The Account That Isn't

ITwo stories

There are two stories about Social Security in America, and you'll hear both within a week of arriving.

The first is that it's socialism that snuck through. A mandatory payment, taken from your wages before you see them, redistributed to people who didn't earn it, wrapped in the language of insurance to keep the objections quiet. The people who tell you this are not stupid. They can point to the formula.

The second is that the money was stolen. You paid in for decades. The government took it, spent it on other things, left behind a drawer of its own IOUs, and now has to borrow from the public to pay you back what it already borrowed from you. It's a Ponzi scheme with a federal seal. The people who tell you this are not stupid either. They can point to the accounting.

Both stories are told with total confidence. Both are told by people who have paid into this program their entire working lives. They cannot both be right, and — this is the part that took me months to accept — neither of them is.

I pay into Social Security. I'm not American. I may or may not ever collect it; the answer turns on a bilateral treaty I'd never heard of until I went looking. That's an odd place to start from, but it has one advantage: I had no inherited story to defend. I just wanted to know what the thing was.

What I found was stranger than either version, and worse in ways neither camp complains about.

IIWhat you think you own

Start with the words, because the words are where this goes wrong.

You make contributions. You have an account. The money sits in a trust fund. What you get back are your earned benefits. Once a year a statement arrives showing what you've paid and what you're owed, formatted like something from a brokerage.

Every one of those words implies property. None of it is.

In 1936 a Bulgarian-born immigrant named Ephram Nestor began paying into Social Security. He paid for nineteen years. In 1955 he qualified. In 1956 he was deported for having belonged to the Communist Party in the 1930s — before membership was grounds for deportation — and his benefits were terminated. He sued, on the reasonable theory that he had bought them.

The Supreme Court disagreed. In Flemming v. Nestor it held that a person covered by the Act has no accrued property right in benefits, and that treating them as one would rob the system of the flexibility Congress deliberately preserved when it reserved the right to alter, amend, or repeal any provision.

Read the reasoning rather than the result. The Court didn't find a loophole. It found the design. The power to take your benefits away was written into the 1935 Act on purpose, and the Court declined to legislate it back out.

What you have is not an asset. It is a promise, revocable by majority vote, from a government to a citizen — and in my case, to a non-citizen who happens to be standing here paying for it.

Which makes the next fact strange. According to the Congressional Budget Office, that promise is the single largest thing the bottom half of American households own.

IIIWhat actually happens to a dollar

Here is the plumbing. It's simpler than the argument around it, which is part of the problem.

You pay 6.2% of your wages. Your employer pays another 6.2% you never see on a payslip, though most economists will tell you it comes out of wages that would otherwise have been yours. If you're self-employed you pay both halves yourself, 12.4%, and you feel it. In 2026 this applies to the first $184,500 you earn. Above that line, nothing.

In 2025, 185 million people paid this tax.

The money goes into two legally separate accounts: Old-Age and Survivors Insurance, and Disability Insurance. OASI is the big one. They are genuinely segregated from general revenue — that part of the popular story is right, and it's statutory.

But the trust funds are not allowed to hold money.

By law, Treasury invests all trust fund income in interest-bearing government securities, and the securities the funds hold are special issues sold only to them. Not cash. Not equities, not corporate bonds, not property. The moment a surplus dollar arrives it is lent to the federal government, which spends it, and what remains in the fund is a bond.

Hold that. It's the hinge of the whole essay and we'll come back to it in Section VI.

Now the flows. In calendar year 2025:

In — $1,449 billion. Roughly $1.3 trillion in payroll taxes. $58 billion from taxing benefits (which we'll get to, and which is stranger than it sounds). $69 billion in interest on those special-issue bonds.

Out — $1,609 billion. Almost all of it benefit payments. Administration cost $7 billion — four tenths of one percent. Whatever else is wrong with this program, it is not the overhead.

Gap — $160 billion. Covered by drawing down reserves, which fell from $2,721 billion to $2,561 billion. Every scheduled benefit was paid in full.

That last line matters. The system has been running a deficit since 2021, and has been paying out more than its non-interest income since 2010, and nobody missed a cheque. Both facts are true at once, and the inability to hold both is most of why the public conversation is broken.

One more thing about the plumbing, and it's the thing least understood: the money does not wait. Payroll tax collected this Tuesday is paid out to a retiree this Wednesday. There is no interval during which your contribution sits somewhere with your name on it, growing. The reserve exists to smooth timing, not to fund you. Your money was spent within days of leaving you, on someone who is already old.

