UNITE MIXTE DE RECHERCHE EN ECONOMIE PUBLIQUE JOINT RESEARCH UNIT IN PUBLIC ECONOMICS. Agricultural trade restrictiveness in the European Union taxes

Transcription

1 UNITE MIXTE DE RECHERCHE EN ECONOMIE PUBLIQUE JOINT RESEARCH UNIT IN PUBLIC ECONOMICS 2004 CAHIERS DE RECHERCHE WORKING PAPERS 2004/03 Agrcultural trade restrctveness n the European Unon taxes and the Unted States J.C. Bureau, L. Salvatc UMR Econome Publque Avenue Lucen Brétgnères Grgnon 16 rue Cl. Bernard Pars Tel. +33 (0) Fax. +33 (0)

2 Agrcultural trade restrctveness n the European Unon and the Unted States Jean-Chrstophe Bureau a, Luca Salvatc b a Insttute for Internatonal Integraton Studes, The Sutherland Center, Trnty College, Dubln 2, Ireland, and Insttut Natonal Agronomque, 16 rue Claude Bernard, Pars cedex 05, France. b Unverstà degl Stud del Molse, Dpartmento d Scenze Economche Gestonal e Socal, Va De Sancts Campobasso, Italy. Correspondng author : Luca Salvatc ADDRESSES: Luca Salvatc Unverstà degl Stud del Molse Dpartmento d Scenze Economche Gestonal e Socal, Va De Sancts Campobasso, Italy Tel: / Fax: e-mal: luca.salvatc@unmol.t Jean-Chrstophe Bureau Insttute for Internatonal Integraton Studes, The Sutherland Center, Trnty College Dubln, Dubln 2, Ireland. e-mal: bureau@tcd.e or bureau@grgnon.nra.fr

3 Abstract. The paper provdes a summary measure of the Uruguay Round tarff reducton commtments n the European Unon and the Unted States, usng the Mercantlstc Trade Restrctveness Index (MTRI) as the tarff aggregator. We compute the ndex for agrcultural commodty aggregates assumng a specfc (Constant Elastcty of Substtuton) functonal form for mport demand. The levels of the MTRI under the actual commtments of the Uruguay Round are computed and compared wth two hypothetcal cases, a deeper cut n hgher tarffs and a unform reducton of each tarff, both leadng to the same average reducton as n the Uruguay Round. Ths makes t possble to nfer how reducng tarff dsperson wll help mprove market access n future trade agreements, and provdes some gudelnes for aggregatng detaled tarffs n trade models. JEL codes: F13, Q17 Keywords: nternatonal agrcultural trade; protecton, tarffs and tarff factors. 2

4 1. Introducton One maor achevement of the Uruguay Round Agreement on Agrculture (URAA) was the prohbton of quanttatve barrers to agrcultural trade, requrng that all such trade take place under a tarff-only regme (except for some specfc derogatons ncludng tarff quotas). Each World Trade Organzaton (WTO) member establshed a base schedule, contanng both preexstng and new tarffs resultng from the converson of non-tarff measures, followng an nternatonal commodty classfcaton scheme (referred to as the Harmonsed System or HS). The adopton of a tarffs-only approach for agrculture was a sweepng reform that went a long way toward subectng agrcultural trade to the same dscplnes appled to other traded goods. However, many authors have ponted out that the URAA agreement acheved only mnor reductons n protecton (Tangermann, 1995). One reason for ths concluson s the rather lax method of converson of non-tarff measures nto ther tarff equvalents. It has also often been ponted out that member countres were allowed sgnfcant flexblty n the allocaton of tarff cuts. For nstance, the tarff cuttng formula was based on a smple average. Thus, by makng rather large percentage cuts n low tarffs, or n tarffs for commodtes that do not compete wth domestc producton, countres could meet the overall 36% average obectve wth only mnmal cuts n poltcally senstve tarffs. The present negotatons under the so-called Doha Development Agenda rase concerns that smlar dluton of the commtments may occur. For that reason, a number of countres have proposed measures to ensure substantal mprovements n market access, usng for example a Swss formula, under whch the level of tarff cut s a functon of the level of the ntal tarff. The July Councl Decson of the WTO states that progressvty n tarff reducton wll be acheved through deeper cuts n hgher tarffs wth flexbltes for senstve products (WTO, 2004). Even though the practcal modaltes of these cuts are stll a matter of negotaton, t s foreseen that a system of bands wth dfferent thresholds wll be used. Smulatons show that such a system results n a tarff structure that s very close to the one obtaned by the Swss Formula. There s some uncertanty, though, about the actual effect of such harmonzng formulas on market openness, compared to commtments based on a radal cut (.e. all tarff lnes cut by the same percentage), or on an average cut as mplemented n the URAA. Ths s one of the ssues addressed n ths paper. All studes on market access run nto some maor dffcultes lnked to data avalablty and nternatonal nconsstences n classfcatons. These emprcal aspects are perhaps the man reason why the varous studes dffer so much when measurng the degree of market access n one gven country. 1 However, methodologcal ssues are also mportant. To assess the overall effect of an uneven reducton n a large number of tarffs, one faces the problem of fndng the approprate ndex. Recent developments n the theory of ndex numbers have led to new ndcators of the aggregate mpact of trade polcy, such as the Trade Restrctveness Index (TRI) and the Mercantlstc Trade Restrctveness Index (MTRI). The MTRI, ntroduced by Anderson and Neary (2003), conssts n estmatng the unform tarff that yelds the same aggregate volume of mports as the orgnal vector of (non-unform) tarffs across a number of mports. We beleve 1 For example, estmates of the EU average agrcultural tarff for agrculture after the Uruguay Round range between less than 9.7% (Gallezot, 2002) and 40% (Messerln, 2001). 3

