Today's payrolls print and the inversion nobody's pricing correctly

Consensus sits at 53,000 to 58,000 depending on whose survey you're reading. Unemployment rate expected flat at 4.1%. Average hourly earnings up 0.2%. None of that is the interesting part of this memo. The interesting part is that for the first time in this cycle, a good number is the bad outcome, and half the desk is still trading it backwards.

Run the logic once, slowly, because it matters for every position sized against this print.

Through 2024 and 2025, weak payrolls meant rate cuts got pulled forward and risk assets rallied on the relief. That reaction function is dead. Waller told a Reuters NEXT audience Thursday that he'd support a hold in September if inflation data confirms cooling — not if the labor market weakens further. Labor weakness stopped being the swing variable months ago. Fed officials have said as much out loud: Governor Barr called the jobs picture "stable" this week, Waller called it "satisfactory," and neither adjective is a compliment — they're a signal that the labor mandate has been quietly benched while the inflation mandate runs the meeting. A strong August print today doesn't get read as "the consumer is fine, buy everything." It gets read as "there's no labor-market excuse left to hold rates, so hike." Weak print, and the excuse to hold reappears. Up is down.

Now overlay what July actually did, because the base effects here are doing more work than most desks have modeled. Payrolls fell 23,000 against a Dow Jones consensus of positive 83,000 — one of the ugliest misses in years. Then the revisions came in worse than the headline: May cut by 66,000 to +63K, June cut by 37,000 to +20K. That's 103,000 jobs erased across two months, after the fact, with no market reaction on the day because everyone was staring at the -23K. The unemployment rate actually fell to 4.1% in that same report, and it fell for the worst possible reason — the labor force shrank by 264,000, household employment dropped 87,000, and the participation rate slid to 61.4%, a level you have to go back to the mid-1970s outside COVID to find. People aren't getting hired. They're leaving the denominator. A falling unemployment rate built on a shrinking labor force is not strength. It's the internals rotting while the headline ratio looks fine, and if today's report shows the same pattern, watch for a similarly muted initial reaction followed by a much uglier realization once someone actually reads past the first table.

Three days before the July print, BLS quietly published its preliminary annual benchmark revision: total nonfarm employment for the twelve months through March was overstated by 79,000 jobs, with private payrolls revised down 178,000 and government jobs revised up roughly 99,000. Retail trade took a 155,000-job markdown. Financial activities lost 76,000. The point isn't the size — a 0.1% adjustment is actually smaller than last year's -898K disaster and below the ten-year average revision magnitude. The point is direction. Bloomberg's economists expected an upward revision of 183,000. They got a downward one. Every economist survey feeding into today's 53-58K consensus is built on a data infrastructure that has now surprised in the wrong direction twice in one summer. Treat the consensus range as a guess dressed up in decimal points, not a number with edges you can trade against confidently.

ADP gave you the appetizer Wednesday: 38,000 private jobs added in August, below the 47,000 consensus, the smallest gain in seven months. Fifth Third's economist is calling for an outright 25,000 decline in today's government print, arguing the July collapse overstated cooling and citing an "outsized decline" tied to TPS cancellation working through the data with a lag most models haven't caught up to. Other shops are calling for an 80,000 rebound on the theory that July's undershoot mean-reverts. The dispersion between those two calls is larger than the entire consensus range from six months ago. Nobody actually knows what number prints in twelve minutes, and every trading desk pretending otherwise is lying to its own risk committee.

Position sizing note: Thursday's rally — Dow +624, Nasdaq +1.4%, 10-year down to 4.768%, 30-year to 5.24% — was built entirely on Waller's conditional, not on any actual data. That rally is now fully exposed to today's number in a way the tape hasn't acknowledged. A soft print in the Fifth Third range validates the hold thesis and probably extends Thursday's move. A print anywhere near the "rebound" camp's 80,000 handle puts a hike back on the table by lunchtime and unwinds a chunk of yesterday's yield decline, dragging equities down with it in the process — precisely the scenario the market spent Thursday betting against without hedging for it. Gold and Bitcoin already told you this isn't resolved: neither rallied on Waller's dovish tilt, which means neither believes the conditional holds. Trust the assets that didn't move over the ones that did.

Desk recommendation: reduce directional conviction into the print, not increase it. This is not a number to have an opinion on before 8:30 AM. It's a number to react to.