Do regional trade agreements increase members' agricultural trade?

Regional Trade Agreements (RTAs) have become a ubiquitous feature of global trade. (1) In 2003, 250 RTAs were notified to the World Trade Organization (WTO) with approximately 180 agreements in force. Remarkably, more than one-half of these agreements occurred between 1995 and 2002. Between January 2004 and February 2005, forty-three additional RTAs were notified to the WTO. By 2010, the world faces the prospect of over 400 RTAs. As Crawford and Fiorentino (2005) noted, we have entered into one of the most prolific periods of RTA formations in recorded history.

The proliferation of RTAs may be due in part to the frustration of negotiators attempting to achieve multilateral free trade. This is particularly true in agriculture. In the Doha Round of trade negotiations, WTO members (particularly developing countries) have made it clear they are unwilling to negotiate on other topics until a suitable agreement on agriculture exists. RTAs may be an attractive alternative to promote agricultural trade. First, Article XXIV (Customs Unions and Free Trade Areas) of the General Agreement on Tariffs and Trade (GATT) commits members to eliminate restrictions on "substantially" all trade within an RTA. Second, RTAs may facilitate deeper integration by liberalizing nontariff barriers including technical standards, food safety concerns, and domestic regulations; areas where the WTO has made very little progress. Finally, RTAs are easier to conclude because they involve fewer negotiating parties.

The gravity model has performed remarkably well as a tool for measuring the impacts of RTAs. (2) Eichengreen and Irwin (1998) called the gravity model "the workhorse for empirical studies to the virtual exclusion of other approaches" (p. 33). However, previous studies investigating the trade flow effects of RTAs have produced conflicting results (Ghosh and Yamarik 2004; Baier and Bergstrand (B&B) 2007). B&B (2007) highlighted the European Union (EU) integration as a case and point: some studies have found positive and significant effects (Tinbergen 1962; Aitken 1973; Brada and Mendez 1985), while others have found insignificant and, in some cases, negative trade flow effects (Frankel, Stein, and Wei 1995, 1996; Frankel 1997; Krueger 2000).

Investigations of the North American Free Trade Agreement (NAFTA), the Mercado Comfin del Sur (Mercosur), the Andean Pact, the Association of Southeast Asian Nations (ASEAN), and the Closer Economic Relations (CER) have similarly produced conflicting results. Frankel (1997) found no trade effects associated with NAFTA or the Andean Pact, but large and significant effects for Mercosur, the ASEAN, and the CER. However, Finger, Ng, and Soloaga (1998) and Soloaga and Winters (1999) found negative trade flow effects for Mercosur, while Krueger (2000) found no effects. Hassan (2001) found negative effects, Krueger (2000) found positive effects, while Sharma and Chua (2000) found no trade effects associated with the ASEAN. Krueger (2000) found positive trade effects in the Andean Pact and the CER, but no trade effects following NAFTA, whereas Dhar and Panagariya (1994) found small trade effects in North America and the EU, but no effects for the ASEAN partners.

It seems almost paradoxical to find such conflicting effects of RTAs on trade. Ghosh and Yamarik (2004) addressed this issue econometrically and showed that cross-sectional gravity equations yielded highly unstable results. B&B (2007) found similarly unstable RTA effects in cross-section regressions. However, B&B (2007) went a step further and showed that previous studies have produced biased results because countries select endogenously into RTAs. Using panel data methods to account for the endogeneity of RTA membership and a theoretically consistent gravity equation, B&B (2007) found that RTAs approximately doubled members' trade using aggregate trade data.

This article adopts B&B's (2007) framework. However, what distinguishes this study from B&B (2007) and many others is the separate consideration of agricultural and nonagricultural trade within RTAs. We contribute in two important ways. First, we demonstrate that the ex post impacts of RTAs on trade depend fundamentally on whether the analysis focuses on agricultural or nonagricultural sectors. A notable feature common to most previous studies is the use of aggregate (merchandise) trade data. (3) We actually know very little about the scope for increased agricultural trade within RTAs, which is why we asked the question in this article's title. However, pre-RTA tariff rates in agriculture were significantly larger than nonagricultural protection. Thus, successful liberalization of agricultural tariffs within RTAs could generate a larger agricultural trade response compared to the nonagricultural sector.

