Coercive institutions exist to safeguard an autocratic ruler in power. Their construction is among the most fundamental tasks he can undertake, and their performance determines his longevity in office and the manner of his exit. These institutions also matter deeply for the millions of ordinary citizens that the autocrat governs, because they affect the levels of violence experienced by those people.

This chapter outlines the logic behind the creation and operation of autocratic coercive institutions. The first half of the chapter begins by reviewing the threats that an autocrat must address to stay in power. It then identifies how different threats lead to the creation of distinct types of coercive institutions, proposing that autocrats face a “coercive dilemma” because managing the risk of a coup and managing the threat of popular unrest call for different optimal levels of fragmentation and social exclusivity: coup-proofing calls for fragmented and exclusive coercive institutions, while managing popular unrest is best accomplished using a unitary and inclusive internal security apparatus. Ultimately, autocrats negotiate this organizational tradeoff by crafting their coercive institutions to address the dominant perceived threat at the time they come to power.

The second half of the chapter turns from using coercive institutional design as the dependent variable to making it an independent variable that explains variations in state violence. I isolate two primary mechanisms by which coercive institutions affect levels of state violence, which I term the intelligence pathway and the incentives pathway. Fragmentation and exclusivity both provide incentives, either social or material, that predispose coercive agents toward higher levels of and less discriminate violence. Both characteristics also hamper the collection, analysis, and transmission of intelligence that might enable more targeted, pre-emptive, and non-violent repression, thereby raising the probability of higher levels of and less discriminate violence. Through these pathways, the coercive apparatus has an independent effect on levels of state violence that cannot be accounted for by other factors.

The fact that authoritarian coercive institutions have varying intelligence capabilities also suggests a set of fairly restrictive conditions under which we should see institutional change, and suggests the likely direction of such change when it occurs. The chapter therefore outlines a set of predictions for when and how coercive institutional change should happen, and contrasts this theory of change with explanations grounded in path dependence and rational design.