures could be move towrds complex system of RoO ( the US nd the EU did it, so why shouldn t we follow in their footsteps! ). Est Asi is now entering phse of very ctive regionl integrtion. Tht integrtion will certinly extend beyond preferences in goods trde with rich vriety of expected gins from regionl integrtion (see e.g. Antkiewicz nd Whlley (2004), nd Bnd nd Whlley (2005)). However, in spite of substntil unilterl triff reductions, preferentil mrket ccess will still be key component of tht regionlism. With preferentil mrket ccess on the rise, nd with more trde tking plce regionlly, the doption of complex product-specific RoO (PSRO) could turn out to be key in the negotitions of triff elimintion mong trding prtners. Thus with trde conflicts likely to occur in the region s the volume of trde continues to increse, this conflict could lrgely ply out in RoO, trde policy instrument tht hs so fr eluded ny discipline from the World Trding System. With the lrge number of trding rrngements under wy, it is very plusible tht lobbying ctivities will resemble those in other PTAs like those involving the US nd EU with lobbies in the most powerful prtners mnging to justify mde-to-mesure RoO to mximize rent-extrction to their benefit t the expense of the weker prtners. This pper rgues tht move in tht direction should be voided. It does so by presenting evidence on the costs of the complex RoO put in plce by the EU nd the US on their trding prtners. Indeed, the experience with RoO elsewhere shows tht RoO diminish the vlue of preferences, in effect reducing the mrket ccess tht the preferences re intended to provide in the first plce beyond wht would pper to be sufficient trnsformtion to confer legitimte origin. Hence, the min messge of the pper for the AFTA prtners is to void the (tempting) trp of moving towrds more complex RoO resembling those observed in other preferentil trding greements (PTAs). To use well-worn expression, RoO should be business friendly rther thn business owned. The pper is orgnized s follows. Section 2 detils the RoO of the EU (these re the rules tht AFTA countries must comply with to benefit from the GSP) nd those under NAFTA ( vrint of these hve been dopted in other FTAs involving the US). Section 3 discusses the economics of RoO nd wht we lerned bout the effects of RoO from NAFTA nd the US experience. Section 4 presents new

4 Rules of Origin for Preferentil Trding Arrngements:~ 291 results tht should be of use for AFTA. Exploiting dt on triff lines where minimum regionl VC is the sole criterion for estblishing origin, we estimte utiliztion rte responses to chnges in preference mrgins nd in the stringency of the VC rule. Section 5 concludes. II. Rules of Origin in US nd EU PTAs With the exception of AFTA (nd few others like ANZERTA nd ANZSCEPsee Tble A1), RoO in PTAs hve two components: regime-wide set of rules nd product-specific rules of origin (PSRO). These re described in the nnex where Tble A1 lists the regime-wide rules nd Tble A2 the PSRO criteri negotited between Mexico nd the US in the context of NAFTA nd the so-clled PANEURO single-list regime used by the EU in ll its PTAs. Tble A1 confirms tht AFTA hs much simpler set of criteri, since it relies only on minimum shre of originting vlue mong prtners (which is equivlent to mximum vlue of non-originting imports). 2 Of course, this minimum originting VC pplies t the product level so in sense it is lso PSRO but becuse it is uniform, in effect it is level-plying field nd AFTA does not hve PSRO. The EU lso uses the VC criterion extensively, though most often this criterion is used long with other criteri (see Tble 1 nd Tble A2). For AFTA, the VC rule is lso its regime-wide rule. AFTA, like the EU PTAs, lso provides for digonl cumultion which is less stringent thn the more usul bilterl cumultion rule pplied in the PTAs described in Tble A1. Under digonl cumultion, countries tied by the sme PTA cn use mterils tht originte in ny member country s if the mterils were originting in the country where the processing is undertken. 3 However, since the domestic content cn be n ggregte of vlue-dded in ny ASEAN member stte, AFTA provides in effect for full cumultion lthough, s noted by Brenton (2006), the rules stipulte tht the finl stge of mnufcture must be crried out in the exporting member stte (wht constitutes the finl process is not defined). Becuse verticl linkges nd outsourcing re very importnt in Asi (see below), full cumultion relxes 2 More precisely, the RoO requires tht the non-originting import content be less thn 60 percent of the FOB price of the product where the vlue of non-originting mterils is bsed upon the CIF import price or the erliest scertined price for products of undetermined origin. 3 See the ppendix for definition of bilterl, digonl nd full cumultion.

5 2 9 2 Tble 1. Distribution of PSRO under NAFTA nd PANEURO % of triff lines No other requirement or EXC TECH or TECH+EXC VC or VC+EXC TECH+VC or TECH+VC+EXC Wholly obt'd Other ddit. Requir. SUBTOTAL (WH) NAF. PAN. NAF. PAN. NAF. PAN. NAF. PAN. NAF. PAN. NAF. PAN. NAF. PAN. No CTC CI CS CH CC Altern Totl Notes: Clssifiction is crried out t the HS-6 triff line level (#5595 lines for the EU nd # 3995 for NAFTA). Ech cell is the percentge of triff lines tht hve the RoO in the corresponding row nd in the corresponding column. CTC = chnge in triff clssifiction with CC = Chnge in Chpter / CH = Chnge in Heding / CS = Chnge in Subheding / CI = Chnge in Item; EXC = Exception to chnge of triff clssifiction; VC = Regionl Vlue Content; TECH = Technicl Requirement. Altern. = Lines for which importers cn choose between lterntive criteri. Source: Cdot et l. (2006b), Tble 2 O liv ie r C d o ț J im e d e M e lo, n d A lb e rto P o rtu g ḻ P é re z

