Thus far I have attempted to illuminate the dynamics of multiparty civil wars by focusing on how warring groups interacted with each other and with their constituents in some notoriously bloody and protracted conflicts. The main body of empirical evidence in this book (Chapters 3–7) draws on the bloody civil wars in Afghanistan and Bosnia. The warring groups in these conflicts exhibited considerable variation in their alliance choices, shares of power, and identities. These variations allowed for controlled comparisons within and across cases, highlighting the dynamics and mechanisms behind both alliance choices and warring group fractionalization, irrespective of whether the civil war was fought across ethnic or nonethnic lines. The results, identified on the level of alliances, warring groups, and subgroups, were then further probed – where data were available – on the municipality and local commander level.

Overall, what I have found is both simple and, on its face, generalizable. Put succinctly, relative power assessments drive both alliance behavior and warring group fractionalization during conflicts. Groups want to be on the side that is victorious, while also ensuring that they get maximum possible returns from their deal as alliance partners. As warring actors become relatively more powerful on the battlefield, they run the risk of alienating their partners, who are afraid that their rising ally cannot credibly commit to sharing power – but they also become less susceptible to within-group splits and takeovers. Variation in groups’ ethnic, religious, or other identities helps explain the building blocks that form warring groups in the first place, and influences the narratives that group elites construct about their allies and adversaries – but shared identity as a variable does not sufficiently explain the variation in alliance choices and fractionalization outcomes. Relative power, instead, explains that variation, in a fashion notably consistent with the neorealist view of alliance behavior in the international system.