Paying for Rivalry
We cost the US–China contest against gross output. The number that actually binds is the discretionary fiscal capacity left after pensions, interest, and the standing state — and demography i...
We cost the US–China contest against gross output. The number that actually binds is the discretionary fiscal capacity left after pensions, interest, and the standing state — and demography is shrinking it. What Rome did to the Seleukids shows the mechanism in its purest form.
TAGS / KEYWORDS: great-power competition · fiscal capacity · US–China · weaponised interdependence · demographic decline · public finance · geoeconomics · political economy · economic statecraft · Rome and the Seleukids
There is a habit, almost universal in the way we talk about the contest between the United States and China, of measuring its cost against the size of the economy. American defence spending runs at roughly three per cent of GDP; the industrial subsidies of the chip wars are a rounding error against a nearly thirty-trillion-dollar economy; even a major shock, we are told, is digestible because the economy is so large. On this accounting the rivalry looks survivable, even comfortable, indefinitely.
I want to argue that this is the wrong arithmetic, and that getting the arithmetic right changes the strategic picture considerably. The mistake is in the denominator. You do not pay for a new carrier, or a subsidy to a domestic foundry, or the standing cost of policing an export-control regime, out of the whole economy. You pay for it out of what is left once the bills you cannot not pay have been paid. And what is left — the discretionary, or free, fiscal capacity of the state — is a far smaller and far more fragile quantity than gross output, and it is being squeezed from several directions at once.
The number that actually binds
A state’s fiscal power is not one quantity but several, and they are easily confused. There is its extractive capacity — the total claim it could in principle place on its economy if it taxed to the limit of political tolerance. This is large, and it is what people implicitly invoke when they say a country “can afford” a contest. But extractive capacity is a reserve, not a budget. Against it stand the commitments already spoken for: the pensions and health entitlements owed to an aging population, the interest owed on the accumulated debt, and the baseline cost of simply running the state and its standing military. What remains after these — the residual a government can actually redirect, in a given year, toward something new — is its free or discretionary fiscal capacity. That residual, not the economy and not even the full tax base, is what new strategic commitments are paid from.
The distinction matters for a reason easy to state and hard to overstate. Mandatory commitments and debt service are, by their nature, the least negotiable lines in any budget; they have first claim. Discretionary spending is what is left to fight over, and the cost of great-power rivalry — incremental military outlay above the existing establishment, the subsidies of supply-chain security, the administration of sanctions and export controls — competes for that residual with everything else a government might want to do. To measure such a cost against gross output is to measure a marginal, discretionary claim against a gross, mostly-committed quantity. It flatters the affordability of the rivalry by an order of magnitude.
There is a second, subtler reason the residual is the right denominator. Because it is the small difference between two large numbers — revenue on one side, committed expenditure on the other — it is far more volatile than either. When output falls in a downturn, revenue falls with it, but the mandatory commitments do not; they often rise. The residual is therefore squeezed disproportionately: a shock that takes a few per cent off output can take a much larger fraction off the slack. The free capacity is geared to the business cycle by a multiple well above one. A burden that looks modest as a share of GDP is not modest as a share of the room actually available to bear it — and it is most binding precisely when a crisis would make a state most want to act.
America is already spending from a margin at or below zero
Put real numbers to the American case and the picture is stark. In its 2026 budget outlook the Congressional Budget Office projects federal revenue at about 17.5 per cent of GDP, mandatory outlays — chiefly Social Security, Medicare, and Medicaid — at about 14.2 per cent, and net interest on the debt at about 3.3 per cent. Set those side by side and the result is arithmetic, not opinion: revenue, less the mandatory commitments, less the interest, comes to roughly zero. Before a single dollar of discretionary spending — defence, infrastructure, research, the whole apparatus of a great power — the committed claims have absorbed essentially the entire revenue of the federal government.
