When the Elderly Outnumber Children, the Pension Bill Comes Due
Two numbers sit less than a percentage point apart. In 2025, one of them became the larger one.
At the end of that year, 223.65 million people in China were sixty-five or older. Children under fourteen numbered 212.76 million.
The old outnumbered the young by 10.9 million, and almost nobody noticed. The two shares differ by eight-tenths of a point: too small for a headline, large enough to mark a border.
For every hundred adults of working age, twenty-three are elderly and twenty-two are children. Supporting the old now costs more than raising the young.
No other country has run this arithmetic at this scale. Whatever answers get found here will be studied everywhere else, Washington included, where the same math is already on the table.
That is not a forecast. It is a closed ledger.
The retirement system was designed when the country was young, on a bet with three legs: each generation would be bigger than the last, paychecks would keep arriving, and people would die on schedule.
All three legs are gone.
Births in 2025 came to 7.92 million. Deaths came to 11.31 million. One year removed 3.39 million people. Life expectancy reached 79.25 years, up from 67.8 in 1981. Benefits run longer while the contributor base shrinks, and both feed the same account.
A Promise Written Against Future Paychecks
The money withheld from a paycheck does not sit in an account with the worker’s name on it. It leaves the same month, to someone who already stopped working.
What a thirty-year-old collects later depends on people who have not been born yet, or who arrived last year.
The system is an IOU against future labor, not a vault. The paper is worth exactly as much as the number of people working when it comes due.
Reserves blur the picture without changing it. In 2025, the enterprise pension fund took in 5.9 trillion yuan and held 7.3 trillion in reserves built over nearly thirty years. A reserve absorbs a bad year. It does not absorb a bad generation.
Building a reserve big enough to matter would mean charging the same workers twice: once for today’s retirees, once for themselves.
The ratio makes the squeeze visible. In 2012, 3.1 workers stood behind each retiree. By 2024, 2.6 did. The missing half-worker does not grow back.
It shows up as a longer contribution period, a bigger fiscal transfer, or a smaller annual increase. Every version lands on somebody.
The transfer is already big. In 2024, the central government steered about 1.1 trillion yuan into pension payments, money that goes straight into current benefits rather than anyone’s account. The guarantor on the IOU has quietly widened from the next generation to the entire tax base.
That guarantee competes with schools, roads, and interest on the debt, and it is paid by people who may never draw a pension of their own.
There is a number hiding inside the dependency ratio. The working-age category runs from fifteen to sixty-four: an age bracket, not a payroll. A twenty-five-year-old delivering packages and a sixty-year-old driving a taxi both count as supporters. Neither may be paying in. A ratio built from birthdays can flatter a system that runs on contributions.
Sixty-five, meanwhile, is a line drawn by statute, not by biology. A healthy seventy-year-old and a frail sixty-year-old live in different countries inside the same border, and the ratio sees neither.
The ratio is not destiny, either. If each worker produces twice as much, 2.6 workers can carry what 5.2 once carried, and the arithmetic eases. Productivity is the escape hatch. It is also the part of the math that care work cannot use, because turning a patient in bed takes the same pair of hands it took a century ago.
Pooling Provinces Does Not Pool Years
When a shortfall opens, the standard answer is national pooling. Provinces with young workers ship surplus to provinces with old ones. That fixes geography.
It does nothing about time. A surplus in 2025 covers a deficit in 2025. It cannot cover a deficit in 2045.
Redistribution across space can hide an overdraft across decades, for a while. Treat the two as one problem and the conclusion turns comfortable: the books balance.
They do not. The liability has moved from one province to another, and from one generation to the next. Moving a bill is not paying it.
A national average also hides the local range, which is where shortfalls actually live.
The Same Sixty-Three, Two Different Bodies
Retirement age started rising in January 2025. Men go from sixty to sixty-three. Women go from fifty and fifty-five to fifty-five and fifty-eight. The phase-in runs fifteen years. From 2030, the minimum contribution period climbs from fifteen years to twenty, six months at a time.
On paper, the arithmetic works: more years in, fewer years out, a narrower gap.
On a body, it reads differently.
Sixty-three in an office means three more years of meetings, a couple more hires trained, a retirement date turned into a consulting contract. Sixty-three on a scaffold means whether the climb still works, whether a back holds for a full shift, whether the knees take the stairs.
Physical capacity depreciates faster than a pension fund. The same three years buys time from one worker and bills the body of another.
None of this argues against a later retirement age. At 79.25 years of life expectancy, sixty is early. The problem is that a rule keyed only to age spreads the cost by occupation, silently and without appeal.
A table of ages has no column for job. A sixty-year-old marathoner and a fifty-five-year-old who cannot squat are filed as the same unit.
143 Yuan a Month
The other half of the ledger is thinner and quieter.
Get Nanmoon’s stories in your inbox
Join Medium for free to get updates from this writer.
In 2025, the minimum pension for urban and rural residents was 143 yuan a month, about twenty dollars. In 2014 it was 55 yuan. Eleven years bought an increase of 88 yuan, eight a year.
