Tariff Rate Quotas with endogenous mode of competition: the case of the EU import regimes for bananas 1

Transcription

1 Tarff Rate Quotas wth endogenous mode of competton: the case of the EU mport regmes for bananas Margherta Scoppola Dpartmento d stud sullo svluppo economco Unverstà d Macerata Pazza Oberdan 3 62 Macerata (Italy) ph: fax: e-mal: scoppola@unmc.t January 29 Abstract: The paper develops a two-stage duopoly capacty constraned model, n whch the mode of competton s endogenous and the constrant s flexble, to nvestgate the mpact of Tarff Rate Quotas (TRQs) and to dentfy the tarff equvalent. The model predcts that the greater the gap between the cost of the lcences plus the n-quota tarff and the out-of-quota tarff, the closer the outcome of the game to the pure Cournot outcome. The tarff equvalent changes accordng to the prevalng mode of competton under the TRQ. The model s used to address the ssue of the tarffcaton of the non-acp TRQ for EU banana mports. The results suggest that under the TRQ frms competed on quantty and that the tarff equvalent s hgher than the tarff ntroduced by the EU n 26. Keywords: mport quotas, olgopoly, bananas JEL code: Q8; F3; L3. I wsh to thank Govann Anana for helpful comments on an earler draft of the paper. Fnancal support receved by the Agrcultural Trade Agreements (TRADEAG) (Specfc Targeted Research Project, Contract no ), and the New Issues n Agrcultural, Food and Bo-energy Trade (AGFOODTRADE) (Small and Medum-scale Focused Research Project, Grant Agreement no. 2236) research projects funded by the European Commsson, and by the European Unon polces, economc and trade ntegraton processes and WTO negotatons research project funded by the Italan Mnstry of Educaton, Unversty and Research (Scentfc Research Programs of Natonal Relevance 27) s gratefully acknowledged.

2 . Introducton A hgh proporton of agrcultural producton and mports n developed countres s currently protected by Tarff Rate Quotas (TRQs). TRQs are two level tarffs wth a lmted volume of mport (the quota) subject to a lower n-quota tarff and subsequent mport subject to a hgher out-of-quota tarff. TRQs are among the most frequently used trade polcy nstruments by developed countres to grant preferental access to products from developng countres. More than 5% of agrcultural products mported by the EU are covered by a TRQ, whle the fgure n the US and Japan s slghtly under %. Nevertheless, the extensve use of TRQs as a tool to lberalze agrcultural trade and to grant preferental access to products from developng countres has rased many questons and several countres wthn the current WTO negotatons have proposed modfcatons to the TRQs regmes to further lberalze trade. One controversal case s the European Unon (EU) TRQs regme for bananas ntroduced n 993. Ths regme has been at the heart of an endless nternatonal conflct - the so called banana war. Wthn the WTO major Latn Amercan banana exporters and the Unted States - whch does not produce bananas, but s the home country of the most mportant multnatonal banana traders challenged the regme (Joslng and Taylor, 23). Followng the 2 agreements reached wthn the WTO between the EU and the major Latn Amercan exporters, n 26 the EU replaced the TRQs system wth a tarff-only regme. Ths new regme was challenged agan wthn the WTO by major Latn Amercan exporters and by the US. The key ssue n the current controversy on bananas s that of establshng whch tarff would leave unchanged prevous mports under the TRQs. Papers on the economcs of TRQs assume, by and large, perfect competton (e.g. Boughner et al., 2; De Gorter and Kluaga, 26; Guyomard and Le Mouël, 23; Anana, 26). Whle ths assumpton reflects reasonably well many manufactured products, where mports and exports are mostly carred out by small frms, t may be unrealstc for agrcultural trade, whch s often domnated by few large frms. Ths paper consders the mplcatons of the exstence of large traders operatng n the nternatonal agrcultural trade; a duopoly capacty constraned model, n whch the mode of competton s endogenous, s used n order to examne the effect of TRQs and the mplcatons for establshng the tarff whch leaves mports unchanged. The lterature on the non-equvalence between quotas and tarffs n non compettve markets whch orgnated wth the contrbuton by Bhagwat (965), has consdered dfferent market structures and settngs. Contrbutons assumng olgopoly have shown that when frms compete on prce, strategc nteractons under quotas and tarffs may be fundamentally 2

