Article XXIV of the GATT: Good, Bad, or Both? Kamal Saggi∗ and Halis Murat Yildiz† Preliminary draft — comments welcome. Abstract Article XXIV of the General Agreement on Tariffs and Trade (GATT) provides an important exception to Article I (most favored nation clause) by permitting countries to enter into preferential free trade agreements (PTAs). Does the existence of Article XXIV undermine multilateral trade liberalization? Would aggregate world welfare be higher if PTAs could not be formed? To answer these questions, this paper endogenizes the formation of free trade agreements (FTAs) in a three-country oligopoly model of intraindustry trade where countries are asymmetric with respect to market sizes and production costs. Two tariff games are contrasted: one where countries have the option to form FTAs and another where they do not. We show that free trade is less likely to occur as an equilibrium outcome under the FTA option game and that the FTA option lowers world welfare when countries are relatively symmetric. Also, the country with the largest market (and/or the highest cost) always benefits from the existence of the FTA option. Keywords: Multilateral Trade Liberalization, Free Trade Agreements, GATT, Intraindustry Trade, Oilgopoly. JEL Classifications: F13, F12. ∗ Department of Economics, Southern Methodist University, Dallas, TX 75275-0496. Phone: 214-768-3274; fax: 214-768-1821; e-mail: ksaggi@mail.smu.edu. † Department of Economics, Ryerson University, 350 Victoria Street, Toronto, ON, Canada M5B 2K3. Phone: 416-979-5000 (ext 6689); fax: 416-979-5289; e-mail: hyildiz@ryerson.ca. 1 1 Introduction By permitting member countries of the World Trade Organization (WTO) to preferential trade agreement (PTA) wherein a group of member countries of the WTO extend tariff concessions to each other that they do not extend to other members, Article XXIV of the General Agreement on Tariffs and Trade (GATT) provides an important exception to the most-favored-nation (MFN) clause (contained in Article I of GATT). Since the notion of nondiscrimination as specified by the MFN clause is at the heart of the WTO system, the existence of Article XXIV has not been without controversy.1 Does the option of forming PTAs adversely affect incentives for multilateral trade liberalization on the part of member countries of the WTO? Wouldn’t the ultimate goal of GATT (i.e. multilateral free trade) be easier to obtain if member countries of the WTO had to strictly abide by Article I and could not form PTAs? The present paper addresses these questions in a three-country oligopoly model of intraindustry trade wherein countries differ with respect to their market sizes and production costs.2 The underlying asymmetry in the model sheds light on underlying characteristics that determine how the PTA option affects individual countries as well as on the welfare effects of PTAs. That the above questions are important is evident from the fact that over 150 PTAs are in force today and almost all major countries participate in some type of a PTA or another (WTO, 2002). The trend toward the formation of PTAs has intensified in recent years — most PTAs have been concluded in the last 10 years or so and since 1995 over 100 such agreements have been notified to the WTO. Prominent examples of PTAs include the North American Free Trade Agreement (NAFTA), the South American Common Market (MERCOSUR), the Association of South East Asian Nations (ASEAN) Free Trade Area, the Andean Pact, and the numerous agreements of the European 1 To minimize the potential harmful effects of PTAs Article XXIV article requires that: (i ) a PTA must cover almost all trade between its members; (ii) PTA members must fully eliminate tariffs and other trade restrictions on each other; and (iii) they should not raise tariffs (or any other trade restrictions) on non-members. 2 These questions are related but not completely equivalent to the question posed by Jagdish Bhagwati (1991): “Are PTAs building or stumbling blocks for multilateral trade liberalization?” The difference is that in our approach, both preferential and multilateral trade liberalization are endogenous and article XXIV is the underlying exogenous factor. See Bhagwati and Panagariya (1999) and Winters (1998) for overviews of the main policy questions in the area. 2 Union with other countries.3 Restricting attention to free trade agreements (FTAs), we evaluate the incentives countries have for multilateral trade liberalization under two scenarios: one in which they can avail of the option to form FTAs and another where this option does not exist. Formally, the paper compares two tariff games. The FTA game proceeds as follows. In the first stage, each country announces a set of countries with whom it wants to form a trade agreement. A bilateral FTA is formed if both countries want the agreement. Similarly, multilateral free trade emerges only when all countries want such an agreement. Next, given the constraints imposed by the policy regime in place, countries simultaneously choose their tariffs to maximize national welfare (defined as the sum of the local firm’s profits, consumer surplus, and tariff revenue).4 Finally, firms compete in the product market in a Cournot fashion where individual country markets are assumed to be segmented. In the no FTA game, in stage one, countries must choose between participating in multilateral free trade or utilizing individually optimal tariffs (i.e. no trade agreement at all). From each country’s perspective, an FTA embodies the following tradeoff. On the one hand, joining an FTA implies that the domestic surplus is lowered relative to the case where (optimally chosen) tariffs can be used. On the other hand, being part of an FTA increases export profits of members in each other’s markets. Since countries are asymmetric, the loss in domestic surplus and the gain from increased market access abroad generally differ across countries. In fact, a country’s loss in domestic surplus from an FTA increases with own market size while the gain in export profits increases with the partner’s market size. As a result, an FTA between the largest and smallest countries is never stable. Furthermore, a country’s gain from increased market access abroad decreases with own cost whereas it increases with a partner’s cost. As a result, the country with the biggest market size and the one with the highest cost are the most FTA desirable partners from the viewpoint of other countries. 3 The phrase ‘regional trade agreements’ is often used to describe trade liberalization amongst a few countries. However, the phrase ‘preferential trade liberalization’ is more apt since such liberalization need not occur only among geographically proximate countries (although it often does). 4 Under bilateral FTAs, a country imposes optimally chosen tariffs on non-members and no tariffs on members. Under the multilateral trade agreement, all countries practice free trade. 3 Our approach is similar to that of Riezman (1999) who also asks whether the PTA option facilitates or hinders the achievement of global free trade. However, there are important differences between the two papers. First, our model differs substantially from Riezman (1999) who utilizes a simple general equilibrium framework with fixed endowment levels. In his model, trade is inter-industry in nature and countries impose tariffs to improve their terms of trade. By contrast, in the oligopoly model presented here trade is intraindustry in nature (as in Brander and Krugman, 1983) and countries use tariffs to extract rents from foreign firms (as in Brander and Spencer, 1984). Second, and more importantly, our model allows us to address issues related to asymmetries between countries; such is not the case in Riezman (1999). For example, our model permits an assessment of whether or not MFN facilitates multilateral free trade.5 While we ask questions related to those examined in Krishna (1998), the conceptual approach of our paper is quite different. Using a similar underlying trade model, Krishna (1998) explores the relationship between FTAs and multilateral trade liberalization. In his model, tariffs are exogenously given, but FTAs are endogenous in the sense that only those FTAs that benefit