Predicting the effects of NAFTA with the BDS, GTAP, and EK models

We use three computable general equilibrium (CGE) models of trade to simulate the removal of tari§ and non-tari§ barriers between U.S., Canada, and Mexico under NAFTA. The models are parametrized with pre-NAFTA data. Our paper has three objectives. First is to evaluate the predictions made by GTAP against the actual post-NAFTA changes in NAFTA trade. Recent literature (Kehoe, 2005; Shikher, 2012) Önds that pre-NAFTA predictions of the e§ects of NAFTA on trade have little correlation with the actual changes. Therefore, we wanted to investigate whether GTAP could have produced accurate predictions if it were used before NAFTA. The second objective is to compare the predictions of GTAP against the predictions of a general equilibrium model based on the Eaton and Kortum (2002) methodology with heterogeneous producers. We wanted to investigate if the two models would produce di§erent predictions and investigate the sources of these di§erences. The third objective is to analyze why the predictions of the BDS model are di§erent from the actual changes in trade. We Önd that the changes in NAFTA trade predicted by GTAP using pre-NAFTA data are much closer to what actually happened after NAFTA than the changes predicted by the pre-NAFTA studies we looked at (Brown, Deardor§, and Stern, 1992; Roland-Holst, Reinert, and Shiells, 1994). GTAPís predictions are also close to those of the model with heterogeneous producers, though the latterís predictions are more closely correlated with the actual changes. We investigate the causes of the di§erences in predictions by BDS, GTAP and a model with heterogeneous producers. Among the various causes we look at are model equations, parameter values, sensitivity to parameter values, choice of base year, and solution methods.