Earlier versions of this article were presented at the Rational Design conference in May 1998 and the 1998 Annual Meeting of the American Political Science Association. For comments on earlier drafts, I thank discussants George Downs, Robert Keohane, and Lisa Martin; the editors of this volume and IO; the anonymous reviewers; and the participants in the Rational Design conferences.

1. See Koremenos, Lipson, and Snidal, this volume.

2. Ruggie 1993.

3. See Curzon and Curzon 1976; Finlayson and Zacher 1983; and Pahre 1999, chap. 10.

4. The data in this are from the Trade Agreements Database, available at ⟨http://www.staff.uiuc.edu/~pahre/tad.html⟩. This database currently contains all trade treaties signed from 1815 to 1913, except for Asian countries east of the Ottoman Empire and west of Japan (exclusive). For a treaty to be included it must make mutual reductions or bindings that directly affect bilateral trade; treaties concerning purely navigational matters or commerce alone (such as treaties granting reciprocal rights of establishment) are excluded, as are the unequal treaties signed by China and some other countries. The sources for the database are discussed on the website and include both official treaty series and secondary sources, such as diplomatic histories.

5. For the analysis of regime spread as an emergent property of the nineteenth-century trade treaty network, see Lazer 1999.

6. Krasner 1983, 1.

7. For a study of how the origins of trade institutions affect their present functioning, see McGillivray, McLean, Pahre, and Schonhardt-Bailey 2001.

8. Dion 1998; compare King, Keohane, and Verba 1994.

9. See Keohane 1990; and Ruggie 1993.

10. Lipson 1986.

11. The figure distinguishes MFN treaties from non-MFN treaties for reasons I explain later. While it excludes Zollverein treaties between, say, Prussia and Bavaria, it does include Prussia's treaties with outside states that had territories within the German Confederation, such as Austria, Denmark (Holstein), Luxembourg/Netherlands, and for a time the United Kingdom (Hanover).

12. See Marsh 1999, chap. 8; Weitowitz 1978; and Werner 1989.

13. For Britain, see Marsh 1999. Belgium lacks a modern secondary literature (but Mahaim 1892), though one can piece together the details from Augier 1906; Marsh 1999; and others.

14. See Lohmann 1997; and Pahre 1994. While similar to player linkage, clustering concerns the number of actors and not the number of issues being negotiated (that is, scope). While such linkage may have interesting implications, this kind of linkage is not my focus here.

15. Morrow, this volume.

16. See Koremenos, Lipson, and Snidal this volume; and Lake 1996.

17. Hoekman 1993. This argument parallels transaction-cost explanations of multilateral negotiations and multilateralism. For example, see Keohane 1984, 90; and Oye 1986, 20. For a critique, see Pahre 1994.

18. Hoekman 1993, 44–45. Compare Bagwell and Staiger 1999.

19. Finlayson and Zacher 1983.

20. Ruggie 1993; especially Burley 1993; and Goldstein 1993.

21. See, among others, Aggarwal 1985.

22. See Lohmann 1997; McGinnis 1986; Pahre 1994; Stein 1980; and Tollison and Willett 1979.

23. See Axelrod 1984; Keohane 1984; Martin 1992; Lohmann 1997; Oye 1986; Pahre 1994; Snidal 1991; and Yarbrough and Yarbrough 1986.

24. See Fearon 1998; Garrett 1992; Garrett and Tsebelis 1996; Krasner 1991; Oatley and Nabors 1998; and Snidal 1985.

25. See Davis 1997, chap. 7; and Henderson 1984.

26. Smith 1980, 182–95.

27. Ibid., 191. See also Marsh 1999, chap. 6–7.

28. I treat MFN as an exogenous regime rule. There are few modern explanations of MFN, but see Bagwell and Staiger 1999. The United States saw a vibrant debate on MFN culminating in the 1920s; for a good example, see Viner 1924.

29. See Grossman and Helpman 1995; Hillman, Long, and Moser 1995; Milner and Rosendorff 1997, app. B; Milner and Rosendorff, this volume; and Pahre 1998.

30. In a more elaborate model, the indifference curves would be elliptoids because home tariffs are more important than foreign tariffs to each government.

