The Trade Effects of Endogenous Preferential Trade Agreements

Abstract

This paper suggests an empirical modeling strategy which takes full account of the structural, non-(log-)linear impact of endogenous trade barriers on trade in new trade theory models. In particular, we use the framework to evaluate and quantify the role of endogenous preferential trade agreement (PTA) membership for trade. Apart from paying attention to structural modeling of the impact of trade policy on trade, the suggested model takes into account both that preferential trade agreement membership is endogenous and that the world matrix of bilateral trade flows contains numerous zero entries. These features are treated in an encompassing way by means of (possibly two-part) Poisson pseudo-maximum likelihood estimation with endogenous binary indicator variables in the empirical model.

The Costs and Benefits of Leaving the EU

“…Trade openness can increase growth rates due to a rise in investment (Wacziarg, 1998), increases in technology diffusion (Sampson, 2013;Wacziarg, 1998), export learning effects (Albornoz, Calvo Pardo, Corcos, and Ornelas, 2012;Egger, Larch, Staub, and Winkelmann, 2011) and greater investment in R&D (Bloom, Draca, and Van Reenen, 2011;Keller, 1999Keller, , 2002). …”

International terrorism as a trade impediment?

“…It turns out that putting extra weight on the numerous zeros in the data with zero-inflated Poisson estimation does not lead to drastically different results from the Poisson models which give equal weights to the zeros as to other levels of trade flows. Log-transformed OLS models (which are suppressed for the sake of brevity) lead to a sizable bias with the data at hand (see Santos Silva andTenreyro, 2006, andEgger et al, 2011, for the general problems with OLS in such contexts). Notice that earlier research relied exclusively on such problematic OLS estimates.…”

Effects of the Endogenous Scope of Preferentialism on International Goods Trade

“…7 For the estimates of ATT E s;s and ATT I s;s to be consistent, we impose three common assumptions: conditional mean 6 The results in Egger, Larch, Staub, and Winkelmann (2011) suggest that these two margins can be analyzed as two separate parts of an integrated model. 7 In general, endogenous treatment effects' problems are ones of missing data.…”