Endogenous Protection in Mercosul: An Empirical Analysis. Germán Calfat *, Renato G. Flôres Jr.** and Maria Cecilia Ganame***

Transcription

1 Endogenous Protecton n Mercosul: An Emprcal Analyss Germán Calfat *, Renato G. Flôres Jr.** and Mara Cecla Ganame*** * Unverstet Antwerpen - UFSIA, Antwerp, Belgum ** EPGE/Fundação Getulo Vargas, Ro de Janero, Brazl *** Unversdad de Córdoba, Córdoba, Argentna (Prelmnary Verson: Aprl 2000)

2 ABSTRACT Poltcal economy vews of trade polces are mportant n order to understand the structure of trade protecton that bears not only on a partcular country but also les behnd most nternatonal agreements. Ths paper analyses the equlbrum structure of protecton n Mercosul, developng emprcal analyses based on the lterature ensung from the sequence of models set forth by Grossman and Helpman snce Not only Mercosul s common external tarff (CET) may be explaned under a poltcal economy perspectve, but the exstence of devatons both at the level of the external tarffs and at that of the nternal ones, make t nterestng to contrast several structures under ths approach. Dfferent general equlbrum frameworks, n whch governments are concerned both wth campagn contrbutons and wth the welfare of the average voter, there exstng a set of organzed specal-nterest groups that care only about the welfare of ther members, are used as the theoretcal bass of the emprcal tests. The nteracton between the government and the nterest groups whch make contrbutons to the government n exchange for protecton - gves as outcome the structure of protecton. We buld a sngle equaton for explanng the CET and two four-equatons systems (one equaton for each member) for explanng devatons from the CET and from the nternal free trade between members. The results (at the two-dgt level) shed an nterestng lght on the sectoral dynamcs of protecton n each country; notably, Brazl seems to ft n better n the model framework, followed by Uruguay. In the case of the CET, and of devatons from t, the nteracton between the domestc lobbes n the four countres plays a maor role. There s also suggeston that, the lobby structure that bd for devatons, be they nternal or external, dffers from the one whch bd for the CET. 2

3 1. Introducton Qute often countres desgn ther trade polces n a way that yelds to pressure from specal nterest groups. Trade negotatons n the nternatonal arena respond smlarly. Poltcal economy explanatons of trade polces are mportant n order to understand the structure of trade protecton that arses not only n a partcular country but also n nternatonal agreements. Foremost among Mercosul s obectves are the elmnaton of nternal tarffs between members and the establshment of common external tarffs wth the rest of the world. 1 The structure of the common external tarff (CET) commtted to by members may be explaned from a poltcal economy vewpont. Moreover, devatons from t as well as from free trade among members, though lastng for a transtonal perod, reflect a protecton beyond that underlyng the agreed structure. Hence, these features should also be analysed from a poltcal economy approach. In an attempt at explanng protecton n Mercosul, Olarreaga and Soloaga (1998) provded emprcal evdence about tarff formaton, by focussng on seven predctons of the correlate of expected cross-sectoral varaton n tarff formaton. These predctons n a partal equlbrum context, where protecton s estmated by ncorporatng explanatory varables addtvely - were based not only on the theoretcal frameworks but also on emprcal grounds. Ths paper departs from the partal equlbrum analyss and consders alternatve general equlbrum formulatons, nspred n the seres of theoretcal models developed by Grossman and Helpman (1994, 1995a, 1995b, 1996) for analysng the settlng on of trade polces n representatve democraces. In ther frst paper, they bult a general equlbrum model n whch governments are concerned both wth campagn contrbutons and wth the welfare of the average voter, and there exsts a set of organzed specal-nterest groups that care only about the welfare of ther members and exchange contrbutons for protecton. The nteracton between the government and the set of nterest groups results n the structure of protecton. Takng ths approach as the 1 The Mercosul agreement has other requrements besdes those related to customs dutes. As t envsages the free movement of goods, servces and factors, measures such as the elmnaton of nontarff restrctons, the adopton of a common trade polcy, and the coordnaton of macroeconomc and sectoral polces are also n progress. 3

4 bass of our specfcatons, we extend t n dfferent ways n order to analyse the structure of protecton n Mercosul. The paper s structured as follows. In secton two, nsttutonal and legal features of Mercosul are descrbed n order to gve general nformaton about the agreement and provde a backgroung for the analyses that follow. Secton three revews the Grossman- Helpman models and dscusses the three extensons, and correspondng specfcatons, made. Secton four descrbes the varables and touches a few mportant econometrc ssues. Secton fve presents the emprcal fndngs and secton sx concludes. 2. About Mercosul: Insttutonal and egal Aspects 2.1 Basc facts Mercosul, the Southern Cone Common Market, was formed on March 1991 by the Treaty of Asuncón, and conssts of Argentna, Brazl, Paraguay and Uruguay. ater, Chle and Bolva became assocate members n October 1996 and March 1997, respectvely, formng a Free Trade Area wth Mercosul. The Treaty of Asuncón n ts Chapter II and the 1994 Ouro Preto Protocol created sx dfferent nsttutons to mplement Mercosul s prncples and purposes. They are the Councl, the premer poltcal nsttuton; the Common Market Group, Mercosul's executve organzaton wth both polcy-makng and admnstratve responsbltes; the Commerce Commsson, responsble for montorng the applcaton of a common commercal polcy; the Jont Parlamentary Commsson, a consultatve body; the Socal- Economc Consultatve Forum, an advsory body, and the Admnstratve Secretarat, whch provdes operatonal support. The Mercosul legal system conssts of the Treaty of Asuncón, ts Protocols, Annexes and related agreements. The effectveness of these documents and the herarchy gven them n each member state depends on the respectve natonal laws. The second level of the legal system conssts of the Councl decsons, Group resolutons and Commerce Commsson drectves, Rowat et al. (1997). The Common Market envsages the free movement of goods, servces and factors of producton, the elmnaton of customs dutes and non-tarff restrctons, the establshment of a common external tarff (CET) and the adopton of a common trade polcy, and the coordnaton of macroeconomc and sectoral polces. These goals are 4

