The model distinguishes the unintended side effect of expressive behavior of members of the general public from the deliberately exerted power of a winning coalition. The latter alone is the one that decides on ousting an incumbent in light of its observation of public unrest and in light of its updated belief in the character of the incumbent. As a result, there are two filters between the degree of relative deprivation of the general public on the one hand and an overthrow of a dictator on the other: the stochastic element in the outbreak of public unrest and the interest of the winning coalition in ousting the incumbent.
Our approach has a rich set of empirical implications that are discussed in the paper. The most important is that public unrest raises the probability of enforced regime changes but that the threat thereof is, if any, only weakly instrumental with respect to an efficient and effective control of public policy by the general public in an autocracy.