Kyle Bagwell and Robert W. Staiger. Revised: November 1993

Transcription

1 Multilateral Tariff Cooperation During the Formation of Regional Free Trade Areas* Kyle Bagwell and Robert W. Staiger Northwestern University The University of Wisonsin and NBER by Revised: November 1993 * We have reeived helpful omments from Phil Levy, Ron MKinnon, and seminar partiipants at The University of Western Ontario International Trade Conferene, Prineton University, and The University of Pennsylvania. Staiger gratefully aknowledges finanial support from Stanford's Center for Eonomi Poliy Researh, and as an Alfred P. Sloan Researh Fellow.

2 Abstrat We explore the impat of the formation of regional free trade agreements on the ability of ountries to maintain low ooperative multilateral tariffs. We assume that ountries an not make binding international ommitments, but are instead limited to self-enforing arrangements. Speifially, we model ooperation in multilateral trade poliy as involving a onstant balane between, on the one hand, the gains from deviating unilaterally from an agreed-upon trade poliy, and on the other, the disounted expeted future benefits of maintaining multilateral ooperation, with the understanding that the latter would be forfeited in the trade war whih followed a unilateral defetion in pursuit of the former. In this ontext, we explore the way in whih the formation of regional free trade agreements upsets the balane between urrent and future onditions, and trae through the dynami ramifiations of these effets for multilateral ooperation. Our results suggest that the emergene of regional free trade areas will be aompanied by a temporary retreat from liberal multilateral trade poliies. Eventually, however, as the full impat of the emerging free trade agreement on multilateral trade patterns is felt, the initial balane between urrent and expeted future onditions tends to reemerge, and liberal multilateral trade poliies an be restored.

3 I. Introdution The reent move toward further regional integration embodied in events ranging from EC1992 to the signing of the North Amerian Free Trade Agreement (NAFTA) to the elimination of tariffs among the Merursor Countries of South Ameria has generated intense interest in the ramifiations of regional trade agreements for the prospets of ontinued multilateral trade ooperation. Sine its ineption, the General Agreement on Tariffs and Trade (GATT) has maintained a somewhat uneasy oexistene with regional exeptions to its Most Favored Nation (MFN) priniple through Artile XXIV whih permits the formation of both free trade areas and ustoms unions (whih also have ommon external tariffs) subjet to ertain stipulations. But the reent proliferation of suh agreements, with the United States in partiular now apparently viewing negotiation of free trade agreements under Artile XXIV as an attrative element of its trade poliy, has triggered the onern that Artile XXIV might be a loophole to the priniple of MFN through whih regional trade agreements ould underut multilateral ooperation. From a historial perspetive, this onern appears unfounded, at least judging from the experiene with the formation and later broadening of the European Community (EC) ustoms union and its impat on multilateral ooperation. Far from underutting enthusiasm for further multilateral liberalization, the future prospet of an integrated EC market devoid of internal barriers but with ommon external tariffs appears to have been a major fator leading to the Kennedy Round of multilateral GATT negotiations initiated in For example, in testimony before the Senate Finane Committee on the Trade Expansion At of 1962 (whih provided U.S. negotiating authority for the Kennedy Round), then Seretary of 1

4 Commere Luther Hodges testified that:...it is generally aepted that if we have to pay the sheduled ommon external tariff rates of the Common Market while our European ompetitors go duty-free, we stand to lose a substantial amount in annual sales we would otherwise be able to make to the expanded EEC market...to avoid this possibility, Mr. Chairman, and gentlemen, the Common Market must be enouraged to implement its announed poliy of liberal trade by making substantial redutions in its external tariffs...there is only one way to aomplish this. We must negotiate a new trade agreement with the Common Market ountries. (p. 34) Similarly, former Seretary of State Christian Herter testified in front of the Joint Eonomi Committee that:...if we are to go in one diretion and Europe go in the other, inevitably, you will find trade barriers growing as between two large free trade areas. With these trade barriers growing, you would find...the slowing down of trade, both imports and exports...so what is the alternative in the piture? The alternative, to my mind, is to reonile our poliies with those of Europe, with a view to inreasing trade on both sides...[p. 12]. Impetus for the Tokyo round of multilateral GATT negotiations initiated in 1974 an be similarly linked in large part to EC enlargement to inlude the United Kingdom and others. Indeed, Bhagwati (1991) has observed that the pereption in the 1960s was one of a general ompatibility between regionalism and GATT, and that only reently did the view that regionalism might be antithetial to GATT gain prominene. Whether or not this newly pessimisti view turns out to be orret, it has sparked a resurgene in the study of regional trade arrangements. Building on the lassi ustoms union analysis of Viner (1950) and the insights of Kemp and Wan (1976), Krugman (1991a,b) has explored the welfare onsequenes of ustoms-union formation when blos produe differentiated produts for both loal onsumption and export, and exploit their power to affet world pries by setting optimal tariffs on the imports of differentiated goods from other blos. Krugman's analysis highlights a basi tradeoff that an emerge when 2

