Democratization reduces the risk of war, but uneven transitions toward democracy can increase the probability of war. Using country-level data on democratization and international war from the period 1875–1996, we develop a general additive statistical model reassessing this claim in light of temporal and spatial dependence. We also develop a new geopolitical database of contiguities and demonstrate new statistical techniques for probing the extent of spatial clustering and its impact on the relationship between democratization and war. Our findings reaffirm that democratization generally does reduce the risk of war, but that large swings back and forth between democracy and autocracy can increase war proneness. We show that the historical context of peace diminishes the risk of war, while a regional context plagued by conflict greatly magnifies it.