The hypothesis that democratization triggers political violence has been proposed repeatedly in the quantitative literature, but remains controversial both with respect to interstate and civil wars. Current empirical research continues to be afflicted by methodological and data problems related to the measurement of democracy and the task of detecting changes in such scores. In order to gain further clarity into the link between democratization and civil war, the current study introduces a new period-finding algorithm that is able to detect periods of democratization and autocratization. This allows for a more flexible way of finding directional changes in governance indicators than is possible with the rigid lag structures commonly employed in previous studies. When regressed on various measures of civil-war onset, the indicator for the initiation of a period of democratization has a strong and robust effect on conflict even in the presence of static measures of regime type. The same applies to autocratization, but its impact is much more sudden than that of democratization. Moreover, we find that the democratization effect is limited to governmental, rather than to territorial, conflicts. Further research will be needed to confirm these results in terms of the relevant causal mechanisms, especially in ethno-nationalist civil wars.