September Jobs: +29,000, and the Revisions Behind It

Analysis. Every figure below is sourced in the box at the end. The facts are laid out; the conclusion is yours.

In the United States, the Bureau of Labor Statistics reported on October 2, 2026, that nonfarm payrolls rose by 29,000 in September and that the unemployment rate was 4.2%. The same release lowered July and August. Over the weekend a Bloomberg tally of past revisions circulated widely online. The question for a reader is what a revision is, and how large the official ones are.

What the BLS published

The September gain of 29,000 followed an average monthly gain of 45,000 over the prior 12 months, according to the BLS. July was revised from +21,000 to −10,000, a cut of 31,000. August was revised from +162,000 to +133,000, a cut of 29,000. Together, the two months are 60,000 lower than first reported.

Other figures in the same release: 7.1 million unemployed, a labor force participation rate of 61.8%, and average hourly earnings up 3.0% over 12 months to $37.81. Health care added 17,000 jobs, below its 12 month average of 33,000, and financial activities is down 129,000 since a peak in May 2025.

Why numbers get revised

The BLS explains that monthly revisions come from additional reports from businesses and government agencies and from the recalculation of seasonal factors. A first estimate rests on the survey responses available at that time. Later responses fill in the picture.

There is a second, annual step. Each year the survey estimates are benchmarked against near complete counts built mostly from state unemployment insurance tax records. On August 28, 2026, the BLS published its preliminary estimate for March 2026: −79,000 for total nonfarm employment (−0.1%) and −178,000 for total private employment. It adds that the revisions of the last 10 years average 0.2% of employment in absolute value, and that official estimates are not changed until February 2027.

A tally, and its limits

A post by Unusual Whales, citing Bloomberg, says initial reports have been revised down in 16 of the last 20 months, removing 793,000 jobs in total. We could not open Bloomberg's underlying data, and other outlets give other totals depending on scope and dates, so treat it as a claim to check against the BLS tables, not as a government number.

The distinction matters for another reason. A revision changes an estimate of past employment. It does not mean that workers who were counted were later found to be out of work. Earlier on this site, the week in five numbers put several of the month's figures side by side.

The arguments, side by side

One reading says the first estimates are the problem: if they are routinely too high, markets and officials act on a picture that later shrinks. Another says that is how a sample survey works, that revisions go both ways over time, and that the BLS is open about them. A third asks whether the survey should lean more on tax records and less on responses from employers.

Each reading rests on the same table, and the table is public.

What to watch

The October employment report is scheduled for Friday, November 6, 2026, at 8:30 am ET. Will July and August move again? How large will the final benchmark be in February 2027, compared with the −79,000 preview? And if first estimates keep being revised in the same direction, which part of the process should change?