US labor share

Q7 forecast — September 14, 2026

Central forecast: the preliminary-release labor share index declines gradually from 93.2 in 2026Q3 to 89.3 in 2031Q3. The corresponding marginal 10–90 intervals widen from 91.5–94.9 to 79.5–98.5.

The evidence cutoff is September 13, 2026, the date of the latest forecast-data snapshot used. The latest direct observation of the named BLS series was updated on September 3, 2026. These are different dates: the more recent forecasting evidence does not constitute a newer BLS observation.

Target and reading

I forecast the seasonally adjusted quarterly index PRS85006173, 2017 = 100, not the percentage of output paid to labor and not an annualized growth rate. The example schema's Q6 identifier and annualized-percent unit are placeholders superseded by the question's explicit instructions. BLS's release narrative gives a labor-share percentage, while the named series is an index; the accompanying first-release dataset is therefore the appropriate numerical target.

I freeze each quarter at its first release. Subsequent revisions to that quarter do not change its resolution. However, a future quarter's first estimate can incorporate methodological or source-data revisions already implemented when it is initially published. I do not attempt to remove those revisions from the published first estimate. Any future rebasing is converted using BLS's published factor, as instructed.

Reference class and base rate, before case-specific adjustments

My reference class is five-year changes in this same index during the post-2000 economy, with quarterly fluctuations and preliminary-estimate uncertainty added separately. Calculations from the current historical series give 18 declines among 22 overlapping Q2-to-Q2 five-year windows beginning in 2000–2021, a descriptive decline frequency of approximately 82%. Their median change is about −3.54 index points, and their range is approximately −7.74 to +4.66. These overlapping observations are not independent trials. The endpoint trend is about −0.65 points annually since 2000Q2, versus −0.40 since 2010Q2. This supports a persistent but uncertain downward drift, rather than either constancy or automatic collapse.

This reference class matches the precise measure and spans different business cycles. Its limitation is that it is current-vintage history, not a complete real-time first-release backtest. I therefore use it as a trend and scale check, not as proof that the submitted intervals have achieved 80% historical coverage.

Current status and near-term anchor

The latest index observations are 96.723 in 2025Q2, 95.727 in 2025Q4, 94.843 in 2026Q1, and 93.446 in 2026Q2. These are September-vintage observations used as forecasting information, not substituted first-release resolutions.

The August 6 preliminary release reported the 2026Q2 labor-share level as 52.9%; September 3 revised it to 52.8%, a series low. This is a concrete comparable case showing why first-release identification matters. September's Table 2 also reports annualized Q2 unit-labor-cost growth of 1.2% against nonfarm-business output-price growth of 7.4%. That gap helps explain the recent decline without requiring an assumption that AI has already transformed the entire economy.

The relevant accounting relationship is labor compensation divided by nominal value added; equivalently, the labor-share index moves with unit labor costs relative to the sector's output-price deflator. Consumer-price inflation is not an interchangeable denominator. BLS documents the compensation and output concepts underlying this measure.

For Q3, my illustrative annualized assumptions are 1.5% unit-labor-cost growth and 2.6% output-price growth. Applying their ratio to 93.446 gives approximately 93.195, motivating the 93.2 median. These are my assumptions, not a published BLS nowcast. They imply substantial normalization from Q2's unusually large price–cost gap. August payroll average hourly earnings grew 3.1% over the year, while the Philadelphia Fed's August 14 survey anticipated Q3 real GDP growth of 2.5% and headline CPI/PCE inflation of 2.3%. These are useful cross-checks, but payroll wages exclude parts of compensation, GDP differs from nonfarm-business output, and consumer prices differ from the required output deflator.

Outside forecasts and expert evidence

Three outside checks influence the path and tails:

• LEAP Wave 7, published April 19, 2026, reports median forecasts of 58% labor share in 2026 and 56% in 2030 across experts, superforecasters, and public participants, using its stated 2024 baseline of 58.3%. Its rationales emphasize slow institutional adjustment and disagreement about whether AI accelerates the historical decline. I use the direction and relative scale, not those percentage levels as interchangeable values of PRS85006173.
• Metaculus's Labor Automation Forecasting Hub, with data dated September 13, reports a national-income labor-share median of 58.9% for 2030, with an interquartile range of 56.73–60.86%, and a 2035 median of 56.62%. This is a related community forecast, not a verified cash-market price or the exact first-release index target. It supports decline with substantial disagreement. I did not identify a usable cash-market quote for this exact series and vintage rule.
• Anthropic's September 2026 Economic Scenario Explorer produces illustrative 2030 labor shares of 59.4%, 56.1%, and 45.2% under modest, substantial, and extreme scenarios, from a roughly 60% starting share. These are conditional model outputs, not probabilities or calibrated forecasts. They justify considering a much lower structural tail, but not replacing the near-term BLS anchor with an extreme scenario. The model also explicitly omits several forces, including policy responses and business cycles.

Path, uncertainty, and causal mechanisms

After the near-term normalization, the median declines approximately 0.8 index points per year. This is moderately more negative than the long-run endpoint trend, but far less negative than mechanically extending the latest year's decline.

My rough five-year scenario weights are 65% ordinary persistence and gradual erosion, 20% stabilization or recovery, and 15% accelerated labor displacement/capital capture. These are my judgmental weights, not estimates from the cited surveys. They inform the tails rather than impose artificial jumps on particular quarters.

The main downward mechanisms are productivity and output prices persistently outrunning compensation, greater substitution of capital for labor, and weak transmission of productivity gains into pay. Higher outcomes can result from wage catch-up, labor scarcity, new labor-intensive tasks, stronger bargaining power, or falling profit margins. Automation does not mechanically lower the aggregate share: the balance between substitution, complementary labor demand, and new tasks matters. These competing mechanisms also appear in the expert rationales and the economic scenarios.

The long-horizon bands are deliberately wider than the historical five-year change distribution. They include uncertainty about the underlying drift, business cycles, compensation measurement, future first-release methodology, and an automation regime change. The lower tail moves farther below the recent historical range because a structural break is possible; the upper tail allows a substantial recovery rather than assuming the current record low is irreversible. These are marginal intervals for each quarter, not an 80% band for the entire 21-quarter path. I have not fitted a complete vintage-error model, so their calibration remains judgmental.

Strongest case against the central estimate

The strongest opposing case is stabilization or recovery, not an even faster decline. The New York Fed's June 24 analysis finds that post-COVID labor-share dynamics resemble earlier cyclical episodes, and that within-industry changes rather than unusual sectoral reallocation explain the decline. Earlier cycles sometimes saw labor share recover later in an expansion. Thus the current low need not mark a new permanent downward regime.

That evidence moves me away from extrapolating the 2025–2026 pace and warrants meaningful recovery probability. I nevertheless retain a declining median because the same-series historical reference class and both outside forecasting exercises favor gradual erosion. The strongest downside objection is that the median gives too much weight to historical adjustment speeds if AI substitutes for expensive cognitive labor much faster than earlier technologies; that possibility is represented primarily in the widening lower tail.

Publication process, incentives, and calendar

BLS must publish the resolving estimate, rather than a company voluntarily acknowledging an effect. Its calendar confirms November 5, 2026 for the Q3 preliminary release and December 8, 2026 for the revision, which will be ignored for Q3 resolution. The calendar also shows that earlier releases were delayed: 2025Q3 preliminary appeared January 8, 2026, and 2025Q4 preliminary March 5. Thus the usual short lag should not be treated as guaranteed, although the specified Q3 date is currently confirmed.

BLS identifies Brett Matsumoto as commissioner beginning August 11, 2026. In his June 10 confirmation testimony, he emphasized statistical independence, advance communication of methodological changes, and risks from declining survey response rates. BLS's quality policy likewise calls for impartiality and advance discussion of major methods changes. I therefore include measurement and timing uncertainty, but make no directional numerical adjustment on an unsupported assumption that officials will suppress or manipulate a low labor-share estimate.

Evidence in the next 90 days that would cause a large update

• Q3 first-release index at or below 91.4 or at or above 95.0: its November publication would resolve Q3 and shift that period's median by more than half its present 10–90 width, which is 1.7 points.
• A persistent baseline shift exceeding about 2.4 index points: new compensation or nominal-output source data that shift the inferred Q4 level that much would move its median by more than half its 4.6-point interval. Revisions would update unresolved quarters only, not rewrite resolved first releases.
• A genuinely structural deployment and income-distribution signal: to move the 2031Q3 median by more than half its 19-point interval would require much stronger evidence than an AI benchmark or an isolated layoff announcement—for example, independently corroborated broad deployment, payroll, and output evidence supporting roughly 20% net displacement of labor spending relative to the baseline by 2031. That would push the long-run median toward or below the present lower tail.
• Conversely, a durable, economy-wide compensation/nominal-output shift supporting a roughly 10-index-point higher long-run level would warrant an equally large upward update. Ordinary monthly payroll surprises would not be sufficient.

