Testing conflicting political economy theories: full-fledged versus partial-scope regional trade agreements.

• Author: Liu, Xuepeng

1. Introduction

   A customs union ... duties and other restrictive regulations of commerce (...) are eliminated with respect to substantially all the trade between the constituent territories of the union or at least with respect to substantially all the trade in products originating in such territories....A free-trade area shall be understood to mean a group of two or more customs territories in which the duties and other restrictive regulations of commerce (...) are eliminated on substantially all the trade between the constituent territories in products originating in such territories.--GATT Article XXIV

   Notwithstanding the provisions of Article I of the General Agreement, contracting parties may accord differential and more favorable treatment to developing countries, without according such treatment to other contracting parties.--GATT Enabling Clause [italics added by author]

   One of the pillars of the GATT/WTO rules is nondiscrimination, which is also called the Most Favored Nation Clause (MFN) in the GATT Article I. It is a commitment that a GATT/ WTO member will extend to another member the lowest tariff rates it applies to any other GATT/WTO members. One prominent exception to the MFN Clause is the existence of regional trade agreements (RTAs), which by their very nature are discriminatory. The proliferation of RTAs since the late 1950s is fundamentally altering the world trade landscape.

   What are the driving forces of regionalism? Besides economic and geographic factors, political economy factors are also important. There are both winners and losers from free trade or RTAs. RTA decisions will be eventually determined by how the preferences of voters, politicians, and interest groups are aggregated in the policy-making process. Therefore, it is important to open the "black box" of the political economy of RTA formation. Different political economy models can have very different predictions. A majority or general interests approach (i.e., the median voter model) in a two-factor Heckscher-Ohlin framework predicts pro-labor trade policy, while a special interests approach (e.g., the lobbying model) based on the argument of the free rider problem predicts pro-capital policy. We test these conflicting predictions in RTA formation and find that different models are more relevant for different types of RTAs.

   Three types of RTAs are permitted under the GATT/WTO: free trade area (FTA), customs union (CU), and preferential trade arrangement (PTA). (1) GATT Article XXIV, introduced in 1947, permits FTAs and CUs as long as such agreements cover "substantially all the trade" in the blocs and do not raise external trade barriers, which are believed essential to promote trade creation and avoid trade diversion. An FTA is an agreement with tariffs eliminated on goods produced in member countries, while a CU is an FTA plus common external tariffs. In 1979 a further exception to nondiscrimination treatment, entitled the Enabling Clause, was agreed upon for developing countries. This clause allows developing countries to exchange partial tariff preference within PTAs. (2) The extent of trade liberalization in PTAs can be very limited. In a famous example, Indonesia, a member of the Association of Southeast Asian Nations (ASEAN) and a country on the equator, excluded major sectors but offered to liberalize imports of snow ploughs from other ASEAN countries (Panagariya 1994). For this reason, we consider PTAs as partial-scope arrangements and FTAs/CUs as full-fledged agreements.

   Previous researchers have typically analyzed the determinants of RTAs in standard binary choice regressions such as probit and logit. We argue that the traditional binary choice model is inappropriate in panel data analysis because of the conditional independence assumption of the RTA dummy variable. We adopt a duration analysis to correct this problem. Our empirical results show that the prediction of the median voter model is supported by FTA/CU formation, while the lobbying model seems to be more relevant in PTA formation. The support for the conflicting political economy models by FTAs/CUs and PTAs is not surprising given the very different natures of these agreements. The limited sector and trade coverage in PTAs encourages lobbying, while the "substantially all the trade" requirement for FTAs/CUs can serve to reduce lobbying activities. Although most of the FTAs/CUs have long staging periods in internal tariff reduction and although some of them may violate the spirit of Article XXIV, they are offered much less scope for exchanging arbitrary preference than are PTAs.

