The geometry of security: Modeling interstate alliances as evolving networks

In this article, it is argued that interstate alliances function as public costly signals of state intentions to cooperate militarily, and as such, they should be expected to influence state expectations within dyads, between dyads, and across time. Accurate statistical modeling of interstate military alliances thus requires that researchers escape the assumption of independent units of observation, which is built into most of the statistical tools currently used by international relations scholars, as such models can be expected to produce unbiased parameter estimates in this domain only if the decisions to create and dissolve interstate alliances are formulated in isolated dyadic bubbles. The use of stochastic actor-oriented models, combined with Markov simulations of network evolution, is shown to be a productive alternative method of modeling interstate alliances, which allows the researcher to avoid the assumption of dyadic independence by incorporating theory-driven assumptions about patterns of extra-dyadic interdependence directly into the functional form of the statistical model. The results demonstrate that triadic patterns of amity and enmity exercise powerful influence over the selection of alliance partners and the evolution of the global alliance network. The results also show that failure to incorporate patterns of extra-dyadic interdependence into our statistical models of interstate alliance decisions is likely to result in biased parameter estimates.