Source URLs and publication/update dates

• BLS historical series via FRED, updated September 3, 2026: https://fred.stlouisfed.org/data/PRS85006173
• BLS Q2 preliminary release, August 6, 2026: https://www.bls.gov/news.release/archives/prod2_08062026.htm
• BLS Q2 revised release, September 3, 2026: https://www.bls.gov/news.release/archives/prod2_09032026.htm
• BLS, Estimating the U.S. labor share, February 2017: https://www.bls.gov/opub/mlr/2017/article/estimating-the-us-labor-share.htm
• BLS August Employment Situation, September 2026 release, accessed September 14: https://www.bls.gov/news.release/empsit.nr0.htm
• Philadelphia Fed SPF, August 14, 2026: https://www.philadelphiafed.org/surveys-and-data/real-time-data-research/spf-q3-2026
• LEAP Wave 7, April 19, 2026: https://leap.forecastingresearch.org/reports/wave7
• Metaculus dashboard, data September 13, 2026: https://labor-hub-tools.metaculus.com/question-statistics
• Anthropic Economic Scenario Explorer, September 2026, version 1.0: https://www.anthropic.com/institute/econ-scenarios
• New York Fed labor-share analysis, June 24, 2026: https://libertystreeteconomics.newyorkfed.org/2026/06/the-post-covid-decline-in-the-labor-share/
• BLS release calendar, accessed September 14, 2026: https://www.bls.gov/schedule/news_release/prod2.htm
• BLS commissioner biography, August 11, 2026: https://www.bls.gov/bls/senior_staff/matsumoto.htm
• Matsumoto Senate testimony, June 10, 2026: https://www.help.senate.gov/imo/media/doc/8aa0f266-cf6d-5110-f6a2-e4957bbe6185/Matsumoto%20Testimony_da31b8ff-3581-4b25-8cab-23908cca731b.pdf
• BLS quality and methodology guidelines, updated April 11, 2024: https://www.bls.gov/bls/quality.htm

Forecast date: 2026-09-14. Most recent evidence: BLS Productivity and Costs, Q2 2026 Revised (2026-09-03); August CPI (2026-09-11); GDPNow (2026-09-10).

Current status

• Resolving series PRS85006173 (nonfarm business labor share, 2017=100). Latest vintage (Q2 2026 revised release, 2026-09-03, bls.gov/news.release/prod2.nr0.htm; full history in bls.gov/web/prod2/labor-productivity-major-sectors.xlsx): 2025Q1 97.87, 2025Q2 96.72, 2025Q3 96.00, 2025Q4 95.73, 2026Q1 94.84, 2026Q2 93.45 (labor share level 52.8%, a record low for the series back to 1947). The index fell 4.4 points in five quarters.
• The index is mechanically unit labor costs ÷ value-added price deflator. The plunge reflects the nonfarm business deflator rising 5.1% (Q1) and 7.4% (Q2) annualized against hourly compensation growth of only ~2–2.6%; unit nonlabor payments +14.8% and nonfinancial-corporate unit profits +43% annualized in Q2. Macro backdrop: US–Iran conflict (now ~7 months) with Brent ~$100–105 in September after ~$84 (Jul) and ~$91 (Aug) (EIA STEO 2026-09-09; CNBC 2026-09-09); PCE inflation 5.3% annualized in Q2 (BEA 2026-08-26); CPI 3.4% y/y, core 2.4% in August; AHE +3.1% y/y, ECI +3.4% y/y — wages are not keeping pace with nominal output; GDPNow for Q3 2026 at 4.4% (2026-09-10).
• Next preliminary releases: Q3 2026 on 2026-11-05; it will incorporate BEA's annual NIPA update released 2026-09-30, which revises 2021–2026 compensation/output.

Vintage/revision noise (matters because resolution is the first-release figure)

From archived BLS releases I computed first-release vs. current values: 2024Q4 prelim 97.28 → now 96.94; 2025Q1 97.06 → 97.87; 2025Q2 97.71 → 96.72; 2025Q3 95.22 → 96.00; 2025Q4 95.26 → 95.73; 2026Q1 95.70 → 94.84; 2026Q2 ~93.5–93.7 → 93.45. Typical |prelim − later| ≈ 0.5–1.0 points. The Sept-2025 annual update alone moved 2024–2025 levels by −0.2 to −1.1 points. So each first-release figure carries ~0.7 pt SD of "vintage noise" on top of true movement.

Reference class / base rates

Historical changes of the index (1947–2026): SD of 1-qtr change ≈1.05, 4-qtr ≈1.7, 8-qtr ≈2.3, 12-qtr ≈2.7, 21-qtr ≈3.2–3.4; since 1990 the 21-qtr change had p10 −6.2, p50 −3.0, p90 +2.0. The series moves in step-downs during shocks (2001–03, 2008–10, 2022, 2025–26) separated by flat plateaus (2010s: 99.4→99.6 over a decade). After the 2022 inflation-driven step-down, there was no rebound; after 1974 and 2008 there were temporary recession-driven spikes followed by declines.

Near-term path

Q3 2026 prelim: strong nominal output growth (GDPNow 4.4% real, deflator likely 3–4%) versus ~3–3.5% compensation growth implies another decline of roughly 0.2–0.8 pts; oil averaged somewhat lower in Q3 than Q2, so the deflator surge should moderate. Median ≈ 93.0 with SD ≈1.05 (Q3 change ± annual-update revision). Q4 2026/Q1 2027: continued modest downward pressure while the war/oil shock and record margins persist; median ≈ 92.6–92.7.

Medium/long term (2027–2031)

Competing forces: (a) mean reversion — oil normalization and wage catch-up could restore ~1–1.5 pts, and a recession (probability perhaps 25–35% over 2027–28) would temporarily spike the share 1–3 pts as profits fall; (b) continued decline — AI/data-center capital deepening, productivity gains accruing to capital, weak wage growth, and a possible longer oil shock. Given the record-low level and the 2010s plateau precedent, I set a roughly flat median through 2028 (~92.5–92.6) drifting to ~92.0 by 2031Q3 (a ~−1.4 pt five-year change, between the flat 2010s and the −3 historical median). Bands widen with horizon to roughly ±4 pts by 2031, matching historical 21-qtr change dispersion plus vintage noise.

Strongest case against

The momentum case: 2022's step-down never reversed and the current episode has the same signature (deflator surge, profit boom, weak wages); with GDPNow at 4.4%, Q3 could print near 92.0 and 2027 could sit at 91–92. Conversely, a fast end to the Iran conflict with oil back to $65 plus a 2027 recession could put 2027–28 values at 95–96. Both are within my 10–90 bands.

Would move me >half the band within 90 days

• BEA annual update (Sept 30) revising 2025–26 compensation/nominal output by >0.7% (shifts all near-term medians ~±0.7).
• Q3 GDP advance (late Oct) with nominal nonfarm output >8% annualized and hours <1% → Q3 median toward 92.3; or nominal output <4% → toward 93.5.
• A ceasefire with Brent < $75 (upside for 2027+), or escalation with Brent > $120 (downside).
• A BLS rebasing announcement (handled via conversion).

Sources: BLS P&C Q2 2026 revised (2026-09-03) and archived releases (prod2_08062026, 06042026, 05072026, 03242026, 03052026, 01292026, 01082026, 09042025, 08072025, 06052025, 05082025, 03062025, 02062025); BLS major-sector xlsx; BEA GDP second estimate Q2 2026 (2026-08-26); BEA annual update notice (Sept 30, 2026); Atlanta Fed GDPNow (2026-09-10); BLS CPI Aug 2026 (2026-09-11); BLS Employment Situation Aug 2026 (2026-09-04); BLS ECI Q2 2026 (2026-07-31); EIA STEO (2026-09-09); ALFRED PRS85006173 vintage page.

Current status vs resolution criteria

Resolving figure is preliminary (first) Productivity and Costs value of nonfarm-business labor-share index PRS85006173, 2017=100, from bls.gov release + accompanying tables.

Latest published vintages:

• Q2 2026 preliminary (Aug 6, 2026, USDL 26-1290): productivity +1.4%, output +1.7%, hours +0.3%, hourly comp +2.7%, ULC +1.3%, labor-share 52.9%, lowest since 1947. Table 2 indexes: prod 120.0, output 127.1, hours 105.9, comp 148.8, ULC 124.0, nonlabor 143.7, deflator 132.6. Implied prelim index =124.0/132.6*100 = 93.5.
• Q2 2026 revised (Sep 3, 2026, USDL 26-1434): productivity +1.4% unrevised, ULC revised to +1.2% on comp 2.6%, labor-share 52.8%, new record low. Table 2: ULC 124.0, deflator 132.7 => 93.4. FRED latest: Q2 93.446, Q1 94.843, Q4 2025 95.727, Q3 2025 95.999, Q2 2025 96.723.
• Q1 2026 preliminary (May 7, 2026, USDL 26-0686): productivity +0.8%, ULC +2.3%, labor-share 54.1%, then record low. Table 2: ULC 124.7, deflator 130.3 => 95.7.
• Q1 2026 revised (June 4, 2026): productivity cut to +0.3%, ULC +1.8%, labor-share 53.7%. Sep vintage: ULC 123.6 / deflator 130.3 => 94.86 matching FRED 94.84.
• Q4 2025 revised (Mar 24, 2026): productivity +1.8% (cut from +2.8%), ULC +4.4% (up from +2.8%), labor-share 54.4%.

