Logo: to the web site of Uppsala University

In the last decade advances in statistics, computing power, and data collection has led to an increased interest in forecasting within the field of peace and conflict research and to the adoption of a wide range of methodological approaches for making such forecasts. By making use of these more powerful forecasting methods researchers have been able to produce accurate predictions, as well as better inferences, of many different types of contentious politics events and to create operational early warning systems for such events. Adapting these forecasting methods to the social world in which politics and political behavior operate, however, is not without its challenges. This dissertation explores a number of methodological issues and advances in peace and conflict research, both inferential and forecasting oriented, through a series of four papers. In the first paper, I explore trends in democratization and autocratization using dynamic simulation. In Paper II, my co-author and I take aim at the difficulty of modeling and making forecasts with data which contains both excess zeroes and extreme-values. We propose an extreme-value and zero-inflated regression model which we use to replicate a study on the effects of UN peacekeepers on violence against civilians. Paper III explores latent variable modeling by using Markov models to make forecasts for escalation and de-escalation of armed conflicts. In the last paper, I investigate the effects of missing data and imputation techniques on the predictive performance of models. The four papers of the dissertation make several contributions to the growing literature of forecasting within peace and conflict research. First, the dissertation contributes to the methodological aspects of conflict forecasting by developing new statistical tools, Paper II, and adapting tools from other fields to different processes of armed conflict and contentious politics, Papers I & III, as well as by evaluating the practical effects of common choices in data pre-processing on the performance of forecasts in Paper IV. Second, the dissertation contributes to new ways of drawing inferences about conflict processes by anchoring the inferences in the latent state of the conflict processes in Papers II & III, and through the comparison of aggregated simulations to the historical record in Paper I. Lastly, the dissertation makes a substantive contribution to the broader field of peace and conflict research in Papers I & II by contributing to the debate on the waves of democratization and autocratization, and by nuancing the impact of UN Peacekeepers on violence against civilians.