5 beng progressvely acheved and, takng nto account the speed of a smlar experment the European Unon, and the great macroeconomc nstablty whch plagued all ts members n the recent past, Mercosul can only but be consdered a successful work n progress. 2.2 Internal trade lberalzaton and the Common External Tarff Artcle 5 of the Treaty of Asuncón set out the agreed Trade beralzaton Programme. Between June 1991 and December 31, 1994, barrers to trade among Mercosul partners coverng close to 95 percent of ntra-regonal trade were elmnated. An adustment regme, agreed n August 1994, led to a progressve phase out of the remanng tarffs on ntra-regonal trade. Senstve tems, contaned n ndvdual exceptons/adustment lsts, were supposed to be phased n between January 1, 1995 and December 31, 1998, for Argentna and Brazl, and between January 1, 1996 and December 31, 1999 for Paraguay and Uruguay. The man lst of goods covered s not extensve, though comprsng (senstve) agrcultural as well as ndustral products, and s estmated to cover 5 percent of ntra trade. However, ths does not nclude the exceptons made for sugar or automobles, the latter beng of partcular mportance n the Argentna-Brazl trade. Mercosul s Common External Tarff (CET) has been n force snce January 1995, coverng 85 percent of the goods traded wthn the group. Durng a transtonal perod, senstve sectors have also been excluded, there exstng common and natonal lsts of exceptons to the CET. The common lst ncludes tems such as captal goods, nformatcs and telecommuncaton products. The natonal lsts allow Argentna, Brazl and Uruguay to have 300 natonal exceptons to the CET whle Paraguay has 399 exceptons. Full mplementaton of the CET s to take place by Besdes ther natonal lsts of exemptons, each country has ts own specal concessonary régmes, where the rates may be reduced below the scheduled values. Examples are nvestment tems and the temporary admsson of goods to be re-exported. These régmes wll be consoldated nto a common Mercosul set of rules (see, for nstance, ard(1997)). 2 A temporary ncrease of the CET by 3 percent ponts on most tarff tems was agreed on n December 1997 to accommodate Brazl's external mbalance and allow Argentna to reduce the levels of ts statstcal tax. 5

6 Non-tarff measures Safeguards were regulated n Annex IV of the Treaty, followng the gudelnes lad down n Artcle XIX of the GATT. Actons aganst thrd countres may be taken by Mercosul as a group or on behalf of a sngle member. On the other hand, safeguard actons by one member aganst another are forbdden. Concernng antdumpng and countervalng measures, there s no agreement yet among members. These actons are allowed and can be n force untl December 2000, when a set of common rules on the Defence of Internal Competton shall then preval. In fact, co-ordnaton of sectoral polces n agrculture, ndustry, energy, transport and labor has been dscussed. Negotatons are under way for the harmonzaton of remanng restrctons on nternal trade n agrcultural products. In March 1998, the scope of the work was expanded to nclude the analyss and dscusson of ndvdual agrcultural and agro-ndustral polces. In order to foster the ratonalzaton of nvestments and the ncrease n the compettveness of frms, the prvate sector may negotate producton-sharng arrangements. Blateral arrangements between Argentna/Brazl, Argentna/Uruguay and Brazl/Uruguay n the automotve sector are n force untl a common sectoral polcy be establshed. 3 The trade arrangement between Argentna/Brazl, n whch the sector benefts from local content plants, allows concessonal entry on vehcles and parts as well as export balancng requrement. Fnally, t s mportant to menton that no mprovement has been reached yet on the free movement of servces, captal and workers. Although there exsts a framework agreement, sgned n December 1997, for freeng trade n servces wthn Mercosul over a ten-year perod, much remans to be done n such areas. 3. The Theoretcal Framework and the Derved Specfcatons 3.1 The basc model The Grossman-Helpman (1994) model can be vewed as a two-stage noncooperatve game n whch nterest groups move n the frst stage, offerng poltcal contrbuton schedules that depend on ther polcy stance, whle the government chooses, 3 At the tme of ths wrtng, there s an ntensve negotaton to reach a common sectoral polcy between Argentna and Brazl. 6