So first-release index fell 95.7 (Q1 prelim) -> 93.5 (Q2 prelim) = -2.2 pts in one quarter, while latest-vintage fell 94.84 -> 93.45 = -1.4 pts. Driver is deflator +7.4% ann in Q2 (+5.1% in Q1) on nonlabor payments +14.8% ann (+9.8% in Q1) vs ULC only +1.2-1.3%, plus unit profits of nonfinancial corps +43.0% ann in Q2, +17.8% y/y, highest since 2021.

Next resolving print: Q3 2026 preliminary scheduled Nov 5, 2026.

Base rate / reference class

Long run: index ~114.6 in 1947 to 93.5 in 2026 = -21 pts /79y = -0.27/yr (-0.07/q). Post-2017 (100 to 93.5 in 9y) = -0.72/yr (-0.18/q). Last 5 quarters (Q1 2025 97.87 to Q2 2026 93.45) = -4.4 pts/yr (-1.1/q), an outlier 4-5x long-run pace.

Quarterly change volatility: ~0.5-0.7 pts std historically; 80% revision band for productivity ±1.1-1.4pp implies labor-share first-release noise ~±0.8-1.0 pt.

I anchor near term on nowcast, distant medians on post-2017 drift (~0.8/yr =0.20/q) – i.e. continuation of capital-biased trend but reversion from 2025-26 outlier – with linear-widening 80% bands (margin 1.0 +0.18*n, n=quarters ahead) giving ~±1.0 near, ±4.6 at 5y (width ~9.2), consistent with random-walk + trend uncertainty.

Causal pathways to higher/lower prelim prints

Down (lower index = lower labor share):

• Productivity stays ~2.1% in current business cycle (Q4 2019-Q2 2026) vs 1.5% prior cycle, on automation, capital deepening, AI data-center investment; GDPNow Q3 tracking 4.4-4.7% with hours +0.3-0.5% implies another strong productivity quarter.
• Hourly comp ~2.6-3.1% nominal, real comp negative in Q2 (-3.3% ann on oil/CPI spike, flat to slight contraction per EY-Parthenon), ECI wages 0.9% q/q, so ULC ~0-1.5%.
• Nonlabor/unit profits continue surging, deflator 3-4%, mechanically depressing ULC/deflator ratio. EY: margins 14.9% of GDP record; “I don’t think there’s a floor,” 50% not a floor if concentration persists. SP Global/CBO both note wages share revised down and AI could further depress share.

Up (higher index):

• Wage catch-up, profit-margin normalization, productivity slowdown (JPM notes productivity ex-data-centers soft), recession raising labor share countercyclically as profits collapse.
• Large downward revisions to output or upward to comp in prelim vintage (Q1 prelim was 0.86 pt above latest vintage).

Central path

Q3 2026 prelim median 92.8 (-0.7 vs Q2 prelim): assumes output ~4%, hours ~0.5%, productivity ~3%, comp ~3%, ULC ~0-1%, deflator ~3.5-4%. Then -0.3 in Q4 2026 to 92.5, then -0.20/q to 88.7 in 2031Q3 (50.2% share at 56.55% 2017 base vs 52.8% now). p90 flat-ish ~93.3-93.8 (stabilization possible); p10 to 84.1 (accelerated AI displacement).

Strongest case against central (higher/stable)

Labor share is at 79-year low; 4-pt/yr drop cannot persist arithmetically without share falling below 50% – unprecedented. Profit surge partly cyclical post-inflation and may mean-revert; real wage growth (+0.6% y/y in Q1, ECI re-accelerating) plus tight labor market (unemp 4.1%, +162k Aug) could lift ULC above deflator. CBO potential-GDP framework still embeds stable factor shares long-run. If so, median should be flat ~93, not 88.7. I keep p90 at ~93.3-93.8 to admit this.

What would move forecast >10pp of interval (>~0.5 near, >~2 far)

• Q3 prelim on Nov 5 printing <91.5 or >93.5 (1+ pt surprise) would shift whole path ~1 pt.
• CPI/ECI/Q3 GDP revisions showing wage acceleration >4.5% or profit collapse would flatten trend.
• BLS benchmark / hours-paid ratio revision back to 2017 or base-year change with published factor.

Sources with dates:

• BLS Sep 3 2026 revised PDF: https://www.bls.gov/news.release/PDF/prod2.PDF
• BLS Aug 6 2026 prelim archive: https://www.bls.gov/news.release/archives/prod2_08062026.htm
• BLS May 7 2026 prelim: https://www.bls.gov/news.release/archives/prod2_05072026.htm
• BLS June 4 2026 revised: https://www.bls.gov/news.release/archives/prod2_06042026.htm
• BLS Mar 24 2026 Q4 revised: https://www.bls.gov/news.release/archives/prod2_03242026.htm
• BLS schedule Nov 5 Q3 prelim: https://www.bls.gov/productivity/
• FRED PRS85006173: https://fred.stlouisfed.org/series/PRS85006173 (Q2 93.446 etc)
• GDPNow 4.4-4.7% Sep 2026: https://www.atlantafed.org/... and https://www.newsquawk.com/headlines/atlanta-fed-gdpnow-q3-2026-44-prev-47-
• Fortune/Yahoo Sep 4 2026 “no floor”, margins 14.9%, share 52.8%: https://finance.yahoo.com/economy/articles/don-t-think-floor-workers-073000433.html
• Reuters Aug 6 2026 record low: https://www.reuters.com/business/us-workers-share-gdp-skids-fresh-record-low-2026-08-06/
• BLS labor-share method: https://www.bls.gov/opub/mlr/2017/article/estimating-the-us-labor-share.htm
• CBO Budget & Economic Outlook 2026-2036 wages share revised down: https://www.cbo.gov/publication/62105 and https://www.cbo.gov/system/files/2026-02/61882-Outlook-2026.pdf

Current status

The BLS nonfarm business labor share index (PRS85006173, 2017=100) is at an all-time series low. Per the BLS API and the September 3, 2026 revised Productivity and Costs release (bls.gov/news.release/prod2.nr0.htm), Q2 2026 = 93.446, corresponding to a labor share of 52.8% of output — "the lowest level in the series, which begins in the first quarter of 1947." The preliminary Q2 2026 release (Aug 6, 2026, bls.gov/news.release/archives/prod2_08062026.htm) put it at ~93.5 (52.9%), so the preliminary-vs-revised wedge is small (±0.1–0.3).

Recent path (revised values, BLS API): 2025Q1 97.871 → 96.723 → 95.999 → 95.727 → 2026Q1 94.843 → 2026Q2 93.446. That is an average decline of -0.88 index points per quarter (≈ -3.5%/yr), roughly five times the fastest sustained decline in the series' 78-year history (e.g., 2000–2005 fell ~8 points over 5 years; the 1947–2025 average drift is ~-0.4 pts/yr).

Mechanism and drivers

The plunge reflects an extraordinary wedge between output prices and labor cost: in Q2 2026 the nonfarm business value-added deflator rose ~7% annualized and unit nonlabor payments +14%, while hourly compensation rose only +2.7% and real hourly compensation fell 3.3%. Nonfinancial corporate unit profits rose 43% annualized (index 183.6, 2017=100). Causes visible in current data: (1) the AI capex/margin boom (record S&P profits, hyperscaler capex driving 6.1% business investment growth per Deloitte); (2) tariff passthrough (Section 301/122 tariffs after the Feb 2026 Supreme Court ruling) and an oil spike from the Middle East conflict (Brent ~$92; CPI 3.4% y/y in July and August 2026); (3) moderating wage growth (AHE ~3.4%, inflation outpacing wages five straight months per Business Insider); (4) a mechanical drag — BLS nonlabor payments include depreciation, so the exploding AI capital stock depresses the labor share even if profits normalize.

