The monthly jobs report from the Bureau of Labor Statistics (BLS) combines two separate surveys: a payroll survey of about 119,000 businesses and government agencies, which gives the headline count of jobs, and a household survey of about 60,000 households, which gives the unemployment rate [3]. They can tell different stories. In September 2026, payroll employment rose 29,000 while household-survey employment rose 406,000, and the unemployment rate was 4.2% [2]. Payroll numbers are revised twice in the following months and once a year against tax records; the preliminary revision for March 2026 was -79,000 [3][4].
What are the two surveys in the jobs report?
The establishment survey, formally the Current Employment Statistics (CES) program, asks employers how many people were on their payrolls in the pay period that includes the 12th of the month. Its sample covers about 622,000 worksites and about 26% of all nonfarm payroll jobs [3]. It produces nonfarm payrolls, average hourly earnings and hours worked.
The household survey, the Current Population Survey (CPS), is run by the Census Bureau for BLS. Interviewers ask about 60,000 households whether each member worked or looked for work in the week that includes the 12th [3]. It produces the unemployment rate, the labor force participation rate and the employment-population ratio.
| Feature | Establishment survey (payrolls) | Household survey (unemployment rate) |
|---|---|---|
| Sample | About 119,000 businesses and agencies, 622,000 worksites | About 60,000 households |
| What it counts | Jobs on payrolls | People, employed or unemployed |
| Self-employed, farm and household workers | Not included | Included |
| People with two jobs | Counted once per job | Counted once |
| Age limit | None | 16 and older |
| 90% confidence range, monthly change | About plus or minus 122,000 jobs | About plus or minus 0.3 points on the rate |
| September 2026 | +29,000 jobs | +406,000 employed; rate 4.2% |
Source: BLS Employment Situation Technical Note and September 2026 release [2][3].
Why do payrolls and the household survey disagree?
They measure different things. A person with two jobs counts twice in payrolls and once in the household survey; a freelancer or farm worker counts only in the household survey; and someone on unpaid leave counts as employed in the household survey but not in payrolls [3]. In September 2026, 8,986,000 people held more than one job, 5.5% of all employed [2].
Sampling error is the other reason. With 60,000 households, the household survey is much noisier month to month. BLS estimates that a reported payroll gain of 50,000 has a 90% confidence range of -72,000 to +172,000, and that a monthly change in the number of unemployed has a range of about plus or minus 425,000 [3]. A gap like September's, 29,000 against 406,000, is within normal noise. Over several months, the two usually tell a similar story, so the trend matters more than one report.
How do revisions and the annual benchmark work?
The first payroll estimate is based on the responses received by the release date. BLS revises each month twice, in the next two reports, as more employers respond, and recalculates seasonal factors each month [3]. In the September 2026 report, July was revised from +21,000 to -10,000 and August from +162,000 to +133,000, a combined -60,000 [2].
Once a year, BLS "benchmarks" payrolls to the Quarterly Census of Employment and Wages (QCEW), a near-complete count of jobs from state unemployment insurance tax records [4]. The benchmark is set for March, and the average absolute revision over the last 10 years has been 0.2% of employment [4].
- March 2025 benchmark (final, February 11, 2026): the seasonally adjusted level was revised down by 898,000, and job growth for all of 2025 was revised from +584,000 to +181,000 [5].
- March 2026 benchmark (preliminary, August 28, 2026): a revision of -79,000, or -0.1%, with private employment revised by -178,000 [4]. The final figure is due with the January 2027 report in February 2027 [4].
What is the birth-death model?
The payroll sample cannot see new businesses right away, and closed businesses simply stop responding. BLS handles this with a two-part adjustment: it imputes employment trends for sampled firms that have gone out of business, and it adds a time-series estimate of net job creation from business openings and closings, based on five years of unemployment insurance data [3]. When business formation changes fast, the model can miss, which is one reason benchmark revisions can be large. Starting with the January 2026 report, BLS changed the model to incorporate current sample information each month [5].
What did the 2025 government shutdown do to the data?
Household survey data for October 2025 were not collected because of the federal government shutdown, so there is no unemployment rate for that month [6]. Payroll data for October were collected late and first published with the November report on December 16, 2025, with collection periods for both months extended [6].