5 welfare is ompared aross equilibria as ostums-union formation proeeds, i.e., as trading blos beome larger in size and fewer in number: The inreasing size of eah blo allows a greater variety of produts to be available duty-free to onsumers within the blo, but the dereasing number of blos leads to higher optimal tariffs on trade among blos. 1 As a result, the welfare onsequenes of a movement towards a world made up of inreasingly larger trading blos will depend on just how important onsumption of within-blo varieties is relative to onsumption of varieties produed outside the blo. When the goods from all the blos enter symmetrially into utility, Krugman shows that welfare in a world of small numbers of big blos is likely to be low, while if blos are formed on the basis of "natural" trading partners who trade disproportionately with eah other, Krugman's analysis shows that blo formation may have more benign welfare onsequenes. Following on Krugman's work, several additional papers (Deardorff and Stern, 1991, and Bond and Syropoulos, 1992) have taken issue with the pessimisti view of trading blos emerging from Krugman's symmetritraders model and have instead provided models of regional blo formation that yield results more in line with the benign view assoiated with his natural blo model. As useful as these papers are in providing insights into the mehanis of the welfare tradeoffs involved in regional blo formation, they nevertheless fail to address what is perhaps the entral question in the debate over the wisdom of forming regional trading blos: What is their impat on the prospets for ontinued multilateral trade ooperation? The answer to this question an not be found in Krugman's model or models like it whih fous on the impliations of blo formation when all blos set non-ooperative tariffs on external trade. 2 This point is made forefully by Fred Bergsten's (Bergsten, 1991) ommentary on 3

6 Krugman (1991b): Regional trading arrangements are learly going to happen...the ruial question is whether these arrangements take plae within the ontext of an effetive and redible global system. If so, they will be - and will be viewed as - supplements to that system...indeed, it is the existene of tariff bindings under GATT (along with the prosriptions of Artile XXIV itself) that prevent blo members from raising barriers toward the outside world to exploit the potential gains desribed by Krugman. (p. 53, author's emphasis). To explore the effets of regional agreements on multilateral tariff ooperation under GATT, one must turn to ooperative models of tariff formation. Papers that do onsider the impat of regional agreements on ooperative multilateral tariffs inlude those of Ludema (1992), Kowalzyk (1990), Kowalzyk and Wonnaott (1991), and Kowalzyk and Sjostrom (1992). These papers fous on the bargaining aspet of international tariff ooperation, and explore how multilateral bargains an be altered by the opportunity to negotiate bilateral deals. However, these papers assume that binding ommitments an be made to enfore the bargaining outome. As we and others have argued elsewhere (Dixit, 1987, Thursby and Jensen, 1984, Bagwell and Staiger, 1990), in the ontext of international agreements, suh as GATT, it is not lear how suh binding ommitments ould be enfored: In pratie, the enforement of agreed-upon behavior under GATT is limited by the severity of retaliation that an be redibly threatened against an offender by its trading partners. 3 A different emphasis is pursued by Levy (1993), who adopts a median-voter model to explore the degree to whih regional options ould undermine politial support for multilateral liberalization. But again, as with other existing work, Levy takes the ability to make binding international ommitments as given. In this paper, we begin an exploration of the impat of the formation of regional trade 4

7 areas on the ability of ountries to maintain ooperative multilateral tariffs when they annot make binding international ommitments, but are instead limited to self-enforing arrangements. We adopt the view, as in Bagwell and Staiger (1990), that enforement issues are entral to an understanding of the dynami behavior of trade intervention in a world where ountries attempt to maintain ooperative trade poliies. Speifially, we view ooperation in multilateral trade poliy as involving a onstant balane between, on the one hand, the gains from deviating unilaterally from an agreed-upon trade poliy, and on the other, the disounted expeted future benefits of maintaining multilateral ooperation, with the understanding that the latter would be forfeited in the trade war whih followed a unilateral defetion in pursuit of the former. In suh a setting, hanges in urrent onditions or in expeted future onditions an upset this balane, requiring hanges in existing trade poliy that will bring inentives bak into line. In Bagwell and Staiger (1990), we adopted this view of tariff-setting to explore the impliations of temporary import surges for the ability to maintain low ooperative tariffs, and interpreted managed trade as a ooperative response to deal with temporary inentive problems in the presene of volatile trade swings. We explore in this and a ompanion paper (Bagwell and Staiger, 1993) the sense in whih the formation of regional trade agreements upsets the balane between urrent and future onditions, and trae through the dynami ramifiations of these effets for multilateral ooperation. A ruial fous of our analysis is the period of transition, during whih the regional agreement is being negotiated, and then implemented. Both beause regional trade agreements typially involve a lengthy staging of tariff redutions, and beause trade patterns 5

8 take time to reflet hanges in trade barriers in any event, there will inevitably be a lag between the onlusion of negotiations and ratifiation of the agreement on the one hand, and final hanges in trading patterns refleting the fully implemented regional agreement on the other. 4 Together with the period of negotiation, this lag reates a period of transition within whih, at least initially, the formation of the regional trade area has its biggest impat on expeted future trade patterns rather than on urrent onditions. It is this basi observation that is entral to our results. To plae our main findings in perspetive, it is helpful first to identify two of the prinipal effets of a regional agreement. One onsequene of suh an agreement is the trade diversion effet, whereby the elimination of internal tariffs between member ountries ats to enlarge intra-member trade volume at the expense of trade between member and nonmember ountries. 5 A seond effet is the market power effet. While the trade diversion effet is shared by both free trade agreements (suh as NAFTA) and ustoms unions (suh as the EC), the market power effet is partiular to the formation of ustoms unions: With the formation of a ustoms union, the member ountries adopt a ommon external tariff poliy, and this in turn enables them redibly to impose higher import tariffs on their multilateral trading partners, should suh punitive tariff ation be desired. 6 In our ompanion paper (Bagwell and Staiger, 1993), we onsider the formation of ustoms unions, and fous on the impliations for multilateral tariff ooperation of the market power effets assoiated with ustoms union formation. In the present paper we onsider regional free trade agreements, where the market power effet does not arise. Thus, our emphasis here is on the trade diversion effets that suh agreements generate, and on their 6

