A Competing Risks Model of War Termination and Leader Change

Recent research suggesting that leader transitions increase the probability of war termination is based on the assumption that leader change is exogenous. However, the exogeneity of leader change needs to be tested, not assumed. This paper uses a bivariate discrete survival model to test the exogeneity of leader change and correctly estimate its partial effect on war termination. The paper extends the analysis by estimating a competing risks model of types of leader transitions. The evidence shows that leader change in large coalition systems never increases the probability of war termination, while leader change in small coalition systems never reduces the probability of war termination. In short, leader transitions in autocratic systems are more likely to bring interstate war to an end than leader transitions in democratic ones. The study also shows that the marginal effect of leader change fades away as the war progresses, and that war has negative duration-dependence.