Abstract: Civil conflict has been a dominant form of organized violence in human history that can result in massive destruction of the economy, society, and human lives. Besides the causes of initial onset, understanding and preventing civil conflict requires explanation of its two-fold variations. First, violence in civil war varies in the frequency and manner of how it is applied even within single conflicts. Why do the frequency and manner of civil war violence vary across subnational localities? What determines where and how violence occurs in civil conflicts? Second, history tells us that the duration and outcomes vary across civil conflicts. What determines the duration and outcome of civil conflicts? More importantly, do the micro-level dynamics of battles significantly influence conflict duration and outcome at the macro level?
This dissertation investigates these twofold questions at the micro- or subnational-level causes and macro- or country-level consequences of violence in civil conflicts. Taking the contributions of previous studies as the point of departure, this dissertation departs from the single-level approach, which has focused on either micro- or macro-level determinants of the variations observed within civil conflicts. The first part of the empirical analysis in this dissertation explores the micro-level causes of civil war violence. Specifically, it aims to disentangle the impacts of two main classes of determinants of civil war violence found in the previous literature —— the first class includes the set of static factors that are mostly exogenous to conflict dynamics, and the second includes the dynamic factors that are largely endogenous to the dynamics of battles. The computational model incorporated with precisely geo-referenced data demonstrates the importance of endogenous diffusion dynamics in determining where and how violence unfolds during civil conflicts. Diffusion dynamics matter in improving our capability to explain and predict insurgent violence, but they matter relatively more in explaining selective violence and less in predicting indiscriminate violence.
The second part of the empirical analysis examines the macro-level consequences of civil war violence by analyzing how the micro-level dynamics of violence translate into the macro-level variations of civil war duration and outcome. Building upon the bargaining model of war, this dissertation explores the associations between micro-level conflict dynamics and macro-level variations in conflict termination. The two chapters in the second part of the analysis posit that the spatio-temporal dynamics of violence that occur during conflict substantially influence when and how civil conflict ends by altering the severity of the underlying bargaining problems. The relative importance of diffusion dynamics depends on how battles diffuse rather than whether battles diffuse because the different diffusion dynamics affect the expectations and underlying power balance between disputants differently. The empirical results provide strong support for the following theoretical claim: while diffusion of battle activities across distant localities substantially lowers the likelihood of conflict termination regardless of outcome types, their diffusion within proximate localities matters less in altering the prospects for domestic peace.