producers in member countries are considered. Krishna (1998) finds that the formation of FTAs undermines support for multilateral trade liberalization. The present paper supports this result because we show that FTAs reduce the likelihood of obtaining multilateral free trade. On the other hand, we argue that in a world in which the FTA option does not exist is not necessarily superior to one where it does exist: there are regions over which multilateral free trade cannot be obtained but welfare improving PTAs are feasible. Shutting out FTAs would imply that such welfare gains would have to be foregone. Furthermore, our analysis of cost asymmetries delivers an interesting insight that we believe is truly novel to the literature: a pattern of FTAs where the lowest cost country forms bilateral PTAs with the other two can be actually welfare superior to multilateral free trade. The ‘stumbling versus building block’ question posed by Bhagwati (1991) has also been analyzed extensively in the literature through models that allow for repeated interaction between countries — see Riezman (1991), Bagwell and Staiger (1997a, 1997b, and 1998a), Bond et. al. (2001), and Bond and Sy5 It is also worth noting that Riezman (1999) uses the cooperative solution concept of the core to illustrates his results via numerical examples whereas we analyze a noncooperative game and analytically derive its sub-game perfect and coalition proof Nash equilibria. 4 ropoulos (1996). In these models, cooperation is required to be self-enforcing in the sense that each country balances the current benefit of deviating from the cooperative tariff against the future losses caused by the breakdown of multilateral cooperation that results from its defection.6 In our model, an FTA is self-enforcing in the sense that it needs to be immune to credible coalitional deviations by both members and non-members. Levy (1997) focuses on political economy considerations that we abstract from and finds that in the monopolistic competition model of intraindustry trade in differentiated goods, PTAs can supplant multilateral trade liberalization.7 Unlike in the present paper, tariffs play virtually no role in Levy’s analysis since only the choice between free trade (bilateral as well as multilateral) and autarky is considered. Freund (2000b) investigates how (exogenous) multilateral trade liberalization affects incentives for preferential trade liberalization in the oligopoly trade model.8 Her main result is that multilateral trade liberalization encourages the formation of PTAs and makes it more likely that such agreements are non-credible. 2 Model We develop a simple oligopoly model of trade in which each country has a unilateral incentive to impose rent extracting tariffs on those trading partners with whom it does not have any trade agreement. There are three countries and two goods: x and y. Countries are asymmetric with respect to their market size and production costs. Good x is produced by a single profitmaximizing firm in each country at a constant marginal cost in terms of the numeraire good y.9 Preferences over the two goods are quasilinear: Ui (xi , yi ) = u(xi ) + yi 6 (1) The literature on PTAs is rather extensive and we only discuss closely related papers. The reader is referred to Bhagwati et. al. (1999) for a collection of many of the important papers in the area. 7 See also Grossman and Helpman (1995) for a political economy model of trade agreements wherein trade protection is a function of contributions made by agents to their respective governments. 8 By contrast, in our approach, all types of trade agreements are endogenously determined. 9 The gains from trade stem from reduced market power in the domestic industry. To this end, the monopoly assumption is not crucial but is the simplest way to represent market power. 5 Furthermore, u(xi ) is assumed to be quadratic: u(xi ) = αi Xi − Xi2 2 P where xi ≡ (xii , xji , xki ) is country i’s consumption profile, and Xi ≡ 3j=1 xji . Note that xji is country i’s consumption of country j’s product (or denotes the output sold by country j’s firm in country i), Due to the quasilinear nature of the utility function, country i’s inverse demand function is given by: X3 xji (2) Pi (xi ) = αi − j=1 where αi represents the market size of each country. Let country 1 have the largest market for good x and country 3 the smallest: α1 ≥ α2 ≥ α3 (3) Consider a three stage game. In the first stage, each country announces a set of countries with whom it wants to form an agreement. An FTA between two countries is formed if both countries want the agreement. Similarly, multilateral free trade emerges only when all countries want such an agreement. Next, countries simultaneously choose their tariffs given the agreement(s) formed in the first stage. Finally, firms compete in the product market in a Cournot fashion and markets are segmented so that each firm makes independent decisions regarding how much to sell in each market (as in Brander and Krugman (1983) and Brander and Spencer (1984)). In the absence of any trade agreement, firm j faces a specific tariff tji when exporting to country i.10 Denote the vector of the tariff schedule of country i by ti ≡ (tji , tki ).11 Since Article I of GATT forbids tariff discrimination, in the absence of any FTAs we restrict attention to the case where tji = tki = ti for i, j, k. In section 5, we evaluate the contribution of MFN to trade liberalization by allowing countries to tariff discriminate even in the absence of any FTAs.12 10 It is obvious that tii = 0 for all i . We also assume that the countries do not impose tariffs on the numeraire good that may be traded internationally in order to balance trade. 12 Note that, relative to discrimination, MFN treatment not only changes the optimal tariffs in the absence of any trade agreements but also changes non-member countries’ optimal tariffs under a bilateral FTA. 11 6 Country j’s effective marginal cost of exporting to country i, denoted by cji , equals: cji = cj + tji (4) where cj denotes the marginal cost of production of country j’s firm. Let country 1 be the lowest cost producer and country 3 the highest: c3 ≥ c2 ≥ c1 (5) Country j’s profit function for exports to country i, denoted by Πji , can be written as: Πji = xji Pi (Xi ) − cji xji (6) First order conditions for profit maximization yields equilibrium output levels:13 P αi − 3ci + j,j6=i cji αi + ci − 3cji + cki and xji = , k 6= i, j. (7) xii = 4 4 Similarly, equilibrium price in country i is easily calculated: P αi + ci + j,j6=i cji Pi = 4 (8) Using (7) and (8), equilibrium profits are: Πii = x2ii and Πji = x2ji (9) The following comparative statics are standard: ∂Πji ∂Πii ∂Πki 3xji xii xki < 0, > 0, and > 0. =− = = ∂tji 2 ∂tji 2 ∂tji 2 (10) Country i’s welfare function, Wi consists of four components: domestic consumer surplus Ui (Xi ) − Xi Pi (Xi ), domestic firm’s profits in home market (Πii ), domestic firm’s export profits (Πij , j 6= i) and tariff revenue: X X Wi = Ui (Xi ) − Xi Pi (Xi ) + Πii + Πij + tji xji (11) j6=i 13 j6=i In order to guarantee market access to all exporting firms in any country under both MFN and tariff discrimination, we assume the following condition 3c2 > 7c3 − 1. 7 At the second stage of the FTA game, countries simultaneously choose their tariffs in order to maximize their own welfare given the agreement(s) formed in the first stage. Since markets are segmented and marginal costs are constant, strategic independence of trade policies obtains. As a result, own tariffs do not affect export profits and each country chooses an optimal tariff schedule, ti ≡ (tji , tki ) in order to maximize: X tzi xzi (tji , tki ) (12) max CSi (tji , tki ) + Πii (tji , tki ) + z6=i Next, we derive optimal tariffs under various FTAs and explore their formation. 3 Formation of trade agreements We have two motivating questions. First, does the fact that countries can form FTAs make global free trade less likely to occur? Second, how does the variation in market size and production cost across countries affect their incentives for forming FTAs? To address these questions, we begin by deriving the sub-game perfect Nash equilibria (SPNE) of the two games. However, to allow countries to deviate as a coalition, we also analyze Coalition Proof Nash equilibria (CPNE) of the two games. While the model permits countries to be asymmetric along two dimensions (production costs and market size), to highlight the role played by each type of asymmetry, we consider each in isolation. 