7 n the second stage, polcy stances maxmzng a poltcal obectve functon that depends on contrbutons and the well-beng of the average voter. The model has the structure of a common agency problem: several prncpals, the specal nterest groups, attempt to nduce a sngle agent, the government, to take an acton that may be costly for hm to perform. The formal framework consders a small and compettve economy, ntegrated by ndvduals wth dentcal, completely addtve preferences, wth each sub-utlty functon dfferentable, ncreasng and strctly concave. The derved demand for good s x = d ( p ) ; and the ndrect utlty at expendture level E and a vector p of domestc prces takes the form V ( p E) = E + s( p),, where s(p) s the consumer surplus. A numerare good s produced wth a labour nput-output coeffcent equal to one, so that the wage rate s also one. Non-numerare goods are produced from labour and sector-specfc nputs. Wth constant returns to scale and the wage rate equal to one, the aggregate reward to sector s specfc factor, π ( p ), depends only on p. The government mplements trade taxes and subsdes, ntroducng a wedge between domestc and world prces. If a domestc prce p exceeds the world prce p *, an mport tarff for a good that s mported or an export subsdy for one that s exported, s mplemented. Conversely, domestc prces below world prces correspond to mport subsdes or export taxes. Government revenue s dstrbuted n a lump-sum fashon. The government cares about total poltcal contrbutons and aggregate welfare. The government's lnear obectve functon s (1) where: G = C ( p) + aw ( p) a 0 a s the government's senstvty to the average voter's well-beng relatve to ts taste for campagn contrbutons; C ( p) s the poltcal contrbuton of the (exstng) lobby n sector ; and W(p) s the aggregate welfare, equal to the sum of the ndrect utltes of all ndvduals (consderng that aggregate ncome s the sum of labour ncome, rewards to the specfc nputs and tarff revenue): 7

8 W n ( p) = l + π ( p ) + N[ r( p) + s( p) ] = 1 (2) 1 = p p d p y p N * where r(p) s the tarff revenue gven by ( ) ( ) ( ) ( ) s the number of consumers. r p, and N An exstng lobby, n sector for nstance, talors the schedule of contrbutons to maxmse the total welfare of ts members. The ont welfare s V = W C, where W s ther gross-of-contrbutons ont welfare, whch s an mmedate adaptaton of (2) above: (3) W ( p) = l + π ( p ) + α N[ r( p) s( p) ] + where l s the labour ncome and α the fracton of populaton who owns some of the specfc nput used n ndustry. Grossman and Helpman consder that the nteracton between the government and lobbes has the structure of a menu-aucton characterzed by Bernhem and Whnston (1986). The equlbrum outcome from ths structure s the same as that from maxmzng the ont surplus (4) Ω = aw + W = a W + ( a + 1) W. In equlbrum, truthful contrbuton schedules nduce the government to behave as f t were maxmzng a socal-welfare functon that weghts dfferently the members of socety. obby groups receve a weght of 1 + a and ndvduals that are not organzed receve a smaller weght of a. Operatng n equaton (4) and defnng varable I as takng value one when the group s organzed or zero otherwse, rewrtten as I 1 f =, Ω can be 0 otherwse (5) n ( a + I ) π + N[ r( p) + s( p) ]( a α ) Ω = ( a + α ) l + + = 1 8

9 where α = α characterzes the share of voters who own specfc factors. By convenently usng Hottelng s emma and the Envelope Theorem, the frst order condton becomes: Ω p = * ( a + I ) y ( p ) + ( a + α )[ y ( p ) + ( p p ) m ] = 0 (6) where m s the dervatve of sector s mport functon m Nd ( p ) y ( p ) Operatng, one arrves at =. * ( p p ) = I α a + α ( p ) y m (7) that gves the structure of protecton. Formula (7), whch s the pont of departure of all econometrc work, may be expressed n terms of the mport elastcty e and the mport penetraton rato z. Knowng * that p = p ( 1 + t ) elastcty s e m = p and m, where t s the ad-valorem tarff rate, and gven that the trade y z = m ( p ) ( p ), equaton ( 7 ) can be rewrtten as (8) t I = α z 1+ t a + α e Ths equaton represents the poltcal equlbrum descrbed n Proposton 2 of Grossman and Helpman (1994). Industres that have hgh mport demand or hgh mport supply elastctes wll have smaller ad valorem devatons from free trade. Ths may be explaned by two reasons. Frst, f a > 0 the ncumbent government may bear a poltcal cost from creatng deadweght loss. Second, even f a = 0, f α > 0 the owners of 9

