Rebel group interactions in multi-party civil wars

Previous studies of civil war have considered how the dynamics of multi-party conflicts differ from those involving a single rebel group. But few studies have examined the interdependent relationships that exist between the rebel groups themselves. Some groups form alliances, while others engage in inter-group conflict. This thesis contributes to the empirical literature on civil conflict by exploring the understudied area of rebel group interactions. Chapter I reviews the literature, highlights the gaps in previous research and outlines the hypotheses to be tested. Chapter II examines the driving forces for rebel group interactions by testing two competing theories using empirical data. Results suggest that power considerations are the primary driving force behind group interactions, but identity considerations are also found to play an important role. Chapter III explores the conditions that facilitate cooperative versus conflictual interactions. A set of theoretical arguments, based on credible commitments, are proposed and these are tested using empirical data. Results suggest that rebels with high levels of alliance credibility are more likely to cooperate with their peers, whereas groups who lack alliance credibility are more likely to engage in inter-rebel violence. Chapter IV examines the effect of interaction strategies on the survival and termination-type of rebel groups. Results of empirical data analysis show that interaction strategies have no effect on group longevity. Groups who engage in inter-rebel violence are more likely to terminate by peace agreement, whereas groups who form alliances are less likely to terminate by peace agreement. Allied groups are also less likely to suffer defeat. Chapter V investigates how computer based simulation techniques may be used to model rebel group interactions. A model of two-sided conflict is developed and extended so that multi-party conflict simulations can be performed. Suggestions regarding the modelling of interaction strategies are proposed and conclusions are drawn, which emphasise the relative advantages of differential equation modelling and agent-based simulation