5 that usng the trade volume as the reference standard s approprate n the context of trade negotatons, snce countres nvolved n the negotaton are nterested n the trade volume dsplacement due to changes n tarffs. Indeed, one of the pllars of the WTO s the prncple of recprocty that can be nterpreted as equvalent mport volume expanson (see Bagwell and Stager 2000). Our contrbuton s the followng: - Frst, assumng a specfc functonal form for the mport demand, we address the problem of assessng the tarff reducton commtments undertaken by the European Unon (EU) and the Unted States (US) under the URAA. In order to do so, we compute the MTRI n 1995 (the frst year of the mplementaton of the URAA) and n 2001 (the end of the mplementaton perod) for 20 agrcultural commodty aggregates. - Second, we compare the effect of the URAA on tarffs wth an alternatve tarff reducton scheme whereby hgh tarffs are cut more dramatcally than low tarffs. - Thrd, we measure the magntude of the dluton effect that could have resulted from the dstrbuton of large and mnmal cuts across tarff lnes. Ths s done by comparng the URAA tarff cuts wth a unform (.e., radal) reducton n tarffs. - Fourth, we compare an ndex based on economc theory, such as the MTRI, to other a- theoretc, ad hoc ndexes of tarff reductons, such as the smple arthmetc average of tarff cuts adopted n the URAA. Our contenton s that much of the emprcal evdence based on these ndexes s nherently flawed. - Fnally, we compute the level of MTRI and not smply the relatve rates of change between two ponts n tme as n Bureau, Fulpon and Salvatc (2000), hereafter BFS. Our results provde a measure of the overall level of protecton across countres, both before and after the URAA. Computng MTRI levels usng a Computable General Equlbrum (CGE) model does not allow the degree of detal suffcent to work wth the actual commtments on agrcultural tarffs. Here, we construct an approxmaton of the MTRI that makes t possble to handle the present EU and US tarff structure, e.g. some 1500 tarff lnes n agrculture. Our results provde ndcators of the degree of market access for 20 aggregated agrcultural products, whch are consstent wth a data set wdely used by trade practtoners, the GTAP dataset (Hertel, 1997). 2. Methodology The practcal and theoretcal defcences of tradtonal tarff ndexes, such as the smple or the trade-weghted average tarff, are well known (Lard and Yeats, 1988; Anderson and Neary, 2003). Indexes such as the TRI and the MTRI have more sold theoretcal foundatons, although the defnton of such ndexes reles on several restrctve assumptons, ncludng the exstence of a compettve equlbrum, a sngle representatve consumer, and fxed world prces (.e., the small 4

6 country assumpton). Because they are derved from the balance of trade functon, the TRI and the MTRI synthesze the overall effect of trade polcy on the economy. 2 The assumpton of fxed (exogenous) world prces s questonable, snce our emprcal analyss deals wth US and EU, two maor traders on the world agrcultural market. However, the small country assumpton helps to guarantee the exstence and unqueness of the ndexes, rulng out counterntutve second best results and s consstent wth a ceters parbus approach (Bureau and Salvatc, 2004). 3 The most theoretcally consstent soluton would be to compute the MTRI as the (scalar) tarff that would yeld the same volume of mports as the ntal tarff structure usng a CGE model (Anderson and Neary, 2003). However, the lmtatons n the number of commodtes n CGE models requre a substantal aggregaton of trade flows and tarffs. 4 Agrcultural tarffs vary wdely even wthn a sngle product aggregate (e.g., wthn a sngle chapter of the HS classfcaton). In addton, tarff reductons under the URAA were taken on the bass of a very detaled lst of tems, and the magntude of tarff cut also vares substantally wthn a product category (Gbson et al. 2001). Therefore, a sgnfcant amount of nformaton on the level of tarff dsperson (and on the change n dsperson over tme) s lost when aggregatng tarffs data up to the level that s consstent wth CGE models aggregates. In order to be able to take nto account the mpacts of changes occurrng on a very large number of fnely dfferentated tarff lnes, we buld on the nsghts of Bach and Martn (2001) who assume a specfc functonal for mport demand. Ther methodology, whch ams to develop tarff aggregators for both the expendture and tarff revenue components of CGE models, can be adapted n order to compute the MTRI Mercantlstc Trade Restrctveness Index Our startng pont s the trade behavour of an economy under perfect competton. When tarffs are mposed, government behavour n collectng tarff revenues and redstrbutng them n lumpsum fashon needs to be ncorporated n the representaton of the economy. Both government and prvate behavours are summarsed by the balance of trade (BoT) functon B (p, u, z). The BoT 2 The balance of trade functon summarzes the outcome of and the consumpton sector, the producton sector and the publc behavour. Equlbrum of the economy s consstent wth a balance of trade that equals an exogenous ncome. Anderson and Neary (1996) and Martn (1997) provde detaled nsghts on the use of the balance-of-trade functon. 3 Anderson and Neary (2003), argue (footnote 8) that there s a ratonale for a ceters parbus trade restrctveness ndex that fxes world prces even when these prces are n fact endogenous. Such a ratonale may be represented by the fact that, by keepng world prces constant, we focus on the component of protecton explaned by natonal polces, and not by the degree of market power of the country. 4 Anderson and Neary (1999) use Anderson s (1998) CGE model whch s unusually dsaggregated as far as the trade structure s concerned. However, even ths model reles on a 4-dgt HS classfcaton, whle the offcal WTO tarff commtments of the EU and the US n the food and agrcultural sector specfy tarffs at the 8-dgt level. 5

