Determining Which Trade Agreement Provisions Matter for Trade

Modern trade agreements contain a large number of provisions besides tari¤ reductions, in areas as diverse as services trade, competition policy, trade-related investment measures, or public procurement. Existing research has struggled with over tting and severe multicollinearity when trying to estimate the e¤ects of these provisions on trade ‡ows. In this paper, we build on recent developments in the machine learning and variable selection literature to develop data-driven methods for selecting the most important provisions and quantifying their impact on trade ‡ows and apply them to a recent database with highly detailed information on the provisions included in trade agreements. We nd that provisions related to technical barriers to trade, antidumping, competition policy, subsidies, trade facilitation, sanitary and phytosanitary measures, and export taxes are associated with enhancing the trade-increasing e¤ects of trade agreements. Interestingly, we nd that the majority of the 305 provision variables in our data have no measurable impact on goods trade, including virtually all provisions that fall within commonly included policy areas such as services, labor markets, and public procurement.