10 specfc nputs to ndustres other than wll bd more to avod protecton n sector the greater the socal cost of that protecton s. Proposton 2 also shows that all sectors that are organzed are protected by ether an mport tarff or export subsdes. Conversely, sectors that are not represented by lobbes are burdened wth mport subsdes or export taxes. The poltcal power of a partcular organzed sector s reflected by the rato of domestc output to mports. In sectors wth large domestc output, the specfc-factor owners have much to lose from free trade. On the other hand, when the volume of mports s low, the economy has relatvely lttle to lose from protecton. Besdes, the smaller the weght the government places on aggregate welfare, the larger n absolute value all trade taxes and subsdes wll be. If a = 0, an nteror soluton remans possble due to the fact that the nterest groups themselves do not want to dstort too much. As the share of voters who are members of one nterest group or another grows, equlbrum share of protecton for organzed sectors declnes. When all voters belong to an nterest group ( = 1) α and all sectors are represented (I =1 for all ), then free trade emerges n all markets as the varous nterest groups neutralse one another. On the other hand, f nterest groups are a neglgble fracton of the votng populaton ( α = 0), then no trade taxes or subsdes wll be appled to goods not represented by lobbes. Indeed, when the potental poltcal contrbutons are too few, they stand lttle to gan from free trade nterventons n sectors other than ther own. The theoretcal model above corresponds to the analyss of one country. In the case of Mercosul, t would be correct to use an approach where more than one country set trade polces cooperatvely. Grossman and Helpman (1995a) have extended the nfluence-drven contrbuton approach to a settng wth two countres that desgn trade polces cooperatvely. The outcome s revealed n the rato of domestc country tarff to foregn country tarff. The equaton of the rato tells that, relatve to free trade, the negotated trade agreement favours the nterest group that has greater poltcal clout between the two countres. Hence, the model explans devatons from common external tarffs and from free trade rather than the structure of common external tarffs. Moreover, the specfcaton of such models wth more than two countres turns to out to be very dffcult and complcated. 10

11 Fnally, Grosman and Helpman (1995b) ntroduce the possblty of ndustry exclusons from free trade, n a free trade area. Ths concept s extended here for the excluson from the CET, as well as, n a case closer to thers, n the nternal Mercosul trade. 3.2 Model extensons Equaton (8) apples to the case of a small country, nternally desgnng ts trade polcy. Mercosul negotatons nvolve drectly or ndrectly nteractons of the lobbes n the four member countres, thus requrng further assumptons and extensons. We consder frst two cases where the smultaneous nteracton s relevant: the establshment of the common external tarff and the natonal devatons from the CET. Then, we try to explan the devatons from nternal free trade wthn member countres 4. Our specfcatons wll broadly consst n addtve extensons of (8). We begn wth the one for the CET. Followng already standard practces, Goldberg and Magg (1999), n equaton (8) the error term s ntroduced addtvely; as t can be thought of as a composte of varables that potentally may affect protecton and have been left out of the theoretcal model, and also as an error n the measurement n the dependent varable. Besdes, the mport demand elastcty s brought beforehand to the left hand sde (lhs). The model to estmate Mercosul s common structure of protecton supposes that four equatons lke (8) are behnd the negotatons, one for each member. Supposng dentcal mport elastctes (by country/member) 5, fnal establshng of the CET mples that the lhs of the four equatons wll be the same. We shall then wrte ths term as f resultng from a weghted average of the four orgnal equatons, gvng rse to the followng CET equaton: 4 As a Common Market, Mercosul must not have nternal tarffs. The exstence of these tarffs are consdered as an nstrument of protecton that mght be explaned by the Grossman-Helpman approach. 5 Ths assumpton s not crucal and s nether the most debatable part of ths extenson. Moreover, n the emprcal estmatons, for lack of better data, the values of the elastctes were actually the same for all the four countres. 11

12 t 1+ t CET CET e = U = A θ I a α + α z + ε = U = A γ θ z + U = A δ θ I z + ε ; γ = α a + α and δ = a 1 + α (9) Equaton (9) expresses the structure of protecton for the unon as whole, wth each country enterng addtvely n the formaton of t CET, the common external tarff for good. Two optons are avalable for the weghts θ. The frst s to consder that, n spte of the clear sze dfferences, n the negotatng table, Mercosul lke the EU gves the same mportance to all ts members, the CET beng the result of reasonably balanced negotatons. Ths makes for settng all θ = ¼. The other, as suggested by Cadot et al. (199) and used by Olarreaga and Soloaga (1998) s to consder that sze matters n the negotatons and weght the countres dfferently by settng the θ equal to the rato of the country s domestc output to total Mercosul output for sector. The poltcal power of each country, n an organzed sector, s reflected by the rato of domestc output to mports; equaton (9) shows that the domestc poltcal envronment constrans the actons that the governments can take regonally and that regonal nterdependence (addtvely) sets the fnal parameters. The weghted sum of the poltcal varables of each country makes the CET dependent on the power of all lobbes n the ntegraton and the structure of protecton determned by ther nteracton. Equaton (8) mples that, for each country, γ < 0, δ > 0 and γ +δ > 0. These sgn relatonshps should contnue to be vald n the aggregate relatonshp (9). Moreover, knowledge of γ and δ allows to compute two key parameters of the model: the fracton of voters represented by lobbes, α = -γ /δ, (10) and the (absolute) weght that the government gves to consumers welfare, a = 1 + γ$ δ$ (11) However, nstead of lookng at a, t s preferrable to make n (1). 12

