The War That Pays

George W. Bush, October 2000: a humbler foreign policy, no nation-building. Barack Obama, 2008: elected on the promise of ending Iraq. Donald Trump, 2016: the endless wars, ended. Different parties, different ideologies, different temperaments, three separate mandates from three separate electorates who all wanted the same thing.

Now the honest accounting, because the lazy version of this argument is refutable in one sentence. The output did change. Bush ran two ground invasions and a decade of occupation. Obama ended the Iraq occupation, returned to Iraq in 2014, left under ten thousand troops in Afghanistan, and in his final full year had American forces conducting strikes in seven countries. Trump ordered one decapitation strike and signed a withdrawal agreement. Those are not the same magnitudes and nobody should pretend they are.

What did not change is that American forces were in combat every year of all three administrations. The mandate was not ignored. It was metabolized. Each president delivered a version of war the electorate had asked to be rid of, in a form the electorate could stop noticing — smaller, remoter, quieter, continuous. The frequency held. What moved was the visibility.

When you vary the input three times and the output shifts only along the axis nobody is watching, you are no longer studying the input. You are studying the machine the input goes into.

Emma Ashford makes this argument in Foreign Affairs — that American policymakers keep getting pulled into military action by an incentive structure that makes force attractive and its short-term costs invisible, and that the fix is to make force harder to reach for. She is right, and the framing deserves to be pushed further than the essay takes it, because the incentive structure is not weather. It was built. Piece by piece, each piece installed for a defensible reason, mostly by people who thought they were reducing the odds of war.

i · the payoff arrives early

Political science named this in 1973, when John Mueller described the rally-round-the-flag effect, and the numbers have been embarrassingly stable ever since.

George H. W. Bush entered the Gulf War around 59 percent approval and peaked at 89 — the highest Gallup had ever recorded. Eighteen months later he was in the thirties and out of a job. His son ran the same curve at higher amplitude: roughly 51 percent before September 11, 90 percent after, a new record, then a slow bleed to 25 percent by the end. Jimmy Carter's approval jumped roughly thirty points in the weeks after the Tehran embassy was seized — for a crisis he did not cause and could not solve, which then destroyed him.

Notice what those three cases have in common. Every one is a crisis that arrived at the president rather than one he went out and manufactured. That matters, because the tempting conclusion — presidents start wars to harvest the approval — is a claim the diversionary-war literature has never settled, and Mueller himself hedges it. Borrowing his authority for something he declines to assert would be exactly the move this piece exists to catch in other people.

The narrower claim needs no hedging and does more work anyway. The rally effect does not manufacture an appetite for force. It raises the price of restraint. In a crisis, the option carrying the visible near-term political cost is doing nothing, and it is the only option on the menu whose bill arrives before the next election. The machine does not reward aggression. It punishes stillness, which is sufficient.

The spike is real. The decay is real. Neither is the point.

The point is the asymmetry between the two clocks. The approval bump arrives within days and clears within a year. The war arrives within days and clears within a generation. A president is graded on a four-year cycle for a decision that finishes settling in forty. Any actor evaluated on the fast clock while making decisions that resolve on the slow one will systematically over-purchase the thing that pays fast.

This is not an accusation about character. Run the honest thought experiment: put a genuine pacifist in the chair, hand them a crisis, a hostile press cycle, an intelligence briefing with a 30 percent probability attached to something terrible, and a menu where one option costs them nothing this quarter. The pacifist does not need to be corrupted. They need only be responsive to the environment they are standing in — which is what we selected them for.

ii · the frictions, removed

Every constraint that once made American war politically expensive has been dismantled. Each removal had a real argument behind it. That is what makes the sequence worth laying out in order.

The draft. Conscription ended in January 1973. Ending it was a genuine expansion of individual liberty and almost nobody wants it back. It also severed war from the median American household. The mass antiwar movement — the one that drove a sitting president out of a renomination fight — dissolved within about two years of the draft's end and has never reassembled at that scale. Not because Americans became less moral in 1975. Because the cost stopped landing on their kitchen tables.