7 represents the external budget constrant, and s equal to the net transfer requred to reach a gven level of aggregate domestc welfare u, for a gven set of domestc prces p, and factor endowments vector z. It summarses the three possble sources of funds for procurng mports: earnngs from exports, earnngs from tarff revenues and nternatonal transfers (Anderson and Neary, 2003). The MTRI reles on the dea of evaluatng trade polcy usng trade volume as the reference standard. The MTRI s defned n terms of the unform tarff τ that yelds the same volume (at world prces) of tarff-restrcted mports as the ntal vector of (non-unform) tarffs. Ths can be expressed wth mport demand functons M, whle holdng constant the balance of trade functon at level B 0 : 0 0 [( 1+τ ) p*, B ] τ : M = M (1) where p * denotes the nternatonal prce vector of the N goods k = (1,,N) and M 0 s the value of aggregate mports (at world prces) n the reference perod. Defne the scalar mport demand as M N * m, (2) ( p p*, B) k = 1 p k I k where Im denotes the uncompensated (Marshallan) mport demand functon and p s the domestc prce vector. Accordngly, the MTRI unform tarff τ would lead to the same volume of mports (at world prces) as the one resultng from the uneven tarff structure, denoted by the N- dmensonal tarff vector t whose elements are t k. That s, the MTRI can be computed by solvng equaton (3) for τ. 5 N k= 1 p * k I m k N 0 * m [ p ( 1+ ), B ] = p I [ p *( 1+ t) 0 * τ, B ]. (3) k = Emprcal estmaton of the MTRI k k Havng defned the MTRI, for the emprcal mplementaton we follow Bach and Martn (2001) modellng demand through a Constant Elastcty of Substtuton (CES) functonal form. Ths functon mposes well-known restrctve assumptons on separablty. Nonetheless t has several emprcal advantages that explan ts use n modellng mport demand (Wnters, 1984). If the utlty functon s homogeneously separable, commodtes may be consstently aggregated (Gorman, 1959). That s, one may form composte commodtes whch may be treated n the same manner as the prmary commodtes. Accordngly, we assume that the overall basket of goods can be parttoned nto J aggregates denoted =1, J, and the utlty functon of the representatve consumer can be wrtten as: = φ u ( x ) u J ( x )), (4) ( 1 1 J U,..., 5 The MTRI derved from equaton (3) provdes a measure of trade restrctveness relatve to a free trade reference, whle BFS computed a unform tarff surcharge measurng changes n the tarff structure from the ntal equlbrum to the new (stll dstorted) equlbrum. 6

8 where φ s contnuous, twce dfferentable, and strctly quas concave and the u are contnuous, twce dfferentable functons, homogeneous of degree one (Lloyd, 1975). When focusng on the J sectoral MTRIs, a convenent (albet restrctve) assumpton s to assume φ to be a Cobb- Douglas functon (mplyng that the expendture functon s also a Cobb-Douglas one n prces wth utlty enterng multplcatvely). In such a case, we avod the ssue of allocaton of consumer expendture across sectors, whch n general equlbrum models, s affected by tarffs wthn a partcular aggregate (Bureau and Salvatc, 2004). In our applcaton, we assume that u s a CES functon n x. Snce the mport volume functon s homogenous of degree zero n the prces of traded goods, the MTRI cannot be calculated (any unform tax would be equvalent to free trade n terms of mports). 6 Ths dffculty of evaluatng the MTRI can be crcumvented f ) there s a desgnated "reference good", so that the prce vectors refer to prces relatve to such a good; or f ) we use the prce of the least dstorted mported good n each sector as the numérare, avodng the need to nclude the domestc good n the subexpendture functon. Snce there are some sectors (such as dary, sugar, beef n the EU, for example) n whch all products face a strctly postve tarffs, usng the least dstorted good n each sector as the reference would not allow to draw meanngful comparsons across sectors and/or countres. We use the popular Armngton (1969) assumpton that mports are mperfect substtutes of domestc goods, and we solve the problem by takng the domestc good as the numérare (Bach and Martn, 2001). 7 We partton the consumpton vector x wthn the th group nto an aggregated domestc good denoted wth a suffx d and N -1 traded goods denoted wth an ndex. 1 ρ ρ ρ β d ( xd ) + β ( x ). (5) u = = 1,..., N Denotng σ 1 = 1 ρ each aggregate s the elastcty of substtuton wthn the group, the expendture devoted to 1 1 σ σ σ 1 1 e ( p, u) = β d ( pd ) + β ( p ) u. (6) The parameters can be calbrated to the ntal values of the expendture shares n the base β data, when all domestc prces are set to 1. After dervng the ndrect utlty functon by nvertng 6 More generally, Neary (1998) shows how the falure to select a reference untaxed good leads to msleadng results n the theory of trade polcy. 7 The assumpton that the domestc good s numérare does not mply that t s exogenous. However, endogenety would requre specfyng market clearng to allow prce determnaton. Here, our goal s to develop a methodology allowng the computaton of tarff aggregators wthout usng a CGE model. 7

9 equaton (6), the Marshallan demand functons of each of the =1,.., N -1 mported goods can be found by Roy s dentty: x = β β d p σ 1 σ ( pd ) + β ( p ) 1 σ e. (7) Denotng P the prce ndex that corresponds to the denomnator of the rght-hand sde, the mport volume functon for the th aggregate, valued at world prces, s N = 1 p * x = p * 1 β P. p σ e wth =1,, N 1. (8) 0 When the ntal total expendture e (expendtures on both domestc and mports n ) s used n expresson (8), we obtan the demand functon at the ntal level of mports. The MTRI unform tarff equvalent τ for each aggregate s found by settng the value of the mport volume functon wth the unform tarff equvalent equal to the ntal value of mports (evaluated at world prces), σ τ P * 0 * 0 p β e = ( ) p I, (9) * p 1+ τ where 0 I applcatons), and P τ are the volume of mports n the ntal perod (.e., 1995 or 2001 n our numercal τ P s the prce ndex: σ 1 σ * 1 σ d ( pd ) + β ( p ( 1+ τ ). (10) = β The unform tarff equvalents for each aggregate commodty are found usng an optmzaton routne n the GAMS package (Brooke et al. 1998), solvng for τ n equatons (9) and (10). The ndcators τ are by themselves relevant for the analyss of trade polcy. In addton, the can be used as aggregate tarffs n any trade model wth a commodty aggregaton and an mport demand structure whch s consstent wth our assumptons. However, t must be acknowledged that they are only an approxmaton of the true (.e., general equlbrum) MTRI tarff 0 equvalent, snce usng ntal total expendture n equaton (9) we gnore the ncome effect due e to the change n tarff revenue. In our applcaton, dealng wth products that are charactersed by 8 low-ncome elastctes n developed countres, we do not expect ths to be a sgnfcant ssue. τ 8 Beghn, Bureau and Park (2003) ntroduce the full expanson effects consstent wth general equlbrum n ther sectoral MTRI, but the mpact on ther emprcal results s very lmted. 8

