Friday recap: NFP 29k, yields rose anyway. Our plan failed on all four.
Written Sunday October 4, covering Friday October 2. September payrolls printed plus 29,000 against a consensus of roughly 85,000 to 90,000, with July and August revised down by a net 60,000; unemployment rose to 4.2 percent, largely on labour force entrants, and hourly earnings rose 0.1 percent (BLS). The first 30 minutes matched our pre-registered weak-print map. By the close the 10-year yield was up about 4 basis points, DXY flat, gold lower and USD/JPY higher: a fail on all four calls, graded in the open. Thursday folded in: ISM Manufacturing 54.5 with Prices at 77.9, initial claims 197,000. One observation consistent with the weak-versus-strong asymmetry hypothesis, not a confirmation.
Written Sunday October 4, covering Friday October 2. This is a recap, published after the fact, not a same-day note. September payrolls came in at plus 29,000 against a consensus of roughly 85,000 to 90,000, with net revisions of minus 60,000 to the prior two months. That is a weak print by any reading, and it landed in the "Weak" bucket of the reaction map we pre-registered before the release. For about thirty minutes the tape moved the way the map said it would. By the 4:00 PM ET close it had reversed: the 10-year yield finished higher, DXY finished flat, gold finished lower, and USD/JPY finished higher. Our plan failed on all four calls. This piece grades it in the open.
The last analysis note in this thread was the September 30 Core PCE piece, where a cool inflation print also failed to rally bonds. There is no Thursday piece; Thursday's releases (ISM Manufacturing and weekly jobless claims) are folded in below rather than backdated.
Thursday, folded in
- ISM Manufacturing PMI, September: 54.5, down from 54.6 in August, the ninth consecutive month of expansion (source: ISM Report on Business via PR Newswire, Oct 1). Consensus was 54.9 (source: CNBC). New Orders 55.3 (from 53.7), Production 56.7 (from 58.3), Employment 52.7 (from 51.2). The line that stands out is Prices: 77.9, up from 71.1.
- Initial jobless claims, week ending September 26: 197,000, down 1,000 from the prior week's revised 198,000; four-week average 200,000. Insured unemployment for the week ending September 19 was 1,701,000 (source: US Department of Labor weekly claims release, Oct 1).
- Thursday's rates and dollar: the 10-year yield closed 5.237 percent, down from 5.293 on Wednesday (source: Yahoo Finance ^TNX daily close). The 2-year Treasury yield fell to 4.78 percent from 4.88 (source: FRED DGS2). DXY closed 102.10 (source: Yahoo Finance DX-Y.NYB daily close).
So Friday morning opened with two things already in the price: a 10-basis-point drop in the 2-year on Thursday, and a manufacturing survey whose activity lines were firm and whose prices line was hot. Low claims said layoffs were not the problem going into payrolls.
The print
From the BLS September Employment Situation, released Friday at 8:30 AM ET (source: BLS Employment Situation):
- Nonfarm payrolls: plus 29,000, against a consensus of roughly 85,000 to 90,000 (Bloomberg median 85,000; MUFG preview 90,000).
- Unemployment rate: 4.2 percent, up from 4.1 percent. BLS attributes the rise largely to labour force entrants rather than job losers.
- Average hourly earnings: plus 0.1 percent month on month to $37.81, plus 3.0 percent year on year.
- Revisions: July revised to minus 10,000 (from plus 21,000), August to plus 133,000 (from plus 162,000). Net minus 60,000 across the two months.
The revisions take away more jobs than the headline adds. Put the three months together and the labour market looks weaker than any single month did when it was first published. Wage growth was soft too. The one part that cuts the other way is the unemployment rate: a rise driven by people entering the labour force is less of a recession signal than one driven by layoffs, and the rate did not reach the 4.3 percent threshold in our map.
After the report, odds of a pause at the October 27-28 FOMC rose to about 85 percent (source: CNBC post-report coverage; meeting dates per the Federal Reserve calendar). Equities rallied: the S&P 500 closed 7,722.72, up from 7,666.45 on Thursday (source: Yahoo Finance ^GSPC).
