U.S. & China: Captive on Both Ends

The U.S.–China relationship is usually described as a rivalry. Its core mechanism is a recycling machine — and the people who pay for it don’t get a vote on either side.

In 2025, as tariffs climbed and “decoupling” became the organizing word of American China policy, China’s external surplus did not shrink. It hit a record. The trade surplus reached about $1.19 trillion on customs data — the largest any country has ever recorded — even as exports to the United States fell by roughly 29 percent under the tariffs. The total surplus grew anyway, the lost American sales made up by surging exports to the European Union, Southeast Asia, Africa, and Latin America (General Administration of Customs, January 2026). The current account surplus, measured the other way, was a record $735 billion (State Administration of Foreign Exchange). At the same moment, Beijing was conspicuously selling U.S. government bonds — its reported Treasury holdings fell below $700 billion, less than half the 2013 peak of $1.3 trillion, dropping it to third among foreign holders behind Japan and the United Kingdom — while buying gold month after month (Congressional Research Service; SAFE).

Set those behaviors side by side and they appear to contradict each other. A country fleeing the dollar does not, in the same breath, ship the world a record surplus that can only be stored, overwhelmingly, in dollars; and a glut supposedly about access to the American market does not reach an all-time high in the very year that access is choked off. The behavior demands an explanation the usual vocabulary doesn’t supply. What kind of system produces a creditor that is feeding the dollar and fleeing it at once, and a surplus that grows precisely as its largest customer walks away?

Three explanations are on offer. The first is the familiar one: this is rivalry, China accumulates claims on the United States as leverage, and the bond-selling is a rival diversifying out of an adversary’s currency. The second is cyclical and is roughly the IMF’s: a country with a deep property bust leans harder on exports to replace lost domestic demand, and the surplus recedes when the bust does. Both are partly true and neither accounts for the pattern. The third does, and it is the one largely absent from the American conversation: the surplus and the dollar hoard are not two decisions but one, joined by an accounting identity, and China cannot stop the second without stopping the first. That makes it a captive creditor — captive not to American capital controls but to its own economic model. If that reading is right, almost everything said about this relationship in public has the direction of dependency backwards, and the stakes of getting it backwards are not academic: they determine who absorbs the cost of two enormous deferred decisions — China’s refusal to let its own households consume, and America’s refusal to pay for its own government honestly — and the answer, on both ends, is the people who never voted for either.

The captivity is in the aggregate, not the composition

A trade surplus is not a pile of money. By the structure of the balance of payments, a country that sells the world more than it buys must lend the difference back to the world — the surplus on the trade account is the accumulation of foreign claims on the capital account. They are the same fact seen from two sides. This is not a theory; it is an identity. Its consequence is that a surplus producer cannot reduce its aggregate foreign-asset position, on net, as a financial choice while the surplus runs, because the surplus is the act of acquiring those assets.

This is where the relationship’s defenders and most of its analysts go wrong, and the error is worth isolating because it looks like a refutation and is actually the point. China obviously can change the composition of what it accumulates. It can buy gold, lend through Belt and Road, settle some trade in renminbi, let the currency appreciate to compress the surplus at the margin, push outflows through sovereign funds rather than official reserves. All of that is real and all of it is happening. None of it touches the aggregate. Run the arithmetic: against a $735 billion current account surplus, China can sell $50–100 billion of Treasuries a year — roughly its recent pace — and its total stock of foreign claims still grows, because the surplus refills the tank faster than the selling drains it. The gold market is far too shallow to take up the difference: all the gold every central bank on earth buys in a year runs to roughly a thousand tonnes — on the order of $100 billion — and China is one of more than twenty buyers competing for it, against a single year’s surplus more than seven times that size (World Gold Council). So the diversification is genuine and the captivity is genuine, and they are not in tension, because they operate on different quantities. “Captive” describes the aggregate position — the requirement to hold something foreign equal to the surplus. The freedom the critics point to is compositional, and composition is not the trap. The trap is that the surplus must go somewhere, the only pool deep enough at that scale is dollar-denominated, and so the marginal flight into gold cannot be an exit whatever else motivates it — sanctions-hedging, duration, signaling — because the door is too narrow to fit a trillion-dollar flow through. The gold accumulation functions as exhaust, not exit: the visible residue of a creditor edging toward a door it cannot pass.