3.1 Equilibrium FTAs: the role of market size From here till section 4, to focus on market size asymmetries we set ci = 0 for all i. In addition, the following normalization is imposed throughout the paper:14 α1 ≥ α2 = 1 ≥ α3 (13) Under the status quo (i.e. no trade agreement), each country imposes its optimal MFN tariff on others. The optimal MFN tariff for country i is found 14 Similarly, from section 4 onwards we focus on cost asymmetries and set α1 = α2 = α3 = 1. 8 by imposing the constraint tji = tki on the above problem and it is given by15 ti (Φ) = 3αi 10 If countries i and j form the FTA (ij ), they abolish tariffs on each other and choose their tariffs on country k’s independently. As a result, under (ij ), the problem in (12) is subject to the constraint tji = 0. The following optimal tariff levels are easily calculated under the different FTAs: tji (jk) = tki (jk) = tji (Φ) = 3αi 10 (14) and αi (15) 7 As might be expected, when an FTA member i reduces its tariff on the other member j to zero, exports of the non-member k decrease. As a result, compared to the case of no agreement (Φ), a members incentive to impose a tariff on the non-member decreases since the non-member becomes a less important source of rent-extraction. This result is known as the tariff complementarity effect in the literature (see Bagwell and Staiger 1997a, 1997b, and 1998a).16 tji (ik) = tji (ki, kj ) = 3.1.1 SPNE of the FTA game Under the FTA game, the strategy set of country i, denoted by Si , consists of four duples of strategies (si ) that involve countries with whom i wants to form an agreement: Si = {{φ, φ}, {j, φ}, {φ, k}, {j, k}} (16) where φ represents country i announcing in favor of ‘no agreement’. In order to conserve notation, each agreement is denoted as follows: (i) Status quo Φ 15 It turns out that even if a country is free to tariff discriminate, under market size asymmetry each country imposes a common tariff on the other two. However, the constraint implied by MFN binds under cost asymmetry (see section 5 for further discussion). 16 A recent report by the World Bank (2000) argues that no clear evidence shows that the formation of a PTA leads member countries to become more protectionist towards non-member countries. While such an outcome may be a consequence of Article XXIV, the tariff complementarity result implies that it could also be optimal for member countries to voluntarily lower tariffs on non-members. 9 — obtains when no two announcements match or when all announce φ; (ii) a bilateral FTA between countries i and j — (ij ) is formed iff two countries announce each other’s names j si and i sj ; (iii) two independent FTAs in which i is the common member — (ij, ik) is formed iff (a) j si and i sj and (b) k si and i sk ; and (iv) free trade — (123 ) obtains iff all countries announce each others’ names.17 As is clear from the discussion above, member of an FTA can sign an independent FTA with the non-member without needing consent of the other member.18 Consider the following announcements: s1 = {2, φ}, s2 = {1, 3}, s3 = {φ, 2} (18) The above strategy vectors give rise to two independent FTAs: (21 ) and (23 ) of which country 2 is the common member. Note also that different strategy vectors may yield the same agreement(s) when there is an asymmetry in terms of countries’ choices. To see this, suppose the strategy vector is given by: s1 = {2, 3}, s2 = {1, 3}, s3 = {φ, 2} (19) Even though country 1 announces country 3, country 3 wants to form an FTA only with country 2.19 As a result, strategy vectors in (19) yields the same agreements (21, 23 ) as the ones in (18). In order to eliminate redundant announcements in the set of SPNE, it is assumed that FTA announcements cost each country ε (where ε > 0 is arbitrarily small). The following lemma summarizes the incentives of countries to form a bilateral FTA: Lemma 1: Country i’s incentive to form a bilateral FTA with country j is increasing in the market size of country j whereas it is decreasing in its own market size. 17 Formally, free trade obtains iff s1 = {2, 3}, s2 = {1, 3}, s3 = {1, 2} 18 (17) As indicated in Furusawa and Konishi (2003), this distinction creates an important difference between an FTA and a CU and leads to a sharp contrast to Yi (1996). 19 Under the open membership rule by Yi (1996), membership is open to all countries. However, this rule does not seem to be appealing for discussions of PTAs since the formation requires consent from both sides. 10 The above lemma implies that the smallest country (3) always has an incentive to form an FTA with the other two countries. Similarly, the medium size country (2) also benefits from an FTA with the largest country (1). The trade-off underlying FTAs is as follows: From each country’s perspective, joining an FTA is costly because its domestic surplus is lowered relative to the case where it is free to use its optimal MFN tariff.20 On the other hand, being part of an FTA increases export profits of members in each other’s markets. Consequently, the size of the larger member country plays a crucial role in determining which bilateral FTAs can arise in equilibrium. What type of FTAs arise in equilibrium? We can show that unless country j is too large relative to country i the bilateral FTA (ij ) is a SPNE: √ (ij ) is SPNE for all αj s.t. 70αj ≤ 3 910αi (20) From the perspective of countries 2 and 3, given that they call each other’s names, the marginal gain of gaining free access to an additional market (i.e. that of country 1) dominates the loss in domestic surplus from having an additional FTA. As a result, when the smaller countries (2 and 3) call each other’s name, they also announce the name of the largest country (1) in order to form two independent FTAs. However, the largest country (1) has an incentive to unilaterally deviate from such a pair of bilateral FTAs (in which it itself is not the common member). This can be explained as follows: Since the FTA partner for country 1 is the common member country (2 or 3), country 1 does not enjoy any competitive advantage over the non-partner country implying that increase in its export profit from the FTA is small. Moreover, country 1 benefits from the tariff complementarity effect when it is not a member of an FTA. W3 (31, 32 ) ≥ W3 (23 ) but W1 (23 ) ≥ W1 (31, 32 ). (21) W2 (21, 23 ) ≥ W2 (23 ) but W1 (23 ) ≥ W1 (21, 23 ). (22) Similarly As a result, the two smaller countries (2 and 3) cannot form two independent FTAs (unless it is global free trade) so that (31, 32 ) and (21, 23 ) are 20 Note that it is not obvious that an FTA is worse for domestic surplus relative to MFN since even under MFN, a country is solving a constrained optimization problem in that it must treat its trading partners the same. However, when two countries are symmetric with respect to cost, it is optimal for the third country to not tariff discriminate so that the optimal MFN tariff is in fact its unrestricted optimal tariff. 