The tax. World War II was financed with war bonds and a top marginal rate of 94 percent. Korea got tax increases. Vietnam got the 10 percent surcharge of 1968 — a war whose price briefly appeared on a pay stub, in the same year public support for it broke. Iraq and Afghanistan were the first major American wars prosecuted alongside tax cuts. The cost did not disappear; it moved from the present tense into the bond market. Debt does not attend town halls. Debt does not vote.

The body bag. Standoff weapons and armed drones removed the last cost that reliably reached the public: American dead, arriving home in daylight. And note who presided over that expansion. The president elected specifically to stop the wars ran the largest targeted-killing program in American history. Hold the personality constant against the machine and watch which one bends. Obama is not the exception that disproves the structural argument. He is the controlled experiment that confirms it.

The vote. The United States last formally declared war in June 1942, against Bulgaria, Hungary, and Romania. Everything since has run on authorizations or bare assertions of Article II. The War Powers Resolution, passed over Nixon's veto in 1973 and explicitly designed as friction, has never once forced the termination of an operation. Read it the way the executive branch actually reads it and it is not a leash — it is a permission slip granting sixty free days, no vote required.

Then there is the 2001 Authorization for Use of Military Force. Passed 420 to 1 in the House, Barbara Lee alone against, in the week after September 11. It has since been invoked for operations in something like twenty countries, against organizations that did not exist when it was written. It has no expiration date.

That last detail is load-bearing, and it deserves to be said plainly: the default state of an American war is "still happening." Inertia runs toward continuation. Stopping requires an affirmative act by people who gain nothing from performing it.

Now line those four up, because they are not four things. Ending the draft did not reduce the number of people who die in American wars. The tax cuts did not reduce the cost; they relocated it into the bond market. Standoff weapons did not reduce the killing, only the fraction of it performed at close range by Americans. Skipping the declaration did not reduce the wars, only the number of names attached to them. Not one of those removals subtracted a dollar or a life from the bill. Every one of them removed the signal — the channel by which a cost reaches somebody positioned to object. The price is unchanged and fully paid. It simply no longer transmits.

Which identifies the fifth friction, the modern one, and the largest.

The record. The executive branch classifies its own body count. Civilian casualty estimates, strike tallies, the legal reasoning authorizing the strikes — the entity conducting the operations decides what the public is permitted to know about the operations, and reviews its own conduct afterward. Independent tallies compiled by journalists and monitoring organizations have run persistently above the official ones, and there is no auditor with subpoena power to reconcile them. This is not an accountability gap in an otherwise sound structure. It is the structure. A system that coheres by controlling what the affected are allowed to know has made the information asymmetry into a load-bearing wall.

iii · the ledger nobody signs

Brown University's Costs of War project put the post-9/11 total above eight trillion dollars, of which roughly 2.2 trillion is future veterans' care running out to 2050. Direct deaths: somewhere around 900,000.

Sit with the composition of that number. The single largest deferred line item is medical care and disability compensation for people who have not yet filed the claims. Linda Bilmes's work on veterans' costs makes the timing explicit — peak spending on a war's veterans arrives decades after the shooting stops, long after the administration that ordered it has been archived, eulogized, and had its library built.

There is no accounting entry anywhere in the federal government that debits a decision-maker. Not one. The cost is real, enormous, and perfectly detached from the person who incurred it. If you designed a procurement system that worked this way — unlimited authority, no budget line, no audit, payment due in forty years to a successor's successor — you would be describing fraud. Applied to war, it is called the separation of powers.

iv · where the circle ends

The restraint argument is almost always made in American costs. Debt. Readiness. Opportunity cost, the money that could have gone to infrastructure or veterans or anything. Ashford's framing does this too, and it is the strongest available lever for changing American behavior, which is precisely why everyone reaches for it.

But look at what a purely American ledger implies. It makes intervention wrong when it is expensive to us. Get the price down — cheaper drones, fewer American personnel, a proxy force with someone else's casualty list — and the objection evaporates. An accounting that only counts one side does not produce restraint. It produces optimization.

On August 29, 2021, in Kabul, a Hellfire missile killed Zemari Ahmadi and nine members of his family, seven of them children. The chairman of the Joint Chiefs called it a righteous strike. Weeks later the Pentagon acknowledged it had killed an aid worker loading water containers into a car. No one was disciplined. Total institutional cost of ten dead civilians: one press conference and a paragraph of regret.