9 impliations for multilateral tariff ooperation. In this regard, we argue that the emergene of regional free trade agreements will be assoiated with temporarily heightened multilateral trade tensions between member and non-member ountries, and onsequently, to a temporary retreat from liberal multilateral trade poliies. This tension arises during the period of transition, when urrent trade flows between member and non-member ountries (and hene urrent inentives to deviate unilaterally from an agreed-upon multilateral tariff) are more or less unhanged at the same time that expeted future trade flows between member and nonmember ountries (and hene the value of maintaining future multilateral ooperation) have diminished. Intuitively, under suh onditions, ountries are apt to onfront long-standing trade disputes more readily, as the risks of a possible trade war no longer pose the deterrent to onfrontation that they one did. Our results suggest, however, that the tension between regional free trade agreements and multilateral liberalization is temporary: Eventually, as the full impat of the emerging regional agreement on multilateral trade patterns beomes felt, the initial balane between urrent and expeted future onditions reemerges, and liberal multilateral trade poliies an be restored. The remainder of the paper is devoted to establishing and elaborating on these points. The next setion sets out the basi model within whih we will study free trade agreements, and establishes several properties in a stationary setting that will be useful in the dynami non-stationary analysis to follow. Setion III then haraterizes the dynami behavior of equilibrium multilateral tariffs in the non-stationary environment of emerging regional free trade agreements. Setion IV derives various omparative statis results and disusses the impliations of our analysis with regard to the design of Artile XXIV. Setion V onludes. 7

10 II. Multilateral Tariff Determination in Stationary Environments In this setion we desribe a simple model of trade between two ountries, under the assumption that the trading environment between the two ountries is stationary through time. In this way, a useful benhmark is reated, against whih we an ontrast a nonstationary model whih allows for the emergene of free trade agreements, as depited in the following setions. A. A Stati Model We onsider two ountries, who trade G goods. To distinguish the ountries, we use an "*" to denote one of the ountries (heneforth referred to as the "foreign" ountry) while the absene of an "*" orresponds to the "no *" ountry (heneforth referred to as the domesti ountry). In order to make our points as simply as possible, we ignore the proess of prodution in the ountries, assuming instead that the respetive ountries are simply endowed with ertain amounts of eah good. Speifially, we assume a symmetri endowment distribution, whereby for eah of G/2 goods (e.g., the even-numbered goods) the foreign ountry has an endowment of two units whereas the domesti ountry has an endowment of zero units; and for eah of the other G/2 goods (e.g., the odd-numbered goods), the situation is reversed, with the foreign ountry having an endowment of zero units and the domesti ountry being endowed with two units. Thus, G/2 of the goods are potentially exported from the foreign to the domesti ountry, and the other G/2 goods follow the opposite trade diretion. In eah ase, the exporting ountry is endowed with two units of the good, and the importing ountry has an endowment of zero units. 8

11 We assume that demand funtions in the two ountries are symmetri aross produts and ountries, and that the demand for any produt i is independent of the pries of other produts j i. Speifially, the demand funtions for produt i ε {1,...G} are given by: (1) C(P i ) = α-βp i ; C(P i* ) = α-βp i* where (α,β)>0, P i is the prie of good i in the domesti ountry and P i* is the orresponding prie in the foreign ountry. Given the symmetry between the two ountries, for any produt i, we may simply speak of "the exporting ountry" and "the importing ountry." Aordingly, let P x i denote the prie of good i in the exporting ountry and P m i give the prie of good i in the importing ountry. We fous here on the determination of (speifi) import tariffs, and so τ m i is used to represent the import tariff levied on good i. It follows that: (2) P m i for eah good i. = P x i + τ m i The struture of the basi model is ompleted with the further requirement of market learing for eah produt i. Using the speifiations of the endowment distribution, (1) and (2), we thus require for every good i that: (3) 2 = α-βp x i + α-β(p x i +τ m i ). Solving (3) for P x i and using (2) then gives the market-learing pries: (4a) ˆP i x (τ i m ) = (α-1)/β - τ i m /2 (4b) ˆP i m (τ i m ) = (α-1)/β + τ i m /2. Finally, letting M i = C(P i m ) denote the import volume of good i, it is diret to use (1) and (4b) to determine good i's market-learing import volume, ˆM i (τ i m ) = C(ˆP i m (τ i m )), whih is 9

12 given by: (5) ˆM i (τ i m ) = 1 - (β/2) τ i m. At this point, given the symmetry of the model, it is apparent that no basis exists for asymmetri tariffs aross produts. Aordingly, we may drop the i supersript, and summarize our analysis so far with the definitions of market-learing export pries, import pries, and import volume for any good. Using (4a), (4b), and (5), these are given by: (6a) ˆP x (τ m ) = (α-1)/β - τ m /2 (6b) ˆP m (τ m ) = (α-1)/β + τ m /2 (6) ˆM(τ m ) = 1 - (β/2) τ m. Notie that under free trade import volume per import good would be one unit, with eah ountry's import volume from the other then amounting to G/2. With (6a)-(6) in plae, we are now ready to define welfare. For either ountry, we represent welfare by the sum of onsumer surplus, produer surplus, and import-tariff revenue. Thus, the domesti ountry's welfare funtion is given by: (7) where τ m is the import tariff levied by the domesti ountry on all imports, and τ m * is the import tariff imposed by the foreign ountry on all imports. The welfare of the foreign ountry is defined in an exatly symmetri fashion. Consider now the optimal import tariff for a ountry. Using (7) and C(ˆP m (τ m )) = ˆM(τ m ), it is diret to verify that: (8) where primes denote derivatives. There are two features of (8) that deserve speial omment. 10

