Four-quarter US productivity growth falls below zero

Where things stand

• Latest release. BLS's Productivity and Costs release for Q2 2026 (revised, Sept 3, 2026) shows four-quarter nonfarm business productivity growth of +2.2% in Q2 2026. The prior quarters were +2.9% (Q1 2026) and +2.5%, +2.5%, +2.1% and +2.0% (2025). The last negative reading was Q4 2022 (−1.4%), and the last first-print negative was Q1 2023 (−0.9%).
• Hours vs. output. Hours worked are growing only about 0.2% a year (weak hiring, flat working-age population), while output is up about 2.5%.
• Next release. The first qualifying release is the Q3 2026 preliminary on Nov 5, 2026. The government is funded through Dec 11.
• Q3 2026. It will be strong. The Atlanta Fed's GDPNow estimate is 3.6% (Oct 15), payrolls grew only weakly (+29k in September), and BEA's Sept 30 annual update revised 2026 first-half GDP growth up (Q1 to 2.5%, Q2 to 2.2%).
• Arithmetic. Q3 2026 would need roughly a −3.7% annualized quarter to turn the four-quarter rate negative. Q4 2026 or Q1 2027 would need a sharp recession starting almost immediately. So the 2026–27 horizons are near zero.

Base rates I computed

Current data (BLS API, series PRS85006091, 1948–2026). 31 of 314 quarters are negative. Of the 21-quarter windows starting from 1985 on, 48% contain a negative. Excluding COVID-era windows (1985–2014 starts), the chance of a negative depends on the window's average growth (μ): 31% at μ=2.0% and 4% at μ=2.5%.

Real-time data (441 monthly ALFRED vintages of the OPHNFB index, 1990–2026). I only counted prints a quarter would actually show in Table 2, which covers about three calendar years. I checked this against the Sept 2, 2015 archived release, whose Table 2 shows 2013 Q1–Q3 at −0.2, −0.8 and −0.2, matching my values.

• Overall: 74% of 21-quarter windows starting 1990–2021 saw a negative print within the horizon; 68% of those starting from 1996.
• High-trend era: 0% for windows starting 1996Q1–2003Q3, but their lowest prints were only +0.06 to +0.67.
• Sensitivity to average growth: the lowest print rises only about 0.3–0.4 points for each point of average growth. Probabilities are about 50% at μ=2.0% and about 27% at μ=2.5%, with a wide spread.

Recessions. In modern recessions the lowest print was:

• 1990: negative in real time.
• 2001: +0.17%.
• 2008: −0.42%, but only after GDP revisions about 1.5 years later.
• 2020: no negative during the recession itself, then −2.55% during the 2021–22 normalization.

So I put the chance of a negative print given a recession at roughly 0.45–0.5.

Synthesis

• Recession route. Polymarket prices a recession by end-2026 at 8% and by end-2027 at 34%. I put the chance of a recession starting by mid-2031 at about 50%, times about 0.45, giving about 0.23.
• No recession. Slowdowns, trend reversion and revisions add about 0.2 × 0.5 ≈ 0.10. Non-recession negatives in 1993–95, 2006–07, 2013–16 and 2022 came with weak trends and hours growth of 1.5%+, which is unlikely now.
• Combined. The structural estimate is about 0.33–0.35; the statistical one is about 0.5. I blend them to about 0.43 by end-2031.
• Timing. Hazards are near zero through mid-2027, rising to about 3% a quarter from 2028 onward.

Strongest case the other way

The real-time base rate is around 70%. Even high-trend windows came within 0.1–0.2 points of zero. Strong productivity runs have usually been followed by slumps (2005–07, 2011–13, 2021–22). Revision noise is rising, recession risk is elevated (3.7% CPI inflation, possible Fed hikes, an energy shock, an AI capital-spending bust), and every print of every quarter counts. All of this argues for 50% or more. Against that, AI and very low labor-supply growth favor productivity, and modern firms cut hours quickly in downturns. That is why I stop at about 0.43.

Sources

• BLS release and Table 2 (Sept 3, 2026): bls.gov/news.release/prod2.nr0.htm and prod2.t02.htm
• Archived release, Sept 2, 2015: bls.gov/news.release/archives/prod2_09022015.pdf
• BLS API series PRS85006091 and PRS85006092
• ALFRED vintages: alfred.stlouisfed.org/graph/alfredgraph.csv?id=OPHNFB&vintage_date=…
• FRED series OUTNFB and HOANBS
• investingLive on GDPNow (Oct 15, 2026)
• CNBC September jobs report (Oct 2, 2026)
• Polymarket recession markets
• Deloitte US outlook (Sept 29, 2026)
• Econbrowser (Sept 30, 2026)
• Fox News on funding through Dec 11

Question. By the last day of each quarter, has BLS reported four-quarter nonfarm business productivity growth below 0% for any quarter?

Definitions.

• BLS: the U.S. Bureau of Labor Statistics.
• Nonfarm business productivity: seasonally adjusted labor productivity (real output per hour worked by all workers) in the U.S. nonfarm business sector, as reported in the BLS Productivity and Costs release, Table 2, “Nonfarm business sector: Labor productivity, hourly compensation, unit labor costs, and prices, seasonally adjusted.”
• Four-quarter growth: the “Labor productivity” column in the section “Percent change from corresponding quarter of previous year,” not the annualized change from the preceding quarter or the change between annual averages.
• Below 0%: a strictly negative published growth rate at BLS's displayed precision. A reported 0.0%, including −0.0%, does not count; an inferred negative value calculated from rounded index levels does not count.

Resolves YES if a qualifying BLS publication reports a strictly negative four-quarter rate for at least one quarter ending after the opening date. Under the Observation period convention, the earliest eligible observation is 2026 Q3. Preliminary estimates, regular revisions, and benchmark revisions all count. A qualifying negative publication establishes YES even if a later revision raises the rate to zero or above, under the Resolution window convention. No attribution to AI is required.

Resolves NO at a horizon if no qualifying publication has appeared by that horizon. Negative annualized quarter-on-quarter growth, negative annual-average growth, declines in other sectors or in multifactor productivity, and new publications repeating or revising observations for quarters ending before the opening date do not count.

Resolving source. BLS's Productivity and Costs release, specifically Table 2 and its archived release editions; the same explicitly reported four-quarter statistic in the release text or Table A1 also counts. A successor BLS release or table counts only if it reports the same sector, productivity concept, seasonal adjustment, and year-over-year comparison. BLS publications alone resolve this question; third-party estimates or reporting do not substitute. Apply the Resolution window and Observation period conventions and resolve on the qualifying publication date under A3, not the observation quarter's end.