3 dfferent (e.g. Harrs, 985; Krshna, 989).These papers, however, mplctly assume that frms behavour and market structure are not nfluenced by the trade polcy. 2 Few papers have analysed mport quotas under olgopoly by means of models n whch the conduct of frms on the market arses endogenously and s determned by the trade polcy. Syropoulos (992) constructed a model where frm behavour s endogenously determned n an nfntely repeated game and analysed the mpact of quotas and tarffs on market conduct. Deneckere et al. (2) developed a capacty constraned duopoly wth homogeneous products n whch the exstence of prce leadershp and the dentty of the leader are endogenous and depend upon frms costs and capactes. They show how conclusons regardng the equvalence of quotas and tarffs may dffer when endogenous conduct s taken nto consderaton. Capacty constraned models are the natural framework for examnng the mpact of quanttatve restrctons under olgopoly. Unlke other capacty constraned models (e.g. Kreps and Schenman, 983; Deneckere et al. 2), n the model developed n ths paper the constrant s not rgd, as the frm can ncrease the mported volume, although f at a hgher tradng cost. The basc dea s that the exstence of a TRQ ntroduces a capacty constrant for tradng frms whch s gven by the number of lcences they have been allocated. However, the frm can adjust ts capacty over tme by ncurrng adjustment costs: ether by acqurng addtonal lcences on the market n the frst stage, or by mportng out-of-quota n the second stage. Based on Magg (996), the model predcts that the greater the gap between the cost of adjustng capacty between the two perods, the hgher the effectveness of the capacty constrant and the closer the outcome of the game to that of Cournot. The replacement of the TRQ wth a tarff, by removng the capacty constrant, may mply a change n the mode of competton and generally leads to an ncrease n the degree of competton. As a consequence, the tarff whch leaves mports unchanged vares accordng to the prevalng mode of competton under the TRQ; n general terms ths s hgher the less compettve the equlbrum under the TRQ,.e. the more effectve the capacty constrant under the TRQ. Therefore, the model suggests that, n order to fnd the tarff whch would leave mports unchanged, t s essental to assess the prevalng mode of competton under the TRQ. The model provdes new nsghts nto the mpact of TRQs and of ther tarffcaton, by showng that two key varables - the out-of-quota tarff and the ntal allocaton of lcences - determne the mode of competton under the TRQ and, therefore, the value of the tarff whch leaves mports unchanged 2 A recent contrbuton by Krshna and Tan (27) analyses, n a compettve settng, the effect of quotas and tarffs consderng the mpact that trade polces may have on the sze of the tradng ndustry. 3

4 The model shows potental for emprcal purposes. Wthn ths framework the mode of competton can be assessed by means of observable parameters,.e. the values of adjustng costs n the two perods; a further advantage s that, whle trade polcy analyss based on the conjectural varaton approach assume a constant conduct parameter, here the possblty that a trade polcy change may change the mode of competton s taken nto account. A calbrated verson of the model s used to address the ssue of the tarffcaton of the EU TRQ on bananas n place untl 26 for non-acp countres. The prevalng mode of competton under the banana TRQ regme and the value of the tarff that would leave mports unchanged are assessed. The results show that, as n the case of multnatonals mports from non-acp countres, the prevalng mode of competton under the TRQ s Cournot for a wde range of parameter values; ths s because the effectveness of the capacty constrant s very hgh manly due to the hgh value of the out-of-quota tarff. The results also suggest that the tarff whch would leave mports unchanged s hgher than the tarff ntroduced by the EU n 26. The paper s organsed as follows. The next secton presents the man hypotheses of the model and the varous outcomes of the game. The thrd secton addresses the ssue of the tarff whch leaves mports unchanged under the dfferent outcomes. In the fourth secton the emprcal exercse on the tarffcaton of the EU TRQ for bananas s presented. The fnal secton offers some concludng remarks. 2. The model The model consders two symmetrc tradng frms mportng a dfferentated product n one country. We assume quadratc utlty functons and, thus, lnear demands. The nverse demand functon n the mportng country for each product s: p α βq λq2 = and p2 = α βq2 λq () wth q and p beng the quantty and the prce of the product traded by frm ; we assume that the own-prce effect domnates the cross-prce effect,.e. that β > λ. The drect demand functons are: q = a b p + b2 p2 and q 2 = a b p2 + b2 p (2) In the mportng country a tarff rate quota s n force wth Q, t and T beng, respectvely, the mport quota, the n-quota tarff and the out-of-quota tarff. 4

5 As for the allocaton of quota lcences, we assume that the government has allocated the quota on an hstorcal bass and that lcences can be transferred between operators. 3 Most studes to date have examned the case of aucton quotas (e.g. Bergsten et al., 987; Krshna, 993) and more recently of lcences-on-demand allocatons (Hraanova et al., 26). Unlke these methods, the hstorcal crtera allocate to mporters the rght to mport a gven quantty at the n-quota tarff wthout payng for the lcences. o In ths model the duopolst holds q lcences. In addton, a large number of small operators hold other lcences, for a total amount q, wth q + q + q = Q. 4 The margnal tradng costs s o o 2 of small operators are above those of large frms, thus, a market for lcences arses; small traders are assumed to behave compettvely both on the product and on the lcence markets. Ths settng s amed at reproducng the workng of the lcence market when lcences are allocated on the bass of hstorcal mports; over tme, traders who become more effcent are wllng to purchase lcences from the less effcent mporters. Small operators are n equlbrum when the prce of the lcence s equal to the value of usng t (.e, the product prce less the margnal costs). If small operators face margnal tradng costsc s = h+ fq s, wth q s beng the quantty mported by small traders and f >, then ther supply of lcences s: L= p ( h+ fq s ) (3) where p s the expected prce of the product. The volume mported by small operators s gven by the dfference between ther equlbrum quantty at the prce p, and the volume of lcences small operators sell to the duopolsts, d q : d p h d qs = qs( p) q = q (4) f By substtutng (4) n the supply of lcence (3) we obtan: L fq d = (5). We assume that the two large frms, whch due to ther low margnal tradng costs are the sole purchasers of lcences, exert market power n the lcence market. In ths settng, the s 3 Lcence transferablty s generally allowed as t s expected to mprove welfare - because lcences are lkely to be fnally used by the lower margnal cost frms - although the common vew that transferable lcences are always welfare mprovng has been questoned n the lterature (e.g., Lott, 987; Krshna and Tan, 999). 4 The ntal allocaton of lcences to frms s exogenous and rent-seekng actvtes to obtan quota rents are not consdered. Hraanova and De Gorter (25) have recently shown how rent seekng by agrcultural tradng frms may affect the lcences market structure. 5

