From Gaddis to Mars: The Decline of War Enters New Territory

Michael Spagat, Brennen Fagan, Stijn Van Weezel

Research output: Working paper

30 Downloads (Pure)


We provide strong evidence for a decline of war beginning around the end of the Korean War. We survey a considerable historical and political science literature that identifies the Korean War as an important turning point in 20th century war. Next, we argue that the Correlates of War Inter-State data, a workhorse of decline-of-war critiques, is fatally flawed and that these critiques have further technical shortcomings. The more appropriate Project Mars database, which covers all conventional wars in the world between 1800 and 2011, shows a dramatic decrease in post-1950 war arrival rates. Evidence for decline remains good even when we restrict attention to wars with killed in action (KIA) counts above the median count for the whole period. When the metric is KIA per 100,000 of world population then evidence for decline is at least strong in all, and overwhelming in most, KIA ranges. Our Bayesian analysis includes an online interactive application that allows choices of KIA measures, war-size ranges and initial beliefs about the plausibility of decline and then display the implied probability distributions over post-1950 arrival rates together with the empirical pre-1950 arrival rates. They robustly show high probabilities in favour of the decline-of-war thesis.
Original languageEnglish
Publication statusUnpublished - Aug 2022

Cite this