Reference class and base rates

Historically, 5-year moves in the index stay within roughly ±8 points (1995–2000: +5.8; 2000–2005: -8.1; 2007–2012: ~-5; 2019–2024: -2.7). Recessions produce an acute spike (2008Q4, 2020Q2) followed by renewed declines in "jobless recoveries" (2001–2003 fell ~6 points through the dot-com bust). The current pace (-17.7 points annualized-equivalent over 5 years) is far outside this range, so pure historical base rates understate downside risk; but forecasters expect normalization: Deloitte's July 2026 baseline (no recession; GDP -1% only in the downside AI-bust scenario) sees oil lower in 2027, tariff passthrough fading, and the Fed hiking in 2026 then cutting by end-2027; CBO expects elevated but moderating inflation through 2029. The Forecasting Research Institute's expert survey (forecastingresearch.org, Mar 2026) finds economists expect a falling labor share alongside rising TFP as the most plausible AI scenario, but with outcomes "not outside the range of historical experience" by 2030.

Path construction (median)

Three-scenario mixture for the 2031 endpoint: (A) AI boom persists, labor share grinds lower with the depreciation drag (~35%): ~84–86; (B) normalization/margin mean-reversion, no recession (~35%): ~90–91; (C) bust/recession with wage catch-up and a labor-share spike (~30%): ~93–96. The mixture median at 2031Q3 is ~90, with the trough (~89.6, i.e., labor share ~50.6%) around 2029. Near term I assume the decline continues but decelerates: -0.8 in Q3 2026 (Nov 5 release), -0.7 in Q4 2026, easing to ~-0.3/quarter by 2027–28 as the deflator normalizes toward 3–4%, then flattening, with a mild recovery in 2030–31 if margins mean-revert and wage growth catches up to inflation.

Uncertainty bands

Near-horizon: recent quarterly changes ranged -0.27 to -1.40 with sd ~0.45, plus preliminary-release noise; the Q3 2026 10–90 band is 91.6–93.6 (only ~15% chance of a quarter-over-quarter rise given five straight declines and elevated deflators). The 21-quarter band widens to 84.4–96.6 (12 points): p10 implies a continued ~-0.43/quarter decline to labor share ~47.7% (plausible if AI displacement broadens per the Fed's July 2026 monitoring note); p90 implies a +3-point rebound (wage-led catch-up, margin normalization, acute recession dynamics). This is wider than the historical ±8-point 5-year range, appropriately, given the unprecedented current pace.

Strongest case against my central estimate

That the current -0.88/quarter pace persists much longer than assumed. The AI buildout is still accelerating (Fed note, July 2026: task horizons doubling every several months; impacts "concentrated and not yet broadened"), inflation is sticky (core CPI accelerating as of August 2026; Fed closer to a hike), and the depreciation drag compounds. Under that view the 2027 medians are 1–2 points too high. Counterweight: profit margins at records mean-revert with high base-rate reliability, oil is expected to fall, tariff passthrough is a one-time level shift, and workers historically claw back nominal wages after sustained inflation — the 5-quarter momentum already embeds the peak of the deflator spike.

What would move me >10 points

The Q3 2026 preliminary release itself (Nov 5, 2026): a print near or above 93.5 (no further decline) would materially raise the whole path; a print near 92 or below would steepen it. Also: hyperscaler 2027 capex guidance (boom vs. cuts), CPI/PCE trajectory and Fed action, ECI/AHE re-acceleration, and NIPA corporate profit data showing margin normalization.

Sources: BLS Productivity and Costs releases (prod2.nr0.htm, 2026-09-03; archives/prod2_08062026.htm, 2026-08-06) and BLS API series PRS85006173; FRED PRS85006173; S&P Global Market Intelligence (2026-06-30); Deloitte US Economic Forecast Q2 2026 (2026-07-01); Fed FEDS Note "The AI Buildout and the Economy" (2026-07-17); CBO Budget and Economic Outlook (2026-02-11); Forecasting Research Institute, "Forecasting the Economic Effects of AI" (2026-03-31); Reuters/CBS/WSJ/Morningstar CPI and jobs coverage (Sept 2026); NYT op-ed (2026-09-07).

Current status. The resolving series is BLS PRS85006173 (nonfarm business labor share, 2017=100), taken from each quarter’s preliminary Productivity and Costs release—not the later revised/FRED vintage.

As of the 3 Sep 2026 revised Q2 2026 release, the index is 93.446, and BLS stated the labor share of output was 52.8%, the lowest since the series begins in 1947 (BLS PDF, 3 Sep 2026; BLS API; FRED PRS85006173). The preliminary Q2 2026 release (6 Aug 2026) had labor share at 52.9%, with Table 2 indexes ULC 124.0 / deflator 132.6 ⇒ index ≈ 93.51 (BLS archive). Q3 2026 preliminary is scheduled for 5 Nov 2026.

Recent current-vintage path (BLS API):

Quarter	Index
2024Q1–Q4	96.882, 96.602, 96.657, 96.940
2025Q1–Q4	97.871, 96.723, 95.999, 95.727
2026Q1–Q2	94.843, 93.446

That is a 4.4-point drop in five quarters, after two years stuck near 96.2–96.9 (2023–24). The 2010s range was roughly 97.6–101.3; 2000 was ~113; 2020Q2 (COVID) peaked at 104.2. Current is far below any post-1947 reading.

Identity used for nowcasts: labor share index ≈ ULC / value-added deflator × 100, so %Δ labor share ≈ hourly compensation − productivity − output deflator (SAAR).

Q2 2026 (revised, SAAR): hourly compensation +2.6%, productivity +1.4%, deflator +7.4% (oil-driven). Real hourly compensation −3.3%. Unit nonlabor payments +14.8%; nonfinancial corporate unit profits +43% SAAR (BLS; Fortune/Daco, 4 Sep 2026).

Q3 2026 nowcast (first release 5 Nov).

• Real GDP: Atlanta Fed GDPNow 4.4% SAAR as of 10 Sep 2026, with PDFP 4.2% and a large inventory contribution (GDPNow slides). Q2 NFB output was +1.7% vs GDP +1.5%, so NFB output in Q3 is likely ~4–5%.
• Hours: payrolls were very weak (prior-12-month average +31k) then August +162k; workweek 34.3 → 34.4. Q2 NFB hours +0.3% SAAR. Q3 hours ~1–2.5% SAAR ⇒ productivity maybe 2.5–3.5% SAAR.
• Wages: AHE +3.1% y/y, +0.3% in August (Employment Situation, 4 Sep 2026); ECI private wages +0.9% q/q in Q2 (~3.6% SAAR); Atlanta Fed Wage Growth Tracker 4.1% in August. NIPA hourly compensation has been running only +2.1% / +2.6% in 2026 H1. Central Q3 hourly compensation ~3.2–3.5% SAAR.
• Deflator: Q2’s 7.4% is not a trend. CPI: Jun −0.4%, Jul +0.1%, Aug +0.4%; energy +16.3% y/y but sequential energy was −5.7% / −1.5% / +2.1%. Q2 GDP price index +6.4%, PCE +5.3% (BEA, 26 Aug 2026). Central Q3 NFB deflator ~3.5–4% SAAR.

Putting those together: 3.3 − 2.8 − 3.7 ≈ −3.2% SAAR ≈ −0.8% q/q ⇒ level ~92.7–93.0. Oil-mean-reversion (deflator back to ~2.5%) could put Q3 near 93.3–93.5; a still-hot deflator plus a productivity boom could put it near 92.3. First-release noise is material: Q1 2026 preliminary labor share was 54.1% vs a later ~53.6% (≈0.9 index points). The 30 Sep 2026 BEA annual NIPA revision will also re-level compensation/output before the 5 Nov release (±0.5–1.5 points historically).

Q3 2026 first-release: p10 91.2 / p50 93.0 / p90 95.1.

Reference class and base rate. (1) Quarterly changes in the 2010s had SD ≈ 1.0–1.2 index points (several ±2 moves); 2015–18 was calmer (~0.4); 2024–26 SD ≈ 0.75. I use ~0.9 as the one-quarter shock SD, plus ~0.5–0.8 of first-release/annual-revision noise. (2) Five-year changes: 2019Q2 100.15 → 2024Q2 96.60 (−3.6); 2020Q2 104.20 → 2025Q2 96.72 (−7.5); 2000Q1 112.8 → 2005Q1 104.7 (−8.1). Post-2000 drift is about −0.7 index points per year. (3) Cyclical: labor share is countercyclical (COVID +2 from Q1 to Q2 2020; 2001 readings 111–113). NY Fed treasury-spread recession probability through Aug 2027 is only ~14% (NY Fed); Conference Board still has ~1.9% GDP for 2026–27. Recession odds by 2031 are still high (~60–75% on a 6–8 year cycle). A recession typically adds 2–5 points for a few quarters.

Path. Near-term median: Q2’s oil-deflator spike fades, but GDPNow-strong output and still-soft wages keep the index from bouncing much—slight further decline then a slow drift. Longer-term median follows a muted version of the post-2000 trend (about −0.4/year from 93.0), not the −1.4/year 2025–26 pace, because that pace was inflated by a 7.4% deflator. AI/capital-biased technical change, record profit margins, and weak real wages (AHE 3.1% vs CPI 3.4%) keep the median below the 2023–24 plateau of ~96.5. Recessions in the mix pull the later median up a bit versus a pure no-recession path (which would be closer to 89–90). Endpoint: p50 ≈ 91.1 in 2031Q3.

