SMU ECONOMICS & STATISTICS WORKING PAPER SERIES Paper No. 06-2007
Rose (2004) showed that the WTO or its predecessor, the GATT, did not promote trade, based on conventional econometric analysis of gravity-type equations of trade. We argue that conclusions regarding the GATT/WTO trade effect based on gravity-type equations are arbitrary and subject to parametric misspecifications. We propose using nonparametric matching methods to estimate the `treatment effect' of GATT/WTO membership, and permutation-based inferential procedures for assessing statistical significance of the estimated effects. A sensitivity analysis following Rosenbaum (2002) is then used to evaluate the sensitivity of our estimation results to potential selection biases. Contrary to Rose (2004), we find the effect of GATT/WTO membership economically and statistically significant, and far greater than that of the Generalized System of Preferences (GSP).