We looked at this graphic on job creation going into Friday's employment data.
It's now updated with August payrolls, and the revisions of the prior two months.
We now have a job market that has averaged just 27k net new jobs over the past four months. For context, the five-year period prior to the pandemic was 182k.
Add this to the OVER-reporting of jobs in the first seven months of the year, and clearly this is not the job market the Fed has been incorporating into its policy making decisions.
And with that, the question is: would the Fed have cut rates earlier this year had they been privy to the numbers in the final revision column of the table above?
Let's do an experiment.
If we ask ChatGPT to play the role of an FOMC voter, with the Fed's publicly stated policy making framework in mind, how would it have voted in its June meeting (at that point-in-time), given the knowledge of the actual payroll numbers (i.e. the final revisions)?
Below is the actual data heading into that June meeting — and the week prior, they would have gotten a reading on May CPI of 2.4%.
Remember, the Fed has admittedly been in a restrictive stance (putting downward pressure on the economy), and claims to be "data dependent" (led by the data).
So, here's the vote from ChatGPT …
On June 18, 2025, with the data you’ve laid out, the FOMC’s decision calculus would look like this:
FOMC Assessment
Inflation:
Employment:
Policy Stance:
-
At 4.33%, policy is restrictive relative to inflation outcomes.
-
Real rates are too high for a labor market already showing cracks.
Decision
Cut by 50 bps (to ~3.8%).
Reasoning:
-
A 25 bps trim risks being behind the curve when the labor market has stalled.
-
Inflation data gives you cover: at 2.4%, you’re not reigniting price pressures.
-
A bold move signals you’re serious about the employment half of the mandate, while still leaving real rates positive.
Bottom line: On June 18, with PCE at target and jobs collapsing, the FOMC would almost certainly deliver a 50 bp cut and open the door to more if the labor data deteriorates further.
So, based on the Fed's policy making framework, and the conditions of the economy based on the actual (final) data, ChatGPT thinks it warranted a 50 basis point cut in June. (xAI's Grok said 50, Google's Gemini said 25 and Anthropic's Claude said 25).
Instead, the Fed held steady, on the false belief (in their words) that the "labor market conditions remain solid."
Now we head into the September Fed meeting next week.
And we now have clear evidence of the deterioration in the labor market. But we have an uptick in inflation, which we may find on Wednesday will be as much as half a percentage point higher than June (as measured by CPI).
Will the Fed make up for its late start (i.e. deliver the 50bps cut it missed)? The market is pricing in only about a 10% chance.
But that positioning may change tomorrow.
The BLS is expected to make a big one-off downward adjustment to the jobs picture, in its scheduled annual "benchmark revision." A year ago, it slashed 818k jobs in this revision. Bessent said over the weekend he thinks it could be a similar number in tomorrow's revision.
