The unemployment rate has six versions. Here is August 2026 in each
The headline 4.1% is U-3. The same BLS release also put the rate at 1.8%, 1.9%, 4.4%, 5.1% and 7.7%. Who each version counts, and what the gaps have meant.
- What happened
- The Bureau of Labor Statistics reported on September 4, 2026 that the August 2026 unemployment rate was 4.1%. Table A-15 of the same release listed five other measures of the same labour market, from 1.8% (U-1) to 7.7% (U-6), all seasonally adjusted.
- Why
- Each measure asks a differently worded question of the same 60,000-household survey. U-3 counts only people with no job who looked for one in the prior four weeks. U-6 adds people who want work but have stopped looking and people working part time who want full-time hours.
- The numbers that matter
- 4.1%U-3, the official rate, August 20267.7%U-6, the broadest measure, August 20263.6 ptsU-6 minus U-3, August 2026; 3.8 pts in August 20254,390,000Employed part time for economic reasons, August 2026
- Why it matters
- Which measure you read changes whether the labour market looks tight or slack, and the broad measures move first when employers cut hours before they cut jobs.
- What happens nextas we saw it on 28 Sep 2026
- The Employment Situation for September 2026, with a new Table A-15, is scheduled for Friday, October 2, 2026 at 8:30 a.m. ET.
The official rate sits in the middle of the range; the two broader measures add people the headline does not count.
| U-1 Unemployed 15 weeks or longer | 1.8% |
|---|---|
| U-2 Job losers and completed temporary jobs | 1.9% |
| U-3 Official unemployment rate | 4.1% |
| U-4 U-3 plus discouraged workers | 4.4% |
| U-5 U-4 plus other marginally attached | 5.1% |
| U-6 U-5 plus part time for economic reasons | 7.7% |
On September 4, 2026, the Bureau of Labor Statistics reported that the unemployment rate for August 2026 was 4.1%. The same release, in a table most coverage never reaches, also put it at 1.8%, 1.9%, 4.4%, 5.1% and 7.7%. None of these is a correction. They are six answers to six differently worded questions, published every month in Table A-15 of the Employment Situation as “alternative measures of labor underutilization”.
The headline is the third of the six, U-3. The narrowest, U-1, counts only people unemployed for 15 weeks or longer. The broadest, U-6, adds people who want a job but have stopped looking and people working part time because they cannot get full-time hours. In August 2026 the distance from narrowest to broadest was 5.9 percentage points.
The six measures, narrow to broad
The measures nest. Each one from U-3 upward adds a group to the top of the fraction and, where that group is outside the labour force, to the bottom as well. That is why U-4, U-5 and U-6 are shares of the labour force plus the added groups.
| Measure | Who is counted | Aug 2025 | Aug 2026 |
|---|---|---|---|
| U-1 | People unemployed 15 weeks or longer, as a percent of the civilian labour force | 1.7% | 1.8% |
| U-2 | Job losers and people who completed temporary jobs, as a percent of the civilian labour force | 2.0% | 1.9% |
| U-3 | Total unemployed, as a percent of the civilian labour force (the official rate) | 4.3% | 4.1% |
| U-4 | Total unemployed plus discouraged workers, as a percent of the labour force plus discouraged workers | 4.6% | 4.4% |
| U-5 | U-4 plus all other people marginally attached, as a percent of the labour force plus all marginally attached | 5.3% | 5.1% |
| U-6 | U-5 plus all people employed part time for economic reasons, same denominator as U-5 | 8.1% | 7.7% |
Seasonally adjusted, U-3 fell from 4.3% to 4.1%, U-6 from 8.1% to 7.7%, and U-2, U-4 and U-5 each fell by 0.1 or 0.2 points. U-1, the share of the labour force unemployed for 15 weeks or longer, edged up from 1.7% to 1.8%. The same table's not-seasonally-adjusted columns show the same pattern: U-3 at 4.5% in August 2025 and 4.3% in August 2026.
August 2025: 8.1% − 4.3% = 3.8 percentage points. August 2026: 7.7% − 4.1% = 3.6 percentage points. Of the 3.6-point gap in August 2026, 0.3 points came from adding discouraged workers (U-4 − U-3), 0.7 points from adding the other marginally attached (U-5 − U-4), and 2.6 points from adding people working part time for economic reasons (U-6 − U-5). The part-time group is the bulk of the difference between the headline and the broad measure.
Who is counted where
The BLS explainer “How the Government Measures Unemployment” sets out all three conditions. Active search means contacting employers or agencies, sending resumes or answering advertisements; reading job postings does not count. Workers expecting recall from temporary layoff are the one exception. People with no job who are not looking are “not in the labor force”, outside U-1, U-2 and U-3 entirely.
That is the footnote definition in Table A-15. Discouraged workers are the subset who give a job-market reason for not looking: no work available, could not find any, or lack the schooling or training. U-4 adds the discouraged; U-5 adds the rest. Table A-16, which is not seasonally adjusted, counted 1,655,000 marginally attached people in August 2026 (1,791,000 in August 2025), of whom 411,000 were discouraged (492,000 a year earlier).
Table A-8 defines the group as people who worked 1 to 34 hours in the reference week for an economic reason: slack work, unfavourable business conditions, inability to find full-time work, or seasonal declines in demand. They are employed, so they are not in U-3. Only U-6 counts them. Seasonally adjusted, there were 4,390,000 in August 2026, down 414,000 from July 2026 and down 365,000 from 4,755,000 in August 2025. Of the August 2026 total, 2,815,000 cited slack work or business conditions and 1,263,000 could only find part-time work.
