The economics literature typically uses counts of casualties as a measure of conflict intensity despite the fact that the units of observation vary considerably in population size. When analysing the impact of conflict on economic growth, the use of counts relies on the assumption that a given number of casualties affects large and small populations in the same way. Using within‐ and between‐country evidence, this paper demonstrates that this standard assumption can be rejected. A per capita model of conflict intensity that captures local effects of violence provides a more consistent empirical framework for both between‐country and within‐country studies.