13 a = β/(1-β), 0 β <1, (12) so that β s now the relatve weght gven to consumers welfare, whle 1-β that gven to the lobbes. From (11) and (12) t follows that: β = (1+γ )/(1+γ +δ ). (13) The z varables, representng the rato of domestc output to mports by each member, (= Argentna, Brazl, Paraguay and Uruguay), are clearly endogenous. A reduced form specfcaton (14) where the Z z y = m = ξ Z + υ 1 varables are such as the number of employees, captal and ndustry concentraton ndexes (see Goldberg and Magg (1999)), s needed. * When the devaton from the CET s examned, the quantty p = p ( 1 + t ) may be looked at as the new world prce, so that, n a gven country, the lobby n sector would be fghtng for an addtonal tarff revenue, per unt of mport, equal to (15) p * (1+t ET ) - p * (1+t CET ) = p * (t ET - t CET ) where t ET s the devant external tarff. Moreover, we suppose that for settng rates dfferent from the CET, domestc lobbes wll have to face the nterests of the other member s lobbes, not necessarly dentcal to thers. In sectors where producers/owners are smlarly organsed n the four countres, obtanng the devaton wll requre an effort lke n (8), but f the combned lobby structure of the other partners dffers substantally from the domestc one, t mght be easer or more dffcult to get the overprotecton reflected n (11). The above consderatons lead to modfy (8), for each country, nto: t ET t 1+ t CET CET e ' ' = γ z + δ1 I z + δ 2 ( I z θ * ' I z ) + ε for every country (16) ' 13

14 The correcton n the rhs answers for the ncremental character of the negotaton, and comes from (11). The term added to the lhs w = I z θ * ' I z ' ' ' tres to account for the nteracton wth the other members lobbes. If w > 0 the domestc lobby s supposed to be more powerful (or aggressve) than the combned others, ths meanng that t wll be more dffcult to wn the overprotecton, as t wll fnd less support for ths n the other members. If w < 0 the reverse apples, and an extra support for ts clams wll come from the other members. Fnally, w = 0 makes the negotaton dentcal to the stuaton n (8). Overall, governments here, a common Mercosul authorty - may consder the devatons from the CET, allowed durng the so-called adustment perods, as necessary n order to make the agreement palatable to opposng nterests. It seems reasonable to suppose that, n ther decson makng, they wll udge each domestc clam aganst the stuaton prevalng n the rest of Mercosul. The nteracton between I z and w tres to capture ths. 6 * Weghts θ are, now, the rato between country s output n sector and the total output, for the same sector, of the three outsde countres. From the prevous dscusson, we expect that δ 2 < 0. We also expect that though there are really no theoretcal grounds to sustan ths. It seems more reasonable to clam that should be postve. There are four equatons equal to (16) and, of course, the four reduced forms (14). These equatons would represent the 'over-protecton' by each country, beyond that already exstng wthn the bloc, relatve to the rest of the world. The fact that several goods were excluded from the elmnaton of tarffs wthn Mercosul shows that ther sectors have poltcal power n the determnaton of tarffs and could perhaps also be explaned n the above lnes. These nterest groups receve protecton not only wth regard to the rest of the world through the CET but also wthn the bloc and, as mentoned above, the excluson of sectors over long perods of adustment may make an agreement poltcally vable. 14

15 Specfcaton of the nternal tarff model should maybe consder the role played by the other members. Notwthstandng, we shall consder that ths stuaton s the one closest to the Grossman and Helpman (1994) context, wth the equlbrum beng truly free trade, and apply t wth the slght modfcaton that varable z - now called z * -wll be the rato of domestc output to mports from Mercosul members: (17) t IT IT 1+ t e I = a = γ α + α z * z * + I δ + ε z + ε for country = A,B,P, U where IT t s the nternal tarff appled by country. Here, there are also four pars of equatons, one for each country. 5. Data and Econometrcs 5.1 The data The data used refer to the four members: Argentna, Brazl, Paraguay and Uruguay. Though t s arguable whether Paraguay should be ncluded, due to the lesser relablty of ts trade statstcs and the small magntude of several observatons, we decded to ncorporate t n order to have a full pcture of Mercosul. The basc data fle s the same used by Olarreaga and Soloaga (1998). They aggregated tarffs and trade data up to the eght-dgt level of the harmonzed system. These were then converted to UNIDO's three dgt level, as ndustral data for Mercosul countres came from UNIDO's three dgt database. We shall now brefly dscuss each of the varables n the model. Protecton measures Tarffs are used to measure protecton. Import-weghted or smple average tarffs consdered, coverng the 27 sectors whch make the three-dgt aggregaton level. It would have been more correct to nclude non-tarff measures as well, n the structure of 6 Of course, other knds of nteractons are possble, and further exploraton of ths s needed. It mght also be argued that (economc) sze matters n ths case and we are not accountng for t. Indeed, exceptons were granted to Paraguay, for nstance, under these grounds. 15

