We thank Jeffrey Smith, Marc Trachtenberg, Deborah Larson, Matthew Baum, Stanley Sheinbaum, the anonymous reviewers, and the editors of IO, David Lake and Peter Gourevitch.

1. In the volume's introduction we define international institution broadly enough to include treaties but not so broadly as to include tacit bargaining; Koremenos, Lipson, and Snidal, this volume.

2. Van Evera 1997.

3. We had pragmatic reasons for sticking to a “two-cornered” fight between our conjectures and the evidence. First, we did not want to expend the energies of our contributors on “meta-” and “-ology” debates; we asked them instead to focus on what the rational perspective could accomplish. Second, we did not feel competent to present, for example, the constructivist or realist conjectures about institutional design. If they “lost” such a fight, we would be accused of rigging the contest through our incompetence. While we welcome third parties to join this intellectual effort, we argue below that there is more payoff from doing so as allies than as adversaries.

4. We thank an anonymous reviewer for reminding us of this point.

5. The Rational Design approach can include nonstate actors and does not inherently restrict the autonomous capacity of international institutions. Multinational firms and transnational interest groups can and do act at the international level and sometimes shape specific institutions, as Mattli makes clear. The Rational Design approach can include any type of intentional actor. In practice, however, states shape nearly all important international institutions, and for simplicity we sometimes limit our analysis to them. We also recognize that states can deliberately choose to endow international institutions with more autonomous powers to serve their purposes. According to Andrew Moravcsik, the principal members of the EU have done exactly that. Moravcsik 1998.

6. According to the influential work of George Stigler, Mancur Olson, Douglass North, and others, rent seeking is central to the organization of the state and state-society relations. Their insights about domestic politics could be extended to international cartels, including producer organizations like OPEC, and perhaps to a wider range of international institutions.

7. Many of these issues are taken up in Lake and Powell 1999.

8. These silences do raise another issue, however: the potential for selection bias. Results from cases where both the independent variable (X) and the dependent institutional variable (Y) are present might be overinterpreted as suggesting a “sufficient” condition that “X leads to Y.” But a “silence” indicates that X can be present without Y necessarily occurring, so the “sufficiency” claim would be wrong. Ideally, we need to compare across cases to understand what other factors determine when X has the hypothesized effect Y, and when it does not. Although we are not aware of any situations where this problem arises, any interpretation of these results needs to be appropriately bounded in this way. Even with silences, however, a preponderance of positive findings still tells us that a factor does influence institutional design under some circumstances.

9. Finally, we stress that the Rational Design conjectures were not retrofitted to the findings. We developed the framework before the case studies began. We subsequently refined our terminology, dropped one conjecture where we found its derivation poorly grounded, and dropped another where we found its logic to be identical to one of our other conjectures. Otherwise the conjectures were all derived before the evaluations. The conjecture we dropped on theoretical grounds dealt with the effects of distribution on centralization and had weak empirical support. The other addressed the effect of distribution on control and had strong empirical support. Dropping these conjectures does not affect the overall results.

10. Abbott and Snidal 2000.

11. Morrow defines noise as “the inability to determine exactly what happened and so to determine precisely the other side's responsibility for the event.”

12. The rationale for this kind of flexibility can be found in Downs and Rocke 1990; they argue that in the face of uncertainty about behavior the optimal strategies are often those that specify thresholds that trigger retaliation.

13. Wendt, this volume, 1028.

14. Ibid.

15. See Koremenos 2000, which analyzes a random sample of 149 international agreements. Similar results on the flexibility of human rights agreements also emerge when we look at renegotiation provisions, withdrawal clauses, and amendment provisions.

16. An additional problem arises because we do not present an exhaustive set of conjectures. Nor, as we state in the introduction, do the ones presented all flow from one carefully-specified model.