13 First, observe that the marginal effet of an import tariff for the domesti ountry is ompletely independent of the trade poliy of the foreign ountry. This arises beause of our assumption of demand independene and beause export taxes are not onsidered. Seond, sine ˆP m` (τ m ) < 1 from (6b), it is lear that a small import tariff improves welfare. For more speifi results, we use (1), (6b) and (6) to re-write (8) as: (9) It follows that W(G,τ m,τ * m ) is onave in τ m. Thus, for any fixed τ * m, the welfaremaximizing response is τ m = 2/3β. Let us now define the stati tariff game to be the game in whih both ountries simultaneously selet an import tariff for all goods, with eah ountry seeking to maximize its own welfare. Sine eah ountry's best-response tariff is independent of the tariff imposed by the other ountry, we have that the Nash equilibrium of the stati tariff game ours when eah ountry selets the import tariff given by: (10) As it is easily verified that W(G,τ,τ) is stritly dereasing in τ, the stati tariff game resembles a Prisoners' Dilemma game: both ountries are better off when there is free trade and are monotonially made better off with any symmetri movement towards free trade, but in the Nash equilibrium the ountries impose positive tariffs and experiene the onsequent lower welfare. Figure 1 illustrates all this by depiting the gains from importing a representative import good in the top left panel, and the gains from exporting a representative export good in the top right panel. Under free trade, these gains would be given by the area under the 11

14 import demand urve (m 1 m 2 m 3 ) and the area above the export supply urve (x 1 x 2 x 3 ), respetively. Under the optimal tariff, the additional gains from importing are given by the net tariff revenue olleted from abroad (m 1 m 4 m 5 m 6 ) minus the dead weight loss triangle (m 6 m 7 m 3 ). Faing the optimal tariff abroad, the redution in the gains from exporting are given by the net import taxes paid by exporters (x 1 x 4 x 5 x 6 ) and the dead weight loss triangle (x 5 x 6 x 3 ). Taken together, when both ountries impose their optimal tariffs, the losses in eah ountry's export market outweigh the gains in its import market, with the net loss for a representative import and export good amounting to the sum of the dead weight loss triangles (m 6 m 7 m 3 + x 5 x 6 x 3 ). The lower panel of Figure 1 depits the domesti and foreign tariff reation urves, with domesti indifferene urves refleting the relative welfare rankings assoiated with reiproal free trade, unilateral optimal tariff setting, and Nash equilibrium tariffs in the stati tariff game. B. A Stationary Dynami Model We now onsider a stationary dynami tariff game, whih is defined by the infinite repetition of the stati tariff game desribed above. In eah period the ountries observe all previous import tariff seletions and simultaneously hoose import tariffs. For the reasons given above, we ontinue to assume that eah ountry applies the same tariff to eah imported good in any given period. The game is stationary in the sense that none of the model's parameters hanges through time. Let δ ε(0,1) denote the disount fator between periods. In order to express our ideas in a simple manner, we fous on a partiular lass of subgame perfet equilibria for the stationary dynami tariff game. Speifially, we onsider 12

15 equilibria in whih (i) symmetri stationary non-negative import tariffs are seleted along the equilibrium path, meaning that in equilibrium the two ountries selet the same import tariff in eah period, and (ii) if a deviation from this ommon tariff ours, then in the next period and forever thereafter the ountries revert to the Nash equilibrium tariffs of the stati tariff game. We then refer to the most-ooperative equilibrium of the stationary dynami tariff game as the subgame perfet equilibrium whih yields the lowest possible equilibrium tariff while satisfying restritions (i) and (ii). The orresponding import tariff is then termed the most-ooperative tariff for the stationary dynami tariff game. 7 In a dynami model, ountries have the possibility of supporting a ooperative tariff, τ with τ <ˆτ N, sine any attempt to raise the urrent-period tariff will be greeted with the retaliatory (Nash) tariff from the trading partner in future periods. Intuitively, a ooperative tariff τ an then be supported in an equilibrium for the stationary dynami tariff game if the one-time inentive to heat is suffiiently small relative to the future value of maintaining a ooperative relationship with the trading partner. To formalize this intuition, let us first examine the inentive a ountry has to heat. For a fixed ooperative tariff τ <ˆτ N, and given the lass of subgame perfet equilibria upon whih we fous, if a ountry is to deviate and selet τ τ, then it will deviate to its bestresponse Nash tariff, ˆτ N. Thus, the gain from heating is given by: (11) Ω(G,τ ) W(G,ˆτ N,τ ) - W(G,τ,τ ). When a ountry heats, however, it also auses future welfare to drop, and we now examine this ost of heating. Define the one-period value to ooperation to be: (12) ω(g,τ ) = W(G,τ,τ ) - W(G,ˆτ N,ˆτ N ). 13

16 Then the ost to heating is δ/(1-δ) ω(g,τ ), sine one a ountry defets and selets a high import tariff, ooperative tariffs are thereafter replaed by the higher Nash tariffs. Using (11) and (12), the fundamental "no-defet" ondition is that the benefit of heating be less than the disounted future value of ooperation, or: (13) Any ooperative tariff τ that satisfies (13) an be supported in a subgame perfet equilibrium of the stationary dynami tariff game. Our interest lies in the most-ooperative tariff, ˆτ, whih is the smallest non-negative tariff that satisfies (13). To haraterize this tariff, we first investigate the properties of Ω(G,τ ) and. Straightforward alulations reveal that: (14) Using (14), it follows that: (15) (16) Thus, a larger number of traded produts, whih orresponds to a greater volume of trade, ats to raise the benefit from a tariff hike, sine the larger tariff is then applied to more units. Notie also that lower ooperative tariffs heighten the inentive to heat, beause a deviation to the Nash tariff then represents a more signifiant tariff inrease. Calulations also reveal that: (17) Using (17), it follows that: 14

