Trade openness and economic growth: an empirical analysis of NAFTA
Date
2019
Authors
Journal Title
Journal ISSN
Volume Title
Publisher
University of Delaware
Abstract
This dissertation investigates empirically the notion that trade openness resulting from deregulation and reduced tariffs required by the North America Free Trade Agreement (NAFTA) promotes the acceleration of economic growth. Methodologically, I employ a two-step instrumental variable (IV) method in which NAFTA adoption is considered as an instrumental variable for trade openness. I use the difference-in-difference (DID) estimation method captured by NAFTA adoption in the first step. And then I take system generalized method of moments (GMM) and Panel Vector Autoregression (VAR) in the second step to estimate a dynamic panel regression. The combination of IV and DID (DDIV) minimizes the negative impact of omitted variables. Therefore, in theory DDIV can improve estimate accuracy. ☐ Empirically I first try to demonstrate the effectiveness of DDIV. On one hand, I set up a test for the parallel trend assumption for DID and verify that the control group and treatment group share a common trend. On the other hand, I use the Arellano estimator to test the endogeneity of IV. The results do not give support for the endogeneity of IV. Therefore, DDIV might be a practicable method for system GMM estimator in a dynamic panel dataset. ☐ Moreover, I compare the system GMM estimator without DDIV and the system GMM estimator with DDIV, both of which display similar conclusions: trade openness has a positive effect on GDP growth at the simultaneous time t at the 5% level of statistical significance whereas the trade openness at lagged time t-1 has a negative effect on current GDP growth at time t at the 10% level of statistical significance. The empirical results also demonstrate DDIV might improve the significance of the coefficient estimates but does not reduce the standard errors. ☐ Furthermore, I set up a panel Vector Autoregression (VAR) Model with DDIV and use the IVGMM two-stage method to estimate the model. That is to say, I still use NAFTA adoption as an instrumental variable and employ DID in the first step. Then I employ a panel VAR model in the second step to estimate the effect of trade openness on economic growth. Panel VAR has the consistent result as system GMM with DDIV: the lagged trade openness at time t-1 is negatively related to current GDP growth at time t at the 10% level of statistical significance. However, such a negative effect is much less than the positive effect of trade openness on GDP growth rate at the simultaneous time. ☐ Finally, I estimate the effect of a shock to trade openness on GDP growth within each of the member countries of NAFTA and between member countries of NAFTA and obtain diverse results. First, I conduct a VAR analysis for each member country of NAFTA. Then I conduct a series of Granger-causality tests which reveal that there are two Granger causality relationships across the member countries of NAFTA: First, the GDP growth rate of the Canada and Mexico Granger-cause the US GDP growth rate. Second, the GDP growth rate of the United States and Mexico Granger-cause the GDP growth rate of Canada. On the basis of VAR analysis of each member country and Granger-causality across member countries, I calculate the effect of a trade openness shock happening in one of the countries on the GDP growth rate of the other NAFTA member countries. A positive shock to trade openness of Canada and Mexico would have a positive effect on US GDP growth.
Description
Keywords
Difference in Difference, NAFTA, Panel VAR, System GMM, VAR