Weakness Is Not Leverage

It is among the most durable beliefs in modern statecraft. That specific conversion — pressure in, concessions out — has a predictive record roughly comparable to astrology. Other mechanisms do work, and we will get to which ones. That one doesn't.

Foreign Affairs recently argued that a materially weaker Russia keeps fighting in Ukraine because Putin sees little reason to negotiate — that degradation has not produced the pressure the theory promised it would. Treated as news, that is a puzzle. Treated as data, it is the base rate. The exceptions are the cases where coercion worked, and they are rarer and more specific than the doctrine admits.

Strip the names off the file and the pattern is legible across every century with decent record-keeping: a state does not surrender when it is losing. A state surrenders when the people who make the decision can survive the surrender — and, as it turns out, when they can believe the survival on offer. Those are two entirely different conditions from "losing," and confusing them is the single most expensive error in the coercion literature.

i · two ledgers, one word

The theory assumes a state is a unitary cost-calculator — a single mind totalling up damage and revising its position when the sum gets too large. This is a convenient fiction for modeling and a catastrophic one for policy.

The state doing the totalling is a set of specific people occupying specific positions, and the costs of continuing are borne almost entirely by other people. Conscripts, not ministers. Provincial towns, not the capital's inner ring. The pensioner whose currency evaporated, not the man who decides whether the war continues. Pressure applied to a country is pressure applied to a population; the deciders sit on the far side of a very effective firewall, and their share of the pain is a rounding error against the thing they actually price.

What they price is the survival of the structure that makes them a subject at all.

Run the two ledgers side by side and the "irrationality" dissolves. On the national ledger, continuing is ruinous — dead men, burned equipment, a decade of compounding poverty. On the regime ledger, continuing is a bet with terrible odds and a nonzero payoff, and the alternative is a payoff of precisely zero: the dock, the noose, the balcony, the culvert outside Sirte. Under those terms a ten-percent chance of muddling through is not stubbornness. It is the dominant strategy. There is nothing to be coerced with, because the worst outcome is already fully priced in.

Every mechanism the doctrine relies on inverts here. Degrading the army removes the institution most capable of removing the leader — juntas depose, ministries don't. Sanctions collapse the private economy and leave the state-adjacent one standing, which increases the share of national wealth flowing through hands the regime controls. Isolation kills exactly the domestic constituency that wanted integration. And an external enemy is the oldest legitimacy subsidy in the catalogue; it has been in continuous use since the Peloponnesian War and it has never once needed a rebrand.

The coercion worked. It just did the opposite of what it was supposed to.

None of this is exculpatory, and the cold framing shouldn't be mistaken for one. Explaining why an aggressor continues is not a claim that both parties face symmetric choices. One of them chose to start. The asymmetry in stakes is real precisely because the aggressor manufactured it — a regime that stakes its survival on a war of conquest has built its own trap, and the fact that the trap now makes it dangerous is a consequence, not a mitigation.

ii · the exit is load-bearing

Look at the cases where coercion actually produced capitulation. The variable isn't the amount of damage. It's whether someone built a survivable exit and made it credible.

Japan, 1945. The war was arithmetically lost by mid-1944 — after Saipan, the home islands were inside bomber range and the shipping that fed the empire was gone. The government fought thirteen more months, absorbed the firebombing of every major city and two atomic weapons, and the deadlock in the Supreme War Council was never about whether Japan could win. It was about the kokutai — whether the imperial institution survived. Unconditional surrender, declared at Casablanca in 1943, had specified that the exit did not exist. Tokyo capitulated within days of receiving a formulation that left the Emperor's position ambiguous enough to be survivable, and the occupation subsequently kept Hirohito on the throne. The exit was built at the last possible moment, in deliberately vague language, and it was the thing that ended the war.

Germany got no such construction and fought to the point where the decision-maker's office was a bunker under a burning capital. Paraguay, in the War of the Triple Alliance, fought until Francisco Solano López was killed in the field at Cerro Corá in 1870, having presided over demographic losses historians still argue about because the plausible range runs from catastrophic to nearly unbelievable. Jefferson Davis was arguing for continued guerrilla resistance in the spring of 1865 while Lee, who had a survivable exit and knew it, took it at Appomattox.

Then the modern controls — including the one that nearly breaks the thesis, which is the reason to run it first.

In 1991, the coalition stopped short of Baghdad and Saddam Hussein accepted a ceasefire within days, because the terms left him in power. He demonstrated as much immediately by crushing the uprisings that followed. Straightforward.

2003 is the case usually filed as no exit was offered, and that is simply false. On March 17, 2003, Bush stood in the Cross Hall and gave Saddam and his sons forty-eight hours to leave Iraq. Exile arrangements had been floated through regional intermediaries besides. The exit was public, explicit, and declined.

So availability alone does not predict the behavior, which means the thesis as usually stated is incomplete — and the missing term is the one this piece has been circling. An exit is not an offer. It is a prediction about the party making it. By March 2003 Saddam held twelve years of evidence on that particular party: a coalition that had stopped short in 1991 and then spent a decade under a declared American policy of regime change, sanctions that never lifted, no-fly zones that never came down. Forty-eight hours is not enough time to reverse a twelve-year record even if the offer was sincere, and he had no instrument for telling sincere from not. He did not refuse an exit. He refused a promise from a party whose promises had a documented expiration date.

Same man, same relative military weakness, opposite behavior across the two wars. The variable that moved was not whether an exit existed. It was whether it could be believed.

