TWO ESSAYS ON THE ESTIMATION OF AGRICULTURAL TRADE FLOWS: MODEL SELECTION AND ENDOGENEITY OF TRADE AGREEMENTS

Abstract

This thesis investigates three econometric issues relevant to the estimation of gravity models of agriculture trade, namely the issues of zero trade flows, heteroskedasticity, and the endogeneity of preferential trade agreements (PTAs). Recent studies have employed the Poisson pseudo-maximum likelihood (PPML) estimator to address zero trade and heteroskedasticity, while a second line of work has argued for better modelling of zero trade through limited dependent variable models, with little consideration for the heteroskedasticity problem. The first essay of this thesis proposes the use of a Box-Cox double hurdle (BCDH) model as a way to address both econometric issues. The second essay in this thesis challenges Sun and Reed’s (2010) conclusion that NAFTA has no impact on members’ agricultural trade. The two-step estimator developed by Treza (1998) provides evidence of substantial endogeneity bias in their estimated effects for nearly all preferential trade agreements, including NAFTA.