10 One may expect that the computaton of an aggregate (.e., for the whole agrcultural sector) MTRI tarff equvalent could be easly performed ntroducng an upper-level demand system. However, the requrement of a reference untaxed good for the computaton of the MTRI tarff aggregator makes the computaton of the same ndex at dfferent levels of aggregaton a trcky ssue. As a matter of fact, f the numérare s a domestc good, the prce (and quantty) ndex to be used at the upper level would nclude both domestc and mported goods, and ths would make the computaton of an upper-level tarff aggregator meanngless. As a consequence, n order to compute an MTRI tarff equvalent for the entre dataset, we defne t as the unform tarffτ that would keep the overall (.e., on all =1,..,J sectors) mport volume equal to the ntal value. Ths can be obtaned by modfyng equaton (9) as follows: p * β d ( p β 1 σ d ) + * p σ * 1 σ β ( ( )) p 1+ τ 0 e = ( ) 1+ τ σ p * I 0. (11) 2.3. Dataset The volumes of mports are taken drectly from the respectve US and EU datasets (US Internatonal Trade Commsson and Eurostat s Comext data). The Schedule XX that the EU and the US submtted to the WTO provdes the base and bound tarffs at the 8-dgt level of the HS classfcaton. The URAA schedule therefore provdes nformaton on tarffs n 1995 (that s, after the Uruguay Round tarffcaton process) and n 2001 and onwards (that s, after the mplementaton of the mandatory 36 percent average reducton n tarffs). The domestc prces are constructed by multplyng the world prce p* by the ad valorem tarff structure (ntal, fnal, or counterfactual tarffs) that we are nterested n. As a result, the measure of market access focuses only on changes n the tarffs ceters parbus, and s not affected by exogenous prce varatons (see BFS, 2000 for detals on the data set). In ths paper, the focus of the analyss s on the tarff reducton commtments. We gnore tarff rate quotas. The EU tarff schedule ncludes 1,764 tarff lnes, whle the US schedule ncludes 1,377 tarff lnes (excludng n quota tarffs). Both the EU and the US apply ther bound tarffs on products traded n a Most Favored Naton (MFN) framework. That s, usng the URAA schedules gves a good mage of the actual tarff structure, although lower tarffs are appled n the framework of preferental agreements that we dd not consder here. For purposes of calculaton, we converted specfc tarffs nto ad valorem equvalents, followng the same conventons as n BFS (2000). The elastctes of substtuton σ that match the lst of aggregates are taken from the GTAP dataset (Dmaranan and McDougall, 2002). Ths comprehensve dataset s wdely used n appled analyss, and researchers mght be nterested n tarff aggregates that match the GTAP classfcaton for smulaton purposes. Moreover, the converson tables from detaled tarff structures to the GTAP sectors are fully avalable, whch makes t possble to aggregate the very detaled lst of tarffs of the URAA Schedule nto a restrcted number of products that correspond to the GTAP system of classfcaton. Fnally, the dataset provdes the nformaton that s necessary for dstngushng between expendtures on domestc products and mports. There s 9

11 however lttle ustfcaton for usng the GTAP elastctes. It actually s qute bothersome that these elastctes are the same for the two countres (Table 1). However, provdng new estmates s certanly out of the scope of ths work. We undertook senstvty tests to examne the effects of dfferent elastcty values on the measurement of MTRI unform tarffs: the results are presented n Secton 6. The orgnal GTAP data set dstngushes J=20 agrcultural and food aggregate products. In order to nclude non-food commodtes lsted n the URAA schedules (manly agrcultural goods lsted n chapters 29 to 53 of the HS classfcaton) we defned an extra aggregate. We gnore one GTAP sector (raw mlk) because there s no trade for the correspondng commodty. Overall, we aggregated 1,764 tarff lnes n the EU (1,377 tarff lnes n the US) at the 8-dgt level of the HS classfcaton up to 20 aggregate products descrbed n Table 1. It s noteworthy that the number of tarff lnes n each commodty aggregate s very uneven. Table 1 shows, for example that there are only three tarff lnes n the aggregate paddy rce, whle the aggregate fruts and vegetable tarff ncludes 183 tarff lnes n the EU schedule. 3. Measures of market access pror to the Uruguay Round Agreement The computaton of the MTRI unform tarff equvalent provdes an estmate of the trade restrctveness of the actual tarff structure. It s calculated for the year 1995 for both the EU and the US, makng t possble to compare the trade effect of the tarff structures pror to mplementaton of the Uruguay Round commtments. The structure of bound tarffs n the EU and the US dffers n several aspects (Table 2). The average non-weghted base tarff was 9.7 percent (12.7 percent f we focus only on the tems wth a nonzero tarff) n the US, whle n the EU the average tarffs were 26.7 percent (31.4 percent, respectvely). In most sectors, the EU average tarff s larger than the US average tarff, the gap beng partcularly wde n the grans, meat, sugar, and rce sectors. In the EU, the trade-weghted average tarff s usually larger than the non-weghted average, whle t s generally the opposte n the US. A trade-weghted average tarff that s smaller than the non-weghted one can result from prohbtve tarffs or may smply mean that larger tarffs are set on commodtes whose demand s partcularly elastc. Here, there s evdence that very hgh tarffs are set n a few sectors where the government s wllng to protect domestc producers from mports, as n the US groundnut or dary sectors. On the other hand, the trade-weghted average s larger than the non-weghted average tarff when low tarffs are set on products whose demand s structurally lmted, ether because these are nche market products (e.g., processed products, pecular types of fruts, beverages, and condments n the EU), or because local producers are compettve (e.g., pg meat and poultry meat). Ths may also mean that hgher tarffs are set on goods wth a relatvely nelastc demand for mports. τ 4. Comparson between the MTRI and a-theoretc ndcators Table 2 shows sgnfcant dfferences between the MTRI and the non-weghted tarff average. Ths s not surprsng, snce the non-weghted tarff average bears lttle relatonshp wth theoretcally sound ndexes lke the MTRI or the TRI. On the other hand, the values for the trade- 10

