Durable goods prices, excluding cars, rose 7.1% over the past three months — the highest rate since 2022 — and 0.8% in July alone, which Morningstar's Preston Caldwell says is "very likely owing to tariffs."
Hold that number. It is the whole fight.
Because the evidence pack cannot even agree on what July CPI was. One wire, in a pre-release forecast, reportedly had economists projecting headline inflation at roughly 3.0%. Morningstar has it at 2.7% headline, 3.1% core. Same month, same BLS release, two different numbers in circulation — and that ambiguity is not noise. It is the designed output of a Fed that stripped out its own forecasts.
The market split is sharper still. Interest-rate swaps put the odds of a September quarter-point hike at roughly a coin toss (E2); CME FedWatch showed a 44.1% probability of a September hike (E3), while other desks priced over a 75% chance of at least one hike by year-end (E5). Bond futures markets price a September cut at 94%, per Morningstar citing the CME FedWatch Tool.
A hike and a cut, priced simultaneously, off one number. That is not a disagreement about data. It is a disagreement about what kind of central bank Kevin Warsh is running.
A hike and a cut, priced simultaneously, off one number. That is not a disagreement about data. It is a disagreement about what kind of central bank Kevin Warsh is running.
Last issue, we covered the July 29 hawkish hold — the 9-to-3 vote, the no-projections doctrine, and what the transcript frames as market participants "learning to play the ball, not the referee." This week the ball landed. The CPI print is the test that coverage teed up, and it tests exactly one thing: whether tariff inflation forces Warsh's hand.
Start with the decomposition, because the number's guts contradict its headline.
Durable goods ex-cars at 7.1% over three months is not a demand story. Nobody's animal spirits are, reportedly, bidding up furniture and appliances in a period where payrolls fell by 23,000. This is a cost being pushed down a supply chain — importers eating a tariff, then passing it to shelves.
Cars, tellingly, barely moved. Caldwell notes automakers "remain reluctant to hike prices for now, despite bearing substantial tariff costs." So the pass-through is uneven, mechanical, and concentrated exactly where import taxes bite hardest. That is the signature of a relative-price level shift, not a broad inflationary dynamic.
Here is the part every macro textbook agrees on: a central bank is supposed to look through a supply shock. A tariff raises the price level once; it does not, by itself, raise the trend. Tightening into it means tightening into an economy that has already been taxed — piling a demand shock on top of a supply shock to punish a number monetary policy didn't create and can't unwind.
The Fed's own framework says look through this. Warsh's July transcript even frames the question elegantly. He asked his colleagues whether price increases from shocks "indicate a broader inflationary dynamic, or do we just focus on them because they are under the bright streetlight."
That is the right question. The problem is that Warsh has demolished the equipment required to answer it credibly.
Looking through inflation is not a mood. It is a communication technology. To look through a hot print, a central bank has to tell the world: we see the number, we expect it to fade, here is our projected path, and we will not react to the level shift. Projections, dot plots, forward guidance — those are the instruments that let a Fed hold its nerve in public.
Warsh threw them out. "The policy statement conveys just the facts. It's steering clear of forecasting." No rolling forecasts, no commentary, markets told to respond "in the direction and magnitude they see fit."
You cannot look through a shock while refusing to say what's on the other side of it. Without a published path, "we're ignoring this print" is indistinguishable from "we're behind the curve." So the Fed loses the ability to look through the one kind of inflation its framework most wants to look through.
That leaves a reaction function that has quietly collapsed into two legs. Hot print: the hawks have cover to hike. Cool print: hold. The cut leg has been amputated.
Why amputated? Because of what a cut would now signify. Warsh spent an entire press conference insisting: "There is no soft inflation target; there is no soft implicit target—not on this Committee's watch. There's only a target, and it's 2 percent."
Now imagine the Committee cuts in September with core CPI at 3.1% and Morningstar's Preston Caldwell reportedly projecting PCE at 2.9%. That cut would not read as looking through a tariff. Stripped of any projection showing inflation returning to 2%, a cut against live core-goods inflation formally ratifies the exact "soft target" Warsh just denied exists.
He has boxed himself in on purpose. The credibility play — five-plus years above target, "this Fed will not waver" — makes easing into a 3-handle print a confession. So data-dependence, in the no-projections regime, becomes a one-directional machine: it hikes on hot, holds on cool, and the market's 94% cut is pricing a Fed that no longer exists.
The labor data is the one genuine force pulling the other way. Payrolls fell 23,000 in July, below every economist's forecast. Caldwell's case for a cut rests on it — and on his point that markets have priced cuts for months, so not cutting "would constitute an effective tightening."
That is real, and it is exactly the input the Committee must weigh against the price data. But weigh is the operative verb. In a hike-or-hold machine, a weak labor market doesn't produce a cut; it produces a hold that feels like mercy. The soft jobs number buys time, not easing.
And the inflation side has an open upside vent: oil. Iran and the Strait of Hormuz keep a war premium in crude, and the CNBC survey reportedly has high oil prices adding an estimated 0.6 percentage points to inflation this year. As long as energy can spike, the hike leg stays live and the hold stays hawkish.
There's a fourth question Warsh flagged that most people skipped: "how much accommodation are we getting from the balance sheet?" The Fed is still "making ample reserves." With the balance-sheet stance unresolved, the Committee has a second lever it can call tightening or accommodation without touching the funds rate — another reason the short-rate cut is not the release valve markets assume.
Which brings us to where the smart money actually is.
Gold ran to $4,350 and, per TradingKey, was reportedly forecast to press toward $4,500. The convenient story blames cooling hike bets — falling yields, lower opportunity cost. But that logic cuts the wrong way: hike odds above 75% on some desks should cap gold, not launch it.
The truer read is, reportedly, in the framing around Warsh's own remark about markets independently repricing the whole curve: central banks are losing their grip on the bond market. Gold at $4,500 is pricing debasement and sovereign-credit risk — a world where the Fed can neither cut cleanly nor control the long end — not a world of gentle rate relief.
And the real bet, the one with teeth, is in Treasuries. CTAs reportedly hold their largest bond short since UBS began tracking in 1990 — roughly $300 million of P&L per basis point on the 10-year, underweights that tripled in the back half of July and haven't budged.
That position is a wager that Warsh means it. That the reaction function really is hike-or-hold, that the 94% cut is wrong, that yields have further to climb.
UBS's Phoebe White calls the setup asymmetric — "there's not a lot of room to add" — and told clients to buy two-year notes on the softer jobs print. She's betting the crowd is too hawkish.
So both fault lines run through the same crack: is the cut leg really gone? If Warsh blinks and eases into tariff inflation, the bond short unwinds violently and the 94% crowd is vindicated. If he means every word of "not on this Committee's watch," what is reportedly the largest bond short in decades is the correct trade and gold's debasement bid is the tell.
The supply shock is wearing a demand shock's clothes. Warsh built a machine that can't undress it — and priced into $4,500 gold and a record bond short is a single question with no dot plot to answer it: does he mean it?