Bands. 10–90 width starts at 3.9 points for 2026Q3 (one-quarter shock + first-release + NIPA revision) and widens to ~14 points by 2031Q3. p10 at five years (~85.3) is “2025–26 pace continues” (labor share ~48% of output). p90 (~99.4) is a recession plus wage catch-up back to the 2017–19 range—not a COVID-sized 104. Historical 80% five-year coverage needs this width; narrower bands would fail in 2020 or 2025–26.

Strongest case against. (1) Bounce: if the deflator fully mean-reverts to ~2.5% while wages hold at 3.5% and productivity is only ~2%, Q3 is ~93.5–94 and the median should sit near 94–96 as the oil shock fades. I partly buy the deflator reversal but offset it with GDPNow-implied productivity. (2) Faster structural decline: five more years at −0.7 to −1.4/quarter would put 2031 in the low 80s; Gregory Daco’s “I don’t think there’s a floor” is the narrative. I put that in the p10, not the median, because 7%+ deflators do not persist and labor share is somewhat mean-reverting at business-cycle frequency. (3) Mean-reversion to 99: the 2010s spent a decade there. Getting back without a labor boom or a deep recession looks unlikely given AI and current profit concentration.

Sources (key): BLS Productivity and Costs 6 Aug 2026 preliminary and 3 Sep 2026 revised; BLS public API PRS85006173 (2010–2026 and 2000–09); BLS Employment Situation 4 Sep 2026; BEA GDP second estimate 26 Aug 2026; Atlanta Fed GDPNow 10 Sep 2026; Atlanta Fed Wage Growth Tracker 10 Sep 2026; BLS CPI Aug 2026 (11 Sep); Fortune/Daco 4 Sep 2026; NY Fed recession probability (data through Aug 2026).

Current status and resolution reading

The latest published value of PRS85006173 is 93.446 for 2026Q2, from the revised Productivity and Costs release on September 3, 2026. This is the relevant jumping-off point for the 2026Q3 calculation, although Q7 resolves each future quarter from its own first/preliminary vintage, not this revised series. The same BLS release reports a nonfarm-business labor share of 52.8%, a record low; productivity rose 1.4%, hourly compensation 2.6%, unit labor costs 1.2%, while the value-added output deflator rose 7.4% and unit nonlabor payments rose 14.8% annualized. BLS also reported nonfinancial-corporate unit profits up 43.0% annualized. (BLS, 2026-09-03; FRED PRS85006173, updated 2026-09-03).

The Q2 preliminary release had labor share at 52.9%, unit labor costs up 1.3%, and hourly compensation up 2.7%; the September revision was small, illustrating the ordinary vintage noise the forecast bands need to include. (BLS, 2026-08-06). The Q3 preliminary release is scheduled for November 5, 2026. (BLS release schedule).

I found no ambiguity material enough to change the reading: forecast the published preliminary PRS85006173 level, in 2017=100 terms.

Near-term forecast: further decline, then stabilization

Economically, the index can be approximated as hourly compensation relative to labor productivity and the output-price deflator. Current information points to another weak labor-share print in Q3:

• Atlanta Fed GDPNow was 4.4% SAAR for Q3 on September 10. (Atlanta Fed).
• August payrolls rose 162,000, aggregate weekly hours rose 0.3% in the month, and average hourly earnings were up 0.3% month over month but only 3.1% year over year. (BLS Employment Situation, 2026-09-04).
• July PCE inflation was 3.7% year over year. (BEA, 2026-08-26). August CPI was up 0.4% month over month and 3.4% year over year. (BLS CPI, 2026-09-11).
• S&P 500 Q3 earnings are expected to grow 28.7% year over year, with a 14.9% net margin; Q4 earnings growth is projected at 26.3%. Those margins are down from Q2's exceptional 17.0% but still far above historical averages. (FactSet Earnings Insight, 2026-09-11).

I therefore put the 2026Q3 median at 92.6, about 0.8 index point below revised Q2, and Q4 at 92.0. Strong real output, moderate wage growth, elevated output prices, and high profits all press the ratio down. I expect the trough around 91.1 in late 2027, as the current profit and inflation impulse slowly passes but earnings growth remains positive.

Medium- and long-run path

The central path then recovers gradually to about 95.0 in 2026Q3, 95.2 at end-2030, and 95.7 by 2031Q3. This is not a forecast of a full return to 2017=100; it is a partial reversal of the unusually sharp 2025–27 decline.

The main sources for that path are:

1. Historical/cyclical base rate. The index was 97.871 in 2025Q1, 94.843 in 2026Q1, and 93.446 in 2026Q2. Moves of 0.9–1.4 points in consecutive quarters are large by post-2017 standards. Over longer periods, the series exhibits both persistent downward episodes and sharp cyclical rebounds when profits fall faster than compensation. The New York Fed finds the post-COVID decline was driven mainly by within-industry changes and broadly followed prior recession-expansion patterns, rather than showing a wholly new sectoral-reallocation dynamic. (NY Fed Liberty Street, 2026-06-24).
2. Macro normalization. The June FOMC median had PCE inflation falling from 3.6% in 2026 to 2.3% in 2027 and 2.0% in 2028, with real GDP still near 2.2–2.3%. (Federal Reserve SEP, 2026-06-17). Lower output-price inflation plus nominal-wage catch-up should stop the present rate of labor-share erosion.
3. Profit-share mean reversion. CBO's February 2026 outlook expected domestic corporate profits to decline modestly as a share of GDP from 2026 to 2031. (CBO, 2026-02-11). That is an upward force on the BLS labor share after the current profit surge.
4. Outside expert view. The March 2026 Forecasting Research Institute survey reported a 2025 baseline labor share of 55.5%. Economists' median unconditional 2030 forecast was 54.3%; the all-respondent median was 54.0%, with scenario medians of 55.0% slow, 53.8% moderate, and 51.3% rapid AI progress. Using the current relation between 52.8% and index 93.446, 54.0% is roughly index 95.6, very close to my 2030–31 central path. (FRI, 2026-03-31, especially Table 39).
5. Measurement caveats. The BLS measure is a gross nonfarm-business ratio with imputed proprietor compensation; depreciation, taxes including tariff revenue, and the GDP/GDI statistical discrepancy can move it without an equivalent change in household capital income. This makes the current record low less structurally definitive than headlines imply and supports some mean reversion. (Tax Foundation, 2026-09-03).

Uncertainty and strongest counterargument

The bands widen from roughly ±1.5–2 points in 2026 to about ±6–7 points by 2031. That is intentionally wider than a simple random-walk extrapolation because the series is exposed to both business-cycle and AI-regime uncertainty.

The strongest case against the central forecast is that the 2026 fall is not mainly a temporary profit/tariff/inflation phenomenon but the beginning of persistent AI-driven capital deepening. The July FOMC minutes reported strong AI-related investment and possible future productivity effects, while the Federal Reserve's July 17 AI note emphasized that aggregate transformation is still uncertain. (FOMC minutes, 2026-08-19; FEDS Notes, 2026-07-17). Anthropic's September scenario work is explicitly not a prediction, but illustrates the downside tail: in its substantial scenario labor share falls from 60% to 56% by 2030, and in the extreme scenario to 45%. (Anthropic, 2026-09). If that kind of transition is underway, the expert survey and my median are too high and the index could remain near or below 89.

The upper tail is also real: an AI-investment bust, recession, or broad margin compression can raise labor share quickly because compensation is stickier than profits. The 2031 p90 of 102.2 covers a substantial but not complete cyclical rebound.

Evidence in the next 90 days that would move this forecast materially includes the Q3 first print itself, BEA's annual NIPA update and Q3 corporate profits, evidence that Q3/Q4 margins remain near Q2's exceptional level, a decisive break in hourly compensation or productivity, and clear signs of either broad AI productivity acceleration or an AI-capex bust.

1. Current Status and Resolution Criteria

The resolving metric is the BLS nonfarm business sector labor share index (series PRS85006173, base year 2017 = 100) as reported in the preliminary (first) Productivity and Costs news release for each quarter.

According to BLS historical series data and news releases:

• Q2 2026 preliminary release (published August 6, 2026): The index was reported at 93.547 (headline labor share percentage of 52.9%, an all-time series low since 1947). In the revised release on September 3, 2026, the Q2 2026 index was revised to 93.446 (labor share 52.8%).
• Q1 2026 preliminary release (published May 7, 2026): The preliminary index was 94.990 (headline labor share 54.1%), subsequently revised to 94.843.
• Historical recent trajectory:
  • 2024: Q1 = 96.882, Q2 = 96.602, Q3 = 96.657, Q4 = 96.940
  • 2025: Q1 = 97.871, Q2 = 96.723, Q3 = 95.999, Q4 = 95.727
  • 2026: Q1 = 94.843 (preliminary 94.990), Q2 = 93.446 (preliminary 93.547)
• Release Schedule: BLS publishes the preliminary report within 40 days of quarter close. The Q3 2026 preliminary release is scheduled for November 5, 2026, and Q4 2026 preliminary for early February 2027.

