The US Bureau of Labor Statistics reported on 2 October 2026 that employers added 29,000 jobs in September and that the unemployment rate rose to 4.2% from 4.1%. Economists had expected roughly 85,000 to 90,000: CoinDesk and Investing's market report cite a consensus of 90,000, while TheStreet cites 84,000. The headline miss is about 60,000 jobs. The revisions to the two months before it were almost exactly the same size, and that is the less-noticed half of the story.

What changed since our last two pieces

On 5 September we argued that the revisions published alongside the August report were small, and that a first print of 162,000 jobs for August was therefore a reasonably reliable number. That August figure has now been revised to 133,000, a cut of 29,000. July, first reported at +21,000, is now -10,000, a cut of 31,000. The two revisions together remove 60,000 jobs (the College Investor, citing BLS figures; CoinDesk reports the same July and August numbers). The point is not that the 5 September piece described the earlier data wrongly. It is that a revision that is small in one month can be large in the next, and one month of small revisions tells you little about the following month.

The second change is in rate expectations. On 25 September we reported that markets priced a 77.5% chance of an October Fed hike. After the report, Polymarket's October hike contract fell to about 16% on 2 October from about 69% on 28 September (Benzinga). CME FedWatch put the October probability at 21.59% (FXStreet), and Investing's market report put it at 12%. The sources do not agree on the pre-report level either: CoinDesk says traders priced 23% before the release, which is far below the roughly 70% that FXStreet and Benzinga describe a few days earlier. We cannot reconcile those numbers from the sources, so we give the post-report range of 12% to 22% and do not assert how much of the fall came from this release rather than earlier data.

The number that matters is the three-month average

A single payroll print is a noisy estimate that gets revised twice. Stacking the three latest months, on the figures as they now stand, gives -10,000 for July, +133,000 for August and +29,000 for September. The average is 152,000 divided by three, or about 50,700 jobs a month. That arithmetic is ours, not BLS's.

The same three months, using the numbers as first published (+21,000, +162,000 and the new +29,000), would have averaged about 70,700. So the revisions alone lowered the three-month average by 20,000 a month (60,000 divided by three), also our calculation. Put another way, the pre-report picture was about 121,000 jobs better than the current one: a 61,000 miss against a 90,000 consensus, plus 60,000 of downward revisions. That comparison mixes a forecast with revised history, so read it as a sense of scale rather than a precise measure.

Is about 51,000 a month weak? That depends on how many jobs the economy needs to keep unemployment steady, and the estimates themselves moved. Reuters surveys of economists, cited by FXStreet, put the break-even job growth at 0 to 50,000 on 4 September and 50,000 to 80,000 on 2 October. A 51,000 average sits at the bottom of the newer range. That is consistent with an unemployment rate that edged up by a tenth of a point, and with a jobless rate that has stayed between 4.1% and 4.3% since March (the College Investor, citing BLS data). It is not consistent with a labour market that is collapsing.

Why bad news lifted risk assets

The market's reaction follows from the Fed's situation. On 16 September the Fed, chaired by Kevin Warsh, raised its policy rate by 25 basis points to 3.75% to 4.00% in a 12-0 vote, its first increase since 2023 (Benzinga; FXStreet). Warsh said the hike reflected doubt that underlying inflation was moving towards 2%. In that setting a weak jobs report matters less for what it says about growth than for what it does to the next hike. A smaller chance of a second increase means lower expected policy rates, which lowers the discount rate applied to equities and makes non-yielding assets such as gold and bitcoin cheaper to hold.

The moves on 2 October fit that reading. The S&P 500 rose 0.89% and the Nasdaq 1.35% (TheStreet). Gold traded around $4,215. Bitcoin rose more than 2% to roughly $86,600 to $87,000 (CoinDesk). The dollar fell against major currencies. The 2-year Treasury yield fell about 7 basis points to 4.71%, and the 10-year fell 6 to 7 basis points to about 5.17%.

It would be easy to assume the 10-year barely moves on news like this. The figures do not support that. The 2-year and 10-year fell by similar amounts, so the curve moved in parallel and the gap between them stayed near 46 basis points (5.17% less 4.71%, our arithmetic). What stayed put was the level: a 10-year yield above 5% is still far above where borrowers such as mortgage lenders would like it. A single weak print can remove an expected hike. It cannot, on this evidence, undo a long-term yield that reflects inflation expectations and term premium. The 30-year yield of about 5.57% reported by Investing's market report tells the same story.

The wage line argues for the lower-hike view

Average hourly earnings rose 0.1% in the month against a forecast of 0.3%, and 3.0% over twelve months against a forecast of 3.2% (CoinDesk). For a central bank worried about inflation, soft wages reduce one channel through which a weak labour market and rising prices could feed each other. New York Fed President John Williams said on 29 September that there was no need for urgency in raising rates again, and Vice Chair Philip Jefferson said policymakers may need more time before the next move (Benzinga). Core PCE inflation also came in below expectations, according to Truflation's Oliver Rust, quoted by Benzinga. Taken together, the payroll miss arrived in a week when the other evidence was already pointing the same way.

What the print does not prove

  • It does not show the economy is entering a recession. One month's first print, even with revisions, is a statistical estimate. The unemployment rate moved by one tenth of a point, which is within the range it has held since March.
  • It does not settle whether the revisions will continue. Revisions are not always downward, and the September figure itself will be revised twice. The 60,000 figure is a measured fact about July and August, not a forecast about September.
  • It does not fix a consensus. Surveys differed by about 6,000 jobs (84,000 against 90,000), and sources disagree about hike odds both before and after the report.
  • It does not show that the Fed is finished raising rates. Polymarket still priced a December hike at about 68% (Benzinga), and FXStreet notes that December hike pricing remained in place after the report.

The counter-argument: jobs set timing, inflation sets the destination

The strongest objection to reading this report as a turning point comes from the rate-setting side. FXStreet's author, Joshua Gibson, argues that payroll reports have shaped when the Fed hikes but not how high rates go; inflation, he writes, sets the destination, and the unemployment rate is the labour-market figure that could change it. Dallas Fed President Logan said that 4.1% is close to the lowest unemployment rate the economy can sustain, according to the same article. On that view a rise to 4.2% is notable mainly because it moves unemployment further from a level the Fed regards as tight, and a hawk could still argue that a 4.2% rate does not by itself justify cutting the path of rates.

That leaves a reading that is narrower than either the relief rally or the recession talk suggests. The September report removed most of the case for an October hike, showed that the underlying trend in hiring is about 51,000 a month on current data, and left the 10-year yield above 5%. Whether December is a live meeting depends on inflation prints that had not yet arrived. The most useful habit for the next release is to read the revisions first, compute the three-month average, and treat the headline as the least reliable number on the page.