11 never SPNE. The only possible FTA that involves two independent bilateral FTAs is the one in which the largest country is the common member (12, 13 ). Finally, global free trade (123 ) is a SPNE if all countries are relatively symmetric with respect to their market size. We summarize these results in the following proposition: Proposition 1: FTAs (12), (13), (23), (12,13), (123) and status quo ( Φ), are all SPNE iff the degree of asymmetry between countries is not too high (i.e. iff α3 ≥ α3 and α1 ≥ α1 ≥ α1 ). When countries are completely symmetric with respect to market size (i.e. αi = 1 for all i), all FTAs are SPNE except for those that involve two independent FTAs.21 Why cannot two independent FTAs such as (12,13 ) arise in equilibrium when countries are completely symmetric? This is because each member country that belongs to only a single FTA (i.e. both countries 2 and 3) has an incentive to unilaterally deviate from its FTA with country 1 to take advantage of the tariff complementarity effect — the gain from tariff reductions a non-member enjoys comes at no cost whereas the gain of free market access in country 1 requires granting free access in return. In effect, each country with a single FTA prefers to free ride on the other two and is better off if the other two form an FTA rather than be part of one itself! Given that there exist multiple SPNE (including status quo Φ), we refine the set of multiple SPNE by deriving Coalition-Proof Nash equilibrium (CPNE) of the FTA game. While the SPNE concept defines equilibrium only in terms of unilateral deviations, CPNE eliminates all non-credible deviations by every conceivable coalition.22 3.1.2 Stable FTAs We first clarify how the notion of CPNE applies in our model and then discuss stable FTAs. 21 Explicit formulae for these market size thresholds are reported in the appendix. As per Bernheim et al. (1987), an FTA is defined to be coalition-proof if and only if it is Pareto efficient among self enforcing agreements. In turn, an FTA is self enforcing if and only if no proper subset (coalition) of countries, taking the actions of its complement as fixed, can agree to deviate in a way that makes all of its members better off. 22 12 The notion of CPNE In order to see how the concept of CPNE is applied in our context, consider the following example: s1 = {2, φ}, s2 = {1, φ}, s3 = {φ, φ} (23) Suppose now that (12 ) is a SPNE and it is immune to any conceivable deviation except for the deviation to (23 ):23 W2 (23 ) > W2 (12 ) and W3 (23 ) > W3 (12 ). (24) In order for (12 ) to be a CPNE, it is necessary to check whether the joint deviation of countries 2 and 3 from (12 ) to (23 ) is credible or not. If there is no incentive for any coalition of countries to further deviate from (23 ), the joint deviation from (12 ) to (23 ) would be credible and (12 ) would not be a CPNE. But what if country 2 — a subset of the countries that deviate from (12) to (23) — do indeed has incentive to deviate further from (23) to (21, 23) taking country 1’s strategy as given (s1 = {2, φ} in the original agreement (12 ))? Suppose W2 (21, 23 ) > W2 (23 ). (25) Then the original deviation from (12 ) to (23 ) would be non-credible. As a result, the FTA (12 ) would be a CPNE. Further note that there can be multiple FTAs that are CPNE simultaneously. Consider the above example again and suppose now that (23 ) is immune to any conceivable deviations except for the joint deviation of countries 1 and 3 to (13 ): W1 (13 ) > W1 (23 ) and W3 (13 ) > W1 (23 ) (26) Suppose further that country 3— a subset of the countries that deviate from (23 ) to (13 )— has an incentive to deviate further from (13 ) to (31, 32 ), taking country 2’s strategy s2 = {φ, 3} in the original agreement (23 ) as given: W3 (31, 32 ) > W3 (13 ) (27) Under such a situation, (12 ) and (23 ) are both CPNE. Finally, following Dutta and Mutuswami (1997), we call an equilibrium FTA stable if it corresponds to a CPNE.24 23 24 It is clear that (12 ) is a SPNE since it is immune to unilateral deviations. Furusawa and Konishi (2002) use a network formation model to endogenize free trade 13 Stable FTAs under market size asymmetry Figure 1 shows stable FTAs in the (α3 , α1 ) space:25 ––Insert figure 1 here–– As seen from figure 1, when the largest country (1) is too large relative to other countries, it does not have an incentive to form an FTA because of the unfavorable trade-off between loss in domestic surplus and gain in export profits. In this situation, the only possible stable agreement is (23 ) unless the smallest country (3) is too small relative to the medium size country (2). Otherwise, no agreement (Φ) is a CPNE. Similarly, if the smallest country (3) is too small, neither country 1 nor country 2 has an incentive to form an FTA with country 3. In this case, if the size of the largest country (1) is similar to the one of the medium size country (2), (12 ) is the stable equilibrium. The following is immediate: Lemma 2: A bilateral FTA between the largest and smallest countries (13) is not stable. It is important to note that only the bilateral FTAs that include the similar size countries, (12 ) and (23 ), are stable. The FTA (13 ) is not stable since 1 and 2 have incentives to deviate from (13 ) to (12, 13 ). This deviation is not credible only within the market size parameter range over which the deviation from (13 ) to no agreement (Φ) is credible. As a result, (13 ) is never a CPNE. From inequalities in (21) and (??), since both (31, 32 ) and (21, 23 ) are not SPNE, they are trivially not in the set of CPNE as well. Moreover, (12, 13 ) is stable if countries 2 and 3 are not too small relative to the largest country (1). Most importantly, free trade (123 ) obtains if and only if market sizes of all countries are similar — in fact, free trade is the unique CPNE when countries are completely symmetric.26 agreements (FTAs). Given any FTA configuration, they examine whether a pair of countries have incentives to sign another FTA, and whether a country has an incentive to cut an existing FTA. A network that is immune to these two incentives is called pairwise stable. 25 Throughout the paper, explicit functions used to obtain the figures can be found in the appendix. Also we limit the range of country 1’s market size as follows while drawing the figures for the case of market size asymmetry: 2 ≥ α1 26 A similar result is also obtained by Furusawa and Konishi (2002) and they conclude that FTAs are building blocs for free trade. However, our approach is different from 14 3.2 Eliminating the FTA option Under the no FTA option, bilateral agreements are not possible so countries can either announce in favor of the status quo or in favor of free trade. The strategy set of country i, denoted by Si , consists of two duples of strategies (si ): Si = {{φ, φ}, {j, k}} (28) Free Trade (123 ) occurs if and only if the following hold:27 W1 (123 ) > W1 (Φ), W2 (123 ) > W2 (Φ) and W3 (123 ) > W3 (Φ). (29) The set of CPNE of the no FTA game is represented in figure 2. ––Insert figure 2 here–– The comparison of figures 1 and 2 leads to the following proposition: Proposition 2: Under both games, free trade (123) is found to be uniquely stable if and only if countries are relatively symmetric with respect to market size. However, free trade (123) is stable over a larger parameter space under the no FTA game than under the FTA game. As seen from figure 2, when countries do not have the option to form FTAs, global free trade (123 ) is stable in the light and dark blue regions while it is stable only in the dark blue region under the FTA game. As a result, free trade is more likely to obtain under the no FTA game. Over the light blue region, countries prefer free trade (123 ) to no agreement (Φ). However, an FTA option provides them an intermediate opportunity to form a trading bloc that is preferable to free trade (123 ). Further analysis of the two games yields some subtle results regarding the welfare implications of FTAs in an asymmetric world. Two distinct regions need to be emphasized. In region I (12 ), (23 ), and (12, 13 ) are stable agreements under the FTA game while free trade (123 ) is the stable agreement Furusawa and Konishi (2002) in an important way: Beside investigating whether FTA formation continues until free trade or not, we also ask whether the option of forming FTAs makes global free trade less likely to occur. 