2. Base Rate and Reference Class

The labor share index (PRS85006173) equals (Unit Labor Costs / Value-Added Output Price Deflator) * 100.

• Over the 1947–2000 era, the series hovered in the 106–116 range.
• Since 2001 (peak of 113.626 in 2001Q1), the series has undergone a sustained secular decline of approximately -0.20 index points per quarter (-0.80 points per year).
• Between 2010 and 2019, the index oscillated in a depressed band between 97.5 and 101.5 (centered near 99–100).
• Following a temporary COVID-era spike (104.2 in 2020Q2), the index declined to ~96.5 in 2023–2024, before accelerating downward in 2025–2026 to reach 93.5 in Q2 2026.
• The standard deviation of single-quarter preliminary index changes is approximately 0.65 index points, with extreme quarter-over-quarter swings rarely exceeding ±1.5 points.

3. Drivers and Causal Pathways

Near Term (2026Q3 – 2026Q4):

• Q3 2026 nowcast: Output growth is tracking strongly (Atlanta Fed GDPNow tracking 4.0%–4.4% annualized real GDP growth as of September 10, 2026). Nonfarm payrolls added 21k in July and 162k in August, with average weekly hours at 34.4. Average hourly earnings grew 0.3% in August (3.1% YoY). Output growth (~3.2%) outpacing hours growth (~1.0%) implies nonfarm productivity growth of around 2.2% annualized.
• With hourly compensation tracking ~3.3%–3.5%, unit labor costs (ULC) are estimated to rise ~1.1%–1.3% annualized. Meanwhile, the value-added output price deflator is tracking ~2.2%–2.5% annualized (headline CPI was +0.1% in July, +0.4% in August).
• Because deflator growth is modestly outpacing ULC growth (partially due to firm pricing power and nonlabor capital payments), labor share is expected to decline slightly or remain near its lows, pointing to a Q3 2026 preliminary index print of around 93.30 (90% interval: [92.20, 94.40]).

Medium to Long Term (2027 – 2031):

• Downside pathways (continued decline / low p10):
  • Rapid advancement and enterprise adoption of agentic AI and workflow automation in white-collar and knowledge-worker sectors (finance, tech, customer operations, legal, administrative), dampening wage bill growth relative to real output.
  • Continued concentration of market share in highly automated "superstar" firms with structurally lower labor shares.
  • Institutional factors: weakened collective bargaining power and implementation of flexible independent contractor rules under the current administration.
  • Under these conditions, the index could drift down to the mid-80s by 2031.
• Upside pathways (mean reversion / high p90):
  • Corporate profit margins reached extreme highs in mid-2026 (nonfinancial corporate unit profits rose 43.0% annualized in Q2 2026, and aggregate corporate profits rose $400.9B). Profit shares typically mean-revert following sharp cyclical run-ups.
  • Macroeconomic downturn or recession: wages are historically stickier than corporate profits. In past recessions (2001, 2008, 2020), labor share exhibited sharp countercyclical spikes.
  • Aging demographics and labor supply tightness in physical, skilled trade, and healthcare occupations could sustain wage pressure.
  • Under a margin-compression or recessionary scenario, the index could rebound into the 96–97 range.

4. Strongest Consideration Against the Central Estimate

The central estimate projects a mild continued downward drift from 93.3 to ~89.2 by 2031. The strongest counterargument is that corporate profit margins are already at historical extremes, and the recent collapse from 97.8 in 2025Q1 to 93.4 in 2026Q2 was an unusually sharp cyclical drawdown driven by a one-off corporate profit surge. If corporate profits mean-revert toward their 2015–2023 average share of gross value added, the labor share index would naturally rebound back to 95–97 rather than continuing to fall. This potential is reflected in our positively skewed upper confidence band (p90 reaching 97.3 by 2031).

5. Potential 90-Day Updates

Over the next 90 days, the forecast would update significantly if:

• The Q3 2026 preliminary Productivity and Costs release on November 5, 2026 prints outside the [92.5, 94.0] range.
• The BEA Advance GDP and corporate profits release in late October 2026 reveals either a sharp drop in corporate profit margins (pushing near-term labor share up toward 94.5+) or an unexpected surge in productivity above 4.0% accompanied by falling real hourly compensation (pushing the index toward 92.0).

Sources Cited

• U.S. Bureau of Labor Statistics, Productivity and Costs, Second Quarter 2026 (Revised, Sept 3, 2026): https://www.bls.gov/news.release/prod2.nr0.htm
• U.S. Bureau of Labor Statistics, Productivity and Costs, Second Quarter 2026 (Preliminary, Aug 6, 2026): https://www.bls.gov/news.release/archives/prod2_08062026.htm
• U.S. Bureau of Labor Statistics, Time Series Database (Series PRS85006173): https://download.bls.gov/pub/time.series/pr/pr.data.0.Current
• Federal Reserve Bank of St. Louis, FRED / ALFRED (Nonfarm Business Sector: Labor Share for All Workers, PRS85006173): https://fred.stlouisfed.org/series/PRS85006173
• Federal Reserve Bank of Atlanta, GDPNow Model Tracker (Sept 10, 2026): https://www.atlantafed.org/research-and-data/data/gdpnow
• U.S. Bureau of Labor Statistics, The Employment Situation — August 2026 (Sept 4, 2026): https://www.bls.gov/news.release/empsit.nr0.htm
• U.S. Bureau of Labor Statistics, Consumer Price Index — August 2026 (Sept 11, 2026): https://www.bls.gov/news.release/cpi.nr0.htm

Q7 — BLS nonfarm business labor share index (PRS85006173, 2017=100), preliminary release values, 2026Q3–2031Q3

1. What the resolving figure actually is, and the current anchor

The BLS Productivity and Costs news release prints the labor share as a percentage (e.g., "52.9 percent"), while the index (2017=100) appears in the accompanying sector data files that are published simultaneously and mirrored to FRED. The index satisfies the identity:

LS index = 100 × (unit labor cost index ÷ value-added output price deflator index)

Verified against the current data: 2026Q2 revised ULC 124.0 ÷ deflator 132.7 × 100 = 93.44 ≈ FRED's 93.446 (https://fred.stlouisfed.org/graph/fredgraph.csv?id=PRS85006173, retrieved 2026-09-14).

Latest quarter (2026Q2), preliminary release of Aug 6, 2026 (https://www.bls.gov/news.release/archives/prod2_08062026.htm): nonfarm business productivity +1.4% SAAR, hourly compensation +2.7%, ULC +1.3%, value-added output price deflator +7.0% SAAR (+4.9% y/y), value-added nonlabor payments +14.0%. Table 2 indexes (2017=100): productivity 120.0, hours 105.9, hourly comp 148.8, ULC 124.0, deflator 132.6 → preliminary LS index ≈ 93.5, labor share 52.9%, "the lowest level in the series which begins in the first quarter of 1947." The Sep 3, 2026 revised release lowered it to 93.446 (52.8%).

So the anchor for the forecast is ≈93.5 (2026Q2 preliminary), not 93.446.

Recent preliminary-vs-revised gaps (relevant because the question resolves on the first release): 2026Q2 preliminary ≈93.51 vs revised 93.446 (+0.07); 2026Q1 preliminary 54.1% ≈95.7 vs revised/final 94.843 (+0.86, driven by a 0.5pp upward productivity revision). First releases have recently printed above the later revision; I therefore do not subtract anything, and I widen bands accordingly. BLS itself notes the third estimate of productivity growth differs from the first by −1.1 to +1.4 pp about 80% of the time.

Release calendar (https://www.bls.gov/schedule/news_release/prod2.htm): Q3 2026 (P) Nov 5, 2026; Q3 2026 (R) Dec 8, 2026. Note the question's timing note says Q4 2026 preliminary comes in "early February 2027," but the observed BLS pattern puts Q4 preliminary in early March (Q4 2025 (P) = Mar 5, 2026). More importantly, the Q4 preliminary release is the one that incorporates the annual CES benchmark and re-benchmarked hours ("Data in this release incorporate the annual benchmark revision… Quarterly measures of hours worked and related measures—including productivity—were revised back at least 5 years," https://www.bls.gov/news.release/archives/prod2_03052026.htm, Mar 5, 2026), and the Q3 (November) release typically absorbs BEA's September annual NIPA update (BEA's 2026 annual update began Sep 30, 2026). Both can shift the level of the index by a few tenths in the first release itself. I treat the number as printed in the first release as the resolving value and add extra width to Q4 and to 2026Q3.