6 duopsonsts are n equlbrum on the lcence market when ther resdual margnal revenue, net of other costs, s equal to L' = L\ q. We assume that L ' + t T. The two large frms face dentcal margnal tradng costs, c, whch are assumed to be constant. Profts of frm are: pq cq tq f q q Π = + < + d d d pq cq Lq t( q q ) f q q q q d d d d pq cq Lq t( q + q ) T( q q q ) f q > q + q As already mentoned, n the frst stage of the game the two frms choose capacty,.e. they decde the quantty to trade; f ths quantty exceeds ther own quota, frms purchase lcences on the market. By choosng capacty, frms commt themselves to mport a certan volume n the second stage. In the second stage we assume that the only way to ncrease mports s to mport out-of-quota. Ths assumpton appears to be realstc n the context of several TRQ regmes wth transferable lcences; lcence transactons usually occur well before the entry of the product nto the mportng country, as frms plan the quantty to trade n advance n order to arrange contracts wth supplers, dstrbutors and provders of logstcal servces. The amount of purchased lcences could thus work as a frm s commtment to trade a certan quantty n the second perod. The effectveness of ths commtment depends on the cost of ncreasng capacty n the second stage, that s, the cost of mportng out-of-quota, whch s tarff T. The hgher T and the hgher s the effectveness of the capacty commtment; on the contrary, f the cost of ncreasng capacty n second stage s equal to that ncurred by frms n the frst stage ( L ' + t = T ), then the capacty chosen n the frst stage does not work as a commtment devce. In the second stage, after observng capacty, frms compete on prce. As Magg (996) has shown, n ths settng the prevalng mode of competton and, consequently, the outcome of the game, depends upon the effectveness of the capacty commtment; as the effectveness of the commtment ncreases, the game moves from a pure Bertrand soluton - when the capacty constrant s not mportant, frms compete on prces - to a pure Cournot outcome - when the capacty constrant s very mportant, frms commt themselves to mport the quantty chosen n the frst stage. The prevalng mode of competton, therefore, depends upon the values of t, T, and q o, that s, t depends upon the value of the key nstruments of the trade polcy. As usual, the game s solved backwards, by frst consderng the second stage decson, when frms compete on prce, gven the capacty chosen n the frst stage. The subgame Bertrand o d reacton curve s a knked curve wth three branches. If q < q + q margnal costs are equal (6) 6

7 to c + t. The frm s mports are lower than the lcences t holds (.e. ts own lcences plus the d o lcences acqured n the frst stage). If q > q + q margnal costs of frms are equal to c + T. The frm s mports are greater than the amount of lcences t holds; therefore there are also o d out-of-quota mports. If q = q + q, the resdual margnal revenue curve crosses the margnal cost curve at ts vertcal segment. The relevant prce combnaton s derved from the demand o d functons by settng a fxed traded quantty q = q + q. Ths subgame admts a unque purestrategy equlbrum, gven by the ntersecton of the two reacton curves. The subgame perfect equlbrum mples that (Magg, 996): B p ( ct, ) f L' + t T< T* p = C p ( c, t, L) f T > T* wth T * beng a crtcal value of the out-of-quota tarff, and p B e p C beng the Bertrand and Cournot equlbrum prces, respectvely. Wth no capacty constrants ( L' + t = T), the full game equlbrum s the outcome of a oneshot Bertrand game wth long run costs equal to (7) c+ t+ L'. When the capacty constrant s effectve (T s hgher than the crtcal level T *) then the full game equlbrum s gven by the Cournot prce.the Cournot soluton s dentfed as the outcome of a one-shot game n whch frms compete on quantty and have long run costs equal to c+ t+ L' (Fgure ). Fgure When the value of the out-of-quota tarff s lower than the crtcal level, but greater than the frst stage capacty costs ( L' + t < T < T* ), the equlbrum prce s the outcome of a one-shot Bertrand game n whch frms compete on prce and have long run costs equal to c + T s the equlbrum prce even f frms do not mport out-of-quota and ncur margnal costs c+ t+ L', whch are lower than c + T. The basc ntuton explanng ths outcome s that when the cost of ncreasng capacty n the frst stage s lower than n the second stage B ( L' + t < T), nvestment n capacty n the frst stage sustans a hgher prce ( p (, ct )) than B the one that would preval n a pure Bertrand one shot game ( p ( ctl,, ') (Magg, 996). The crtcal level T * above whch the mode of competton swtches from Bertrand to B C Cournot s defned mplctly by p (, ct) = p (,, ct L'). As for the Bertrand prce holdng fxed the rval s prce: B p, the reacton curves may be obtaned by maxmsng (6) by. Ths 7