31. Nash 1950.

32. It is ironic that A might rationally anticipate the outcome of its own myopic behavior. One way to square this circle is to suppose that a rational institutional designer anticipates that subordinate organizations will behave in a boundedly rational way; another is that today's government might anticipate myopia on the part of a successor government.

33. Both public presentation and anonymous review of this model have yielded this reaction.

34. Using backwards induction instead of myopia for this “intuitive” approach yields the same results: A withholds concessions from C at the AC node in order to make them at the AB node.

35. If B does not insist on parity, A could get away with making sufficiently few concessions to B that it might end up as well off as it does under clustering; however, I cannot see any reason why B might be nice to A in this way.

36. To see that the concessions in the AC negotiations do not affect the NBS, consider that the outcome of the AC negotiations can be washed away with a monotonic transformation of A's and B's utility functions without affecting the AB negotiation problem. Because the NBS is resistant to monotonic transformations of the utility functions, this has no effect.

37. Conditional MFN, practiced mostly by the United States in the nineteenth century, has ambiguous status here. The ostensible American insistence on additional concessions in exchange for generalizing each MFN reduction often did not matter, since the United States had a single-column tariff. Even under the Dingley two-column tariff, all trading partners received the lower tariff within a few years. For discussions, see Fisk 1903; Lake 1988; O'Halloran 1994; and Viner 1924. Such examples mean in effect that conditional MFN is more an empirical question than a theoretical one.

38. Deardorff and Stern 1992. Multilateral bargains have higher transaction costs because they include irrelevant negotiations; A could participate in matters that affect only B and C.

39. Pahre 2001. Two modern explanations focus on MFN-cum-reciprocity as a renegotiation-proof bargaining rule or as a way to increase the sanction against defection. Viner's classic analysis would also fit well into any modern theory of political economy. See Bagwell and Staiger 1999; Pahre 1994; and Viner 1924.

40. Williams 1972, 258–61.

41. Rephrased, this argument maintains that non-MFN is sufficient for nonclustering. A statement of this form is logically equivalent to the statement that MFN is necessary for clustering: ⟨∼x → ∼ z⟩ ↔ ⟨x ← z⟩.

42. I am making logical claims of the form that ⟨x ∩ y⟩ → z and x ← z. MFN and economic structure are each an “insufficient but necessary part of an unnecessary but sufficient” (inus) condition. For an influential philosophical explanation of why we would expect most causes to take this form, see Mackie 1965 and 1974; see also Kim 1971. For a probabilistic critique, see Trenholme 1975; and Friedman 1980. The inus formulation concedes that we cannot know whether a given cause is truly necessary for a given effect without knowing all other imaginable theories that might explain that effect. This is clearly an impossible standard. Phrased differently, this formulation concedes that MFN is only a necessary condition for clustering within this theory. Since we do not have other theories of clustering, we can treat MFN as a necessary condition here.

43. Braumoeller and Goertz have recently proposed techniques for testing necessary conditions quantitatively; the dichotomous variables here allow for a simpler approach. Braumoeller and Goertz 2000.

44. Annual treaty initiations provide an adequate but not ideal measure. States cluster by negotiating simultaneously with many states, and by making sure that they reach agreement on particular tariff lines simultaneously. These agreements need not be reached in the same year, and unrelated issues sometimes drag out the final treaty. This annualized data series treats agreements reached in January and February as clustered, those reached in December and January as unclustered. These limitations can be overcome through sensitive use of the data, supplemented by qualitative evidence.

45. Because the null hypothesis of this test is that the data are Poisson distributed, the test might wrongly code some actual clustering as Poisson distributed. However, for almost all cases in Table 1 the confidence level for rejecting the Poisson null is either less than .10 or greater than .90 (these latter levels are not shown as such in the table). For these data, then, wrongly accepting the null is unlikely to be a problem. Only Serbia presents a serious question of inference in these data.

46. See Braumoeller and Goertz 2000. While this necessary condition would not be falsified even if none of the MFN treaties were non-Poisson distributed, such a condition would not be very interesting or useful.

47. Testing a correlative claim between MFN and clustering would raise more complicated issues, especially since both MFN and clustering are dichotomous variables.