2. Why the series has been falling, and the arithmetic that drives the path

Decomposition (y/y, 2026Q2): LS growth = hourly comp growth − productivity growth − VA deflator growth = 3.7 − 2.2 − 4.9 = −3.4%/yr, matching the observed move from 96.723 (2025Q2) to 93.446 (−3.4%). The decline is not a wage story — it is a price-level shock: a deflator running +4.7 to +7.0% SAAR in H1 2026 (tariff pass-through plus the spring energy spike) against hourly compensation growing only ~2–2.7% SAAR. Real hourly compensation fell 3.1% SAAR in Q2 and was −0.1% y/y.

Long-run context: LS index fell at ~−0.66%/yr since 2000 and ~−0.98%/yr since 2019Q4 (99.573 → 93.446). After the 2022 shock it fell ~2.2 points and then plateaued at 96.2–96.9 for eight quarters (2023Q1–2024Q4) before resuming its decline in 2025. Quarterly changes have σ ≈ 1.0 index point; 20-quarter changes have mean ≈ −1.2, σ ≈ 3.5, range −8.1 to +5.7. The series is strongly countercyclical (2020Q2 104.2; 2008Q4 106.1).

3. Current macro environment (as of 2026-09-14)

• Oil/war: the US–Iran conflict has intensified; Brent passed $100 on Sep 9 and was ~$107–113 in the days before my cutoff (https://www.theguardian.com/business/2026/sep/09/oil-prices-rise-iran-war-brent-crude-inflation-higher-interest-rates; https://www.cbsnews.com/live-updates/iran-war-us-strait-of-hormuz-oil-gas-price-strikes/). Trump stated the war and high oil prices will not end "until after the November races" (https://www.bbc.com/news/articles/c5ydld0gykpo, ~Sep 10, 2026). This directly raises the Q3 and Q4 2026 output-price deflator — the term that is crushing the labor share.
• Inflation: CPI-U 3.4% y/y in August 2026 (core 2.4%, energy +16.3%). The EIA STEO (Sep 3, 2026, before the latest spike) had the GDP deflator at +1.2% SAAR in Q3 2026 and +2.7% in Q4, drifting to ~1.5–2.1% SAAR through 2027 with Brent ~$90 in 2H26 and ~$74 in 2027 — i.e., forecasters expect the price shock to fade, which is the main mean-reversion channel.
• Labor market: unemployment 4.1% (Aug), July NFP −23k with −103k of net prior revisions, 12-month average payroll growth only ~+31k/month (https://www.reuters.com/business/economy/us-nonfarm-payrolls-fall-july-unemployment-rate-eases-41-2026-08-07/). AHE +3.1% y/y and decelerating; ECI wages ~2.9–3.4%. Slack is building → nominal compensation growth should keep decelerating, which is bearish for the labor share unless the deflator falls faster.
• Policy: Fed Chair Warsh, funds 3.50–3.75%, with a hike widely expected at the Sep 16 meeting (after my cutoff); June SEP: PCE 3.6/2.3/2.0, core 3.3/2.5/2.1, GDP 2.2/2.3/2.2, UER 4.3. Tariff regime is chaotic (SCOTUS struck the IEEPA tariffs in Feb 2026; refunds made net tariffs −$3.5B in Q2 vs +$71B in Q1; replaced by Sec 301/338 measures).
• Commentary consensus: record-low labor share, record profit margins (~14.9% of GDP), and analysts explicitly not expecting near-term wage catch-up (PIMCO Macro Signpost, Sep 2026: ULC ~1% y/y vs core PCE 3.3% is the profit-share story; AI substitutes labor, weak unions). Fortune (Sep 4, 2026): "I don't think there's a floor." CNBC (Sep 13, 2026): wage growth slowing and failing to keep pace with inflation, with AI's role drawing attention (https://www.cnbc.com/2026-09-13/ai-jobs-pay-inflation.html). Counterpoint: Tax Foundation (Sep 2026) argues part of the measured decline is a compositional/measurement artifact — it doesn't change the published index, but it does caution against extrapolating the trend indefinitely.

4. Reference class and the median path

Reference class = the historical distribution of quarterly and 20-quarter changes in PRS85006173 (σ ≈ 1.0 per quarter; −8.1 to +5.7 over 20 quarters), combined with the identity LS growth = comp − productivity − deflator and the EIA/Fed disinflation path.

Median logic, quarter by quarter:

• 2026Q3 (Nov 5 release): deflator still hot (oil >$100 for most of September, tariff pass-through) at ~+3.5 to +4.5% SAAR; hourly comp ~+3.0%; productivity ~+2 to 2.5% (GDPNow had Q3 real GDP at +4.4%, and a productivity surge lowers ULC and hence the labor share). LS change ≈ −2.5 to −3% SAAR → 92.7. Additional level risk from the BEA annual NIPA update folded into this release.
• 2026Q4: oil stays elevated through the midterms per the administration's own framing; CES benchmark lands in the March 2027 first release → 92.0 with a wider band.
• 2027: inflation normalizes (EIA: Brent ~$74, CPI drifting to ~1.2% y/y by 2027Q4), so the deflator-driven decline ends, but hourly comp decelerates with slack and AI-enabled productivity holds near 2%, so LS keeps drifting down ~−0.5%/yr → 91.6 / 91.3 / 91.1 / 90.9.
• 2028–2031: structural drift of about −0.3 to −0.4% per quarter annualized (~−0.6%/yr): capital deepening and AI labor substitution, weak bargaining power, and elevated margins, offset by partial real-wage catch-up as the price shock is fully absorbed. Median ends at 89.0 in 2031Q3.

Overall the median path implies −0.95%/yr from 2026Q2 to 2031Q3, essentially the post-2019Q4 trend rate and only modestly steeper than the post-2000 average — deliberately not extrapolating the current −3.4%/yr, which I treat as a one-off price-level shock rather than a new trend.

5. Bands

Built from cumulative σ ≈ 1.05 × √n for the quarterly-change process, plus a growing trend/regime-uncertainty term, plus explicit width for benchmark and preliminary-release noise (±1.9 at 2026Q3, widening to −7.0/+6.6 at 2031Q3). The upper tail is where a recession lives: LS is countercyclical and rose 4–5 points within a year in 2008–09 and 2020, so a downturn in 2027–2031 plausibly pushes the index to 94–96. The lower tail is the AI-displacement/persistent-inflation scenario (−2%/yr sustained → low 80s). Historical 20-quarter σ of ~3.5 would justify ±4.5; I use ~±6.5 at five years because the sample contains no episode of AI-scale labor substitution and because the resolution is on a noisy first-release number.

6. Strongest case against my central estimate

Mean reversion plus a cyclical rebound. The labor share plateaued for two full years after the 2022 shock; the current decline is caused by a relative price shock (tariffs, oil), and once energy normalizes — EIA has Brent at $74 in 2027 and CPI at ~1.2% y/y by 2027Q4 — the deflator term flips from −3.4%/yr drag to roughly neutral, while nominal compensation keeps growing ~3.5%. If productivity simultaneously slows with the weaker labor market, LS growth = comp − prod − deflator could turn positive, and a recession (a hawkish Fed hiking into an oil shock, with payrolls already negative in July) would produce a 3–5 point spike. That is the single most likely way my median path proves too low; I carry it in the p90 (93.9–95.6) rather than the median, because a recession is not the modal single path and spikes partially reverse.

7. What would move this forecast materially (>1 index point on near horizons, >half the band on distant ones)

• A US–Iran ceasefire / Hormuz reopening that drops Brent below $75 → Q3–Q4 2026 deflator much lower → raise 2026Q3/Q4 medians by 0.8–1.5.
• A further oil spike (>$130) or a new tariff escalation → lower Q3 2026–Q2 2027 by 1–2.
• The Sep 16, 2026 FOMC decision/SEP: an aggressive hiking cycle plus a labor-market break would raise the odds of a 2027 recession → shift mass to the p90 and raise 2027–2028 medians.
• Evidence that AI is producing a genuine productivity acceleration (Q3/Q4 2026 productivity prints >3% y/y with comp flat) → lower the whole path.
• A sharp re-acceleration in the ECI/AHE (comp >4.5% y/y) → flatten or reverse the path.
• The size of the BEA annual NIPA update (Sep 30, 2026) and the CES benchmark in the March 2027 release: a level shift >0.4 in the published index would mechanically move every subsequent period.
• Any BLS announcement of a base-year rebasing (none as of Sep 14, 2026) — resolution would require converting with BLS's published factor.
• A funding lapse after Dec 2026 delaying BLS releases (timing only; I assume the first actual publication resolves the period).