6 Rules of Origin for Preferentil Trding Arrngements:~ 293 Tble 3. Protection, the PSRO Restrictiveness Index, nd Utiliztion rtes in Textiles nd Apprel 3. Triff Peks nd the PSRO index R-Index vlue,b NAFTA PANEURO Triff peks c 6.2 (257) 5.18(780) Low triffs d 4.8 (1432) 3.92(3241) Totl number of triff lines / Source. Cdot et l. (2006b), tble 3. b/ RoO indexes re unweighted indexes. c,d /Triff peks (low triffs) re clculted for ll triff lines tht exceed 3 times (one third of) the verge GSP triff level. 3b: Preferences nd Utiliztion Rtes in Textiles & Apprel (HS-11) (verges, unweighted) #Observtions e Utiliztion rte Preferentil Mrgin NAFTA (2001) GSP (2004) ACP (2004) 16555(HS-8) (HS-6) 1370 (HS-8) 1175(HS-6) considerbly the requirements of stisfying origin. Thus under full cumultion, the more developed higher lbor cost countries cn outsource lbor-intensive, lowtech, production stges to less developed lower-wge prtners whilst mintining the preferentil sttus of the good produced in the low-cost loctions. Tble 1 tbultes the criteri used for PRSO under PANEURO nd NAFTA. It shows tht the two regimes use lrge list of criteri. These typiclly include technicl requirements (such s the triple trnsformtion requirement in textiles nd pprel which requires tht pprel must be woven from originting fbric nd from originting yrn). The criteri lso include exceptions nd llownces. NAFTA relies more hevily on chnge of triff clssifiction (CTC), though often ccompnied by other criteri. PANEURO relies mostly on VC nd on the Wholly obtined (WH) criterion (this criterion is not used in NAFTA since Mexicns export fr fewer griculturl exports thn GSP nd ACP countries). The key issue is how RoO ffect the utiliztion of preferences (i.e. do importers end up using the MFN sttus when preferences would suggest they should clim for preferentil sttus)? Tble 2 nd Figure 1 show verge utiliztion rtes computed for different thresholds of minimum preferentil mrgins for NAFTA nd for the two EU preferentil schemes, the Generlized system of preferences (GSP) for which ASEAN countries (except Singpore qulify) nd the Cotonou

7 294 Olivier Cdot, Jime de Melo, nd Alberto Portugl-Pérez Figure 1. Averge utiliztion rtes for different thresholds of preferentil mrgin rtes Tble 2. Preferences nd their utiliztion in the NAFTA nd EU mrket(un-weighted dt) GSP(92) b ACP(37) b NAFTA c τ 4% 50.2% (1297) d 92.5% (1627) d 87.0% (1239) d τ 8% 52.5% (91) 94.3% (892) 86.0% (558) τ 12% 66.2% (44) 96.4% (566) 82.8% (287) MFN PREF PREF τ i = ( t i t i ) ( 1 + t i ) is the preference mrgin. b Computed t the HS-8 triff line level for GSP nd ACP. c Computed t the HS-6 triff line level for NAFTA. d Number of triff lines between brckets. Source: Authors clcultions preferences which benefit the former colonies of the Afric, Cribben nd Pcific (ACP) regions. 4 Since there re dministrtive costs (see below), only utiliztion rtes bove 4% re considered in Tble 2 which lso shows in prenthesis the 4 The reson for distinguishing between GSP nd ACP utiliztion rtes is tht ACP preferences hve been in effect longer, tht ll triff lines qulify for ACP sttus, tht some GSP-eligible countries lso qulify for EBA but hve chosen not to use the more fvorble sttus, perhps becuse these preferences were only in effect since 2002 or finlly becuse of slightly more stringent regime-wide rules for GSP-eligible countries thn for ACP countries. Also ll AFTA countries except Singpore qulify for GSP sttus.

8 Rules of Origin for Preferentil Trding Arrngements:~ 295 number of triff lines over which utiliztion rtes re computed. As reference, the verge preferentil mrgin (computed over triff lines with positive triffs) ws 4.5% for NAFTA (lmost ll triffs hd been eliminted on NAFTA trde by 2001), 2.4% for GSP-eligible countries nd 4.6% for ACP countries (not eligible for EBA sttus). EU dt is for 2004, when 62% of trde for GSP-eligible countries took plce t zero triffs nd over 80% for ACP countries (some ACP lso benefiting from EBA sttus t zero triffs in the EU mrket). Severl ptterns stnd out. First, utiliztion rtes by Mexicn exporters re systemticlly higher thn utiliztion rtes by eligible exporter to the EU s GSP. This could reflect severl fctors including closer proximity nd knowledge of the US mrket, but lso less stringent RoO rules or greter dministrtive costs for EU schemes thn for NAFTA (e.g. relting to certifiction methods). 5 The differences in utiliztion ptterns lso probbly reflect prtly differences in the composition of exports. Second, utiliztion rtes re systemticlly higher for ACP countries thn for GSP countries (this is why the number of triff lines is lwys greter for ACP countries thn for GSP countries t ny triff line, even though there re fewer ACP (78) thn GSP (92) countries). 6 Agin, it could be tht better knowledge of procedures (sme lnguge for former colonies) re fctor, but the combintion of higher preference mrgins nd better mrket ccess due to less stringent rules of origin must lso be determining fctor. Regrding RoO, the regime-wide rule for ACP countries re less restrictive tht for GSP qulifiers since they benefit from full cumultion rther thn digonl cumultion nd they hve more fvorble tolernce rule (15% tolernce for ll triff lines insted of 10% nd n exclusion of ccess to the tolernce rule for chpters 50 to 63 which cover textiles nd pprel (T&A) for the group of 92 GSP-eligible countries). Third, s cn be seen from Tble 2 nd Figure 1, utiliztion rtes do not rise monotoniclly with preference mrgins, s one would expect. Indeed, utiliztion rtes fll s preference mrgins increse for NAFTA nd lso for the GSP-eligible 5 For exmple, under NAFTA certifiction is crried out by the privte sector nd does not hve to be crried out for ech trnsction, wheres under PANEURO regime there is double certifiction method (privte nd public) for ech trnsction. See Estevdeordl nd Suominen (2006) for further discussion nd comprison of certifiction methods cross PTAs. 6 Of the 78 ACP countries, 41 qulify for Everything But Arms (EBA) sttus which gve these countries virtully duty-free ccess to the EU mrket. However, in 2004, most ACP countries qulifying for EBA sttus continued to request ACP sttus. Hence when we spek of ACP countries, we refer to the 78 ACP countries.

