The August 2026 employment report shows a weakening labor base and only limited protection of household purchasing power. The published raw data show fewer employed people, a smaller labor force and substantially more people outside the labor force than in December 2025, even as the civilian noninstitutional population increased. Goods-producing industries are delivering the strongest real weekly earnings gains. Private services—the largest part of private payroll employment—also remain positive on the NSA year-ending calculation, but trail goods production per employee.
ATN uses not seasonally adjusted data as the primary basis of the numerical employment analysis. Many market desks prefer the raw published series because it has not been smoothed through a seasonal-adjustment model. Recurring cyclicality is addressed by using year-ending, year-over-year and longer-period comparisons. A separate section beneath the raw analysis compares the SA headline with NSA data and explains why ATN prefers the NSA approach.
The January 2026 household survey did introduce new Census population weights. That is explained briefly as a technical comparability note, but this report retains the official published NSA figures available to market participants.
This two-part analysis examines the numerical employment picture first, followed by real weekly earnings per employee and across the private payroll workforce. Particular attention is given to the 26-week unemployment-insurance cliff, which turns long-term unemployment into an immediate household-income and consumption problem.
August employment report: the key findings
- NSA is the primary employment basis: ATN uses the published raw estimates and mitigates recurring seasonality through year-ending and longer-period comparisons. SA employment figures are confined to a separate comparison section.
- The raw YTD picture is negative: from December 2025 to August 2026, the civilian labor force fell by 675,000, employment fell by 1.053 million and the number outside the labor force rose by 1.273 million.
- Population continued to grow: the civilian noninstitutional population increased by 599,000 over the same period while participation declined from 62.1% to 61.7%.
- Long-term unemployment is an income cliff: 1.946 million people had been unemployed for 27 weeks or longer—the point at which regular unemployment insurance has expired for many workers.
- Goods lead services: NSA year-ending real weekly earnings are positive across the major groups, but goods production outperforms the much larger private service sector per employee.
Sources: BLS August employment release, BLS Table A-1, BLS Table A-12 and U.S. Department of Labor unemployment-insurance information.
Part 1: Employment numbers, participation and the labor base
Methodology: NSA as the primary analytical basis
The principal analysis uses not seasonally adjusted data. Seasonally adjusted figures are produced through a statistical model designed to remove recurring calendar and seasonal patterns. The result can be useful for headline month-to-month comparisons, but it is also sensitive to model assumptions, recalculated seasonal factors and later revisions.
ATN therefore prefers the published raw NSA series commonly monitored by market desks. Cyclicality is mitigated through year-ending, year-over-year and longer-period comparisons. The SA headline is examined later in its own section so that the two approaches can be compared without combining them inside a single calculation.
The household survey measures people as employed, unemployed or outside the labor force. The establishment survey measures payroll jobs and provides industry earnings and hours. A person holding two payroll jobs can appear twice in the establishment count but only once in household employment. Self-employment and other coverage differences also matter. This article uses the household survey for the labor-force analysis and the establishment survey for the earnings analysis.
The YTD shadows in the published NSA data
The published NSA figures show a materially negative change from December 2025 to August 2026:
| Measure | December 2025 | August 2026 | YTD change |
|---|---|---|---|
| Civilian noninstitutional population | 274,816,000 | 275,415,000 | +599,000 |
| Civilian labor force | 170,723,000 | 170,048,000 | −675,000 |
| Employed | 163,720,000 | 162,667,000 | −1,053,000 |
| Unemployed | 7,003,000 | 7,381,000 | +378,000 |
| Not in the labor force | 104,094,000 | 105,367,000 | +1,273,000 |
| Participation rate | 62.1% | 61.7% | −0.4 percentage point |
Source: BLS December 2025 NSA household data and BLS August 2026 Table A-1.
The central signal is difficult to dismiss: the civilian population increased by 599,000, yet employment fell by more than one million and the labor force contracted by 675,000. At the same time, the number outside the labor force rose by 1.273 million. A larger population is therefore supporting a smaller employed base, weakening the volume of labor earnings available to the household economy.
The rise in people outside the labor force is consistent with an aging population, but aging is not the only possible cause. Retirement, education, disability, caregiving and labor-force discouragement can all contribute. The economic consequence is nevertheless the same: people outside the labor force do not contribute current labor earnings.
