Luke John Keele, Dissertation Advisor/Co-Advisor Glenn Hunter Palmer, Committee Member Christopher Jennings Fariss, Committee Member Guido Cervone, Special Member
Keywords:
conflict genocide forecasting prediction mass violence
Abstract:
Forecasting conflict, for many years the domain of qualitative analysts, pundits, and amateur Nostradamuses, has become increasingly practical thanks to the application of quantitative models. Government groups view these models as key to developing early warning systems for conflict and related violence. As a result, much of the early work in this area was spearheaded government-funded academic research and has only recently begun to spread more generally, even into the private sector. Previous and current models tend to focus on generating probabilistic forecasts at a country-year level. This approach works for some problems, but not everything fits into a country-level format. With a focus on violence against civilians as the outcome of interest, I explore how to improve existing early warning models and develop models to generate forecasts at a subnational and subannual level.