17 (18) (19) Thus, a greater volume of trade enhanes the future disounted value of ooperative tariffs, but higher ooperative tariffs lower the disounted value of future ooperation. The determination of the most-ooperative tariff is now niely illustrated by Figure 2. Observe in Figure 2 that the no-defet ondition (13) is satisfied for all τ ε[ˆτ,ˆτ N ]. These are the set of tariffs that are supportable as subgame perfet equilibrium tariffs for our stationary dynami tariff game, given the lass of equilibria upon whih we fous. Solving (13) for the tariff that gives equality yields the most-ooperative tariff, whih is given by: (20) Two observations are immediately apparent. First, note that ˆτ is dereasing in δ, with ˆτ = 0 at δ=3/5 and ˆτ = ˆτ N at δ = 0. This dereasing relationship is intuitive: as δ inreases, the disounted value of future ooperation is enhaned, and so a lower tariff an be supported (despite the onsequent greater inentive to heat). This proess is easily illustrated in Figure 2, if one imagines inreases in δ ausing an upward shift in. To avoid ases in whih the most-ooperative tariff orresponds to either of the extreme polar outomes of free trade or the non-ooperative tariff ˆτ N, we assume δ ε (0,3/5) in all that follows. Seond, observe that the most-ooperative tariff is independent of G. In other words, the most-ooperative tariff level between two ountries trading many goods would be the same as that between two ountries trading few goods, provided that the ountry pairs are 15

18 otherwise idential and that for both trading pairs the number of goods traded is onstant through time. This "neutrality" feature of our dynami stationary model is purposeful, as we will use this feature in the next setion to larify the essential onsequenes of the formation of free-trade agreements for multilateral tariff determination. C. Summary We may summarize the findings of this setion with the following proposition: Proposition 1: (a) For the stati tariff game, the Nash equilibrium ours when eah ountry sets an import tariff of level ˆτ N = 2/3β on eah imported good. (b) For the stationary dynami tariff game, the most-ooperative equilibrium ours when eah ountry sets the most-ooperative import tariff on eah imported good. III. The Formation of Free-Trade Areas We turn now to a dynami model in whih, at some point in time, the foreign and domesti ountries enter into free-trade agreements with other (unmodeled) ountries. We fous on how the formation of their own separate free trade areas affets the ability of the domesti and foreign ountries to ontinue to ooperate multilaterally. Our modeling approah is to assume that the free-trade agreements result in the foreign and domesti ountries trading fewer goods with eah other, as more of their trade is diverted into the 16

19 respetive free-trade zones. Thus, the formation of free-trade areas marks a real hange in the multilateral trading environment between the domesti and foreign ountries, and our goal here is to examine the ramifiations of this non-stationarity for the multilateral tariffs that these ountries are able to support. A. The Free-Trade Agreement Model We again assume that the domesti and foreign ountries set import tariffs simultaneously in eah of an infinite number of periods. The game we onsider in this setion, however, differs from the stationary dynami tariff game onsidered above, in that the strutural environment within whih the two ountries trade is now assumed to hange through time. Speifially, we envision a trading relationship that passes through three phases. In phase 1, the foreign and domesti ountries trade G goods with one another, just as desribed in the previous setion. They are aware, however, that a time may ome at whih it beomes politially feasible for eah to negotiate respetive free-trade agreements with other ountries. Phase 2 orresponds to a transition phase, in whih the foreign and domesti ountries ontinue to trade all G goods with one another, but in whih eah of these two ountries has already begun disussion with other ountries about future free-trade agreements. Finally, in phase 3, the free-trade agreements are fully implemented, the foreign and domesti ountries now trade less with one another as they eah divert some trade to their respetive free-trade partners, and these new trading patterns are stationary into the infinite future. We model this trade diversion effet by assuming that the domesti and foreign ountries trade only G-F 1 goods in phase 3. 17

20 We hoose not to rigorously examine the politial proess, the diret welfare benefits of the free trade agreements for the domesti and foreign ountries, the welfare of other (free-trade-partner) ountries, or the free-trade-agreement negotiation proess. Instead, we assume simply that in any period, if free-trade disussions have not yet begun, then there is a probability of ρ ε(0,1) that they will begin (for both ountries and their respetive partners) in the next period. Thus, if the ountries are in phase 1 at date t, then ρ is the probability of being in phase 2 at date t+1. We model the transition from phase 2 to phase 3 in a similar way: if the various ountries have already begun negotiating their respetive freetrade agreements, then there is a probability of λ ε(0,1) that all free-trade agreements will be finalized and that implementation will be omplete by the beginning of the next period. Thus, if ountries are in phase 2 at date t, then λ is the probability of being in phase 3 at date t+1. This set-up has two features that require speial omment. 8 First, we assume that all free-trade-agreement disussions begin at the same (random) date, and that the free-trade agreements also are all ompleted and implemented at the same (random) date. These assumptions are not intended to be interpreted literally; rather, they ensure that the foreign and domesti ountries fae symmetri situations, and this in turn onsiderably simplifies the analysis. Seond, we assume that the enatment of the respetive free-trade agreements results in the domesti and foreign ountries trading fewer goods. This assumption appears to onflit with Artile XXIV of GATT, whih requires free-trade agreements to apply to substantially all trade, suggesting that F=G. This onflit is only superfiial, however. For example, it might be assumed that the (unmodeled) ountries that partiipate in the respetive 18

