Jobless Claims 206K vs 205K Est — the Four-Week Average Has Now Risen Four Weeks Running

Fundamentals · 2026-09-03

Initial claims 206,000 for the week ending August 29, up 2,000 and roughly in line with consensus, with the prior week revised up 1,000 to 204,000; the 4-week average rose to 207,250, a fourth consecutive increase and 4.1 percent above its August 1 low of 199,000; unadjusted claims rose 30 where seasonal factors expected a fall of 1,226 — a 1,256 miss in the wrong direction, the mirror image of last week; insured unemployment 1,779,000, up 8,000, rate steady at 1.2 percent but the unadjusted rate fell to 1.1 percent; zero states posted an increase above 1,000 for a second straight week; work-sharing claims ticked up 82 but sit 39.1 percent below last year; claims remain 12.7 percent below the 236,000 of a year ago.

What Is This?

• What it is: The Labor Department's weekly count of unemployment insurance activity, released Thursdays at 8:30am ET. Initial claims count people filing after a separation from an employer — the cleanest high-frequency read on layoffs. Insured unemployment, or continued claims, counts weeks of benefits actually claimed and lags by a week. These are administrative counts from every state program office rather than a survey, so there is no sampling error, but weekly seasonal adjustment is difficult and the series is volatile.
• Why it matters: It is the last labor data before tomorrow's August payroll report, and it closes a week in which every other labor measure deteriorated: ISM manufacturing employment with a net of just +0.3, JOLTS showing hires down 278,000 with implied net employment of -18,000, and ADP private payrolls at +38,000 with education and health services accounting for more than the entire gain. Claims is the one series that has stayed benign. This release is about whether it still is.
• How to read it: The weekly print is noisy; the 4-week average is the trend. Read the unadjusted number against what the seasonal factors expected, because that gap is what actually drives the headline. And keep the distinction firmly in mind: claims measure firing, not hiring. A low number says people are not being let go. It says nothing about whether anyone is being taken on.

Want the full explainer? Economic News Events, Explained breaks down this release and every other one we cover, in plain English.

Summary

Initial claims came in at 206,000 for the week ending August 29, up 2,000 from a prior week revised up 1,000 to 204,000. Against a consensus of roughly 205,000 that is as close to a non-event as this series produces, and at 206,000 claims remain 12.7 percent below the 236,000 recorded in the comparable week of 2025. On the level, the layoff picture is still historically benign.

The trend is a different matter. The 4-week moving average rose to 207,250, its fourth consecutive increase. It bottomed at 199,000 in the week ending August 1 and has climbed 8,250, or 4.1 percent, in the four weeks since. The weekly prints have sat in a narrow 204,000-to-212,000 band for five weeks while the average has ground steadily higher, which is what a slow drift looks like before it becomes a trend.

That matters more than usual because of who cited it. Six days ago at Jackson Hole, Kevin Warsh listed "unemployment claims, on a four-week average" as an "empirically robust real-time indicator" and said it sat near its lowest level in decades. On the level he was right and remains right. On direction, that specific measure has now risen every week since, including the week he said it.

The mechanics underneath flipped hard this week, and this is the finding. Last week's release showed unadjusted claims falling 3,231 when the seasonal factors expected a decline of only 207 — the raw data came in dramatically better than the calendar said it should. This week unadjusted claims rose 30 when the factors expected a decrease of 1,226. That is a 1,256 miss in the wrong direction, and it is the reason the headline rose despite a raw number that barely moved. Two consecutive weeks, opposite signs, same series. The continued-claims side did the same thing more mildly: unadjusted insured unemployment fell 33,225 where the factors expected a fall of 40,733, so it also underperformed the seasonal norm by 7,508.

Insured unemployment rose 8,000 to 1,779,000 for the week ending August 22, though the prior week was revised down 7,000 first, so the level is lower than last week's published figure. The 4-week average here actually fell 5,000 to 1,781,750 — the two averages are moving in opposite directions, initial claims up and continued claims down. The seasonally adjusted insured unemployment rate held at 1.2 percent, but the unadjusted rate slipped to 1.1 percent, down a tenth, its first sub-1.2 reading in this run.

Year-over-year, every headline series remains substantially better than 2025. Initial claims are down 12.7 percent, the 4-week average down 10.1 percent, insured unemployment down 8.2 percent, and continued weeks claimed across all programs down 8.7 percent at 1,794,843 against 1,966,150. Nothing in this release is a distress signal on its own terms.

