Jobless Claims 197K vs 201K Est — Second Sub-200K Week, 4-Week Average Falls Again
Fundamentals · 2026-09-24
Initial jobless claims fell 1,000 to 197,000 in the week to 19 September, against a 201,000 consensus — a 4,000 beat and a second straight week below 200,000, with the prior week revised up 2,000 to 198,000. The 4-week average fell 1,750 to 202,250, its third straight decline. Continuing claims rose 2,000 to 1,719,000 against ~1,750,000 expected, with the prior week revised down 13,000 to 1,717,000, and their 4-week average fell to 1,744,000, the lowest in the 54 weeks the release tables. Insured unemployment rate 1.1%, unchanged. Against a year ago: initial claims -10.0%, the 4-week average -14.6%, continuing claims -10.3%. Unadjusted claims rose 10,243 when the seasonals expected 10,475, so this week's headline is clean. Kentucky (+1,041, manufacturing layoffs) was the only state up more than 1,000; Michigan reversed last week's jump with -2,169. No state triggered Extended Benefits.
What It Changes
- Nothing in the rate path, and that is the point. The market went into 08:30 pricing an October hike at roughly 73% on a 58.4 composite PMI and a 10-year at 5.1%; a sub-200,000 claims print removes the one input that could have argued for waiting.
- This week's headline is not a seasonal artifact, unlike last week's. Raw claims rose almost exactly as much as the seasonal factors expected — a 232-claim gap against last week's roughly 8,100 — so 197,000 is the labour market, not the Labor Day adjustment.
- The 4-week average is now the story. Three straight declines, 207,500 to 202,250, and it sits 3,250 above its 2026 low of 199,000 from the week to 1 August.
- Firing is still absent; hiring is still unmeasured here. This confirms "low-fire" for a third straight week and says nothing about the "low-hire" half, which sits in payrolls and JOLTS.
Impact
- USD — Slight bullish — another tight labour print supports the hike pricing that has carried DXY to a two-month high, though claims rarely move the dollar alone.
- DXY came into the print at 101.27 (TradingView, about 06:45 ET), up 0.57% on Wednesday alone, with USD/JPY 158.79 at its overnight high — the rate differential is doing the work, and this print keeps it where it is.
- Four data points this morning compete for the same rate read: Barkin at 08:00, Hammack at 08:50, Paulson at 10:10, and the 7-year auction at 13:00, which is the bigger test of the long end.
- US Indices (ES / NQ / YM / RTY) — Slight bearish — good labour news reads as confirmation of the hawkish path while the 10-year sits above 5.1%, and that pressure lands hardest on duration.
- NQ was down 0.94% against its settle before the print and RTY closed Wednesday -1.8%; a claims number that supports hikes does not help either, and RTY carries the most rate sensitivity of the four.
- Gold (GC) — Slight bearish — gold is trading as a rates instrument; a labour market with no slack keeps real yields supported, and GC was already below $4,300 before 08:30.
Inside The Number
Initial claims for the week ending 19 September came in at 197,000, down 1,000 from the previous week's revised 198,000, against a consensus of 201,000. The prior week was revised up from 196,000, so the headline decline is smaller than a straight comparison with last Thursday's print would suggest — but the level is lower than consensus and below 200,000 for a second straight week.
Where 197,000 sits in 2026. It is the third-lowest reading of the year, behind 189,000 (18 July) and 190,000 (25 April). Six of 2026's 38 weeks have printed below 200,000, and two of them are the last two. The 2026 average to date is 209,947. The comparable week of 2025 printed 219,000; today is 10.0% lower.
The seasonal mechanics are neutral this week, and that matters more than the headline. Unadjusted claims rose 10,243, or 6.7%, to 163,811. The seasonal factors expected 10,475, or 6.8%. The two are within 232 claims of each other. Last week the raw decline beat the seasonal expectation by roughly 8,100 claims, and that gap produced the sub-200,000 headline; this week there is no gap, so the second sub-200,000 print is the stronger evidence of the two. Unadjusted claims are 9.5% below the comparable 2025 week of 180,992.
The 4-week average is where the trend shows. It fell 1,750 to 202,250, with the prior average revised up 750 to 204,000. The sequence since the last week of August runs 207,500, 206,250, 204,000, 202,250 — three consecutive declines. Against the year-ago 236,750 it is 14.6% lower.
Continuing claims were flat, and the flat is at a low. Insured unemployment for the week ending 12 September rose 2,000 to 1,719,000 against a ~1,750,000 consensus — 31,000 below expectations. The prior week was revised down 13,000, from 1,730,000 to 1,717,000, which makes 1,717,000 the lowest reading in the 54 weeks the release tables and 1,719,000 the second-lowest. The 4-week average fell 13,000 to 1,744,000, also the lowest in that window. Revisions moving down for a second week in a row is a coherent picture, not a noisy one.