16 protecton. These are appled by Mercosul members, Argentna, for nstance, havng qute frequently used ths nstrument aganst Brazl. The fact that non-tarff measures are not consdered underestmates the structure of protecton; however, gven the problem n combnng tarff and non-tarff barrers n a sngle ndex, ths was left for a future mprovement. Elastctes Though estmates of trade elastctes exst, n dfferent aggregaton levels and for vared perods and scenaros (see, as an example, Zn (1988) for Brazl), we dd not fnd a homogenous set, referrng approxmately to the same date. As other authors, we took Shells et al. (1986) estmates of US mport demand elastctes at the ndustry level as proxes to the Mercosul countres s elastctes. As Goldberg and Magg (1997) do, mport demand elastcty of good,e, s brought on the left hand sde of the estmatng equaton to consder a measure of error n the elastctes as a component of the measurement error of the dependent varable. Export supply elastctes are not consdered n the analyss due to the mpossblty to obtan relable data on these elastctes. Poltcal organzaton dummes In prncple, one should be able to dentfy the organzed ndustres by lookng at the contrbuton level of the respectve group. However, data on contrbutons of lobbes are not avalable for Mercosul countres. The I dummes must then be created wth the help of proxes. Olarreaga and Soloaga (1998) used an ndustry concentraton ndex to nfer whch sector s an organzed group. Ths ndex s calculated as (number of frms n the whole economy)/(number of frms n sector ). If concentraton s hgh, the free rdng ncentves could be solved and frms mght form a lobby. The poltcal organzaton dummy s constructed somewhat arbtrarly, consderng a value of 1 whenever the ndex s above a threshold of 50 whch represents 2 percent of the total number of frms n the economy -, otherwse the dummy s set equal to zero. Ths leads to a few awkward results, partally due to the aggregaton level used; the most conspcuous beng a zero for the transport materal sector. Also, for sectors such as beverages (ISIC 313); textle (321); wearng apparel, except footwear (322); footwear, except rubber or plastc (324); furnture, except metal (332) and prntng and publshng (342), whch present a low ndex of concentraton, there exsts the suspcon that, n 16

17 some of the four countres, they overcame the free rdng ncentves and bd for protecton. We combned the concentraton ndex wth... Import penetraton rato The mport penetraton rato s perhaps the worst measured varable n our model. The reason may be that the numerator and denomnator come from dfferent statstcal sources, rasng the suspcon that the dfferent measurement errors may sometmes oddly combne n the rato. 5.2 A few ponts on the econometrcs Model (9) s a one-equaton system, whle models (16) and (17) are fourequatons systems. In the three cases there are 27 observatons and varables z - or z * - are consdered as endogenous,.e. correlated wth the shocks n the model. To solve ths problem, a reduced form lke (14), where the mport penetraton rato s wrtten as a functon of exogenous varables that may ndrectly affect protecton, must be estmated. Followng prevous emprcal exercses lke Goldberg and Magg (1999), we started by consderng as exogenous varables the number of employees, captal and the ndustry concentraton ndex. Table A.1 n the Annex shows the results: wth the excepton of Brazl, the fts are not good (the R 2 are also low). A man reason for ths s the fact that the output-mport rato, for all countres but Paraguay, has a very skewed dstrbuton, wth one or two values for each country standng as true outlers. We expanded the set of exogenous and, n model (16), removed an outler sector from the Argentnan data. When estmatng the systems, we took also nto account the correlaton between the shocks for the same sectors n dfferent countres, what, takng nto account the prevous estmaton of the output-mport rato, amounts to usng the three-stage least squares technque. Moreover, heteroskedastcty correctons were also employed. Fnally, all equatons were estmated wthout a constant, so that care must be taken n the nterpretaton of the R 2 n the next secton. 6. Results 17

18 The results from estmatng the model that represents the structure of protecton of the CET, there are fve equatons to be estmated. Table 1 reports the coeffcents of the trade protecton equaton (9), correspondng to the unform weghtng case, whch proved superor to the proportonal one. Ths mght sgnal that, for establshng the CET, the four members had a balanced nfluence. The sgns and the t statstcs of coeffcents γ and δ for Brazl and Uruguay are sgnfcant and consstent wth the Grossman-Helpman model. But for Brazl s γ, all the other estmates are statstcally sgnfcant at the 5% level. The other mplcaton of the Grossman-Helpman model, γ +δ >0, s also statstcally confrmed, at the 10% level, for the same two countres. Estmates for Paraguay present sensble sgns, but are not statstcally sgnfcant. The odd result concerns the estmates for Argentna whch, though sgnfcant, present changed sgns; ther sum, however, s statstcally accepted as postve at 10%. For the countres wth wholly sgnfcant results, we computed the values of β and α. They are, α = 2/3 and β = for Brazl α = 0.86 and β = for Uruguay. These values ndcate that lobbes are more present n Uruguay than n Brazl, though both governments gve a neglgble weght to them n settng trade polces. The frst pont seems reasonable: the small dmensons of the Uruguayan economy make t easer to get producers wth smlar nterests together; as for the second, t mght mean that, at the level of aggregaton we are workng, strong lobbes lke the automotve sector are dluted wthn the bgger ndustry. The fndngs broadly support the model predctons, wth poltcally organzed sectors clearly recevng more protecton from Brazl and Uruguay. The structure of the CET s thus the combned result of the pressure of senstve sectors that ask for protecton n each of the member countres. When sector s organzed n all countres of the agreement, the power to obtan protecton from the rest of the world s enforced and a hgh common external tarff may be mposed. Conversely, when organzaton dffers wthn the Mercosul, the varous nterest groups are less able to brng together ther power at the negotatng table. 18