Now subtract the baseline military establishment, on the order of three per cent of GDP, and the residual against which any new strategic commitment must be measured is not merely thin; it is firmly negative. The marginal cost of the rivalry is not being paid from a current surplus. It is being shifted into debt and into future revenue claims — financed at the margin by deficit expansion, by higher taxes later, by inflation, or by the displacement of other discretionary spending.
This is not a claim that the United States is bankrupt, or that it lacks strategic capacity. It plainly retains enormous extractive capacity: it can raise taxes, reprioritise, and draw on the deepest capital markets and most flexible private economy in the world. The point is narrower and more uncomfortable. On the federal accounts, the discretionary room out of which competition is supposed to be funded is at or below zero, and the gap is being closed by borrowing. The CBO’s own projections show the deficit at around 5.8 per cent of GDP, debt held by the public rising from roughly 101 per cent today toward 120 per cent within a decade, and net interest climbing past the cost of defence and then past all non-defence discretionary spending combined. Borrowing extends the constraint; it does not abolish it. The capacity to borrow is itself bounded by the rising-interest dynamic the debt path sets in motion — flexibility that is real, but finite, and shrinking.
China: the same logic, harder to see
China presents the same structural problem in a less legible form. The symmetric calculation cannot honestly be performed, because so much of the relevant expenditure and contingent liability sits outside the headline general-government accounts: large off-budget funds, the quasi-fiscal activity of state-owned enterprises, and above all the borrowing of local-government financing vehicles. Read the official books alone and China’s fiscal position looks almost conservative.
The International Monetary Fund’s most recent Article IV assessment is where the hidden claims are brought partly into view. On a narrow official reading, general-government debt is around 61 per cent of GDP. On the Fund’s augmented reading — which expands the perimeter to include the off-budget funds and the financing vehicles — it is around 117 per cent, and the augmented measure of net government borrowing runs at something like minus 13 per cent of GDP. The gap between the two figures, some fifty-six points of GDP, is not a separately measured stock of hidden debt to be quoted as a fact; it is the size of the perimeter adjustment — the measure of how much is happening off the visible books.
The honest claim about China is therefore not a number but a direction. No defensible point estimate of its discretionary residual is available on public data. But the augmented accounts imply, unambiguously, that the headline fiscal position overstates its true discretionary capacity: the largest non-discretionary claims are precisely the ones that sit outside the official perimeter, so the real free capacity is smaller — and its compression by an aging population sharper — than the published figures suggest. The disciplined response is neither to leave the question blank nor to fill it with a spurious figure, but to bracket it by accounting perimeter and to name the unbooked claims that make the residual smaller than the headline accounts imply.
One further asymmetry is worth stating, because it changes how the constraint bites. In a market-financed system like the American one, borrowing capacity fails through the bond market — through rising yields and, at the limit, a buyers’ strike. In a state-directed system it fails differently and more quietly: through the deterioration of bank balance sheets, the refinancing stress of local governments, the pressure on the capital account, and the financial repression of households whose savings are conscripted to fund the state. The wall is no less real for being reached by a different road.
Why the slack gets spent anyway
If the discretionary residual is so thin, why do states keep spending it down on rivalry, rather than husbanding it? The answer is not irrationality but political economy, and it is the part of the story that turns an accounting constraint into a binding one.
The benefits of rivalry spending are concentrated and the costs are diffuse. A defence programme, a subsidised foundry, a protected strategic industry delivers visible gains to identifiable beneficiaries who organise to defend them; the cost is spread thinly across millions of taxpayers, none of whom has much incentive to fight it. Debt makes the asymmetry worse, because it shifts the cost onto future taxpayers who do not vote in the present bargain at all. And the spending that competes with rivalry for the residual — the entitlements owed to the old, the interest owed to creditors — happens to be the most politically protected and least cuttable spending a modern state carries. The result is a standing bias toward drawing the residual down rather than defending it.