27 Note that the set of SPNE and CPNE are identical in the no FTA option game. 15 under the no FTA game. In region II, (12 ), (23 ), and (12, 13 ) are stable agreements under the FTA game, whereas no agreement (Φ) is the stable outcome under the no FTA game. Note that the former (latter) region refers to the situation in which countries are relatively symmetric (asymmetric) with respect to the market size. Given these two distinct regions, we explore the following questions: Which countries benefit (suffer) the most from the FTA option and what are the world welfare implications of FTAs? Proposition 3: A comparison of the FTA game and the no FTA game yields the following results: (i) over region I, the FTA option has beneficial welfare effects for the largest country (1) but harmful effects for the smallest country (3) as well as for the world as a whole; (ii) the welfare impact on the medium size country (2) is ambiguous and it depends on the size distribution of countries: if the smallest country (3) is quite small and the medium size country (2) is similar in size to the largest country (1), the FTA option has beneficial welfare effects for the medium sized country; however, if the largest country (1) is too large and the medium size country (2) is similar in size to the smallest country (3), the FTA option is harmful for the medium sized country (2); and (iii) over region II, the FTA option has beneficial welfare effects for all countries (and therefore for the world as a whole). Proposition 3 implies that the largest country necessarily benefits from the FTA option. As indicated in lemma 1, smaller countries always have incentive to form an FTA with the largest country (1). This gives the largest country (1) a special position in the FTA game that it ends up exploiting in equilibrium in that stable FTAs are those that benefit the largest country (1). On the other hand, country 3 has the weakest position in the FTA game since it offers free access to the smallest market. This makes the no FTA game preferable for the smallest country (3) over region I. Note also that, over region I, if the smallest country (3) is too small relative to the medium size country (2) and the largest country (1) is similar in size as the medium size country (2), the stable FTAs are (12 ) or (12, 13 ) both of which are superior for the medium size country relative to free trade (123 ). However, if the largest country (1) is too large relative to the medium size country (2) and the smallest country (3) has a similar size as the medium size country (1), the resulting FTA is (23 ), which is inferior for the medium size country (2) relative to free trade (123 ). Furthermore, from the world welfare point of 16 view, free trade is always preferable to stable FTAs over region I. However, in region II, all forms of stable FTAs are preferable to no agreement (Φ). As a result, when countries are relatively symmetric (asymmetric) with respect to the market size, FTA game has harmful (beneficial) impacts on the world welfare. We next compare the two games under cost asymmetry. 4 How cost asymmetry matters As in the case of asymmetric market sizes, we investigate the following questions under cost asymmetry. First, how does a country’s incentive to form FTAs depends upon its own cost of production relative to others? Second, does the option of pursuing FTAs undermine incentives for multilateral trade liberalization? We begin with the FTA game and then analyze the no FTA game. Throughout the analysis we assume that αi = 1 for all i and we normalize c1 = 0. 4.1 Equilibrium FTAs As per Article I of GATT (MFN), we impose the following constraint on the problem in (12) under no agreement (Φ): tji = tki = ti (30) As a result, the optimal MFN tariff (ti (Φ)) under no agreement (Φ) is: P 3 − j cj (31) ti (Φ) = 10 The following tariff levels are found under the different FTAs: tji (jk) = tki (jk) = 6 − 2ci + 3ck − 7cj 20 and tji (ik) = tji (ki, kj ) = 17 3 − 9cj + 7ck − ci 21 (32) (33) 4.1.1 Multiplicity of SPNE Recall that a country’s incentive to form an FTA is determined by two countervailing forces: an FTA increases export profits but it lowers domestic surplus. If country i’s production cost is sufficiently lower than that of country j then country j’s loss in domestic surplus from the FTA (ij) outweighs the beneficial effects of improved access to country i’s market and the FTA (ij) is not a SPNE. Lemma 3: Country i’s incentive to form an FTA with country j is increasing in country j’s cost whereas it is decreasing in own cost and that of country k. The above lemma implies that the low cost countries always has an incentive to form an FTA with higher cost ones and the higher their cost the stronger is this incentive. Given the above incentives, the highest cost country’s (3’s) decision is the crucial one in determining equilibrium FTAs. Moreover, once country i announces country j ’s name, it has always has an incentive to independently announce country k’s name as well. This is because an additional FTA results in an increase in the export profits due to the better competitive position country i enjoys against each FTA partner in the other partner’s market and this positive impact is always larger than the additional loss in domestic surplus caused due to the second FTA: W1 (12, 13 ) ≥ W1 (12 ) and W1 (12, 13 ) ≥ W1 (13 ) (34) W2 (21, 23 ) ≥ W2 (12 ) and W2 (21, 23 ) ≥ W2 (23 ) (35) W3 (31, 32 ) ≥ W3 (13 ) and W3 (31, 32 ) ≥ W3 (23 ) (36) However, country 3 has an incentive to unilaterally deviate from an agreement that involves two independent FTAs in which the other two countries are common members since country 3 is at a competitive advantage — it does not gain much in its export markets while it loses a lot in its own market: W3 (12 ) ≥ W3 (12, 13 ) and W3 (12 ) ≥ W3 (21, 23 ) for all c2 , c3 (37) Consequently, inequalities in (34), (35), (36) and (37) together imply that (12, 13 ) and (21, 23 ) are never SPNE. The only possible agreement that involves two independent FTAs is the one in which the country 3 is the common member (31, 32 ). 18 Proposition 4: FTAs (12), (13), (23), (31, 32) and (123) and status quo ( Φ), are all SPNE iff the degree of cost asymmetry between countries is not too high (i.e. iff c3 ≤ c3 and c2 ≥ c2 ≥ c2 ). Furthermore, if countries are completely symmetric with respect to cost, all FTAs are SPNE except for those that involve two independent FTAs. Two independent FTAs cannot arise because when countries are completely symmetric, the countries that belong to only FTA have a unilateral incentive to become a non-member: once again, the tariff complementarity effect that these countries enjoy when they are not a member is greater than the gain from forming an FTA. Next, we describe stable FTAs under cost asymmetry. 4.1.2 Stable FTAs Figure 3 shows stable FTAs under cost asymmetry in the (c3 , c2 ) space. ––Insert figure 3 here–– The following corollary is immediate: Corollary 1: No agreement ( Φ), and FTAs (23) and (31, 32) are not in the set of CPNE although they belong to the set of SPNE. In other words, agreements that involve two independent FTAs are never a CPNE. Recall that (12, 13 ) and (21, 23 ) are not even SPNE, so that they cannot be CPNE as well. If (31, 32 ) is formed, the joint deviation of countries 1 and 2 to free trade (123 ) is credible unless country 1 is too efficient relative to country 2. Moreover, if the country 1 is too efficient relative to country 2, then country (2) has an incentive to deviate form (31, 32 ) to (13 ) and this deviation is credible. Consequently, (31, 32 ) is never a stable agreement. Furthermore, the deviations from no agreement (Φ) to (12 ) and (23 ) are credible over the entire region so that no agreement (Φ) is never a stable equilibrium although it is a SPNE. Finally, the deviation of countries 1 and 2 deviate from (23 ) to (21, 23 ) is credible unless they are almost symmetric with respect to their production costs. When all countries have relatively similar costs, they jointly deviate from (23 ) to global free 19 trade (123 ) and this is a credible deviation since (123 ) is immune to any conceivable deviations over this range. As a result, (23 ) is never a CPNE. Figure 3 demonstrates that over the dark blue region, there are two CPNE: (12 ) and (123 ). This region is determined by two non-credible deviations. First, the joint deviation of countries 1 and 3 from (12 ) to (13 ) is not credible since country 1 always has an incentive to deviate further to (12, 13 ) (see inequality in (36)). Second, the deviation from (12 ) to (123 ) is not credible since countries 2 and 3 have incentives to deviate further to (23 ). As a result, (12 ) is a CPNE. Finally, over the same region, the only possible deviation from free trade (123 ) is the joint deviation of countries 2 and 3 to (23 ) but this turns out to be non-credible since country 3 has an incentive to further deviate to (31, 32 ). Over the green region, there are two CPNE: (12 ) and (13 ). Three credible deviations from (13 ) determine the boundaries of the green region: (i) Countries 2 and 3 deviate to (31, 32 ), (ii) Country 3 deviates to no agreement (Φ) and (iii) All countries deviate to free trade (123 ). The yellow region is determined only by two credible deviations: (i) Country 2’s deviation to no agreement (Φ) and (ii) joint deviation of all countries to free trade (123 ).28 It is important to note that the region over which free trade (123 ) is stable is determined by the unilateral deviation of country 3 from (123 ) to (12 ). As a result, country 3’s choice is pivotal for the stability of free trade (123 ). Next, we examine which agreements arise in the CP NE if countries do not have an FTA option. 