12 weghted average tarffs are often qute close to those gven by the MTRI tarff. Ths emprcal fndng converges wth those of Anderson and Neary (2003) and Bach and Martn (2001) who show that the trade-weghted average tarff s a lnear approxmaton to the tarff aggregator based on the expendture functon. In other terms, the trade-weghted average tarff plays the same role as the Laspeyres prce ndex n consumer theory, provdng a fxed-weght approxmaton that underestmates the true heght of tarffs because t neglects substtuton nduced by tarff changes. In the partcular case of a CES aggregator functon, the trade-weghted average tarff corresponds to constant expendture shares. Constant shares correspond to the specal case of a Cobb-Douglas subutlty functon, where =1. In such a case, we observe the followng result (proof n the Appendx). σ PROPOSITION 1. In the base equlbrum (that s, wth all domestc prces equal to 1), the MTRI unform tarff concdes wth the trade-weghted average tarff when the CES aggregator functon becomes Cobb-Douglas. Ths proposton clarfes the lnkage between our MTRI estmates, usng a CES aggregator functon, and the trade-weghted ndex. Snce the values of the n the GTAP data set rank between 2.2. and 3.8, t s not surprsng that the MTRI unform tarffs for can be rather close to the trade-weghted average tarffs. The MTRI unform tarff s more lkely to be hgher than the trade-weghted average tarff the more elastc s the demand for tarff-constraned mports. On the bass of emprcal calculatons wth a CGE model, Anderson and Neary (2003) confrm ths basc nsght. 9 Our emprcal estmate of the MTRI leads to smlar conclusons. In our specfc case of a CES aggregator functon, we can derve the condtons under whch the MTRI exceeds the trade-weghted ndex: PROPOSITION 2. In the base equlbrum (that s, wth all domestc prces equal to 1), () the tradeweghted average tarff overestmates the MTRI unform tarff when σ <1 (σ denotes, the elastcty of substtuton of the CES aggregator functon); () the trade-weghted average tarff underestmates the MTRI unform tarff when σ > 1. Lookng at Tables 1 and 2, t s also obvous that the MTRI and the trade-weghted ndex gve very smlar results when the number of tarff lnes n the aggregate s very small, or when there s lttle dsperson n tarffs wthn an aggregate. Fgures n Table 2 show that the percentage varaton between the MTRI and the trade-weghted average depends postvely on the standard error of tarffs, somethng that s confrmed by elementary descrptve statstcs. For the aggregates wth a large number of products, the gap between the two ndexes can be very large. Gven our assumpton on substtuton elastctes, n the dary sector, for example, the trade- σ 9 Anderson and Neary (2003) proved the followng: The MTRI unform tarff exceeds the trade-weghted average tarff f: () the compensated arc elastcty of demand for the composte tarffed good exceeds one; () the composte tarffed good s normal; and () the trade expendture functon s mplctly separable n tarffed and other goods. 11

13 weghted average underestmates the trade restrctveness of the pre-uraa tarff structure by 29 percent n the US and by 9 percent n the EU. Ths s also the case n the cattle sector and n the beverages sector n the EU (underestmaton of 29 percent and 23 percent respectvely), and n the olseeds sector n the US (underestmaton of 40 percent). Overall, for sx aggregate EU products out of twenty, the trade-weghted average underestmates the MTRI by more than 10 percent. In bref, the trade-weghted tarff can only be a satsfactory approxmaton of more theoretcally consstent ndcators of market access under very specfc condtons and for specfc values of the substtuton elastctes. In more general cases, when the aggregate ncludes a large number of heterogeneous tarff lnes and elastctes dffer from unty, the trade-weghted average s a poorer ndcator of the restrctveness of the tarff structure. 5. Impact of the Uruguay Round and counterfactual scenaros The computaton of the MTRI for the year 2001 makes t possble to evaluate the trade restrctveness of the tarff structure that results from the URAA. Followng BFS, we also want to assess the relatve effects of reducng the tarff average and tarff dsperson. We smulated two other tarff reducton schemes n addton to the actual reducton mplemented by the EU and the US. The three cases are called Uruguay Round commtments, Swss Formula, and unform tarff reducton, respectvely. In the three cases, we start from the same tarff structure n 1995 (that s, 95 the ntal vector p s the same for each case), but the three schemes lead to three dfferent vectors for the year These may be summarzed as follows: 2000 Uruguay Round commtments. The prce vector p s the one that results from the bound tarffs n year The resultng tarff structure reflects the oblgaton of a 36 percent nonweghted average reducton, but wth no constrants placed on the mx of reductons to acheve the overall average (except that each tarff lne must be reduced by at least 15 percent) Swss Formula. In ths case, we calculate the prce vector p that would have resulted from t 2001 a harmonzng tarff reducton (hgher tarffs subect to larger cuts, as t should be the case n the present round accordng to the July 2004 compromse). The formula s gven n equaton (15) and the parameter C s chosen to obtan the same non-weghted average reducton of 36 percent n tarffs as specfed n the URAA. Comparng the value of the MTRI-unform tarff equvalents wth those that actually result from the URAA, we can assess the mpact of commtments that would have focused more on reducng tarff dsperson than the actual URAA tarff cuts = Ct /( C + t ). (15) Unform (.e. radal) tarff reducton. Under ths scheme, we assume that a unform 36 percent reducton s appled to all tarff lnes. Ths wll obvously result n the same average reducton as specfed under the URAA, but t does not permt countres to allocate the adustment across commodtes. The comparson of the values of the MTRI-unform tarff equvalents wth those that actually result from URAA commtments therefore measures the magntude of the dluton effect that resulted from the dstrbuton of large and small or mnmal cuts across tarff lnes. 12

