Missing data in the structural gravity: estimation bias of preferential trade agreements due to the omission of internal trade

In the last decade, there has been an intense development in trade models aiming to explain the determinants of bilateral trade. A seminal theoretical and methodological contribution is Anderson and van Wincoop (2003), who introduced the concept of multilateral resistance and structural gravity. However, there is still an important gap between the theoretical developments of the structural gravity model and its empirical applications. Two main issues come from the presence of zeros in bilateral trade and missing data for internal trade flows (own production oriented to the own market). The presence of zero trade flows has been considered in Santos Silva and Tenreyro (2006) and Helpman, Melitz and Rubinstein (2008). The consequences of omitting internal transactions have not been much studied, even when its relevance may be greater due to a significant heterogeneity across countries’ openness. The objective of the paper is to analyze and characterize the consequences from omitting internal trade in the estimation of trade proximities (inverse trade costs) and on the values of multilateral resistances, which in turn will affect the comparative statics effects derived from different trade policy measures.