Sources

• BLS, Productivity and Costs, Second Quarter 2026, Preliminary, Aug 6, 2026 — https://www.bls.gov/news.release/archives/prod2_08062026.htm
• BLS, Productivity and Costs, Fourth Quarter and Annual Averages 2025, Preliminary (benchmark), Mar 5, 2026 — https://www.bls.gov/news.release/archives/prod2_03052026.htm
• BLS, Schedule of Releases for Productivity and Costs — https://www.bls.gov/schedule/news_release/prod2.htm
• FRED PRS85006173 (revised data through 2026Q2) — https://fred.stlouisfed.org/series/PRS85006173
• EIA, Short-Term Energy Outlook, Sep 3, 2026 — https://www.eia.gov/outlooks/steo/
• Federal Reserve, June 2026 Summary of Economic Projections — https://www.federalreserve.gov/monetarypolicy/fomcprojtabl20260617.htm
• The Guardian, Sep 9, 2026 — https://www.theguardian.com/business/2026/sep/09/oil-prices-rise-iran-war-brent-crude-inflation-higher-interest-rates
• BBC, ~Sep 10, 2026 — https://www.bbc.com/news/articles/c5ydld0gykpo
• Reuters, Aug 7, 2026 (July employment report) — https://www.reuters.com/business/economy/us-nonfarm-payrolls-fall-july-unemployment-rate-eases-41-2026-08-07/
• CNBC, Sep 13, 2026 — https://www.cnbc.com/2026-09-13/ai-jobs-pay-inflation.html
• Fortune, Sep 4, 2026; PIMCO Macro Signpost, Sep 2026; Tax Foundation, Sep 2026 — https://taxfoundation.org/blog/capital-is-not-taking-half-of-americas-income-labor-share/

What is being forecast

The resolving figure is the first (preliminary) release of BLS series PRS85006173, Nonfarm Business Sector Labor Share for All Workers, index 2017=100, SA quarterly, taken from the Productivity and Costs release in which that quarter first appears (Q3 2026: 2026‑11‑05; Q4 2026: early Feb 2027; …; Q3 2031: ~Nov 2031).

Current status (verified from primary sources)

FRED/BLS vintage of 2026‑09‑03 (https://fred.stlouisfed.org/series/PRS85006173, data table https://fred.stlouisfed.org/data/PRS85006173):

Quarter	Index (latest vintage)
2024Q4	96.940
2025Q1	97.871
2025Q2	96.723
2025Q3	95.999
2025Q4	95.727
2026Q1	94.843
2026Q2	93.446

Two important calibrations from the releases themselves:

1. Percent ↔ index mapping. The Sep 3, 2026 revised release (USDL 26‑1434, https://www.bls.gov/news.release/PDF/prod2.PDF) states the labor share "was 52.8 percent in the second quarter of 2026, the lowest level in the series, which begins in the first quarter of 1947." So 93.446 ↔ 52.8%, i.e. the 2017 benchmark ≈ 56.5% of output. The preliminary Q2 figure was 52.9% (Voronoi, Aug 14, 2026: "fell to 52.9% of value added in Q2 2026… the latest quarter is preliminary and can be revised") → ≈ 93.62 index.
2. Preliminary-vs-final wedge is real and can be ~1 index point. Q1 2026 preliminary (May 7, 2026, https://www.bls.gov/news.release/archives/prod2_05072026.htm) reported the labor share at 54.1% ("the lowest recorded value since the series began in 1947") → ≈95.75 index. The June 4 revised release put it at 53.7% (Quartz, https://qz.com/us-worker-productivity-first-quarter-2026-revised-lower-060426) and the latest vintage is 94.843 (53.6%). So Q1's first release sat ~0.9 index points above the eventual value; Q2's sat ~0.17 above. Average wedge ≈ +0.5, with dispersion of ±0.5–1.0.

Mechanism. The 2025–26 drop is a nominal-price/wage squeeze, not a compensation collapse: in Q2 2026 nonfarm business hourly compensation rose only 2.6% (annualized) while the value-added output price deflator rose 7.4% and real hourly compensation fell 3.3%. Productivity (+1.4%) continues to outrun real pay, so the labor-share ratio keeps falling. Reuters (Aug 6, 2026) framed it as an "ongoing productivity boom" producing record lows; CNBC (Sep 13, 2026) notes wage growth "struggling to keep up with inflation," with AI's role drawing attention (https://www.cnbc.com/amp/2026/09/13/ai-jobs-pay-inflation.html). Uprise RI (Sep 4, 2026) confirms the 52.8% revised print.

Reference class and base rate

Historical 5‑year (20‑quarter) log changes in this series (my own tabulation from the FRED table): 1947–52 −2.8%, 1952–57 +1.2%, 1957–62 −1.5%, 1962–67 −1.8%, 1967–72 +0.4%, 1972–77 −2.5%, 1977–82 +4.0%, 1982–87 −1.3%, 1987–92 −0.6%, 1992–97 −3.6%, 1997–2002 +1.6%, 2002–07 −4.1%, 2007–12 −5.4%, 2012–17 +1.5%, 2017–22 −1.9%, 2020–25 −4.1%. Mean ≈ −1.0%, sd ≈ 2.6%; the tail is roughly ±5%. Longer-run drifts: −0.27%/yr since 1947, −0.69%/yr since ~2001, −1.44%/yr over the last 5 years, −3.4%/yr over the last year. The last 4 years (−7.2% log) already exceed the worst historical 5‑year window, so the recent regime is genuinely outside the historical envelope — which argues for centring on a continued but decelerating decline rather than extrapolating −3.4%/yr, and for bands wider than the historical sd.

Path construction

Near term (2026Q3): take the latest Q2 vintage (93.446), apply a ~0.8‑point quarterly decline consistent with the recent run rate (Q4'25→Q1'26 −0.88; Q1→Q2 −1.40), then add the observed preliminary wedge (+0.25) → median 92.9. The 10–90 band (±1.7% log, ±1.6 index points) reflects both the possible continuation of ~1‑point quarterly drops and preliminary-value noise of ~1 point.

Medium/long term: I model quarterly log declines decelerating from ~0.55% to ~0.15% per quarter as inflation normalises and nominal wages partially catch up, giving a cumulative ≈ −6.3% over 20 quarters → median 87.05 for 2031Q3 (a −1.3%/yr rate, in line with the last decade's drift but far below the last year's). Bands widen roughly as √h plus a drift-uncertainty term, reaching ±6–7% in log (80.6–93.6) at 2031Q3 — wider than the historical ±5% 5‑year spread, justified by (a) the current deviation from the historical envelope, (b) live regime uncertainty over AI-driven labour displacement, and (c) the possibility of policy/monetary reversal. The 2031Q3 band implies ~10% chance the labor share is at or above today's level within five years.

Causal pathways

• Toward lower values: AI-driven productivity acceleration without matching real wage growth; continued inflation outpacing nominal compensation (as in Q2 2026, deflator +7.4% vs compensation +2.6%); capital deepening and rising markups; weak union/ bargaining power.
• Toward higher values / stabilisation: inflation normalisation with wage catch-up (the 1972–80 pattern); a recession (the series spikes countercyclically — 2020Q2 reached 104.2, 2009 Q1 102.0); labour scarcity/demographics; policy response to a record-low labor share (minimum wage, profit-sharing, AI-related redistribution); a genuine downward revision in the AI-productivity narrative.

Strongest case against my central estimate

That the 2025–26 collapse is largely a transitory inflation artifact: if the 4–6% inflation bout ends and nominal wages catch up (as in 1980, when the labor share recovered ~4% in five years), the series could stabilise in the low 90s and drift sideways rather than continue to 87. That is exactly why my p90 stays near 93–94 rather than projecting continued decline, and why my median embeds strong deceleration (−6.3% over five years vs −7.2% over the last four). The opposite risk — that AI-driven displacement makes −3%/yr the new normal — is why the p10 sits at ~80.6, ~14% below today.

What would move the forecast >10 points of probability (or >half the band)

• The 2026‑11‑05 Q3 2026 preliminary print: a value at or below 91.5 (decline accelerating) would pull the whole path down several index points; a value above 94 (stabilisation) would pull it up.
• A decisive break in core/CPI inflation toward 2% together with ECI/ECEC wage growth holding ≥4% (wage catch-up ⇒ labor-share stabilisation/reversion).
• Evidence that the productivity surge is AI‑driven and accelerating (e.g., 3%+ four-quarter productivity growth persisting through 2027) with flat real wages ⇒ faster decline.
• Any BLS methodology/base change (rebasing factor application), or an annual benchmark revision that materially re-levels the series.
• A recession signal (labour market break) — historically a strong positive shock to this ratio.

Ambiguities / reading adopted

I read "preliminary (first) release" as the figure published in the news release in which the quarter first appears, using that release's vintage (which can differ from later vintages by ~0.2–0.9 index points, as Q1 2026 demonstrated), and I express everything in the 2017=100 index of PRS85006173, converting any labour-share percentage quoted in the release text using the 2017 benchmark (≈56.5% of output).