Almost every country in the world is a member of at least one preferential trade agreement (PTA). These agreements govern more than half of all world trade. Identifying the effects of PTAs on trade is challenging due to self-selection bias: countries choose to become members. To address the selection issue, this paper builds a comprehensive dataset, and uses the blocking estimator from the causal inference framework. The paper shows that, after accounting for selection, PTAs increase bilateral trade by 48% fifteen years after entry into force. The effects phase in gradually, with one-third being realized five years prior to the agreement. Anticipation effects are only present for non-natural trading partners—geographically distant countries that trade little and have a low probability of signing an agreement. Other methods that do not account for selection substantially overestimate the effects. Equipped with the empirical estimates, this paper builds a model to analyze the general equilibrium effects of a recent important agreement.

International trade agreements are considered to be beneficial for their members, both from the game-theoretic perspective and in the data. Yet, very few country pairs in the world have agreed to full trade liberalization. This paper provides an explanation to why this is the case, focusing on the very nature of bilateral trade negotiations. We model a real-world feature of trade negotiations---the concessions countries are willing to make to achieve a reduction in a trade barrier---as a game with strategic side payments. These reciprocal contingent concessions can alter the negotiation outcome such that the agreement will not always be possible. We derive conditions under which efficient tariffs can be reached in a general framework. We further explore a canonical model of trade to derive and test empirical predictions on the trade agreement formation.

This paper takes a long-run perspective and analyzes the evolution of trade agreements from the `first wave of globalization'—the beginning of the 19th century—to modern times. I collect a novel dataset containing information about trade agreements dating back to 1815. I present the information from the texts of the treaties in two different ways. First, by reading the treaty texts, I identify and codify the key design features that were agreed upon by the signatories. Second, I analyze the evolution of the design of the commercial treaties, the emergence and the prevalence of the key features in different time periods, and the geographical distribution of the treaties. The results show that many of the features of the contemporary trading system were developed long before the General Agreement on Tariffs and Trade (GATT)—the foundation of the modern trading system—was signed. Moreover, it emerges from the analysis that many of the commercial provisions in the treaties appeared, adjusted and developed along with the political and economic developments of the different historical periods.

Most studies of preferential trade agreements (PTAs) use gravity models with PTAs as binary variables. This approach leaves out substantial amount of metadata and semantic information embodied within the texts. Over the last several decades the PTA texts have evolved, creating a considerable variation in the content and structure across agreements. This paper constructs a novel dataset with individual characteristics of PTAs and a document-term matrix. Using text-mining techniques and predetermined intended semantic meanings, the paper builds a continuous index measuring heterogeneity in legal enforceability across PTAs.