The aims of this paper are (a) to construct probability models for peacetime and wartime alliance formation, (b) to provide international political interpretations of the probability models, and (c) to obtain some policy implications from these models.
Since Horvath and Foster published their seminal work in 1963 on probability models of wartime alliance sizes, no further research has been undertaken to elaborate their ideas (except for Yamamoto, 1974; and Yamamoto and Bremer, 1980), even though varied probability models have been applied to the analysis of war occurrences (see Richardson; Singer and Small, 1972; and Wilkinson) and of peacetime alliance formation (see McGowan and Rood, and Siverson and Duncan). Horvath and Foster found that the frequency distribution of wartime alliance sizes is accounted for by a Yule distribution which has a long tail (meaning that large alliances are likely to form in war) and that a Yule distribution can be derived from a process in which a war alliance will acquire new members with.a probability directly proportional to its current size (i.e., the number of participating nations in the alliance) and in which the members of alliances are not allowed to leave until war ends.