19 Table 1: Estmaton of the Common External Tarff model Coeffcent t-statstc Prob. γ A 23.5E δ A -21.4E γ B -1.2E δ B 1.8E γ P -2.0E ns δ P 3.1E ns γ U -39.5E δ U 45.9E R-squared 0.56 Table 2 reports the results on the devatons from the CET. The sgns of the δ 2 coeffcents are all consstent and sgnfcant. The estmated γs are also all negatve and, but for Paraguay, statstcally sgnfcant. The estmated δ 1 s are postve for Argentna, Brazl and Uruguay, but sgnfcant only for Brazl. These fndngs ndcate a clear nteracton between the domestc lobbes n the bddng for the exceptons to the CET, however, only n Brazl the expected structure of protecton s wholly sgnfcant, the model fndng a weaker support n Argentna and Uruguay and almost none n Paraguay. Though the sgns of the coeffcents n Argentna and Uruguay are as expected, the domestc lobby coeffcents are not statstcally sgnfcant. Ths rases the suspcon that a dfferent structure of nternal lobbes mght bd for protecton beyond that one obtaned n the CET. In fact, some sectors may want to form a pressure group to nfluence the desgn of the CET but may not be nterested n nfluencng the structure of devatons from t. 7 7 The cost of bddng for protecton could then be greater than the beneft from t. 19

20 Table 2: Estmaton of the External Tarff Devatons model Coeffcent t-statstc Prob. γ A -1.43E A δ E ns A δ E R-squared 0.37 γ B -2.20E δ 1 B 2.57E δ 2 B -0.52E R-squared 0.28 γ P -3.44E ns δ 1 P -0.27E ns δ 2 P -0.65E R-squared 0.12 γ U -0.26E U δ E ns U δ E R-squared 0.49 The results of a senstvty analyss consderng an alternatve specfcaton of the lobby representaton dummy, n whch several sectors are not consdered as pressure groups to nfluence the structure of the external tarff for Argentna (322, 332, 342), Brazl (321, 342), Paraguay (313, 321, 322, 332, 342) and Uruguay (321, 324, 332) present a slght mprovement. The Argentnan δ 1 becomes sgnfcant at 10%, and the one for Paraguay changes to the correct sgn, though remanng not sgnfcant. Therefore, through the second specfcaton of the external tarff model, t may be seen that the power of the nterest groups to bd for protecton would also be present, at least n one of the four countres that make up Mercosul: Brazl. Interest groups that are organzed and have much to lose wth the agreement acheve protecton from the government of ther respectve country. However, ths over-protecton n some senstve 20

21 sectors happens durng a transtonal perod. All members countres have to converge to common external tarffs. Fnally, t s mportant to note that the devatons are not postve n all countres. Paraguay and Uruguay wll converge upwards to the CET n several sectors. In these small and relatvely more open economes, organzed sectors may fal to obtan protecton due to the fact that ther ratos of domestc output to mports are low. As Grossman and Helpman (1994) explan, f the share of voters belongng to an nterest group s hgh, the equlbrum share of protecton for organzed sectors s low, as the varous nterest groups neutralze each another. Interest groups wth a low rato of domestc output to mports, whch n the model - represents ther power, fal to obtan protecton wth a relatvely hgh α. As estmated above, Uruguay s α was equal to Table 3 reports the results of the wthn Mercosul tarff model, whch ams at explanng the devatons from free trade among member countres. The sgns of the coeffcents γ and δ are consstent wth the theoretcal model n the cases of Argentna, Brazl and Uruguay. The estmated γs are negatve and statstcally sgnfcant for these three countres. The δs are postve but sgnfcant only for Brazl. The estmates for Paraguay present unexpected sgns and no sgnfcance. The relaton γ + δ > 0 seems vald, at 10%, for Argentna; but the null γ + δ 0 cannot be reected at ths level ether for Brazl or Uruguay. As n the prevous model, these fndngs may suggest that a dfferent structure of nternal lobbes mght bd for protecton wthn the free trade area, what would call for a reapprasal of the dummes I. Table 3: Estmaton of Internal Tarff model Coeffcent t-statstc Prob. γ A -1.11E δ A 2.23E ns R-squared 0.21 γ B -0.05E δ B 0.05E R-squared 0.37 γ P 1.82E ns δ P -3.00E ns R-squared 0.09 γ U -1.97E