The insured unemployment rate held at 1.1% on covered employment of 153,732,307, against 1.3% a year ago. It has now printed 1.1% for three consecutive weeks after spending most of 2026 at 1.2%.
The all-programs total fell hard, and that one wants caution. Continued weeks claimed across every program fell 102,149 to 1,588,612 for the week ending 5 September, almost all of it in regular state programs (-102,629). That is unadjusted data for the week before Labor Day, and the seasonally adjusted insured count for the same week fell 48,000 — so a large part of the raw drop is calendar. Against a year ago the total is 11.3% lower.
Against the Fed. The Committee hiked last week on the premise that the labour side of its mandate is "in good shape", and New York's Williams said this morning at about 06:00 ET that the labour market is "broadly in balance" while calling another hike by year-end "reasonable". Two more weeks of claims data now agree with both. Nothing in this release constrains an October move.
The Internals
Initial claims, seasonally adjusted unless marked, week ending 19 September:
Measure · This week · Prior week · Change · Year ago · vs year ago
Initial claims, adjusted · 197,000 · 198,000 · -1,000 · 219,000 · -10.0%
Initial claims, not adjusted · 163,811 · 153,568 · +10,243 · 180,992 · -9.5%
4-week moving average, adjusted · 202,250 · 204,000 · -1,750 · 236,750 · -14.6%
Consensus for the headline · 201,000 · n/a · Actual beat by 4,000 · n/a · n/a
Prior week as first reported · 196,000 · n/a · Revised up 2,000 · n/a · n/a
Continuing claims and the insured rate, week ending 12 September:
Measure · This week · Prior week · Change · Year ago · vs year ago
Insured unemployment, adjusted · 1,719,000 · 1,717,000 · +2,000 · 1,916,000 · -10.3%
Insured unemployment, not adjusted · 1,552,283 · 1,565,654 · -13,371 · 1,718,818 · -9.7%
4-week moving average, adjusted · 1,744,000 · 1,757,000 · -13,000 · 1,926,250 · -9.5%
Insured unemployment rate, adjusted · 1.1% · 1.1% · 0.0pp · 1.3% · -0.2pp
Insured unemployment rate, not adjusted · 1.0% · 1.0% · 0.0pp · 1.1% · -0.1pp
Consensus for continuing claims · ~1,750,000 · n/a · Actual below by 31,000 · n/a · n/a
Prior week as first reported · 1,730,000 · n/a · Revised down 13,000 · n/a · n/a
The seasonal adjustment, this week against last:
Measure · This week · Last week
Actual unadjusted change · +10,243 · -24,630
Change the seasonals expected · +10,475 · -16,515
Gap · 232 fewer than expected · About 8,100 more decline than expected
Read · Headline is clean · Headline was flattered by the adjustment
The 4-week average, recent run:
Week ending · Initial claims · 4-week average
22 August · 204,000 · 205,750
29 August · 207,000 · 207,500
5 September · 207,000 · 206,250
12 September · 198,000 · 204,000
19 September · 197,000 · 202,250
Where the levels sit within 2026:
Reading · Value · Context
Initial claims · 197,000 · Third lowest of 38 weeks in 2026, behind 189,000 and 190,000
Weeks below 200,000 in 2026 · 6 · The last two are two of them
2026 average to date · 209,947 · Against 219,000 in the comparable 2025 week
2026 low for the 4-week average · 199,000 · Week to 1 August; now 3,250 above it
Insured unemployment · 1,719,000 · Second lowest in the 54 weeks tabled; lowest is 1,717,000 a week earlier
Insured unemployment 4-week average · 1,744,000 · Lowest in the 54 weeks tabled
All programs, continued weeks claimed for the week ending 5 September, not adjusted:
Program · This week · Prior week · Change · Year ago
Regular state programs · 1,561,544 · 1,664,173 · -102,629 · 1,753,149
Short-time compensation and worksharing · 13,897 · 13,396 · +501 · 22,046
Former federal civilian employees · 5,105 · 5,444 · -339 · 8,168
Newly discharged veterans · 4,975 · 4,715 · +260 · 4,495
State additional benefits · 3,073 · 3,018 · +55 · 2,596
Extended benefits · 18 · 15 · +3 · 14
Total, all programs · 1,588,612 · 1,690,761 · -102,149 · 1,790,468
State And Program Detail
The state detail covers the week ending 12 September, one week behind the headline. That is the week that printed 198,000 nationally.
Only one state rose by more than 1,000: Kentucky, +1,041, which the state attributed to "layoffs in manufacturing industry." The rest of the largest increases were small — Hawaii (+213), Arkansas (+201), South Carolina (+147) and Massachusetts (+114).