4.2 No FTA game Figure 4 shows stable FTAs under cost asymmetry in the (c3 , c2 ) space. ––Insert figure 4 here–– A comparison of figures 3 and 4 shows that if countries can not form FTAs, free trade is achieved in the light blue region in figure 4; while it is not the case under the FTA option. Thus, we have the following result. 28 It is important to note that the region over which free trade (123 ) is stable is determined by the unilateral deviation of country 3 from (123 ) to (12 ). As a result, country 3’s choice is pivotal for the stability of free trade (123 ). 20 Proposition 5: Under both games (FTA and no FTA), free trade (123) is uniquely stable if and only if the countries are relatively symmetric with respect to their production costs. However, even under cost asymmetry, free trade (123) is stable over a larger parameter space under the no FTA game than under the FTA game. The analysis of figure 3 and figure 4 provides additional results regarding the welfare implications of FTAs. Before doing that, it is important to note again that (12 ) and (13 ) are the stable agreements under the FTA option over the yellow region and the theory offers no guidance as to which agreement ((12 ) or (13 )) might actually be observed. We suppose that the ‘natural’ trading block (12 ) is the one that obtains as the stable agreement.29 Two distinct regions should be examined: region I: (12 ) is the stable agreements under the FTA option while free trade (123 ) is achieved as the stable agreement under the no FTA option, region II: (12 ) is the stable agreement under the FTA option whereas no agreement (Φ) is the stable outcome under the no FTA option. Given these distinct regions, we explore the following questions: Do the efficient or inefficient countries benefit from the FTA option and what are the world welfare implications of having an FTA option? Proposition 6: A comparison of the FTA game and the no FTA game yields the following results: (i) over region I, the option to form FTAs has beneficial welfare effects for the highest cost country (3) but harmful welfare effects for the other two countries as well as the world as a whole. (ii) over region II, the FTA option has beneficial welfare effects for all countries (and therefore the world as a whole). Proposition 6 implies that the highest cost country always benefits from the FTA option. This result reflects the fact that market access to the highest cost country does not come at too high a price for the lower cost countries who are at a competitive advantage. However, this stronger incentive of low cost countries actually turns into an advantage for the high cost country who ends up exploiting it in equilibrium. In fact, for the least efficient country, (12 ) 29 If the stable agreement we observe over yellow region is (13 ), we have three distinct regions: region I: (13 ) is the stable agreement under the FTA game while free trade (123 ) is achieved as the stable agreement under the no FTA game, region II: (12 ) is the stable agreement under the FTA option while free trade (123 ) is the stable one under the no FTA option, and region III: (12 ) is the stable agreement under the FTA option whereas no agreement (Φ) is the stable outcome under the no FTA option. 21 is always preferable to the stable agreements (123 ) and (Φ) under the FTA option and no FTA option respectively. However, over region I, countries 1 and 2 prefer free trade (123 ) that is not stable under the FTA option. As a result, these countries are worse off with the FTA option over region I and it is the lowest cost country that loses the most from the FTA option. Over Region II, when (12 ) is a CPNE, since there is no unilateral deviation of countries 1 and 2 to Φ, they prefer (12 ) to Φ. Because of the tariff complementarity effect, 3 also prefers (12 ) to Φ. Consequently, world welfare is higher under (12 ) than under Φ. Next, we examine the role of the most favored nation (MFN) principle in the formation of trade agreements. 5 Role of MFN in trade agreements If Article I (MFN) were not applicable, the problem in (12) under no agreement (Φ) is no longer subject to the symmetric treatment constraint in (31) and countries are free to tariff discriminate. The following optimal discriminatory tariff (tji (Φ)) is obtained under no agreement (Φ): tji (Φ) = 6 − 2ci + 3ck − 7cj 20 It is important to note that inefficient (efficient) exporters benefit (suffer) from tariff discrimination relative to MFN: ck − cj (38) tji (Φ) − ti (Φ) = > 0 iff ck > cj 4 It is straightforward to argue that a non-member country’s welfare under a bilateral FTA is higher when its trade policy is not constrained by MFN. The following lemma explains the role of MFN in the two games: Lemma 4: A country’s incentive to deviate from any agreement to no agreement ( Φ) or to the ones in which it is a non-member is higher under tariff discrimination relative to MFN. The above lemma argues that the region over which the agreements that are stable under MFN treatment changes. Figure 5 shows stable agreements under the FTA option when countries are free to discriminate. 22 ––Insert figure 5 here–– Similarly, stable agreements under the no FTA option are represented in figure 6. ––Insert figure 6 here–– A Comparison of figures 3-6 yields the following result: Proposition 7: MFN contributes to trade liberalization in two ways: (i) Under both games, the region over which free trade (123) is stable is larger under MFN than under tariff discrimination; (ii) under the FTA game, over the region where free trade is not stable, (12) and (13) are stable over a larger cost parameter space under MFN relative to tariff discrimination. Since country 3 is pivotal and its welfare under (12 ) is higher under discrimination than under MFN, the region over which free trade (123 ) is stable is larger under MFN treatment. The same argument is valid for the no FTA option as well since the deviation of country 3 from free trade (123 ) to (Φ) determines the region over which free trade (123 ) is stable. Over the region where free trade is not stable, since the deviations of the less efficient member countries (2 and 3) from (12 ) and (13 ) to no agreement (Φ) are binding for the stability of these bilateral FTAs, the region over which they are stable is larger under MFN relative to tariff discrimination. The following insight emerges from the above analysis: Article I (MFN treatment) facilitates the achievement of global free trade independent of the role of Article XXIV. However, when countries are not allowed to practice Article XXIV (no FTA option), the impact of MFN on the stability of free trade is relatively larger. 6 Concluding remarks This paper contributes to the long-standing debate regarding the effect of FTAs on multilateral trade liberalization by analyzing two games: one where countries have the choice to pursue both FTAs and global free trade and another where they cannot form FTAs. We find that the option to pursue 23 FTAs does lower the likelihood of achieving global free trade. In this sense, the exception made available by Article XXIV is in conflict with the GATT’s main goal of achieving global free trade. However, since both types of trade liberalization is endogenous in our model, we also find that the option to form FTAs leads to welfare improving trade liberalization when global free trade cannot be achieved. In this sense, Article XXIV supports GATT’s main goal — it can be better to have some preferential trade when multilateral liberalization is infeasible. In fact, the underlying asymmetry in our model delivers a surprising insight: the option to pursue FTAs can actually lead to an outcome that is welfare preferred to free trade when the FTAs formed favor low cost producers relative to high cost ones. Our analysis further sharpens the stumbling block versus building block debate by highlighting conditions under which the answer goes one way or another. When the underlying asymmetry between countries is relatively small, the option to pursue FTAs does more harm than good. On the other hand, when countries are quite asymmetric, multilateral free trade is harder to obtain and the FTAs are actually be desirable from a world welfare perspective. Finally, we show that Article I (MFN treatment) facilitates the achievement of global free trade independent of the role of Article XXIV. However, when countries are not allowed to practice Article XXIV (no FTA option), the impact of MFN on the stability of free trade is relatively larger. 