Hero image for "The "Record High" Income Number Comes With a $1,040 Asterisk"

The "Record High" Income Number Comes With a $1,040 Asterisk


The Census Bureau announced this week that real median household income hit $87,460 in 2025, up 2.6% from 2024 and the highest figure since the Bureau started tracking it in 1967. That's a real number, and it's genuinely good news. But the press release version of this story — "highest ever" — skips the part where the Bureau itself reports the estimate with a margin of error of ±$1,040.

That margin matters because of what it's being used to argue. The same report notes that 2024's median of $85,210 "was not statistically different" from 2019's pre-pandemic level — a gap of roughly $2,000 that the Bureau itself declined to call meaningful. If a $2,000 gap wasn't statistically distinguishable in that comparison, a "record" framed as a single point estimate above a previous point estimate deserves the same skepticism, even when the new number is more comfortably outside the range. The headline treats the 2025 figure as a clean fact. The Bureau's own methodology treats it as a range.

Then there's the split between pre-tax and post-tax income, which the coverage mostly ignores. Post-tax median household income was $76,060 in 2025, up 3.1% from 2024 — but not a record. It still hasn't caught up to 2020 or 2021, when pandemic stimulus payments inflated post-tax income. So depending on which of the two numbers you quote, "highest ever" is either true or four years stale. Nobody picking the headline stat mentions which one they're using.

The inequality context gets buried further down, and it's the part I'd actually lead with. From 1967 to 2025, income at the 10th and 50th percentiles rose about 56%, while the 90th percentile rose about 121%, per the same Census report. The ratio between the top and bottom of the distribution widened from 9.23 to 13.06 over that 58-year window. That's a trend with a stated time frame, which is more than the "record high" line can say for itself.

None of this means the number is wrong. It means the single figure being repeated everywhere is doing work that a ±$1,040 interval and a 58-year trend line were built to caution against.