Americans work fewer than 2.75 hours a day on average

Where things stand

• Latest value. The newest ATUS release (2025 annual averages, published June 25, 2026) shows the target figure, Working and work-related activities, civilian population, Total, at 3.32 hours a day (Table 1; summary).
• That release doesn't count. It came out before the opening date, and its reference period ended before it too.
• Which releases can count. Only the annual results for 2026 through 2030 can qualify. BLS publishes them each late June (every release since 2006 has come out in June or July), so they arrive in mid-2027, 2028, 2029, 2030 and 2031. The 2031 results land after the window closes.
• What YES needs. A displayed value of 2.74 or less: a fall of 0.58 hours, about 17%, from a level that is already the lowest on record.

The history I computed

I downloaded the BLS LABSTAT files (download.bls.gov/pub/time.series/tu): the annual series TUU10101AA01011987, the quarterly series TUU10101QA01011987, and the series for employed and non-employed people. I paired them with CPS employment-to-population ratios from the BLS API (LNU02300000 and LNS12300000).

• The series. Annual values run 3.69 (2003), 3.81 (2007), 3.50 (2010), 3.61 (2019), 3.56 (2023), 3.43 (2024), 3.32 (2025). There is no 2020 annual value. Even the pandemic quarters bottomed at 3.12 (Q4 2020).
• Year-to-year changes. The standard deviation is 0.08. The largest fall was −0.20, in 2009.
• ATUS relative to the CPS jobs rate. For 2003–2023, ATUS hours divided by the CPS employment rate stayed between 5.83 and 6.11. It then fell to 5.71 (2024) and 5.56 (2025).
• Where the 2024–25 drop came from. Hours per employed person were flat at about 5.4 a day. What fell was the ATUS-implied employment rate. It ran about 2.5 points above the CPS rate in 2003–2022 (SD 0.45), was 4.5 points above in 2023, and only 0.6 points above in 2025.
• This looks like a survey effect. The CPS employment rate barely moved: 60.3% in 2023, 59.7% in 2025, and about 59.2% on average for January–September 2026. Meanwhile ATUS response fell from 36.9% to 32.4% to 25.7%, and only about 6,100 people were interviewed in 2025. ATUS weights adjust for age, sex, race, ethnicity, education and children, but not for employment status (User's Guide, Sept 2026; Technical Note). Part of the 2025 drop may also be noise from the 2025 government shutdown.
• The bar. With non-employed people averaging 0.11 hours and employed people 5.43, a value of 2.75 needs an ATUS-implied employment rate of about 49.6%. Even with drift in how the survey measures work, it would take something like a CPS employment rate near 50–54%. The record low for any annual average is 56.8%, in 2020.

Paths to YES

1. A severe AI-driven (or depression-scale) jobs collapse by 2029–30. Mainstream forecasts are modest:
   • Metaculus' Labor Automation Hub projects AI cutting employment by about 0.7–1.9% by 2030 (launch release).
   • An NBER study puts its central 2030 unemployment estimate at 5–6% (tech-insider, Sept 2026).
   • Kalshi's thin market on December 2030 unemployment above 10% is quoted at 9¢ bid / 10¢ ask (Kalshi API, October 2026). Even 10% unemployment would give an ATUS value of only about 2.9–3.0.
2. The 2024–25 slide continues. Falling response rates could keep pushing the survey's measured work time down. Extending the recent pace in a straight line would hit about 2.72 in the 2030 data. But the ratio was stable for 20 years, and BLS would likely notice and fix a widening gap with the CPS.
3. Combinations. Moderate drift plus a recession landing in 2029 or 2030.

Paths to NO

• The labor market is soft but not collapsing. September 2026 payrolls rose 29,000 and unemployment was 4.2% (CNN, Oct 2, 2026).
• ATUS continues and may stabilize. BLS plans an ATUS AI module for 2027–28 (Federal Register, July 10, 2026), and the House bill funds BLS at $716.5M (COSSA).
• Discontinuation also means NO. A small risk that the survey is cut, or stops publishing, removes the chance of a qualifying release.

Model and forecast

I ran a Monte Carlo model with these parts:

• the ATUS/CPS ratio starting near 5.58, with an uncertain persistent drift (−0.03 ± 0.06 a year) plus year-to-year shocks;
• sampling noise of 0.065 hours;
• the CPS employment rate trending down 0.25 points a year, a 15% yearly chance of a recession, a 5% chance of a severe AI collapse and a 20% chance of a moderate one;
• a 2% yearly chance that ATUS is discontinued.

The base case gives 5.8%. Plausible alternative settings range from 2.2% to 11.5%, and with no AI shock and no survey drift it is 0.4%. I shade the base case down slightly, to about 4.5% by the end of 2031, because a growing gap between ATUS and the CPS is something BLS would probably correct.

Almost all of that probability sits on the 2029 and 2030 data releases (June 2030 and June 2031). I assume about 90% of each release lands by June 30 and the rest in the third quarter. The 2026 and 2027 data are essentially ruled out, because the 2026 jobs rate is already known to be about 59%.

Strongest case against my estimate: the recent 0.12-hour-a-year decline, worsening response, and fat tails on AI disruption could make 10% or more defensible. I discount this because the bar needs an outcome with no historical precedent.

Question. By the last day of each quarter, has the American Time Use Survey reported that the civilian population aged 15 and over spent on average fewer than 2.75 hours a day on working and work-related activities?

Definitions.

• Reported average: the annual-average value in the U.S. Bureau of Labor Statistics (BLS) American Time Use Survey (ATUS) annual release, Table 1, “Time spent in primary activities and percent of the civilian population engaging in each activity, averages per day by sex,” row “Working and work-related activities,” column “Average hours per day, civilian population — Total.” A renamed or renumbered successor table counts only if it reports the same population, activity coverage, and annual all-days measure.
• Population and days: the ATUS civilian noninstitutional population aged 15 and over, both sexes combined, including employed and nonemployed people and those who did none of the activity on their diary day; the average covers all days of the week, not only workdays.
• Working and work-related activities: the published composite category, including working, job-related activities, other income-generating activities, job search, and related travel, as defined in the ATUS Technical Note. The narrower “Working” row is not the target measure.
• Fewer than 2.75 hours: the published table value is strictly below 2.75; a value of exactly 2.75 does not count. Use the table's displayed precision, not an inferred unrounded value.

Resolves YES if BLS publishes a qualifying annual-average value below 2.75 on or before the quarter-end. Initial estimates and officially revised estimates both count. Apply the Observation period convention: the annual reference period must end after the opening date. No finding that AI caused the decrease is required.

Resolves NO if no qualifying publication exists by the quarter-end. Subgroup averages, averages only for people who engaged in the activity, weekday-only or weekend-only averages, partial-year estimates, private calculations from ATUS microdata, forecasts, and confidence-interval bounds do not count. Revisions or republications solely for annual periods ending before the opening date do not count. A later upward revision does not undo a qualifying YES under the Resolution window convention.

Resolving source. BLS-authored ATUS annual releases, specifically Table 1, its archived release versions, and official BLS corrections or revisions to that table; definitions are supplied by the ATUS Technical Note. This is a narrower source requirement than Public confirmation (A2). Apply Resolution date (A3), the Resolution window, and the Observation period conventions; use the publication date of the qualifying release or revision, not the reference year's end.