The headline looks calm and the composition does not. US unemployment was 4.2% in September 2026, against 3.6% on average in 2023. It has eased this year partly because fewer people are in the labour force.
What sits under the headline, to September 2026:
- Longer searches. The median spell out of work is 11.5 weeks, up from about 9 weeks on average in 2023. About 27 per cent of the unemployed have been out for six months or more.
- Low hiring. About 4 in 10 people on unemployment benefit now reach the end of it, against 1 in 3 in 2019. Layoffs are still low, so this is a hiring problem.
- Pay. Hourly pay has trailed prices since April 2026, mostly because of energy. The saving rate is down to 4.1 per cent (August, after BEA's revision of 30 September).
- Few new jobs. September added 29,000 and the last three months averaged 51,000. How many the economy now needs is its own question: how many jobs the US needs each month.
How this started. The June 2026 analysis found long-term joblessness (27+ weeks) had nearly doubled since 2023 to 2.0 million people, with the long-term share back to 27.5%. The share was 17.8% just after Covid and peaked at 45.5% in the financial crisis. That is the slow-grind shape of 2008, this time with no recession to explain it.
Before comparing any two of these numbers, know which survey each came from. Payrolls count jobs and the household survey counts people, and both carry error bars wide enough to swallow a normal month: why the two US jobs surveys disagree.
Track the BLS/FRED series live: https://www.tigzig.com/tremor. Analysis: the September 2026 update https://www.tigzig.com/post/us-jobs-september-2026-update-oct2026, the 24 September deck https://www.tigzig.com/post/us-jobs-calm-before-storm-sep2026 and the June analysis https://www.tigzig.com/post/us-unemployment-long-term-hidden-jun2026.
Building something like this? How I work covers the rates, the availability and what I take on.