21 free-trade agreements do not trade in the remaining G-F goods. More generally, free-trade agreements yield trade-diversion onsequenes that are more pronouned for some goods than others, and our model represents a situation in whih trade is strongly diverted into freetrade unions for F goods, but the extent of trade diversion is minor for the other G-F goods. 9 In any ase, for this free-trade agreement game, we examine a lass of subgame perfet equilibria, in whih (i). along the equilibrium path, in any given phase of the game, the foreign and domesti ountries selet a ommon import tariff for all goods at all dates within the phase; and (ii). if at any point in the game a deviation from the equilibrium tariff for the orresponding phase ours, then in the next period and forever thereafter the two ountries revert to the Nash equilibrium tariffs of the stati tariff game. For suh equilibria, there will be three ooperative tariff levels, with eah orresponding to a different phase. Let τ 1, τ 2 in phases 1, 2 and 3, respetively. and τ 3 refer to the ooperative tariff levels We again look for a most-ooperative equilibrium, and we solve for the orresponding most-ooperative tariffs, ˆτ 1, ˆτ 2 and ˆτ 3, in a reursive fashion. Speifially, we first identify the no-defet ondition for phase-3 and find the lowest tariff that an be supported in this phase in an equilibrium of the desired lass. Having thus solved for ˆτ 3, we next turn to phase 2, represent the relevant no-defet ondition for this phase, and then solve for the most-ooperative tariff at this phase. Finally, with ˆτ 2 and ˆτ 3 then determined, we haraterize the no-defet ondition for phase 1 and solve for the lowest tariff that doesn't invite heating. This tariff is ˆτ 1. Sine the disounted value of future ooperation rises as 19

22 the level of future tariffs falls, and sine urrent tariffs are minimized by hoosing future tariffs whih maximize the disounted value of future ooperation, solving reursively as we do for the lowest sustainable tariff in phase 3 first, followed by phase 2 and phase 1, provides the lowest tariff sustainable in eah phase of the model. This, then, outlines the basi struture of our model as well as the method by whih we will haraterize the most-ooperative tariffs. The next step is to formally derive the nodefet onditions for eah of the three phases. Let us begin with phase 3. At this point, the domesti and foreign ountries trade only G-F goods, and the future is stationary, so the no-defet ondition for phase 3 is: (21) This has the same form as (13), the no-defet ondition in our stationary model, exept that the number of goods traded is now G-F rather than G. But sine ˆτ, whih as defined by (20) is the most-ooperative tariff in our stationary model, is independent of the number of goods traded, it follows immediately from (20) that: (22) is the most-ooperative tariff that an be supported in phase 3. The fat that the most-ooperative tariff in phase 3 when G-F goods are traded and the future is stationary, ˆτ 3, is equal to the most-ooperative tariff in our stationary model when G goods are traded, ˆτ, reflets one again the "neutrality" property of our model with regard to (stationary) levels of the volume of trade. This property suggests an intuitive benhmark from whih our results an be measured. That is, if the formation of free trade agreements were to ome as a omplete surprise and were implemented instantaneously, then 20

23 ˆτ would be the most-ooperative multilateral tariff up until the instant that the free trade agreements arose, at whih point F goods would beome non-traded multilaterally and the same level of multilateral tariff ooperation (ˆτ 3 = ˆτ ) would ontinue to prevail thereafter. In this ase, the presene or absene of free trade areas would be ompletely irrelevant for the level of multilateral tariff ooperation sustainable. This benhmark helps to emphasize the fat that, in our model, the emergene of free-trade areas an have an effet on the mostooperative multilateral tariff levels only in so far as the transition proess assoiated with their formation is expliitly onsidered. 10 We turn now to phase 2. The no-defet ondition in this ase is: (23a) where r indexes the period at whih phase 3 begins, with r=1 meaning that phase 3 begins in the next period, and where q and k orrespond to periods within phases 2 and 3, respetively. 11 Notie also that all G goods are traded between the two ountries in phase 2, as refleted by the use of G in Ω. With some further algebra, the phase-2 no-defet ondition may be re-written in an easier-to-use form: (23b) where V 2 is understood to be the expeted disounted value to ontinued ooperation, as viewed in phase 2. The lowest tariff satisfying (23b) defines ˆτ 2, whih we haraterize below. Before doing this, however, we represent the phase-1 no-defet ondition: (24a) where s indexes the period at whih phase 2 begins, with s=1 meaning that phase 2 21

24 begins in the next period, and where t represents periods within phase 1. Using (23b), we may re-write (24a) as: (24b) where V 1 gives the expeted disounted value to ooperation, from a phase-1 perspetive. The smallest tariff satisfying (24b) is then defined to be ˆτ 1. B. Charaterization of the Most-Cooperative Tariffs We now investigate the properties of ˆτ 1, ˆτ 2 and ˆτ 3, in order to determine the onsequenes of bilateral free-trade agreements for multilateral tariff determination. In partiular, we seek to rank the relative magnitude of the three most-ooperative tariffs, and in this way to assess the impat of the formation of regional free trade areas on the ability to maintain low multilateral tariffs. We haraterize these tariffs reursively, through a sequene of lemmas. To begin, let us reall from (22) that: Lemma 1: Thus, for the permissible range of δ, the phase-3 most-ooperative tariff lies between freetrade and the Nash tariff. Consider next the transition or phase-2 tariff, ˆτ 2. In order to haraterize this tariff, some features of the funtion V 2 (τ 2 ; λ,δ,f) must be determined. We have the following: (25a) 22

25 (25b) V 2 (ˆτ N ;λ,δ,f) > 0 (25) The first two observations follow easily from (19) and (23b) and also (17) and ω(g,ˆτ N ) = 0, respetively. To prove (25), note that (18) implies: With these observations in plae, we are now prepared to haraterize ˆτ 2, whih is the lowest tariff for whih Ω(G,τ 2 ) V 2 (τ 2 ;λ,δ,f). As Figure 3 illustrates, ˆτ 2 must lie stritly between ˆτ 3 and ˆτ N : Lemma 2: 0 < ˆτ 3 < ˆτ 2 < ˆτ N. In words, a high multilateral tariff is required while free-trade agreements are being negotiated and implemented, but one the agreements are fully implemented and the pattern of trade reflets fully the hanged onditions brought about by the regional agreements, the multilateral tariff rate delines. The intuition underlying Lemma 2 is atually quite diret. If the trade volume (number of traded goods) between the domesti and foreign ountries were stationary through time, then the two ountries ould support a tariff level of ˆτ 3. But, in the transition phase of the free-trade agreement game, the two ountries reognize that their urrent trade volume exeeds that whih will obtain between them one the free-trade agreements are finalized and 23