9 296 Olivier Cdot, Jime de Melo, nd Alberto Portugl-Pérez Figure 2. Growing Verticl integrtion in the World Economy countries over certin rnge. This is due to composition effect. In the T&A sector where preference mrgins re round or bove 10%, utiliztion rtes re low. However. GSP-eligible countries get much less preferentil ccess thn ACP countries in the T&A sector (see Tble 3b). Since these sectors re chrcterized by specific mnufcturing requirements, lobbying ctivities by the domestic industries in the high-income prtner influence the specifiction of the rules, nd the utiliztion rtes re low even though preferentil mrgins re high. 7 Bsed on n nlysis of utiliztion rte in tht sector for NAFTA, Anson et l. (2005) nd Cdot et l. (2005 nd 2006), conclude tht the PSRO requirements in tht sector were designed to leve Mexicn exporters on their prticiption constrint, i.e. leving only mrginl rents for Mexicn exporters. Are these utiliztion rtes low? Certinly not by AFTA stndrds which show utiliztion rtes in the 5%-10% rnge for the lte 1990s for the Philippines, the 7 Brenton nd Imgw (2005, Box 9.2, p. 20) mention prticulrly egregious cse in which NAFTA s RoOs for pprel specify tht imported fbric must be of subheding or , if hnd-woven, with loom width of less thn 76cm, woven in the United Kingdom in ccordnce with the rules nd regultions of the Hrris Tweed Assocition, Ltd, nd so certified by the Assocition. Such specific rules, which clerly smck of firm-specific lobbying re, however, reltively uncommon.

10 Rules of Origin for Preferentil Trding Arrngements:~ 297 country with the highest preference utiliztion rte (see Bldwin (2006), Figure 5). Bldwin rgues tht this is becuse the bulk of intr-asean trde is in the computer/mchinery (HS-84) nd electricl equipment (HS-85) sectors where preferentil ccess mrket is negligible. However, the use of VC criterion could lso contribute to this low rte of utiliztion of preferences even if t first sight meeting this requirement would not pper to be too difficult since most prts re indeed produced in the region, t lest for the computer hrd-disk exmple shown in his Figure 8. However, the very high import content of exports for the Asi region shown in Figure 2 suggest tht AFTA utiliztion rtes could be low under the present VC limit. To drive home the importnce of trde in intermedites, consider the following exmple reclling tht Jpn is not currently member of AFTA. According to the Asin input-output dt provided in Bldwin (2006, Tble 1), for ll the middleincome countries (Indonesi, Philippines, Mlysi, Thilnd), on verge 35%- 40% of intermedites re sourced from outside the AFTA group. Tke then n ctivity with 10% vlue-dded (this vlue-dded Figure is consistent with the deepening of inter-industry flows) nd 40% intermedites non-originting. Then, originting vlue for this ctivity would be, 64%. But for ctivities where 60% of mterils would be non-originting, originting vlue would fll to 46%, brely bove the stipulted 40% minimum requirement stipulted under AFTA. One might lso note tht, in the cse of the EU, the preference rte for triff lines where vlue-content only criterion is used is lower thn for the other triff lines (see Tble 5). Returning to the EU nd US PTAs, the rther low utiliztion rtes for reltively high preference mrgins suggests tht restrictive PRSO could be the culprit. Indeed, it could well be tht more restrictive PSRO re pplied to the triff lines with the highest preferentil mrgins. To see if this is the cse, we use n ordinl restrictiveness index tht tkes vlues in the rnge 1 r i 7 with incresing vlues corresponding to more restrictive rule. 8 Though not menble to quntifiction like effective rtes of protection, the R-index is intended to ply the sme role s n 8 The index ws first proposed by Estevdeordl (2000) for NAFTA. The construction of the R-index is illustrted in the ppendix nd discussed in further detil in Cdot et l. (2006). In ddition to the inevitble rbitrriness involved in setting up the observtion rule, the R-index hs other shortcomings. In prticulr, it does not control for the degree of preferences nd for the chrcteristics of the different ctivities: stisfying CTC involving CH for intermedite ctivities is likely to be esier thn if it is to be stisfied for finl good ctivity. Finlly, it is n ordinl rther thn crdinl mesure. For description of some of the shortcomings of the R-index, see Ersmus et l. (2006).

