Denunciations in autocratic regimes: how and when do they work

States often adopt policies encouraging citizens to denounce their peers to gather information. Are these policies effective? How and when are they effective? We developed a theoretical framework to answer these questions. The state seeks to screen aligned from misaligned citizens. Citizens have information about their peers. Each citizen then decides whether to denounce their peer to the state. The state then decides whether to conduct investigations to obtain additional evidence and decides whether to punish citizens. Our analysis reveals that policies encouraging denunciation are effective when the regime is actually quite popular. Further, these policies work when misaligned citizens are difficult to unmask: the information available to citizens about their peers is noisy, but not too noisy. Guided by the insights from our theory, we evaluate the effectiveness of current policies adopted by the Chinese state.