There is one apparent leak in this — private capital flight, legal and illegal, by Chinese households and firms — and it turns out to reinforce the argument rather than puncture it. Flight changes who holds the foreign claim, moving it from the state’s reserves to a private Chinese balance sheet, but the national constraint binds regardless of where the flight lands or in what currency: the identity pins the country’s net foreign accumulation to the surplus no matter who ends up holding the claim (and much of the flight lands in dollar assets anyway — U.S. real estate, dollar deposits). More to the point, capital flight is Chinese savers doing privately what the state does officially: fleeing the suppressed returns of a financial system engineered to hold household income down. It is the same impulse to escape, leaking through a different valve. It confirms the pressure rather than relieving it.

The household pays first, and pays certainly

Why is the surplus so large that it cannot be absorbed at home? Because the model is built to prevent home absorption. Chinese household consumption sits near 39 percent of GDP against a world average around 63 percent, with a national savings rate above 43 percent (World Bank, 2023–24). Even the most generous revision — crediting China with social transfers and housing services that international statistics miss, lifting the figure into the high 40s or low 50s (Peterson Institute; Zhang Jun) — still leaves one of the lowest consumption shares of any major economy. The gap is starkest in production: Federal Reserve research finds China produces roughly 35 percent of the world’s manufactured goods while consuming about 15 percent of the world’s output. A country that builds a third of what the planet makes and buys a seventh has to send the difference abroad. That difference is the surplus, and it is the product of a deliberate, decades-old arrangement that holds household income — wages, the safety net, the returns savers are permitted to earn — below what households produce, and routes the gap into investment and the state-linked corporate sector. Chinese economists name the causes without euphemism: a low labor share, weak social security, thin consumer credit, the residency rules that wall tens of millions of workers out of urban services (Wan and Luo, China & World Economy, 2025).

So the first and largest transfer in this story has already happened, and it happened entirely inside China, before any dollar left the country. The household was taxed through foregone consumption: income that was produced but never received, or received in forms that were not safely spendable. That cost is certain and complete. It does not depend on anything the United States does. This is the correction to the tempting straight-line story in which the Chinese saver’s loss arrives only at the end of the chain via American inflation. The saver’s principal loss is upstream and unconditional. What flows abroad is the proceeds of that domestic expropriation, held not by the household but by the state, in SAFE’s reserves.

That distinction matters for what comes next, because it means the American end of the wire operates on the state’s balance sheet, not directly on the household’s.

America’s path of least resistance is a tax nobody votes for

The United States carries a federal debt above $38 trillion, of which foreigners hold roughly 31 percent — about $9.2 trillion — and of those holdings some 42 percent sit with official, governmental hands rather than private investors (Congressional Research Service, 2025). A government with debt that large and rising has five historical ways to bring the ratio down: grow out of it, tax or cut its way out, default, inflate by surprise, or run a steady, low-grade financial repression — hold the rate it pays below the economy’s nominal growth rate long enough that the debt shrinks relative to GDP without being repaid in real terms.

The record on which path gets chosen is not ambiguous about its appeal. Carmen Reinhart and Belén Sbrancia, studying how the advanced economies retired their enormous World War II debts, found that from 1945 to 1980 real interest rates in the U.S. and U.K. were negative roughly half the time, quietly liquidating government debt at an average of three to four percent of GDP a year — compounding to a 30–40 percent reduction over a decade, larger than most explicit austerity ever achieves (Reinhart and Sbrancia, NBER 2011; Economic Policy 2015). They named it precisely: financial repression is “a tax on bondholders and savers.” Its decisive political property is that it requires no legislation. Default is a vote Congress must own; austerity is a vote; a wealth tax is a vote. Repression is a transfer executed silently through the interest-rate structure — low-salience, deniable, continuous. It is the path that happens when the others require courage.

This is the leg to state with its limits in full view, because it is where a careful reader pushes hardest, and the push is correct. The United States is not running financial repression now. The Federal Reserve has been fighting inflation, not capping yields into it; real yields are positive, which is the signature’s opposite. More: the machinery that made 1945–80 work — a captive domestic holder base under capped deposit rates and a closed capital account — is largely gone. With an open capital account and a powerful financial sector, repression today is harder to impose and noisier when attempted, and the post-2020 inflation has raised the political salience of real yields in a way that lets the bond market discipline faster than it did in the 1950s. So “America will inflate away its debt” is not a present fact; it is a structural hypothesis about where fiscal arithmetic bends when every alternative keeps requiring a vote. The clean falsifier runs the other way: if real yields stay positive and the premium investors demand widens as the debt grows, the bond market is winning, and the resolution is open crisis or genuine fiscal reckoning rather than quiet liquidation. The crucial thing — and the reason this leg does not carry the essay’s weight — is that the conclusion survives either branch. If repression comes, the foreign creditor is taxed slowly and silently. If repression fails and the debt reprices disorderly, the foreign creditor is taxed quickly and visibly through a mark-to-market loss. The mechanism and the timing diverge; the loss to existing holders does not.