7 Appendix All supporting calculations and proofs not provided in the text are given below. Market Size Asymmetry First, we report the welfare levels of countries under different types of agreements for the general case: Wi (Φ) = 40α2i + α2j + α2k 100 35α2i + 8α2j 40α2k α2i + α2j α2k Wi (ij ) = + , Wk (ij ) = + 98 100 100 49 2 2 2 2 4(αj + αk ) 5α2j α2k 11αi α + , Wj (ij ), (ik) = i + + Wi (ij ), (ik) = 32 49 16 14 49 24 (39) (40) 11α2i + 2(α2j + α2k ) (41) Wi (ijk) = 32 In the following proofs, the critical α1 and α3 levels can be found following the proof of proposition 3. Lemma 1 Ii (ij ) is defined as the incentive of a country i to form an FTA with a larger country j : Accordingly, the following is immediate: Ii (ij ) = Wi (ij ) − Wi (Φ) = As a result: 351α2j − 210α2i 4900 (42) ∂Ii (ij ) ∂Ii (ij ) > 0 and <0 ∂αj ∂αi Proposition 1 Part (i): When countries are asymmetric (α1 > α2 = 1 > α3 ): (12 ) is SPNE for all α1 s.t. α1 ≤ αa1 (13 ) is SPNE for all α1 s.t. α1 ≤ αa1 α3 (23 ) is SPNE for all α3 s.t. α3 ≥ αa3 (12, 13 ) is SPNE if a-) αc3 ≥ α3 ≥ αf3 for all α1 s.t. αe1 ≥ α1 ≥ αh1 and b-) α3 ≥ αc3 for all α1 s.t. αd1 ≥ α1 ≥ αh1 and finally, (123 ) is SPNE if α3 ≥ αe3 for all α1 s.t. α1 ≤ αb1 . As a result, (Φ), (12 ), (13 ), (23 ), (12, 13 ) and (123 ) are SPNE simultaneously if α3 ≥ α3 and α1 ≥ α1 ≥ α1 and this condition is restated as the following combination: a-) αg3 ≥ α3 ≥ αa3 αh1 for all α1 s.t. α1 ≤ αa1 α3 25 and b-) α3 ≥ αg3 for all α1 s.t. αb1 ≥ α1 ≥ αh1 . Part (ii): When countries are completely symmetric (α1 = α2 = α3 = 1); (ij) is a SPNE since no member has an incentive to deviate unilaterally: Wi (ij ) − Wi (Φ) = Wj (ij ) − Wj (Φ) = 141 >0 4900 (123) is a SPNE since no country has an incentive to deviate unilaterally: Wi (123 ) − Wi (Φ) = 39 >0 800 Figure 1 Figure 1 represents the following set of CPNE:30 (i) Φ is a CPNE if α3 ≤ αa3 for all α1 s.t. α1 ≥ αa1 . (ii) (12 ) is a CPNE if α3 ≤ αb3 for all α1 s.t. α1 ≤ αa1 . (iii) (23 ) is a CPNE if α3 ≥ αa3 for all α1 s.t. α1 ≥ αb1 . (iv) (12, 13 ) is a CPNE if (a) αc3 ≥ α3 ≥ αb3 for all α1 s.t. α1 ≤ αc1 and (b) αd3 ≥ α3 ≥ αc3 for all α1 s.t. α1 ≤ αd1 . (v) (123 ) is a CPNE if α3 ≥ αe3 for all α1 s.t. α1 ≤ αb1 . Lemma 2 Given 2 ≥ α1 ≥ α2 = 1 ≥ α3 : (13 )−→(1 and 2)−→(12, 13 ): 1 and 2 always deviate: W1 (12 , 13 ) − W1 (13 ) ≥ 0, W2 (12 , 13 ) − W2 (13 ) ≥ 0 and 2 has no incentives to deviate further but 1 has an incentive to deviate further to Φ if α1 ≥ αd1 and to (12 ) if α1 ≥ αe1 . Consequently, the initial deviation of countries 1 and 2 from (13 ) to (12, 13 ) is not self enforcing if (a) α3 < αc3 and α1 ≥ αe1 and (b) α3 ≥ αc3 and α1 ≥ αd1 . However, over this region 1 deviates from (13 ) to Φ if α1 > αa1 and it is self-enforcing. As a result, (13 ) is never a CPNE. Figure 2 30 Detailed derivations are available upon request. 26 (i) Φ −→(1, 2 and 3 )−→(123 ): 3 always deviates, 2 deviates if α1 > αf1 and 1 deviates α1 > αg1 . As a result Φ is a CPNE if α1 > αg1 . (ii) (123 )−→(1 or 2 or 3 )−→ Φ: 3 never deviates, 1 deviates if α1 < αf1 and 2 deviates if α1 > αg1 . As a result, F T is a CPNE if α1 ≤ αg1 . Proposition 2 It is immediate from figures 1 and 2. Proposition 3 Over region I: When (12 ) is a CPNE (if α3 ≤ αb3 for all α1 s.t. α1 ≤ αa1 ): W1 (12 ) − W1 (123 ) = 15 21α23 3α21 + − >0 224 784 400 15α21 21α23 3 + − >0 224 784 400 9α2 33(1 + α21 ) W3 (12 ) − W3 (123 ) = 3 − <0 160 784 15(1 + A21 ) 39A23 W W (123 ) − W W (12 ) = + >0 1568 800 When (23 ) is a CPNE (if α3 ≥ αa3 for all α1 s.t. α1 ≥ αb1 ): W2 (12 ) − W2 (123 ) = W1 (23 ) − W1 (123 ) = 9α21 33(1 + α23 ) − >0 160 784 15α23 21α21 3 + − <0 224 784 400 15 21α21 3α2 W3 (23 ) − W3 (123 ) = 3 + − <0 224 784 400 15(1 + α23 ) 39α21 W W (123 ) − W W (23 ) = + >0 1568 800 When (12, 13 ) is a CPNE (if (a) αc3 ≥ α3 ≥ αb3 for all α1 s.t. α1 ≤ αc1 and (b) αd3 ≥ α3 ≥ αc3 for all α1 s.t. α1 ≤ αd1 ): W2 (23 ) − W2 (123 ) = W1 (12 , 13 ) − W1 (123 ) = 27 3α21 33α23 − >0 224 784 15(1 + α23 ) >0 784 33 3A23 W3 (12 , 13 ) − W3 (123 ) = − <0 224 784 15(1 + A23 ) W W (123 ) − W W (12 , 13 ) = >0 1568 W2 (12 , 13 ) − W2 (123 ) = Over region II: When (12 ) is a CPNE, since there is no unilateral deviation of countries 1 and 2 to Φ, they prefer (12 ) to Φ. Because of the tariff complementarity effect, 3 also prefers (12 ) to Φ. When (23 ) is a CPNE, since there is no unilateral deviation of countries 2 and 3 to Φ, they prefer (23 ) to Φ. Because of the tariff complementarity effect, 1 also prefers (23 ) to Φ. When (12, 13 ) is a CPNE, since there is no unilateral deviation of country 1 to Φ, it prefers (12, 13 ) to Φ, and the following holds: 21α21 51α23 3 + − >0 400 4900 70 21α21 51 3α2 W3 (12 , 13 ) − W3 (Φ) = + − 3 >0 400 4900 70 It is trivial that since the welfare of all countries rises, world welfare increases as well. W2 (12 , 13 ) − W2 (Φ) = Critical α1 and α3 levels: p p √ √ 330 + 330α23 3 910 b 6 182 d 2 390 + 390α23 a c α1 = , α1 = , α1 = , α1 = 70 21 35 35 p p √ √ 210 − 196α23 210 + 210α23 h 2 770 6 182α3 f g e α1 = , α1 = , α1 = , α1 = 35 14 15 55 √ √ √ √ 910 b 5 182 c 5 d 154 αa3 = , α3 = , α3 = , α3 = 39 156 4 22 r √ √ √ 4070 f 770 182 g 11 αe3 = , α3 = , α3 = 10 110 858 1357 Lemma 3 28 Given that 15 ≥ ck ≥ cj ≥ ci = 0 and 3cj > 7ck − 1, when Ii (ij ) = Wi (ij ) − Wi (Φ), Ii (ij ) −768 + 218cj + 669ck < 0, = ∂ci 3150 Ii (ij ) 3882 + 1471cj − 6879ck = >0 ∂cj 22050 and −156 − 2293cj + 657ck Ii (ij ) < 0. = ∂ck 7350 For the following proofs, the critical c2 and c3 levels can be found at the end of the appendix. Proposition 4 The proof of part (ii) is the same as the proof of part (ii) of the proposition 2. Part (i): When countries are asymmetric ( 15 ≥ c3 ≥ c2 ≥ c1 = 0 and 3c2 > 7c3 − 1): (12 ) is SPNE for all c2 s.t. c2 ≤ ca2 (13 ) is SPNE for all c2 s.t. c2 ≥ cb2 (23 ) is SPNE for all c2 s.t. c2 ≥ cc2 (31, 32 ) is SPNE for all c2 s.t. c2 ≤ cd2 and finally, (123 ) is SPNE for all c2 s.t. c2 ≥ ce2 . As a result, (Φ), (12 ), (13 ), (23 ), (12, 13 ) and (123 ) are SPNE simultaneously if c3 ≤ c3 ∼ = 0.095 and cd2 ≥ c2 ≥ ce2 . Figure 3 Figure 3 represents the following set of CPNE:31 a-) (12 ) is a CPNE if (i) c2 ≥ ce2 and c2 ≥ cf2 and (ii) c2 ≤ ce2 and c2 ≤ ca2 . b-) (13 ) is a CPNE if c2 ≤ ce2 , c2 ≥ cb2 and c2 ≥ cd2 . 31 Detailed derivations are available upon request. 