REGIONAL TRADE AGREEMENTS REVISITED

The gravity model is a workhorse for econometric studies of the impact of regional trade agreements (RTAs). Despite its initial lack of theoretical basis, the model has been successfully derived from various trade theories. The latest theoretical derivation by Anderson and van Wincoop (2003) reveals that prior gravity studies have made the critical error of omitting the multilateral resistance variable, which results in biased estimates. Other recent studies have highlighted empirical issues with the commonly used procedure of log-linearizing the gravity model and estimating the parameters using Ordinary Least Squares (OLS) regression. Silva and Tenreyro (2006) point out that this method yields inconsistent estimates in the presence of heteroskedasticity. Helpman, Melitz and Rubinstein (2004) show that the concomitant practice of dropping observations with zero trade values (because the log-linearized model is not defined for such observations) will also give rise to biased results. To deal with these two issues of inconsistency and bias, we estimate the gravity model in its multiplicative form. Both cross-sectional and panel data analysis are performed, employing Poisson Pseudo Maximu