The effects of integration agreements in Western Hemisphere trade, 1970–2014

Abstract

The paper analyzes the effects of four regional integration agreements (Common Market of the South [MERCOSUR], Andean Community [ANCOM], Central American Common Market [CACM] and North America Free Trade Agreement [NAFTA]) on bilateral trade in 19 countries from the Western Hemisphere for the period 1970–2014. For this purpose we estimate different gravity models to control for trade creation and diversion, export diversification and intra-industry trade using OLS log-linearized gravity model and Poisson pseudo-maximum-likelihood panel data estimators that allow controlling for zero-value trade flows. We find trade creation for ANCOM, MERCOSUR and CACM and trade diversion for NAFTA and MERCOSUR countries. Export diversification negatively affects bilateral trade in all American agreements, while intra-industry trade has contributed to trade expansion in ANCOM and the opposite for NAFTA, MERCOSUR and CACM. Global supply chains may help us explain these results. Finally, we find anticipatory effects on trade several years before the signing of the agreements, but only NAFTA countries seem to be natural trading partners in the region while the rest of Latin American regional agreements have not resulted in a comprehensive, profound and consolidated common market.

Keywords:

• Economic integration
• trade creation and diversion
• intra-industry trade
• product differentiation
• gravity model

JEL Classifications:

• C23
• F14
• F15

Acknowledgments

The authors are grateful to anonymous referees for excellent comments that have improved this paper substantively.

Disclosure statement

No potential conflict of interest was reported by the authors.

Notes

1. See Kahouli and Maktouf (2014, 2015) for a review of some recent studies that used gravity models to account for RTAs’ effects on trade.

2. Estimation is possible because we have 2n² country pairs but only 2n country dummies.

3. Incorporating time dummies together with country dummies is also possible because the total set of country and time dummies is 2nT, below the total 2n² observations.

4. See Redding and Venables (2004), Rose and Van Wincoop (2002), Feenstra (2002), for a discussion about country-fixed effects. Also, Baltagi, Feng and Kao (2015) have a discussion about exporter-time and importer-time fixed effects in addition to country-pair fixed effects in order to control for endogeneity among the regressors. We have not included these sets of variables although they improve the accuracy of the estimates, because they do not allow to estimate trade diversion variables (Magee, 2008).

5. According to Halvorsen and Palmquist (1980), in the presence of semi-logarithmic equations, the estimated coefficient of dummy variables does not directly reflect their effects on the dependent variable. To calculate the real effect of these variables in trade we need to apply the exponential function to the estimated coefficient minus one (e^γ − 1).

6. The Herfindahl–Hirschman index measures the degree of market concentration, according to the following formula: where HHIxit is the Herfindahl–Hirschman trade concentration index for total exports reported by country i index at time t, xnt is the value of exports of product n in time t; and Nt is the number of products at the three-digit SITC, Revision 3 level. Xt = ∑Ntn = 1xnt. It has been normalized to obtain values from 0 (minimum concentration) to 1 (maximum concentration)

7. The Grubel–Lloyd index defined by Grubel and Lloyd (1975) measures intra-industry trade through the following equation: where X_ijk and M_ijk are, respectively, exports and imports of industrial product k between countries i and j at time t. Then, a weighted average can be obtained of all the k products in order to get an aggregated GLI index for each country.

8. Available at http://faculty.haas.berkeley.edu/arose/RecRes.htm#Trade (http://faculty.haas.berkeley.edu/arose/datagravstataWEO.zip)

9. The interpretation of dummy coefficients is not direct and their effect on trade must be calculated by transforming the estimates with (e^γ − 1).