26 fully implemented. Thus, as ompared to the stationary environment that supports ˆτ 3, the ountries pereive the expeted disounted value of future ooperation now to be lower (as (25) states). Hene, to maintain ooperation, the inentive to heat also must be diminished, and this is aomplished by endogenously reduing the volume of trade in the transition phase with the seletion of a higher import tariff. In this way, the antiipation of the eventual trade diversion assoiated with the free-trade-agreement results in the seletion of a high import tariff during the period of time over whih the agreements are negotiated and implemented. 12 We turn next to the initial phase-1 tariff, ˆτ 1, whih is the lowest tariff for whih Ω(G,τ 1 ) V 1 (τ 1 ;ρ,λ,δ,f). To haraterize this tariff, we first reord the following: Lemma 3: ω(g,ˆτ 2 ) > ω(g-f,ˆτ 3 ). This lemma states that the per-period equilibrium value of ooperation delines as the ountries move from the transition phase 2 to the final phase 3. Intuitively, two offsetting effets are involved in this alulation: the per-period value of ooperation will be greater in the transition phase relative to the final phase due to the relatively greater volume of trade assoiated with that phase, but this value is also diminished by the presene of high transition-phase tariffs. The lemma, whih is proved in the Appendix, indiates that the diret effet of higher trade volume in the transition phase outweighs the indiret effet of high transition-phase equilibrium tariffs. With lemma 3 in plae, some observations about the V 1 funtion may be made: 24

27 (26a) (26b) (26) The first observation again follows diretly from (19) and (24b). To prove (26b), one an use the definitions of V 2 and V 1 given in (23b) and (24b), respetively, to show that: where the inequality follows from Lemma 3. Finally, for (26), note that Lemma 2, (18) and (19) imply that: With these observations established, we may now return to Figure 3 and onlude that ˆτ 3 < ˆτ 1 < ˆτ Summarizing: Lemma 4: ˆτ 3 < ˆτ 1 < ˆτ 2. Thus, the initial-phase tariff is higher than the final-phase tariff, although it is not as large as 25

28 the tariff that ours in the transition phase. To gain some intuition for this lemma, let us first onsider why tariffs are higher in the transition phase than the initial phase. As the proof to (26b) suggests, this is in fat a onsequene of the onlusion of Lemma 3, that the per-period equilibrium value of ooperation delines as the ountries pass from the transition to the final stage. An essential differene between the initial and the transition phase is that the lower-stakes final stage is expeted to arrive sooner when the ountries are already in the transition phase. This effet raises the disounted expeted value of ooperation in the initial relative to the transition phase, and as a onsequene a lower tariff an be supported in the initial phase. Consider next the finding that the initial-phase tariff exeeds the final-phase tariff. Reall that the final-phase tariff, ˆτ 3, is the minimal tariff whih an be supported in a stationary dynami environment. This tariff, however, annot be supported in the initial phase of the free-trade agreement model, and for two reasons. First, higher tariffs are required in the transition phase of the free-trade agreement model, and the prospet of this diffiult transition proess redues the expeted disounted value of ooperation relative to that found in the stationary dynami setting. Seond, free-trade agreements eventually will be established, and so the volume of trade will drop relative to that whih would be maintained in a stationary dynami model. Both of these effets, whih are eah used in the proof of (26), at to lower the expeted disounted value of ooperation for the initial phase of the free-trade agreement game to a value that is below that whih would be found in a stationary dynami setting. Thus, a high initial-phase tariff is required, in order to slow the volume of trade and redue the inentive to heat

29 Observe that an interesting orollary of Lemmas 3 and 4 is that the per-period equilibrium value of ooperation delines in eah suessive phase of the free-trade agreement game: Corollary 1: ω(g,ˆτ 1 ) > ω(g,ˆτ 2 ) > ω(g-f,ˆτ 3 ). The first inequality uses ˆτ 1 < ˆτ 2 from Lemma 4 and (19), while the final inequality is just a restatement of Lemma 3. This orollary will be useful below, when we do omparative statis. Our main results may now be summarized in the following proposition: Proposition 2: For the free-trade agreement game, in the most-ooperative equilibrium, the domesti and foreign ountries set the most-ooperative import tariffs, ˆτ 1, ˆτ 2 and ˆτ 3, in phases 1, 2 and 3, respetively, on eah imported good. Furthermore, 0 < ˆτ 3 < ˆτ 1 < ˆτ 2 < ˆτ N. Figure 4 reflets the result of Proposition 2 by depiting the multilateral tariff during the pretransition, transition, and post-transition phases of the free trade agreement model. As illustrated, regional free-trade agreements are detrimental to multilateral tariff liberalization, in the sense that the prospet of eventual free-trade agreements and the onsequent diversion of trade puts upward pressure on multilateral tariffs both before free-trade agreement negotiations begin and, partiularly, during the negotiation and implementation phase. On the other hand, however, one the agreements are in plae, low multilateral tariffs again an 27

