In Dangerous Alliances: Proponents of Peace, Weapons of War, Patricia Weitsman develops a theoretical framework which synthesizes major insights of the alliance literature to better explain alliance formation and alliance cohesion. Weitsman takes "threat" to be her main explanatory variable, and in so doing proceeds from Stephen Walt's seminal work, The Origins of Alliances, which argues that states balance against threats rather than against power.

In the first step of her theory, Weitsman argues that variation in threat level leads to different types of alliance formations-hedging, tethering, balancing, or bandwagoning. She argues that there is a curvilinear relationship between threat level and alliance type. Low threat levels lead states to form low-commitment hedging alliances with both friends and potential adversaries, medium threat levels lead states to form tethering alliances with their adversaries in the hopes that through their mutual alliance adversaries can restrain one another, high threat levels lead states to balance against their adversaries, and very high threat levels lead states to bandwagon with the threatening adversary.

In the second step of the theory, Weitsman explains alliance cohesion with threat source; that is, cohesion varies depending on whether the greater threat comes from inside or from outside the alliance itself. According to the theory, a combination of low internal threat and low external threat leads to moderate or low cohesion, low internal threat and high external threat yields moderate to high cohesion, high internal threat and low external threat leads to low or no cohesion, and finally high internal threat and high external threat makes cohesion "difficult though not impossible" (p. 26).

Weitsman then looks for empirical support for her arguments by process-tracing eight case studies: the two Leagues of the Three Emperors (1873 and 1881), the Dual Alliance between Austria-Hungary and Germany (1879), the Triple Alliance (1882), the Franco-Russian Alliance (1891/94), the Triple Entente between France, Russia, and Great Britain (1907), and the Central Powers and the Triple Entente from 1914-1918.

Dangerous Alliances is important reading for anyone studying alliances. It makes sense of seemingly contradictory claims in the broader alliance literature, it usefully delineates the range of possible alliance types, and it introduces several new analytic concepts such as hedging and tethering.

A closer look at the empirical cases, however, raises doubts that alliance type varies with threat level in the way the theory predicts. In the "Summary of Propositions" Weitsman writes, "at relatively high levels of threat, states will seek to ally with their adversaries" by tethering (p.30). But in the cases, tethering occurs sometimes under conditions of "moderate" threat levels (e.g., Austria-Hungary with Germany in the Three Emperors' League 1873-1878 (table 3.1-3.2, p. 45, 47)) and sometimes under conditions of "high" threat levels (e.g., Austria-Hungary and Russia in the Three Emperors' League 1881 (tables 3.3-3.4, p. 57, 59) and Italy and Austria-Hungary in the Triple Alliance of 1882 (tables 4.3-4.4, p. 87-91)). Balancing never occurs under "high" threat levels, as the theory proposes (p.30), but rather only under "very high" threat levels (Germany against France in the Three Emperors' League 1881, the Dual Alliance 1879 (tables 4.1-4.2, p. 74, 76), and the Triple Alliance 1882). "Very high" threat levels is supposed to cause bandwagoning according to the theory (p. 30). Bandwagoning, however, does not seem to occur at all in any of the cases. Instead, in the several cases of "very high" threat levels, the states in question all chose a balancing alliance. Moreover, although the relationship between threat level and alliance type is theorized to be curvilinear, the study has at least one case where two alliance types were used in combination against the same threat. Faced with a "high" threat from Russia, Austria-Hungary simultaneously pursued balancing against Russia in the Dual Alliance with Germany from 1879-1914 (p. 55) and tethering with Russia in the Dreikaiserbund from 1881-1887 (p. 73). The broader problem here is that if the same threat level can lead to several different or even a combination of alliance types, then threat level alone cannot account for variation in alliance type.

One reason for inconsistencies between the cases and the theory could be that other variables are doing the causal work. A shortcoming of this study is that it does not consider alternative variables. Threat level is certainly a plausible reason for states to form alliances in the first place, but which type of alliance formed might be determined by something completely different, such as the costs of the alliance type and the availability or reliability of alliance partners. A state might pursue tethering rather than balancing in the face of a high threat because it has no available balancing partners. A state might decide to pursue balancing rather than bandwagoning against a very high threat because bandwagoning is too costly. Or a state might choose to pursue both tethering and balancing because it cannot rely on its balancing partner. Because the study does not control for these alternative variables, it is difficult to know how to evaluate their relative explanatory value.

A second reason why inconsistencies between the cases and the theory might exist is because the threat level variable is difficult to operationalize. Weitsman measures threat level as a function of military expenditures, military size, industrial and technological resources, population, proximity of the homeland, proximity of colonial holdings, and intentions (p. 34). In the study, an increase or decrease in any one of these variables constitutes a respective increase or decrease in threat. But what Weitsman does not tell us is what constitutes moderate, high, or very high threat levels in the study. We do not know what the levels are relative to or what the coding rules are. Not knowing what counts as a high or low threat level is particularly problematic for deriving alliance behavior predictions from Weitsman's theory. Weitsman's theory should, for example, be able to tell us something about the alliance behavior between the US and Europe today. Many analysts are wondering why, contra to traditional balance of power theory, Europe is not balancing against the US. In order to do so, however, we would have to know what data would constitute which level of threat.

The second major argument of Dangerous Alliances-that threat source explains alliance cohesion-is well-supported by the empirical cases. As with part one of the argument, however, the strength of the empirics is unsettled by the lack of consideration for alternative variables and by the difficulty in operationalizing cohesion.

The empirics show that low internal threat is a necessary but insufficient condition for alliance cohesion. Even with low internal threats, however, our cooperation theories tell us that collective action could be difficult. Given that alliances are a type of cooperative behavior, it is surprising that Weitsman does not take the usual collective action variables into account. Agreement on common goals and strategies could be hindered by an unequal distribution of costs or responsibilities, by alliance size and fears of free-riding, or by lack of shared values. Conversely, an equal distribution of costs or responsibilities, small alliance size, or highly shared values might promote cohesion even when the internal and external threat levels are both low. The study shows that threat source is a plausible variable, but it does not tell us how well high external threat compares to other important and relevant variables.

Perhaps more dissatisfying with this part of the argument is Weitsman's operationalization of cohesion. Weitsman defines cohesion as agreement on goals and strategies for achieving those goals (p. 35). "When states agree on mutual goals and strategies toward those ends, I will interpret cohesion as high. When states cannot come to any agreement over their goals and strategy, or if they undermine their allies, I will interpret cohesion as low" (p. 36). Unfortunately, this dichotomous operationalization leaves a lot of room for ambiguity. The very creation of any alliance is an agreement on some goals and strategies, so all alliances should be considered cohesive (to some degree or another), as defined here. In addition, although the term cohesion seems to capture the quality of an alliance, the definition used here makes no distinction between reaching agreement on difficult issues or reaching agreement on simple issues-any kind of goal agreement seems sufficient to count as cohesion and is treated the same by the definition. As Weitsman recognizes, cohesion is a difficult concept to operationalize, but the result of that thorniness is uncertainty about what the dependent variable is really assessing.

Despite these shortcomings, Dangerous Alliances takes an important step forward in our understanding of alliances. Moreover, it lays a firm foundation upon which variable refinement and further testing can be carried out. Future studies should take it as a building block within the alliance literature.

Lora Viola, University of Chicago

Back to top