Ingco (1995) found that agricultural protection was fifteen times greater than manufacturing tariff levels in 1992. By 2001, Gibson et al. (2001) found that agricultural tariffs were still twelve times higher than nonagricultural protection. (4) Tariffs alone, however, are only part of the story. In the Canada-U.S. Trade Agreement (CUSTA) and NAFTA, for example, tariff rates on many agricultural products were low before the agreements entered into force. As tariffs are reduced within the multilateral framework, nontariff barriers become increasingly important. The WTO allows countries to adopt sanitary and phytosanitary (SPS) measures and Technical Barriers to Trade (TBT) to protect plant, animal, and human health as well as the environment. Thilmany and Barrett (1997) called attention to the importance of regulatory barriers in agri-food trade that may or may not be obviated by RTAs. The recent food safety concern associated with Bovine Spongiform Encephalopathy (BSE) in Canada and the United States is a prime example of cross-border trade being halted, even within RTAs. This article sheds light on the effectiveness of RTAs in dismantling agricultural tariff and nontariff barriers.

Second, we demonstrate that the trade flow effects of RTAs depend on the specific agreement, and on the length of the implementation (i.e., phase-in) period. (5) It may take several years or even longer than a decade, in the case of agriculture, before the trade flow effects of RTAs are measurable. Whether and to what extent RTAs increase members' agricultural trade and how long it takes before actual effects on trade occur is an important empirical question addressed in this article.

For example, Canada's supply managed products (dairy, poultry and eggs) and U.S. dairy, sugar, tobacco and peanut programs were effectively excluded in the CUSTA, which entered into force on 1 January 1989, but contained a ten-year phase-in period. (6) NAFTA was signed in 1994, but required two separate bilateral agreements for agricultural trade with Mexico and a fifteen-year phase-in period ending 1 January 2008. Mercosur was signed in 1991, but continues to protect wheat, wool, and some fruit and vegetable products. The ASEAN entered into force in 1992, but initially excluded agriculture. It was not until 1996, that members agreed to the gradual elimination of agricultural barriers by 2004. At the other extreme, the CER between Australia and New Zealand, and the EU permit free trade in agriculture. EU members also benefited from a harmonized Common Agricultural Policy (CAP) that maintains high agricultural tariffs against nonmembers, potentially creating a strong intraregional bias. (7)

To our knowledge, only three published studies have examined the effects of RTAs on members' agricultural trade (Zanhiser et al. 2002; Furtan and van Melle 2004; Vollrath, Hallahan, and Gelhar 2006). Furtan and van Melle (2004) and Zanhiser et al. (2002) focused only on NAFTA trade. Vollrath, Hallahan, and Gelhar (2006) investigated the socio and geopolitical forces influencing land-based and processed food trade. The authors controlled for regional similarities within the EU, NAFTA, and Mercosur and found some evidence that these agreements increased members' agricultural trade. However, Vollrath, Hallahan, and Gelhar (2006) and Furtan and van Melle (2004) used cross-sectional data, which are known to be plagued by the endogeneity concerns of B&B (2007). None of these studies provided a benchmark comparison to nonagricultural trade, nor did they consider the phase-in periods that characterize nearly all RTAs (B&B 2007).

The Gravity Model

A typical log-linear gravity equation to investigate the trade flow effects of RTAs is

(1) [MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII]

where [X.sub.ij] is the value of trade from country i to country j; [GDP.sub.i] ([GDP.sub.j]) is the gross domestic product of the exporting (importing) country as a proxy for economic size; and [D.sub.ij] is the distance between countries i and j used to proxy for transportation costs. Researchers usually experiment with other geographic and preference similarities such as whether the two countries are contiguous ([Adj.sub.ij]), share a common language ([Lang.sub.ij]), or are landlocked ([LL.sub.ij]). (8) [RTA.sub.ij] is a dummy variable indicating the existence of a regional trade agreement between countries i and j.