For 2026 the floor rises by twenty yuan, to 163.
Roughly 188 million people collect it. In most cities 143 yuan covers neither a month of utilities nor one clinic visit paid out of pocket, let alone a decent bag of rice.
For them the word pension is a courtesy. Real support comes from money sent by children, a spouse’s pension, a plot farmed at seventy, and credit at the local shop.
That is a second intergenerational transfer, run not by the state but at the kitchen table, and it never appears in a budget.
The inversion cuts unevenly. Formal-sector retirees have 2.6 workers behind each pension. Residents drawing the minimum mostly stand behind themselves. The same demographic shift means a slower raise for one group and a thinner grocery list for the other.
A retirement system is judged at the bottom, not the top. The 143-yuan floor says more about the country than the 15.9-percent headline.
Both ends of the budget now pull. Since 2025, families with a child under three get 3,600 yuan a year, 300 a month. The payment is warranted, and it is also a tell: at the floor, the state now pays more for a toddler than for an eighty-year-old.
The middle funds both, and the middle is shrinking.
The child payment points the same claim in the other direction. It assumes money now produces adults later, which is a fair assumption and not a guaranteed one.
Capital Moves; Caregivers Do Not
One constraint is harder than money.
About 320 million people are sixty or older, and roughly 40 million are disabled or semi-disabled. Certified care aides number 500,000. At one aide to four people who need care, the shortfall reaches 9.5 million.
Long-term care insurance covered 308 million people by the end of 2025 across ninety-two regions, with a nationwide rollout set for 2026. At the end of 2024, it had 8,800 designated providers and about 300,000 care workers.
Coverage is a rule. Reimbursement is a payment. Neither turns a body in bed.
A pension pays cash. It does not pay someone to be turned every two hours, which is why care is a separate bill and a harder one.
Turning, bathing, feeding, managing a tube, preventing a pressure sore: machines do parts of it badly and should not do all of it.
Care also happens in a room, at a bed, on a schedule, which keeps the shortage local even when the money is national.
Pension money can be stretched by transfers, equity allocations, and investment returns. In 2025, 2.31 trillion yuan of state capital was transferred into the system, and the national social security fund held 3.34 trillion in assets. Both sums are large. Set against 5.9 trillion in annual revenue, they are a few months.
A caregiver is not a balance-sheet entry that can be moved. A twenty-year-old weighs wages, dignity, promotion, and how much the shift hurts by the end of the day. That decision gets made in a hallway, not a ministry.
Aging does not stop at the treasury. It stops at the hiring notice.
No Single Payer
South Korea built a five-tier long-term care system before its aging curve steepened, matching home visits, day care, and cognitive support to dependency levels, and spending 580 million dollars on care technology. Public approval is high. Premiums and fiscal pressure rose anyway.
European systems fold farmers into statutory pensions through subsidies and rural mutual-aid schemes, and the Nordic version runs on high taxes and universal benefits. They work at a scale of ten million elderly, not 320 million.
The American mix leans on Social Security, personal accounts, and farm subsidies, which assumes deep household savings and a mature private market. That assumption does not travel either.
The lesson is not a program. No single payer has carried the load anywhere. Government alone runs out of budget. Families alone run out of children. Markets alone run out of customers who can pay, which is the quietest failure of the three.
Who Signs
Every argument arrives at the same question, usually without saying it out loud: whose name goes on the note.
A later retirement age charges the person close to retiring; a higher contribution rate charges the person working. A slower benefit increase charges the person already retired. A private account charges whoever still believes in a long horizon.
Each claim is defensible. Each group feels wronged. The ledger that would settle it has never been published.
Someone born in 1965 retired around 2025, after thirty-odd years of contributions, with a promise that payments would continue until death.
Someone born in 1995 turned thirty in 2025 and is paying for that promise now. When their turn comes, the signature may belong to one of the 7.92 million children born in 2025.
One system, two experiences. The older cohort feels repaid. The younger cohort feels drafted. Neither is wrong, because the terms were never written down: how much each cohort pays, how much it draws, for how long, and who absorbs the residual.
The politics are lopsided by construction. Benefits arrive now, to people who vote. Costs arrive later, to people who are not yet born or not yet eligible to vote. A system running on that asymmetry does not need a crisis to fail. It needs an election.
Without a ledger, the debate is only position. With one, it can be a contract.
A real ledger would fit on one page: contributions per cohort, benefits per cohort, the retirement age applied, the years of payment expected, the residual assigned to the treasury. Every figure would be contested. Precision is not the point; having a shape is.
Contracts can be revised. An inverting population is a reason to revise, not a reason to pretend. But the revision has to say which risk sits with the government, which with the individual, which with the market, and which no one can carry.
Ambiguity is not a compromise. It is a deferral, and deferrals are paid by whoever has the least room to refuse.
The residual usually lands on the household, which is another name for whoever has no paid alternative.
A pension is not a sum.
It is a signature against the future, and signatures change.
