The simple version

ADP, a payroll processing company, reported that private-sector employment increased by 38,000 jobs in August. Coverage of that number tends to frame it as an early read on the government employment report that follows.

It is an early read on something, and not on that. The two counts come from entirely different places: one from anonymized payroll records, the other from a government survey of a sample of establishments.

Neither is the draft version of the other, and when they disagree, the disagreement can be fully explained by who was counted.

The numbers

• ADP reported private-sector employment increased by 38,000 jobs in August, in its National Employment Report published September 2, 2026, and the Bureau of Labor Statistics has scheduled the August Employment Situation for Friday, September 4, 2026, at 8:30 a.m. Eastern (ADP; BLS)
• ADP states in the same release that the July total was revised from 44,000 to 46,000 (ADP)
• ADP describes August as the slowest pace of job creation since January, and reports that manufacturing, professional services, and information shed jobs while education and health care, construction, and leisure and hospitality showed hiring (ADP)
• ADP describes its report as an independent measure of the labor market based on the anonymized weekly payroll data of more than 26 million private-sector employees in the United States (ADP)
• ADP states the report is produced by ADP Research in collaboration with the Stanford Digital Economy Lab (ADP)
• BLS states that its employment release presents statistics from two monthly surveys: the household survey measures labor force status including unemployment, and the establishment survey measures nonfarm employment, hours, and earnings by industry (BLS)
• The most recent published Employment Situation covers July, when nonfarm payroll employment changed by -23,000 and the unemployment rate was 4.1 percent (BLS)
• ADP's report covers private-sector employment; the government's headline payroll figure covers both private and government employment (ADP; BLS)
• ADP's release makes no reference to the government series and states no relationship between the two counts (reviewed 2026-09-02)

Two different populations

The payroll company's count is built from payroll records rather than from asking anyone. In its own words, the report is an independent measure of the labor market based on the anonymized weekly payroll data of more than 26 million private-sector employees.

That is a genuine strength and a specific limitation, and it is worth stating both without editorializing. Payroll records are administrative data rather than survey responses, which removes one whole category of error. They also describe the employees whose payroll runs through that company, which is a large group and not the same group as every employee in the country.

The government count works the other way around. BLS describes its establishment survey as measuring nonfarm employment, hours, and earnings by industry, drawn as a sample intended to represent the whole economy. That addresses the population question and introduces sampling error, nonresponse, and the revisions that follow as more responses arrive, which we have written about separately.

So one is a complete count of a partial population and the other is a partial count of a fuller one. Neither of those is a defect. They are different designs answering the same question, and this article does not claim either is more accurate.

Why the preview framing persists

The payroll company's report arrives before the government's and covers the same month, which makes the preview framing almost irresistible in coverage.

The timing is real. The inference is not. A figure that arrives first is only a forecast of a figure that arrives second if the two are measuring the same thing, and these measure overlapping but distinct populations by fundamentally different methods.

There is also a scope difference that is easy to miss. The government's headline payroll number includes government employment. The private processor's figure does not. Comparing them directly means comparing a total against one component of that total, before any question of method arises.

Neither publisher helps here. ADP's release does not mention the government series at all, and states nothing about how the two relate. So any claim about a typical gap between them, or about which direction one usually misses in, is not coming from either source.

The Real Cost lens on which count includes you

The useful question is not which number is right. It is which population each one describes, because that determines what anyone can conclude from it.

• If your employer's payroll runs through that provider, changes at your workplace sit in its count as records rather than as an estimate
• If you work for a public employer, you are outside the private figure entirely and inside the government's headline number
• If you are self-employed or work outside traditional payroll, neither payroll count sees you, though the separate household survey may
• The household survey is where the unemployment rate comes from, which is why that rate and the payroll figure can move in ways that look contradictory
• No national figure of either kind describes an individual, an industry, or a region

That is the practical translation. Two numbers describing August will differ, both can be accurate, and the gap between them is mostly a fact about who was counted rather than about what happened in August.

What this means

When two figures describing the same month disagree, the first question is what each one counted. Both publishers document their own methods, and the answer is usually in the population rather than in an error by either.

The broader habit applies well beyond employment. Private data and government data frequently measure the same subject with different coverage, and the private version often arrives first precisely because it is not a survey. Faster and narrower is a real trade rather than a free lunch.

What this is NOT

This is not a prediction of Friday's employment report or of any economic figure, and no forecast or consensus estimate appears in this article. This is not a claim that either count is more accurate than the other; they measure different populations by different methods and each publisher documents its own approach. This is not a characterization of the labor market as strong or weak, and the descriptions of sector movements are ADP's own words about its own data rather than a judgment by this article. ADP is named as the publisher of its own report and is not evaluated as a company, and nothing here is a statement about it as an investment. No comparison figure, ratio, or typical divergence between the two series is computed or asserted, because neither publisher publishes one. National figures do not describe any individual, industry, or region. This is not investment or financial advice of any kind.

Sources

• ADP, National Employment Report, private-sector employment increased by 38,000 jobs in August, September 2, 2026: https://mediacenter.adp.com/2026-09-02-ADP-National-Employment-Report-Private-Sector-Employment-Increased-by-38,000-Jobs-in-August
• ADP, National Employment Report: https://adpemploymentreport.com/
• U.S. Bureau of Labor Statistics, Employment Situation: https://www.bls.gov/news.release/empsit.nr0.htm
• U.S. Bureau of Labor Statistics, Employment Situation release schedule: https://www.bls.gov/schedule/news_release/empsit.htm
• U.S. Bureau of Labor Statistics, Current Employment Statistics methodology: https://www.bls.gov/opub/hom/ces/