Wall Street has spent the week fixated on Kevin Warsh's first Jackson Hole address, scheduled for Friday at 10:00 a.m. Eastern. At that exact hour, the Bureau of Labor Statistics will publish a number that could matter just as much to the September rate debate — and almost nobody has it circled.
What Drops at 10 a.m. Friday
The BLS releases its preliminary estimate of the annual benchmark revision to establishment survey data on August 28, alongside first-quarter 2026 figures from the Quarterly Census of Employment and Wages. The benchmark exercise swaps the monthly payroll survey's sampled estimates for the QCEW's near-census of employer tax records, covering roughly 95% of U.S. jobs. This year's revision applies to the employment level as of March 2026, and the final adjustment will be folded into the January 2027 jobs report next February.
That sounds like accounting housekeeping. It has not behaved like it. The last three benchmark revisions all pointed the same direction: payrolls overstated by 266,000 jobs in the 2023 revision, 589,000 in 2024, and roughly 900,000 in last year's preliminary estimate covering April 2024 through March 2025 — the largest downward mark on record at the time.
Why This One May Break the Pattern
The consensus expectation of another gut-punch may be wrong. Guy Berger, chief economist at Homebase, argues the incoming QCEW data points the other way. "The most likely outcome based on what we currently know is a small upward revision," he wrote, noting the census data suggests the payroll survey has been undercounting rather than overcounting. He adds a caveat in the same direction: because QCEW figures themselves tend to get revised up, the eventual February number will probably land more positive than Friday's estimate.
Berger is careful about what that would and would not change. The unemployment rate still rose from 4.18% to 4.26% between March 2025 and March 2026. A friendlier benchmark rewrites the level of job growth, not the trajectory.
The Data Underneath
Thursday's weekly claims report reinforced the standing picture of a frozen but not collapsing labor market. Initial claims for the week ending August 22 fell 4,000 to 203,000, below the roughly 208,000 economists expected, though the four-week moving average rose 1,250 to 205,500. Continuing claims eased to 1.778 million from a revised 1.796 million.
The monthly data is where the softness lives. July payrolls fell 23,000 against expectations of an 83,000 gain — the first outright decline of this cycle — while May and June were revised down a combined 103,000. The three-month average job gain now sits near 20,000. Unemployment ticked down to 4.1%, but through a labor force participation rate that slid to 61.4%, a five-year low.
The Stakes for September
Warsh inherits a federal funds rate at 3.50%–3.75% with inflation near 3.4%, and futures markets price roughly one-in-three odds of a September increase. A benchmark revision showing the labor market was stronger than reported hands the hawks ammunition. Another nine-figure downgrade hands it to the doves. The August employment report follows on September 4, with early estimates clustered around 12,000 jobs.
Sources: U.S. Bureau of Labor Statistics, U.S. Department of Labor, Macro Mostly (Guy Berger), Kansas City Fed, Trading Economics