11 298 Olivier Cdot, Jime de Melo, nd Alberto Portugl-Pérez effective rte of protection: indicte the restrictiveness tht must be met by product to obtin originting sttus. Tble 3 shows tht triff lines with triff peks (triffs lines three times or more the verge triff) tht is triff lines where preferentil mrgins re the highest do indeed hve higher vlue for the R-index thn those with low triffs. This is the cse for both NAFTA nd the EU. The conclusion from this first look t preferentil mrgins nd utiliztion rtes is tht, even for non-negligible preferences, utiliztion rtes cn be quite low, nd tht PSRO re, t lest prtly, designed to discourge the use of preferences. This is prticulrly true for the T&A sector where multiple criteri re the norm. Once more, one might sk how relevnt these observtions re for AFTA. At this stge lmost ll intr-asean trde is in middle-products (see Snyl nd Jones (1982), the finl product being sold to the outside world. But the picture would chnge rther drmticlly with the proposed FTAs between Jpn nd AFTA members nd between Kore nd other AFTA members. Then, the resulting pttern of trde would be closer to the one currently observed between the US nd the EU nd their Southern trding prtners. Here, differences in endowments mtter with the resulting pttern of trde being of the inter-industry rther thn intr-industry pttern. As soon s inter-industry trde involves the sourcing of frction of intermedites from the outside world nd there is potentil for regionl trde in the finl product (e.g. in pprel nd other light industry products), the experience of the EU nd US becomes relevnt. Est Asin trde hs grown remrkbly rpidly in the lst 15 yers through the breking up of the mnufcturing process or the slicing up of the vlue chin. With triff cuts on discrimintory bsis in the Chin-ASEAN FTA nd the Jpn- ASEAN bilterls on the tble, tensions re likely to pper in the region. One likely element of these tensions will be the negotition over Rules of Origin (RoO). So why re RoO in AFTA pprently less stringent thn elsewhere? First, until recently Asin regionlism hs been more bout coopertion thn bout preferentil trde (APEC under the egis of the US ws specificlly set up to void preferentil trde nd the formtion of n Asin trde bloc ), with much of the Region s integrtion in the world economy driven by unilterl triff reductions. Second, much of the regionl trde hs been prt of the development of the Asin mnufcturing mtrix whereby the lbor-intensive stges of production initilly crried out in Jpn nd then Kore hve been outsourced to the lower-wge countries in the region for finl exporting to the EU nd US mrkets. In this unusul set-up reltive to the other globl trding ptterns, intr-regionl trde in

12 Rules of Origin for Preferentil Trding Arrngements:~ 299 politiclly sensitive finl products where protection is highest ws insignificnt. Hence, the politicl-economy forces tht would usully led to lobbying ctivity resulting in the complex RoO observed elsewhere ws not observed. III. The Economics of RoO RoO re often justified s n instrument to promote industriliztion in lowincome countries (or vlue-dded-generting ctivities). In fct, they impose costs on the intended beneficiries of preferences. In ddition, contrry to wht is climed, rther thn help countries industrilize, they end up suppressing trde, n effect tht could be prticulrly importnt in AFTA countries where much frctionliztion of production tkes plce (Ersmus et l. (2006) provide vivid exmples bsed on cse studies in SADC where RoO hve been climed to be development tool for low-income members). The suppression effect of RoO cn be esily visulized if one considers tht the production of finl product involves continuum of ctivities, with the rnge of ctivities being ordered in decresing lbor-intensity. For exmple the sowing of buttons on shirt could tke plce in low-income Cmbodi, the weving in Indonesi, the production of yrn in Indonesi. Then imposing minimum regionl VC would suppress trde with the outside prtner (Bngldesh) nd redirect investments to the region. As result, production costs would increse nd efficiency would fll. Below we present generl decomposition formul of the costs of RoO using minimum VC. This decomposition motivtes the review of evidence tht follows nd serves justifiction for the new estimtes presented in section 4. A. The Costs of RoO: decomposition Suppose tht producer in the Philippines wishes to sell shirt, X i, under preferentil ccess in nother AFTA member, Thilnd. This shirt is produced with originting intermedite goods (i.e. with intermedites from other AFTA members nd from domestic sources),, but lso with inputs from nonoriginting sources, Z j N, sy from Bngldesh or Indi. Let i stnd for the shirt nd j for the intermedites, sy textiles. Omit ny txes pid domesticlly, nd lbel ech input with its own price since intermedites re differentited products. Let VA i represent the pyments to lbor (nd to cpitl or profits). Then the vlue of finl sles is broken down between pyments to vlue dded nd to intermedite products: Z j o

13 300 Olivier Cdot, Jime de Melo, nd Alberto Portugl-Pérez O O N N P i X i = VA i + P j Zj + P j Zj (0.1) To obtin originting sttus, intermedite purchses from domestic producers of textiles nd from textile producers in the regionl prtner countries re counted s originting i.e. P j O Zj O counts s regionl vlue-dded. Then, unrestricted originting vlue-dded expressed s percent of unit sle (evluted t FOB prices in the cse of AFTA) is given by the expression: v i * = ( VA i + ) P i X i P j O Zj O (0.2) Now ssume tht to stisfy origin requirements to sell under preferentil sttus in Thilnd, the producer in the Philippines is forced to source more originting inputs thn he would under optiml circumstnces (i.e. in the bsence of the PSRO). If we let n * denote the initil unconstrined (i.e. optiml) choice, nd c constrined choice, then modeling the effects of binding vlue content boils down to modeling the implictions of: v i c > v i * (0.3) In the cse of AFTA, C ( v i 40% ) is the uniform criterion used to qulify for originting sttus (lthough producers lso hve the option of using other qulifying requirements, notbly in the T&A sector where the lterntive PSRO were used becuse the vlue-content rule ws found to be too restrictive). Figure 3, dpted from Krishn (2006) shows the costs implictions of forcing producer to use more originting inputs thn he would otherwise. Assume competitive environment with constnt- returns-to- scle technology where the producer who tkes his cost-minimizing decision on the bsis of given prices. When he is unconstrined in his purchse of intermedites, (i.e. when z ( z ROW z AFTA ) = z * ), his unit costs re given by c = c 0. When he is constrined (i.e. when z = z C ), his unit costs re higher nd given by c = c 1. Being forced to use greter shre of originting inputs is equivlent to forcing him physiclly to use greter proportion of originting intermedites. As result his unit costs of production increse to c 1. Let by this distortion. To this, one must dd dministrtive costs, D 1 0 c i ( = c i c i ) be the cost induced. Though not shown in Figure 3 to sve spce, it would be esy to do so in similr frmework by reinterpreting the c i A