29 c-) (123 ) is a CPNE if c2 ≥ ce2 . Corollary 1 (i) Φ −→(1 and 2)−→(12 ): 1 always deviates, 2 deviates if c2 < ca2 . As a result, the deviation happens if c2 < ca2 and it is self-enforcing. Similarly, Φ −→(2 and 3)−→(23 ): 2 always deviates, 3 deviates if c2 > cc2 . As a result, the deviation happens if c2 > cc2 and it is self-enforcing. These two conditions (c2 < ca2 , c2 > cc2 ) cover entire region so that Φ is never a CPNE. (ii) (23 )−→(1 and 2)−→(21, 23 ): 2 always deviates, 1 deviates if c2 > g c2 . As a result, the deviation happens if c2 > cg2 and it is self-enforcing. Furthermore, if c2 < cg2 , all countries deviate from (23 ) to free trade (123 ) and it is a self-enforcing deviation since free trade is strong Nash when c2 < cg2 . Consequently, (23 ) is never a CPNE. (iii) (31, 32 )−→(1 and 2)−→(123 ): 1 always deviates, 2 deviates if c2 < h c2 . As a result, the deviation happens if c2 > ch2 and it is self-enforcing when c2 < ch2 . Note also that when c2 > ch2 , it is not even a SPNE. Consequently, (31, 32 ) is never a CPNE. Figure 4 (i) Φ −→(1, 2 and 3 )−→(123 ): 1 and 2 always deviate, 3 deviates if c2 > 111c493 −13 . As a result Φ is a CPNE if c2 ≤ 111c493 −13 . (ii) (123 )−→(1 or 2 or 3 )−→ Φ: 1 and 2 never deviate, 3 deviates if c2 < 111c493 −13 . As a result, (123 ) is a CPNE if c2 ≥ 111c493 −13 . Proposition 5 It is immediate from figures 3 and 4. Proposition 6 Region I refers to the region over which c2 ≤ ce2 and c2 ≤ ca2 and (12 ) is a CPNE under FTA option while free trade (123 ) is a CPNE under no FTA option. Region II refers to the region over which c2 ≤ 111c493 −13 and (12 ) is a CPNE under FTA option while no agreement (Φ) is a CPNE under no FTA option. The following welfare comparison is immediate: Over Region I: W1 (123 ) > W1 (12 ) for all c2 and c3 W2 (123 ) > W2 (12 ) for all c2 and c3 W3 (123 ) < W3 (12 ) iff c2 ≤ ce2 30 and W W (123 ) > W W (12 ) for all c2 and c3 . Over Region II, when (12 ) is a CPNE, since there is no unilateral deviation of countries 1 and 2 to Φ, they prefer (12 ) to Φ. Because of the tariff complementarity effect, 3 also prefers (12 ) to Φ. Since welfare of all countries are higher under (12 ) than under Φ, so is the world welfare. Lemma 4 It is obvious that welfare of the non-member country under a bilateral FTA and also the welfare of countries under no agreement (Φ) is lower under MFN relative to discrimination since they are constrained to treat their importers symmetrically. The proof immediately follows this fact. Figure 5 The CPNE condition for (123 ) is determined by the deviation of the least efficient country (3) to (12 ). Since MFN only changes the welfare under Φ, we obtain the same condition for (123 ) to be a CPNE: (123 ) is a CPNE if c2 ≥ cj2 . On the other hand, (12 ) is a CPNE if (a) c2 ≥ cj2 and c2 ≥ ck2 , (b) c2 ≤ cj2 and c2 ≤ cm 2 . (13 ) is a CPNE if (a) c2 ≤ cj2 , c2 ≥ cn2 and c2 ≥ cp2 . Figure 6 Under discrimination: (i) Φ −→(1, 2 and 3 )−→(123 ): 1 always deviates, 2 deviates if c2 ≤ cq2 and 3 deviates if c2 ≥ cr2 . As a result Φ is a CPNE if c2 < cr2 . (ii) (123 )−→(1 or 2 or 3)−→ Φ: 1 never deviates, 2 deviates if c2 > cq2 , and 3 deviates if c2 < cr2 . As a result, (123 ) is a CPNE if c2 ≥ cr2 . Proposition 7 It is immediate from the comparisons of both figure 3 with figure 5 and figure 4 with figure 6. Critical c2 and c3 levels: p − 4683 + 15 109984 − 233016c3 + 147889c23 5376c 3 ca2 = 2219 31 p + 5 −10143 + 59682c3 + 147889c23 156 − 1792c 3 cb2 = 657 p −1526c3 − 3882 + 15 58681 + 113890c3 + 24857c23 cc2 = 1471 p 1317 + 961c3 − 4 110880 − 305340c3 − 95970c23 cd2 = 4399 p −807 + 2029c3 + 16 2295 − 8625c3 + 9385c23 ce2 = 291 p 38829 − 1463c3 − 120 72058 + 6090c3 + 154756c23 cf2 = 66703 p −843 + 441c3 + 4 44730 − 80220c3 − 60480c23 g c2 = 559 p 573 + 833c3 − 8 1386 + 6762c3 + 12201c23 ch2 = 1775 p 807 − 2029c3 − 2 200535 − 1450170c3 + 2043935c23 cj2 = 689 p 47649 + 64687c3 − 15 4824862 + 16471560c3 + 20470744c23 ck2 = 168133 √ 3(3857 − 5 359086) m (2 + c3 ). c2 = 41839 √ (3857c3 − 5 359086c3 ) n c2 = −2 + 423 p 1317 + 961c3 − 1774080 − 4885440c3 + 4931010c23 cp2 = 4399 p 213 + 179c3 − 5 924 − 924c3 + 3086c23 cq1 = 571 p 87 − 179c3 − 5 426 − 2592c3 + 3086c23 cr1 = 79 32 References [1] Bagwell, Kyle, and Robert. W. Staiger. “Multilateral Cooperation During the Formation of Free Trade Areas.” International Economic Review, 1997b, 38, 291-319. [2] Bagwell, Kyle, and Robert. W. Staiger. “Regionalism and Multilateral Tariff Cooperation.” In John Pigott and Alan Woodland, eds., International Trade Policy and the Pacific Rim, 1998a, Macmillan, London. [3] Bagwell, Kyle, and Robert. W. Staiger. “Will Preferential Agreements Undermine the Multilateral Trading System?” Economic Journal, 1998b, 108, 1162-1182. [4] Bernheim, Douglas B., Bezalel Peleg and Michael Whinston. “Coalitionproof Nash equilibria I. Concepts.” Journal of Economic Theory 42, 1-12, 1987. [5] Bhagwati, Jagdish. The World Trading System at Risk, 1991, Princeton University Press, Princeton, NJ. [6] Bhagwati, Jagdish, Arvind Panagariya, and Pravin Krishna, eds., Trading Blocs, 1990, The MIT Press, Cambridge, MA. [7] Bhagwati, Jagdish, Arvind Panagariya. “Preferential Trading Areas and Multilateralism — Strangers, Friends, or Foes?” in Bhagwati, Jagdish, Arvind Panagariya, and Pravin Krishna, eds., Trading Blocs, 1999, The MIT Press, Cambridge, MA. [8] Bond, Eric W., and Constantinos Syropoulos. “The Size of Trading Blocs: Market Power and World Welfare Effects.” Journal of International Economics, 1996, 40, 411-437. [9] Bond, Eric W., Syropoulos, Constantinos, and Winters, L. Alan. “Deepening of Regional Integration and Multilateral Trade Agreements.” Journal of International Economics, 2001, 53, 335-362. [10] Brander, James A., and Barbara J. Spencer. “Tariff Protection and Imperfect competition.” In ed. H. Kierzkowski Monopolistic Competition and International Trade, 1984, Oxford University Press, Oxford. 33 [11] Chang, Won and Winters, L. Alan. “How Regional Blocs Affect Excluded Countries: The Price Effects of MERCOSUR.” American Economic Review, 2002, 92, 889-904. [12] Dutta, Bhaskar and Suresh Mutuswami. “Stable Networks.” Journal of Economic Theory 76, 322-344, 1997. [13] Ethier, Wilfred J. “Regionalism in a Multilateral World.” Journal of Political Economy, 1998, 106, 1214-1245. [14] Freund, Caroline. “Multilateralism and the Endogenous Formation of Preferential Trade Agreements.” Journal of International Economics, 2000b, 52, 359-376. [15] Grossman, Gene M. and Elhanan Helpman. “The Politics of Free-Trade Agreements.” American Economic Review 85, 667-690, 1995. [16] Furusawa Taiji and Hideo Konishi. “Free Trade Networks.” unpublished manuscript, 2003. [17] Kennan, John and Raymond Riezman. “Optimal Tariff Equilibria with Customs Unions.” Canadian Journal of Economics, 1990, 90, 70-83. [18] Kennan, John and Raymond Riezman. “Do Big Countries Win Tariff Wars?” International Economic Review, 1988, 29, 81-85. [19] Knetter, Michael M. “International Comparisons of Pricing to Market Behavior.” American Economic Review, 1993, 83, 473-486. [20] Krishna, Pravin. “Regionalism and Multilateralism: A Political Economy Approach.” The Quarterly Journal of Economics, 1998, 113, 227251. [21] Krugman, Paul R. “The Move Toward Free Trade Zones.” in Policy Implications of Trade and Currency Zones: A Symposium Sponsored by the Federal Reserve Bank of Kansas City, Federal Reserve Bank of Kansas City, Kansas City, 7-41, 1991. [22] Levy, Philip I. “A Political-Economic Analysis of Free Trade Agreements.” American Economic Review, 1997, 87, 506-519. 34 [23] Olivier Cadot, Jaime de Melo, and Marcelo Olarreaga. “Regional Integration and Lobbying for Tariffs Against Nonmembers.” International Economic Review, 1999, 40, 635-658. [24] Richardson, Martin. “Tariff Revenue Competition in a Free Trade Area.” European Economic Review, 1995, 39, 1427-1437. [25] Riezman, Raymond. “Can Bilateral Trade Agreements Help Induce Free Trade?” Canadian Journal of Economics 1999, 32, 751-766. [26] Riezman, Raymond. “Dynamic Tariffs with Asymmetric Information.” Journal of International Economics 1991, 30, 267-283. [27] Winters, L. Alan. “Regionalism versus Multilateralism.” In Baldwin, R.E., Cole, D., Sapir, A., and Venables, A.J., eds., Market Integration, Regionalism, and the Global Economy, 1998, Cambridge University Press, London. [28] The World Bank. Trade Blocs, 2000, Oxford University Press, Oxford. [29] The World Trade Organization. International Trade Statistics, 2002, Geneva. [30] Yi, Sang-Seung. “Endogenous Formation of Customs Unions under Imperfect Competition: Open Regionalism is Good.” Journal of International Economics 41, 153-177, 1996. 35 α1 No Agreement (23) (12) (12, 13) Free Trade α3 Figure 1: Equilibrium FTAs under Market Size Asymmetry α1 No Agreement Free Trade under only no FTA option Free Trade under both scenarios α3 Figure 2: Free Trade under Market Size Asymmetry c2 (12) (13) Free Trade Lower threshold of c2 c3 Figure 3: Equilibrium FTAs under Cost Asymmetry and MFN c2 Free Trade under only no FTA option Free Trade under both scenarios No Agreement Lower threshold of c2 c3 Figure 4: Free Trade under Cost Asymmetry and MFN c2 (12) (13) Free Trade Lower threshold of c2 c3 Figure 5: Equilibrium FTAs under Cost Asymmetry and discrimination c2 Free Trade under only no FTA option Free Trade under both scenarios No Agreement Lower threshold of c2 c3 Figure 6: Free Trade under Cost Asymmetry and discrimination