Files
Abstract
How do armed groups choose their allies? While early research presented security as armed groups’ overriding motivation, more recent work has contended that factors like ideological overlap and state sponsorship also influence who allies with whom. In this paper, I test several of these arguments using a novel dataset of armed group alliances in Burma from 1950 through 2016, which represents the country’s rebel alliances as an evolving social network. I draw on several theories of armed group cooperation to specify a dynamic, mixed-membership stochastic blockmodel–and then use that model to predict armed group alliances out-of-sample. Assessing the model’s predictions alongside qualitative evidence, I find mixed support for theories that attribute alliances to state sponsorship and shared ideology. Knowing armed groups’ state sponsors does not help predict who will ally with whom. While ideological overlap helps predict alliances somewhat, qualitative analysis of the model’s successful predictions reveals ways that existing theories fall short. I find that theories framing ideological constituencies and state sponsors as commitment devices lack clear empirical implications at the micro level. Moreover, I argue that this focus on commitment devices has distracted researchers from simpler theories with clearer empirical implications. I conclude by suggesting that future research into armed group alliances should use more qualitative evidence and descriptive social network analysis.