ADP estimates a 7,000-job August decline outside education and health

By Cumulant Research | September 2, 2026

[ADP's August estimate](https://mediacenter.adp.com/2026-09-02-ADP-National-Employment-Report-Private-Sector-Employment-Increased-by-38,000-Jobs-in-August) says US private employers added a net 38,000 jobs. Education and health servicesEducation and health servicesA broad private-industry group combining educational services with health care and social assistance. [BLS](https://www.bls.gov/news.release/archives/empsit_08072026.htm) added 45,000. Subtracting the second number from the first leaves a 7,000-job decline across the other nine industry groups combined.

A contribution larger than the total is not a mathematical mistake. Think of a group restaurant bill: one diner can pay more than the final check if another diner receives a refund. Education and health contributed 45,000 jobs, while the gains and losses in the other groups reduced the final increase to 38,000. [ADP industry data](https://mediacenter.adp.com/2026-09-02-ADP-National-Employment-Report-Private-Sector-Employment-Increased-by-38,000-Jobs-in-August)

The decline was not universal. Construction added 12,000 jobs, leisure and hospitality added 16,000, financial activities added 6,000 and other services added 6,000. Manufacturing lost 17,000, professional and business services lost 16,000, natural resources and mining lost 5,000, trade, transportation and utilities lost 5,000, and information lost 4,000. Five of the 10 industry groups therefore grew and five contracted. [ADP August report](https://mediacenter.adp.com/2026-09-02-ADP-National-Employment-Report-Private-Sector-Employment-Increased-by-38,000-Jobs-in-August)

Calling the 7,000 decline a hiring collapse would mix up two measurements. Payroll change is the net result after workers enter and leave payrolls. JOLTSJOLTSThe Job Openings and Labor Turnover Survey, which separately estimates openings, hires, quits, layoffs and other departures. [BLS](https://www.bls.gov/news.release/archives/jolts_09012026.htm) separately counts hires and separations, while ADP estimates the resulting employment level. [BLSBLSThe US Bureau of Labor Statistics, the federal agency that produces the government's monthly Employment Situation and JOLTS reports. [BLS](https://www.bls.gov/bls/empsitquickguide.htm) JOLTS definitions](https://www.bls.gov/news.release/archives/jolts_09012026.htm)

A bathtub is the useful analogy. Hiring is water entering through the tap, separations are water leaving through the drain, and payroll change is the rise or fall in the water level. A nearly level tub could mean both flows are weak or both are strong.

The freshest direct evidence on those flows predates ADP's August estimate by one month. In July, BLS estimated 7.3 million job openings, 5.1 million hires and 5.1 million total separations. It also estimated 3.1 million quits and 1.7 million layoffs and discharges, describing each national total as little changed. Professional and business services hiring fell by 188,000, but nationwide layoffs did not jump. [BLS July JOLTS](https://www.bls.gov/news.release/archives/jolts_09012026.htm)

That combination is consistent with a low-hirehireA worker added to an establishment's payroll during the month, including a newly hired or rehired worker. [BLS JOLTS](https://www.bls.gov/news.release/archives/jolts_09012026.htm), low-fire market in which employers retain many current workers but create relatively few openings. It is not proof of that explanation, and July turnover cannot determine what happened in August. [St. Louis Fed explanation](https://www.stlouisfed.org/on-the-economy/2026/jun/its-still-business-cycle-young-adult-workers-low-hire-low-fire-labor-market)

ADP's weekly tracker also remained positive before the monthly release. Its four-week moving average rose from 8,250 jobs per week through July 25 to 9,500 through August 1 and 11,750 through August 8. Because the windows overlap, those readings cannot be added together as a separate monthly estimate. [ADP NER Pulse](https://mediacenter.adp.com/2026-08-25-ADP-National-Employment-Report-Preliminary-Estimate-for-August-8%2C-2026)

The latest available BLS payroll report covers July, not August. It estimated that total private payrolls increased by 30,000 while private education and health increased by 25,000. The implied change across the remaining private industries was a gain of 5,000. [BLS July Employment Situation](https://www.bls.gov/news.release/archives/empsit_08072026.htm)

That is useful continuity but not confirmation. ADP and BLS used different data and measured different months. ADP redesigned its report in 2022 as an independent employment measure rather than a forecast of the BLS result, and the two estimates can diverge in any given month. [ADP methodology](https://www.adpresearch.com/research/how-representative-is-adp-employment-data)

The apparent precision of the 5,000 residualresidualThe amount left after subtraction, here calculated as total private employment minus private education and health employment. is misleading. BLS reported a 90 percent confidence intervalconfidence intervalA range around a sample estimate that communicates statistical uncertainty instead of treating the point estimate as exact. [BLS](https://www.bls.gov/charts/employment-situation/otm-employment-change-by-industry-confidence-intervals.htm) of plus or minus 113,000 around July's 30,000 total-private estimate and plus or minus 51,000 around the 25,000 education-and-health estimate. A valid interval for their difference cannot be obtained by subtracting or simply combining those two margins because the estimates are statistically related. [BLS confidence intervals](https://www.bls.gov/charts/employment-situation/otm-employment-change-by-industry-confidence-intervals.htm)

We tested whether ADP concentration months have historically predicted a later contraction. Because ADP retooled the report as an independent measure in 2022, we examined 42 possible signal months from August 2022 through January 2026. January is the last month for which a complete six-month BLS outcome was available by the July 2026 data cutoff. [ADP methodology](https://www.adpresearch.com/research/how-representative-is-adp-employment-data) [ADP and BLS series](https://fred.stlouisfed.org/data/ADPMNUSNERSA)