Technical note on the January population controls
The household survey is a sample, not a literal census of every worker. Each response is weighted to represent part of the national population. In January 2026, Census and BLS changed those population weights using updated demographic and migration estimates. NSA data are free of seasonal smoothing, but they still use these survey weights.
BLS published an illustrative December calculation using the new weights to show the size of the statistical break. It was not a second December survey result and it did not replace the official December figures. ATN therefore retains the official published NSA data in the table above—the figures that were available to market participants—while noting that part of the change across January reflects reweighting rather than solely real-world labor flows.
The year-ending NSA comparison confirms the weakness
Comparing August 2026 with August 2025 removes much of the recurring calendar effect without introducing a seasonal-adjustment model:
| Measure | August 2025 | August 2026 | Year-ending change |
|---|---|---|---|
| Civilian noninstitutional population | 274,001,000 | 275,415,000 | +1,414,000 |
| Civilian labor force | 171,035,000 | 170,048,000 | −987,000 |
| Employed | 163,288,000 | 162,667,000 | −621,000 |
| Unemployed | 7,747,000 | 7,381,000 | −366,000 |
| Not in the labor force | 102,966,000 | 105,367,000 | +2,401,000 |
| Participation rate | 62.4% | 61.7% | −0.7 percentage point |
Source: BLS Table A-1, NSA columns.
The unemployment count declined, but so did employment and the labor force. This is not a clean expansion in opportunity: fewer people were classified as unemployed partly because substantially more people were outside the labor force. The year-ending data therefore reinforce the YTD warning rather than presenting a contradictory monthly signal.
Why the Non-Labor-Force Population Is Rising: America Is Aging
The rise in the number of people outside the labor force has a major structural component: the U.S. population is aging. In August 2026, Americans aged 65 and over represented 23.7% of the civilian noninstitutional population aged 16 and over, but they accounted for 50.4% of everyone outside the labor force. Including those aged 60 to 64, people aged 60 and over represented approximately 58.4% of the non-labor-force total.
The longer-term change is even clearer. Between 2019 and 2025, the civilian noninstitutional population increased by 14.478 million, of which 8.763 million—or approximately 60.5%—came from the 65-and-over group. Over the same period, the number of people aged 65 and over who were outside the labor force increased by 7.661 million, exceeding the 7.210 million net increase across all age groups.
However, aging does not explain all of the August 2026 weakness. Year over year, the civilian noninstitutional population increased by 1.414 million, while the non-labor-force population increased by 2.401 million and the labor force declined by 987,000. The demographic trend explains much of the long-term direction, but the latest figures also indicate a broader decline in participation.
Read the full demographic analysis: America Is Aging Out of the Labor Force: What the August 2026 Data Really Shows.
Sources: BLS August 2026 employment status by age, BLS 2025 annual age data and BLS 2019 annual age data.
Separate comparison: NSA versus the seasonally adjusted headline
The primary analysis above uses NSA figures. The table below now places the official SA calculation underneath it, solely to show how strongly the seasonal model changes the apparent YTD picture:
| Measure | Published NSA change | Published SA change |
|---|---|---|
| Civilian noninstitutional population | +599,000 | +599,000 |
| Civilian labor force | −675,000 | −1,718,000 |
| Employed | −1,053,000 | −1,246,000 |
| Unemployed | +378,000 | −472,000 |
| Not in the labor force | +1,273,000 | +2,317,000 |
| Participation rate | −0.4 percentage point | −0.8 percentage point |
Sources: BLS December 2025 household data and BLS August 2026 Table A-1.
The contrast is substantial. The two methods even produce opposite directions for unemployment: the raw NSA count rises by 378,000, while the SA series falls by 472,000. The model also more than doubles the apparent contraction in the labor force and the increase outside it.
This does not prove deliberate falsification. Seasonal adjustment is an openly documented statistical process intended to smooth recurring patterns. It does demonstrate why the adjusted headline should not be treated as if it were the observed raw number. Seasonal factors are estimated, can be recalculated and may be revised.
Why ATN prefers NSA
ATN prefers the NSA series because it preserves the published survey movement before seasonal modelling. Its weakness—recurring cyclicality—is handled through August-to-August year-ending comparisons and longer observation windows. This allows the reader to see the raw data, understand the comparison period and reproduce the result without depending on a seasonal model that can materially change both the scale and direction of the signal.