That isn't a scandal. That is what the program is.

IVThe formula

To know what you'll be paid, you need three steps.

First, your AIME. Social Security takes your 35 highest-earning years, indexes each to national wage growth through the year you turn 60, adds them up and divides by 420 months. If you worked fewer than 35 years, the missing years enter as zeros and pull the average down. Careers interrupted for caregiving — overwhelmingly women's careers — are punished here, not for earning little, but for earning nothing at all in a given year.

Second, the bend points. Your AIME is cut into three slices. For someone first eligible in 2026, the cuts fall at $1,286 and $7,749. The first slice comes back at 90 cents on the dollar, the second at 32, the third at 15. Those three percentages have not changed since 1979.

Graph it and the line literally bends: steep, then shallow, then almost flat.

Third, when you claim. The result is what you'd get at 67. Take it at 62 and it's permanently cut by about 30%. Wait until 70 and it's roughly 124%.

Now watch what that structure does to real careers. SSA's actuaries publish five hypothetical workers — four built from the earnings patterns of actual insured workers, plus one who earns the cap every year. These are their figures for someone reaching 67 in 2027, in wage-indexed 2026 dollars:

Note that the lowest replacement rate is 75.5%, not 90%. The 90/32/15 figures are the marginal factors applied to slices of your earnings — nobody's whole benefit replaces 90% of anything. It's a distinction most explanations skip, and it matters: the formula is progressive, but not as dramatically as the headline percentages suggest.

Now the other half of the question — what you paid to get it. Here I'll use the Congressional Budget Office rather than my own arithmetic, because CBO models real work histories and real mortality tables. Its finding: people in the bottom fifth of lifetime household earnings receive about 2.5 times what they paid in payroll taxes. The middle fifth, about 1.5. The top fifth, about 1.0 — they get back roughly what they put in, before accounting for the fact that they put it in decades earlier.

CBO also names the two forces pulling against each other. The progressive formula pushes the ratio up for low earners, and so does the fact that lower earners are more likely to draw disability benefits. Pulling the other way: higher earners live longer, and so collect for more years.

That second force is larger than it sounds. The gap in life expectancy between the richest and poorest 1% of Americans is 14.6 years for men and 10.1 for women, and it widened through the 2000s — between 2001 and 2014 the top 5% gained roughly two to three years while the bottom 5% gained essentially nothing. Congress sets the monthly benefit. Mortality sets the number of months. Nobody legislated the second, and it runs against the intent of the first.

Still, the program is progressive. Meaningfully so. Which brings us to the point of this section.

No savings product on earth behaves like this. An annuity pays out in proportion to what you paid in. This pays 90 cents on your first dollars and 15 on your last. Two workers, one earning three times the other, do not get three times the benefit — they get nowhere near it. A high earner subsidises a low earner, by design, according to a formula written into statute and unchanged in forty-seven years.

You don't have to take my word for any of it. The machinery says what it is. It is a transfer.

One footnote, and it's a peculiar one. Social Security is not means-tested — your benefit depends on what you earned decades ago, not on what you have now. Warren Buffett collects. But there is a means test, bolted on through the tax code: above a combined income of $25,000 single or $32,000 joint, part of your benefit becomes taxable; above $34,000 and $44,000, up to 85% does.

Those thresholds were set in 1984 and 1994 and have never been indexed to anything. In 1984 you needed real money to cross them. Today a retiree with a modest IRA and an average benefit crosses routinely. America bolted a means test onto its non-means-tested program, forgot to index it, and let inflation widen it for forty years. Nobody voted for that. It simply happened.

VWho actually gets it

Ask an American who receives Social Security and they will say: retired people.

In June 2026 the program paid 71.3 million people about $138 billion for the month. Retired workers were 76.7% of them.

Which means nearly one in four beneficiaries — 16.6 million people — is not a retired worker.

Seven million are disabled workers. Nearly six million are survivors: widows and widowers, and a small, specific category called widowed mothers and fathers. Two million are spouses.

And 3.7 million are children.

Children of retired workers, children of workers who died, children of workers who became disabled. More people than live in Los Angeles, drawing Social Security cheques, and I have never once heard an American mention them. The program that everyone calls retirement is also the country's largest life insurance scheme and its largest disability scheme, and it does not advertise either.

The averages hide the shape of it. A retired worker gets $2,084 a month. A child of a disabled worker gets $531.

