Multilateral treaties’ success depends in large part on decisions made during their drafting and negotiations. Lack of support from key states, weak or non-binding commitments, and sweeping reservations often doom treaties to ineffectiveness or worse. Challenges to treaty effectiveness have inspired significant bodies of research in international law and relations. Yet existing research in these fields has given little systematic attention to negotiations or to the political origins of treaties generally. This manuscript aims to improve our understanding of treaty-making through both theory development and empirical analysis. I first develop a positive decision-theoretical model of the factors that states consider in drafting, negotiating, approving, and ratifying multilateral treaties. The model takes account of states’ right to opt-out of a treaty and that right’s several implications: that treaty-making entails a three-stage decisional process unique in democratic lawmaking, and that treaty externalities and the quantity and character of future members both affect states’ decisional logic during negotiations. These phenomena have not been fully appreciated in either the legal or international relations literatures, much less formally theorized. I then apply these insights to analyze real-world drafting efforts. Using a novel technique, I code the drafting states’ recorded positions based on two treaties’ negotiating histories, and I use them to estimate states’ ideal points on multiple issues. My findings demonstrate that this method can predict states’ ratifications and reservations with considerable accuracy. The analysis provides new insights into how international law is created and implemented, and under what circumstances it meaningfully affects later state behaviors. Specifically, the issues that divide states differ across treaties, and I find evidence that states’ preferences for particular treaty provisions coincide with those we would expect of utility-maximizing states. That the state positions predict subsequent behavior implies that treaty negotiations yield a rich trove of relatively authentic revealed state preferences. That suggests that, in addition to fueling theory and data-generation, these methods and the insights they provide could also be used to aid ongoing treaty negotiations in the future.