Endogenous tariff formation has been the subject of theoretical studies that attempt to determine the fundamental economic variables that influence the structure of industry protection implemented by international trade policy makers. An empirical analysis of endogenous tariff formation under the framework of a regional trade agreement implemented by the Andean Community Group is offered in this dissertation. Econometric models for the group??s common external tariff (CET) and for individual country tariff deviations with respect to the CET are estimated. The analysis is based on cross-sectional industrial and trade data for 1996, collected at four digit level of aggregation. The level of aggregation refers to the specific definition of industrial sectors included in the International Standard Industrial Code (ISIC). While previous studies on another regional integrated group in South America (MERCOSUR) use data at the three digit level, the aggregation used in this research implies a significant increase in the sample size, and also a more homogeneous specification regarding the composition of the industrial sectors under analysis. The causal links among the variables are obtained by using the directed acyclical graphs (DAGs) approach. This allows for a refined search for causal relationships. The approach is particularly appealing for the analysis of endogenous trade protection since it allows analyzing economic systems that involve policy intervention. The empirical analysis supports several of the classic theoretical models on trade protection. The results are consistent with the equity concern model, which suggests that governments tend to protect industrial sectors that employ a significant number of low wage unskilled workers. The estimated models also support the interest group and the adding machine theoretical formulations. However, a rather interesting result derived from the DAG analysis is the feedback interaction that seems to operate between tariffs and policy variables. The current literature restricts the estimation of trade protection by imposing tariffs as the dependent variable with no reverse effect from this variable to the policy variables. Our results challenge this unidirectional causality view, since an effect from tariffs to the policy variables shows up in most of the estimated specifications.
