Nobody Can Save a Nursing Shift

Strip away the acronyms and there is only one problem

Every retirement system ever built, anywhere, is a machine for one task: turn roughly forty years of working into twenty or thirty years of not working, without anyone starving at the end.

Easy to say. Brutal to engineer, because nearly every input is unknown at the moment you need it. Nobody knows how long they will live. Nobody knows what markets will do, what inflation will do, whether they will be laid off at 52, or whether they will need a hundred and fifty thousand dollars of long-term care at 81.

So a retirement system is not really a savings product. It is a scheme for allocating risk, and the only question it genuinely answers is this one: when the unknowns go badly, who eats the loss?

A pension says the employer and its fund. A 401(k) says you. Social Security says everyone at once, by formula, forever. That is the whole taxonomy. Everything else is implementation detail.

America changed retirement systems almost by accident

Mid-century, the model for a lot of American workers had three legs: Social Security, an employer pension, and personal savings. Inside a pension, retirement planning was close to automatic. You did not choose a contribution rate, pick funds, or estimate your own date of death. Many private pensions were also backstopped, within statutory limits, by the Pension Benefit Guaranty Corporation, so even the failure of your employer did not necessarily mean the failure of your retirement.

It had real gaps. Coverage was uneven and concentrated in large employers and union shops, and the formulas quietly punished anyone who moved. But for the workers inside it, the hard parts were somebody else's job.

Then the 1970s happened. Congress passed ERISA in 1974 after a run of pension failures, setting federal funding and fiduciary standards for private plans. Four years later the Revenue Act of 1978 added a modest provision to the tax code: section 401(k). It was not conceived as the thing that would replace the American pension. It was a rule about deferring compensation. Treasury regulations proposed in 1981 made clear that ordinary salary deferrals qualified, and adoption went vertical.

Employers noticed something about the new vehicle that had nothing to do with taxes.

Under a pension, the employer promises the outcome. Under a 401(k), the employer provides a container. Everything that determines whether the container is full at 67 now lives somewhere else: your contribution rate, whether you enroll at all, your fund choices, market returns, fees, withdrawals, job changes, and how long you happen to live. The Government Accountability Office describes this in exactly those terms, as a shift of retirement risk and responsibility from employers toward workers.

The scale is easier to see in plan counts than in any argument about it. GAO found that in 1975 the country had about 103,300 defined-benefit plans and about 207,700 defined-contribution plans. By 2015 there were about 45,600 defined-benefit plans and more than 648,200 defined-contribution plans. Labor Department filings tell the same story from the worker's side: active participants in private-sector defined-benefit plans fell from 27.2 million in 1975 to 11.3 million in 2022, while active defined-contribution participants rose from 11.2 million to 92.6 million.

What "just save more" is quietly asking of people

By the 1990s, without anybody announcing it, the United States had enrolled its workforce in a social experiment with a very clear question: can ordinary households successfully run a forty-year investment program by themselves?

To land well you have to get a job that offers a plan, enroll, contribute enough, capture the match, pick reasonable investments, avoid high fees, resist cashing out at every job change, keep contributing through recessions and emergencies, raise your rate as your income rises, and then, at the very end, convert a lump sum into an income stream that has to last an unknown number of years. Roughly ten decisions, each compounding, across four decades.

Two workers earn the same salary. One auto-enrolls at 25 at ten percent with a match and never touches it. The other starts at 38 at four percent, withdraws twenty thousand dollars after a layoff at 45, and pauses contributions a few times. At 65 their balances can differ by hundreds of thousands of dollars, and from the outside their careers looked identical.

But the deeper problem with "save more" is not discipline. It is arithmetic. In March 2025, 22 percent of workers in the lowest wage quartile participated in a defined-contribution plan. In the highest quartile, 64 percent did. Fewer than half of that lowest quartile had access to such a plan at all.

You cannot automatically invest money somebody does not have.

That one sentence eats most of the reform proposals that end with "and then people should contribute more." The households that most need to build retirement wealth are, reliably, the ones with the least left over to build it with.