14 Rules of Origin for Preferentil Trding Arrngements:~ 301 Figure 3. Rule of Origin nd Costs isoqunt s n isovlue (i.e. it is drwn to reflect constnt vlue of output). Let * 0 0 * then P i = P i = 1 nd choose the constnt vlue, v i so tht v i = P i Xi = 1. With zero profits, unit costs will equl unit vlue 0 0 v i = c i nd the producer will produce X = 1 in equilibrium. With the product price unchnged, the effect of n increse in unit production costs is to mke the producer unprofitble, unless he is compensted by sufficient preferentil mrket ccess. Finlly, consider rents nd mrket power. Recent estimtes reviewed below lso indicte tht only frction of the rent ssocited with mrket ccess ccrues to exporters nd tht the shre of rents they receive is lessened by the presence of RoO. Cll the loss of rent trnsfer ssocited with RoO,. Then per unit µ i µ i production cost in the presence of RoO will be the sum of costs in the bsence of C RoO nd complince costs, c i : 0 D A 0 C c i = c i + c i + c i + = c i + c i µ i (0.4) Under these conditions, the Philippine producer cn either export to Thilnd under the Common Preferentil Extended Triff (CPET) scheme where the preferentil triff rte will be given by: τ i = MFN PREF t i t i MFN 1 + t i

15 302 Olivier Cdot, Jime de Melo, nd Alberto Portugl-Pérez or under MFN sttus. Under MFN sttus, unit price will be 1 nd under preferences,. Since unit profits re given by =, if u i denotes 1 + τ i π i P i c i whether or not CPET preferentil sttus is requested: D A τ i c i + c i + µ i u i = 1 D A τ i < c i + c i + µ i u i = 0 (0.5) In prctice, mny heterogeneous firms export t the HS-8 level. Thus one will observe distribution of utiliztion rtes t the most disggregted HS-8 level. Since there is no dt to mtch firms with utiliztion rtes, the estimtion of utiliztion rtes nd their correltes in effect ssumes tht ll firms exporting t the HS-8 level re in effect, identicl. This evidence is now reviewed. B. Evidence on the Costs of RoO It is difficult to estimte the costs ssocited with RoO becuse these costs re not observble. Estimting the costs of RoO is further complicted by the fct tht, with the exception of the estimtes presented in section 4 below, ll econometric estimtes so fr hve been obliged to rely on dummy vribles to cpture the presence of RoO requirement (i.e. one cn only note the presence or the bsence of, sy, n exception or technicl requirement). Finlly, dt on utiliztion rtes re only vilble for NAFTA (nd more recently for the EU preferentil schemes). As result, sttisticl evidence reltes to the EU nd US schemes, nd furthermore it is prtil. Below re the highlights of the econometric evidence. Distortionry Costs Most of the evidence rests on isolting the effects of vrious PSRO criteri on utiliztion fter controlling for the influence of preferentil mrgins. Crrère nd de Melo (2006) ssume tht the utiliztion of preferences for product line i is positive function of the difference between the triff preference mrgin,, nd (unobserved) totl complince costs, c i (expressed s percentge of unit price) ssocited with pplying the RoO criteri. Tht is they suppose tht. ( ); f (). > 0, nd tht = ( ) g (). > 0. These ssumptions u i = f τ i c i c i g RoO i led to n estimble reltion of the form: τ i u i = λ + ατ i + θ + k k RoO ik ε i (0.6)

16 Rules of Origin for Preferentil Trding Arrngements:~ 303 where Roo ik is vector of dummy vribles cpturing the presence of PSRO (technicl requirements, chnge of triff clssifictions, exceptions). Results from estimting on dt from NAFTA by the uthors (nd by Cdot et l. (2006) for the EU) indictes tht utiliztion rtes re positively relted to preferentil mrgins, nd negtively to the presence of PSRO. Crrère nd de Melo combined their estimtes with R-index vlues to compute n estimted d-vlorem equivlent of totl RoO complince costs (dministrtive nd due to higher input costs). Their estimtes rnge from 3.5% for chnge of chpter to over 15% for combintions of RoOs involving technicl requirements. Even if the estimtes point in the sme direction, it is difficult to gther sense of robustness from estimtes derived from reltion like (0.6). The reson is tht there is so much heterogeneity nd unobservbles influencing preference uptke tht estimtes re quite sensitive to the inclusion of dummy vribles controls (this is especilly the cse for estimtes bsed on GSP preferences extended by the EU to mny trding prtners). This sid, ll estimtes point in the sme direction. When multiple criteri re used to determine origin, utiliztion rtes re, ceteris pribus lower. Moreover, results bsed on estimtes regressing utiliztion rtes on dummy vribles cpturing the presence of different types of PSRO, suggest tht technicl requirements tend to be ssocited with high complince costs, wheres chnges of triff clssifiction, even t high degrees of ggregtion (like chnge of chpter) do not seem to hve comprbly strong effects. Of course these suggestive conclusions must be tken very cutiously s these instruments re in most cses used in combintion with ech other nd their effects re only cptured by dummy vribles, mking it difficult to drw inferences on their individul effects. However they re consistent with the csul observtion tht technicl requirements cn be esily mnipulted by specil interests, nd suggest tht brod relince on chnge of triff clssifiction s one-size-fits-ll origin criterion would be more desirble direction to go. Administrtive Costs An lterntive to the bove is to use the R-index to crry out non-prmetric estimtion of upper nd lower bound estimtes following inspirtion from erly work by Herin (1986) for EFTA. Tble 4 summrizes the results of n exercise in Cdot et l. (2006b). By reveled preference, for hedings with u i =100%, the C C verge preference mrgin, τ i, is n upper-bound, c U, for complince costs (s c i