Autocracies are not exempt from this logic; they merely run a different version of it. A regime whose legitimacy rests on growth and on visible national strength has its own powerful reasons to keep spending on the symbols and instruments of rivalry even as the fiscal room contracts — and the state-directed financial system gives it the means to hide the bill for longer, in the balance sheets of banks and the obligations of state enterprises, than an open system could. The capacity is drawn down all the same; it is simply harder to see it happening.
Weaponised interdependence: making the other side pay
This is where the contest acquires its characteristic modern weapon. Henry Farrell and Abraham Newman gave the phenomenon its name — weaponised interdependence: the use of chokepoints in the networks that bind the world economy, the dominance of a few firms in advanced semiconductors, the concentration of critical-mineral processing, the centrality of the dollar in global payments, as instruments of coercion. A state that sits astride such a chokepoint can deny a rival access, or surveil the flows that pass through it.
Seen through the fiscal lens, the strategy has a sharper logic than mere denial. The point of weaponising interdependence is not only to withhold a good from the rival; it is to force the rival to consume its discretionary capacity on defensive reconfiguration. Confronted with a chokepoint, the targeted state must build alternative supply chains, stockpile against disruption, subsidise domestic substitutes, stand up new payment channels, and pay the standing cost of evading or enforcing restriction. Every one of these is a claim on free capacity. The deep purpose of the manoeuvre is to make the other side spend down its residual — to raise the rival’s numerator while protecting one’s own.
The weapon has a self-erosion problem, and it is instructive. The more aggressively the dominant power uses its chokepoint — the dollar-clearing system, say — the stronger the rival’s incentive to build an alternative, and the faster the very leverage erodes. Coercion through a network tends to degrade the network’s value, which is why the most potent instruments are also the ones most dangerous to wield routinely. The same dynamic explains why third-party theatres — the sanctioned economies of Russia and Iran, for instance — matter to the central contest without being equal to it: they are the proving grounds and the workshops where techniques of evasion and alternative settlement are developed, lowering the value of the incumbent infrastructure for everyone. They are channels through which the cost of network power is revealed and contested, not a third pole in the rivalry.
It is worth noting how unsettled the American instrument has lately become. In February 2026 the Supreme Court ruled that the emergency-powers statute could not be used to impose the across-the-board tariffs, forcing the administration onto other authorities while the measures levied since 2018 stayed in force. The point for the present argument is not the legal detail but its fiscal signature: a tariff is, among other things, a tax, and the 2025 campaign amounted to the largest American tax increase as a share of GDP since the early 1990s. The cost of contesting the relationship shows up at home, on the coercer’s own ledger, even when it is aimed abroad.
The compressor: the denominator is shrinking
So far the residual has been treated as a fixed, if small, quantity. It is not. It is contracting in both principals, for a reason that has nothing to do with the rivalry itself: population aging.
The relationship between aging and growth is now estimated with some precision. Using variation in the predetermined component of aging across American states, economists have found that a ten per cent increase in the share of the population aged sixty and over reduced the growth of per-capita output by around five and a half per cent — shaving something like a third of a percentage point off annual growth, with roughly a third of the effect running through slower employment and two-thirds through slower productivity. That estimate must be handled with care: it is identified across US states, where mobility and federal transfers net out part of what would not net out across nations, and applying it to China is an extrapolation rather than a transplant. Taken for what it is — a bounded magnitude for the size of the mechanism, and a directional claim about the comparison — it is enough.
What matters for fiscal capacity is that aging compresses the residual twice over. It lowers the trend growth of the revenue base, through both the employment and the productivity channels; and, because the fixed claims of pensions and health care rise exactly as the working-age share falls, it enlarges the mandatory commitments that must be honoured before any residual is free. The slack is squeezed from both ends at once — the revenue end growing more slowly, the commitment end growing faster — and, by the gearing argument above, the proportional effect on the residual is larger than the proportional effect on output. Aging is, in this precise sense, a structural compressor of the very fiscal room out of which rivalry is paid.
