Which Payroll Numbers Did the Investment Committee Actually See?

Compare the October and November 2026 US payroll releases to separate new jobs from revisions and reconstruct the figures available at the meeting.

In this article

A committee meeting after the 8:30 a.m. Eastern release on October 2, 2026 could have recorded 81,000 jobs a month as the average US payroll gain for August and September. On November 6, the Bureau of Labor Statistics put the same two-month average at 123,000. The economy had already lived through both months by the first meeting. The additional 42,000 jobs per month came from revisions to the estimates. Keeping both releases lets a portfolio manager distinguish what the analyst knew from what the department learned later. October 2 employment report · November 6 employment report, archived PDF

Rebuild the numbers on the old slide

Both releases measure seasonally adjusted monthly changes in total nonfarm payroll employment. The date in a column below is the date the number became available in that release. The date in a row is the month employers were describing. Keeping those two dates separate prevents a plausible-looking historical chart from putting November information in October’s meeting.

Payroll monthAvailable October 2Available November 6What changed
August 2026+133,000+186,000Revised up by 53,000
September 2026+29,000+60,000Revised up by 31,000
October 2026Awaiting release+165,000First estimate for a new month
August and September together+162,000+246,000An additional 84,000 for the same two months

The two-month growth estimate rose by more than half. September’s gain more than doubled. Nevertheless, hiring still slowed from August to September in both releases. An October memo that described deceleration could survive review. A memo that estimated recent hiring at roughly 80,000 a month needs an updated figure. Those are two different edits to an investment thesis. BLS November revision detail

Distinguish a new month from a revised trend

Suppose the November committee packet needs an August–October average. With the values available November 6, the three months sum to 411,000 jobs, or 137,000 per month. Appending October’s 165,000 to the two figures copied from the previous packet yields 109,000 per month. That apparent shortfall of 28,000 per month is exactly the 84,000 of revisions spread over three months.

Both calculations can be reconstructed. Only the first describes everything BLS had disclosed by the November meeting. The second is useful as a bridge showing the incremental effect of updating the old cells. A chart labelled “three-month average” with no release date gives a reader little chance of spotting the difference. For a chart of first estimates, go back to each month’s initial release. Even the August value in the October packet had already passed through a revision.

BLS attributes regular monthly revisions to additional establishment reports and recalculated seasonal factors. An upward revision is evidence about the estimate of past employment. It is insufficient to say which industries or companies drove the change without checking the corresponding industry data. Also keep the unemployment rate on its own line. BLS obtains it from the household survey, which measures people on a different basis from payroll jobs. BLS technical notes and tables

Preserve the decision as well as the series

For this committee, the most useful archive has two entries. The October entry retains the dated release, the August–September average and the analyst’s contemporaneous conclusion. The November entry links back to it, records the 84,000 revision and adds the October first estimate. A reviewer can now ask what changed in the evidence and whether the position rationale depended on the direction or the size of the slowdown.

If the question is which payroll series a researcher could download on a past date, start with the St. Louis Fed’s ALFRED archive. Its PAYEMS series stores employment levels in thousands. Subtract consecutive monthly levels within the same vintage to calculate a monthly change. ALFRED’s real-time periods use calendar dates, so for a meeting on the release day also check BLS’s publication time. A date alone cannot establish what the committee knew at 8 a.m. ALFRED payroll vintages · St. Louis Fed real-time period documentation

For a recurring review we recommend an AllMind Data Room for the dated release files and the committee’s saved memos. A reviewer can scope questions to those documents and follow the citations. Limit the October reconstruction to material available by its cutoff. A continuously updated data feed alone cannot establish what was on the analyst’s screen that morning. The arithmetic here comes from the two archived public releases. For carrying these decisions into slides, see our macro presentation workflow.