17 304 Olivier Cdot, Jime de Melo, nd Alberto Portugl-Pérez Tble 4. Complince Cost Estimtes (percentge of unit costs) PANEURO NAFTA Row Averge Averge τ i Totl Complince Costs (c C ) τ i u i =0 1 0<u i <1 2 u i =1 3 c L C C = τ i = 4.7 c L = τ i = 0.3 c = τ i = 7.2 c = τ i = 6.2 c U C C = τ i = 8.2 c U = τ i = 7.1 Distortionry cost (c D ) R i 2&u i 95% 5 c D = τ = 3.8 c D = τ = 4.4 R i 2&u i 90% 6 c D = τ = 3.9 c D = τ = 4.3 Administrtive cost estimte (c D ) 7 c D = D c c U = = 3.4 c D = D c c U = = 3.3 c D = D c c U = = 1.8 c D = D c c U = = 1.9 τ i = 0 & Request c 8 # 691 lines # 1089 lines b τ i 3% & Request c 9 # 2173 lines # 1972 lines b Source: Cdot et l. (2006b) Tble 6. Notes: Dt for NAFTA refer to Mexicn exports to the US for Dt for PANEURO refer to request by ACP countries for Cotonou sttus (dt for GSP countries were not used becuse EBA sttus ws only in plce since 2001 ). Trde-weighted estimtes. Totl number of triff lines (t HS-6 digit): b Totl number of triff lines (t HS-6 digit): 3555 c Only for PANEURO cnnot be greter thn the preferentil mrgin, τ i ). Likewise, for hedings with u i = C 0%, the preference mrgin gives lower-bound estimte, cl. For the remining sectors with 0% < u i < 100%, ssumptions must be mde. One could rgue tht, heterogeneity of firms notwithstnding, firms would be indifferent to exporting under the preferentil regime. Then, n pproximtion of complince costs would be given by the verge rte of triff preference computed for the remining sectors, i.e. on the smple 0% < u i < 100%. Applying this resoning, we obtin the estimtes in rows 1-3 in Tble 4. The R-index cn then be used to brekdown the dministrtive cost component in totl complince costs. Ignore rents, i.e. ( µ i = 0) so tht totl complince costs, c C C D A, re given by c i = c i + c i, both expressed s percentge of unit price. Assume tht dministrtive costs would be negligible for firms on their

18 Rules of Origin for Preferentil Trding Arrngements:~ 305 prticiption constrint, i.e. for (0%<u i <100%), provided tht firms would lso be confronted with low vlues of r i, i.e. vlues corresponding to chnge of triff clssifiction t the subheding level, CS. This corresponds to r i 2 (not much pperwork is involved in "proving" chnge of subheding). Hence, clculting preference mrgins for utiliztion rtes close to 100% (sy u i =95% or u i =90% to be on the sfe side when r i 2 ), gives n upper bound of the distortionry D component, c i. These estimtes re given in row 7. As noted by Cdot et l., the rther lrge differences between the PANEURO nd NAFTA estimtes, especilly when using unweighted dt, suggest cution in interpreting these estimtes since too much weight is probbly given to outlier observtions with smll trde volumes. Nonetheless, the higher estimtes for the EU re coherent with likely higher dministrtive costs (perhps ssocited with certifiction), nd ultimtely with lower utiliztion rtes. The estimtes in Tble 4 indicte the difficulty of coming up with plusible estimtes of the costs of RoO, whether one relies on prmetric or non-prmetric methods. For exmple, how should one interpret the number of requests for preferentil sttus in rows 8 nd 9 when triff preferences re nil? Negligible dministrtive costs?. Or if dministrtive costs re non-negligible, then neither should one observe requests for preferentil sttus when preferentil mrgins re low, here ssumed to be τ. 9 i = 3% In ny cse, even if there is much hetero-geneity in the dt, dministrtive costs re not negligible, justifying the proposl to do wy with origin requirements for preferentil mrgins below n mount round 5%. Mrket power To get hndle on the welfre effects of RoO, one hs to fctor in the rent element ssocited with preferences nd their distribution between the exporting nd importing country. This implies estimting the pss-through effect of triffs on consumer prices (i.e. the extent to which preferences trnslte into higher producer price for exporters). Olrreg nd Özden (2005) for AGOA preferences, nd Özden nd Shrm (2006) for CARICOM preferences estimte tht between one-third nd one hlf of triff reductions re pssed on to producers. However, prt of the border price increse could just reflect n increse in the complying costs discussed bove. Cdot et l. (2005) link the pss-through effects of preferences to Mexicn producers in the Textiles & Apprel sector under NAFTA 9 Using threshold estimtion techniques, Mnchin (2004) estimtes for the sme dt tht preferentil sttus is not sked when τ i < 3 4%

