Type of Work:Working Paper
We consider the estimation of the ‘effectiveness’ of states in international relations. Noting that pairwise contest data is the norm in applied research, we motivate a straightforward Bradley-Terry statistical model for this problem from first principles, which will allow for a closer integration of theoretical and statistical practice for scholars in this subfield. The essence of this approach is that we learn about the latent abilities of states from observing conflict outcomes between them. We demonstrate the novelty and appeal of this set-up with reference to previous attempts to capture estimands of interest, and show that for many questions of concern—especially regarding ‘initiation effects’—our approach may be preferred on theoretical and statistical grounds. We show how the model may be used in application, and how its results may be interpreted, with a commonly used international relations conflict data set.