A month counted as a signal when ADP total private employment increased but education and health increased by more than the total. Seven of the 42 months qualified. For every month, we calculated BLS private employment outside education and health and asked whether that residual was lower three and six months later. [ADP total private](https://fred.stlouisfed.org/data/ADPMNUSNERSA) [ADP education and health](https://fred.stlouisfed.org/data/ADPMINDEDHLTNERSA) [BLS total private](https://fred.stlouisfed.org/data/USPRIV) [BLS education and health](https://fred.stlouisfed.org/data/USEHS)

Three of the seven signal months, or 43 percent, were followed by contractions at each horizon. Among the other 35 months, 16, or 46 percent, were followed by contractions. The signal therefore did not outperform the ordinary-month baseline in this sample. The three-percentage-point difference is not meaningful evidence of protection either, given the very small number of signals. [Underlying ADP and BLS series](https://fred.stlouisfed.org/data/ADPMNUSNERSA)

The individual cases show why the average is inconclusive. The residual fell after the April 2024, August 2024 and February 2025 signals, but rose after the September 2023, October 2025, December 2025 and January 2026 signals. The same three-versus-four split appears at both horizons. [ADP and BLS series](https://fred.stlouisfed.org/data/USPRIV)

The historical pattern identifies poor employment breadth, but it did not predict subsequent contraction better than the baseline.

ADP's scale is substantial: its report draws on anonymized weekly payroll information covering more than 26 million private-sector workers. But it begins with a convenience sampleconvenience sampleA sample assembled from available participants, such as ADP clients, rather than employers randomly selected from the entire economy. [ADP methodology](https://www.adpresearch.com/research/how-representative-is-adp-employment-data) of ADP clients rather than a random census of every employer. ADP says its sample leans toward larger establishments, manufacturing and the Northeast. [ADP report](https://mediacenter.adp.com/2026-09-02-ADP-National-Employment-Report-Private-Sector-Employment-Increased-by-38,000-Jobs-in-August) [ADP methodology](https://www.adpresearch.com/research/how-representative-is-adp-employment-data)

ADP reweights its observations by industry, state and employer size using QCEWQCEWThe Quarterly Census of Employment and Wages, an administrative count covering more than 95 percent of US jobs. [BLS](https://www.bls.gov/cew/overview.htm) records. That adjustment addresses differences the model can observe, but it does not turn the sample into a complete census or eliminate monthly estimation error. [ADP methodology](https://www.adpresearch.com/research/how-representative-is-adp-employment-data)

The more comprehensive QCEW records also arrive too late to answer an immediate news question. QCEW covers more than 95 percent of US jobs, but ADP notes that the data are released five or six months after a quarter ends. [BLS QCEW overview](https://www.bls.gov/cew/overview.htm) [ADP methodology](https://www.adpresearch.com/research/how-representative-is-adp-employment-data)

Recent benchmark evidence illustrates the revision risk. On August 28, BLS estimated that its March 2026 total-private employment level would eventually be revised down by 178,000, including a 96,000 downward revision in private education and health. The implied adjustment outside that category is 82,000 downward. This concerns the March employment level, not jobs lost in August, and BLS said the final revision would enter the official estimates in February 2027. [BLS benchmark summary](https://www.bls.gov/news.release/prebmk.nr0.htm) [BLS benchmark table](https://www.bls.gov/news.release/prebmk.t01.htm)

ADP also publishes education and health as one combined group. Its August table cannot show how the 45,000 gain was divided among private schools, health-care providers and social-assistance organizations. [ADP industry table](https://mediacenter.adp.com/2026-09-02-ADP-National-Employment-Report-Private-Sector-Employment-Increased-by-38,000-Jobs-in-August)

The August BLS Employment Situation is scheduled for September 4 at 8:30 a.m. Eastern. It will provide the first independent monthly payroll estimate covering the same period as ADP's August report. [BLS release calendar](https://www.bls.gov/schedule/news_release/empsit.htm)

• Corroborated, BLS private employment outside education and health is zero or negative, and at least five of the nine other comparable industry groups are flat or declining.
• Not corroborated, The BLS residual is positive and most of the other comparable industry groups expand.
• Mixed, The total residual and the industry count point in opposite directions.

This decision rule is prospective: it was set before the BLS release so the interpretation cannot be rewritten after the result is known. Even a corroborating first estimate would remain preliminary because BLS revises monthly payroll estimates as additional employer reports arrive and later benchmarks the series to QCEW records. [BLS methods](https://www.bls.gov/bls/empsitquickguide.htm)

No market move can substitute for that economic evidence. A price change after the ADP release would show how traders repriced the news, not whether employment outside education and health actually contracted.

The narrow answer is yes within ADP's August estimate: 38,000 total private jobs minus 45,000 in education and health equals a 7,000 decline across the other industry groups combined. Five of those nine groups declined and four grew. [ADP August report](https://mediacenter.adp.com/2026-09-02-ADP-National-Employment-Report-Private-Sector-Employment-Increased-by-38,000-Jobs-in-August)

The broader answer is not yet. July's BLS residual was near zero but highly uncertain, July turnover data did not show a national surge in layoffs, and the historical concentration signal failed to beat a simple baselinebaselineThe comparison result used to judge whether a signal performs better than an ordinary month.. [BLS July report](https://www.bls.gov/news.release/archives/empsit_08072026.htm) [BLS JOLTS](https://www.bls.gov/news.release/archives/jolts_09012026.htm) [Historical series](https://fred.stlouisfed.org/data/ADPMNUSNERSA)

The defensible finding is therefore limited but consequential: ADP recorded an August contraction outside education and health, revealing unusually narrow employment growth. The evidence does not yet prove that the wider private labor market entered a sustained contraction.