Now the part I did not expect to write.

The official poverty rate among Americans 65 and over is about 10%. Without Social Security, under the Supplemental Poverty Measure, it would be around 48%. The program lifts roughly 17 million older adults out of poverty, and provides the majority of income for half of all retirees.

I came to this subject annoyed. That number is the reason I stopped being only annoyed. Whatever else this is — and it is several things it doesn't admit to — it is also the most effective anti-poverty program in American history, and any account of it that leaves that out is selling something.

Which makes the last group in this section the sharpest thing in the essay.

To get anything at all you need 40 quarters of coverage — about ten years of work. Thirty-nine gets you nothing. Not a reduced benefit: zero. You don't get the taxes back either.

This is the only genuine cliff in the entire system. Everything else is a slope. And 2.6 million Americans aged 60 or older are standing on the wrong side of it — about 3.3% of that age group. Half of them are immigrants who arrived at 50 or later. Most of the rest simply never strung together ten years of covered work.

The numbers on that group are worse than I expected. 54.3% of them live below the poverty line, against 5.8% of everyone else their age. 63% are women. 36% never finished high school, against 9% of beneficiaries.

So the people locked out are disproportionately poor and disproportionately female — which is to say, precisely the people the 90% replacement band exists to protect. A progressive formula is worth nothing to someone who never reaches the gate.

They are not, incidentally, left with nothing at all. There is a separate programme — Supplemental Security Income — which is means-tested, paid from general revenue rather than payroll tax, and administered by the same agency from the same buildings. It pays 7.3 million people an average of $738 a month.

That is America's actual old-age welfare program. It is small, it is poor, it is fought over, and almost nobody can tell you it exists.

VIThe fund that couldn't fund

In 1937, two years after the Act passed and three years before the first cheque, Senator Arthur Vandenberg started asking an awkward question about the reserves piling up in the new system. His concern was that the government would not truly save them — that it would use them.

He got an advisory council out of it. What the council produced, in 1939, was something else entirely: a recommendation to add benefits for spouses, children and survivors, and to start paying two years earlier than planned. Congress enacted it. In doing so it converted an individual retirement scheme into a family insurance scheme, and redistributed benefits toward early participants and away from later ones, before a single monthly payment had been made.

It also renamed the pot. The Old-Age Reserve Account became the Old-Age and Survivors Insurance Trust Fund.

Reserve means an accumulation. Trust means a promise held on someone's behalf. Nobody explained the difference to anyone, and the savings vocabulary carried on unamended.

Whether 1939 formally moved the system to pay-as-you-go is genuinely disputed — Social Security's own historians say the law never stated it and that the "contingency reserve" language was a political fudge. Advocacy sources state it as settled fact. It isn't. What is not disputed is that the large accumulating fund envisioned in 1935 never materialised as designed.

Then, in 1983, America tried again.

The Greenspan Commission raised payroll taxes above the level needed to pay current benefits, deliberately, to build a surplus ahead of the baby boomers' retirement. That was the stated purpose. And the surplus, by law, had to be lent to the government that collected it.

Here is the thing I keep returning to.

You cannot pre-fund a national retirement system by lending money to yourself.

A country genuinely pre-funds future consumption in one of two ways: by increasing what it can produce, or by accumulating claims on foreigners. Buying your own government's debt does neither. It is a transaction between two of your own pockets. In 2032 the bonds will be redeemed, and to redeem them the Treasury must raise real money — from taxes, or from borrowing. Exactly as it would have had to if the trust fund had never existed.

The 1983 reform was not a fraud. The money went precisely where the law required. It's worse than fraud in a way: it was an incoherence, visible from the outside, raised by a sitting senator forty-six years earlier, and legislated around twice.

Two honest caveats. First, the alternative was not obviously better: a trust fund large enough to matter, buying equities, would have made the US government one of the largest shareholders on earth, with governance problems nobody had solved. Congress refused that, deliberately, and the refusal is defensible. Second, whether the trust fund surpluses encouraged looser spending elsewhere is contested among economists and I'm not going to pretend it's settled.

But note that "governments cannot invest" is not a law of physics. Canada faced the same arithmetic in the 1990s and built the CPP Investment Board, which holds real assets in global markets. It was a choice. America has made the other choice, repeatedly, and continues to.

So: 2032.

The OASI trust fund is projected to run out of bonds in the last quarter of 2032. This is where both opening stories die.

