Briefs914 words, 4 minutes
The August 2026 jobs report as it appears in FRED, in five paragraphs
A dated reading of the 2026-09-04 Employment Situation through four FRED series retrieved the next day: payrolls, unemployment, participation, and earnings.
The Bureau of Labor Statistics published the Employment Situation for August 2026 on Friday 2026-09-04. We pulled four monthly series from FRED the following day (PAYEMS, UNRATE, CIVPART, CES0500000003, all since January 2024) and ran the attached script, which computes the monthly payroll change, its three-month average, and the twelve-month change in hourly earnings. This brief reads that table in five paragraphs, checks it against the release's own sentences, and says what a single month can and cannot support. It is a description of one vintage of the data on one day, and nothing in it is a forecast.
The table
| month | payrolls (thousands) | monthly change | 3-month average | unemployment rate | participation | avg hourly earnings y/y |
|---|---|---|---|---|---|---|
| 2025-09 | 158,548 | +76 | +23 | 4.4 | 62.5 | 3.85% |
| 2025-10 | 158,408 | -140 | -45 | n/a | n/a | 3.92% |
| 2025-11 | 158,449 | +41 | -8 | 4.5 | 62.5 | 3.93% |
| 2025-12 | 158,432 | -17 | -39 | 4.4 | 62.4 | 3.73% |
| 2026-01 | 158,592 | +160 | +61 | 4.3 | 62.1 | 3.66% |
| 2026-02 | 158,436 | -156 | -4 | 4.4 | 62.0 | 3.70% |
| 2026-03 | 158,650 | +214 | +73 | 4.3 | 61.9 | 3.43% |
| 2026-04 | 158,798 | +148 | +69 | 4.3 | 61.8 | 3.57% |
| 2026-05 | 158,861 | +63 | +142 | 4.3 | 61.8 | 3.34% |
| 2026-06 | 158,892 | +31 | +81 | 4.2 | 61.5 | 3.38% |
| 2026-07 | 158,913 | +21 | +38 | 4.1 | 61.4 | 3.24% |
| 2026-08 | 159,075 | +162 | +71 | 4.1 | 61.6 | 3.09% |
df["payroll_change"] = df["PAYEMS"].diff()
df["payroll_3m_avg"] = df["payroll_change"].rolling(3).mean()
df["ahe_yoy"] = 100 * (df["CES0500000003"] / df["CES0500000003"].shift(12) - 1)
One: the headline, and the check
Payrolls rose by 162,000 in August to 159,075,000, and the release says the same thing in its first sentence: "Total nonfarm payroll employment increased by 162,000 in August". The unemployment rate in the file is 4.1 percent, and the release says it "was unchanged at 4.1 percent". Two independent routes to the number (the FRED CSV and the BLS text) agreeing is the minimum check for a brief like this, and it passed. The three-month average of payroll gains is 71,000, because the June and July changes in this vintage are 31,000 and 21,000.
Two: the revisions are in the file already
Those June and July figures are not the ones first published. The release states that "the change in total nonfarm payroll employment for June was revised up by 11,000, from +20,000 to +31,000, and the change for July was revised up by 44,000, from -23,000 to +21,000." The FRED series carries only the current vintage, so the file retrieved on 2026-09-05 already shows +31 and +21, and the previously published -23,000 for July exists nowhere in it. The three-month average of 71,000 is therefore a number that could not have been computed a day earlier, and the August figure of 162,000 will itself be revised twice. Anyone comparing this table to one they saved in August is comparing two vintages, and the differences are revisions, not errors.
Three: participation
The participation rate in the file is 61.6 percent for August, up from 61.4 in July, and the release describes it the same way: it "edged up to 61.6 percent in August but is down by 0.5 percentage point since January." The table shows that decline directly: 62.1 in January, 61.4 by July. A falling participation rate and a falling unemployment rate can coexist, because people who leave the labour force are counted in neither, and the unemployment rate's move from 4.4 in February to 4.1 in July happened while participation fell 0.6 points. The script does not decompose that, and a single brief should not pretend to.
Four: earnings
Average hourly earnings for all private employees are up 3.09 percent over the year in the file. The release, working from the same series, says "over the year, average hourly earnings have increased by 3.1 percent", and the monthly rise of "10 cents, or 0.3 percent, to $37.75". The twelve-month rate has come down from 3.93 percent in November 2025 to 3.09 now, which is the clearest trend in the table and the one least affected by monthly noise, since it is a ratio of two levels twelve months apart rather than a difference of two adjacent months.
Five: the gap in October
The UNRATE and CIVPART cells for October 2025 are blank in the FRED file, and the script reports them: "months with no UNRATE value: ['2025-10']". The payroll series does have an October value. FRED carries the household-survey gap the same way the BLS does, and the script's rolling average is written on the payroll series, which is continuous, so nothing in the table is computed across the hole. A reader building a twelve-month change in the unemployment rate should notice that the October 2026 figure, when it arrives, will have no base to compare against.
What one month supports
The file says: 162,000 jobs in August, an unemployment rate of 4.1 percent for the second month, participation at 61.6 percent, earnings growth of 3.1 percent over the year, and a three-month payroll average of 71,000 after upward revisions to the two prior months. The payroll change over the last twelve months in this vintage is +603,000; the 2025 calendar-year total is +116,000; 2026 to date is +643,000.
It does not say whether August is a turn or noise. The monthly changes in this table run from -156,000 to +214,000 inside seven months, each will be revised twice, and the annual benchmark revision to the level is still ahead. What can be read is the direction of the slower series: earnings growth lower each quarter, participation lower since January, the unemployment rate lower since February. The next release, for September, is what will say whether August's payroll figure survives its first revision, and this brief will not be updated to guess.