You may have read that October’s jobs report “smashed” expectations. Employers added 165,000 jobs, against a forecast of about 65,000, and unemployment fell to 4.1%. The two months before were revised up, too.
But that 165,000 is a first draft. The Bureau of Labor Statistics will revise it in each of the next two reports, and again in its annual benchmark revision, a once-a-year check against tax records that cover nearly every employer. After the last few years, many people now ask the same thing first: will this one shrink too?
We checked every first estimate since 1979 against what it became two months later. The typical revision was about 41,000 jobs, up or down.
October’s 165,000 is really a range
Eight in ten revisions fell between 69,000 down and 97,000 up. Add that spread to October and the final number probably lands between about 96,000 and 262,000.
Even the low end of October’s likely range is above the forecast.
96,000–262,000
43,000–287,000
The BLS’s own error margin is wider, because it measures something else: the survey reaches only a sample of employers, so any month’s number could be off by about 122,000 jobs either way. By both measures, the low end is still above the 65,000 forecast. First estimates rarely flip from a gain to a loss: 31 times in 571 months.
Four years of downward revisions
What has changed is the direction. For most of the record, first estimates were a bit too low and got revised up. Since 2022, the average year has gone the other way.
Revisions leaned down every year from 2022 to 2025, the longest run since 1979.
In 2025, the average month was cut by about 58,000 jobs. Only 2008 (73,000) and 2020 (60,000) were worse, and both were recessions. The annual benchmark was bigger still. Its preliminary estimate, published in September 2025, cut job growth in the year to March 2025 by about 911,000. (Shaded years had a recession. The tall bar in 2021 is the reopening year, when first counts kept falling short.)
That’s why I stopped reading the first number as a fact. I treat it as a guess that may be cut again, and I check the direction against the four-week average of weekly jobless claims, which was about 195,000 at the end of October. I trust claims more because they’re real filings, counted every week.
Does a downward lean mean a recession is coming?
Not on its own. The two biggest downward years, 2008 and 2020, were recessions. But 2001 was a recession and its revisions leaned up, and 2023 through 2025 leaned down without one. So read a downward lean as a reason to wait for the revision, not as a warning.
This year looks different again. So far in 2026, the average cut is small, about 8,000 a month, and the two latest revisions went up:
The two months before October ended higher than first reported.
| Month | First | Second | Third |
|---|---|---|---|
| August | 162,000 | 133,000 | 186,000 |
| September | 29,000 | 60,000 | Dec. 4 |
| October | 165,000 | Dec. 4 | Jan. 8 |
The next annual benchmark looks calmer too. Its preliminary estimate in August trimmed the year to March 2026 by only about 79,000 jobs.
What it means for rates and your money
A strong report raises the odds of another Fed rate hike in December, after the one in September. That would mean higher rates on credit cards and other loans that move with the Fed.
For paychecks, the news is less cheerful: average hourly pay was up 3.1% from a year earlier in October, still behind the 3.7% rise in prices through September. Markets will react to the headline either way. But the decision itself rests on several months of data, and those months will have been revised by then.
I’ve never made a big money move on a jobs headline. Last summer, on the day a weak report cut the two months before it, I put off a planned fund purchase by a week and left my monthly investing alone. It made almost no difference, and that’s the point: if no large decision rides on one number, a revision can’t hurt much.
The first revision to October comes with the November report on December 4, five days before the Fed’s December decision. Our recession warning lights will update with it. If October stays well above 65,000 after that revision, the pickup is real. If it slides toward the low end of the range, revisions are still running down.
Sources
Method: For every month since January 1979, we took the Bureau of Labor Statistics' first estimate of the change in nonfarm payrolls and its third estimate, published two months later, and measured the difference. We then looked at the size of those revisions, the share that went down, and the average for each year. Recession years come from NBER dates via FRED. Data: BLS; calculated by MKViz. Code and data
- Bureau of Labor Statistics, Nonfarm Payroll Employment: Revisions between over-the-month estimates, 1979-present
- Bureau of Labor Statistics, Employment Situation Technical Note
- Federal Reserve Bank of St. Louis (FRED), Unemployment Rate
- Federal Reserve Bank of St. Louis (FRED), NBER based Recession Indicators
- CNN, The US economy added 165,000 jobs in October and the unemployment rate fell