Milošević, read that way, stops being an anomaly. He capitulated after seventy-eight days of NATO bombing in 1999 while under indictment from The Hague — which the doctrine says should have locked him in to the end. But the indictment was an international instrument and his power was domestic, and in the spring of 1999 there was no precedent for an international instrument reaching inside a sovereign state to collect a sitting head of government. He was not being irrational about his exit; he was betting on the absence of a precedent that had not yet been set. He lost the bet, and the precedent was set on his body: overthrown the following year, extradited in 2001, dead in a cell in 2006.

Which is roughly when the price of every future exit went up, permanently, for everyone. That is the part policy keeps refusing to metabolize. Exits are not renewable if you keep defaulting on them. Qaddafi surrendered his weapons program in 2003 in exchange for rehabilitation, was rehabilitated, and was pulled from a drainage culvert and killed in 2011. Every regime that watched that sequence updated. The credibility of an off-ramp is not a property of the offer; it is a property of the offerer's record, and that record is now a matter of public documentation, available to precisely the people you would least like to have consulted it.

This is what the conflict-resolution literature has been saying since the 1980s and what coercion policy keeps not hearing. I. William Zartman's account of ripeness requires two conditions, not one: a mutually hurting stalemate and a perceived way out. Everyone remembers the hurting. The way out gets treated as a detail to be worked out later, which is like calling the load-bearing wall a decorating question.

iii · what actually ends these

Robert Pape's survey of strategic air campaigns found that punishment aimed at a population almost never coerces a government; what works, when anything does, is denial — making the military objective physically unattainable, which changes the calculation of the people holding guns rather than the ones holding office. Thomas Schelling had already named the deeper problem in 1966: coercion is a bargain, and a bargain requires the threatener to be able to promise restraint as credibly as harm. A threat with no accompanying assurance isn't coercion. It's just damage. The sanctions data tells the same story from another direction — the most-cited tally puts full success at about a third, the scholars who re-scored those same cases put it far lower, and on either count the subset where the demand touched regime survival performs worse than the average.

There is one more break in the frame, and honesty requires putting it before the recommendations rather than after. Every case above is a state that could in principle be militarily defeated outright — Paraguay, the Confederacy, Japan, Serbia, Iraq, Libya. Pape's dataset is conventional air campaigns against non-nuclear states. Russia is not in that category, and the difference is one of kind, not degree.

A nuclear arsenal is itself a permanent, self-supplied exit. It guarantees the regime's physical survival independent of anything anyone offers or withholds, which does two things at once: it removes denial as an option in the strong sense, since an objective cannot be made unreachable in a state that cannot be invaded, and it devalues every off-ramp anyone constructs, because a guarantee you hold yourself dominates a guarantee extended by a party with Qaddafi on its record. To recommend denial against a nuclear power on evidence drawn entirely from non-nuclear ones would be the same category-borrow this piece opened by indicting. So the limit gets stated plainly: the historical record establishes what ends wars against states that can be beaten. It does not establish what ends wars against states that cannot. The sample is thin, recent, and nobody has run it to conclusion — which is the strongest available argument for not finding out.

Inside that limit, the realistic menu has three items.

Build a real exit and eat the injustice of it — which means someone who deserves a cell gets a dacha instead, and the war stops years earlier than it otherwise would. This is morally unbearable and periodically correct, and the fact that it is unbearable is why it is so rarely chosen and so often, in retrospect, the option that would have saved the most lives. Its precondition is not generosity but credibility, and credibility is a stock that has been drawn down for twenty years.

Or pursue denial rather than punishment — make the objective unreachable and let that fact propagate to the people who would have to keep reaching. Slower, more expensive, less satisfying, and the best hit rate in the set against opponents who can be defeated. Against one who cannot, it degrades from a theory of victory into a theory of exhaustion, which is item three wearing better clothes.

Or accept that you are not negotiating at all — you are waiting for the internal structure to fail, on its own schedule, at a cost paid mostly by people who never had a vote in any of it. Which is a legitimate strategy. It's just not the one anybody says out loud, because saying it out loud requires admitting the timeline is measured in years and the currency is other people's children.

The regime, meanwhile, is not behaving irrationally. It is behaving with perfect internal coherence — every decision follows cleanly from the premise that the structure must persist, and the structure holds together beautifully right up until it doesn't. A system can be flawlessly self-consistent and still monstrous, and it stays self-consistent exactly by refusing to count anyone outside itself.

Which is the point at which the instrument should be turned around, because the mechanism has been running on both sides of the table this whole time. The two-ledger structure was never a description of authoritarian regimes specifically. It is a description of decision-makers behind a firewall, and Western ones have their own. The cost of offering an exit lands on the politician who offers it — the accusation of rewarding aggression, the coalition partner who walks, the opponent with the clip. The cost of not offering one lands on Ukrainians. Two ledgers, two sets of people, and the deciders are not on the one that gets debited.

So coercion doctrine persists for the same structural reason the war does: it is internally coherent for the people who hold it. Every step follows from the premise, the premise is never audited, and the arithmetic balances perfectly as long as the people it lands on go uncounted. The difference between the two systems is real and it is moral rather than structural — one of them chose to start a war of conquest, and nothing here narrows that gap. But the machinery is the same machinery. Diagnosing someone else's is easy enough that it gets published weekly. Recognizing your own is the part that almost never happens, because the ledger that would reveal it is the one nobody in the room is on.

Prediction, since that's the job: the next round of commentary will again describe a weakening adversary's refusal to negotiate as inexplicable, propose more pressure as the remedy, and arrive at the same result. It will do this because "build them an exit" is politically unsayable and "we are waiting for a collapse we cannot schedule" is worse. The doctrine survives not because it works but because the alternatives are unspeakable in public — which is a fact about the speakers, not about the war.