8 max Π q (8) = pq ( c + T ) p Frst order condtons are: 2 2 a 2 b p + b p + b( c+ T) (9) whch yelds the reacton curve: p a + b p + b ( c + T ) 2 2 = () 2b Because frms are symmetrc, the reacton curve of frm 2 s dentcal to (). The Bertrand prce equlbrum p B ( c+ T) s: a + b ( c + ) p T B B = p2 = 2b b 2 The Cournot reacton curves are obtaned by solvng the maxmsaton problem: max Π = qp ( c+ tq ) Lq (2) d q gven that d o q = q q, frst order condtons are: α 2 βq λq ( c + t) 2fq + 2fq = (3) whch yeld: q = Gven that: 2 α λq ( c+ t) + 2 fq 2( β + f ) ( b b2 ) 2 2 b2 2 a b b2 α =, β = and λ = (5) b b b b b 2 reacton curves n the prces space can be obtaned by substtutng n equaton (4) demand C functons (2) and parameters (5). The Cournot equlbrum prce p ( ctl,, ') s: C C Δ+Φ ( c+ t 2 fq ) p = p2 = Σ where ΔΦΣ>,, depend upon parameters of the demand and of margnal costs functon of small operators. 5 B C When parameters satsfy all constrants and T < T *, p < p whch B C mples q > q. 2 The value of T * determnes the equlbrum under the TRQ. As already mentoned, T * s B C dentfed as the tarff satsfyng the dentty p ( ct, ) = p ( ct,, L'). From () and (6) we obtan: 5 More detals are ncluded n the Appendx. () (4) (6) 8

9 Ψ+ cκ+ ( t 2 fq ) ϒ T* = Ν (7). wth ΨΚϒΝ>,,, dependng upon demand and margnal cost functons parameters. The mode of competton n ths model, thus, depends upon the key varables of the trade polcy: the ntal allocaton of lcences between small and large mporters and the values of the n-quota and out-of-quota tarffs. Equaton (7) shows how the larger q, the lower T *, the hgher the probablty of a Cournot outcome. The reason s that a hgh share of lcences allocated to large traders dmnshes the amount of lcence transactons, reduces the prce of lcences and, therefore, ncreases the gap between the costs of adjustng capacty between the two perods, that s, the capacty constrant. Hence, the model suggests that, f the share of lcences allocated to large traders s suffcently hgh, then the capacty constrant wll be stronger and the equlbrum on the product market wll be Cournot. If ntally less effcent traders hold a large share of lcences, an ncrease n the number of lcences allocated to more effcent operators may shft the mode of competton from Bertrand to Cournot. Ths result adds a further argument to the dea - already developed n the lterature on mport quotas (e.g., Krshna et al. 994) - that competton n an ndustry may decrease wth the concentraton of lcences; n ths model ths occurs because hgh concentraton ncreases the effectveness of the duopolsts capacty constrant. The value of the n-quota tarff postvely affects T * : a hgh n-quota tarff reduces the effectveness of the capacty commtment and, thus, the probablty of a Cournot outcome. As for the out-of-quota tarff, f ths s hgher than the crtcal value T * then the capacty commtment s effectve and frms compete on quantty, whle Bertrand prevals n the opposte case. Even when the out-of-quota tarff s prohbtve and there are no out-of-quota mports, the value of ths tarff nfluences the equlbrum under the TRQ. Ths result dffers from the predcton of perfect compettve statc models; n the latter, when there are no outof-quota mports, the value of the out-of-quota tarff never nfluences the equlbrum under the TRQ (Boughner et.al, 2; De Gorter and Kluaga, 26). 3. The tarff equvalent of TRQs wth endogenous mode of competton A change n the trade polcy nstrument may affect the mode of competton. In ths model a tarff means that there s no capacty constrant; frms n the frst stage can ncrease ther capacty by sustanng the same cost as n the second stage,.e. by payng the tarff. In other 9

10 words, a tarff represents the case n whch L =, T = t and thus L' + t = T. Under these crcumstances, the outcome of the game s the one-shot Bertrand soluton. If t s the tarff n force n the tarff-only regme, then the new equlbrum prce s: p Bt p a + b ( c + t ) Bt = 2 = (8) 2b b2 The mpact of the tarffcaton may dffer accordng to the equlbrum prevalng under the TRQ regme whch depends upon the values of the n-quota tarff, the out-of-quota tarff and the ntal allocaton of lcences. Let us assume that demand and cost condtons are such that frms do not mport out of the quota Q, even though they mport more than the quota they hold and, thus, they purchase lcences from small operators at prce L. We can dstngush two dfferent cases. Case a): If under the TRQ regme L ' + t T < T*.e. there s a weak capacty constrant before tarffcaton, the outcome of the game s dentfed by the one-shot Bertrand game wth costs equal to c+ T (equaton ). The tarff whch leaves quanttes unchanged s the tarff whch B B Bt Bt satsfes q ( p ) = q ( p ). From () and (8) t s clear that ths s equal to the out-of-quota tarff T. Ths s the tarff that would leave unchanged the prce as well. Wth mport quotas, when frms compete on prce, tarffs and quotas are not equvalent, as the rgd quanttatve restrcton sustans a hgher prce than the one that would preval wth a tarff mplyng the same level of mports (Harrs, 985; Krshna 989). However, the quanttatve restrcton due to the TRQ s flexble. For each level of mports t s possble to fnd a tarff equvalent whch s the sum of the prce of the lcence, the value of the n-quota tarff and the olgopsonstc rent of frms. Ths s shown n Fgure 2 whch represents the Bertrand equlbrum of Frm under the TRQ regme when the frm purchases lcences on the market ( o q q > ) and when L ' + t = T < T*,.e. there s no capacty constrant. Small operators beneft from the quota rent whch s equal to the prce of the lcences pad by the large frm; the large frm benefts from olgopsonstc and olgopolstc rents and from the quota rents generated by ts own lcences. After tarffcaton frms contnue to compete à la Bertrand: the tarff e t ncludes the n-quota tarff, but also the quota rent of small operators and the olgopsonstc rent. Fgure 2

