Seventy-Five Years of Almost Leaving

Every clause of that is defensible. Every clause has been defensible since 1951. That is the finding, and it is not a compliment.

Dwight Eisenhower took command in 1951 as the first Supreme Allied Commander Europe and treated the deployment as a splint, not a skeleton. He said — in testimony, in letters, in remarks he kept repeating for the next fifteen years — that if American troops were still stationed in Europe for European defense a decade later, the project would have failed by definition. The decade elapsed. Eisenhower was by then President. He did not withdraw them.

Mike Mansfield spent nine years, 1966 through 1975, introducing measures to cut the American presence in Europe roughly in half. He was Senate Majority Leader. He had the argument, the seniority, and the floor, and he brought it back annually the way you re-file a bill you expect to eventually pass. The 1971 version got a real vote and lost. Richard Nixon announced a doctrine in 1969 whose premise was that allies would supply their own manpower; the doctrine was applied in Asia and quietly declined to travel to the Rhine.

The pattern is not that the United States stays in Europe. The pattern is that the argument for leaving is itself a permanent institution — older than most of the bases, more durable than any of the administrations that have hosted it, and structurally incapable of producing the outcome it proposes.

The Argument Is Load-Bearing

Count the iterations. Eisenhower's ten-year clock, 1951. The Nixon Doctrine, 1969. Mansfield, annually, for the better part of a decade. The post–Cold War drawdown, which was real: roughly 315,000 American personnel in Europe in 1989, under 100,000 by the late nineties, around 65,000 by the early 2010s. Donald Rumsfeld's Global Posture Review in 2004, which closed installations across Germany on exactly this logic. The 2012 pivot to Asia, which was the same essay with a new destination. Trump's July 2020 order to pull 12,000 troops out of Germany, reversed by the Biden administration in February 2021. And now, this week, the case again, in the journal of record for American foreign-policy consensus.

Eight recognizable iterations across seventy-five years. The troop level today sits above where it stood in 2013.

The usual reading of this is failure — good idea, bad politics, blocked by inertia and lobbying and the Atlanticist priesthood. That reading requires believing that seventy-five years of consistent non-adoption is an accident that keeps recurring.

There is a more interesting reading, and I want to be honest about how much weight it can hold. It is that the essay is not failing. It is functioning.

A commitment that cannot be withdrawn produces no leverage. A commitment that is withdrawn also produces no leverage, because a spent threat is worth nothing the morning after. On that logic, the only configuration that generates continuous influence is a commitment permanently under review — visibly, publicly, in respectable venues, on a cycle frequent enough that no European planning ministry can ever treat the American presence as a fixed input.

The problem with that reading is the evidence for it. The obvious test case is 2020, the one time an administration moved from argument to order: 12,000 came out on paper, and within seven months the order was reversed and the ambiguity was back. That looks like a corrective firing. It is also fully explained by an election — a different administration with opposite priors took office and undid a predecessor's decision, which is a thing that happens constantly and requires no hidden mechanism at all. I cannot point to the corrective firing inside an administration, which is what would separate the two stories. So the leverage reading is available and unproven, and anybody selling it harder than that is selling.

What survives without the mechanism is smaller and, I think, more damaging.

An arrangement that everyone denounces and nobody exits is usually an arrangement that everyone wants. Washington would rather not surrender control over European security policy, and has spent seventy-five years building institutions that ensure it doesn't have to. Europe would rather not pay what autonomy actually costs — not in money, which is now flowing freely, but in the political work of standing up a command structure, a nuclear guarantee, and a foreign policy that two dozen sovereign governments can agree to use in a crisis. Both parties need to be publicly on record wanting the opposite of what they are doing. Washington has to be visibly considering departure, or it is a patron. Europe has to be visibly pursuing autonomy, or it is a protectorate.

The essay is the instrument that lets both of them keep saying it. Not a maintenance schedule that Washington imposes on Europe — a script that both sides have lines in, and that neither has any incentive to stop reading.

This is why the debate never resolves. It is not stuck on the way to a decision. The irresolution is the product.

What Actually Moves the Number

The stated variable is European readiness — capability, spending, political will, the 2 percent line and now the 5 percent line. Run the history against it and the correlation does not appear.

American force levels in Europe fell hardest between 1990 and 1995, when European capability was declining too, because the Soviet Union had dissolved and the American budget was being cut. They fell again after 2003, when the assets were needed in Iraq and Afghanistan. They rose after 2014 and again after 2022, when Russia moved. In none of these cases did the number track anything Europe did. It tracked what Washington needed elsewhere, and what Moscow did.

