The Week Ahead: The Scorekeeper Erased Last Season

Friday morning, the Bureau of Labor Statistics mentioned, in the tone you'd use to admit you dented someone's car, that the American economy created 79,000 fewer jobs than it had been reporting for the twelve months through March. Private payrolls came off by 178,000. Economists surveyed by Bloomberg had expected the revision to go the other way, up 183,000, which means the forecast of the correction missed by more than the correction did.

The same morning, Kevin Warsh stood up at Jackson Hole and explained that the Fed's job right now is to focus on prices. Not jobs. Prices. Markets took about an hour to move September rate-hike odds from roughly 35% to a coin flip.

So here's the week the calendar built. The Fed talks Tuesday, Wednesday, and Thursday. Then it goes silent for ten days. And the number that settles the whole argument arrives Friday at 7:30 AM CT from the agency that just told you its last year's numbers were wrong.

Last week's damage report is in the tide chart edition.

Everything dated lives here. The prose that follows explains why it matters, not when it happens.

The pre-meeting blackout starts Saturday, September 5, and runs through September 17, which is the Fed's way of saying that once the arguing gets expensive, everybody shuts up. That makes this the last week anyone in the building can say a word before the Fed decides whether to raise interest rates on September 16.

They are using it. Governor Michael Barr speaks Tuesday at 8:05 AM CT. Governor Christopher Waller does a Reuters interview Thursday at 7:30 AM CT. Both are Board governors, which means both vote, which means both sentences count. Cleveland's Beth Hammack is on the calendar for Thursday at 2:00 PM CT. She votes, too, and has already said she thinks it will take more than one 25-basis-point move because "one 25 basis point move probably doesn't do a whole lot for the economy." That is not a woman looking for an excuse to hold. Sandwiched between them, Wednesday at 1:00 PM CT, the Beige Book arrives with twelve districts' worth of anecdotes about what businesses are actually charging, which is the only inflation data the Fed collects by asking humans instead of counting receipts.

What could go wrong is the boring version: everybody repeats what Warsh already said, nothing moves, and you have spent a week reading tea leaves that all say the same thing.

Once a year, the BLS checks its monthly payroll survey against actual tax records, which is roughly the difference between counting the crowd and counting the tickets. On Friday, it published the preliminary check for the year through March 2026, and the tickets came up 79,000 short. Total private payrolls were overstated by 178,000. Average monthly job growth across that stretch is now closer to 11,000, a number that would embarrass a medium-sized county.

Preliminary benchmark revisions have now cut employment estimates lower in seven of the last eight years, so this is less a scandal than a habit. Which is its own problem, because the Fed sets the price of your mortgage using a number that has been wrong in the same direction eight years running.

Then, on Friday, you get a fresh one. Consensus lands somewhere between +45,000 and +58,000 depending on whether you read the calendar estimates or the Reuters poll, against July's outright loss of 23,000 jobs. The unemployment rate is seen at 4.1% or 4.2%, and the fact that professional forecasters cannot narrow a labor market to a tenth of a point tells you what kind of visibility everyone is working with. Wednesday's ADP print and Tuesday's JOLTS openings are the appetizers, and both have a long history of pointing the wrong way.

What could go wrong: a hot wage number. Average hourly earnings are seen at 3.3% year over year, up from 3.2%, and Warsh has made clear which half of the mandate he is reading.

Broadcom reports fiscal Q3 Wednesday after the close, and it is the single most important corporate event of the week by a distance nobody is contesting. The company guided to roughly $16 billion in AI semiconductor revenue for the quarter, which is the sort of number that used to be an entire company's revenue. The real question is not the print. It is whether management raises its fiscal 2027 AI target above $100 billion, because declining to raise it last quarter is what sent the stock down despite a beat.

Underneath that sits the thing this letter has been circling for three weeks. Bloomberg reported on August 20 that Broadcom is in talks to borrow more than $60 billion to finance AI chip capacity. In talks. Not raised, not closed, not signed. But add Alphabet's completed $25 billion notes offering on August 10 and CoreWeave's loss widening on interest costs, and the AI buildout has quietly stopped being a spending story and become a borrowing story. Broadcom went into its own earnings week at 368.79, up 0.09% on the week, which is the market's way of saying it has absolutely no idea.

The software slate around it already has a mood. Salesforce reported on August 26, and the stock rose 22.58% the next session, which is not a normal reaction to enterprise software. Palo Alto Networks, MongoDB and GitLab report Tuesday, Snowflake Wednesday, and Zscaler, DocuSign and Lululemon close it out Thursday.

Here is the week's genuine oddity. Markets spent Friday repricing the odds of a rate hike from about a third to a coin flip, and volatility went down. The VIX closed at 14.43, its lowest close of the week, against 15.13 seven days earlier. The S&P 500 ETF finished the week up 0.47%, the Nasdaq proxy up 0.42%, the Dow proxy up 0.53%.

The tape did register it, just not where the headlines look. Small caps took the hit, with the Russell 2000 ETF down 1.40% on the week and 1.35% of that arriving Friday alone, which is what happens when the companies that borrow at floating rates hear the word "hike." Gold fell 3.24% Friday on the heaviest volume of its week, roughly 25 million shares, because a Fed that fights inflation is a Fed that does not need you to own a metal. The 10-year Treasury yield closed Thursday at 4.674%, and Friday's quote came back frozen, so we are not printing one.

What could go wrong is simple arithmetic. A VIX at 14.43 means options are cheap, and options are cheap right before the last jobs report of a live rate decision only if somebody is wrong. We walked through how the market sold its insurance in Friday's edition.

Ranked by how much they can actually move your money.

Friday, September 4, 7:30 AM CT. The August employment report.

It is the last major data release before the Fed decides on September 16; it arrives hours before the blackout, and it is the only number this week with the standing to move the odds on its own.

If payrolls come in soft and wages are cool, the hike case weakens, the front end rallies, and small caps get their money back. If payrolls are fine and average hourly earnings print 0.2% on the month or better, then Warsh has his cover, a September hike stops being a coin flip and starts being a plan, and you find out how much of this year's rally was borrowed against cheap money.

The honest posture, and we say this most weeks because it keeps being true: do not be a hero before the print. There is no edge in guessing a number that professional forecasters currently disagree about by 13,000 jobs and a tenth of a point of unemployment. Let it land. The trade is in the ten minutes after, not the ten hours before.

The Fed spends this week explaining itself, then locks the door. The BLS spent Friday quietly erasing a year of good news. And on Friday morning, it hands you a brand-new number and asks you to believe it.

Somebody will be very confident about all this by Thursday night. Do not be that person.

See you Tuesday.

Follow us on X and Instagram for daily market updates

Markets don't wait, and neither do we. Follow Tracking the Trade for fast takes, the numbers that matter, and the occasional meme that hits a little too close to your portfolio.