Numerator: 7,031,000 unemployed + 1,655,000 marginally attached + 4,390,000 part time for economic reasons = 13,076,000. Denominator: 169,777,000 labour force + 1,655,000 = 171,432,000. Ratio: 7.6%, against the published 7.7%. The gap is rounding plus one inconsistency in our inputs: the BLS publishes the marginally attached count only without seasonal adjustment. This is a check on the arithmetic, not a replacement for the published figure.
Where the numbers come from
All six measures come from one survey, and it is not the one that produces the payroll number. The unemployment rate comes from the Current Population Survey, the household survey. The 162,000 payroll gain for August 2026 comes from the Current Employment Statistics survey, the establishment survey. They cover different populations and can disagree in a given month without either being wrong.
The BLS explainer says the survey has run every month since 1940, that Census Bureau interviewers contact about 60,000 eligible households, and that this covers approximately 110,000 individuals. Questions refer to the reference week, usually the week that includes the 12th. The population is everyone aged 16 and over in the civilian noninstitutional population: no upper age limit, but people in prisons, nursing homes and other institutions, and those on active military duty, are excluded. Table A-1 puts it at 275,415,000 in August 2026, of whom 169,777,000 were in the labour force and 7,031,000 unemployed.
7,031,000 ÷ 169,777,000 = 4.14%, published as 4.1%. Participation is the labour force over the population: 169,777,000 ÷ 275,415,000 = 61.6%, up from 61.4% in July 2026. The employment-population ratio is the employed over the population: 162,746,000 ÷ 275,415,000 = 59.1%. In August 2025 the two stood at 62.3% and 59.6%; our piece on the 61.6% participation rate covers that decline.
The Technical Note says the establishment survey covers about 119,000 businesses and government agencies, representing approximately 622,000 worksites, for the pay period including the 12th. It counts a person with two payroll jobs twice; the household survey counts each person once, and also includes the unincorporated self-employed, unpaid family workers, agricultural workers, private household workers and people on unpaid leave, none of whom appear in the payroll count.
The Technical Note gives a 90-percent confidence interval for the monthly change in nonfarm employment of plus or minus 122,000 and, at an unemployment rate of around 6.0 percent, about ±0.3 percentage point for the monthly change in the unemployment rate and ±425,000 for the change in the number unemployed. A one-month move of 0.1 point in any of the six measures is inside that noise, which is why the table above shows years, not months; see one month is not a trend.
What the U-3 to U-6 spread has meant
U-6 adds employed people whose hours have been cut and jobless people who have stopped searching. Employers tend to trim hours before headcount, and job seekers give up after the search has already gone badly, so both groups can grow while U-3 is flat. In August 2026 the largest added group, part time for economic reasons, fell by 414,000 on the month and 365,000 on the year, which is why the spread narrowed while the headline rate held at 4.1%. This is the mechanism, not where any measure goes from here.
The BLS explainer records that some have argued the official measure is too restricted to capture the breadth of labour market problems, and that the alternative measures were developed in response. But U-6 folds together people with no work and people working 34 hours a week. Each measure is a precise answer to a precise question. Know which question your number answers, and compare it only with the same measure in earlier months.
What it means for your money
For a household, Table A-15 is a check on the headline. Losing hours is closer to what U-6 measures than to what U-3 measures, because U-3 does not count hours at all. In August 2026, 4,390,000 people were working part time when they wanted full-time hours: 2.7% of the 162,746,000 employed, and none of them counted as unemployed.
The arithmetic is plain. A 40-hour job cut to 30 hours is a 25% cut in pay with no change in the unemployment rate. Table A-8 shows 2,815,000 involuntary part-timers cited slack work or business conditions, the category that grows when employers shorten weeks. That is one reason an emergency fund is sized in months of spending rather than around the single event of losing a job: the more common shock is a smaller paycheck, not no paycheck.
Because all six measures share one survey and its error bars, none is informative on its own in a single month. The U-6 to U-3 spread is a yearly gauge, and in the twelve months to August 2026 it narrowed by 0.2 points.
Sources
- Table A-15. Alternative measures of labor underutilization, The Employment Situation, August 2026Bureau of Labor Statistics · Sep 28, 2026 · Primary source
- The Employment Situation — August 2026U.S. Bureau of Labor Statistics · Sep 4, 2026 · Primary source
- How the Government Measures UnemploymentBureau of Labor Statistics · Sep 28, 2026 · Primary source
- Employment Situation Technical Note, August 2026Bureau of Labor Statistics · Sep 28, 2026 · Primary source
- Table A-1. Employment status of the civilian population by sex and age, The Employment Situation, August 2026Bureau of Labor Statistics · Sep 28, 2026 · Primary source
- Table A-8. Employed people by class of worker and part-time status, The Employment Situation, August 2026Bureau of Labor Statistics · Sep 28, 2026 · Primary source
- Table A-16. People not in the labor force and multiple jobholders by sex, not seasonally adjusted, The Employment Situation, August 2026Bureau of Labor Statistics · Sep 28, 2026 · Primary source
- Schedule of release dates for principal economic indicatorsU.S. Bureau of Labor Statistics · 2026 · Primary source
Related analysis
The unemployment rate held at 4.1%. So did the reason to check the other number.
Labor force participation was 61.6% in August, half a percentage point lower than in January. People who stop looking for work stop being counted as unemployed.
August added 162,000 jobs. The average of the prior twelve months was 31,000.
The strongest hiring month in a year sits on top of a twelve-month average five times smaller. Both numbers are true, and only one of them is a trend.
July's jobs number went from −23,000 to +21,000. That is not a rounding error.
The Bureau of Labor Statistics revised two months of hiring up by a combined 55,000. One of them changed sign, and it is the figure this publication built an article around.
How to check whether a financial number you just read is still true
Every economic figure has two dates, a publisher, a revision history and an error bar. Four of those five are usually missing from the article you read it in. Here is how to find them in about a minute.