The two principals are on different timetables, and the asymmetry is strategically decisive. By the static measures — the share over sixty-five, today’s dependency ratio — the United States is actually the older society. China’s disadvantage is dynamic. Its old-age dependency ratio — the number of people over sixty-five for every hundred of working age — stands at around twenty-one today against the American thirty; but it more than doubles, to roughly fifty-two, by 2050, overtaking the American figure, which rises only to about thirty-seven, near mid-century. China’s working-age population peaked in 2015 and is set to fall by more than a hundred million by 2050, with no migration to cushion it; it confronts the prospect of growing old before it has grown rich. The United States ages later and more slowly, and possesses a partial offset China lacks — net immigration — but that offset is a contested policy choice, not a demographic constant. A rivalry comfortably affordable against today’s free capacity may be a great deal less so against the compressed capacity of the 2030s, and the side whose capacity contracts faster meets the tighter constraint first.
There is one genuine escape worth naming, and naming its limit. A sufficiently large acceleration of productivity — from automation, or from artificial intelligence — could offset the growth drag of aging and loosen the constraint. But even a full productivity offset would address only one of the two channels. It would raise the growth of the revenue base; it would do nothing to reduce the pension and health claims of an older population, which rise with the dependency ratio whether or not output per worker grows faster. The compressor therefore has a floor that no productivity miracle can remove: the expenditure channel of the squeeze survives even the optimistic case.
What Rome did to the Seleukids
The mechanism I have been describing — capacity borne by surplus rather than by gross, and the base that regenerates the surplus shrunk at the same time — is not new. It appears in its starkest premodern form in the settlement Rome imposed on the Seleukid empire after defeating Antiochus the Great at Magnesia in 190 BCE. The terms, fixed at Apameia in 188, are worth recalling, because they are a controlled experiment in fiscal coercion stripped of all modern complication.
Antiochus was made to pay an indemnity of fifteen thousand talents of silver — an immense sum, payable not at once but in annual instalments stretched over years. He was made to surrender the whole of Asia Minor north and west of the Taurus mountains, among the richest tax-yielding territory in his realm. His navy was cut to a token squadron, his war elephants given up, and twenty hostages taken to Rome — among them his younger son, the future Antiochus IV. Rome did not annex the Seleukid state, garrison it, or govern it. It did something cheaper and, in the long run, more corrosive: it bound the empire to a recurring payment that consumed its surplus, while stripping away the provinces whose revenues had generated that surplus in the first place.
The two blows compounded. The indemnity was a standing claim on discretionary capacity, payable year after year; the territorial loss shrank the base from which any such capacity could be regenerated. Capacity drained, and the engine of regeneration cut, together. The consequence was not sudden collapse but a slow fiscal hollowing. Antiochus himself died the year after the treaty, in 187, killed by locals while stripping the treasury of a temple in the east — a king reduced to robbing shrines to make his payments, which is as exact an image of a fiscal death as the ancient record provides. The dynastic instability that followed, and the empire’s long decline, cannot be reduced to Apameia alone; but the settlement set the fiscal terms within which that decline unfolded.
I am not predicting that the contest between the United States and China will end as the Seleukid empire did. The homology is structural, not prophetic, and the differences are enormous: fiat money and central banks that the ancient world had no equivalent of; an interdependence so deep that each side’s prosperity is hostage to the other’s; nuclear weapons that cap the stakes; and a standing diplomacy that offers exits a Hellenistic king never had. What the ancient case offers is not a forecast but clarity. It shows the mechanism in its skeleton — a great power drained not by conquest but by a recurring claim on its surplus and a shrinking of its base — free of the modern noise that ordinarily hides it. The logic is the same; the setting, the symmetry, and the available exits are not.