11 Case b): If under the TRQ regme T > T *,.e. there s a strong capacty constrant, before tarffcaton C the outcome of the game s the Cournot prce p (,, ctl '). In ths case tarffcaton nduces a change n the mode of competton; by shftng the outcome from Cournot to Bertrand t ncreases the degree of competton on the market. The tarff t e whch leaves quanttes the C Bt same as before tarffcaton s the one whch satsfes q = q whch mples Bt e p (,, ctl') = p (, ct), that s, the tarff whch has been defned as the crtcal tarff T *. C Therefore, when the out-of-quota tarff s suffcently hgh, the tarff t e s the crtcal tarff (equaton 7) above whch frms compete on quanttes. Ths tarff captures the n-quota tarff, the prce of the lcence, the olgopsonstc rent and part of the olgopolstc rent under the TRQ regme. Ths s because tarffcaton, by shftng from Cournot to Bertrand, mples a decrease n the olgopolstc rent of frms. These results have a number of nterestng mplcatons. The tarff t e changes wth the prevalng mode of competton under the TRQ. Therefore, when addressng the ssue of the tarffcaton of TRQs one must beforehand analyse the prevalng mode of competton between frms under the TRQ whch ultmately depends upon the value of the out-of-quota tarff wth respect to T * and L ' + t. As aforementoned, one mportant factor s the ntal dstrbuton of lcences to small and large operators; a hgh proporton of lcences allocated to small operators ncreases the prce of lcences and reduces the probablty of a Cournot outcome and the tarff equvalent beng T *. The out-of-quota tarff also nfluences the value of the tarff equvalent, even f there are no out of-quota-mports. When T s suffcently hgh ( T > T* ) then the tarff equvalent s T *, whle f T < T* the tarff equvalent s exactly the value of the out-of-quota tarff. 4. The case of the EU mport regmes for bananas A calbrated verson of the model s used here to show how the mode of competton under the TRQ and the tarff equvalent can be emprcally assessed wthn ths framework. The case of the EU mport regme for bananas represents an llumnatng example of how the ssue of the tarff equvalent of TRQs may well become a relevant one at the nternatonal level. Before 26 two TRQs were n force, one for ACP countres and the other one for non-acp countres; whle the ACP n-quota tarff was zero, a postve n-quota tarff was appled to non-acp countres. The out-of-quota tarffs were very hgh and n the whole perod durng

12 whch the TRQ regme was n force, there were vrtually zero out-of-quota mports. Lcenses, whch were allocated through the hstorcal allocaton crtera, were transferable, albet wth certan restrctons, among operators. Ths regme was at the heart of a long standng trade dspute wthn the WTO (Joslng and Taylor, 23). In 26 the EU replaced the non-acp TRQ wth a tarff whle mantanng a duty-free mport quota for ACP bananas. The new regme was challenged agan wthn the WTO by major Latn Amercan exporters and the US. A formal complant by Ecuador n November 26 was followed by one from Colomba, Panama and the US n 27 6 (WTO, 26, 27a, 27b and 27c). Many papers have assessed the mpact of the EU mport regmes for bananas by assumng perfect competton (e.g. Guyomard et al. 999; Guyomard and Le Mouël, 23; Anana, 26), although varous authors acknowledge the fact that the concentraton rato n ths ndustry s very hgh 7 and that the gap between the exportng countres FOB prces and the EU CIF mport prces s exceptonally hgh. There are few emprcal studes on the degree of market power n the banana tradng ndustry whch are rather old and provde contrastng results (Deodhar and Sheldon, 995; Herrmann and Sexton, 2). Few papers have assumed mperfect competton: Mc Corrston and Sheldon (996) and Mc Corrston (2) have developed a vertcally-related markets model n whch the extent of mperfect competton s assessed by means of the conjectural varaton approach. All studes perform comparatve statc under the assumpton that everythng remans constant after the trade polcy change, and thus gnore possble changes n frms behavour, although Deodhar and Sheldon (995) have found evdence that the ntroducton of the TRQs caused the ndustry to behave n a Cournot manner, thereby stressng the fact that trade polcy changes may affect the mode of competton. Unlke the conjectural varaton approach, the mode of competton s here assessed on the bass of observable parameters,.e. the cost of adjustng capacty n the frst and second perod; comparatve statc s performed by allowng the mode of competton to change wth the trade polcy, whle n studes usng the conjectural varaton approach the mode of 6 In December 27 and May 28 two WTO panels ruled aganst the EU tarff regme. 7 More than 5% of EU banana mports are concentrated n the hands of two frms, Dole Food and Chquta, whch dfferentate ther product by the means of brands; these large frms are vertcally ntegrated throughout the tradng channel from the exportng countres to the EU market; they perform shppng, transportaton and also rpenng n the EU and generally sell the product drectly to EU wholesalers (Taylor, 23; UNCTAD, 23). 2