14 Comparng the values of the MTRI-unform tarff equvalents of the 2001 tarffs (frst column n Table 3) wth the MTRI-unform tarff equvalents of the 1995 tarffs (thrd column n Table 2), we can assess the actual mpact of the URAA n terms of market access. The URAA ndeed reduced each of the 20 MTRI-unform tarffs both n the EU and n the US. Because both the varance and the mean of tarffs decrease (see Tables 4 and 5), t s not surprsng that the MTRI unform tarff also moves n the same drecton for all aggregates, as well as at the aggregate level, confrmng that the URAA ncreased market access. Ths s a consequence of the commtment to reduce each tarff lne by at least 15 percent. The absolute values of the reductons are much smaller n the case of the US, as could have been expected gven the low values of the MTRIunform tarff equvalents n the base perod (see Table 2). Ths s also consstent wth the BFS results suggestng that the Uruguay Round led to a larger ncrease n market access n the EU than n the US. We now turn to the counterfactual scenaros n Table 3. If the Swss Formula had been appled, the Uruguay Round would have led to a consderable ncrease n the EU market access as measured by the overall MTRI. In the EU, the Swss Formula would have led to a dramatc decrease n trade restrctons n hghly protected sectors such as grans, meat, and dary, as well as n sectors characterzed by a hgh tarff dsperson, such as fruts and vegetables. The US market also would have been more open at the aggregate level (see Table 4), but n several cases the Swss Formula does not provde a sgnfcant mprovement n market access compared to the other schemes (unlke n the EU). Clearly, the mpact of choosng such or such tarff reducton formula s larger n the EU than n the US. Tables 3, 4, and 5 show that the URAA ncreased access to the market n a way that s very comparable to what would have resulted from a unform tarff reducton n most sectors. Ths means that both countres have not allocated tarff cuts n a very strategc way. The dluton of the tarff reducton effect was partcularly lmted n the EU, as could have been expected snce most tarffs were cut by 36 percent and no tarff was reduced by less than 20 percent. 6. Comparson wth prevous results and senstvty The comparson of the MTRI-unform tarff equvalents between the URAA commtments and the counterfactual scenaros confrms BFS s conclusons that the dluton of tarff cuts has had overall a lmted mpact on market access n the EU. It also confrms that harmonzng formulas would have resulted n much larger market access than that whch occurs wth the Uruguay Round dscplne, especally n the EU. 10 In order to check the consstency of the numercal results wth those of BFS, we need to compute, for the entre dataset, the unform tarff surcharge (.e., the extra rate to be appled to the non-unform tarff vector of tarffs n perod 1), whch compensates the non-unform change n the tarff structure (see Secton 2.1). In practce, the overall MTRI unform tarff factor surcharge s obtaned by solvng for µ n equaton (12): 10 Remember that the tarff reducton procedure agreed upon on July , and whose modaltes are presently under negotaton, reles on deeper cuts n hgher tarffs. The effects are, n practce, very smlar to the Swss Formula. 13

15 p * β d ( p β 1 σ d ) + 1 p β 1 1 ( p ( 1+ µ )) σ σ 0 e = ( ) 1+ µ σ p * I 0. (12) Gven the dfferences n the methodologcal approaches followed here and n BFS (2000), the results presented n Table 6 are surprsngly smlar. Only n the case of the Swss Formula s the dfference substantal. Ths s the scenaro that mples the largest change n tarffs; n such a case, then, the hgher substtutablty mpled by the CES functonal form leads to a hgher mpact. Fnally, we turn to senstvty analyss of smulaton results, n order to check to what extent the value of the substtuton elastctes affect the MTRI computaton. As t was mentoned n Secton 2.3, even though the elastctes extracted from the GTAP dataset are wdely used by appled analysts around the world, ther relevance s questonable. There are several reasons to beleve that the GTAP elastctes are low, compared to what s consstent wth recent econometrc estmates of mport elastctes (see e.g. Hummels, 1999, Erkel-Rousse and Mrza, 2002). In order to assess the senstvty of the results to the choce of the parameters of the CES functon we computed the overall MTRI-unform tarff equvalents, makng dfferent assumptons about the values of the substtuton elastctes (Table 7). The elastctes are assumed to range from onethrd to three tmes the orgnal values. Even though the rankng among dfferent scenaros remans the same for the varous assumptons, the MTRI s obvously qute senstve to the degree of substtuton between products, a result consstent wth Proposton 2. Snce the large values of the ndex are more senstve to the assumpton on substtuton, the results are more affected by changes n the s n the EU than n the US, where the agrcultural sector s less protected. σ 7. Concluson The results of our comparson of tarff ndexes and tarff reducton scenaros should be used wth cauton n polcy analyss. Indeed, our fgures for tarffs after the Uruguay Round do not correspond to the actual EU and US protecton. The reason s that, for the purpose of comparson between scenaros, the world prce was kept the same as n the ntal (1995) stuaton. In addton, the actual protecton of EU and US agrculture s clearly overestmated because we focused on the MFN tarffs. That s, we gnore preferental tarffs, whch affect roughly one thrd of the value of EU mports. However, by computng sectoral ndexes, our approach provdes a more detaled assessment of the market access mprovement n the EU and the US than prevous studes (e.g. BFS). Because we manage to approxmate the MTRI unform tarff wthout usng a CGE model, we are able to take nto account the large number of dfferent tarffs that characterze the agrcultural sector n most WTO countres. Our computaton of the absolute level of the MTRI shows that access to the EU market s stll far more restrcted than to the US market, at least for countres that do not beneft from preferental treatment. On a non-weghted bass, the overall average tarff on agrcultural and food products was 26.7 percent n the EU and 9.7 percent n the US n 1995, whle the trade-weghted average tarff was, respectvely, 25.5 percent and 3.3 percent. The MTRI-unform tarff measures a degree of trade restrctveness of 32.4 percent for the EU and 3.5 percent for the US (see Tables 3 and 4). 14