The state detail is the quietest it has been. For a second consecutive week no state posted an increase above 1,000, and this week only one state posted a decrease above that threshold — California at -1,049, with no comment attached. Last week four states cleared it on the downside. The distribution has compressed to the point where nothing anywhere in the country is moving enough to require an explanation. The largest increases were Illinois +631, New York +514, Texas +258, Michigan +229 and Massachusetts +173; the largest decreases California -1,049, Kentucky -583, New Jersey -519, Florida -475 and Minnesota -281. Michigan, which swung -2,446 a fortnight ago on manufacturing layoffs, is back to a trivial +229.

Two sub-populations are worth tracking. Work-sharing — short-time compensation, where employers cut hours across a workforce instead of laying people off — came in at 14,751, up 82 on the week after last week's 864 decline. On a base of 14,751 that uptick is noise. The annual comparison is not: work-sharing is 39.1 percent below the 24,226 of a year ago, a deeper decline than last week's 37 percent. Employers are still not reaching for the tool that precedes layoffs.

The federal picture continues to normalize, with former federal civilian employees filing 337 initial claims against 515 a year ago, down 34.6 percent, and their continued claims at 5,691 against 8,128, down 30.0 percent. The exception, again, is veterans. Newly discharged veterans filed 413 initial claims, up 28 on the week and up 15.0 percent against 359 a year ago, with continued claims also up slightly year over year. These are small and volatile numbers — the y/y comparison was negative last week and is positive this week — so no single reading means much, but it is the one sub-population that has not improved for several weeks running.

Put the week together and the split is now explicit. Firing remains rare: 206,000 claims, 12.7 percent below last year, a 1.0 percent layoff rate in Tuesday's JOLTS, and not one state with a meaningful increase. Hiring has fallen apart: JOLTS hires down 278,000 with implied net employment of -18,000, ADP at +38,000 with one sector accounting for more than the whole gain, and ISM manufacturing employment at a +0.3 net. The one thing that changed this week is that the firing side stopped improving. A 4.1 percent rise in the four-week average off a low is not deterioration in any meaningful sense — but it is the first time in this cycle that the layoff channel and the hiring channel have moved the same way at once.

The Internals

Initial claims, seasonally and non-seasonally adjusted:

Measure · Aug 29 · Aug 22 · Change · Aug 15 · Year ago

Initial claims, seasonally adjusted · 206,000 · 204,000 · +2,000 · 207,000 · 236,000

Initial claims, unadjusted · 170,626 · 170,596 · +30 · 173,017 · 196,712

4-week moving average, adjusted · 207,250 · 205,750 · +1,500 · 204,250 · 230,500

Insured unemployment, week ending August 22:

Measure · Aug 22 · Aug 15 · Change · Aug 8 · Year ago

Insured unemployment, adjusted · 1,779,000 · 1,771,000 · +8,000 · 1,796,000 · 1,937,000

Insured unemployment, unadjusted · 1,737,781 · 1,771,006 · -33,225 · 1,794,032 · 1,892,212

4-week moving average, adjusted · 1,781,750 · 1,786,750 · -5,000 · 1,788,250 · 1,944,500

Insured unemployment rate, adjusted · 1.2 percent · 1.2 percent · unchanged · 1.2 percent · 1.3 percent

Insured unemployment rate, unadjusted · 1.1 percent · 1.2 percent · -0.1 · 1.2 percent · 1.2 percent

What the seasonal factors expected against what happened. This is the reversal from last week:

Series · Actual change · Seasonal factors expected · Gap

Initial claims, unadjusted, week ending August 29 · +30, or 0.0 percent · -1,226, or -0.7 percent · 1,256 worse than expected

Initial claims, unadjusted, week ending August 22 · -3,231, or -1.9 percent · -207, or -0.1 percent · 3,024 better than expected

Insured unemployment, unadjusted, week ending August 22 · -33,225, or -1.9 percent · -40,733, or -2.3 percent · 7,508 worse than expected

The four-week average, which is the number that has actually changed:

Week ending · Initial claims, adjusted · 4-week average · Direction of the average

July 18 · 189,000 · 208,000 · Falling

July 25 · 198,000 · 203,250 · Falling

August 1 · 200,000 · 199,000 · The low

August 8 · 212,000 · 199,750 · Rising

August 15 · 207,000 · 204,250 · Rising

August 22 · 204,000 · 205,750 · Rising

August 29 · 206,000 · 207,250 · Rising, a fourth week

Four consecutive increases, 8,250 or 4.1 percent off the August 1 low, while the weekly prints have held a 204,000 to 212,000 band for five weeks.