The SA figures remain useful for understanding the headline likely to move markets immediately. The NSA series is preferred here for evaluating the underlying labor base and household-income capacity.
Unemployment by education: the improvement is uneven
| Educational attainment | August 2026 | December 2025 | August 2025 |
|---|---|---|---|
| Less than a high school diploma | 4.1% | 6.0% | 6.2% |
| High school graduate, no college | 4.4% | 4.0% | 4.3% |
| Some college or associate degree | 3.9% | 3.6% | 3.4% |
| Bachelor’s degree or higher | 3.0% | 2.6% | 3.0% |
Sources: August Table A-4 and December release, Table A-4. These published comparisons cross the population-control change. December-to-August also compares different points in the calendar cycle, which is why the August-to-August reading receives greater weight.
The largest rate improvement is among people without a high school diploma. Yet their NSA employment level fell from 8.092 million in August 2025 to 7.522 million in August 2026, alongside a smaller labor force. That means the lower unemployment rate alone cannot be described as proof of a hiring recovery in this group. The rate increased for high-school graduates and those with some college, while the graduate category was unchanged from a year earlier. The outcome is mixed.
The crucial 26-week unemployment-insurance cliff
The most important duration figure is the number unemployed for 27 weeks or longer. This is not merely another percentage in the report: it broadly marks the point at which regular state unemployment insurance has ended.
The U.S. Department of Labor states that regular unemployment insurance can be paid for a maximum of 26 weeks in most states. Several states provide fewer weeks. Additional benefits may become available during periods of high unemployment, but they depend on statutory extension triggers and are not a permanent continuation of ordinary state benefits.
| Measure | August 2025 | August 2026 | Change |
|---|---|---|---|
| Unemployed 27 weeks or longer | 1,938,000 | 1,946,000 | +8,000 |
| Share of unemployment | 25.0% | 26.4% | +1.4 percentage points |
Source: BLS Table A-12, NSA columns.
In August 2026, nearly 1.95 million people had therefore been unemployed beyond the normal maximum duration of regular benefits in most states. Some will have reached the benefit limit earlier because their state provides fewer than 26 weeks or because they did not qualify for the maximum entitlement.
The BLS duration data do not identify which individuals received unemployment insurance, and total household income may include savings, another earner or other assistance. Nevertheless, the worker’s own labor earnings are absent and regular wage-replacement benefits may have expired. This creates a direct household cash-flow cliff, forcing greater reliance on savings, credit or other household income while reducing potential consumer spending.
Long-term unemployment is also harder to reverse. Skills can deteriorate, professional networks weaken and employers may become more reluctant to hire workers after a prolonged absence. The 27-week figure is therefore more economically significant than the unemployment rate alone: it measures a group facing both labor-market detachment and the likely exhaustion of its principal temporary income support.
Conclusion—numerical employment analysis
The published NSA evidence remains negative. Both the YTD and year-ending comparisons show a contracting employed base, declining participation and a sharp increase in the number outside the labor force while the civilian population continues to grow.
The decline in total unemployment does not overturn that conclusion because employment and labor-force membership also declined. Most importantly, almost 1.95 million people have been unemployed for at least 27 weeks, placing many at or beyond the ordinary unemployment-insurance limit. The central risk is therefore not merely statistical unemployment—it is the loss of household cash flow, purchasing power and attachment to the workforce.
Part 2: Real weekly earnings and the economy’s spending power
One employee’s purchasing power versus the whole payroll
An unemployment rate cannot tell us whether pay packets are keeping up with living costs. Equally, an average pay packet cannot tell us whether total labor income is growing. Both the earnings average and the number of payroll jobs matter.
Real weekly earnings per employee are nominal average weekly earnings divided by a consumer-price index. Aggregate real weekly payroll is employment multiplied by average weekly earnings, then divided by the same price index. The latter measures purchasing power across the covered payroll jobs, not total national income or the income of every employed person.
For example, if average real earnings rise 1% while employment falls 1%, aggregate real payroll is roughly unchanged: 1.01 × 0.99 = 0.9999. A stronger average can therefore coexist with a weak total. Conversely, adding jobs can lift total purchasing power even when the average worker’s gain is very small.