Nothing stops. Payroll taxes keep arriving every payday from 185 million people. At depletion, that incoming money covers 78% of scheduled retirement benefits. The combined funds, counting disability, last until 2034 at 83%.

But 22% comes off. Automatically. Not as a risk or a projection — as current law, in six years, applied to current retirees and future ones alike. Against the average retired worker's cheque of $2,084, that's about $460 a month, gone.

The panic is wrong. The reassurance is wrong. The truth is that America has legislated an automatic benefit cut and then spent decades arguing about whether the program is "going bankrupt," which it can't, instead of about the cut, which is scheduled.

The Trustees are blunt about what would close the gap if Congress acted this year: raise the payroll tax from 12.4% to 16.65%, or cut benefits 25.2% across the board, or 30.3% for new beneficiaries only. Wait until depletion and the tax number becomes 17.3%. Those are not proposals. They are arithmetic.

Behind all of it sits demography. In 1960 there were 5.1 covered workers for every beneficiary. In 2025 there were 2.6. By 2040 it's projected at 2.3.

Be careful with that series, though — it's the most abused statistic in the debate, and I nearly misused it myself. The commonly quoted 16.5-to-1 for 1950 is close to meaningless: the program was thirteen years old and most people hadn't qualified yet. In 1945 the figure was 41.9, which tells you nothing about anything. Most of the early collapse is a young scheme maturing, not a society aging. The honest series starts around 1974, when SSA's own analysis notes the ratio settled and stayed between 3.2 and 3.4 for the next thirty-five years.

The second thing the raw number hides is where it stops. The ratio does not fall forever. Under the Trustees' intermediate assumptions it declines to about 1.9 by 2075 — and then holds there, flat, through 2100. This is not a spiral. It's a one-time step down to a new level, driven by a birth-rate change that has already happened. That distinction matters enormously for how you think about the problem: America is not facing an accelerating burden, it is facing a permanent shift to a higher one.

And there's a counter-argument worth sitting with. Count all dependents — children as well as the old — and America carried a heavier load in 1960, when the boomers were in school, than it's projected to carry in 2030. The country has supported this many non-workers before. It supported them through schools and households rather than federal cheques, which is a real difference. But "we have never had to carry this" isn't true.

VIIForced to save, or forced to transfer

There's an objection to all this that sounds like a first-principles objection and isn't: they force you to pay.

Every tax is compulsory. That can't be the complaint.

The real complaint is narrower and much harder to answer. A country organised more thoroughly around private property than any other on earth runs its largest single program on a mandatory payment that generates no property right whatsoever. You cannot direct it, exit it, borrow against it, or leave it to your children. You are compelled to fund an asset you are legally barred from owning. Flemming v. Nestor said so, and the Court said the barrier was the point.

I came into this assuming Australia — where I'm from — simply did it better, and that the comparison would be a rout. It isn't, and the reason is instructive.

Australian superannuation is also mandatory. Employers must pay a percentage of wages into a retirement account, and workers cannot opt out. Australia doesn't coerce less. It coerces in a different direction.

Australians are forced to save. Americans are forced to transfer.

The Australian version produces a real, owned, inheritable asset with your name on it, invested in real markets, which grows and which passes to your children when you die. It also charges fees for decades, exposes you to markets you didn't choose and may not understand, produces a large and well-documented gap between what men and women accumulate, allows people to draw it down badly, and offers nothing at all to someone who never worked. Australia patches that last hole with a separate means-tested Age Pension — which is to say, with welfare.

The American version produces no asset, no inheritance, and no market exposure. It cannot crash. It cannot be mismanaged by you. It pays until you die however long that takes, adjusted for inflation every year, and it pays your widow and your children if you die early. And it can be reduced by an act of Congress, and in 2032 it will be.

Neither of those is obviously the better deal. They're different bets about which risk you'd rather run: the market's, or the legislature's.

Canada took a middle path in the 1990s, part pay-as-you-go and part genuinely invested in global assets through an arm's-length board. Chile went all the way to individual private accounts in 1981 and has spent the last two decades dealing with the consequences of low coverage and low replacement rates.

The question I'd put to an American isn't which system is better. It's the one their own system never asks: do the countries whose bottom half accumulate real, inheritable assets end up with a less concentrated distribution of wealth — or does the failure just move somewhere else?

I don't think that's settled. But America has never seriously tried to find out, and the reason isn't economic.