A contest over whose margin runs out first
Assemble the pieces and the shape of the argument is plain. The sustainable intensity of the rivalry is governed by a ratio: the recurring direct fiscal cost of the competition, set against the free fiscal capacity available to bear it. Demography is shrinking the denominator of that ratio in both principals, earlier and faster in China. Weaponised interdependence is the means by which each side tries to raise the numerator of the other’s ratio — to force the rival to spend more of its slack on defensive reconfiguration — while protecting its own. The broader strategic-economic costs of the contest — the deadweight of tariffs, the private cost of rewiring supply chains, the welfare losses of fragmentation — do not all fall directly on the public budget, but they narrow the residual indirectly, by eroding the tax base, raising the subsidies the state must fund, and shifting the political incidence of the burden.
The competition, in its fiscal skeleton, is therefore a contest over whose discretionary capacity is exhausted first — conducted against a background in which both sides’ capacity is structurally contracting. This reframes the strategic question. It is no longer only “what does the competition cost?” but “against a shrinking denominator, how long can the cost be borne, and by whom first?” That is a question the gross-output accounting cannot even pose, because it holds fixed the very quantity that is moving.
What this argument does not claim
It is worth being explicit about the limits of the case, because a structural argument is only as honest as the conditions it states for its own qualification.
The claim is a framework and a homology, not a forecast. It does not assign probabilities to outcomes or predict a date at which either state’s capacity gives out. The central quantity — free fiscal capacity — is only partially observed, especially for China, and the intellectually honest treatment is to model it as a latent variable inferred from its causes and indicators, not to fake a point estimate the data cannot support. Several developments would genuinely qualify the argument, and they deserve naming. A sustained productivity surge large enough to outrun the expenditure channel of the demographic squeeze would loosen the constraint — though, as I have argued, it cannot remove the floor. A political reordering that made entitlements cuttable would enlarge the residual, at a political cost few governments will pay. A migration regime that offset aging would slow the American compression, if the politics allowed it. And a genuine de-escalation of the rivalry would remove the recurring claim altogether, which is, after all, the cheapest fiscal policy available to either side.
What the argument insists on is direction and structure, not destiny. It says that the cost of the rivalry should be measured against the awkwardly small quantity of discretionary fiscal capacity rather than against gross output; that this quantity is being compressed by forces neither side controls; and that the strategy of weaponised interdependence is best understood as an attempt to accelerate the compression on the other side. None of that requires knowing how the contest ends. It requires only looking at the right number.
Conclusion
The deepest error in how we reason about great-power competition is a category mistake about money: treating a marginal, discretionary cost as though it were paid from the whole economy, when it is paid from the thin and shrinking margin the economy leaves over. Correct the mistake and the contest looks different — tighter, more time-bound, more clearly a race against an internal clock than a stable equilibrium that can be held indefinitely.
Rome understood the principle without the vocabulary. It did not defeat the Seleukid empire by destroying it. It bound the empire to a payment that consumed its surplus and stripped away the territory that surplus came from, and let the arithmetic do the rest. The modern contest is incomparably more complex, and its weapons are subtler; but at its fiscal core it may turn less on who commands the larger economy than on who can keep paying from a contracting margin — and on who can force the other to pay more from theirs. That is the arithmetic to watch. The denominator is moving, and it is moving against everyone.
A note on sources. This essay condenses a longer, formal working paper. The American fiscal figures are drawn from the Congressional Budget Office’s 2026 Budget and Economic Outlook; the Chinese figures from the International Monetary Fund’s 2025 Article IV consultation, which sets out the augmented-accounting perimeter for off-budget and local-government liabilities. The estimate of aging’s effect on growth is from Maestas, Mullen, and Powell’s study of US states; the demographic projections are from the United Nations’ World Population Prospects 2024, with cross-checks against the World Bank. The concept of weaponised interdependence is Henry Farrell and Abraham Newman’s. The Roman–Seleukid settlement is recorded in Polybius and Livy. The framework’s stock-and-flow distinctions, its treatment of free capacity as a latent variable, and the demographic and weaponised-interdependence arguments are developed at length, with full citation, in the underlying paper.