Which is the general case, and there is a long ledger behind it. Britain occupied Egypt in 1882, announcing a temporary presence contingent on the restoration of order. Order was restored several times. The British left in 1956, seventy-four years later, after Suez broke the capacity to stay. Soviet forces did not leave Eastern Europe when the stated mission concluded; they left when the state that garrisoned them stopped existing. Garrisons end when the garrisoner's position changes, not when the host graduates.

The counter-cases are worth naming, because they mark the actual exit conditions rather than the advertised ones. Austria got all four occupying powers out in 1955, ten years in, by treaty — and the price was permanent neutrality, which is to say Austria bought its way out by removing itself from the board entirely. Nobody in Europe is offering that, and nobody is being asked for it. The Philippine Senate voted the American bases out in 1991, which worked, and then Manila invited the Americans back under a new agreement in 2014 and expanded it in 2023, which tells you what a host-initiated ejection is worth over a thirty-year horizon.

So the honest version of the claim is not that garrisons never end. It is that the level is set by the garrisoner's global constraints and never by the host's condition.

Which means every European conversation about readiness is a conversation about a variable that is not in the equation — and European defense ministries have had seventy-five years and several excellent analytical staffs in which to notice this. They have noticed. The readiness debate continues anyway, because moving the American number was never its function. Its function is to demonstrate effort, which is what a junior partner owes a senior one, and what a domestic electorate needs to see before it will fund anything. That is not a category error the arrangement imposes on Europeans. It is a performance whose audience they have correctly identified and whose price they have agreed to pay, in preference to the alternative.

The Remedy Deepens the Condition

There is a further turn in the current round, and it is the part that should embarrass everyone involved.

Europe is in fact rearming. Budgets are up across the continent, in some cases dramatically. And a large share of that money is buying American systems — F-35s in Germany, Patriots and HIMARS across the eastern flank, Polish procurement at a scale that made Warsaw one of the larger export customers in the American defense industrial base. The prescribed cure for dependency is being administered as a subscription upgrade.

Sovereign capability means being able to fight without permission. A fleet of aircraft whose software, sustainment, and munitions supply run through Fort Worth is a capability held on license. Europe can spend its way to a much larger military and arrive with less strategic autonomy than it started with, because the new equipment ties the logistics tail more tightly to the same supplier. Burden-sharing, executed as procurement, is not the exit from the arrangement. It is the renewal.

And it is being renewed voluntarily, with alternatives on the table. Dassault will sell you a Rafale. Saab will sell you a Gripen. Germany's F-35 order replaces the Tornado in the nuclear-sharing role — a mission that by definition requires an aircraft the Americans will certify to carry their weapons. That is not a procurement decision that got captured by a vendor. It is one where the dependency was the requirement, written into the specification before any competition began. Nobody put a gun to Berlin. Berlin bought the tie deliberately, because the tie is the thing it actually wants.

Which means the essay's remedy, taken seriously and funded generously, produces a Europe that is harder to leave — and that this is a feature to the people funding it.

What Would Count as a Break

A genuine deviation looks like a force reduction that survives a Russian escalation — a cut announced in a calm year and held through a violent one. Every reduction in the record so far has reversed on contact. It also looks like a major European procurement decision that goes to a European platform over a cheaper or better American one, accepting the capability cost to buy the independence.

And a third marker, aimed at the reading I have offered rather than the one I rejected. The bargain holds because the autonomy argument is always addressed to the other party — Europeans make it in Washington, to prove they are worth trusting; Americans make it in Brussels, to prove they might leave. It would break the first time a European head of government made the autonomy case at home, to their own voters, as the reason to spend the money and accept the risk — and won an election on it. Nobody has tried. That is the datum.

Those three things together would mean the structure moved and not just the rhetoric. They would be worth marking, because they would be new.

Absent them, the prediction is boring and I am confident in it: the essay runs again. Somewhere between 2029 and 2031, in this journal or its nearest competitor, under a headline that means leave Europe to the Europeans, citing the 2026 version as evidence that the moment has at last arrived. The troop number will be within twenty thousand of where it is today, in whichever direction Moscow and the Pacific have jointly decided.

Seventy-five years of almost leaving. The names change, the destination for the redeployed assets changes, the acronym for the burden-sharing target changes. The trajectory doesn't, because the trajectory was never the point. Staying was expensive. Leaving would have been cheap. Perpetually considering leaving turned out to be the only one of the three that both parties could sign, and it has now been in continuous production since Truman was in office.