A critique of the Richardson equations

We review, and extend, one of the classic dynamic models of conflict in economics by Richardson (1919) and Boulding (1962). Restrictions on parameters are relaxed to account for alliances and for peace-keeping, yielding new dynamic patterns of conflict. In addition, we explore an incrementalist version of the model as well as a stochastic one and show how this affects its theoretical stability properties. Using Monte Carlo techniques as well as time series analyses based on GDELT data (for the Ethiopian-Eritreian war, 1998-2000), we also assess the empirical usefulness of the model. It turns out that the simulations fail to converge in a large number of cases, and that one important prediction of the model is not borne out by the data. We therefore conclude that the Boulding-Richardson equations are of limited use for modelling (de-)escalation in dynamic conflict.