Sector-specific bilateral trade and currency unions

Empirical estimations of the gravity equation for international trade have proven to be useful tools for explaining the pattern of bilateral, aggregate trade ((see Evenett and Keller (2002)). The development of a commodity-specific gravity equation by Anderson and Yotov (2010) permits bringing the useful tool of gravity to explaining trade of particular commodities or classes of goods. This paper brings to disaggregate trade the analysis brought to previous panel studies of worldwide, aggregate, bilateral trade (see Glick and Rose (2002), Baier and Bergstrand (2007), Baier and Bergstrand (2009), and Head et al. (2010) among others). In contrast to previous studies such as Lambert and McKoy (2009) and Vollrath and Hallahan (2011), this paper uses a gravity equation designed specifically for disaggregate, rather than aggregate, trade. The paper gives particular emphasis to trade within currency unions, such as the CFA zones in Africa, the East Caribbean Currency Union (ECCU), the Eurozone, and the Rand zone around South Africa. This paper estimates the commodity specific gravity equation using the Pseudo-Poisson Maximum Likelihood estimation in order to explain the determinants of trade for agricultural trade and manufacturing trade separately. The paper uses a panel of all available country pairs ranging from 1976 to 2010 (for agricultural trade) and from 1980 to 2010 (for manufacturing trade). Data come from the UN COMTRADE database, the TRAINS database (both maintained by WITS), and the UNIDO INDSTAT database. The most significant determinant of agricultural trade is whether or not a country pair consists of a former colonial power and one of its former colonies. Though this relationship also matters for manufacturing trade, the effect is moderate when compared with agricultural trade. However, two countries being former colonies of the same colonizers is a better predictor of manufacturing trade than of agricultural trade, where “better” signifies that the marginal effect is larger for manufacturing trade than for agricultural trade. Additionally, regional trade agreements, a common currency, and a common language are better predictors of manufacturing trade than of agricultural trade. The effect of a common currency on trade differs widely across currency unions and between types of traded goods. The West African Economic and Monetary Union (UEMOA), the Rand zone, US-Dollarized countries, and the anchor-client relationship of India-Bhutan exhibit modestly higher levels of intraunion agricultural trade than otherwise similar countries. In contrast, the ECCU, the Central African Economic and Monetary Union (CEMAC), the UEMOA, the Australian Dollar zone (Australia, Kiribati, and Tonga) and India-Bhutan demonstrate significantly higher levels of intraunion trade in manufactured goods. The Rand zone demonstrates a modestly higher level of intraunion trade in manufacturing. The paper further disaggregates manufacturing trade into 2 groups of ISIC 3 2- digit classifications: 15-19 and 20-37. The determinants of trade in the former set of goods, consisting of food products and textiles, resemble the determinants of trade in agricultural goods generally. However, a trade agreement is a more important factor for goods classified 15-19 than for goods classified 20-37, in contrast to the results from the more aggregated models. The paper also isolates trade in textiles for analysis. Textile trade is defined as trade in goods belonging to division 17 in the ISIC Revision 3 classification system. The determinants of textile trade are largely similar to the determinants of trade for the broader category of goods with codes 15-19. Currency unions, regional trade agreements, and colonial heritage tend to matter more for the broader category of trade than for textiles.