Year over year, the level comparisons remain firmly better:

Measure · Current · Year ago · Change

Initial claims, adjusted · 206,000 · 236,000 · -30,000, or -12.7 percent

Initial claims, unadjusted · 170,626 · 196,712 · -26,086, or -13.3 percent

4-week average, adjusted · 207,250 · 230,500 · -23,250, or -10.1 percent

Insured unemployment, adjusted · 1,779,000 · 1,937,000 · -158,000, or -8.2 percent

Insured unemployment, unadjusted · 1,737,781 · 1,892,212 · -154,431, or -8.2 percent

Continued weeks, all programs · 1,794,843 · 1,966,150 · -171,307, or -8.7 percent

State And Program Detail

Continued weeks claimed across every program, week ending August 15:

Program · Aug 15 · Aug 8 · Change · Year ago

Regular state programs · 1,766,626 · 1,789,706 · -23,080 · 1,926,595

Short-time compensation and workshare · 14,751 · 14,669 · +82 · 24,226

Federal civilian employees, UCFE · 5,691 · 6,042 · -351 · 8,128

Newly discharged veterans, UCX · 4,814 · 4,798 · +16 · 4,718

State additional benefits · 2,931 · 2,912 · +19 · 2,443

Extended Benefits · 30 · 9 · +21 · 40

Total, all programs · 1,794,843 · 1,818,136 · -23,293 · 1,966,150

No state was triggered onto the Extended Benefits program, and the national caseload rose from nine claims to thirty — a number too small to carry meaning, but worth noting because it had been at single digits.

Initial claims in the federal programs, week ending August 22:

Program · Aug 22 · Aug 15 · Change · Year ago · Year-over-year

Former federal civilian employees, UCFE · 337 · 390 · -53 · 515 · -34.6 percent

Newly discharged veterans, UCX · 413 · 385 · +28 · 359 · +15.0 percent

Veterans are the only sub-population running above last year on either measure.

The largest state increases in initial claims, week ending August 22. None reached the 1,000 threshold that would require a comment:

State · Change

Illinois · +631

New York · +514

Texas · +258

Michigan · +229

Massachusetts · +173

The largest decreases, of which only one cleared 1,000:

State · Change · State-supplied comment

California · -1,049 · No comment

Kentucky · -583 · Below the threshold, no comment required

New Jersey · -519 · Below the threshold, no comment required

Florida · -475 · Below the threshold, no comment required

Minnesota · -281 · Below the threshold, no comment required

Last week four states posted decreases above 1,000 and none posted increases above it. This week one state cleared it on the downside and none on the upside. The distribution of state-level movement has narrowed to the point where the national number is being made by hundreds of small moves rather than any identifiable event.

Where insured unemployment is concentrated. The national rate is 1.2 percent adjusted; these states ran above it for the week ending August 15:

State or territory · Insured unemployment rate

New Jersey · 2.6 percent

Puerto Rico · 2.6 percent

Rhode Island · 2.2 percent

Massachusetts · 2.1 percent

Minnesota · 2.0 percent

Oregon · 2.0 percent

Washington · 1.9 percent

California · 1.8 percent

Connecticut · 1.8 percent

Nevada · 1.7 percent

New York · 1.7 percent

Pennsylvania · 1.7 percent

The composition is nearly identical to last week, with California and Connecticut easing a tenth and Minnesota slipping from 2.1 to 2.0. The insured unemployment rate is calculated against covered employment of 153,732,307.

Impact on USD

• Mixed, lean bearish — the level is benign but the four-week average has now risen four weeks running, and it lands on top of a week of deteriorating hiring data.
• Unadjusted claims rose 30 where seasonal factors expected a 1,226 fall, exactly reversing last week's beat — the raw data underperformed the calendar for the first time in a month.
• Against it: 206,000 is 12.7 percent below last year, insured unemployment is down 8.2 percent, and the unadjusted insured rate fell to 1.1 percent.
• Warsh cited the four-week average by name six days ago as evidence of full employment. The level still supports him; the direction no longer does.
• Second-order regardless: tomorrow's payroll report, with a +65,000 consensus, decides the week.