And the holder, in the official slice, is disproportionately the surplus recyclers — China foremost.

The same maneuver, billed to the voiceless

Here the two ends connect, but along a more exact path than the straight line. The Chinese household’s loss is already booked, in consumption it never got to make. The proceeds sit in the state’s foreign claims. When the United States eventually resolves its debt — by repression or by repricing — the loss lands on the state that holds those claims: SAFE, the People’s Bank. Whether and how that second loss reaches the household depends on how the Chinese state absorbs it — through inflation, a weaker safety net, reduced services, or a levy on domestic deposits. An authoritarian state with command over its own price level, its own banks, and its own social spending is structurally well-equipped to pass such a loss down. So the household is positioned to pay twice: once certainly, in foregone consumption, and once contingently, when the state socializes the erosion of the savings that foregone consumption financed. The bagholder is the same in both rounds; only the second round runs through the state’s books on the way.

This is why the bilateral-hostage framing — China holds our debt, China has leverage over us — is not merely incomplete but inverted. The creditor here is the exposed party. The feared weapon, China selling Treasuries to punish Washington, has a specific defect: its recoil scales with its payload. A large enough sale to move U.S. yields cuts the value of China’s own remaining stockpile in the same stroke, and sustained selling means winding the surplus down, which forces the domestic adjustment China is trying to avoid. So the holding confers real disruptive capacity — a one-time sale can spike yields and signal displeasure — and essentially no coercive capacity, because compulsion requires sustained pressure and sustained pressure here is sustained self-harm. It is a deterrent at most, never an instrument of demand. The gold-buying is what it looks like when a holder understands this and reaches for a door.

The pattern worth naming, because the rivalry conversation never does, is that both governments are running the same maneuver — and the symmetry sits in the deferral, observed on both sides today, not in the resolution mechanism, which is asymmetric. China defers the fight over household income; the United States defers the fight over honest fiscal arithmetic. Both are present-tense facts: China’s $735 billion surplus is the suppressed consumption, and America’s chronic deficit financed by foreign capital is the avoided reckoning. Only the terminal form of the American deferral is contingent. And the asymmetry, looked at squarely, sharpens rather than softens the indictment: the United States has the genuinely available exit — it can raise taxes, a thing merely politically hard — and declines it, while China’s exit requires not just political will but a structural rebuild of its growth model. The American evasion is more clearly a choice; the Chinese one is closer to a trap of its own construction. Each evasion funds the other. China’s refusal to consume manufactures exactly the captive creditor base America’s refusal to tax relies on; America’s willingness to absorb the world’s surplus is exactly what lets China postpone rebalancing.

No one is colluding. There is no table at which this was agreed, which is worse, not better — a structural collaboration has no negotiating venue at which to be unwound. It is two domestic equilibria, each independently sustaining the condition the other needs, locked into a circuit that neither built on purpose and neither can exit alone. That is why decoupling rhetoric runs so far ahead of any actual disentanglement: the coupling is load-bearing for both regimes’ evasions, and the 2025 numbers show what “decoupling” actually does when only the external wire is cut. American imports fell, Chinese exports to America fell by nearly a third — and the surplus hit a global record, because the glut simply found other buyers. Cutting the wire from outside does not break the circuit. It only moves where the current grounds. The European Union and Southeast Asia, now absorbing the redirected flood, are discovering this in real time and beginning to raise the same complaints the United States did.

One thing makes the arrangement less durable than its longevity implies, and it is structural rather than cyclical. The dollar’s usefulness as the world’s store of surplus rests on the security and institutional order the United States built and is now, in other domains, visibly drawing down; every use of dollar dominance as an instrument of pressure teaches captive creditors to want out. The central banks’ steady gold accumulation is the residue of that lesson. But the reason there is no alternative sink is not a passing fact that financial innovation will fix — it is the same structure that creates the imbalance. To be the world’s surplus sink, an economy must do two things at once: run persistent deficits to supply the safe asset, and keep an open capital account so the asset trades freely. The eurozone could supply the depth but will not run the deficits; China could in principle run them but cannot open its account without surrendering the controls that hold household returns down. The only candidates to replace the dollar are disqualified by the very conditions that make the imbalance persist. So the system does not break; it limps. And “limps” is the dark reading, not the comforting one: it means the costs migrate, slowly and continuously, to the parties who cannot refuse them and mostly cannot see them, for as long as no one is forced to call it.

