Delegated launch authority

Why would a state make itself safer by giving away its own power to decide on retaliation?

Every instinct says the gravest decision a state can make should sit with one accountable person, be revocable at any moment, and be reached slowly. Yet nuclear command arrangements have repeatedly gone the other way: authority written down in advance for subordinates to use under stated conditions, procedures that fire on warning rather than on damage, and at least one reported system built to work with nobody left to consult. Is that recklessness dressed as doctrine, or is there an argument in which keeping the decision to yourself is the dangerous option?

Begin with what a retaliatory threat has to do. It has to be believed before anything happens, because that belief is the entire product; retaliation itself accomplishes nothing, since by the time it would occur the thing it was meant to prevent has already happened. So the threat is peculiar: it must be credible at a moment when carrying it out would be pointless.

Now take the adversary's side and plan the attack. If retaliation requires a specific person to give an order, what is the highest-value target on the map? Not the missiles — the person, and the wires: the aerials, relays and airborne command posts that carry an order from a decision to a launcher. Decapitation is not an atrocity added on top of a first strike; it is the move that makes a first strike make sense. A leadership that cannot order retaliation has been disarmed without a single silo being hit.

Notice what this does to the original instinct. Concentrating authority in one accountable, always-consultable person does not make the arrangement cautious. It creates a single point whose destruction wins the war, and thereby manufactures the incentive to strike first and strike fast. The safeguard has become the vulnerability.

What removes that incentive? Anything that makes retaliation independent of the survival of the decider. There are three families of answer, and they are worth separating. Harden the force, so the weapons survive whatever happens — submarines at sea being the strongest version, since an ocean cannot be targeted. Harden the communications, so an order still gets through. Or give away the decision — write the authority down in advance for named commanders under stated conditions, or arrange the response to proceed on warning, or build a device that completes the sequence without anyone.

Only the third is what we are asking about, and its logic is the strangest. The state is not delegating because subordinates decide better. It is delegating because an adversary who cannot find a decision to destroy has nothing to gain from destroying anything. What is surrendered is discretion, and discretion was exactly what the adversary was hoping to eliminate. This is the general shape of commitment: you gain bargaining power by removing your own ability to choose, because an option you retain is an option someone else can attack, buy, or wait out.

Two things follow, and both are uncomfortable. First, a commitment device deters nothing unless the adversary knows it exists — a secret doomsday machine is a pure liability, which is why states have signalled such arrangements more than concealed them, and why the reported Soviet system remains partly acknowledged and partly disputed in its actual functioning. Second, every mechanism that raises the certainty of a response also raises the chance of a response nobody authorised. This is a genuine dilemma, not a solved problem: weapons must always work when properly ordered and never fire otherwise, and essentially every device that improves one requirement degrades the other. Compressing the decision into the minutes an intercontinental missile takes to arrive — on the order of half an hour, and much less from close in — means false warnings must be adjudicated at speed, and the Cold War produced more than one occasion on which a warning was judged false by a human who might have been wrong either way.

Two clarifications hold the argument together. Delegation is not the only route to a credible second strike — it is one answer to the decapitation problem, and arguably the worst of the three, which is why states that could afford survivable submarines leaned on those instead. Reading pre-delegation as a preference rather than as a symptom of thin alternatives gets the history backwards.

And "giving away the power to decide" flatters the arrangement. Nothing about it is unconditional in practice; the conditions are written, narrow and hedged. What is really given away is the adversary's confidence that a decision-maker is required at all — a claim about the adversary's uncertainty, which is why the design lives or dies on what the other side believes rather than on what the mechanism does.

• How the same commitment logic appears in ordinary institutions — fixed-rule central banks, constitutional entrenchment — where the point is also to remove one's own discretion.

Every term the collection defines is gathered in the glossary.