   This article is related to the work of Dutt and Mitra (2002) and Liu (2008), who find empirical support for the median voter model (pro-labor) in trade policies and FTA/CU formation, respectively. This article distinguishes partial-scope from full-fledged RTAs and provides a better understanding of the conflicting distributional forces behind RTA formation. We show that whether trade policies are pro-labor or pro-capital depends on a number of parameters. In the case of RTA formation, it depends on the types of RTAs, the sector coverage of RTAs, governments' ideological preferences and concerns for special versus general interests, etc.

   To reinforce our confidence in the political economy interpretation of our results, we provide the following auxiliary regressions. First, we confirm that the main findings of this article are indeed driven by distributional effects of RTAs. Within a two-factor framework (capital and labor), this distributional effect should be stronger for country pairs with more different capital-labor ratios (K/L). We run regressions by subsamples based on the difference between two countries' K/L. As expected, the support for both models is stronger in the subsample with more different K/L compared to those with similar K/L. Second, the pro-labor (pro-capital) trade policy orientation is closely related to the left (right) ideology of a government. If the regressions in this article indeed capture the political ideology of the governments, the support for the median voter model (lobbying model) should be stronger when the two countries in a pair have left-oriented (right-oriented) governments, compared to the rest of the sample. We confirm this with strong empirical evidence. Third, we relate the support of PTAs for the lobbying model to the estimated weight on social welfare by Gawande, Krishna, and Olarreaga (2008). We find stronger (weaker) support for the lobbying model in the PTA regressions when restricting the countries to those with lower (higher) weights on social welfare. Finally, we investigate further the relevance of RTAs' sector coverage to our findings. We argue that our results may be driven by the very different trade coverage requirements of Article XXIV and the Enabling Clause. Better sector coverage leaves less room for lobbying and makes governments care more about the general interests. We explore an alternative way to support this argument by distinguishing different FTAs/CUs based on their coverage of service sectors. Based on this measure we find that the support for the median voter model is stronger when agreements have better coverage.

   The above analyses show that we indeed capture the conflicting distributional forces in RTA formation rather than some spurious correlations. Two cautions in the political economy interpretations, however, are worth noting. First, the support of the median voter model should not be taken too literally in the context of direct democracy and single-peaked preference. We can have exactly the same pro-labor policies if a government is simply concerned about inequality, as demonstrated in Limao and Panagariya (2007). Therefore, the median voter model should be taken as a short-hand way to capture the general concerns for popular support, general interests, or income equality. Second, the lobbying model considered in this article is based on the argument of the free rider problem, as described in Rodrik (1986), and is different from the "protection for sale" approach described in Grossman and Helpman (1994), which explains the cross-industry trade policies and offers different predictions for organized and unorganized industries.

   The findings in this article offer some policy implications. Although the support for the median voter model by the FTAs/CUs does not necessarily imply good governance, the support for the lobbying model by PTAs does raise some concerns. An RTA heavily influenced by lobbies usually leads to welfare loss as a result of trade diversion ("static welfare impact") and prevents the move from regionalism to multilateralism ("dynamic time path"). According to the classic analysis of Viner (1950), an RTA is welfare improving if the trade creation effect dominates the trade diversion effect. Manipulations by special interest groups representing import-competing sectors are often considered a source of trade diversion and "stumbling block" effect. (3) The recent literature of endogenous trade agreement formation (e.g., Maggi and Rodriguez-Clare 1998, 2007; Mitra 2002) indicates that trade agreements may serve as a commitment device to tie the hand of a politically motivated government or a government with low bargaining power. Ornelas (2005a, b) shows that FTAs lowering the domestic market shares of import-competing sectors reduce their incentives of lobbying for higher external tariffs ("rent dissipation" effects). (4) The results in this article imply that a full-fledged RTA, when compared to a PTA, can serve as a better commitment device and has stronger rent dissipation effects and, hence, is more likely to be beneficial to multilateral liberalization.

   The rest of the article is organized as follows. Section 2 discusses the theoretical predictions. In section 3, we explain the testing strategy and econometric methodology. Section 4 describes the data. In section 5 we report...