The US labour share falls below 40%

Where things stand

BLS's revised release for Q2 2026 (Productivity and Costs, September 3, 2026) and the Economics Daily article of September 15, 2026 put the nonfarm business labour share at 52.8%. That is the lowest reading since the series began in 1947. The index PRS85006173 stood at 93.446 (2017=100), which I pulled from the BLS API. On that vintage, 40% corresponds to an index of about 70.8. The next releases are Q3 2026 preliminary on November 5, 2026 and revised on December 8, 2026, then one release every quarter. The last quarter that can be published inside the window is Q3 2031, due out in November or December 2031.

To resolve YES, the share has to fall 12.8 points: a 24% drop, or a log change of −0.278. Spread over 21 quarters, that is about −5.3% a year.

Base rate

I parsed all 318 quarterly percentages (Q1 1947 to Q2 2026) from the BLS Economics Daily table. The file is saved as /workspace/ls_pct.csv.

• The largest fall over any 21-quarter window was 5.5 points (log −0.090), from Q1 2001 to Q2 2006. The largest fall over four quarters was 2.9 points.
• The standard deviation of 21-quarter changes is 1.8 points.
• The required move is about three times the record and roughly seven standard deviations.
• Even if the past year's near-record pace continued (log change of −3.5% a year), the share would be about 43.9% by Q3 2031.

Under any historical pattern, the chance is effectively zero. The probability comes almost entirely from an unprecedented structural break.

Paths to YES

1. An explosive AI-driven boom: output soars while pay stagnates. Anthropic's September 2026 scenario paper (Korinek, Jones and others) has an "extreme" scenario: GDP growth of 15% a year by 2030, unemployment of 11.9%, and the labour share falling from 60 to 45.2%. That is a 25% relative drop, which would take the BLS measure to about 39.8% in early 2030, right at the threshold.
   • With very inelastic capital it falls to about 36%; with perfectly elastic capital, to about 43%.
   • Their "substantial" scenario only reaches about 49.4% on the BLS measure, and the median member of the public implies about 50%.
   • Some of the paper's reviewers called the extreme case a thought experiment.
   • A 2026 survey by the Forecasting Research Institute (Karger and others) asked directly about this BLS measure. Median forecasts for 2030 were 54.3% for economists with no condition on AI progress, and 52% assuming rapid AI progress. AI experts assuming rapid progress gave a median of 50% (middle half of answers: 45.4–52.6). Without conditioning, the middle half of all groups' answers sat between 53 and 55%.
   • So a share below 40% by 2030 sits in the far tails, even for AI experts.
2. A real-wage collapse driven by inflation (stagflation of the kind seen in Turkey). The US has no precedent for this. Inflation is 3.7% (September CPI), driven by an energy shock that might take another point or two off the share, not thirteen.
3. Benchmark or method changes. These have historically shifted the level by one or two points. A replacement series counts only if it keeps the all-worker definition.
4. A publication error. Negligible.

Paths to NO

NO is the default. Recessions usually raise the labour share because profits fall, so a "doom loop" of AI layoffs and falling demand pushes the wrong way. Part of the recent decline is cyclical: oil-driven price increases, and tariff revenue that inflates the output figure the share is divided by (Haver; Tax Foundation). Research from the New York Fed finds that the post-COVID decline follows earlier cyclical patterns.

Estimate

I put about 2–4% on an extreme-type transformation arriving by 2030–31. Given that, I put 40–70% on BLS reporting a quarter below 40% by Q3 2031. Adding a small allowance for other paths gives about 1.6% cumulative by the end of 2031. The yearly hazard is close to zero through 2027, because the share would have to fall about 25% within five quarters. It then rises through 2028–29 and flattens in 2030–31, the period when the extreme scenarios would cross 40%.

Strongest case against this estimate: AI capabilities are advancing fast (Kalshi priced OpenAI reaching AGI before 2030 at about 53% in March 2026), the share is falling at a near-record pace while profit margins hit records, and any single quarter or revision counts. If I adopted the tail views of AI experts, the estimate could be 3–5%. Economists' views imply less than 0.5%. I sit between the two.

Sources

• BLS Economics Daily, 9/15/2026: https://www.bls.gov/opub/ted/2026/labor-share-at-its-lowest-level-52-8-percent-in-second-quarter-2026.htm
• BLS Productivity and Costs, Q2 revised, 9/3/2026: https://www.bls.gov/news.release/prod2.nr0.htm; release schedule: https://www.bls.gov/schedule/news_release/prod2.htm
• BLS API, series PRS85006173
• Forecasting Research Institute (Karger et al. 2026), Table 39: https://forecastingresearch.org/research/economic-effects-of-ai
• Korinek et al. (Anthropic, September 2026), Tables 3 and 5: https://www.anthropic.com/institute/econ-scenarios
• Fortune, 9/4/2026; Tax Foundation, 9/2026; Haver Analytics; New York Fed Liberty Street Economics posts of 6/24/2026 and 10/15/2026; news coverage of the September 2026 CPI (10/14/2026).

Question. By the last day of each quarter, has BLS reported a U.S. nonfarm business labour share below 40% for any quarter?

Definitions.

• Labour share: BLS's seasonally adjusted, all-worker labour compensation share of current-dollar output in the nonfarm business sector, expressed as a percentage. Compensation includes employee pay and benefits and BLS's estimate of proprietors' labour compensation; this is not an employee-wages-only share.
• Percentage, not index: The threshold is 40% of output, not 40 index points. BLS series PRS85006173, “Nonfarm Business Sector: Labor Share for All Workers,” is published as an index (currently 2017=100), not a percentage of output. An index observation alone cannot meet the threshold. Conversion counts only when BLS supplies the percentage-level benchmark or conversion factor applicable to that same data vintage; no conversion using a rounded historical percentage from another vintage counts.
• Quarter: A calendar quarter, not an annual average or a rolling period. The Observation period convention applies; the earliest eligible observation is 2026 Q3.

Resolves YES if a qualifying BLS publication reports an eligible quarterly labour share strictly below 40%, or publishes data that establish that result using the conversion defined above. Initial, revised, and benchmark-revised estimates all count. Use the most precise value BLS publishes in that publication or its accompanying data; a value reported only as 40.0% does not qualify. No attribution to AI is required. Once a qualifying publication establishes YES, a later upward revision does not reverse it.

Resolves NO for a horizon if no qualifying publication has appeared by that horizon. Forecasts, hypothetical estimates, annual values, unadjusted observations, other sectors, and third-party estimates do not count. A revision covering only quarters ending before the opening date does not count.

Resolving source. U.S. Bureau of Labor Statistics publications and public data only: the Productivity and Costs release and accompanying Major Sector Productivity and Costs data, or BLS publications explicitly reporting the same percentage measure, such as The Economics Daily table “Labor share (percentage of output)” in the September 15, 2026 article. The related index is PRS85006173; the definition is documented in Estimating the U.S. labor share, Monthly Labor Review, February 24, 2017. An officially designated replacement series counts only if it preserves the nonfarm-business, all-worker compensation share of current-dollar output. This is a narrower resolving-source rule than Public confirmation (A2). Apply Resolution date (A3) and the Resolution window convention to the first qualifying BLS publication, not the quarter measured.