The tape: first thirty minutes versus the close
| Market | Pre-release (05:00 ET) | 08:30-09:00 ET bar | 4:00 PM ET close |
|---|---|---|---|
| 10Y yield | 5.237% | n/a | 5.277% (+4 bp) |
| DXY | 101.93 | 101.8 | 101.93 (flat) |
| Gold | 4,213 | 4,248 | ~4,162 to 4,177 (down) |
| USD/JPY | 157.56 | 157.2 | ~157.8 (up) |
Sources: pre-release levels from our editorial/plans/2026-10-02.md plan file, committed at 05:14 ET; intraday from Yahoo Finance 30-minute bars (12:30 UTC); closes from Yahoo Finance daily data (^TNX, DX-Y.NYB, GC=F, USDJPY=X). The gold range reflects the difference between the front-month future (4,162.30 settle) and spot. The 10-year figure is per market close data; Treasury's official daily yield (FRED DGS10) for October 2 had not been published at the time of writing and is the number to confirm against. Brent futures closed about $102.25 and EUR/USD about 1.126 (Yahoo BZ=F, EURUSD=X).
The first half hour was the textbook weak-print reaction: gold up more than $30, DXY softer, the yen stronger. All of it was gone by the close.
Grading Friday's plan
Before the release we wrote down a reaction map and committed it to the repository at 05:14 ET. The rule was simple: if payrolls came in below 40,000 or unemployment hit 4.3 percent or higher, expect the 10-year yield down at least 6 basis points, DXY down, gold up, and USD/JPY down, graded at the 4:00 PM ET close against the pre-release levels. Payrolls at 29,000 triggered the Weak bucket on the payrolls leg alone. Here is the grade.
| Call | Expected | Actual at close | Grade |
|---|---|---|---|
| 10Y yield | down 6 bp or more | up ~4 bp (5.237 to 5.277) | Fail |
| DXY | down | flat (101.93 to 101.93) | Fail |
| Gold | up | down (4,213 to ~4,162-4,177) | Fail |
| USD/JPY | down | up (157.56 to ~157.8) | Fail |
The map failed on all four at the close. The direction was right for about thirty minutes and then reversed. We graded at the close because that is the rule we wrote before the number came out, and changing the grading window after seeing the result would make the exercise worthless. A map that is only right for half an hour did not describe how the market priced the print, and we are recording it as a failure.
The thesis test. The plan also logged a hypothesis carried over from the September 30 reference piece: under the current reaction function, a weak print moves the 10-year down by less than a strong print would move it up. Friday is one observation consistent with that: a clearly weak print did not rally the long end at all, and the yield finished 4 basis points higher. But one print is one observation. It is not confirmation, and we did not get a strong print to compare it with. The plan said this would be graded over several prints, and it will be.
Two corrections to our own earlier work. First, the September 30 piece used a 155,000 consensus for September payrolls. The surveys we checked Friday morning (Bloomberg median 85,000, MUFG 90,000) did not support that figure, and the plan used the survey numbers. Second, that piece also gave a percentage split across Fed outcomes that was not tied to a named market source. We are not carrying it forward. From here, any probability cited in this thread comes from a named source such as CME FedWatch.
What we do not know yet
Why the reaction reversed is not established, and we are not going to pretend otherwise. There are three explanations that fit the facts we have, and each can be tested:
- The front end and the long end split. Higher pause odds are a dovish front-end repricing. If Treasury's official curve for October 2 shows the 2-year down while the 10-year rose, that is a bear steepener, and it points to term premium (compensation for holding duration) doing the work rather than policy expectations. If the 2-year rose too, the market did not believe the pause either, and this explanation is wrong.
- The composition took the edge off. The unemployment rise came largely from new entrants, wages rose 0.1 percent, and Thursday's ISM Prices index sat at 77.9. A market focused on inflation risk may have treated the payrolls miss as less informative than the headline suggested. If the October CPI release shows goods prices following ISM Prices higher, this explanation strengthens. If it does not, it weakens.
- Positioning and risk appetite. Thursday's 10-basis-point drop in the 2-year means some of the weak-data trade was already on before the release. Equities rallying 0.7 percent fits a "good news for the Fed path, sell duration, buy risk" reaction. This one is the hardest to falsify with public data, and we hold it most loosely.
The reference piece published alongside this one sets out these mechanisms in general terms.
What would change the read
No unconditional call. Two conditional statements, each with what would prove it wrong:
- If the official October 2 Treasury curve confirms the 2-year fell while the 10-year rose, then the asymmetry hypothesis gets a cleaner second data point: policy expectations moved dovish and the long end still did not rally. Wrong if the official 2-year also rose on Friday.
- If the next payrolls release (October data, due November 6 per the BLS release schedule) is strong and the 10-year rises by clearly more than Friday's 4 basis points, then that is the matching observation on the other side of the asymmetry. Wrong if a strong print moves the 10-year by about as much as, or less than, Friday's weak print did.
Nothing on this site is investment advice or a recommendation to trade. Most retail traders lose money. Reaction maps published here are falsifiable hypotheses, graded in public, not signals.