Civilian control

Why does an armed force take orders from an unarmed government?

A minister signs an order cutting the army's budget, or sacking its most popular general, and the army complies. On the face of it this is absurd: the institution with every rifle in the country obeys a person who has none, and could stop obeying this afternoon. Peter Feaver put the puzzle sharply — the institution created to protect the community is the one institution capable of destroying it. So what is actually holding, and is it one thing or several?

The tempting first answer is law: the constitution says the army obeys. But a constitution is a document, and the question is precisely why armed men treat a document as binding when they need not. Law cannot be the ground floor here; it is one of the things being explained.

So try a harder question. Suppose a colonel is genuinely considering a coup. What is the decision actually like? Not "should I seize power" — he cannot seize anything alone. It is "will enough others go along?" His brigade must move, other units must not move against him, the capital's garrison must sit still, the air force must not fly. Every one of those officers is asking the same question about him. Nobody's answer is fixed in advance; each depends on what each believes the others will do.

That is a coordination problem, and it has a striking property: the belief is self-fulfilling in both directions. If everyone expects the plot to fail, joining is suicide, so nobody joins, so it fails. If everyone expects it to succeed, standing aside is suicide, so they join, so it succeeds. This explains an otherwise odd feature of coups — the rush to broadcast, to announce victory before it exists. That is not vanity but an attempt to move everyone else's expectation, because the expectation is most of the mechanism.

Now see what routine obedience is doing. Every ordinary day in which orders are followed is evidence to every officer that everyone else expects orders to be followed. The habit is not merely a habit; it is the public signal that keeps the coordination problem unsolvable. Which is why the dangerous moments are the ones that scramble the signal — a lost war, a disputed election, a government that half the officer corps already regards as illegitimate.

What else is at work? Professional socialisation, and it is not decoration. Samuel Huntington's argument for what he called objective control is a bargain: give officers real autonomy over their own professional sphere — doctrine, promotion, how to fight — and in exchange they adopt an ethic of political abstention, treating intervention as a breach of the profession itself rather than as a risky option. Morris Janowitz doubted the sphere could be kept that clean, since senior officers are inevitably political actors, and the argument between those positions is still live.

Then there is deliberate structural work, sometimes called coup-proofing: dividing command so no one officer controls enough force, maintaining parallel armed bodies with separate chains — presidential guards, gendarmeries, party militias — rotating officers before they build personal followings, and running counterintelligence inside the ranks. Each measure keeps the plotter's coordination problem hard. James Quinlivan's point about this is worth keeping, though: these measures buy regime security at the price of military effectiveness, since an army designed so its parts cannot combine is an army that fights badly.

Finally, legitimacy, which enters through the ranks. The colonel must also ask whether his soldiers would fire on a crowd, or arrest a president most of them regard as lawful. Where the answer is no, the guns he nominally commands are not available for that purpose. Civilian control rests as much on what the bottom of the hierarchy will do as on what the top intends.

The most common misreading is to look for the institution that does the work — the constitution, the oath, the civilian defence ministry. There is no such single institution. What exists is a set of overlapping and partly redundant mechanisms: internalised professional norms, the coordination problem, divided and fragmented force structure, the perceived legitimacy of the government, and monitoring. Any one of them can be weak in a country where control nevertheless holds, which is what makes single-factor explanations sound plausible and travel badly.

Two honest qualifications. First, civilian control is not the same thing as democratic control. A one-party state can exercise firm control over its army through the party apparatus, with commissars and parallel political hierarchies, and no election anywhere in the arrangement. Second, none of this is a solved problem. Coups have been common across much of the world within living memory, their frequency has fallen markedly since the end of the Cold War, and why it fell — external sanction, diffusion of norms, better coup-proofing, or something else — is genuinely contested.

• Why coup-proofing measures that protect a regime tend to make its army worse at fighting.
• How militaries influence policy without ever disobeying — delay, leaks, and shaping the options offered.

Every term the collection defines is gathered in the glossary.