This article examines one of the most important trends in international legal governance since the end of the Cold War: the rise of "soft law," or legally non-binding instruments that are given legal effect through domestic law or internationally binding agreements such as treaties. Scholars studying the design of international agreements have long puzzled over why states use soft law. The decision to make an agreement or obligation legally binding is within the control of the states negotiating the content of the legal obligations. Basic contract theory predicts that parties to a contract would want their agreement to be as credible as possible, to ensure optimal incentives to perform. It is therefore odd that states routinely enter into agreements establishing international rules and regulatory standards in a wide range of subject areas, from banking to arms control to the environmental protection, that are not legally binding.I begin by proposing a new definition of soft. Although previous definitions have often distinguished soft law from hard law on the basis that only the latter is legally binding, no one has explained what distinguishes soft law from purely political arrangements. By contrast, this article defines soft law as those obligations that, while not legally binding, are given some legal effect through separate legal instruments. The Nuclear Suppliers Group Guidelines, for example, are not legally binding on states, but arguably give content to legally binding obligations in the Nuclear Nonproliferation Treaty, and are also given domestic legal effect by statute and regulation. This new definition helps us understand what is "legal" about soft law, while at the same time allowing us to analyze differences between hard and soft law. I then argue that states use soft law as a way to delegate authority over the content of legal rules and regulations to states that possess a particularly strong interest in those rules. Making an agreement non-binding lowers the penalty associated with deviating from the existing legal rules, and thus encourages states with a significant interest in the content of legal rules to unilaterally innovate. This oligopolistic approach to the evolution of legal rules can, under certain circumstances, lead to more efficient legal rules by avoiding the hold-up problem involved in renegotiating contracts in which every state effectively exercises a veto over potentially efficient amendments. As such, the choice between soft law and hard law implicates a tradeoff between the procedural equity inherent in the doctrine of sovereign equality, and the efficiency of legal rules. Soft law privileges the latter over the former.