This article assesses the robustness of the liberal or ‘Kantian’ peace propositions by challenging two common practices: pooling data for different geographic regions, and using conflict at any level as a proxy for interstate war. The findings indicate that there are substantial differences between regions in the effects of democracy, economic interdependence, and international organizations. Conflict (all MIDs) and war have considerably different relationships to these key variables, and to each other, across regions. While I do not argue that these results undermine the general Kantian peace propositions, they do represent powerful qualifications that provide insight into theoretical foundations and raise related questions of specification error. They also point to the continuing importance of concepts such as security communities and norms as liberal factors distinct from the Kantian variables.