Why do dictatorships sometimes allow opposition groups to publish media, but at other times forbid opposition media? I propose a theory that dictators tolerate opposition media selectively in order to limit protests. I formalize the theory in a signaling model, from which I derive several testable empirical implications. I illustrate the logic of the theory with a case study of the Ben Ali dictatorship in Tunisia during its first five years (1987-1992), based in part on interviews I carried out in Tunisia. I show how the theory explains variation in Ben Ali's willingness to allow opposition media, across both time and opposition groups. To test the implications of the model quantitatively, I construct a panel dataset on ten Arab countries with authoritarian regimes during 1992-2017. The data measure which regimes allowed opposition groups to produce media in which years and are based on my research on a wide range of opposition groups (of various ideologies and legal statuses) and of media (including newspapers, websites, and TV channels).

I find that dictators allow opposition media when their regimes are most likely to survive an uprising, in order to signal their strength to citizens and discourage them from protesting. In particular, in years when authoritarian regimes experience strong economic performance – including low unemployment, high economic growth, and plentiful revenue from oil and natural gas – they are far more likely to permit opposition media. After the "Arab Spring" uprisings of 2011 revealed that the region's authoritarian regimes were more vulnerable to mass unrest than they previously appeared, those regimes became much less likely to tolerate opposition media. By advancing a new theory and analyzing original empirical evidence, this study contributes to our understanding of why media freedom varies in authoritarian regimes.