That is not a scandal. A scandal implies a malfunction. This was the system charging the accurate price — which is what it had been built to charge.

Coherence purchased by excluding the people it lands on is not coherence. The people who bear the overwhelming majority of the cost of American force have no input into a single mechanism named in this piece: no vote in the authorization, no line in the budget, no representation in the polling that produces the rally effect. And the exclusion runs one layer deeper than the ledger, which is the part usually missed. They are left out of the data before they are left out of the accounting, because a cost that no independent body is permitted to count cannot be priced by anyone at all. Widen the circle or the redesign just relocates the damage somewhere it cannot be heard from.

v · redesigning the room

The antiwar movement has spent sixty years appealing to conscience, and it has been outperformed by a tax code. Conscience is not the variable. The cost function is.

Count them, independently. Quarterly, mandatory, statutory: dollars, American casualties, and foreign civilian casualties, compiled by a body that does not report to the people conducting the operations and cannot be classified out of existence by them. This belongs first, not last. Every other item on this list is arithmetic performed on numbers, and right now the operator supplies its own numbers.

Put a floor under the dead. A statutory compensation obligation triggered by any acknowledged civilian death, with a real floor rather than the discretionary condolence payments currently available, drawn from the conducting command's operating account and not from a supplemental appropriation. This is the one mechanism that puts a foreign life on the same ledger as a munition and makes them compete. It converts Kabul from one press conference into a line item.

Invert the default. Every use-of-force authorization expires in twelve months and requires an affirmative, recorded re-authorization vote to continue. Today, ending a war takes an act of Congress and continuing one takes silence. Flip it. Make silence end things and make continuation cost a roll-call vote in an election year.

Price it in the present tense. An automatic, un-waivable surtax triggered by sustained hostilities. Not primarily to raise revenue — to make the war a line on a pay stub while it is being fought, rather than a line in a debt table after it is over.

Sunset the money with the authority. Appropriations expire when the authorization does. No war outliving its permission slip on residual funds.

Be clear about what the middle three are, because this essay has just spent a section indicting the currency they are denominated in. A surtax, a roll-call vote, and an appropriations sunset are all American costs, and American costs alone do not produce restraint — they produce substitution. Satisfy that cost function cheaply enough and the answer is somebody else's infantry at the end of somebody else's runway. That is why the first two items are not decoration. Counting the dead and paying for them are the only mechanisms on the list that get more expensive as the war moves further from the payer. Without them, the rest is a speed bump that teaches the driver a better route.

None of this requires anyone to become a better person. That is the entire design principle. And none of it will pass, because it would have to be enacted by the people currently enjoying an arrangement in which they receive full credit for supporting the troops and never have to put their name on a war.

vi · deviation, marked

The machine is not destiny, and the honest way to prove it is to point at the times it jammed.

January 1991. The Senate authorized the Gulf War 52 to 47. Five votes. That is what a real authorization looks like — a vote a senator has to defend for the rest of their career, taken in public, in advance, on a war that had not started yet.

August 2013, and this one is cleaner abroad than at home. On the 29th, the British Parliament voted down participation in strikes on Syria — the first time in more than two centuries a prime minister lost a war vote. A legislature was asked to own a decision on the record and it did, and the answer was no. Mechanism, functioning.

What happened in Washington was not that. Obama referred the question to Congress, and Congress quietly declined to schedule it. The strike died of neglect. Say that plainly rather than filing it under legislative accountability, because avoidance is a different instrument from refusal: a body that will not vote is not a body that votes no, and only one of those is what the reforms above prescribe. The uncomfortable finding is that friction of almost any kind — including cowardice — can jam the machine. That is weaker evidence for roll-call votes than it looks like, and stronger evidence for the general proposition that this thing stops when something physically obstructs it.

The clean examples are old. Both involved the same instrument: someone with a name and a district forced to say yes on the record before anything was launched. That instrument still exists. It has simply fallen out of use, in the way that things fall out of use when nobody with power finds them convenient.

The pattern here is not that Americans love war. It is that nobody in the American decision chain has ever personally been billed for one, and nobody outside it has ever been allowed to send the invoice. Build a room where both are possible, and the personalities can stay exactly as flawed as they already are.