30 be supported. IV. Comparative Statis and Artile XXIV The free-trade agreement model developed in the previous setions has a variety of parameters, and it is important to assess the sensitivity of the most-ooperative tariffs to these parameters. Moreover, several of the parameters of the model an be assoiated loosely with GATT poliy toward free trade agreements as embodied in Artile XXIV. Hene, in this setion we present omparative statis results and disuss their impliations with regard to the design of Artile XXIV. Examining the respetive no-defet onditions (21), (23b) and (24b), it is apparent that the most-ooperative tariffs have the following funtional dependenies: ˆτ 3 = ˆτ 3 (δ), ˆτ 2 = ˆτ 2 (λ,δ,f) and ˆτ 1 = ˆτ 1 (ρ,λ,δ,f). A. Comparative Statis on ρ The parameter ρ reflets the speed with whih free-trade negotiations will begin. It affets only the phase-1 tariff, ˆτ 1, and it is lear from Figure 3 that ˆτ 1 is inreasing in ρ if V 1 (τ 1 ;ρ,λ,δ,f) dereases in ρ at τ 1 = ˆτ 1. Intuitively, if a higher value for ρ lowers the expeted disounted value of ooperation, then a higher tariff is required to hold the inentive to defet in hek. We are therefore led to onsider the following alulation: 28

31 where the inequality follows from Corollary 1. We thus have: Lemma 5: Intuitively, as ρ rises, the transition and final stages are enountered earlier, and this diminishes the disounted expeted value of ooperation in the initial phase, sine the perperiod equilibrium value of ooperation delines in eah suessive phase. A higher initial tariff is therefore required. We may onlude that a greater prospet of entering the freetrade agreement negotiation proess ats to raise the initial-phase tariff. B. Comparative Statis on λ An inrease in λ affets both ˆτ 2 and ˆτ 3, and a higher λ means that the free-trade negotiation and implementation proess will take less time one it begins. Arguing as above, ˆτ 1 and ˆτ 2 inrease with λ, if V 1 (τ 1 ;ρ,λ,δ,f) and V 2 (τ 2 ;λ,δ,f) deline in λ at τ 1 = ˆτ 1 and τ 2 = ˆτ 2, respetively. It is diret to show that: and 29

32 Using Corollary 1, it follows that the latter partial derivative is negative, and thus that ˆτ 2 inreases with λ. Using this finding and Corollary 1 one more then gives that the first partial derivative is also negative, so that ˆτ 1 inreases with λ. Summarizing, we have: Lemma 6: The intuition here is also quite simple. A higher value for λ expedites the transition into the final phase, where ooperation is less valuable. Hene, ˆτ 2 must rise to maintain ooperation in the transition phase. Finally, from the perspetive of the initial phase, a greater value for λ lowers the expeted disounted value of ooperation, both beause it expedites entry into the lower-stakes final phase and beause it raises transition-phase tariffs along the way. We may therefore onlude that a more rapid free-trade agreement negotiation and implementation proess will ause higher multilateral tariffs in the initial and transition phases. C. Comparative Statis on F One may think of F as refleting the degree of trade diversion orresponding to the formation of free-trade areas. In other words, for larger F, there will be greater trade diversion following the implementation of the free-trade agreements. As has been previously established (see (20)), ˆτ 3 is independent of F. To evaluate the dependene of ˆτ 2 and ˆτ 1 on F, observe that: 30

33 and Thus, as F grows, V 2 is diminished, and so ˆτ 2 rises. This in turn implies that V 1 also falls with F, so that ˆτ 1 must inrease with F. Summarizing, we have that: Lemma 7: The essential intuition is quite lear. When F is big, a free-trade agreement will ause a big loss in trade volume between the foreign and domesti ountries. This means that the disounted expeted value to ooperation is diminished, and so higher tariffs are required to blok the inentive to heat. In other words, free-trade blos with a high degree of trade diversion require greater adjustment in initial- and transition-phase multilateral tariff rates. D. Comparative Statis on δ The patiene of trading partners is aptured by the parameter δ. It has been established previously (see (20)) that ˆτ 1 delines with δ. To verify that ˆτ 2 and ˆτ 3 also deline with δ, observe that: 31

34 where the inequality follows sine ˆτ 3 delines in δ. Thus, we have that ˆτ 2 also delines in δ. Using these findings, and re-writing V 1 as it is diret to verify that eah braketed term is inreasing in δ. Thus, and so ˆτ 1 also delines in δ. Colleting these results, we have: Lemma 8: E. Summary We may now summarize the preeding lemmas as follows: 32

35 Proposition 3: For the free-trade agreement game the most-ooperative tariffs satisfy the following relationships: (i) ˆτ 1 (ρ,λ,δ,f) is dereasing in δ and inreasing in ρ, λ, and F. (ii) ˆτ 2 (λ,δ,f) is dereasing in δ and inreasing in λ and F. (iii) ˆτ 3 (δ) is dereasing in δ. The general message of this proposition, then, is that parameter hanges that make the eventual free-trade agreements more imminent (higher ρ,λ) or more trade-diverting (higher F) result in higher multilateral tariffs up until the point that the agreements are atually implemented. 15 F. Artile XXIV It is interesting at this point to assess the nature of the prosriptions plaed on free trade agreements by Artile XXIV in light of these results. In partiular, Artile XXIV requires that (i) free trade agreements provide for the elimination of duties on "substantially all" trade, and that (ii) any interim agreement whih serves as a stepping stone to the free trade agreement be of "reasonable" duration. The eonomi rationale for ondition (i) has been the subjet of a long and inonlusive literature, but it an be argued (see Bhagwati, 1991) that the main intent of this ondition is simply to redue the frequeny with whih free trade agreements are negotiated. At the same time, ondition (i) limits the ability of ountries to define what trade is and is not overed in the free trade agreement, and in so doing limits the ability of ountries to seletively liberalize setors where the main impat of liberalization would be to divert trade 33