13 competton s assumed to be constant; a further advantage s that ths framework, dfferently from the conjectural varaton model, s game-theory founded (Magg, 996). 8 The assumpton of a symmetrc duopoly wth dfferentated products does not appear to be too unrealstc for the EU market for non-acp bananas. Two frms, Dole Food and Chquta, have domnated EU mports from non-acp countres and account for smlar market shares n the EU and world-wde. Both frms have been nvestng over tme n product dfferentaton strateges (Read, 986; Taylor, 23) and sell the product on the market wth a well known brand. Ther tradng costs are roughly smlar as both of them export manly from neghbourng Latn Amercan countres (Costa Rca, Panama and Colomba) and export to the EU through a full ntegrated marketng chan (UNCTAD, 23; Taylor, 23). Untl 26, several small hstorcal mporters used to sell ther lcences to the two large frms. One of the man emprcal dffcultes s that ths model requres frms data whch are not always avalable. The problem of the lack of data has been dealt wth n dfferent ways. As for frms, mports nto the EU from non-acp countres have been assessed on the bass of the Chquta s quarterly company reports for the year 25 and all other avalable nformaton. 9 Unlke other papers (e.g. Guyomard et al. 999; Guyomard and Le Mouël, 23; Anana, 26), the EU wholesale prce, nstead of the CIF prce, has been consdered. Ths s because the CIF prce s lkely to reflect ntra-frm transactons and, therefore, to be a transfer prce, often manpulated by frms for fscal purposes. The wholesale prce of Chquta bananas n the EU market s from FAO (26). Frms costs have been obtaned by addng up the varous costs the frm ncurs when tradng the product from the Latn Amercan supplers to the EU market. These nclude the FOB prce of the countres Chquta exports from (.e. Panama, Costa Rca, Colomba), the nsurance, freght and other relevant costs from these supplers to the EU borders and the cost of rpenng and transportaton n the EU markets. The model has been calbrated usng a range of values of prce elastctes, correspondng to dfferent values of the degree of brand dfferentaton. As for the lcence market, the prce of the non-acp mport lcence s the one reported n recent years n the specalsed press; a set of ntal lcence allocatons has been consdered, rangng from the hypothess that all 8 Crtcsms of the use of the conjectural varaton model n trade polcy analyss can be found, among others, n Helpman and Krugman (989). 9 Chquta reports the amount of mports n Europe; ths ncludes mports comng from all sources (both ACP and non-acp mports) and sold n all European countres (both EU and non EU). Ths fgure has been corrected by subtractng mports of Chquta from the Ivory Coast and mports of Chquta nto non EU countres. Both fgures were estmated on the bass of market shares of Chquta avalable n UNCTAD (23) and data on countres mports/exports provded by FAOSTAT. 2 λ The degree of brand dfferentaton s defned as ρ =. 2 β 3

14 d lcences are allocated to large operators ( q = ), to the opposte one ( q = ). Detals on the nput data are reported n Table A of the Appendx. Fgure 3 Fgure 3 shows the value of the crtcal tarff as a functon of the degree of brand dfferentaton. For values of the degree of brand dfferentaton below.45 the crtcal tarff s below the out-of-quota tarff and the outcome of the game s the Cournot one; thus, the tarff equvalent s the crtcal tarff. Ths result s consstent wth the conclusons reached by Deodhar and Sheldon (995) who found emprcal evdence of quantty competton under the TRQ regme n the md nnetes. Above ths value for the degree of brand dfferentaton, the mode of competton s Bertrand and the tarff equvalent s the out-of quota tarff T. Fgure 4 Fgure 4 reports the crtcal tarff as a functon of the share of lcences held by large traders. If ths s low (lower than 5% n the Fgure), the amount of lcences purchased, and ther prce, s hgh and the capacty constrant weak; the outcome of the game s Bertrand and the tarff equvalent s the out-of-quota tarff. On the contrary, when the duopolst holds a sgnfcant share of lcences, then the lcence prce s low, the capacty constrant becomes effectve and frms compete on quantty; the tarff equvalent n ths case s the crtcal tarff. Overall, the Fgures show that the tarff equvalent s, by and large, hgher than the value fnally set by the EU (76 /t equal to 22 US$/t at the 26 average exchange rate) and ths holds for a wde range of elastcty values. Ths result dffers from prevous papers whch suggest lower tarff equvalents (e.g. Guyomard and Le Mouël, 23; Anana, 26). Such dfferences are explaned manly by two factors: a) unlke other studes, ths paper takes nto consderaton olgopsonstc and olgopolstc rents; b) ths paper takes nto account the change n the mode of competton among frms and ths makes the tarff equvalent hgher than the one predcted by models assumng perfect competton. The emprcal exercse also shows that olgopolstc and olgopsonstc rents account for more than 7% of the EU prce, whle frms costs, ncludng the prce of the lcence, for less than 3% (Table A2 of the Appendx). Thus, a major part of the observed gap between FOB and EU domestc prce may be due to these rents. 5. Conclusons Ths paper has used a two stage capacty constraned duopoly model, n whch the mode of competton s endogenously determned and the constrant s flexble, n order to nvestgate 4