Year-ending NSA real earnings: goods production leads
For consistency with the numerical employment analysis, the principal monetary calculation also uses not seasonally adjusted data and compares August 2026 with August 2025.
| Private-sector industry | Per employee | Aggregate payroll |
|---|---|---|
| Total private | +1.81% | +2.33% |
| Goods-producing | +2.14% | +2.80% |
| Construction | +1.93% | +3.20% |
| Manufacturing | +1.90% | +2.18% |
| Private service-providing | +1.69% | +2.18% |
Calculation basis: ATN calculations using NSA payroll employment from BLS Table B-1b and NSA nominal average weekly earnings from BLS Table B-3b, deflated by the rounded 3.4% annual all-items CPI increase. These are approximate ATN calculations rather than an additional official BLS real-earnings series.
The annual real gain per employee is calculated as [(current weekly earnings ÷ year-earlier weekly earnings) ÷ 1.034 − 1] × 100. For aggregate payroll, the earnings ratio is also multiplied by the matching employment ratio before deflation. Construction and manufacturing are components of goods-producing employment and should not be added to that total.
The NSA year-ending calculation is positive across every major group in the table. Goods production remains the strongest broad sector per employee, while construction records the largest aggregate gain because both real weekly earnings and payroll employment contribute.
Private services also show positive year-ending purchasing-power growth on the NSA basis, but the per-employee gain trails goods production. Because services account for the overwhelming majority of private payroll employment, even a modest difference in real earnings growth materially affects the purchasing power of the household sector.
These figures are consistent with firmer earnings in productive industries, but they do not by themselves prove reshoring, stronger management expertise or higher physical output. Hours, occupational mix, wage rates and employment all influence average and aggregate earnings.
Separate SA earnings comparison
BLS reports that real average weekly earnings for all private employees increased 0.2% from July to August and 0.3% from August 2025, seasonally adjusted. Real hourly earnings fell 0.1% over the month, but a longer average workweek supported weekly purchasing power. For production and nonsupervisory employees, real weekly earnings fell 0.1% over the month and rose 0.1% over the year.
BLS uses CPI-U for all employees and CPI-W for production and nonsupervisory employees. An analysis that applies CPI-U to both groups is a separate calculation and must be labelled accordingly. Source: BLS Real Earnings, August 2026.
The difference from ATN’s NSA calculation is material: the official SA annual result for all private employees is +0.3%, whereas the approximate NSA year-ending calculation above is +1.81%. This reinforces the methodological conclusion from Part 1. The seasonal model can materially change the magnitude of the result, so ATN uses the NSA year-ending calculation as its primary analytical measure and presents the SA release only as a separate headline reference.
Both measures describe changing groups of jobs rather than the pay history of one continuously employed person. A different mix of higher- and lower-paid jobs can move an average even without an equivalent pay change for every worker.
Why services matter so much
On an NSA basis, August private service-providing employment was 114.546 million out of 136.425 million private payroll jobs: approximately 84.0% of private payroll employment. That is not 84% of the entire labor market, which also includes government employment and workers outside the payroll survey. Source: BLS Table B-1b.
Such a large weight means the service-sector result dominates the household picture. Its real year-ending gain is positive, but it trails goods production per employee. Sustained improvement requires real earnings, paid hours and employment to strengthen across services as well as goods production.
Energy is an important squeeze, but the duration is uncertain
August CPI rose 0.4% over the month and 3.4% over the year. Energy rose 2.1% over the month and 16.3% over the year; gasoline accounted for more than one-third of the monthly increase in the overall index. Core CPI, excluding food and energy, rose 0.3% over the month and 2.4% over the year. Source: BLS August CPI release.
That pattern supports the view that energy is placing a material burden on purchasing power. It does not establish that every component of inflation comes from energy, or that the effect will necessarily be brief. Even if the inflation rate later slows, households do not recover the earlier loss in purchasing power unless earnings catch up or relevant prices decline.
Conclusion—monetary and real-earnings analysis
Real weekly earnings remain too weak and too uneven to indicate a broad inflationary wage boom. Goods-producing industries show the strongest purchasing-power gains, while the much larger private service sector remains positive but trails goods production on a per-employee basis.
The distinction between one employee and the entire workforce is critical. A modest gain in average real weekly earnings does not guarantee stronger aggregate household purchasing power if employment, paid hours or labor-force participation are contracting. The numerical employment weakness identified in Part 1 therefore limits the monetary benefit of higher average pay.
