The US labour share falls below 50%

Where things stand

• Latest reading. BLS's revised Q2‑2026 Productivity and Costs release (Sept 3, 2026) puts the nonfarm business labor share at 52.8%, the lowest since 1947. The Sept 15 TED article confirms this. The next release, the Q3‑2026 preliminary, is due Nov 5, 2026. A continuing resolution funds the government to Dec 11, so a shutdown is unlikely to delay it.
• How the measure works. Checked against the BLS API: the labor share index PRS85006173 is exactly unit labor costs divided by the implicit price deflator. The percentage equals the index × 0.56505 (that factor is pinned down by the one‑decimal TED values). To read below 50% at one decimal (49.9 or less), the index must fall under about 88.4, versus 93.45 now.
• A likely upward revision first. BEA's annual update (Sept 30, 2026) revised Q1‑2026 compensation up using Quarterly Census of Employment and Wages (QCEW) data. Comparing ALFRED vintages from before (2026‑09‑15) and after the update:
  • total compensation for Q2‑2026 is up 0.77%;
  • private wages are up 1.11%;
  • nonfarm business value added is up only 0.24%.
  • So the Q2‑2026 labor share should be revised up about 0.7%, to roughly 53.2%.
• Q3‑2026 nowcast. Compensation is growing about 3.3% annualized (July–August BEA data), against nominal output growth of about 6–7% (Atlanta Fed GDPNow 3.6% real growth plus moderate prices). That implies Q3‑2026 at about 52.7% ± 0.45. Falling below 50% from there needs a further decline of about 5.4% (log).

Base rate

I computed this from the BLS series for 1947–2026.

• How often the share fell at least 5.5% within the next 20 quarters: 8.7% of all starting quarters, 17% of starts since 1990, and only 6% when starting near a record low.
• Long‑run drift: about −0.45%/yr since 1990 and −0.68%/yr since 2000. More recently it has been faster: −0.8%/yr since 2019 and about −1.15%/yr since 2022 (after revision).
• Two statistical models fitted to the series (a random walk with drift, and a smooth trend plus a persistent cycle) behave very differently at five years.

The key additional path: BEA counting firms' in-house data as investment

• The plan. BEA's director wrote on Apr 20, 2026 that BEA plans to start treating companies' in‑house data production as investment ("capitalizing" it) in the 2028 comprehensive update of GDP.
• The size. BEA's experimental estimates (working paper WP2026‑4) put this investment at $875B in 2024, of which $709B is private. The for‑profit part is about 3% of nonfarm business output.
• Why it lowers the share. It adds output but no compensation, so the labor share falls by roughly that percentage.
• Precedent. The 2013 change that counted R&D as investment did the same thing. Elsby, Hobijn and Şahin reported a 2010–12 average of 58.3% before the revision; today's vintage shows about 56.2%.
• Timing. BEA's recent updates have come out in late September, and BLS adopted the 2023 comprehensive update in its next release. So the drop would most likely show up in the November 2028 BLS release, which revises every quarter including 2026Q3 onward.

I assumed about a 58% chance this reaches BLS data by late 2028 and about 76% by 2031, with a size of 2.3% ± 0.9% (allowing BEA to narrow the scope).

Simulation

I ran a Monte Carlo that mixes the two models fitted to the 1947–2026 series, in roughly 40/60 proportions.

• Trend scenarios: rising or flat 34%, −0.5%/yr 37%, −1.1%/yr 19%, −1.8%/yr 10% (the steep ones represent faster AI‑driven decline).
• It includes uncertainty about the starting level and noise between published versions of the same quarter (preliminary and revised estimates both count).
• It applies the data‑capitalization shift to all published quarters when that event occurs.

Results:

• Without the data change: about 30% by end‑2031.
• With it: about 0.7% by end‑2027, 19% by end‑2028, 50% by end‑2031.

I shade these down a little for model risk.

Paths and risks

• Toward YES: falling below 50% before 2028 needs an unusually fast decline, such as AI‑driven productivity with lagging wages or continued stagflation in which prices outpace pay. From late 2028, the BEA change plus even a modest trend makes a reading below 50% plausible.
• Strongest case for NO: declines this large from a record low are historically rare (6%). Some of the 2025–26 drop came from tariffs (customs duties equal to about 1.2–1.5% of output) and the oil shock from the Iran war, and both could unwind. The Fed is raising rates, so a 2027–28 recession that temporarily lifts the share is plausible. BEA could also delay or shrink the data change; that alone would cut the 2031 probability to roughly 30%.

Sources

• BLS release: https://www.bls.gov/news.release/prod2.htm (Sept 3, 2026)
• BLS TED article (Sept 15, 2026)
• BLS API, series PRS85006173, PRS85006113, PRS85006143
• ALFRED vintages of PRS85006173, COE, GDP, A132RC1, A358RC1Q027SBEA
• BEA NipaDataQ/M flat files
• BEA annual‑update article (SCB, Sept 2026) and Q2 third‑estimate release
• BEA blog on data as an asset (Apr 20, 2026)
• BEA working papers WP2026‑4 (with experimental‑estimates XLSX) and WP2026‑7
• Elsby, Hobijn and Şahin (FRBSF working paper 2013‑27)
• GDPNow (Oct 8, 2026); news coverage of the Fed hike, September CPI and the September jobs report

Question. By the last day of each quarter, has BLS reported a nonfarm business labour share below 50% for any quarter?

Definitions.

• BLS: the U.S. Bureau of Labor Statistics.
• Nonfarm business labour share: BLS's seasonally adjusted quarterly measure of labour compensation as a percentage of current-dollar output in the U.S. nonfarm business sector, including BLS's allowance for the labour compensation of self-employed workers. The threshold concerns the percentage level, not a percentage change or a normalized index. In particular, BLS series PRS85006173, “Nonfarm Business Sector: Labor Share for All Workers,” is an index (currently 2017=100), not itself the percentage to compare with 50.
• Quarter: a calendar quarter. The Observation period convention applies: only quarters ending after the opening date qualify, beginning with 2026 Q3.

Resolves YES if a qualifying BLS publication reports a labour-share level strictly below 50% for at least one eligible quarter. Preliminary estimates and subsequent revisions both count. Use the most precise percentage level BLS publishes in that release or its accompanying data; a displayed value of 50.0% alone does not qualify. A later revision above the threshold does not undo a YES. No attribution to AI is required.

Resolves NO for a horizon if no qualifying publication has appeared by that horizon. Annual averages, unadjusted observations, other sectors, forecasts, private estimates, and reports or revisions solely concerning quarters ending before the opening date do not count. An index value below 50 does not count, and no percentage level may be inferred by anchoring the index to a rounded historical percentage.

Resolving source. BLS publications only: the quarterly Productivity and Costs preliminary and revised releases (release), their accompanying major-sector historical data workbook, or another BLS publication explicitly reporting the same quarterly percentage measure, including The Economics Daily. The measure is illustrated in BLS's September 15, 2026 article and table. This is a narrower source requirement than Public confirmation (A2); third-party reporting alone cannot resolve it. Apply the Resolution window convention and Resolution date (A3): resolution is dated to the qualifying BLS publication or dated data release, not the quarter measured. An officially designated successor series counts only if it retains the nonfarm-business scope and compensation-share-of-output concept.