15 the workng of TRQs and ther removal. The results emphasze the role played by two varables the out-of-quota tarff and the ntal allocaton of lcences n determnng the value of the tarff equvalent. Unlke prevous studes on TRQs, the out-of-quota tarff n ths model, by nfluencng the strategc nteracton between frms, s crucal n determnng the tarff equvalent even when the out-of-quota tarff s prohbtve and there are no out-of-quota mports. As for the lcences allocaton, ths paper has consdered the case of the hstorcal allocaton methods wth a market for lcences, under the assumpton that large traders exert olgopsony power and that the supply of the lcences s upward slopng. Under these crcumstances, the prce of the lcences s lower the hgher ther concentraton n the hands of large traders. Thus, the model suggests that a low concentraton of lcences n the hands of large traders prevents Cournot competton. The model also shows that the tarff equvalent changes wth the mode of competton under the TRQ: ths wll become hgher the lower the degree of competton under the TRQ. Fnally the theoretcal framework proposes an approach to assess the mode of competton whch s based on observable varables,.e. the costs of adjustng capacty n the frst and n the second stage: n the specfc case of TRQs, the mode of competton can be assessed on the bass of the prce of lcences and of the nquota and out-of-quota tarffs. The case of the non-acp TRQ regme for banana mports to the EU s an example of a prohbtve out-of-quota tarff whch, by ncreasng the effectveness of the capacty commtments, has nfluenced frms behavour; the emprcal exercse suggests that the tarff equvalent s hgher than the one suggested by other studes, possbly because t takes nto account olgopolstc and olgopsonstc rents. The tarff equvalent s also much hgher than the tarff ntroduced by the EU n 26; the model, thus, predcts that mports from non-acp countres should be well above (and the prce well below) the 25 levels. Ths could help to explan the sgnfcant observed ncrease n mports from non-acp supplers snce 26. The assumpton of perfect competton, when markets are domnated by large traders, may thus result n an underestmaton of the TRQ tarff equvalent. The emprcal use of the model, although prelmnary, has shown ts potental when addressng ssues whch tradtonal perfect competton statc frameworks tend to be unable to explan satsfactorly. The gap between the mport and export prce s here explaned by the consderable olgopolstc and olgopsonstc rents whch account for about 7% of the market prce. In concluson, the efforts to take nto consderaton the role of large traders n the emprcal analyss of trade polcy may, n many crcumstances, contrbute to a better overall understandng of the phenomena n queston. 5

16 References Anana, G., 26. The 25 WTO Arbtraton and the New EU Import Regme for Bananas: A Cut too Far? European Revew of Agrcultural Economcs 33, Baghwat, J.N., 965. On the Equvalence of Tarffs and Quotas. In: R. E.Baldwn (ed), Trade Growth and the Balance of Payments: Essays n Honour of Gottfred Haberler, 53-67, Chcago. Bergsten, C.F., Ellott, K.A., Schott, J.J., Takacs W.E Aucton Quotas and Unted States Trade Polcy. Polcy Analyses n Internatonal Economcs No. 9. Insttute of Internatonal Economcs, Washngton DC. Boughner, D., de Gorter, H. and I. Sheldon. 2. The Economcs of Tarff-Rate Quotas n the Agrcultural Agreement n the WTO, Agrcultural and Resource Economcs Revew 2, De Gorter, H., Kluaga, E., 26. Reducng Tarffs versus Expandng Tarff rate Quotas. In: K.Anderson, W.Martn (eds), Agrcultural Trade Reform and the Doha Development Agenda, Palgrave Macmllan and the World Bank, Washngton DC. Deneckere R.J., Kovenock D. and Sohn Y.Y, 2. Quota and Tarffs wth endogenous conduct. Industral Organzaton 9, Deodhar, S.Y., Sheldon, I.M., 995. Is Foregn Trade (m)perfectly compettve? An analyss of the German market for banana mports. Journal of agrcultural economcs 46, FAO (26) Banana statstcs 25, Rome: FAO Guyomard, H., Laroche, C. and Le Mouël C Impacts of the Common Market Organzaton for Bananas on the European Unon Markets, Internatonal Trade, and Welfare. Journal of Polcy Modelng 2, Guyomard, H., Le Mouël, C., 23. The New Banana Import regme n the European Unon: A Quanttatve Assessment. The Estey Centre Journal of Internatonal Law and Trade Polcy 4, Harrs, R., 985. Why Voluntary Export Restrants are Voluntary. Canadan Journal of Economcs, 8, ; Helpman, E., Krugman, P., 989. Trade Polcy and Market Structure, the MIT Press, Cambrdge Mass. Herrmann, R., Sexton, R. J., 2. Market Conduct and the Economc Impacts of A Tarff-Rate Quota Polcy: The European Banana Case In: Moss, C., Rausser, G.B., Schmtz, A., Taylor, T. and Zlbermann, D. (eds) Agrcultural Globalzaton, Trade and the Envronment, Kluwer Academc Press, Dordecht, pp Hraanova, J., De Gorter H., 25. Rent Seekng wth Poltcally Contestable Rghts to Tarff-rate Import Quotas. Revew of Internatonal Economcs 3, Hraanova, J., De Gorter H., Falck J., 26. The Economcs of Admnsterng Import Quotas wth Lcence-on-demand n agrculture. Amercan Journal of Agrcultural Economcs 88,

