A Degree Didn't Stop Paying. It Stopped Paying More.

On October 9 the Federal Reserve published the 2025 Survey of Consumer Finances, the study of American family finances it has run every three years since 1989. It is not a jobs report and it got almost no coverage. It contains the most useful thing anybody published about American pay this year.

Here is the headline, and the headline is already two numbers. Between 2022 and 2025, real median family income rose 7 percent, to $82,200. Over exactly the same period, from exactly the same survey, real mean family income fell 6 percent, to $145,200.

Both are correct. Neither is a revision of the other. The median is the family in the middle; the mean is the total divided by the count. When the middle rises and the average falls, income has not gone up or down. It has moved inward.

Where it moved from

The survey publishes income by the education of the person it classifies each family under. In constant 2025 dollars, before tax, 2022 then 2025:

Families whose reference person has no high school diploma: median $35,400 to $36,300. Mean $46,200 to $52,300.

High school diploma: median $58,000 to $60,600. Mean $81,000 to $82,700.

Some college: median $65,700 to $69,000. Mean $94,600 to $95,900.

College degree: median $128,900 to $129,200. Mean $265,000 to $232,900.

Read down the median column and every single group gained. Read down the mean column and three gained and one lost - the college-degree group, by about 12 percent. The largest mean gain anywhere in the table, about 13 percent, belongs to the families with the least formal education.

Those percentage changes are arithmetic on the table's own levels. The table prints levels only.

What that does and does not mean

It does not mean a degree stopped paying. The levels say the opposite, loudly. A college-degree median of $129,200 is still three and a half times the $36,300 median for families whose reference person never finished high school. Nothing in three years touched that.

What moved is the top of the college distribution. A mean falling 12 percent while its own median holds flat is the signature of the upper tail coming down, not the middle going anywhere.

So the return to a degree is intact at the median and compressed at the top. That is a different sentence from either "degrees are worthless now" or "the gap is closing", and it is the only one the table supports.

Where the premium went

The Federal Reserve's own committee record offers a partial answer, and it has to be read as a separate instrument rather than as more of the same number. The minutes of the September meeting, published on October 7, record that strong demand for skilled workers in artificial-intelligence-related sectors was driving strong wage gains for those workers.

That is a committee's qualitative read of recent months. This is a triennial survey of family income. They cannot be added together.

But set side by side they describe a single shape. A credential that used to lift the whole top of the distribution now lifts a slice of it. The average came down because most of the top did not move, while the gains concentrated somewhere narrower than a diploma.

Why this is worth more to an employer than the monthly jobs number

A payroll figure tells an employer how many people are working. It says nothing about what a job is worth to the household behind it, and nothing about which households are gaining.

This survey does. Over three years the gains landed mostly in the middle and at the bottom of the education distribution, and the familiar assumption - that pay rises fastest where credentials are highest - did not hold.

For anyone setting a wage band on a role that does not require a degree, that is not a social observation. It is a market condition. The families you are hiring from had their income rise, and expectations rise with income.

It also cuts against the quiet logic of a job posting that asks for a degree the work does not need. If the pay premium at the top of the credentialed distribution is compressing while the middle of every group gains, then a degree requirement is buying an employer less than it did three years ago - and it is still screening out the same applicants it always did.

And one more finding belongs here, because it says why the requirement survives anyway. Asked what they would prefer regardless of whether a role calls for a degree, 80 percent of American employers chose the candidate who has one, 31 percent strongly. Gallup and the Lumina Foundation published that on October 19, from 2,500 employers who make hiring decisions. It is an opt-in, unweighted panel with no margin of error calculated, so it is what employers say and never a measurement of who gets hired. The premium at the top is compressing. The preference is not.

The cautions that travel with every number above

Education of the reference person is not the education of everybody earning in the family. This is a family income measure classified by one person's schooling, which means it is not a return to a degree for an individual and must never be quoted as one.

The survey is triennial. There is no 2023 observation and no 2024 observation, so there is no path between the two columns - only a start and an end, three years apart, with a different labour market at each end.

And the method is only partly recoverable. The press release publishes no sample size at all. The full report publishes one: 4,367 families interviewed in 2025, against 4,602 in 2022. Neither document states a fielding period or a response rate.

One line to hold

The question an employer or a jobseeker reaches for is what this kind of work pays now. The survey says that depends on something nobody asks: whether you mean the family in the middle of a group or the average of the whole group. Between 2022 and 2025 those two answers moved in opposite directions.

A degree didn't stop paying. It stopped paying more.