VIIIWhat it actually is

I wrote an essay a while back — Cheap Things, Expensive Life — about why a country that has become measurably richer feels poorer, and it ran into something it didn't have room to chase. The Congressional Budget Office, when it measures family wealth, counts the present value of your future Social Security benefits as an asset you own.

The effect is enormous. Counting it, the top 10% of American families held 60% of all wealth in 2022 and the bottom half held 6%. Strip it out and the top tenth holds nearly 70% while the bottom half holds 3%. For families in the bottom quarter of the distribution, accrued Social Security is roughly half of everything they have.

Both measures are defensible. They answer different questions. But the reason the second one exists is the thing this essay has been circling.

Social Security wealth cannot be spent. It cannot be sold. It cannot be borrowed against. And it cannot be inherited.

A family whose largest asset is several hundred thousand dollars of accrued Social Security passes on precisely none of it. The money arrives monthly, is consumed monthly, and stops at death. Whatever the next generation starts with, it does not start with that.

Which means the single largest thing the bottom half of America owns is the one thing that resets to zero every generation.

And it is the thing most exposed to 2032. CBO models this directly, in an appendix almost nobody reads: because Social Security wealth is a far larger share of what poorer families hold, a benefit cut reduces their wealth by proportionally more than it reduces anyone else's. The automatic reduction now written into law is, in wealth terms, regressive. It falls hardest on the people for whom this promise is nearly everything.

That is not a scandal. Nobody stole it, and the program does exactly what it was built to do — it has kept something like 17 million old people out of poverty this year alone, and by any reasonable accounting that matters more than the arguments I've made against it. But the mechanism that prevents destitution is the same mechanism that prevents accumulation. It converts what might have been savings into a promise, and promises don't compound across generations. One design decision, two consequences, and America only ever discusses the first.

I don't think the country was lied to about the money. The money went where Roosevelt said it would.

I think it was lied to about the category — and not by accident. When a critic complained to Roosevelt that the payroll tax was regressive, he agreed about the economics and said the taxes had never been a matter of economics at all. They were politics, all the way through. The contributions were there to give people a legal, moral and political right to their pensions, so that no politician could ever scrap the program.

It worked. It is still working. The vocabulary of savings has protected this thing for ninety years against everyone who wanted it gone.

But there's a cost to a country believing its largest transfer program is a savings account, and the cost is coming due in 2032. You cannot have a sensible argument about how to adjust a transfer if half the room believes it's adjusting a contract, and the other half believes it's cancelling a theft. Both of them are arguing about a thing that doesn't exist.

The account isn't an account. It never was.

It's a promise a country made to itself, and the only thing standing behind it is the willingness of the people currently working to keep paying for the people who currently aren't.

That's not nothing. It might even be more solid than a market. But it isn't what anyone thinks they bought, and America has spent ninety years making sure of that.

Sources & method

Program mechanics, beneficiary composition and financial operations: Social Security Administration — monthly statistical snapshot (June 2026), the 2026 OASDI Trustees Report, and SSA's historical and policy archives. Benefit formula parameters and the taxable maximum for 2026: SSA and Congressional Research Service (Social Security: Benefit Calculation Overview, IF11747). Replacement rates and scaled-worker figures: SSA Office of the Chief Actuary, Actuarial Note 2026.9 (June 2026), Table C, with earnings patterns from Actuarial Note 2026.3. Lifetime benefit-to-tax ratios: Congressional Budget Office, CBO's 2024 Long-Term Projections for Social Security. Wealth measurement: CBO, Trends in the Distribution of Family Wealth, 1989 to 2022 (October 2024), including Appendix C on payable benefits. Poverty figures: Census Bureau Supplemental Poverty Measure. Life expectancy by income: Chetty et al., JAMA 315(16), 2016. Never-beneficiaries: SSA, Population Profile: Never Beneficiaries, Aged 60 or Older, 2024 (released May 2024; MINT8 microsimulation on 2023 Trustees intermediate assumptions).

A note on secondary sources. While researching this I found several widely-read personal-finance sites publishing 2026 bend points that were actually 2025's, and a 2026 taxable maximum of $176,100 — also 2025's. Every parameter in this essay comes from SSA, CBO, CRS or the underlying academic paper. If you check my numbers, check them there.

Where sources disagree, I've said so in the text. Two disputes are worth flagging again: whether the 1939 amendments formally established pay-as-you-go financing (SSA's own historians say the law never said so), and whether trust fund surpluses enabled looser federal spending elsewhere (unsettled among economists). I've taken no position on either.