TL;DR
The standard account of the pension shift says coverage collapsed: employers dropped defined-benefit plans, individuals were left holding market risk, fewer people are protected. The first and last parts are wrong, and the Federal Reserve’s own survey says so. In 2022 just over two-thirds of working-age families participated in a retirement plan — the highest level since 2010 — and participation rose across every income group between 2019 and 2022, driven by IRA and DC growth everywhere and by increased DB participation among families below the 90th percentile.1 So the system is not excluding people. What it does instead is transmit outcomes in proportion to the balance you already hold. Among families who have these accounts, the mean combined IRA and DC balance reached $331,400 in 2022. For the top decile of earners it rose to $913,300. For the bottom half it fell, from $66,600 to $54,700 — over three years when the stock market rose substantially.1 The risk did move. It did not move by locking anyone out.
An emptied-out room. The argument about pensions is usually about who is in the building. Photo: Lisa from Pexels.2
The Story, and What’s Wrong With It
The narrative is familiar enough to recite. Companies used to promise a pension. The Revenue Act of 1978 created the 401(k). Employers realised a defined-contribution plan let them replace an open-ended liability with a fixed contribution, and over four decades the guaranteed pension was replaced by an account whose value depends on markets. Risk that once sat on a corporate balance sheet now sits on yours.
The mechanism in that story is real. A DB plan owes you a defined income; a DC plan owes you whatever is in the account. That is a genuine transfer of who bears investment and longevity risk, and it happened.
But the story is usually told with an implied second claim — that this left people uncovered — and on the best available household data, that part does not hold.
What the Household Data Says
The Federal Reserve’s Survey of Consumer Finances counts a family as a retirement-plan participant if they have any of an IRA, an account-type DC pension through an employer (401(k) or 403(b)), or a DB pension through an employer. Its Box 1 focuses on working families with a reference person aged 35 to 64 — old enough to have finished education, young enough not to have retired.1
Three findings, none of which fit the collapse account.
Participation rose across the income distribution between 2019 and 2022.1
Overall participation was at its highest level since 2010.1
And the composition of that increase is the surprise: it was driven by IRA or DC growth across all income groups and by increased participation in DB plans for families below the 90th percentile.1 Defined-benefit coverage went up for most of the distribution over those three years.
Families remain about twice as likely to hold an IRA or DC plan as a DB plan.1 The long-run shift toward DC is not in dispute. What is in dispute is the idea that its consequence is exclusion.
Where the Risk Actually Landed
Among families that hold IRA or DC assets, the mean combined balance was $331,400 in 2022, with the gains concentrated in the top half of the income distribution.1
Break it apart and the mechanism becomes visible:
- Top decile: rose more than 10% to $913,300
- Upper-middle: rose more than 10% to $226,700
- Bottom half: fell from $66,600 to $54,7001
Read the third line against the period it covers. Between the 2019 and 2022 surveys, major stock indexes rose substantially — the same report notes robust growth in the median and mean value of stock holdings across all major income groups.1 In a rising market, the mean retirement balance of participating families in the bottom half of the income distribution went down.
That is the actual shape of the risk shift, and it is not the shape the popular argument describes.
Why a Neutral-Looking Account Isn’t Neutral
A defined-contribution plan is a proportional instrument. It returns a percentage. A 20% market gain adds $2,000 to a $10,000 balance and $180,000 to a $900,000 one, and both people were offered exactly the same product on exactly the same terms.
Three things then compound in the same direction.
Contribution capacity. You can only defer income you don’t need. A family with no slack contributes nothing in a bad year, and there is no mechanism to make that up.
Withdrawal under stress. A DC balance is reachable. When a household with no other buffer hits a job loss or a medical event, the retirement account is the buffer — which is one plausible reading of a mean balance falling in a rising market, though the survey I am citing does not decompose it that way.
Employer match as a multiplier. A match is a percentage of what you put in, so the subsidy is largest for those contributing most.
None of this requires anyone to be excluded. The 401(k) is available, participation is at a twelve-year high, and the outcome still diverges — because the instrument’s output is a function of an input people have wildly unequal amounts of.
This is the retirement version of a pattern this site keeps running into. The same exam is not the same chance when preparation is purchasable, and advice is differentially takeable depending on whether you have the resources to act on it. Uniform access, non-uniform capacity, divergent outcome — and no rule anywhere that treats anyone differently.
What About Ayn Rand
Accounts of this shift usually open with intellectual history: Rand’s The Fountainhead, “rational selfishness,” the Chicago School, a young Alan Greenspan in Rand’s circle, and then — the implication goes — four decades of policy.
I have deliberately left that out of the argument, and it is worth saying why. It is a real and interesting story about who read what. It is not evidence that those ideas caused these policies. The 401(k) arrived through a technical provision of a 1978 revenue bill, not a philosophical conversion, and the corporate move to DC plans is adequately explained by wanting to convert an unpredictable liability into a fixed cost. Ideas can supply permission for a change that accounting was already pushing for, and telling the story the other way round is very hard to check.
I’d rather stand on a number in a Federal Reserve table than on a chain of influence.
Where I’d Hold This Loosely
Five limits, and the first is a real weakness in my evidence.
One survey, one three-year window. 2019 to 2022 includes a pandemic, emergency fiscal transfers, hardship-withdrawal rule changes, and an unusual labour market. The bottom half’s falling balance may reflect withdrawals under those specific conditions rather than a structural feature of DC plans. I cannot separate those from this source, and the direction of the long-run trend is not established by a single interval.
Second, I could not read the administrative data. The Department of Labor’s Form 5500 series is the authoritative record of the DB-to-DC transition in plan counts and participants, and both it and the Congressional Research Service summary of it refused every route I tried. So this post makes no claim at all about the long-run trend in plan counts — the thing the standard narrative is actually about. I have described the mechanism and shown one recent cross-section. That is less than I wanted.
Third, “participation” is a low bar. Holding an IRA with $500 in it counts. A participation rate at a twelve-year high is compatible with a great many people being nowhere near secure, and I have used it only to rebut the exclusion claim, not to suggest adequacy.
Fourth, mean balances are dragged by the top of each group. The bottom half’s $54,700 mean will sit far above its median, so the typical participating family in that group is worse off than even this comparison suggests. That cuts in favour of my argument, which is exactly why I should flag that I have not seen the medians.
Fifth, and against my framing: a DC plan genuinely does some things a DB plan did not. It is portable in a labour market where people change employers often, it does not vanish when a single employer fails, and it is inheritable. The people most harmed by DB plans were those who left before vesting — which was a great many workers, disproportionately women and the mobile. “The old system was better” is a claim about averages that conceals its own distribution, and I have not checked that one either.