Evidence Framework

Documented in public records (Tier 1)

China’s 2025 trade surplus of about $1.19 trillion (customs; record), exports to the U.S. down ~29% while total exports rose via redirection to the EU, ASEAN, Latin America; current account surplus a record $735 billion — General Administration of Customs and State Administration of Foreign Exchange, January–February 2026; 2024 goods surplus ~$992 billion.
China’s U.S. Treasury holdings below $700 billion (from a $1.3 trillion 2013 peak), now third among foreign holders; foreigners hold ~31% of U.S. federal debt (~$9.2 trillion), ~42% of it official/governmental — Congressional Research Service, 2025.
China household consumption ~39% of GDP vs. ~63% world average; national savings >43% of GDP — World Bank, 2023–24.
China ~35% of global manufacturing output vs. ~15% of global consumption — Federal Reserve research (via Coalition for a Prosperous America, 2026; advocacy intermediary flagged); consistent with CFR/Setser on the “stealth” surplus.
Financial repression liquidated U.S./U.K. debt at ~3–4% of GDP/year, 1945–80, real rates negative ~half the time; “a tax on bondholders and savers” — Reinhart & Sbrancia, NBER w16893 (2011); Economic Policy 30:82 (2015).
U.S. federal debt above $38 trillion; Fed fighting (not capping into) inflation, positive real yields, as of early 2026.

Reasonable inferences from documented facts (Tier 2)

China is a captive creditor in the aggregate: it cannot reduce its total foreign-asset position while the surplus runs (BoP identity), though it retains compositional freedom (gold, BRI, RMB, appreciation) that does not reach the aggregate — supported by the arithmetic of surplus size versus diversification pace, and by persistent ~$3.3 trillion reserves.
The household’s primary loss is domestic and already realized (foregone consumption); the loss from U.S. debt resolution falls first on the state’s reserves and reaches the household only if the state socializes it — inference from the consumption/savings data plus who holds the claims.
The Treasury holding confers disruptive but not coercive capacity (recoil scales with payload) — inference from valuation mechanics of a large concentrated holding.
Capital flight relocates the foreign claim from state to private hands without changing the national total, and reflects savers fleeing the same suppressed returns — reinforcing, not refuting, the captivity reading.

Structural hypotheses requiring additional evidence (Tier 3)

The U.S. eventually resolves its debt primarily via repression / negative real returns rather than tax, cut, or default. Move to Tier 2 if: the Fed shifts to yield suppression as debt rises, term premia are held down, regulatory Treasury demand is tightened, inflation is tolerated above target. Falsified if: real yields stay positive and term premia widen as debt grows. (Either branch preserves the conclusion that the foreign holder takes a loss; only the mechanism and timing change.)
The relationship is best read as mutual deferral rather than rivalry. Move to Tier 2 if: both sides keep prioritizing the imbalance’s maintenance over its resolution. This is a structural reading of emergent (not coordinated) behavior.

Alternative explanations considered

“Underconsumption is overstated.” Adjusted for transfers-in-kind and imputed housing, China’s consumption reaches the high 40s–low 50s, and the surplus is only ~2.2% of GDP (PIIE; Zhang Jun). Insufficient: even fully adjusted, China’s share is among the lowest of any major economy; the small surplus-to-GDP ratio understates the absolute flow (a record $1.19 trillion the world must absorb regardless of the ratio), and the 35%-vs-15% production/consumption gap is the structural measure the ratio hides — the point CFR makes about the surplus’s statistical “invisibility.”
“Rivalry and leverage.” Insufficient: the holding cannot be used coercively without self-inflicted loss and forced domestic adjustment; the gold-buying fits exit, not coercion.
“It’s cyclical (property bust → export reliance).” Insufficient: the low consumption share and surplus model predate the bust by two decades; the 2025 record surplus arrived with the property sector still impaired and the U.S. market largely closed.

What resolution would require, and what to watch

Beijing: shift income to households at scale — raise the labor share, build the safety net, extend consumer credit, reform residency rules excluding migrant workers. The only move that shrinks the surplus without a crash.
Washington: choose the debt path openly — tax, cut, or explicitly accept inflation — rather than defaulting into repression by inertia. Unlike Beijing’s, this exit is available; it is merely unpopular.
Observers: track two numbers. China’s household consumption share (a structural climb toward the mid-50s means the captivity is loosening at its source) and the U.S. real-yield/term-premium signature (positive yields with widening premia = the bond market resisting repression; pinned-negative real yields with suppressed premia = repression begun).

Unresolved questions

The master hinge: does China’s consumption share rise structurally, or stay pinned? Every path routes through it.
How the U.S. debt resolves — quiet liquidation, disorderly repricing, or honest adjustment — which sets how the foreign creditor takes its loss, not whether.
Whether any sink other than the dollar can ever be deep enough, given that the only candidates are disqualified by the same conditions that create the imbalance.