Access is narrower than the three-legged stool implies, too. As of March 2025, 70 percent of private industry workers had access to a defined-contribution plan and 14 percent had access to a defined-benefit plan. And the generational split is stark: in the Federal Reserve's 2025 household survey, 52 percent of adults 65 and older had a defined-benefit pension, compared with 20 percent of adults aged 25 to 54. A large share of today's retirees are partly protected by a system their children will not inherit.

Meanwhile, 35 percent of non-retirees said their retirement savings were on track. That figure does not mean the other 65 percent face destitution; plenty of them have pensions, home equity, a spouse with assets, or simply time. But it is a good measure of how widely the anxiety is distributed, and the anxiety is not irrational. It is an accurate read on how many variables got handed over.

One of the things we handed over cannot be held by one person

Most of the risks in that migration are at least the kind of thing effort improves. Contribute more, choose cheaper funds, do not cash out. Fine.

One of them is different, and it is the one nobody talks about at parties.

You do not know when you will die.

This is not a morbid observation. It is a budgeting problem with no available solution. Retire at 65 with seven hundred thousand dollars. Die at 70 and you saved far too much and spent decades you did not need to spend carefully. Live to 102 and you saved nowhere near enough. Same person, same discipline, same portfolio. The difference between comfort and catastrophe is decided by a fact that is unknowable in advance and impossible to hedge alone.

An individual facing this has exactly two bad options: underspend for thirty years to cover a tail they probably will not reach, or spend at a reasonable rate and find out.

A pool does not have this problem. Not because pools are clever, but because averages behave.

Then I noticed my own model of this was wrong

I had been thinking about retirement as storage. You fill a container during the working years, you drain it afterward, and the policy question is how to help people fill it faster.

But nothing is stored.

A trillion dollars sitting in retirement accounts does not feed anybody. Retirees in 2060 will eat food grown in 2060, take medicine manufactured in 2060, and live in houses that people in 2060 maintain. Every single thing a retiree consumes is produced by the economy that exists at the moment of consumption. Money is the mechanism that lets them lay claim to a slice of it. It is a ticket, not a pantry.

Which means you can pre-fund a claim. You cannot pre-fund the goods.

Put it as concretely as possible. You can put dollars in a vault. You cannot put a nursing shift in a vault. Nobody can: not a 401(k), not a pension, not Social Security, not a gold bar, not a bitcoin wallet. The care you receive at 88 will be given by a person who is right now a child, or not yet born, using equipment nobody has manufactured. There is no version of saving that reaches them.

This is not a nihilistic reframe. It is a clarifying one, because it tells you exactly what a retirement system can and cannot do.

It can decide how reliable your claim is, and who absorbs the damage if it does not pay in full. It cannot manufacture the goods in advance. Every real argument about retirement policy is an argument about the first thing, conducted by people who believe they are arguing about the second.

Social Security stops being confusing once you stop picturing an account

Almost all of the confusion about Social Security comes from imagining that there is a giant account somewhere with your name on it. There is not, and there was never supposed to be.

It is social insurance. You and your employer pay payroll taxes while you work, and those revenues largely finance benefits for the people receiving benefits now. Surpluses historically accumulated in trust funds invested in special Treasury securities. Picture three generations: the working generation's taxes largely fund the retired one, and the generation now in school will fund the workers when their turn comes.

That does not make your benefit unrelated to your work. Social Security tracks your earnings history and runs it through a formula. Higher lifetime earnings produce higher benefits, but the formula is progressive: it replaces a larger share of earnings for lower-wage workers than for high-wage ones. That design choice tells you what the program is for.

It is also not only a retirement program. It pays survivor benefits when a covered worker dies and disability benefits when someone can no longer work, and it pools longevity risk about as well as anything ever constructed. Live to 105 and it does not inform you that your account ran dry at 91. The check keeps coming.

Which is also the source of its funding problem. A pay-as-you-go system is comfortable when many workers support few retirees, and strains when people live longer and that ratio falls. According to the 2026 Trustees Report, the retirement and survivors trust fund's reserves are projected to be depleted in the fourth quarter of 2032, at which point continuing income would cover about 78 percent of scheduled benefits. Combined with the disability fund, reserves run to 2034 with about 83 percent payable.