The shutdown also shows how revisions work. October 2025 payrolls were first reported at -105,000, including a 162,000 drop in federal employment [6]; after later revisions, FRED shows a decline of 140,000 [7]. The November 2025 unemployment rate was first published at 4.6% and now stands at 4.5% after BLS's annual revision [6][8].
DataPorium's monthly payroll and unemployment series have a gap from October 2025 to May 2026, and its 2025 payroll levels predate the February 2026 benchmark, so comparisons across that gap should use the revised official series [1][7].
What does the 2026 jobs data show so far?
| Month (2026) | Payroll change | Unemployment rate |
|---|---|---|
| January | +160,000 | 4.3% |
| February | -156,000 | 4.4% |
| March | +214,000 | 4.3% |
| April | +148,000 | 4.3% |
| May | +63,000 | 4.3% |
| June | +31,000 | 4.2% |
| July | -10,000 | 4.1% |
| August | +133,000 | 4.1% |
| September | +29,000 | 4.2% |
Source: FRED, total nonfarm payrolls and unemployment rate, data as of October 2, 2026 [7][8]; June to September also in DataPorium Economic Indicators [1].
Total nonfarm payrolls were 159.0 million in September 2026, 496,000 more than a year earlier, an average of about 41,000 a month [7]. The unemployment rate has stayed between 4.1% and 4.4% all year [8]. Average hourly earnings rose 3.0% over the year to $37.81, and the labor force participation rate was 61.8% [2].
Payrolls count jobs from about 119,000 employers; the unemployment rate comes from about 60,000 households, so one month's gap between them is usually noise.
When does the jobs report come out?
The report is usually released at 8:30 a.m. Eastern Time on the first Friday of the month. The BLS schedule lists the October 2026 report for Friday, November 6, 2026, and the November 2026 report for Friday, December 4, 2026 [9]. The latest figures and history appear on DataPorium's economic indicators page [1].
Key takeaways
- The payroll survey samples about 119,000 businesses and agencies; the household survey about 60,000 households [3].
- In September 2026, payrolls rose 29,000 while household employment rose 406,000; unemployment was 4.2% [2].
- A monthly payroll change has a 90% confidence range of about plus or minus 122,000 jobs [3].
- The March 2025 benchmark cut payrolls by 898,000; the preliminary March 2026 benchmark is -79,000 [4][5].
- There is no October 2025 unemployment rate: the household survey was not collected during the shutdown [6].
Frequently asked questions
What is the difference between the payroll and household surveys?
The payroll survey counts jobs reported by about 119,000 employers; the household survey counts people in about 60,000 households and produces the unemployment rate [3]. In September 2026 they showed gains of 29,000 and 406,000 [2].
Why are jobs numbers revised?
More employers respond after the first release, so each month is revised twice, and once a year payrolls are matched to unemployment insurance tax records. July and August 2026 were revised down by a combined 60,000 [2][4].
What is the BLS benchmark revision for 2026?
BLS announced a preliminary benchmark revision of -79,000, or -0.1%, for March 2026 on August 28, 2026; the final revision comes with the January 2027 report [4].
Why is there no October 2025 unemployment rate?
Household survey data for October 2025 were not collected because of the federal government shutdown, so BLS has no October 2025 unemployment rate [6].
Sources & References
- [1] DataPorium Economic Indicators (nonfarm payrolls and unemployment rate)
- [2] U.S. Bureau of Labor Statistics, The Employment Situation - September 2026 (October 2, 2026)
- [3] U.S. Bureau of Labor Statistics, Employment Situation Technical Note (October 2, 2026)
- [4] U.S. Bureau of Labor Statistics, CES Preliminary Benchmark Announcement (August 28, 2026)
- [5] U.S. Bureau of Labor Statistics, The Employment Situation - January 2026 (February 11, 2026)
- [6] U.S. Bureau of Labor Statistics, The Employment Situation - November 2025 (December 16, 2025)
- [7] Federal Reserve Bank of St. Louis, FRED, All Employees, Total Nonfarm (updated October 2, 2026)
- [8] Federal Reserve Bank of St. Louis, FRED, Unemployment Rate (updated October 2, 2026)
- [9] U.S. Bureau of Labor Statistics, Schedule of Employment Situation releases (accessed October 13, 2026)