48. The United Kingdom, though substantively important in international trade, was relatively unimportant to the treaty regime because it chose to rely on unilateral free trade accompanied by MFN treaties and only occasional bargaining over tariff lines after 1860. For details, see Marsh 1999.

49. The early date for Portugal reflects the Methuen Treaty with Britain (Portugal denounced the treaty in 1836) and not a general policy of trade treaties. Sweden-Norway was not very active despite signing an occasional treaty throughout the century, and Russia negotiated trade treaties only very rarely until the 1890s.

50. The countries of Latin America, not discussed here, provide many more such examples.

51. See Kasaba 1988, chap. 3; Pamuk 1987; and Shaw and Shaw 1977, 71–106.

52. Brauneder and Lachmayer 1980, 112–33.

53. See Andersen 1958; and Jones 1970, 72–90.

54. See Black 1943; and Crampton 1997.

55. See Lazer 1999; and Pahre 2001.

56. von Bazant 1894, 28–29.

57. For a review of these models and discussion of how they relate to economic theory, see Deardorff 1984.

58. See Lindberg 1983; Marsh 1999, chap. 8; and Matis 1973, 51.

59. Marsh 1999, 137.

60. Another trick that states used to combine protectionism with MFN treaties was greater tariff differentiation. Sweden used differentiation of iron export duties to good effect to force Germany to make major concessions. See Lindberg 1983; and Werner 1989, chap. 1.

61. Examples of the historiography include Coppa 1970 and 1971; Friedman 1978; Lindberg 1983; Howe 1997; Marsh 1999; Platt 1968; Rogowski 1989; Smith 1980; Weitowitz 1978; and Werner 1989.

62. Examples include Marsh 1999; and Weitowitz 1978.

63. Dion 1998.

64. “als eine Gratisprämie für alle Staaten, die Gegenkonzessionen nicht machen können oder wollen, eine Klausel, durch welche jede Zollverabsetzung, die irgend einem Staate für irgend eine Konzession gewährt wird, sofort und ipso facto auch allen anderen Staaten zugute kommt.” Cited in von Bazant 1894. 34–35.

65. “daβ bei kooperativen Handelsvertragsverhandlungen mit Italien … das Maβ unserer Konzessionen, nicht nur durch die uns direkt angebotenen italienischen Konzessionen, sondern auch durch die Aussicht bestimmt wird, weitere Konzessionen indirekt durch den österreichisch-italienischen Vertrag zu erhalten.” Cited in Weitowitz 1978, 144.

66. See Marsh 1999, chap. 5–6; and Weitowitz 1978, 58–59.

67. Weitowitz 1978, 55–56.

68. “denn sie ermöglichte, die widerstreitenden Interessen der Agrarier und Industriellen in Öster-reich-Ungarn zu beschwichtigen. Der Wiener Regierung wurde es hierdurch leichter gemacht, Deutschland industrielle Konzessionen zu gewähren, in der begründeten Hoffnung, Zollvorteile auf dritten Märkten zu erlange.” Weitowitz 1978, 56.

69. See Marsh 1999, chap. 8; and Weitowitz 1978, 115–16.

70. Weitowitz 1978, 83–84, 154–55.

71. Lindberg makes exactly this argument for Sweden a decade later. Lindberg 1983.

72. Weitowitz 1978, 91, 104.

73. Ibid., 114–15.

74. Ibid., 136, 143.

75. “Was er ‘Pression’ und ‘Drohung’ nennt, ist also nichts anderes als der durch die Kooperationsidee und die Meistbegünstigung geschaffene, untrennbare Zusammenhang der verschiedenen Verträge, den er selbst mit Annahme unserer Bedingung anerkannt hatte.” Cited in Weitowitz 1978, 144.

76. For two otherwise different examples among many, see Curzon and Curzon 1976; and Ruggie 1993.

77. Marsh 1999.

78. Winham 1986, 62–63.

79. See Finlayson and Zacher 1983; and Bagwell and Staiger 1999.

80. Of course, quota negotiations would require a different kind of model than one focused on tariff setting.

81. Aggarwal 1985.

82. See Rosendorff and Milner, this volume.