The Gravity Equation: To Pool or Not to Pool-A Critical Assessment

I. IntroductionThe gravity model has become a central empirical tool in international trade, with robust predictive power and increasing theoretical refinement.While its empirical use dates back to Tinbergen (1962), the gravity equation has undergone a profound transformation over the past two decades.The foundational work of Anderson & Van Wincoop (2003) introduced multilateral resistance terms, providing a micro-founded structure that links bilateral trade flows to relative trade costs in a general equilibrium framework.This structural interpretation is complemented by Eaton & Kortum's (2002) Ricardian model, which derives trade patterns from stochastic productivity differences, and embeds the gravity equation within a richer general equilibrium framework.Parallel to these theoretical developments, empirical advances have addressed important econometric limitations.Notably, Silva & Tenreyro (2006) demonstrated that the commonly used log-linear OLS specification of the gravity equation is inconsistent in the presence of heteroskedasticity and zero trade flows.They proposed the Poisson Pseudo-Maximum Likelihood (PPML) estimator as a robust alternative, which has since become the gold standard in empirical