Elections and the Development of Civil War: An Empirical Study of Sri Lanka

Abstract

ABSTRACTPresent empirical research on the connection between elections and civil war often treats causality as a one-way effect because it focuses exclusively on either the onset or the consequence of civil war. What is omitted in the literature is the endogenous nature of the election–conflict nexus throughout a civil war. This article dedicates itself to filling that gap. Treating administrative units in conflict as part of a rebellion network, we apply a stochastic actor-oriented model (SAOM) to the case of Sri Lanka. We find that, on the one hand, rebellion is less likely to occur in Tamil regions if national winners in presidential elections enjoy high local approval ratings (selection effect). On the other hand, regions involved in the rebellion network converge in terms of their local support for the national winner of the presidency (influence effect). Overall, our model-based simulation analysis shows that the influence effect has a larger impact on the endogenous relationship than the selection effect.