16 The reason why the MTRI gves a dfferent pcture s that the hgh tarffs n the US are set on a restrcted set of partcular goods. In contrast, n the EU, most of the commodtes mported n large quanttes face sgnfcant MFN tarffs. On the methodologcal sde, the MTRI unform tarff and the trade-weghted ndex tend to move closely together when the number of commodtes s small, and when the dsperson of tarffs s low. In other cases, the trade-weghted ndex underestmates the true mpact of the tarff structure on market access, as measured by the MTRI. When we aggregate a large number of tarffs, or when the dsperson s large, the two ndexes dffer sgnfcantly. The dfference that we observe between the MTRI unform tarff and the non-weghted tarff average suggests that models that rely on aggregate tarffs constructed as smple averages use poor estmates of the actual tarff structure. Ths bas s lkely to affect a large number of studes, as t s common practce to construct aggregate tarffs as smple averages of the detaled tarffs appled by custom offcers, who sometmes work at a level of detal correspondng to the 8, 10 dgt level (or even 14-dgt n the case of the EU). Constructng the aggregate tarffs used n trade models as trade-weghted averages s more satsfactory. However, when aggregatng a large number of goods wth a large tarff dsperson nto a sngle commodty, ths method also results n sgnfcant bas, usually an underestmaton of the aggregate tarff, as measured by the MTRI. The computaton of the absolute levels of the MTRI ndex makes t possble to compare the strateges n the allocaton of tarff reductons takng nto account the dfference n the ntal (bound) tarffs of the EU and the US. We were also able to assess the consequences of emphaszng reductons n tarff dsperson n terms of gettng a (more) level playng feld between the EU and the US. Overall, our results confrm the ntuton by BFS: ) the dluton of tarff cuts n the URAA had a lmted mpact on overall market access; ) n the Doha round, a harmonzng formula would provde a sgnfcant mprovement n market access, whch would not be the case wth an average reducton. The July 2004 Decson states that hgher tarffs wll face larger cuts. Nevertheless, f the senstve sectors are excluded from ths dscplne, the market access mprovement wll be lmted, especally n the US, where hgh tarffs are concentrated n a few sectors. The behavoural parameters make a great dfference n our smulated outcomes. In the past, Shoven and Whalley (1984 p 1047) argued for the establshment of an elastcty bank n whch elastcty estmates would be archved, evaluated by groups of experts. Such a proposal could be more realstc now that we have examples of wdely accessble global databases, and that much progress has been made n the area of aggregaton and demand system estmaton: ths greatly reduces the dmensonalty of the demand systems that can be used n trade models. 11 In bref, ths paper provdes a summary measure of the Uruguay Round tarff reducton commtments n the EU and US, takng nto account the mpact of changes n a large number of 11 As ponted out by an anonymous referee, Lewbel (1996) developed smple tests for aggregaton based on tmeseres propertes of prce data, whle Capps and Love (2002) have used these tests and showed that relable demand systems can be estmated from the aggregates. 15

17 tarff lnes. The mpacts of alternatve tarff-cuttng procedures were evaluated usng the MTRI as the tarff aggregator. We were able to compute the ndex for partcular commodty aggregates wthout usng a CGE model, but we assumed a specfc functonal form for mport demand. Such an approach s easy to mplement: t requres only nformaton on tarffs, mport values, and total expendture on each commodty, n addton to the knowledge of the parameters of the demand functon). Ths comes at a cost, namely the need to specfy a tarff aggregator functon, whch also requres restrctve assumptons. Trade weghted ndexes underestmate protecton. But whle a CES based MTRI provdes an operatonal measure more theoretcally consstent, the results are nherently senstve to the assumpton on product substtuton, on whch there s stll lttle relable nformaton avalable. 16

18 Appendx Proofs of Propostons Proof of Proposton 1. Wth all domestc prces equal to 1 n the base equlbrum, equaton (9) becomes * p β β d ( p 1 σ d ) + β [ ] * 1 σ * ( p ( 1+ τ ) p ( 1+ τ ) 1 σ = p β. (A.1) * When =1, the MTRI unform tarff equvalent s σ 1 β d τ = 1. (A.2) * β p Recallng that n the base equlbrum p *, from the above equaton we obtan 1 = 1 + t βt τ =. (A.3) β Ths proves the proposton. Proof of Proposton 2. We frst wrte equaton (9) as follows (droppng the ndex for the sake of smplcty) p * 0 * 0 β e = (, ) = 0 *( 1+ ) p I F τ σ. (A.4) p τ P τ σ Assumng that there s a soluton F ( τ, σ ) = 0 and F σ ( τ, σ ) 0 Theorem n the neghbourhood of ( 0,σ τ ) there exsts a functon τ f ( σ ) dτ dσ F τ = f ( σ ), then by the Implct Functon = and τ =. (A.5) Fσ The dervatves of the F functon are F τ < 0, F σ > 0. Indeed, whch s negatve; whle σ F σ = e 0 * β p ( + ) ln τ s postve, snce 1 ( 1+ τ ) P τ σ P τ σ 1 0 τ e P F τ = σ β, 2 ( + ) * 1 τ ( ) p 1+ τ τ τ τ P 0 P ln P 0 ( 1 ) ( 1 ) ( 1 ) and σ > 0. Because the dervatves have opposte sgns, τ + τ + τ + τ ncreases wth σ (at least n the case where there exsts a soluton). Recallng from Proposton 1 17

19 that the MTRI and the trade-weghted average tarff concdes for σ = 1, the result follows mmedately. All ths s llustrated n Fgure 1, drawn n the space of unform tarff τ and elastcty of substtuton σ. The trade-weghted average tarff, whch corresponds to pont A, s drawn as a straght lne, as t does not change accordng to the vale of the elastcty of substtuton. We need to locate the ponts correspondng to the MTRI unform tarff. They le on the locus F 0, whch from the prevous results must be upward slopng. At pont B, whch corresponds to σ = 1, we know from Proposton 1 that τ must be equal to the trade-weghted average tarff. Apparently, when the elastcty of substtuton s lower than 1, the trade-weghted average tarff overestmates the MTRI unform tarff, whle the opposte s true for elastcty values greater than 1. Ths fndng s fully consstent wth our emprcal results, snce all the elastctes used n the calculaton exceed 1 (Table 1) whle the trade weghted average tarff never exceeds the MTRI unform tarff (Table 2). 18

20 Acknowledgements The authors thank Chrstan Bach who kndly provded useful samples of some of hs GAMS programs. They also thank two anonymous referees, as well as Alex Gohn and Lnda Fulpon for helpful comments, but the usual dsclamer apples. For ths research J.C. Bureau and L. Salvatc benefted from a grant of the Italan Mnstry of Unversty and Technologcal Research (Research Program of Natonal Scentfc Relevance on The WTO negotatons on agrculture and the reform of the Common Agrcultural Polcy of the European Unon ). J.C. Bureau dd part of the research at CARD, Iowa State Unversty and at the IIIS, Trnty College Dubln. Senorty s shared equally between authors. 19