19 306 Olivier Cdot, Jime de Melo, nd Alberto Portugl-Pérez directly to rules of origin, nd show tht once one tkes into ccount RoO, the pss-through effect flls from 80% to bout 50%. They lso show tht US producers of intermedites re ble to retin substntil shre of the rents generted by Mexicn triff preferences. They conclude tht in North-South PTAs where the preferentil mrgin in the Northern prtner re rther low, it is likely tht RoO re less development tool to prevent screwdriver ssembly thn circuitous wy of rising the profits of upstrem producers by creting cptive mrket for them in prtner countries. At this stge, AFTA integrtion hs been mostly gered towrds supplying products to the outside world, so the rent trnsfer ssocited with preferentil ccess hs not been n issue. However, s preferentil ccess is bound to increse with the current FTAs under negotition involving Jpn nd Kore with AFTA members, protectionist pressures will increse. A pttern of verticl trde long the lines described here would then be likely to emerge between the richer nd poorer members in the PTA nd the distribution of rents might be mostly tilted towrds the more developed prtner. IV. Vlue Content Rules nd Preference Utiliztion: Estimtes from the EU Preferentil Mrket Schemes Currently AFTA hs VC rule. How restrictive is this rule, nd how might utiliztion rtes rect to chnge in the VC rte from the current mximum limit of 60% for non-originting imports to lower (or higher) rte? We drw here on the EU dt to estimte wht these effects might be. As explined below, becuse of the gret heterogeneity in the dt, we confine our estimtes to triff lines tht rely on VC rte s the only PSRO. Tble 5 describes the distribution of the minimum locl vlue content 10 when this criterion is pplied long with the distribution of minimum threshold vlues. Currently minimum originting vlue rnges from 50% to 85%. 11 There is 10 In fct, under the current PSRO, the criterion is specified s mximum non-originting imports expressed s percentge of the ex-works fctory price, i.e. of unit price t the fctory. Since the EU is currently considering move towrds simplifiction of RoO procedures tht might result in single criterion requiring minimum originting vlue, we hve expressed here this criterion in terms of minimum originting vlue. 11 For the wholly obtined criterion, originting vlue hs to be 100%. Since this criterion only pplies to griculturl products nd is not relevnt to the AFTA context, is it not considered here.

20 Rules of Origin for Preferentil Trding Arrngements:~ 307 Tble 5. Vlue-Content Requirements in EU PTAs nd utiliztion rtes RVC rte b Frequency Percent u i (GSP) d u i (ACP) d 50 4, % (2.6) 42% (2.6) 60 14, % (1.1) 18% (2.7) % (2.3) 29% (3.0) 80 1, % (2.9) 93% (4.3) % (3.6) 0% (3.7) Totl 19, N.A. N.A. c τ i 5% 36% 34% c τ i 10% 34% 58% Frequency rtes re for triff lines where vlue-content criterion is the sole criterion for obtining origin b The Figures in the column indicte the minimum regionl vlue content to stisfy the sufficient processing threshold c From Tble 2 d Utiliztion rte nd verge preference rte in prenthesis bunching round the 60% vlue but there is still some vrition in VC rtes cross triff lines, giving hope tht the dt will enble identifiction of the effects of vritions in VC rtes on utiliztion rtes. The Tble shows tht utiliztion rtes re usully higher for ACP thn for GSP, with lrge vritions in utiliztion rtes in spite of reltively smll vritions in preferentil mrgins tht re quite low. For AFTA, the closest VC rte is the 50% RVC rte where the utiliztion rte is round 40% for preferentil mrgin of 2.6%. Following the discussion in section 3.1, ssume tht the profit rte for product line i, π i is n incresing function of the preferentil mrgin, τ i, nd decresing vlue of the originting vlue content, vc i,: π i = α 0 + α 1 π i + β 1 vc i + ε i (0.7) where the nticipted signs for the prmeters or interest re: > <. α 1 0; β 1 0 There is no observble mesure of this benefit, so it is ltent vrible. Wht is observed insted is the utiliztion rte, u i which obeys the following lw: u i = 0 if π i < π 0 f( π i ) π 0 < π i < π 1 1 if π 1 < π i (0.8) for some function f nd bounds π 0 nd π 1. These bounds, which re not known,

21 308 Olivier Cdot, Jime de Melo, nd Alberto Portugl-Pérez cn be thought of s vlues of the preferentil tretment s net benefit tht re respectively so low ( π 0 ), nd so high ( π 1 ), tht ll firms (heterogeneity cross firms is ignored) in the sector end up mking the sme decision (using or not using the preferentil tretment). For simplicity, ssume liner form for f, so tht follows (0.7) in the rnge where it is not censored t zero or one. This leds to the model: u i u i * = α 0 + α 1 π i + β 1 vc i + ε i * u i = 0 if u i 0 * u i = u i * if 0 u i 1 (0.9) * u i = 1 if u i 100 where u i * is the ltent utiliztion rte nd u i is the observed one. Turning to the estimtion of (0.9), to void giving excessive weight to triff lines with smll trde flows, triff preferences were ggregted using imports s weights from the HS-8 to the HS-6 level where the vlue content rtes re specified. In ddition, to minimize mesurement error, we took n verge of utiliztion rtes over the whole period for which dt were vilble (2002 to 2004) nd we considered seprtely GSP (including EBA) beneficiries nd ACP beneficiries becuse of the systemticlly differentil utiliztion rtes detected in Tble 2 nd Figure 1. As mentioned bove, to reduce the effects of omitted vrible bis, we only considered triff lines where VC ws the only criterion used. Agin, this would correspond most closely to the current RoO in Asin PTAs. Finlly becuse of dministrtive costs, we lso restricted our estimtes to triff lines with preferentil mrgins in excess of 2% or 5%. The results pper in Tble 6. The overll fit for the models summrized in the likelihood-rtio vlues re resonble (see row 6). Averge estimted vlues for preferences rtes pper in row 9 nd for utiliztion rtes in row 10. Note tht the utiliztion rtes re low for these triff lines, suggesting tht vlue content criteri re rther restrictive. Estimted utiliztion rtes re slightly higher for GSP-eligible countries thn for ACP countries. Since estimtes re influenced by omitted vribles, it is best not to speculte on this smll difference in utiliztion rtes cross smples. Turning to the estimtes of the individul vribles, ll signs of estimted coefficients re s expected. Estimtes in columns 1 nd 4 re crried over the whole smple, with the other columns referring to seprte estimtes for GSP nd