22 δ U 1.83E ns R-squared 0.32 Brazl, then, stands out as the more consstent country, n terms of the unformty of ts lobbes n all three stuatons analysed. It s however tellng the much lower value of ts coeffcents, sgnallng that the resultant protecton s lower than n the other members what s ndeed true n the case at stake. Indeed, all members devate from free trade n only few sectors. Brazl devates n 3 of the 27 (aggregate) sectors that are beng consdered; Paraguay and Uruguay devate from free trade n a greater percent than Brazl and Argentna but the devatons represent less than 5 percent of the total tarff lnes. 8 Summng up the three models, n the case of Brazl, though the government attaches consderable weght to socal welfare, protecton arses n the external tarff structure of Mercosul. The multplcty of lobby representatons n the country s not so hgh as to neutralze the power of the nterest groups that stand to lose. The pattern of protecton n Mercosul, through the CET, external and nternal tarff models s qute consstent wth the basc predctons of the theory. Explanaton of protecton n Uruguay s qute dfferent. Though the degree of lobby representaton and the relatve weght of socal welfare are hgh, protecton exsts. The reason must le n the fact that there are mportant sectors that are not organzed. The case of Argentna deserves further thoughts. 6. Concludng Remarks The Grossman-Helpman model seems to be a parsmonous, promsng approach to explan the structure of protecton that sprngs from the Mercosul agreement. We consdered n our analyss two of the three varables that explan protecton: the poltcal organzaton dummy and the rato of domestc output to mports. Trade (mport) elastctes were ncorporated n a partcular, less precse manner. Import elastctes are on the left hand of the equaton n order to consder ts measure error as a component of the measurement error of the dependent varable. Export elastctes were 8 As Olarreaga and Soloaga (1998) ndcate, n 1996 devatons from nternal trade only corresponded to 0.2 percent of total lnes for Brazl, 2.5 percent for Argentna, 3.3 percent for Paraguay and 4.4 percent for Uruguay. 22

23 not used n the analyss due the fact that no relable estmates were found. Therefore, term-of-trade effects dd not enter n the argument. The estmatons suggest that some sectors receve protecton both through the CET and ts devatons. The common external tarff model gves evdence about protecton n Brazl and Uruguay. Furthermore, the external tarff model supports protecton to senstve groups n Brazl. The structural parameters confrm that a certan degree of protecton n Brazl and Uruguay does exst. Though n Brazl there s a relatvely hgh degree of lobby representaton, ths s not so hgh as the one necessary to neutralze the power of nterest groups. Uruguay presents a dfferent pcture. The degree of lobby representaton s hgher, but protecton may exst due to the fact that all sectors are not represented and the output/mport rato s very hgh for some. To conclude, an mportant result s acheved: there s a dstnct pattern of protecton that depends on whether or not the sector s organzed. The postve correlaton between protecton and the mport penetraton rato (the nverse of the domestc output to mports rato) consdered n partal equlbrum analyss apples only to the non-organzed sectors; wthn organzed sectors ths correlaton s negatve. 23

24 ANNEX: The Reduced Form for the Output-mport Rato Table A.1: Estmaton of the reduced form of the output-mport rato equaton Coeffcent t-statstc Prob. Argentna Constant ns Employment A ns Concentraton Index A ns Captal A 1.36E ns Brazl Constant Employment B Concentraton Index B Captal B -6.21E Paraguay Constant ns Employment P ns Concentraton Index p Captal P 4.38E ns Uruguay Constant ns Employment U ns Concentraton Index U Captal U ns 24

25 REFERENCES Braga, Helson C.; Santago, Glda M. C. and Ferro, uz C. M Estrutura da Proteςão Efetva no Brasl:1985. Pesqusa de Planeamento Economco 18 (3): Dxt, Avnash; Grossman, Gene and Helpman, Elhanan 'Common Agency and Coordnaton: General Theory and Applcaton to Government Polcy Makng'. Journal of Poltcal Economy 105 (4): Goldberg, Pnelop K. and Magg, Govann 'Protecton for Sale: An Emprcal Investgaton'. NBER workng paper # Greene, Wllam H Econometrc Analyss. Prentce-Hall. Grossman, Gene and Helpman, Elhanan 'Protecton for Sale'. Amercan Economc Revew 84(4): 'The Poltcs of Free Trade Agreement'. Amercan Economc Revew 85(4): 'Trade Wars and Trade Talks'. Journal of Poltcal Economy 103(4): 'Electoral Competton and Specal Interest Poltcs'. Revew of Economc Studes 63(2): Helpman, Elhanan 'Poltcs and Trade Polcy'. NBER workng paper # Hllman, Arye 'Declnng Industres and Poltcal-Support Protectonst Motves'. Amercan Economc Revew 72(5):

26 ard, Sam 'Mercosul: Obectves and Achevements'. Staff Workng Paper IPRD World Trade Organzaton, Geneva, Processed. Mayer, Wolfgang 'Endogenous Tarff Formaton'. Amercan Economc Revew 74(5): Olarreaga, Marcelo and Soloaga, Isdoro 'Endogenous Tarff Formaton: The Case of Mercosul'. The World Bank Economc Revew 2(2): Olarreaga, Marcelo; Soloaga Isdoro and Wnters, Alan 'What's behnd Mercosul's CET'. Trade Semnar at the World Bank. Rowat, Malcolm; ubrano, Mchele and Porrata, Rafael 'Competton Polcy and Mercosul'. World Bank techncal paper # 385. Shells, Clnton; Stern Robert and Deardoff Alan 'Estmates of the Elastctes of Substtutons between Imports and Home Goods for the Unted States'. Welwrtschaftlches Archv 122: Trade Polcy Revew of Argentna World Bank Trade Organzaton WT/TPR/S/47. Zn, Alvaro A. J Funςões de Exportaςão e de Importaςão para o Brasl. Pesqusa e Planeamento Economco 18 (3):