21 References Anderson, J.E., and J.P. Neary A New Approach to Evaluatng Trade Polcy. Revew of Economc Studes. 63, Anderson, J.E., and J.P. Neary The Mercantlstc Index of Trade Polcy. Internatonal Economc Revew. 44(2), Armngton, P.A A Theory of Demand for Products Dstngushed by Place of Producton. Internatonal Monetary Fund Staff Papers, 16. Bach, C.F., and W. Martn Would the Rght Tarff Aggregator Please Stand Up? Journal of Polcy Modellng. 23(6), Bagwell, K., and R.W. Stager GATT-Thnk. NBER Workng Paper, W8005. Natonal Bureau of Economc Research, Cambrdge, MA. Beghn, J.C., J.C. Bureau. S.J Park Food Securty and Agrcultural Protecton n South Korea. Amercan Journal of Agrcultural Economcs, 85,3, Brooke, A., D. Kendrck, A. Meeraus, and R. Raman GAMS, A User s Gude. GAMS Development Corporaton, Washngton D.C. Bureau, J.C., L. Fulpon, and L. Salvatc Comparng EU and US Trade Lberalsaton under the Uruguay Round Agreement on Agrculture. European Revew of Agrcultural Economcs. 27(3), Bureau, J.C., and L. Salvatc WTO Negotatons on Market Access n Agrculture: a Comparson of Alternatve Tarff Cut Proposals for the EU and the US. Topcs n Economc Analyss & Polcy, Vol. 4. No. 1, Artcle 8, 2004 ( Capps, O. Jr., and H.A. Love Econometrc Consderatons n the Use of Electronc Scanner Data to Conduct Consumer Demand Analyss. Amercan Journal of Agrcultural Economcs. 84(3), Dmaranan, B., and R.A. McDougall Global Trade, Assstance, and Producton: The GTAP 5 Data Base. Center for Global Trade Analyss, Purdue Unversty. Erkel-Rousse, H., and D. Mrza Import Prce Elastctes Elastctes: Reconsderng the Evdence. Canadan Journal of Economcs. 35, Gallezot, J Accès au marché agrcole et agro-almentare de l'unon européenne: Le pont de vue du négocateur à l'omc et celu du douaner. Econome Rurale. 267, Gbson, P., J. Wano, D. Whtley, and M. Bohman Profles of Tarffs n Global Agrcultural Markets. USDA Agrcultural Economc Report 796, US Department of Agrculture, Economc Research Servce, Washngton D.C.. Gorman, W.M Separable Utlty and Aggregaton. Econometrca. 27, Hertel, T.W. (Ed.) Global Trade Analyss : Modelng and Applcatons. Cambrdge Unversty Press, Cambrdge, UK. Hummels, D Toward a Geography of Trade Costs. GTAP Workng Paper 17, Global Trade Analyss Proect, Purdue Unversty. Lard, S., and A. Yeats A Note on the Aggregaton Bas n Current Procedures for the Measurement of Trade Barrers. Bulletn of Economc Research. 40(2), Lewbel, A Aggregaton Wthout Separablty: A Generalzed Composte Commodty Theorem. Amercan Economc Revew. 86(3), Lloyd, P.J Substtuton Effects and Bases n Nontrue Prce Indces. The Amercan Economc Revew. 65(3),

22 Martn, W Measurng Welfare Changes wth Dstorsons. In Francos, J.F., and K.A. Renert (Eds.), Appled Methods for Trade Polcy Analyss: A Handbook. Cambrdge Unversty Press, Cambrdge UK, pp Messerln, P Measurng the Costs of Protecton n Europe: European Commercal Polcy n the2000s, Insttute for Internatonal Economcs, Washngton D.C.. Neary, J.P Ptfalls n the Theory of Internatonal Trade Polcy : Concertna Reforms of Tarffs and Subsdes to Hgh-Technology. Scandnavan Journal of Economcs. 100(1), Shoven, J.B., and J. Whalley Appled General-Equlbrum Models of Taxaton and Internatonal Trade: An Introducton and Survey. Journal of Economc Lterature. XXII, Tangermann, S Implementaton of the Uruguay Round Agreement on Agrculture by maor developed countres, UNCTAD, Geneva. Wnters, L.A Separablty and the Specfcaton of Foregn Trade Functons. Journal of Internatonal Economcs. 17: WTO Doha Work Programme. Draft General Councl Decson of 31 July WT/GC/W/535, World Trade Organzaton, Annex B. 21

23 Table 1. GTAP agrcultural commodtes and HS-8 tarff lnes Commodtes 1 GTAP Classfcaton Number of EU tarff lnes Number of US tarff lnes Elastctes of substtuton Paddy rce Wheat Cereal grans Vegetables, fruts, nuts Olseeds Sugar cane, sugar beet Plant based fbers Other crops Cattle, sheep, goats, horses Other anmal products Raw wool, cocoons and har Meat: cattle, sheep, goats, horses Other meat products Vegetable ols and fats Dary products Processed rce Sugar Other food products Beverages and tobacco Nonfood tems (goods lsted n URAA, beyond Chapter HS 24) other Note : Raw mlk (GTAP code 20) s excluded because of absence of trade. 2.2

24 23 Table 2. Base tarffs (year 1995, actual bound tarffs) Commodtes Non-weghted average tarff (%) Trade-weghted average tarff (%) MTRI tarff (%) Coeffcent of varaton of tarffs EU US EU US EU US EU US Paddy rce Wheat Cereal grans Vegetables, nuts fruts, Olseeds Sugar cane, sugar beet Plant based fbers Other crops Cattle, sheep, goats, horses Other products anmal Raw wool, cocoons, har Meat: cattle, sheep, goats, horses Other meat products Vegetable ols and fats Dary products Processed rce Sugar Other food products Beverages tobacco and Nonfood tems Note: All three tarff ndces compare the actual tarff structure wth free trade. See text for detals. 23