17. Fearon and Wendt point out that parallel critiques of rational choice in terms of bounded rationality (such as Simon's) have since been significantly clarified and narrowed by subsequent work on incomplete information games. See Fearon and Wendt 2001; and Simon 1957. Ongoing game-theoretic work in the evolutionary tradition is exploring the implications of limits to rationality—including ignorance about possible outcomes. Fudenberg and Levine 1998; and Rubinstein 1998.

18. Shubik makes an almost identical argument regarding the application of game theory to explain economic activity. Shubik 1982.

19. Walt 1999.

20. A stylized fact is a summary statement of a circumstance or relationship that is observed across a wide range of situations and is taken either as an input into a modeling exercise or as something to be explained through it.

21. Pahre (pers. comm. with Koremenos and Snidal) offers the interesting observation that the more formal articles address only one or two of the conjectures, whereas the nonformal articles address between four and nine conjectures. This reflects the power of formal research in exploring the logic of a particular conjecture and the power of empirical cases in examining evidence on multiple conjectures within a case.

22. Mearsheimer 1994/95.

23. Power is notoriously difficult to conceptualize and theorize—both in the traditional and in the formal literature. An example of the latter is the attempt to treat patience (that is, a low discount rate) as a proxy for power in bargaining models, which we find unpersuasive from a substantive perspective.

24. This example is from Kim 1999.

25. Mitchell and Keilbach, this volume, 898.

26. This example draws on Wheelbarger 1999.

27. We thank an anonymous reviewer for bringing up this point.

28. This is not meant to denigrate our focus on design. Quite the contrary, design matters if the substantive provisions are going to come into effect and stay in effect.

29. Of course, one of our design features, scope, is substantive in nature. Indeed, the determination of scope may depend on the availability of powerful states to keep items off the agenda (agriculture out of GATT, for example) or to force them on the agenda (services in the WTO, for example).

30. We view factors such as a state's position in the international economy or its position with respect to the international distribution of power as constraints and environmental conditions, not as preferences. Over time, of course, the environment may condition state preferences. The Rational Design conjectures, however, are comparative static hypotheses.

31. Our categories are not intended to pigeonhole each article uniquely. For example, Richards' consideration of domestic factors fits both “gross characteristics” and “exogenous shocks.”

32. For an in-depth discussion of related issues, see Lake and Powell 1999, especially the introduction and the chapters by Peter Gourevitch and Jeffry Frieden.

33. Rogowski 1989.

34. Rosecrance 1986.

35. Koremenos 2001a.

36. Wendt 1992.

37. Calvert 1995.

38. Koremenos 2001b.

39. Koremenos 1999b.

40. See Knight 1992; and Gruber 2000.

41. A highly instructive example is the long-standing difficulty that states have had in agreeing on televisions standards—even though all would benefit from a common standard—because of future implications for competitiveness. See Austin and Milner 2001; and, more generally, Abbott and Snidal 2001.

42. This echoes Krasner's change of regime and change within a regime. Krasner 1983.

43. Axelrod and Keohane 1986, 251.

44. For an overview of the complementarities and tensions between rational choice and evolutionary approaches, see Kahler 1999.

45. For a fuller discussion, see the volume's introduction.

46. Abbott and Snidal 1998 and 2000. At a more technical level, Wendt misconstrues the implications of the “theory of the second best.” Second best theory does not recommend that we stop trying to “maximize expected utility.” Instead, it provides guidance on certain pitfalls in doing so.

47. In fact, one of the exciting areas of research in international relations has been on the role of normative factors in affecting international outcomes through persuasion and direct action. Margaret Keck and Kathryn Sikkink make the important point that normative actors must act strategically and take into account issues of institutional design if they are to achieve their goals. In short, normative actors need good designs. Keck and Sikkink 1998.

48. For similar reasons, there is no need to presume that institutional designs must be all embracing. The best designs, like property rights in the market, often establish broad parameters within which individual actors determine specific outcomes through their actions.

49. Townsend 1988.