Notice what that 78 percent actually is, though. It is the system stating in advance, in public, a decade early, exactly who eats the loss and how much. That is more than almost any other retirement arrangement in the country will tell you.

If you were building this from scratch, one rule does most of the work

Carry the risks you can actually control. Pool the dice.

Run that rule over the list and the sorting is not especially controversial.

The left column responds to effort and preference, and you are the only person who knows what you want. The right column is dice. Making a single household the insurer of its own tail risk is the most expensive possible arrangement, because the household has to either self-fund the worst case or gamble.

So here is a system that is neither a pension nor a 401(k). Consider it a sketch, not a bill.

Instead of accumulating dollars, you accumulate units of future living standard. Every month a share of your earnings buys some. There is no balance to refresh, because you are not buying dollars. You are buying a standardized claim on real consumption in retirement: housing, food, utilities, transportation, basic care, and some discretionary cash on top.

And a unit is not defined as "two thousand dollars a month." It is defined relative to the economy. Something along the lines of: enough purchasing power to buy a set fraction of median working-age consumption. If groceries double, the claim moves with them. If the economy becomes dramatically richer over forty years, retirees participate in that instead of watching it from outside.

Isn't that just a pension? No, and the difference is precisely what pensions got wrong. A pension is a promise from one employer. This is a claim you own outright. Changing jobs does not matter. Being self-employed does not matter. You could accumulate units driving a truck, running a plumbing company, freelancing, or working at a bank, and the unit follows the person rather than the employment relationship.

Isn't that just a 401(k)? No, and the difference is precisely what 401(k)s got wrong. There is no portfolio for you to manage and no lump sum for you to mis-spend. Behind the scenes the issuing institutions obviously hold assets: equities, bonds, infrastructure, farmland, housing, energy. But you are not running that portfolio, any more than you tell your homeowner's insurer how to invest your premiums. What you bought was a transfer of risk.

At 67 you hold, say, 0.82 units. The system does not hand you nine hundred thousand dollars and wish you luck. It pays 0.82 units of consumption every year for as long as you are alive. Die at 73 and you received it until 73. Live to 108 and you received it until 108. The person who dies at 72 helps finance the person who reaches 104, which is not a tragedy or a subsidy. It is the definition of insurance.

Stop making 67 a rule. Make it a price.

Right now the retirement age is a number in a statute, and every argument about raising it is a fight over whether to take something away from everybody at once.

In a unit system it is an exchange rate. Your units convert at a factor that depends on when you start drawing. Leave early and each unit pays less per year, because the same claim has to stretch across more expected years of payments. Leave late and it pays more. Nobody has to tell you when to retire. The price tells you what the choice costs, and you decide whether it is worth it.

With one necessary exception, and it matters more than the mechanism does. A person who leaves at 58 because they would rather golf and a person who leaves at 58 because their spine is finished after three decades of roofing are not making the same decision. Disability has to be insured separately, on its own terms, or a clean price mechanism quietly bills the people it should be protecting.

How do you make catastrophe close to impossible for one person?

Not by making one mechanism excellent. That is the instinct, and it is the wrong one.

Engineers do not protect critical infrastructure by saying "this generator is extremely reliable." They say: if A fails, B starts. If B fails, batteries carry the load. If the batteries fail, the emergency generator runs. What makes the system survivable is not the quality of any component. It is that the failure modes do not overlap.

A retirement system can be built the same way, and mostly is not.

Decide in advance whose retirement absorbs a genuine disaster

There is one more piece, and it is the one nobody wants to write down, which is exactly why it should be written down early.

If something truly bad happens, the system should not pretend every promise is equally binding. It should have declared an order of losses long beforehand: food, basic housing, health care, and utilities protected first, then basic discretionary income, then enhanced benefits, and luxury retirement consumption absorbing the hit first.

Say it plainly, because it is the whole point. Somebody expecting ten thousand dollars a month might receive less. In exchange, somebody who depends on the system to buy food does not get pushed below the survival floor because wealthier retirees hold contractual claims that happen to sit ahead of theirs in line.

Every system has this ordering. Most have it by accident, discovered during the crisis, decided by whoever's contract was drafted more aggressively.