Michigan is the answer to last week's open question. It led the country with +2,075 in the week to 5 September, which this desk flagged against the auto inventory build. This week it fell 2,169, with the state citing "fewer layoffs in manufacturing industry." One week up, one week down by the same amount: that is a distortion unwinding, not the inventory overhang reaching the labour data. Kentucky is now the manufacturing state to watch, and one week of +1,041 is not a trend.
The largest decreases were California (-4,809), Texas (-2,948), New York (-2,341), Michigan (-2,169) and New Jersey (-1,597), followed by Illinois (-1,563), Connecticut (-1,234) and Pennsylvania (-1,199). New York cited fewer layoffs in transportation and warehousing, health care and social assistance, and information; Pennsylvania cited manufacturing, administrative and waste services, education and health care.
The advance state data for the headline week has one outlier. Unadjusted initial claims in the week to 19 September rose in most states, as the seasonals expected, led by California (+2,522) and Hawaii, which more than doubled from 1,382 to 2,922 (+1,540). The release carries no state comment for advance figures, so there is no stated cause yet; next week's release will say whether Hawaii was a one-week event.
The insured unemployment rate map, for the week ending 5 September: New Jersey 2.3% highest, then Massachusetts 1.9%, Puerto Rico 1.9%, Washington 1.8%, California 1.7%, Nevada 1.7%, Oregon 1.7%, New York 1.6% and Rhode Island 1.6%. A week earlier New Jersey and Puerto Rico were both at 2.6%, so the top of the map came down.
Federal programs are below year-ago levels on both lines now. Former federal civilian employees filed 362 initial claims, down 38 on the week and against 635 a year ago. Newly discharged veterans filed 358, down 140 and now below the year-ago 420 — last week this was the one federal line running above its year-ago level, and that has reversed.
Extended Benefits stay off. No state triggered "on" for the week ending 5 September, with 18 continued weeks claimed nationally against 14 a year ago.
Short-time compensation is 37% below a year ago. Worksharing claims rose 501 on the week to 13,897, against 22,046 a year earlier. Firms are neither cutting staff nor cutting hours in any volume.
Against This Morning's Open
- The Open had the numbers right and the prior stale — The Open tabled claims at a 201K consensus against a 196K prior, and continuing claims at 1.750M against 1.730M; both priors were revised this morning, to 198K and 1.717M.
- It resolved the way The Open said it would lean. The Open wrote that a 201K print "would be another low number from a labour market that Williams just called balanced, which adds weight to the hike case" — it printed 4,000 lower than that.
- The Open's fulcrum call stands. It said claims were "the only print likely to move anything" before the 13:00 7-year auction and the Xi meeting, which remain the day's two tests; a 4,000 beat does not change that ranking.
What This Sets Up
- Next claims — Thursday 1 October, covering the week to 26 September, the first reading after this one with no seasonal distortion in either direction.
- Whether the 4-week average makes a new 2026 low. It needs to fall below 199,000; with 207,000 dropping out of the average next week, a print near this week's level would take it to about 199,750.
- The payroll survey week is already in. The week to 12 September is the September payroll reference week, and claims printed 198,000 in it against 207,000 in the August reference week (week to 15 August), which says firing did not rise into the survey.
- Whether Kentucky repeats. It is the only state to report manufacturing layoffs above 1,000 this week; a second week would matter more than the first.
What Is This?
- What it is: The Unemployment Insurance Weekly Claims report, published by the Department of Labor every Thursday at 08:30 ET. Initial claims count new filings for unemployment benefits in the week ending the previous Saturday — people who have just lost a job — and are the most current labour indicator the US produces. Continuing claims (insured unemployment) count people still receiving benefits and run one week further behind. The insured unemployment rate divides continuing claims by covered employment, currently 153.7 million. The report also carries state detail, federal-employee and veteran programs, and Extended Benefits status.
- Why it matters: It is the highest-frequency hard labour data there is, arriving weekly with a five-day lag when payrolls arrive monthly with a three-week one, so it is where a deterioration would show first. Right now it sits at the centre of the policy argument: the Fed hiked on the premise that the labour market can absorb it, and markets are pricing another hike in October, and claims is the series that would show that premise failing first.
- How to read it: It measures firing, not hiring — a low number means few people are being let go, not that the labour market is healthy. The weekly number is volatile and the 4-week average is the signal, especially around holidays. Compare the unadjusted change with what the seasonals expected — when the two diverge, the adjustment is producing the headline. State detail lags the headline by one week. And the raw count is not comparable across decades, because covered employment has grown; the insured unemployment rate is the better historical measure.
Want the full explainer? Economic News Events, Explained breaks down this release and every other one we cover, in plain English.
_For informational purposes only. Not investment advice._