17 Joslng, T. E., Taylor, T. G. (eds) 23. Banana Wars. The Anatomy of a Trade Dspute. CABI Publshng, Wallngford. Kreps, D., Schenkman J., 983. Quantty Precommtment and Bertrand Competton Yeld Cournot Outcomes. Bell Journal of Economcs 4, Krshna, K., 989. Trade Restrctons as Facltatng Practces. Journal of Internatonal Economcs, 26, 25-7 Krshna, K., 993. The Case of the Vanshng Revenues: Aucton Quotas wth Olgopoly, In: Krshna, K., Theory, polcy and dynamcs n nternatonal trade: Essays n honor of Ronald W. Jones, Cambrdge Unversty Press, Cambrdge, New York and Melbourne, Krshna, K., Erzan R., Tan, L.H., 994. Rent Sharng n the Mult-Fbre Agreement: Theory and Evdence from US Apparel Imports from Hong Kong.Revew of Internatonal Economcs, 2, Krshna, K., Tan, L.H., 999. Transferable Lcences versus nontrasferable lcences: what s the dfference? Internatonal Economc Revew, 4, Krshna, K., Tan, L.H., 27. Trade Polcy wth heterogeneous traders: do quotas get a bum rap? Workng paper 34, Natonal Bureau of Economc Research, Cambrdge MA. Lott J. R., 987. Lcencng and non-transferable rents. Amercan Economc Revew 77, Magg, G., 996. Strategc Trade Polces wth Endogenous Mode of Competton. Amercan Economc Revew 86, Mc Corrston, S., 2. Market Structure Issues and the Evaluaton of the Reform of the EU Banana Regme. World Economy. 23, Mc Corrston, S., Sheldon, I., 996. The Effects of Vertcal Integrated Markets on Trade Polcy Reform, Oxford Economc Papers. 48, Read, R., 986. The banana ndustry. In:M. C. Casson (ed) Multnatonals and World Trade. Vertcal Integraton and the Dvson of Labour n World Industres. Allen & Unwn, London. Syropoulos, C., 992. Quanttatve restrctons and tarffs wth endogenous frm behavour. European Economc Revew 36, Taylor, T.G., 23. Evoluton of the Banana Multnatonals. In: T.E Joslng, T.G. Taylor, Banana Wars. The Anatomy of a Trade dspute, CABI Publshng, Cambrdge MA. UNCTAD, 23. Major developments and recent trends n nternatonal banana marketng structures. Study prepared by the UNCTAD secretarat, UNCTAD/DITC/COM/23/, 24 November 23 WTO, 27a. European communtes Regme for the mportaton of bananas. Request for consultatons by Colomba, WT/DS36/G/L/88, 26 March, Geneve. WTO, 27b. European communtes Regme for the mportaton, sale and dstrbuton of bananas. Recourse to Artcle 2.5 of the DSU by the Unted States. WT/DS27/83, 2 July, Geneve. 7

18 WTO, 26. European communtes Regme for the mportaton, sale and dstrbuton of bananas. Recourse to Artcle 2.5 of the DSU by Ecuador. WT/DS27/65/Rev., 29 November, Geneve. WTO, 27c. European communtes Regme for the mportaton of bananas. Request for Consultatons by Panama. WT/DS364/ G/L/822, 27 June, Geneve 8

19 p 2 p C (c,t,l ) p B (c,t) p B (c,t,l ) p Fgure : The full game p L p B c+t+l =T Olgopolstc rents L Quota rents frm Olgopsonstc rents c+t+l c+t Quota rents small operators t e =T MR(p 2 B ) D(p 2 B ) c q o q B q Fgure 2: Tarff equvalent and rents under Bertrand 9

20 6 4 Cournot Bertrand T* 2 T*,T 8 6 T Degree of dfferentaton Fgure 3: The tarff equvalent as a functon of the degree of dfferentaton per Ed= e ES=2 9 8 T T*, T 7 Bertrand Cournot 6 T* % lcences held by large operators Fgure 4: The tarff equvalent as a functon of lcences held by large operators 2

21 Appendx The Cournot prce The reacton curves n the prce space are obtaned by substtutng demand functons (2) and parameters (5) n (4): q = α λq ( c+ t) + 2 fq 2 2( β + f ) α λ( a b p + b p ) c e+ 2fq a b p + b p = 2 2 2( β + f ) 2 2 α ab ( + b) λ b 2 2 Gven that = and = β + 2f 2( b + f( b b ) 2β + 2f 2( b + f( b b ) 2 we obtan the followng reacton curve n the prce space: ( b + Z) a+ ( b b )( c+ e 2 fq ) + b p ( b + Z) = 2 2 2b b2 + bz p 2 wth Z = f b b ( ) 2 The Cournot prce s: C C Δ+Φ ( c+ t+ e 2 fq ) p = p2 = Σ wth Δ= ( b + Z) a, Φ= ( b b ) and Σ = 2 b b + bz b ( b + Z) Gven that b > b 2, Z, ΔΦand, Σ are all greater than zero. The crtcal tarff B C The crtcal tarff s dentfed as the tarff satsfyng the dentty p ( ct, ) = p ( ct,, L'). From () and (6) we obtan: ab ( 2 + bz ) + cb ( b2)( b2 + bz ) + ( t 2 fq)( b b2)(2 b + bb 2 b2 ) T* = b( b b )(2 b + b + Z) whch can be wrtten as : 2 2 Ψ+ cκ+ ( t 2 fq ) ϒ T* = Ν 2 wth Ψ= ab ( + bz), 2 Κ= ( b b )( b + bz), ϒ= ( b b )(2 b + bb b ) and Ν= b( b b )(2 b + b + Z) 2 2 Gven that b > b 2, ΨΚϒΝ,,, are all greater than zero. 2

22 Table A. Input data (US$/ton, tons ) Fgure 3 Fgure 4 Table A2 Calbraton: mn max mn max Prce elastcty Cross prce elastcty Degree of dfferentaton Lcences allocated to large frms (q ) 475, , , ,773 Purchased lcences (q d ) 4, 4, 875,773 4, Quantty mported by the frm Wholesale prce Prce of the lcence (L ) Other data Tradng costs (c ) In quota tarff (t ) Out-of-quota tarff (T ) 875,773, Table A2: Prce, quanttes and rents under the TRQ Crtcal tarff (T* ) 246 Cournot prce (p c ) 2,74 Cournot quantty (q c ) 723,976 Lcence prce (L ) 47 Olgopolstc rent,488 Olgopsonstc rent 47 22