SASIGNAL ATLASCross-industry intelligence / Research desk
SIGNAL ATLAS / RESEARCH DESK

A first jobs print is revised twice before it is final

BLS revision tables show the second and third payroll estimates typically move the total by tens of thousands of jobs.

The signal

This method entry sits in Signal Atlas's cross-sector method desk, and it starts with the number reporters quote within minutes of a jobs release: the change in nonfarm payroll employment for the month just ended. The Bureau of Labor Statistics' own page on revisions to that estimate is explicit that the headline number is not final. Each month's payroll count is published once as a preliminary estimate, then revised twice more before being fixed at an annual benchmark, and the same page publishes the mean and mean-absolute size of those revisions, split between 1979 to 2003 and 2003 to the present.

The evidence

For the more recent period, the page reports a seasonally adjusted mean revision from the first estimate to the second of 7,000 jobs and from the second to the third of 0, with the full first-to-final mean absolute revision at 51,000 jobs; for 1979 to 2003 the comparable first-to-final mean absolute revision was 61,000. A second BLS page, on what seasonal adjustment is, explains why the published monthly figure is adjusted at all: the technique "attempts to measure and remove the influences of predictable seasonal patterns," such as weather, harvests, holidays and school schedules, so a reader can see the "cyclical, underlying trend" rather than a repeating seasonal wobble. The same first-print caution applies outside employment data: the Census Bureau's quarterly retail e-commerce release publishes both a seasonally adjusted and an unadjusted estimate for the same quarter, each carrying its own stated margin of error, an explicit acknowledgement that a single quarter's move can sit inside its own margin.

Timeframe and confidence

Editorially, a first print of a monthly employment change should be read as a provisional estimate with a stated historical error band measured in tens of thousands of jobs, not as a fixed fact. The revision tables cover decades of prior months, a solid basis for the general expectation of movement, but they do not say how far this specific month's print will move, only how far the average one has.

What would change the reading

The reading sharpens once the second and third estimates for the same reference month are published, roughly one and two months later, because a claim resting on a single first print is resting on the estimate the agency itself expects to move the most.

A single monthly print is one observation in a series the publishing agency itself expects to revise. Reading it as settled fact skips the rounds of revision its own documentation describes.

Source trail

  1. Nonfarm Payroll Employment: Revisions between over-the-month estimates, 1979-presentwww.bls.gov · Source publication: not established · Retrieved 2026-09-16

    States the mean and mean-absolute size of the second and third revisions to the payroll employment estimate, by period.

  2. What is seasonal adjustment?www.bls.gov · Source publication: not established · Retrieved 2026-09-16

    Explains why the published monthly employment figure is seasonally adjusted and what the technique removes.

  3. Quarterly Retail E-Commerce Sales Reportwww.census.gov · Source publication: not established · Retrieved 2026-09-16

    Shows a second official series publishing both a seasonally adjusted and an unadjusted estimate with a stated margin of error each quarter.

Event date
No single event
First source date
Unknown
Source-record publication
Not supplied — draft retained
Preparation
2026-09-16

Read across the evidence

Markets & measurement · All 100 historical records