The arrangement is stable only because both sides refuse to move, and that is its hidden fragility. The threat is not the familiar one of China dumping Treasuries. It is subtler and more binding: the system survives precisely because neither party attempts an honest exit, and the first one who does detonates the other’s deferral. Let China genuinely rebalance toward its own households and its surplus drains away — and with it America’s most reliable captive financier, just as the bill on U.S. debt comes due. Let the United States genuinely consolidate its fiscal position and the supply of safe dollar assets contracts — just as China most needs somewhere to store a surplus it has not yet unwound. Each reckoning is domestic, and each is survivable on its own. What neither government will do is go first, because going first means facing your own voters while the other side keeps deferring. So the wire stays live, and the bill runs quietly to the savers in Tianjin and the savers in Toledo — two groups who will never be told they are funding each other, sitting at opposite ends of one transfer, and who are, between them, the only parties to the arrangement with no vote on it.

METADATA (author review — not for publication)

v3 polish pass (two-model line review): accepted the precision and concision edits — “existing holders” (repricing hits the mark-to-market of current holders, not new buyers), “structurally” not “exceptionally” (drops an undefended cross-country comparative), “on net” on the aggregate claim, “at that scale” on the dollar-pool claim, currency-independent framing of capital flight (the identity binds regardless of where flight lands), and trims to the opening and symmetry paragraphs. Added a verified quantification to the gold point (~1,000 tonnes / ~$100B total annual central-bank buying vs. a $735B surplus; World Gold Council) and made that point motive-independent: gold cannot be an exit whatever drives it, because the market is too shallow — sharper than either the original “exactly what it looks like” or the reviewer’s suggested “consistent with.” Rejected two soften-traps: kept “detonates” in the close (the ending is about triggering the other side’s reckoning, not merely exposing it — “forces into the open” trades the mechanism for the theme), and did not trim the repression leg despite a length flag, because its length is the graceful-degradation defense and the conclusion must survive both its branches.

Revision basis (v2): multimodel review pass (six models). Rejected the dominant convergence — captive→constrained, collaboration→interdependence, leverage→self-limiting — as softening bias inconsistent with house discipline (commit in prose, hedge in the falsification apparatus). Converted each hedge-request into a precision-increase instead.

Sharpeners incorporated:

Transmission step (single best reviewer point, a divergence): split the cost into the certain, already-realized domestic loss (foregone consumption) and the contingent, state-borne loss from U.S. debt resolution that the state can socialize down. Fixes the “too-linear chain” complaint and hardens the moral architecture.
Aggregate vs. composition: made the captive claim precise — captivity is in the aggregate (= surplus), diversification is compositional and irrelevant — with arithmetic ($50–100B Treasury sales vs. $735B surplus; gold market too shallow).
Symmetry relocated: the symmetry is in the deferral (observed both sides now), the asymmetry in the terminal mechanism; the U.S. having the easier exit (tax) and refusing it strengthens the “evasion” framing rather than weakening it.
Leverage: sharpened “suicide vest” to recoil-scales-with-payload — real disruptive capacity, no coercive capacity.
Capital flight reframed from hole to reinforcement (savers fleeing the same suppression; relocates the claim, not the national total).
No-sink given its structural reason (a sink-supplier must run deficits + keep an open account; the only candidates are disqualified by the conditions that create the imbalance) — replacing the weaker “no current substitute” hedge with a stronger structural claim.
Ending upgraded to the double-bind (the first honest exit detonates the other’s deferral) — a more precise threat than “China dumps Treasuries.”
New empirical spine from verification: 2025 trade surplus a record $1.19T while U.S.-bound exports fell ~29% — the glut redirecting past the U.S. and growing, which preempts the “it’s bilateral” reading and shows decoupling-from-outside only reroutes the current.

Deliberately not done: did not give “fiscal consolidation” co-equal narrative weight (kept the asymmetry: repression needs no vote, which is its whole appeal) — but conceded the U.S. exit is more available, which sharpens the indictment. Did not add a broad scenario section (the double-bind ending does that work more precisely). Did not concede “no possible substitute” (gave the structural reason instead).

Brittleness: four independent evidence lines (BoP identity + consumption data; production/consumption gap; Reinhart-Sbrancia mechanism; reserve/Treasury/gold behavior). The repression leg is explicitly conditional and the conclusion survives both its branches.

Visibility mode: B. No constraint-framework vocabulary in the text. Confidence tracks source strength (identity + Tier-1 data written hardest; repression leg written conditional).