Wages Didn't Stall. They Lost a Race.

On October 21 the Bureau of Labor Statistics published what American workers usually earn in a week. The median was $1,232, up 1.5 percent over the year. The same release prints, four tables later, that in constant dollars the same median fell 2.1 percent.

That is not a contradiction and neither figure is a revision of the other. It is one number measured against two different things: itself a year ago, and what a dollar bought a year ago. Prices rose 3.5 percent over the same period, by the agency's own comparison. A raise of 1.5 against a rise of 3.5 is a pay cut that arrives looking like a raise.

What the release actually says

Median usual weekly earnings of full-time wage and salary workers, third quarter 2026: $1,232. Third quarter 2025: $1,214. In constant 1982-84 dollars the same two quarters read $367 and $375.

The percentage fall is this Desk's arithmetic on those two printed levels. The agency publishes no real percent-change column at all, which is worth knowing before anybody quotes one.

On the seasonally adjusted series the median fell from $1,258 in the second quarter to $1,239 in the third. That is a different comparison on a different basis and it is not netted against the annual figure.

A second instrument says the same thing on a different measure

Average hourly earnings across private employers, adjusted for prices, fell 0.7 percent over the year to September and 0.5 percent over the month. That is a different survey, a different unit and a different cadence from the weekly median above, and the two are never added together. What matters is that they point the same way. One measures what a typical full-time worker takes home in a week; the other measures what an hour buys across every private payroll in the country. Both lost ground.

The ladder is intact and the whole ladder moved down

For workers 25 and over, by schooling: less than a high school diploma $819. High school graduates, no college, $993. Some college or an associate degree, $1,140. Bachelor's degree and higher, $1,752.

The premium is real and it is not in question. What moved is the floor under all four rungs, because the deflator does not care which rung a worker is standing on.

There are two races, and young workers are losing both

The same release, the same quarter, the same table: median usual weekly earnings for full-time workers aged 16 to 24 are $773. For workers 25 and over they are $1,307. That is a gap of 1.69 times at the median, on one instrument, with nothing merged.

Read that beside the first race and the position is worse than either figure alone suggests. A young worker is not only running behind prices; they are running behind the rest of the labour market while they do it. The first race is against the deflator and everybody is in it. The second race is against seniority, and losing the first one compounds the second, because a percentage raise on $773 buys less recovery than the same percentage on $1,307.

One caution has to travel in the same breath, and it cuts against the figure rather than for it: this is a full-time measure, and part-time work is a far larger share of employment at those ages than at any other. So $773 describes young people in full-time jobs, not young workers generally, and the young workers it leaves out are mostly the ones earning least.

Britain measures it from the other end and finds the same shape

On October 22 the Office for National Statistics published how many British jobs pay badly. Across all employee jobs, 2.8 percent are low-paid - hourly pay below two-thirds of the median. Among employees aged 16 to 21, the figure is 33.6 percent.

That is a different country, a different instrument and a different definition from anything above, and it is provisional until October 2027. It is cited here for its shape and not for its level. The same bulletin also separates by hours: 6.6 percent of part-time jobs are low-paid against 1.3 percent of full-time ones. Which is the other half of the American caution, measured.

Why this matters to somebody setting pay

An employer who benchmarks a wage band against last year's median and adds two percent has not held position. They have cut the band in real terms and will find out at the second interview, from candidates who are doing this arithmetic in a grocery aisle rather than in a spreadsheet.

The entry band is where that bites hardest, because it is the band with the least room above the floor and the most competition from the part-time market underneath it.

And for the worker, the useful thing is not the number. It is knowing that the feeling of running to stand still is a measurement, not a personal failure.

The cautions that travel with every number above

These are individual weekly earnings from the household survey, wage and salary workers only, all self-employed excluded. They are not family income and must never be set against the family-income figures from the Survey of Consumer Finances, which is a different unit on a three-year cadence.

Medians are interpolated from fifty-dollar-wide intervals, so some estimates move sluggishly for arithmetic reasons rather than economic ones.

There is no fourth-quarter 2025 estimate at all. The October 2025 collection was lost to the federal funding lapse, so any comparison spanning that quarter cannot be made.

What this spine still needs

A wage-band benchmark somebody actually uses, so the employer paragraph rests on a practice rather than an inference. And a Houston-specific real-pay figure, which does not currently exist on any readable source and may be the honest limit of the piece.

Wages didn't stall. They lost a race. Two of them, and the people at the start of a working life are in both.