Shifting sands: Explaining and predicting phase shifts by dissident organizations

Why does a dissident group go through phases of violence and nonviolence? Many studies of states and dissidents examine related issues by focusing on structural or rarely changing factors. In contrast, some more recent work focuses on dynamic interaction of participants. We suggest forecasting state-dissident interaction using insights from this dynamic approach while also incorporating structural factors. We explore this question by offering new data on the behavior of groups and governments collected using automated natural language processing techniques. These data provide information on who is doing what to whom at a directed-dyadic level. We also collected new data on the attitudes or sentiment of the masses using novel automated techniques. Since obtaining valid and reliable time-series public opinion data on mass attitudes towards a dissident group is extremely difficult, we have created automated sentiment data by scraping publicly available information written by members of the population and aggregating this information to create a pollof opinion at a discrete time period. We model the violence and nonviolence perpetrated by two groups: the Tamil Tigers in Sri Lanka and the Moro Islamic Liberation Front in the Philippines. We find encouraging results for predicting future phase shifts in violence when accounting for behaviors modeled with our data as opposed to models based solely on structural factors.