22 Rules of Origin for Preferentil Trding Arrngements:~ 309 ACP recipients. Since the vrible entered for the VC is the minimum regionl vlue content, the negtive sign estimte indictes tht, s expected, higher minimum regionl vlue content lowers utiliztion rtes. Since coefficients derived from Tobit estimtion indicte the expected mrginl effect of the coefficients on the ltent utiliztion rte, u i, we computed the 12 expected mrginl effect of the VC on the observed utiliztion rtes, u i. Row 8 trnsltes this estimte into corresponding chnge in the estimted utiliztion rte for reduction in the mximum regionl content from 60% to 50%. For exmple, in the cse of the GSP estimtes in column 2, this reduction would rise the estimted utiliztion rte by 2.5 percentge point (row 8 column 2) from the estimted verge of 16.7 percent. The corresponding estimte for the ACP grouping is 8.2 percentge points, n estimte on the high side since it predicts tht the utiliztion rte by more thn 50%. The lst three columns report results for smller smple including only triff lines with preference rtes exceeding 5%. While the pttern of signs is unchnged, coefficient vlues often vry by lrge mrgin, suggesting tht the estimtes re not very stble. Concentrting on the estimted effects of chnge in the mximum regionl content from 60% to 50%, one gets the plusible result tht utiliztion rtes re less sensible to chnge vlue content rte t triff lines with higher preference rtes. As finl exercise, the bottom of the Tble estimtes wht utiliztion rtes would be for preferentil rtes in ACFTA. Chinese nd Philippine verge MFN triff rtes (row 13 nd 15) re used to proxy for the extent of likely preferentil mrket ccess for ASEAN exports into Chin nd the Philippines, or vice-vers. In this illustrtive simultions, the VC is set to the AFTA level (40% in row 12), which is likely to be the VC rte under ACFTA. Results of this simultion re reported on rows 14 (Chin) nd 16 (Philippines) t the bottom of Tble 6. As expected, predicted utiliztion rtes re higher when using the Chinese MFN triff, which is more thn twice higher thn the Philippine one. 12 * Tobit coefficients re the expected mrginl effects of the coefficients on the unobserved u i. Unlike in the clssicl liner model, they cnnot be interpreted directly. Insted, expected mrginl effects on the observed u i must be clculted by correcting the tobit coefficients for the censorship. For continuous regressors, the mrginl effects re the Tobit function s derivtives nd thus give the effect of infinitesiml increses in regressors on the dependent vrible. For binry regressor (dummy vrible), the mrginl effect gives the effect on the dependent vrible of n increse from zero to one in the regressor. Becuse of the Tobit s nonlinerity, mrginl effects vry with the level of ll regressors. By convention, they re estimted t their men vlues.

23 310 Olivier Cdot, Jime de Melo, nd Alberto Portugl-Pérez Tble 6. Determinnts of Utiliztion Rtes for EU Preferentil Schemes (Triff lines withvlue-content Rule only) Row 1 (GSP+ACP) 2 (GSP) 3 (ACP) 4 (GSP+ACP) 5 (GSP) 6 (ACP) Dependent vrb. u i u i u i u i u i u i Constnt (4.52)*** (4.99)*** (10.68)*** (2.36)** (6.38)*** (tr. Pref.)(GSP) (12.74)*** (10.89)*** (5.35)*** (2.80)*** (7.47)*** (7.14)*** (11.66)*** (3.06)*** (3.56)*** (6.96)*** (tr. Pref.)(ACP) (10.00)*** (6.06)*** Number of Obser Likelihood Rtio (0) (0) 222.3(0) 47.48(0) 25.5(0) 71.68(0) Smple restriction 7 ( τ i 2% ) ( τ i 2% ) ( τ i 2% ) ( τ i 5% ) ( τ i 5% )( τ i 5% ) 1 0 rvc 8 2.0% 2.5% 8.2% 5.2% 1.7% 7.9 i = rvc i 10 u i Men mrgin Men Utiliztion Rte Men vlue content Simulted utiliztion rtes Assumed vlue for rvc i Men MFN triff (Chin) Predicted utliliztion rte Men MFN triff (Philippines) Predicted utliliztion rte Source: Authors Computtions. See text. The reported likelihood rtio follows chi-squred distribution with two degrees of freedom (χ 2 (2)). The p-vlue of this sttistic is reported in brckets. Becuse of the very different pttern of trde between the EU nd its GSP nd ACP prtners nd the ruling ptterns of trde in AFTA, it is difficult to conclude whether these ex-nte simulted utiliztion rte responses to chnges in VC nd in preferentil mrgins could be used s guides to the likely effects under AFTA. Furthermore, the estimted coefficients vlues re sensitive to the choice of