There is a limit, and it is not a financial one

Suppose that in 2060 there are a hundred million retirees and thirty million workers, harvests fail, energy is scarce, and output falls by a third. No pension, no 401(k), no trust fund, no bitcoin wallet, and no unit scheme creates the missing goods. You can reshuffle claims. You cannot conjure calories.

So the strongest honest promise a retirement system can make is not "you will always have enough." It is this:

As long as society can produce enough for every old person to live decently, no financial, institutional, demographic, or personal mistake will stop you from receiving your share of it.

That is dramatically stronger than what either a pension or a 401(k) offers. It is also visibly weaker than a guarantee, and the visibility is a feature. A system that states where it ends can be reasoned about, argued with, and reinforced. A system that pretends to have no end just fails somewhere nobody was looking.

You can reduce the dependence, though, and this is where the argument gets genuinely interesting.

Compare two retirees holding the same amount of value. The first owns nine hundred thousand dollars in financial assets. The second owns a paid-off efficient house, solar generation with storage, a share of farmland, a stake in utilities and infrastructure, and diversified ownership of productive businesses. The first still has to exchange dollars for whatever the future economy charges. The second has already acquired some of the things that make what they need. If electricity gets expensive, owning generation offsets the cost. If rents explode, owning the house shields you. If automation makes businesses extraordinarily productive, owners participate in the gain instead of only paying the higher prices.

There is a ladder here, from fragile to durable. Cash depends on inflation behaving. Bonds depend on borrowers and a currency. Stocks own productive companies. Globally diversified assets reduce dependence on any one country. Direct ownership of housing, energy, food production, and infrastructure reduces it further still.

But even the top rung is not independence. Solar panels need replacement inverters. Roofs need somebody to manufacture shingles and somebody willing to climb up there. A robotic caregiver needs components, electricity, software updates, and a supply chain. You have not escaped the future economy. You have moved from "I need the economy to prosper so my investments hold their value" to "I need civilization to keep maintaining the physical systems I already own." That is a meaningfully lower bar. It is not a different game.

And this reframe lands hardest on the argument everyone is actually having, the one about demographics. Fewer workers per retiree sounds fatal to a pay-as-you-go system. But the ratio that matters is not workers divided by retirees. It is productive capacity divided by what people need. If one worker in 2070, with machines, produces what five produce today, then the frightening ratio is simply the wrong ratio to be frightened of. The way retirees participate in that is by owning a share of the machinery, which is why the deepest version of the unit is not really a security unit at all. It is a claim, acquired across a working life, on a diversified slice of civilization's productive capacity.

The quiet version

So what about the Reddit comment I started with?

Probably overstated in its specifics, and pointing at something real anyway. It almost certainly will not look like tens of millions of elderly Americans with literally nothing, because Social Security is still there and 91 percent of retirees over 65 receive it, and many households own homes.

The plausible version is quieter and much worse at announcing itself. People retire later than they planned. Some keep working part time, not as a lifestyle choice. Adult children start sending money. Standards of living drop by a notch nobody declares. A house becomes the only real asset, and it is not a liquid one. A single long-term-care episode destabilizes everything. And people who expected retirement to mean independence end up almost entirely dependent on the one piece of the system that still pools risk.

Here is the statistic I keep turning over. In the Federal Reserve's 2025 survey, 73 percent of adults aged 55 to 64 had a tax-preferred retirement savings account. That sounds like good news, and in a sense it is. But having something and having enough are entirely different questions, and the 401(k) era is the first one in which the second question is put to each household privately, at the end, with no appeal.

We are about to get the first real readout. Roughly 2025 through 2040 is the first stretch in which large numbers of Americans retire after spending essentially their entire careers inside the 401(k)-dominant system rather than the pension-dominant one. Forty years of an unannounced experiment, reporting its results.

But the thing I would want anyone to carry out of all this is smaller and more useful than a forecast.

So the question to ask of any retirement plan, including your own, is not how much have I saved.

What claim did I buy, who has to honor it, and what happens to me if they cannot?

Nobody can save a nursing shift. Everyone is buying a promise that one will be there. The only genuine design choice in the whole subject is how many independent things have to go wrong before the promise breaks, and for the last forty years the answer has been trending toward one.