Impact on US Indices (ES / NQ / YM)

• Mixed — layoffs this rare protect household income, but the four-week drift plus this week's hiring collapse is an earnings signal.
• Nothing here changes the consumer's income base materially, which is the argument against reading it as a growth scare.
• The sector read is in the absence of movement: no state showed a meaningful layoff increase, so the manufacturing and professional-services weakness in JOLTS and ADP is showing up as *not hiring* rather than firing.
• Work-sharing down 39.1 percent year over year says employers still are not preparing to cut, which caps the downside from here.
• Tomorrow dominates: a soft payroll print against this benign a claims backdrop is the combination that would reprice the front end.

Impact on Gold

• Slight bullish — a labor market softening at the edges trims the hawkish case and pulls expected real yields marginally lower.
• The four-week average turning up while hiring measures fall is the early shape of the growth-scare channel the metal trades.
• Offsetting: layoffs at these levels are nowhere near recessionary, and the Fed's stated focus is prices rather than employment.
• Conditional: gold's read arrives tomorrow morning. This print on its own is not a catalyst.

What To Watch

• August employment situation — Friday, September 4, 8:30am ET. FactSet consensus is +65,000 including government. ADP said +38,000 private, JOLTS implied -18,000 for July, and ISM manufacturing employment netted +0.3.
• The four-week average. It has risen from 199,000 to 207,250 in four weeks. A fifth increase would make this a trend rather than a drift.
• Next weekly claims — Thursday, September 10, 8:30am ET. Watch whether the unadjusted series underperforms its seasonal expectation for a second week.
• August CPI — Friday, September 11. With ADP base pay at 3.2 percent, wages are not the inflation problem, which narrows what a hike would fix.
• FOMC decision and dot plot — Tuesday and Wednesday, September 15-16. Three officials dissented for a hike in July; this week's labor data is the strongest counterargument they have faced.
• Work-sharing and the veterans series. Both are small, but work-sharing is the leading indicator of employer distress and veterans are the one sub-population not improving.

TLDR

Initial Jobless Claims (week ending August 29, released September 3):

• Initial claims: 206,000 (vs ~205,000 est, 204,000 prior revised up from 203,000) — essentially in line, and 12.7 percent below the 236,000 of a year ago
• 4-week moving average: 207,250, a fourth consecutive increase, up 8,250 or 4.1 percent from its August 1 low of 199,000
• Unadjusted claims rose 30 where seasonal factors expected a fall of 1,226 — a 1,256 miss the wrong way, exactly reversing last week's 3,024 beat
• Insured unemployment: 1,779,000 for the week ending August 22, up 8,000, though the prior week was revised down 7,000 first
• Insured unemployment rate: 1.2 percent adjusted and unchanged, but the unadjusted rate fell to 1.1 percent, its first sub-1.2 reading in this run
• Zero states posted an increase above 1,000 for a second straight week, and only California cleared it on the downside at -1,049
• Work-sharing: 14,751, up 82 on the week but 39.1 percent below last year — a deeper annual decline than last week's 37 percent
• Former federal employees: 337 initial claims against 515 a year ago, down 34.6 percent; continued claims down 30.0 percent
• Newly discharged veterans: 413 initial claims, up 15.0 percent year over year — the only sub-population not improving
• Continued weeks claimed, all programs: 1,794,843, down 8.7 percent from 1,966,150 a year ago

Firing is still rare and hiring has fallen apart, and this is the first week the two moved together. The level says nothing is wrong: 206,000 claims, 12.7 percent below last year, not one state with a meaningful increase, and work-sharing 39 percent below 2025. The trend says something changed: the four-week average has risen four consecutive weeks off its August 1 low, and unadjusted claims came in 1,256 worse than the seasonal norm after beating it by 3,024 the week before. Six days ago Warsh cited that exact four-week average as evidence of full employment — the level still supports him, the direction no longer does. Set against JOLTS hires down 278,000, ADP at +38,000 with one sector carrying the whole gain, and ISM manufacturing employment at a +0.3 net, tomorrow's payroll report at +65,000 consensus carries more weight than any single print this year. Payrolls September 4, CPI September 11, FOMC September 15-16.

_For informational purposes only. Not investment advice._