The report points to an economy in which energy costs are doing more immediate damage to purchasing power than excessive real wage growth. Higher interest rates cannot directly create energy supply, and further tightening risks weakening employment and household cash flow while the inflation source remains substantially external to wages.
Does this justify the Federal Reserve’s rate increase?
On September 16, the FOMC raised its federal funds target range by 25 basis points to 3.75%–4.00%. Its statement cited elevated inflation and the objective of returning it to 2%. It did not identify a mechanical requirement to follow the 13-week Treasury yield. Source: September FOMC statement.
The strongest criticism of tighter policy in this setting concerns its transmission: higher borrowing costs cannot directly produce more oil or remove a physical supply disruption. They can, however, restrain spending and investment while households are already dealing with weaker real purchasing power.
That is a policy trade-off, not proof that interest rates are irrelevant. Monetary policy affects demand, credit conditions and expectations. It can influence whether an initial cost shock becomes persistent across a wider range of prices. The Fed’s explanation of monetary-policy transmission describes these broader channels.
Nor do weak real wages, by themselves, rule out wage-related inflation pressure. Real earnings already subtract consumer inflation; assessing the cost pressure arising from compensation requires nominal pay, productivity and unit labor costs. A rise in pay backed by equivalent productivity gains has different implications from one without those gains.
ATN’s assessment: these earnings figures provide little evidence of a broad household purchasing-power boom. They strengthen the case for scrutinizing the growth cost of further tightening. They do not establish that the Fed was forced to hike by Treasury bills, that wages are the only valid reason to raise rates, or that monetary policy has no influence on the persistence of inflation.
What futures and equity traders should watch next
The relevant market question is whether modest payroll and earnings gains can continue while energy inflation eases. The following are conditional scenarios, not predictions of a particular trading session:
| Development | Economic implication | Market focus |
|---|---|---|
| Energy eases; jobs and paid hours hold up | Real spending power can recover without an acceleration in nominal wages. | A more supportive growth-and-inflation mix for broad equities; watch Treasury yields. |
| Energy remains high; services earnings stay weak | Household budgets remain squeezed and discretionary spending becomes more vulnerable. | Consumer exposure, margins and the relative performance of energy producers. |
| Employment and hours weaken together | Aggregate real payroll can deteriorate even if the average wage remains firm. | Growth concerns can outweigh any benefit from lower expected interest rates. |
| Broader inflation persists despite weak purchasing power | The policy trade-off becomes more difficult. | For NQ and other rate-sensitive assets, monitor yields and earnings expectations together. |
The next scheduled national employment report, covering September, is due October 2 at 8:30 a.m. ET / 7:30 a.m. CT. September CPI and real earnings follow on October 14 at the same time. Check the BLS release calendar for scheduling updates.
Report conclusion: the labor base and household cash flow remain under pressure
The published NSA data do not show a broad labor-market recovery. From December through August, employment fell by 1.053 million, the civilian labor force contracted by 675,000 and the number outside the labor force rose by 1.273 million, even as the civilian population increased by 599,000. The August-to-August comparison confirms the same underlying weakness while reducing ordinary seasonal distortion.
The separate SA comparison explains why the official headline can tell a different story. Seasonal adjustment materially changes both the size and, in the case of unemployment, the direction of the movement. ATN therefore prefers NSA data for structural analysis and uses year-ending comparisons to manage cyclicality, while retaining SA only as a clearly separated market-headline reference.
The most urgent finding is the 26-week income cliff. Nearly 1.95 million people had been unemployed for 27 weeks or longer. In most states, regular unemployment insurance does not continue beyond 26 weeks, and several states provide fewer weeks. For many workers, long-term unemployment therefore coincides with the loss of their principal temporary wage replacement, intensifying pressure on savings, credit and household consumption.
Real weekly earnings provide only partial protection. Goods-producing industries show stronger gains, while services dominate private employment and remain positive but trail goods production on a per-employee basis. A rise in the average wage cannot fully support the economy if the number of employed people, paid hours and labor-force participation continue to weaken.
The combined conclusion is clear: the economy is not experiencing a broad wage-driven purchasing-power boom. It is experiencing a weaker labor base, a serious long-term-unemployment cash-flow problem and an energy-led squeeze on real income. Those conditions make the growth cost of further monetary tightening increasingly important.